NASA/CP—2002-211506

Proceedings of the 36th Aerospace

Mechanisms Symposium

April 2002
 The NASA STI Program Office . . . in Profile

Since its founding, NASA has been dedicated to • CONFERENCE PUBLICATION. Collected
the advancement of aeronautics and space papers from scientific and technical
science. The NASA Scientific and Technical conferences, symposia, seminars, or other
Information (STI) Program Office plays a key part meetings sponsored or cosponsored by
in helping NASA maintain this important role. NASA.

The NASA STI Program Office is operated by • SPECIAL PUBLICATION. Scientific,

Langley Research Center, the Lead Center for technical, or historical information from
NASA’s scientific and technical information. The NASA programs, projects, and missions,
NASA STI Program Office provides access to the often concerned with subjects having
NASA STI Database, the largest collection of substantial public interest.
aeronautical and space science STI in the world.
The Program Office is also NASA’s institutional • TECHNICAL TRANSLATION. English-
mechanism for disseminating the results of its language translations of foreign scientific
research and development activities. These results and technical material pertinent to NASA’s
are published by NASA in the NASA STI Report mission.
Series, which includes the following report types:
Specialized services that complement the STI
• TECHNICAL PUBLICATION. Reports of Program Office’s diverse offerings include
completed research or a major significant creating custom thesauri, building customized
phase of research that present the results of data bases, organizing and publishing research
NASA programs and include extensive data results . . . even providing videos.
or theoretical analysis. Includes compilations
of significant scientific and technical data and For more information about the NASA STI
information deemed to be of continuing Program Office, see the following:
reference value. NASA’s counterpart of peer-
reviewed formal professional papers but • Access the NASA STI Program Home Page
has less stringent limitations on manuscript at http://www.sti.nasa.gov
length and extent of graphic presentations.
• E-mail your question via the Internet to
• TECHNICAL MEMORANDUM. Scientific [email protected]
and technical findings that are preliminary or
of specialized interest, e.g., quick release • Fax your question to the NASA Access
reports, working papers, and bibliographies Help Desk at 301–621–0134
that contain minimal annotation. Does not
contain extensive analysis. • Telephone the NASA Access Help Desk at
301–621–0390
• CONTRACTOR REPORT. Scientific and
technical findings by NASA-sponsored • Write to:
contractors and grantees. NASA Access Help Desk
NASA Center for AeroSpace Information
7121 Standard Drive
Hanover, MD 21076
NASA/CP—2002-211506

Proceedings of the 36th Aerospace

Mechanisms Symposium

Proceedings of a conference held at the

Sheraton City Centre Hotel and
cosponsored by the NASA Glenn Research Center and
Lockheed Martin Space Systems Company and
organized by the Mechanisms Education Association
Cleveland, Ohio
May 15–17, 2002

National Aeronautics and

Space Administration

Glenn Research Center

April 2002
 Acknowledgments

The high quality of this symposium is a result of the work of many people, and their efforts are gratefully
acknowledged. This extends to the voluntary members of the symposium organizing committee representing the
eight NASA field centers, LMSSC, and the European Space Agency. Appreciation is also extended to the session chairs,
the authors, and particularly the personnel at GRC responsible for the symposium arrangements and the publication
of these proceedings. A sincere thank you also goes to the symposium executive committee who is responsible for the
year-to-year management of the AMS, including paper processing and preparation of the program.

Trade names or manufacturers’ names are used in this report for

identification only. This usage does not constitute an official
endorsement, either expressed or implied, by the National
Aeronautics and Space Administration.

Available from
NASA Center for Aerospace Information National Technical Information Service
7121 Standard Drive 5285 Port Royal Road
Hanover, MD 21076 Springfield, VA 22100

Available electronically at http://gltrs.grc.nasa.gov/GLTRS

 Preface

The Aerospace Mechanisms Symposium (AMS) provides a unique forum for those active in
the design, production, and use of aerospace mechanisms. A major focus is the reporting of
problems and solutions associated with the development and flight certification of new
mechanisms. Organized by the Mechanisms Education Association, the National Aeronautics
and Space Administration and Lockheed Martin Space Systems Company (LMSSC) share the
responsibility for hosting the AMS. Now in its 36th year, the AMS continues to be well attended,
attracting participants from both the United States and abroad.

The 36th AMS, hosted by the Glenn Research Center (GRC) in Cleveland, Ohio, was held
May 15, 16, and 17, 2002. During these 3 days, 32 papers were presented. Topics included
deployment mechanisms, tribology, actuators, pointing and optical mechanisms, International
Space Station mechanisms, release mechanisms, and test equipment. Hardware displays during
the supplier exhibit gave attendees an opportunity to meet with developers of current and future
mechanism components.

NASA/CP—2002-211506 iii
 Table of Contents

Agenda.............................................................................................................................................ix

Organizing and Advisory Committees............................................................................................. xiv

Asteroid Sampling Mechanism for MUSES−C

Ken Higuchi, Jun’ichiro Kawaguchi, Akira Fujiwara, The Institute of Space and Astronautical
Science; Toshiaki Okudaira, NEC TOSHIBA Space Systems; and Satoru Yajima,
Sumitomo Heavy Industries ........................................................................................................ 1

A Description of Mechanisms Used in the Low Resolution Airglow and Aurora Spectrograph
and Special Sensor Ultraviolet Limb Imager
Philip Kalmanson and Russell Starks, Praxis, Inc.; and Stefan Thonnard and
Kenneth Dymond, Naval Research Laboratory ......................................................................... 13

The IRAC Shutter Mechanism: Residual Magnetism and the Rotary Solenoid
Scott Schwinger, Claef Hakun, Gary Brown, and Ken Blumenstock,
NASA Goddard Space Flight Center ........................................................................................ 15

A Cryogenic Half-Wave Plate Module to Measure Polarization at Multiple FIR Passbands

Timothy S. Rennick, John E. Vaillancourt, Roger H. Hildebrand, and Stephen J. Heimsath,
The University of Chicago ........................................................................................................ 27

Tape-Spring Rolling Hinges

Alan M. Watt and Sergio Pellegrino, University of Cambridge.................................................. 37

Material Property Effects on Coaxial Cable Mechanical Failure

R.B. Pan, J.B. Chang, C.C. Wan, Y.R. Takeuchi, The Aerospace Corporation;
R. McVey, Boeing Satellite Systems; and I. Chen, Raytheon Systems ..................................... 51

Design and Testing of the CRISP Tracking Mirror Cover and Release Mechanism
Jeffrey Lees and Ed Schaefer, Johns Hopkins University Applied Physics Laboratory............... 63

Zero Deadband, Multiple Strut Synchronized Hinge for Deployable Structures

Matthew Botke, David Murphy, Thomas Murphey, AEC-Able Engineering;
and Peter Sorensen, Kollabra................................................................................................... 77

DC Motor Selection for Space Mechanisms

David B. Marks and Richard A. Fink, Moog Inc......................................................................... 89

Design and Manufacturing of Extremely Low Mass Flight Systems

Michael R. Johnson, Jet Propulsion Laboratory...................................................................... 103

Design of a Robotic Wrist and Tool-Exchange Mechanism for Satellite Servicing

Russell D. Howard, University of Maryland .............................................................................. 117

Operational Improvements of a Pyrotechnic Ultra Low Shock Separation Nut

Alexander Luna, Hi-Shear Technology Corporation............................................................... 131

NASA/CP—2002-211506 v
Alternatives to Pyrotechnics—Nitinol Release Mechanisms
Andrew Tuszynski, Hi-Shear Technology Corporation............................................................. 137

System Characterization and Motor Step Verification Through Rotary Acceleration Signals
Scott Starin and Fred Crosno, CDA InterCorp ......................................................................... 141

High Resolution Standard Proximity Sensor

Roger Blaser, Vibro-Meter SA................................................................................................. 147

Qualification of the Inflatable Sunshield In Space (ISIS) Mast

David J. Rohweller, TRW Astro Aerospace ............................................................................. 155

ISS S/A Deployment—The Highs and Lows of EVA Contingency Capability, A Designer’s
Perspective
Bert Haugen, Malcolm Ferry, and Kevin Klein, Lockheed Martin Space Systems Company.. 169

The Lightweight Deployable Antenna for the MARSIS Experiment on the Mars Express
Spacecraft
Geoffrey W. Marks and Michael T. Reilly, TRW Astro Aerospace; and Richard L. Huff,
University of Iowa.................................................................................................................... 183

MABE: High-Precision Tip/Tilt Mechanism Based on Magnetic Bearing Technology

N. Loix and J.Ph. Verschueren, Micromega Dynamics; and L. Scolamiero, ESA/ESTEC...... 197

Development of a Fine Steering Mirror Assembly

Ruben Nalbandian and Jeff Williams, Moog Inc. ................................................................... 207

A Description and Performance of Mechanisms Used in the High Resolution Ionospheric

Thermospheric Spectrograph
Phillip Kalmanson, Praxis Inc.; Hsiung Chou, Atlantic Online; Russell Starks, Praxis Inc.;
and Kenneth Dymond and Stefan Thonnard, Naval Research Laboratory.............................. 215

Design of an EVA-Capable Four-Point Restraint System

Armond Asadurian and Jim Hammond, Moog Chatsworth Operations; Kazuo Hamada and
Hiroshi Irikado, NEC TOSHIBA Space Systems; and Michio Ito and Miho Endo, NASDA....... 229

Intricacies of Using Kevlar Cord and Thermal Knives in a Deployable Release System:
Issues and Solutions
Alphonso C. Stewart and Jason H. Hair, NASA Goddard Space Flight Center........................ 243

Release Mechanisms on the Gravity Probe-B Relativity Mission

Sean McCully and Dennis St. Clair, Lockheed Martin Space Systems Company................... 257

Conceptualization and Design of a Mechanical Docking System

Troy Nilson and Mitch Wiens, Starsys Research Corporation................................................... 265

Automating Separation System Reliability Testing

Ryan L. Perroy, Planetary Systems Corporation...................................................................... 279

Random Vibration Failure Mechanism of a Conrad Bearing Using Crown Ball-Retainer

Klaus Engel, COM DEV International; Regina Kwiatkowska, COM DEV Space Group;
and Mihai Vladimirescu, COM DEV International .................................................................. 291

NASA/CP—2002-211506 vi
Comparison of Several Different Sputtered Molybdenum Disulfide Coatings for Use in Space
Applications
Robert L. Fusaro, NASA Glenn Research Center; and Mark Siebert, University of Toledo ..... 305

Space Ball Bearing Load Capacity With Dry Lubrication

A. Borrien, J.B. Mondier, and E. Conde, CNES ...................................................................... 319

Tribological Properties of a Pennzane-Based Liquid Lubricant (Disubstituted Alkylated

Cyclopentane) for Low Temperature Space Applications
Clifford Venier, Pennzoil-Quaker State Company; Edward W. Casserly, Penreco Company;
William R. Jones, Jr., and Mario Marchetti, NASA Glenn Research Center; Mark J. Jansen,
Sest, Inc.; and Roamer E. Predmore, NASA Goddard Space Flight Center............................ 331

ISS SGANT Group Level Offloading Test Mechanism

Xilin Zhang, EMS Technologies Canada, Ltd........................................................................ 339

Relationship Between the Difficulty Index and the Evaluation Accuracy of Ground Deployment
Testing
Hironori Ishikawa and Akira Meguro, NTT Corporation............................................................ 353

NASA/CP—2002-211506 vii
 Agenda

WEDNESDAY, 15 MAY 2002

8:00 WEDNESDAY PRESENTERS' BREAKFAST—Hope Room on 7th Floor

8:00 CHECK-IN AND REFRESHMENTS—Registration Office and East Ballroom on 6th Floor

9:00 INTRODUCTORY REMARKS

Fred Oswald, Host Chairman
NASA Glenn Research Center, Cleveland, OH
Stuart H. Loewenthal, General Chairman
Lockheed Martin Space Systems Company, Sunnyvale, CA

CENTER WELCOME
Dr. Donald Campbell, Center Director
NASA Glenn Research Center, Cleveland, OH

9:30 SESSION I—INSTRUMENT MECHANISMS

Lucio Scolamiero, Session Chair
ESA/ESTEC, Noordwijk, The Netherlands
• Asteroid Sampling Mechanism for MUSES-C
Ken Higuchi, Jun’ichiro Kawaguchi and Akira Fujiwara, The Institute of Space and
Astronautical Science, Sagamihara, Japan; Toshiaki Okudaira, NEC TOSHIBA Space
Systems, Yokohama, Japan; Satoru Yajima, Sumitomo Heavy Industries, Niihama, Japan
• A Description of Mechanisms used in the Low Resolution Airglow and Aurora
Spectrograph and Special Sensor Ultraviolet Limb Imager
Philip Kalmanson, Russell Starks and Kenneth Dymond, Praxis, Inc., Alexandria, VA; Stefan
Thonnard, Naval Research Laboratory, Washington, D.C.
• The IRAC Shutter Mechanism: Residual Magnetism and the Rotary Solenoid
Scott Schwinger, Claef Hakun, Gary Brown and Ken Blumenstock, NASA Goddard Space
Flight Center, Greenbelt, MD
• A Cryogenic Half-Wave Plate Module to Measure Polarization at Multiple FIR Passbands
Timothy Rennick, John Vaillancourt, Roger Hildebrand and Stephen Heimsath, University
of Chicago, Chicago, IL

11:30 LUNCH BREAK

12:45 SESSION II—DEPLOY

Bert Haugen, Session Chair
Lockheed Martin Space Systems, Sunnyvale, CA
• Tape-Spring Rolling Hinges
Alan Watt and Sergio Pellegrino, University of Cambridge, Cambridge, United Kingdom
• Material Property Effects on Coaxial Cable Mechanical Failure
Robert Pan, J.B. Chang, C.C. Wan and Y. R. Takeuchi, The Aerospace Corporation,
El Segundo, CA; R. McVey, Boeing Satellite Systems, El Segundo, CA; I. Chen,
Raytheon Systems, El Segundo, CA
• Design and Testing of the CRISP Tracking Mirror Cover and Release Mechanism
Jeffrey Lees and Ed Schaefer, Johns Hopkins University Applied Physics Laboratory,
Laurel, MD
• Zero Deadband, Multiple Strut Synchronized Hinge for Deployable Structures
Matthew Botke, David Murphy and Thomas Murphey, AEC-ABLE Engineering, Goleta, CA;
Peter Sorensen, Kollabra, Palo Alto, CA

2:45 BREAK

NASA/CP—2002-211506 ix
3:00 SESSION III—ACTUATORS, ROVERS, ROBOTS AND POSTER PREVIEW
Karl Anderson, Session Chair
Moog Inc., Chatsworth, CA
• DC Motor Selection for Space Mechanisms
David Marks and Richard Fink, Moog Inc., Research Triangle Park, NC
• Designing and Manufacturing of Extremely Low Mass Flight Systems
Michael Johnson, Jet Propulsion Laboratory, Pasadena, CA
• Design of a Robotic Wrist and Tool-Exchange Mechanism for Satellite Servicing
Russell Howard, University of Maryland, College Park, MD
• Operational Improvements of a Pyrotechnic Ultra Low Shock Separation Nut
Alexander Luna, Hi-Shear Technology Corporation, Torrance, CA
• Alternatives to Pyrotechnics - Nitinol Release Mechanisms
Andrew Tuszynski, Hi-Shear Technology Corporation, Torrance, CA
• System Characterization and Motor Step Verification through Rotary Acceleration Signals
Scott Starin and Fred Crosno, CDA InterCorp, Deerfield, FL
• High Resolution Standard Proximity Sensor
Roger Blaser, Vibro-Meter SA, Fribourg, Switzerland

6:30–9:30 RECEPTION AND DISPLAYS AT THE SHERATON—West Ballroom on 6th Floor

Invited component and software suppliers display current products and provide tutorials.

NASA/CP—2002-211506 x
THURSDAY, 16 MAY 2002

7:30 THURSDAY PRESENTERS' BREAKFAST—Hope Room on 7th Floor

8:30 SESSION IV—DEPLOY II

Gary Heinemann, Session Chair
AEC-Able Engineering, Goleta, CA
• Qualification of the Inflatable Sunshield In Space (ISIS) Mast
David Rohweller, TRW Astro Aerospace, Carpinteria, CA
• ISS S/A Deployment—The Highs and Lows of EVA Contingency Capability, A Designer’s
Perspective
Bert Haugen, Malcolm Ferry and Kevin Klein, Lockheed Martin Space Systems,
Sunnyvale, CA
• The Lightweight Deployable Antenna for the MARSIS Experiment on the Mars Express
Spacecraft
Geoffrey Marks and Michael Reilly, TRW Astro Aerospace, Carpinteria, CA; Richard Huff,
University of Iowa, Iowa City, IA

10:00 BREAK

10:15 SESSION V—INSTRUMENT MECHANISMS II

Martha Milton, Session Chair
NASA Marshall Space Center
• MABE: High-Precision Tip/Tilt Mechanism Based on Magnetic Bearing Technology
Nicolas Loix and J.Ph. Verschueren, Micromega Dynamics SA, Angleur, Belgium; Lucio
Scolamiero, ESA/ESTeC, Noordwijk, The Netherlands
• Development of a Fine Steering Mirror Assembly
Ruben Nalbandian and Jeff Williams, Moog Inc., Chatsworth, CA
• A Description and Performance of Mechanisms used in the High Resolution Ionospheric
Thermospheric Spectrograph
Phillip Kalmanson and Russell Starks, Praxis, Inc., Alexandria, VA; Hsiung Chou, Raytheon
STX, Lanham, MD; Kenneth Dymond and Stefan Thonnard, Naval Research Laboratory,
Washington, D.C.

11:45 LUNCH BREAK

1:00 SESSION VI—RELEASE AND DOCKING MECHANISMS

Robert Fusaro, Session Chair
NASA Glenn Research Center (ret), Cleveland, OH
• Design of an EVA-Capable Four-Point Restraint System
Armond Asadurian and Jim Hammond, Moog Inc., Chatsworth, CA; Kazuo Hamada and
Hiroshi Irikado, NEC TOSHIBA Space Systems, Japan; Michio Ito and Miho Endo,
NASDA, Japan
• Intricacies of Using Kevlar Cord and Thermal Knives in a Deployable Release System:
Issues and Solutions
Alphonso Stewart and Jason Hair, NASA Goddard Space Flight Center, Greenbelt, MD
• Release Mechanisms on the Gravity Probe-B Relativity Mission
Sean McCully and Dennis St. Clair, Lockheed Martin Space Systems Co., Sunnyvale, CA
• Conceptualization and Design of a Mechanical Docking System
Troy Nilson and Mitch Wiens, Starsys Research, Boulder, CO
• Automating Separation System Reliability Testing
Ryan Perroy, Planetary Systems Corporation, Silver Spring, MD

3:30 BREAK

NASA/CP—2002-211506 xi
3:45 SPECIAL PRESENTATION: ROSE PARADE FLOAT MECHANISMS
The under-the-hood details on what makes those beautiful, animated, short-lived floats
dazzle audiences around the world every January 1st.
Michael Johnson, Jet Propulsion Laboratory, Pasadena, CA

5:30–10:30 SYMPOSIUM BANQUET AT GREAT LAKES SCIENCE CENTER

Walk or bus to Science Center

7:00 Dinner and displays

10:00 Return to the hotel

NASA/CP—2002-211506 xii
FRIDAY, 17 MAY 2002

7:30 FRIDAY PRESENTERS’ BREAKFAST—Hope Room on 7th Floor

8:30 SESSION VI—TRIBOLOGY AND GROUND TEST

David Osterberg, Session Chair
Honeywell Inc., Glendale, AZ
• Random Vibration Failure Mechanism of a Conrad Bearing using Crown Ball-Retainer
Klaus Engel, Regina Kwiatkowska and Mihai Vladimirescu, COM DEV, Ltd., Cambridge,
Canada
• Comparison of Several Different Sputtered Molybdenum Disulfide Coatings for Use in
Space Applications
Robert Fusaro, NASA Glenn Research Center, Cleveland, OH; Mark Siebert, University of
Toledo, Toledo, OH
• Space Ball Bearing Load Capacity with Dry Lubrication
Andre Borrien, Jean-Bernard Mondier and E. Conde, Centre National D'Etudes Spatiales,
Toulouse, France
• Tribological Properties of a Pennzane-Based Liquid Lubricant (Disubstituted Alkylated
Cyclopentane) for Low Temperature Space Applications
Clifford Venier, Pennzoil-Quaker State Company, The Woodlands, TX; Edward Casserly,
Penreco Company, The Woodlands, TX; William Jones, Jr. and Mario Marchetti, NASA
Glenn Research Center, Cleveland, OH; Mark J. Jansen, Sest, Inc., Middleburg Hts., OH;
Roamer Predmore, NASA Goddard Space Flight Center, Greenbelt, MD
• ISS SGANT Group Level Offloading Test Mechanism
Xilin Zhang, EMS Technologies Canada, Ltd., St. Anne-de-Bellevue, Canada
• Relationship between the Difficulty Index and the Evaluation Accuracy of Ground
Deployment Testing
Hironori Ishikawa and Akira Meguro, NTT Network Innovation Laboratories, NTT
Corporation, Kanagawa, Japan

11:30 BREAK

11:45 PRESENTATION: AN OVERVIEW OF GRC

12:15 TECHNICAL SESSIONS CONCLUSION

• Presentation of the Herzl Award
• Closing Remarks

1:30 BUSES DEPART HOTEL FOR GRC TOUR

1:30–3:30 GLENN RESEARCH CENTER FACILITY TOUR

NASA/CP—2002-211506 xiii
 Organizing Committee

Fred B. Oswald, Host Chairman, NASA GRC

Stuart H. Loewenthal, General Chairman, Lockheed Martin

Edward A. Boesiger, Operations Chairman, Lockheed Martin

Obie H. Bradley, Jr., NASA LRC

Michael J. Eiden, ESA/ESTEC
Carlton L. Foster, NASA MSFC
Claef F. Hakun, NASA GSFC
Christopher P. Hansen, NASA JSC
Alan C. Littlefield, NASA KSC
Edward C. Litty, JPL
Ronald E. Mancini, NASA ARC
Fred G. Martwick, NASA ARC
Stewart C. Meyers, NASA GSFC
Martha E. Milton, NASA MSFC
Wilfredo Morales, NASA GRC
Minh Phan, NASA GSFC
John F. Rogers, NASA LRC
Donald R. Sevilla, JPL
Mark F. Turner, NASA ARC

Advisory Committee

David F. Engelbert, NASA ARC (ret)

Robert L. Fusaro, NASA GRC (ret)
John W. Redmon, Sr., NASA MSFC (ret)
Alfred L. Rinaldo, Lockheed Martin (ret)
William C. Schneider, NASA JSC (ret)
Bowden W. Ward, Jr., NASA GSFC (ret)

NASA/CP—2002-211506 xiv
 Asteroid Sampling Mechanism for MUSES-C

Ken Higuchi∗, Jun’ichiro Kawaguchi*, Akira Fujiwara*, Toshiaki Okudaira**, and Satoru Yajima
+

Abstract

An asteroid mission of the Institute of Space and Astronautical Science (ISAS) is devoted to acquiring a
sample on an asteroid surface and then returning it to the earth. Its sampling system, named “sampler”, is
a unique idea and is a subsystem of the MUSES-C spacecraft. Sampler has many engineering
challenges, but with the final purpose of the sampling being scientific, the sampler structures and
mechanisms have been developed under the scientific requirements and restrictions. The sample should
not be contaminated by artificial matter or natural matter originated on the earth during manufacturing,
integration, cruising, sampling, reentry, and recovery. The sampler has three projectors, and each
projector fires a projectile in order to fragment an asteroid surface into small pieces. The sample is stored
in a container and it is retrieved using a reentry capsule. The sampler involves many linear and rotational
sliding mechanisms, and they should work in a dusty or crunchy environment after a long cruise period in
a space environment. The function of each mechanism has been verified by ground tests. Sliding tests in
a sandy and crunchy environment verified the function and the limitations of the sampler, and high-
frequency shock tests verified that the sample container is airtight during the reentry.

Introduction

A sampling system on an asteroid surface was developed to meet engineering and scientific requirements
after a long cruise in space environment. In order not to reduce the scientific value of the sample, the
structures and mechanisms of the sampler should meet many scientific requirements such as
contamination, sealing, and dismantle process. Popular space lubricants are not welcomed, and hard-
anodized aluminum coated by polytetrafluoroethylene (PTFE) is used at the surface of sliding parts. The
sampling mechanism is to work after a two-year cruise in space, and the sample in the reentry capsule is
to be retrieved after four-year cruise. However, the sampling mechanism needs to work in a presumably
crunchy environment, because the projector makes dust, sand and fragments of rocks fly up. The
functions of sliding parts in crunchy environments need to be verified by tests. The mechanisms are
considered to be as simple as possible; no electromagnetic motor is used. Projectors to make samples
and a movable horn to collect samples are adaptive to some topographical and physical uncertainty of the
environment on an asteroid.

MUSES-C stands for a series of missions performed by the Space Engineering Spacecraft launched by
MU rocket of ISAS, and the C means the third mission of this series. The primary goal of the MUSES-C
mission is to acquire and verify technology necessary to retrieve samples and to bring back them to the
earth. The spacecraft uses a solar-powered electrical propulsion system, and an autonomous navigation
and guidance to a small target in deep space. The spacecraft collects a sample during the touch-and-go
on the asteroid surface. A small robot lander will be also delivered to the surface. In the final phase, a
small reentry capsule conducts the reentry flight from the interplanetary earth return trajectory. The
scientific returns are promising even if the sample amount is very small.

The design conditions of the sampler are discussed, then the sampling sequence is illustrated, and the
operational tests are presented.

∗
The Institute of Space and Astronautical Science, Sagamihara, Japan
** NEC TOSHIBA Space Systems, Yokohama, Japan
+
Sumitomo Heavy Industries, Niihama, Japan
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 1
 Design Conditions

Sampler has one or more projectors, and each projector fires a projectile in order to fragment an asteroid
surface into small pieces as sample. The fragmented sample is collected and conducted into a storage
vessel called ‘catcher’ through a horn and a duct by its own multi-reflection. The sample obtained in every
sampling projection is stored separately in the partitioned catcher. After all the projection, the catcher
moves into a container in a reentry capsule to retrieve the sample on the earth. The container is airtight in
order not to leak volatile matter of the sample during cruise and also to prevent an invasion of atmosphere
during reentry and recovery.

SPCECRAFT
SIDE PANEL
REENTRY SEPARABLE
CAPSULE SAMPLE
CATCHER
TRANSPORTATION

AIRTIGHT
CONTAINER
REVOLBABLE AND
SPACECRAFT RETRACTABLE DUCT
BASE PANEL

PROJECTOR(S)
FRAGMENT
AS SAMPLE
LASER RANGE FINDER
SAMPLE TO DETECT
COLLECTION TOUCHDOWN OF
HORN MOVABLE HORN

ASTEROID SURFACE

Figure 1. Outline of Sampling

Scientific restrictions are as follows. The amount of sample is sufficient for scientific analysis and the
sampling process must not degenerate the physical properties of the sample. The sampler is equipped
with up to three projectors, each of which can project one projectile. The size of the catcher and the size of
duct from the surface to the catcher should have room for fragments of up to 10 mm in diameter. The
fragments of different sampling attempt position on an asteroid surface are stored separately. The sample
should not be contaminated by artificial matter and natural matter originated on the earth. The artificial
contamination matter comes from thrusters of the spacecraft, projectors of a gunpowder type, sampler
structure surface, sliding mechanisms, manufacturing and handling processes, and so on. The cleanliness
of the environment during handling and integration processes is monitored, and witness plates are
attached to the sampler to monitor the contamination during the manufacturing, handling, and the
integration at the launching site. The materials and the surface treatment of the structure and mechanisms
of the sampler should be selected not to reduce scientific merit.

There are also many engineering restrictions and conditions. Anchoring the spacecraft for sampling will
not be available. Sampling involves hovering, touchdown, and lift-off of the spacecraft. The spacecraft and
the total system should be kept safe during the sampling sequence, although the sequence of actions
proceeds autonomously without any command from the earth. The sampling action must not damage the

NASA/CP—2002-211506 2
spacecraft. Each firing of projector is triggered by the detection of touchdown on the asteroid surface. The
projector should be designed to minimize the contamination to the sample from combustion gases.
Sampling sequence should be executable safely. For safety, the sampler subsystem should match the
other subsystems of the spacecraft and mission such as AOCS, RCS, reentry capsule, structures, and
thermal systems. Sampler horn does not interfere with any range of vision of the other instruments for
navigation and science. High-speed fragments and reflected projectile may not damage spacecraft
including large solar array panels. Structures and mechanisms are adaptive to an expected slope and
roughness of the surface, even if the spacecraft remains the transversal velocity. Sampling mechanism is
to work after two-year cruise in space environment. Sampling actions will be carried out in presumably
crunchy environment with dust, sand, and fragment. The transportation system of the catcher from the
duct position to the container in the reentry capsule does not become a cause of disturbances at the
separation of the reentry capsule from the spacecraft. Every mechanism and sequence works steadily in
the asteroid environment of high vacuum and micro-gravity. The best effort should be made so that the
occurrence of a malfunction in sampling sequence can minimize the effects on the rest of the mission.
The sampler system should be lightweight structure, consume a small amount of electric power, stays
within the spacecraft, stay within the allowed envelope at launching, and withstand the mechanical and
thermal environment of the launching, cruising, and sampling phases. The sample in the reentry capsule
should be recovered without loss of scientific merit in a four-year cruise after launch. The clearance of
every sliding part must be verified by sliding tests. A design of sampling system must permit some
uncertainty of environments on the asteroid such as temperature, gravitation, hardness, roughness, slope,
and constituent of the surface.

Structures and Mechanisms

-3
The outline of sampling is illustrated in Figure 1. A projectile of 5x10 kg needs to be fired with velocity of
300 m/s to fragment the target surface. The projector method may be able to meet the variation of
conditions of the surface, which may be made of solid metal, rock, or sand. Sample fragments, and also
-6 3
projectiles by some chance, are caught in the catcher. The inside volume is about 60x10 m . The
fragments of different sampling points on the surface are stored in a partitioned catcher. The structure of
the whole sampler is shown in Figure 2. It consists of two major portions, a horn mechanism assembly
and a sampling mechanism assembly. Figure 3 shows the components. The horn, which is a deployable
structure, is useful not only because it is compact at launching but also because it works as a part of
touchdown sensor repeatedly, and because it is adaptive to the variation in the slope and roughness of
asteroid surface. However, the spacecraft remains the transversal velocity. The extended length of the
horn is 1 m from the bottom of spacecraft.

Usually, space mechanisms work under a high-vacuum and clean environment soon after launch. On the
other hand, the sampling mechanism needs to work in a presumably crunchy environment, because the
projector may make hard fragments and the surface may possibly be sandy. Therefore, the lubrication of
the mechanism in a bad environment is very important. However, due to requirements of scientific
analyses, typical solid lubricants such as molybdenum disulfide, gold, silver, and lead, are not permitted
where samples may confront or pass by. Hard-anodized aluminum coated by PTFE, which does not have
a fatal influence on the samples, is used on the surface of sliding parts. The surface of aluminum
structural elements that contact the sample is coated with purified aluminum, according to scientific
requirements.

All of the mechanisms work by spring forces for simplicity. No electromagnetic motor is used. The spring
forces are released by shape memory alloy (SMA) driven actuators, a non-explosive actuator (NEA), and
a pyrotechnic wire cutter. The magnitude of spring forces is restricted, because the spring forces and the
actuation shock are directly related to the weight of the structures and mechanisms of sampler, although
the mechanisms need enough force to work in a crunchy environment after long exposure duration in
space environments. The weight of sampler is shown in Table 1.

The catcher is stored in a container in the reentry capsule at the final phase of sampling. The container
should be kept airtight against the shock environments during reentry and landing.

NASA/CP—2002-211506 3
 SIDE PANEL OF SPACECRAFT

REENTRY
CAPSULE
CONVEYER GUIDANCE CYLINDER

REVOLVING AND RECEDING

MECHANISMS

FLANGE OF SPACECRAFT

PROJECTOR

BASE PANEL OF SPACECRAFT

EXTENDIBLE HORN
CLAMP

HORN I/F PLATE

FIXED HORN

DUST SHIELD
ALLOWED ENVELOPE
AT LAUNCHING

MOVABLE HORN

Figure 2. Components of Sampler

NASA/CP—2002-211506 4
 SAMPLER

HORN ASSEMBLY

SAMPLING ASSEMBLY HORN STRUCTURE AND MECHANISM

CATCHER ASSEMBLY FIXED HORN

SAMPLE SEPARATION MECHANISM EXTENDIBLE CLOTH HORN

OF REVOLVING TUBE
MOVABLE HORN
RETRACTION MECHANISM OF
SAMPLE CONDUCTION TUBE
CLAMP AND RELEASE
MECHANISM
CATCHER TRANSPORTAION
MECHANISM AND STRUCTURE
EXTENSION MECHANISM AND
LATCH AND SEAL MECHANISM COMPLIANT STRUCTURE
OF CONTAINER
DEPLOYABLE SHIELD

PROJECTOR

Figure 3. Structure of Sampler

Table 1. Weight of Sampler

Classification Mechanisms Weight
-3
[10 kg]
Sampling Mechanism Portion
Catcher Mechanism 70
Sample Separation Mechanism 390
Sample Conduction Tube Retraction Mechanism 250
Sample Conveyer Mechanism 1520
Latch and Seal Mechanism 970
Electrical and Thermal Wiring 350
Bolts and Nuts 100
Horn Mechanism Portion
Horn Mechanism 4650
Projectors 450
Electrical and Thermal Wiring 300
Bolts and Nuts 100
Total 9150

Metallic parts of the fixed horn and the movable horn are made of aluminum alloy plate. The materials and
the thickness of the horn are determined based upon penetration tests of projectile. The other metallic
structural parts are made of aluminum alloy and magnesium alloy. Extendible and adaptive horn is made
of cloth of high-tension polyarylate fiber (VECTRAN) and a double-helical spring of glass fiber reinforced
plastic (GFRP). The selection of cloth materials needs the penetration tests and much consideration to
long and severe space environments.

NASA/CP—2002-211506 5
 Sampling Sequence and Characteristics of Sampler

The horn is kept retracted at the launch in order to satisfy the allowed envelope, and is extended after
launch. The spacecraft descends toward an asteroid, and a projector is fired just after perception of
touchdown. Firing is triggered by the perception of displacement of the extended movable horn. The
displacement is detected by a laser range finder. The movable horn is to cover the crater caused by a
projectile, and the fixed horn and the conduction tube are to collect and conduct fragments into a catcher
using multi-reflection; no additional power is necessary to conduct the fragments. The horn is also
adaptive to comply with the slope and roughness of the surface, and is to protect the spacecraft against
high-speed fragments and reflected projectiles. To this end, the movable horn has a deployable shielding
brim made of cloth. The brim is deployed at the same time as the extension of the movable horn. The
schematic diagram is illustrated in Figure 4(a).

FOLDED UNDER DEPLOYMENT FULLY DEPLOYED

CONDUCTION
SPACECRAFT BASE PANEL TUBE

CLAMP FIXED
HOLDER HORN

DEPLOYABLE
STRUCTURE

MOVABLE HORN
EXTENDIBLE
CLOTH HORN
DEPLOYABLE DUST SHIELD

FIXED DUST SHIELD

Figure 4. Sampling Sequence, (a) Extension of Horn

The sample obtained in every sampling attempt is stored in the catcher separately by a revolving duct
mechanism. After the first projection, the aperture of the revolving duct, which is open to the first sample
room A, rotates 120 degrees to close the room A and to be open to the next sample room B, and the
catcher become ready for the next sampling. After each sampling attempt, the spacecraft lifts-off as soon
as possible to avoid a crash at the asteroid surface. After the final sampling attempt, the aperture of the
revolving duct rotates another 120 degrees and confronts a blind wall in the catcher. The conduction tube
together with the revolving duct recedes from the catcher, and the catcher moves into the container
through the backboard of the reentry capsule. The rear abrator of the reentry capsule and the lid of
container are attached to the catcher, and are conveyed together with the catcher. The lid is latched
-3
automatically at the end of the transportation. The length of the conveyer guidance cylinder is 230x10 m,
-3
and the transportation distance is 160x10 m. The sealing force to compress the double O-ring is obtained
by other springs installed inside the lid. The initially clamped springs are released by an action of NEA.
The container is kept airtight until it is dismantled on the earth. Only the seal mechanism of the container
works by using spring forces, which are released by an NEA. An NEA is used not only because the space
for an actuator in the lid is extremely limited, but also because the sample should be kept away from
combustion gases of a pyrotechnic actuator, and because large actuation shock by pyrotechnic devices
has bad effects on the electronic instruments adjacent to the seal mechanism in the reentry capsule. The
remaining conveyer force is released in order not to cause disturbances at the separation of the reentry
capsule from the spacecraft. The reentry capsule is separated using a spin-separation mechanism. The
schematic diagram is illustrated in Figures 4(b)-(d).

NASA/CP—2002-211506 6
 Z
SPCECRAFT
REENTRY PANEL
CAPSULE X
CATCHER Y

ROOMS A AND B

REVOLVING
CONDUCTION SEPARATION
TUBE MECHANISM
CONTAINER

SEPARATION OF SAMPLE SAMPLING CONCLUDED

FOR ROOM A AND ROOM B (CATCHER IS CLOSED)

Figure 4. Sampling Sequence, (b) Sample Separation Process

X Z

CONVEYER GUIDANCE CONTAINER LID AND

CYLINDER REAR ABRATOR

CONDUCTION TUBE IS CLOSED AND CATCHER IS TRANSPORTED INTO

RETRACTED DOWNWARD. CONTAINER, AND CONTAINER LID IS
READY FOR TRANSPORTATION OF CATCHER LATCHED TO REENTRY CAPSULE.

Figure 4. Sampling Sequence, (c) Catcher Storage Process

SEAL ACTUATOR
POWER CABLE CONSTRAINT WIRE

CYLINDER
BACKWARD LID

CONTAINER IS SEALED BY SEAL MECHANISM. RESIDUAL CONVEYER FORCE IS

THEN, POWER CABLE FOR SEAL ACTUATOR RELEASED.
IS CUT. READY FOR CAPSULE SEPARATION

Figure 4. Sampling Sequence, (d) Container Sealing Process

NASA/CP—2002-211506 7
 Sliding Test

The sampler has been tested after the system vibration test onboard the spacecraft, and the whole
function has been verified. Also all actions including operations of SMA and pyrotechnic actuators have
been completed successfully in a thermal vacuum chamber.

A fundamental sliding test of the materials by test pieces has been carried out in a vacuum environment
before the manufacturing of the mechanism. Sliding parts of the sampler system have been also tested in
a sandy environment. The sliding test of the mechanism in a dusty or crunchy environment has two
purposes; one is to prove the operation of the sampler, and the second is to find the limitation of operation
under the dust. The clearance of every sliding part has been verified by sliding tests using fragments of
basalt with a wide range of grain size distribution. Basalt is a typical hard rock. The dust used in the test is
the collected fragment made by impacting projectile into basalt, and the particle size distribution of them is
expected to be the same as the actual sample on an asteroid if the asteroid is made of basalt. Excessively
severe conditions in the amount of dust may be involved, in order to know the limitation of function. The
test consists of the following four steps.

(1) The operation using the predetermined spring is repeated 100 times without dust, and the
variation of necessary operating force is measured.
(2) The mechanism with dust is operated using the predetermined spring, and the function is
verified for categorized particle size. The amount of dust, the particle size, and the relative
directions between sliding and gravity are varied. The number of test combination of these
parameters is about 100.
(3) The necessary operating force without dust is measured after the test step (2).
(4) After the test step (3), the necessary operating force for categorized particle size is measured.
The amount of dust, the particle size, and the relative directions between sliding and gravity are
varied.

The sliding tests have been applied to the revolving and retracting part, where the sample is separated
into each room by a revolving duct and the duct is retracted before conveyance of the catcher, and the
transportation part, where the catcher together with the rear abrator are conveyed linearly into the reentry
capsule. The results of sliding test steps (1)-(4) of the revolving and retracting part are shown in Tables
2(a)-(d), respectively. The results of sliding test steps (1)-(4) of the linear transportation part are shown in
Tables 3(a)-(d), respectively.

Table 2. Sliding Test of Revolving and Retracting Duct, (a)

Variation of Sliding Drag Force after Repeated Operation; Test Step (1)
Number of Operation Revolving Drag Force Retracting Drag Force
(*1) [N] [N]
10 16.9 13.0
20 29.8 14.7
30 26.7 10.4
40 32.3 16.5
50 40.4 19.2
60 24.5 20.0
70 41.0 17.6
80 41.6 18.4
90 33.7 25.4
100 31.8 21.0
(*1) Without Dust

NASA/CP—2002-211506 8
 Table 2. Sliding Test of Revolving and Retracting Duct, (b)
Operation Check under Dust Environment; Test Step (2)
Dust Axis of Gravity at Revolution Test (*1) Direction of Gravity at
Retraction Test (*1)
Particle Size Dust Amount X-axis X-axis Y-axis Y-axis Z-axis Z-axis X-axis Y-axis Z-axis
-6 -6 3
[10 m] [10 m ] (*2) (*3) (*2) (*3) (*2) (*3)
<53 1 O O O O O O O O O
10 O O O O O O O O O
53-125 1 O O O O O O O O O
10 O O O O O O O O O
125-250 1 O O O O O O O O O
10 O O O O O O O O O
250-420 1 O O O O O O O O O
10 O O O O O O O O O
420-590 1 O O O O O O O O O
10 O O O O O O O O O
590-840 1 O O O O O O O O O
10 O O O O O O O O O
840-1600 1 O O O O O O O O O
10 O O O O O O O O O
Mixed 10 O O O O O O O O O
70 O x O x O x x x x
O: Normal Operation
x: Locked
(*1): For the Direction of Gravity, See Figure 4(b).
(*2): 0deg-120deg Revolution
(*3): 120deg-240deg Revolution

Table 2. Sliding Test of Revolving and Retracting Duct, (c)

Variation of Sliding Drag Force after Operation Check under Dust Environment; Test Step (3)
Revolving Drag Force [N] Retracting Drag Force [N]
Before Test Step (2) After Test Step (2) Before Test Step (2) After Test Step (2)
22.4 23.9 17.9 16.6

Table 2. Sliding Test of Revolving and Retracting Duct, (d)

Variation of Sliding Drag Force for Dust Particle Size; Test Step (4)
-6
Particle Size (*1) [10 m] Revolution [N] Retraction [N]
Without Dust 23.9 16.6
<52 23.9 16.6
53-125 19.0 20.9
125-250 14.6 17.9
250-420 11.3 20.4
420-590 12.2 20.5
590-840 6.3 20.9
840-1600 14.7 33.0
Mixed 9.3 22.2
-6 3
(*1) Dust amount is 1x10 m .

NASA/CP—2002-211506 9
 Table 3. Sliding Test of Linear Conveyer, (a)
Variation of Sliding Drag Force after Repeated Operation; Test Step (1)
Direction of Horizontal Vertically Upward
Operation
Loading Position Center Eccentric on Eccentric on Center Eccentric
Upper Part Lower Part
Measurement Before After Before After Before After Before After Before After
(*1) (*2) (*1) (*2) (*1) (*2) (*1) (*2) (*1) (*2)
Drag Force (*1) 3.8 19.7 6.1 52.1 4.0 10.6 10.2 10.2 12.3 24.0
[N]
(*1): Before 100 Operations without Dust
(*2): After 100 Operations without Dust

Table 3. Sliding Test of Linear Conveyer, (b)

Operation Check under Dust Environment; Test Step (2)
Dust Direction of Gravity at
Sliding Test (*1)
Particle Size Dust Amount X-axis Y-axis Z-axis
-6 -6 3
[10 m] [10 m ]
<53 1 O O O
10 O O O
53-125 1 O O O
10 O O O
125-250 1 O O O
10 O O O
250-420 1 O O O
10 O O O
420-590 1 O O O
10 O O O
590-840 1 O O O
10 O O O
840-1600 1 O O O
10 O O x(*2)
Mixed 10 O O O
10 O O O
O: Normal Operation
x: Locked
(*1): For the Direction of Gravity, See Figure 4(c).
(*2): Locked at the first attempt, Normally operated at the second attempt.

Table 3. Sliding Test of Linear Conveyer, (c)

Variation of Sliding Drag Force after Operation Check under Dust Environment; Test Step (3)
Direction of Horizontal Vertically Upward
Operation
Loading Position Center Eccentric on Eccentric on Center Eccentric
Upper Part Lower Part
Measurement Before After Before After Before After Before After Before After
(*1) (*2) (*1) (*2) (*1) (*2) (*1) (*2) (*1) (*2)
Drag Force (*1) 12.0 24.2 28.8 x – 10.1 – 9.7 – x
[N]
x: Locked
– : No Test
(*1): Before Test Step (2)
(*2): After test step (2)

NASA/CP—2002-211506 10
 Table 3. Sliding Test of Linear Conveyer, (d)
Variation of Sliding Drag Force for Dust Particle Size; Test Step (4)
Dust Direction of Operation
Horizontal Vertically Upward
Dust Particle Size Loading Position
-6
Amount [10 m]
-6 3]
[10 m
Center Eccentric Eccentric Center Eccentric
on Upper on Lower
Part Part
0 Without Dust 24.2 x 10.1 9.7 x
1 Mixed 21.1 41.6 8.5 23.2 x
10 16.7 40.6 16.8 x x
1 <53 8.4 18.8 4.8 9.3 21.3
53-125 11.1 21.6 7.4 12.9 29.0
125-250 13.1 23.7 8.4 36.1 x
250-420 22.2 42.9 7.4 48.9 x
420-590 21.0 31.5 4.6 60.0 x
590-840 37.8 84.2 18.3 73.0 x
840-1600 37.7 70.8 22.2 53.2 x
x: Locked

Steps (1) of both of the parts resulted in the sliding drag force increasing as the operation is repeated,
roughly speaking. It should be noted in steps (4) of both of the parts that the dust of small particle size
might decrease the drag, which seems to be similar to a function of roller. The sliding tests results showed
that the mechanism worked under most of the dust conditions including excessive cases, and that only in
some cases it did not work, where the dust amount is unrealistic or the eccentric amount of the force
application point is unrealistic. The tests indicated the environmental limitation for operation and the
necessary operating force.

Pyrotechnic Shock Test

The container should be kept in vacuum in order not to reduce the samples scientific value. The container
suffers four types of shock load after it is sealed. The four types of shock are (1) capsule separation from
spacecraft, (2) parachute jettison from the capsule, (3) parachute deployment, and (4) landing. The shock
levels of them are shown in Figure 5. The former two are caused by pyrotechnic actuators, which are
installed in the reentry capsule system. The latter two shock environments are lower than the compressive
force capability of the seal spring as shown in the figure. The capsule separation shock (1) is enveloped in
the parachute jettison shock (2). This means that the container airtight has only to be verified against the
parachute jettison shock (2).

The high-frequency shock is applied to the container flange, and the vacuum in the container is monitored.
A helium leak detector is also used to monitor inside the container. The container was tested both in
atmosphere and in helium atmosphere. The test result says that the container lid does not open even if
the shock of parachute jettison is applied.

NASA/CP—2002-211506 11
 10000
Acceleration [Gsrs]

1000

100
Acceleration at Sampler Container
inside the Reentry Capsule
(1) Capsule Separation
(2) Parachute Jettison
(3) Parachute Deployment
(4) Landing
Seal Spring Capability
10
1 10 100 1000 10000
Frequency [Hz]

Figure 5. Mechanical Environment of Reentry Capsule

Conclusion

Sampler, the asteroid sampling mechanism for MUSES-C, has been developed and tested in prescribed
environments. Due to scientific requirements, only the selected lubricants are usable at the surface of
sliding parts, and the surface coating of structure that may touch sample has been selected. The function
of mechanisms has been verified, and the operation limit has been obtained. The seal of the sample
container has also been verified.

NASA/CP—2002-211506 12
 A Description of Mechanisms used in the Low Resolution Airglow and Aurora Spectrograph and
Special Sensor Ultraviolet Limb Imager

This paper was withdrawn because it was not approved before the publication deadline.

NASA/CP—2002-211506 13
 The IRAC Shutter Mechanism: Residual Magnetism and the Rotary Solenoid

* * * *
Scott Schwinger , Claef Hakun , Gary Brown , Ken Blumenstock

Abstract

The IRAC Shutter mechanism was originally presented in the paper, “A Low Power Cryogenic Shutter
th
Mechanism for Use on Infrared Imagers” at the 34 Aerospace Mechanisms Symposium, May 2000. At
that time, the shutter was believed to be performing flawlessly and there was every indication it would
continue to do so. In early spring of 2001, the calibration shutter, a rotary solenoid designed to be fail-safe
open, remained in a closed state with no power to the electromagnetic coils. The ensuing investigation,
subsequent testing, proposed remedy, and lessons learned are the focus of this paper.

Introduction

The Infrared Array Camera (IRAC) is one of three instruments on board the Space Infrared Telescope
Facility (SIRTF), the fourth “Great Observatory”. The SIRTF instruments will provide imagery, photometry,
and spectroscopy of astronomical bodies of interest over a spectral range of 3.6 to 160 µm. IRAC will take
1
images in four bands centered on 3.6, 4.6, 5.8, and 8.0 µm wavelengths . It is a cryogenic instrument
thermally coupled to a superfluid Helium dewar with an operational temperature of 1.4 Kelvin (K).

Shutter Background and Description

The IRAC shutter mechanism is required to block incoming
light energy, producing a dark internal environment, and to
allow viewing of internal calibration sources for IRAC to
calibrate its detectors. The two most important requirements
for the shutter, aside from the level of light attenuation the
shutter is to provide, are to exhibit low power dissipation to
limit cryogen depletion, and to be fail-safe open to avoid
loss of mission in the event of a failure in the closed state.
These two requirements factored heavily into the shutter
design and will be addressed further. The mechanism
(Figure 1) is approximately 0.15 m (5.9 in) long and has a
mass of 1.25 Kg. It translates a mirrored panel through an
arc of 0.663 rad (38 deg). The shutter is required to
attenuate incoming radiation by 1.0E+06. On the inside
portion of the panel, mirrored surfaces are diamond turned
to reflect an on-board calibration source through the
instrument optics onto the focal plane detectors. This
shutter is required to be fail-safe open, have a lifetime of
20,000 actuations, operate at 1.4 K, and dissipate less than
5 mW average power (0.5J/100s). It is required to be
redundant, and therefore, has two separate actuators, Side
A and Side B, located on opposite ends of the mechanism,
driving a common shaft as shown in Figure 2. Separate
electronic cards drive the Side A and Side B actuators Figure 1. IRAC Shutter Mechanism
independently. Only one card can be powered at a time and
only one actuator is required to close the shutter.

The suspension system of the shutter relies on a torsion flexure to provide axial stiffness, torsional
preload and restoring torque upon actuation (Figure 3). This flexure is chemically etched from a 0.660-

*
NASA Goddard Space Flight Center, Greenbelt, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 15
mm (0.026 in) thick sheet of Beryllium Copper (BeCu) 25AT. The flexure is attached to the shaft of the
shutter at its center point and is fixed at either end by components that fix them in rotation. As the shutter
closes, a restoring torque is produced in each half of the flexure. The nominal torsional preload in the
flexure in the open state is approximately 24.7 mN•m (3.5 in•oz). Closed, the flexure produces 49.4
mN•m (7 in•oz). Bushings at either end of the shaft provide radial stability. They limit radial motion of the
shaft during vibration and closing operations. Since radial forces approach zero with no current applied to
the actuators, friction at the bushing interfaces approaches zero as the shutter opens.
Electromagnetic Coil

Housing End Cap

Retainer Closeout

Rotor

Stator Stator
Shaft

Counterweight Shaft Hub

Figure 2. Shutter mechanism cross-section

The actuators are variable reluctance rotary solenoids designed at GSFC for the IRAC project. All
magnetic components are made of Hiperco 50A, a soft magnetic material with high magnetic saturation,
which was heat treated to minimize coercive force. With no permanent magnets in the circuit, the
actuation of the motor is insensitive to voltage polarity across the coil (i.e. these motors always pull, they
never push). The design of these motors consists of two stators and a rotor housed within an
electromagnetic coil. A closeout cylinder then surrounds the coil and joins the two stators at either end
(Figure 4) thus closing the magnetic circuit. The coil consists of 11500 turns of 38 gage 99.99% pure
copper wire. Coil resistance drops
( Side Cross-section)
from 2700 Ω at room temperature
to 17 Ω at 1.4 K. Low power
dissipation is achieved by virtue of
this resistance change and the
ability to drop the current input
Flexure End Square necessary to hold the shutter
Threaded Flexure Hub Cross-section
Piece Flexure
Clamp closed. The lower hold current is a
product of magnetic tabs that
extend normal to the stator faces
(Figure 5). These tabs are
contacted by the rotor in the
closed state and complete the
(Front View)
Mirror Mounting Surface magnetic circuit in the motor. The
closed magnetic circuit requires
dramatically lower current to
Figure 3. Shutter suspension system achieve the required flux level and
resultant torque to maintain a
2
closed state against the restoring torque of the torsion flexure. Since power to the coils is I R, lower
current means less power dissipation. The drive circuitry provides approximately -60 mA to close the

NASA/CP—2002-211506 16
shutter. The current is then reduced to -2.5 mA to maintain the closed position while the instrument runs
calibration procedures. Current is then removed to allow the shutter to open.

Tabs
Stator

Rotor

Stator

Latching Tab

Closeout

Electromagnetic Coil

Stator Rotor Stator

Figure 4. Exploded view of shutter actuator Figure 5. Shutter rotor and stator

Description of Failure

During ground testing at the SIRTF system level, the shutter was found to remain closed with no current
applied at the conclusion of one of the test sequences, thus violating the fail-open requirement. An
investigative team was assembled to determine if there were, in fact, a failure, the extent, repeatability,
and cause of the failure, as well as how to resolve it.

Events Leading Up To The Failure

During a review of the IRAC operations, it was discovered that the shutter was operated in an illogical
manner that, while not damaging hardware, stressed certain electrical components more than necessary.
The original shutter electronics and software design engineers had long since left NASA, so the design
philosophy of the command sequence is unclear. The electronics are primarily a closed loop current drive
with a solid state relay in series with the shutter coil output. When the relay is opened, the feedback
electronics are zeroed out and held in a low state. When the relay is closed, the system uses feedback
and a PID controller to track the current level command. The command is a 12-bit word and is scaled
such that the electronics can output up to +/-80mA in 40µA steps. The original opening sequence in
V3.13 of the software was:
- Simultaneously command zero current and open the relay (with 5mA of holding current flowing)
- 0.2 second later, close the relay
- 0.1 second later re-open the relay
This opening and closing of a relay driving current into an inductive load seemed inappropriate. Another
problem with the original software was that the sensor was not being operated properly. To correct the
sensor operation and reduce the stress on this relay, the shutter opening command sequence was
changed in V3.2 to:
- Command the current to zero (Wait at least 0.1 second)
- Open the relay
Soon after the implementation of V3.2, the failure occurred. Therefore, the first suspect was the software.
However, the investigation showed the software sequence that provided smoother, gentler current
profiles simply uncovered the low torque margins upon opening that were due to residual magnetism in
the shutter.

Investigations on Flight Hardware

The failure was first diagnosed while processing data taken during the last functional test before breaking
configuration. The first action was to thoroughly examine all data and telemetry taken during that test.
There was no evidence of improper commands or current anomalies, so the detailed investigation began.

NASA/CP—2002-211506 17
The next step was an attempt to determine the actual state of the shutter. Since the impedance of the
magnetic circuit changes, resistance, inductance, and capacitance measurements were made on the two
shutter coils and the sensor coils to determine the state. The shutter was verified to be open. However,
the shutter coils were inadvertently measured, including changing scales on the resistance meter, before
those of the sensor. In later investigations, it was determined the current spike produced by changing the
scale on the multi-meter had the ability to open the shutter if it did remain closed. It was never determined
whether the shutter opened at the end of the testing, during the move of the dewar, or during the
impedance measurements, but testing the sensor coil proved to be a valid means to verify position when
the instrument was powered off.

A set of tests, run on 6/3/01, instrumented the voltage and current waveforms to verify the functionality of
the electronics and to capture the differences in behavior using the various software versions. These
waveforms suggested that the electronics were healthy and appropriately responding to all commands
sent. They also verified the differences between the command software versions and gave insight to the
behavior of the shutter under various conditions. During the attempt to capture full opening and closing
waveforms, it was discovered that using a short pull-in and hold cycle would reproduce the failure
somewhat consistently. Figure 6 shows the current and voltage profiles of an actuator that fails to open at
zero current and one that opens properly. This command sequence proved invaluable in further

Figure 6. Current and voltage profiles for an actuator that fails to open at zero current
(left) and one that opens properly (right)

investigating the anomaly. The first round of tests included capturing the waveforms, sending individual
commands to prove the closed shutter could be opened by sending current of the opposite polarity,
measuring the pull-in and release currents on the shutter, and sending various static current levels to
properly calibrate the current output and telemetry test points for updating the ground data system. Figure
7 summarizes the release current data for the flight shutter. The results of the tests on 6/3/01 showed that
the hold currents on Side B were lower than expected, but did not require opposite polarity current to
open. This was supported by the fact that Side B operations NEVER led to the shutter remaining closed.
Side A current measurements showed that a slight current in the opposite polarity opened the shutter
reliably. The results also showed that the pull-in currents were virtually unchanged from previous data.
This suggested that the spring and shutter coils were healthy. Finally, the calibration tests revealed a +50
µA Side A offset and a +360 µA Side B offset when zero current was commanded. The current output
responded linearly to commands and the telemetry feedback equations were updated from a generic
approximation equation to side-specific equations which fit the data.

Once these tests were completed, a new and final version of the software was written. The new
command sequence added a commandable current pulse in the opposite polarity to counteract residual
magnetism, more time for current commands to settle, and more flexibility in command parameters. The
Version 3.21 opening sequence is as follows:
- Turn on the position sensor

NASA/CP—2002-211506 18
 - Command zero current then wait 1 second
- Send a commandable "kick current", +2.5mA, (Wait a commandable time, 3 seconds)
- Command zero current then wait 1 second
- Read the sensor and store the value
- Open shutter relay and turn off the sensor
This software version was implemented and the
electronics were instrumented on 6/14/01. The
waveforms verified proper operation and the failure
never repeated. Other tests were run on the flight
instrument including reverse polarity closings, Reverse polarity releases
unexpected loss of power, and attempting to use Side B
to release the shutter after being deliberately stuck
closed using Side A. The reverse polarity tests showed
pull-in and hold current comparable to the nominal
operations. This verified that the magnetic circuit fully
saturates regardless of polarity when the high, 60mA, Like polarity releases
pull-in current is applied. Side B always released during
a sudden loss of power, but Side A could be made to
stay closed under certain conditions. Unfortunately,
Side B operations were unsuccessful in releasing the
shutter once Side A was stuck due to residual
magnetism. This ultimately became the a concern of Figure 7. Release current data for the
SIRTF management and prompted the study into IRAC flight shutter
mitigation approaches which could be implemented on
the flight system.

GSFC Actuator Testing and Analysis

Considerable effort went into testing the 2 spare units at GSFC to duplicate the failure experienced on the
flight unit. The impacts of software and electronics board performance were eliminated as root causes of
the failure through extensive characterization. In order for the mechanism to remain closed with zero
current the following equation must be satisfied:
TRestoring Spring ≤ TResidual Magnetism + TFriction

Both the magnetic and frictional sources of torque in the mechanism which could cause the mechanism to
stick with zero current where investigated at length at both room and cryogenic temperatures. The
actuator components were also tested separately to characterize the release currents and torque
generated in an actuator without the influence of being in the more complicated shutter assembly. This
section of the paper will summarize the testing history, magnetic analysis efforts, and some of the key test
results obtained at GSFC. The results of these efforts confirm that the flight failure is due to intimate
contact between rotor and stator tab surfaces and the torque generated by residual magnetism.

History
Our initial suspicions were that the problem was related to residual magnetism in the rotary solenoid. The
theoretical possibility that enough residual magnetism could be developed in the actuator to overcome the
restoring torque of the spring, and thus not allow for failsafe operation, was an aspect of the design and
was anticipated from the beginning of the design process. The electronics were designed to allow for
reverse polarity currents to be applied to the actuator to cancel any residual magnetism and the potential
need for a degauss command was identified. The problem during initial testing prior to delivery to the
IRAC instrument was that we did not observe a zero current stick condition on any of the 4 units
developed at GSFC including the flight unit. However, we did see lower than desired hold currents on the
flight unit during initial assembly and testing. A non-magnetic shim was placed in the magnetic circuit to
maintain a fixed gap in the circuit in the closed condition. At the time, it was thought that this shim
provided enough coercive force on the magnetic circuit to effectively reduce the residual magnetism in the
actuator components after the 60 mA pull-in current was applied. After reassembly with this shim, desired
release currents were achieved. Unfortunately, after the flight failure occurred, a closer look into the

NASA/CP—2002-211506 19
placement of the shim revealed that it was not placed optimally and was subject to potential magnetic
shorting due to small variations in assembly or shifting of gaps during shutter operation.

Considerable frustration was endured during initial post failure testing

which showed very little effect of magnetics on the ability of the
shutter to remain closed with zero current. Upon examination of the
rotor and stator parts, a slight burnishing of the rotor where it
contacted the stator tabs was revealed (Figure 8). Clearly the majority
of the tab surface was not in intimate contact with the rotor. Only
when the contacting surfaces of a set of actuators were lapped
together (Figure 8) did it became apparent that our initial suspicions
were most likely correct. Approximately 12 µm (0.0005 in) of material
was removed during the lapping process. When these components Unlapped Burnishing
were assembled into a spare mechanism, the unit consistently failed
to open with zero current applied. This was a breakthrough in the
testing program. It showed that small variations in the nature of the
contact between the rotor and the stator have a significant impact on
residual magnetism and, therefore, the residual torque that the
actuator could exhibit.

Magnetic Analysis and Parameters

Considerable magnetic analysis was done on the actuator by Lapped
constructing 2-D and 3-D BEM models. Unfortunately, these models
proved of no use in understanding “small air gap” behavior. It was
suspected that large flux levels could be developed when the gap Figure 8. Rotor pre and
between rotor and stator tabs approaches zero. Since laboratory post lapping
measurements on spare units failed to duplicate flight residual
magnetic torques necessary to overcome the flexure torque (measurements were less than 14 mN•m (2
in•oz) compared with the flexure return torque of 50 mN•m (7 in•oz)) and result in a “stuck” condition, the
question arose as to what the theoretical maximum residual magnetic torque was, and moreover, how
this behavior was related to air gap distance. Stator/rotor alignment is one factor affecting the minimum
air gap and tab surface quality is another.

One analytic solution was utilized which assumed all the material, Carpenter Hiperco 50A, was
magnetized and that all the reluctance was in the air gap. Another case was run for the case of zero air
gap wherein the reluctance of the Hiperco was used to determine the maximum magnetic flux, Bmax.
Hiperco 50A, an alloy of nearly 50/50 iron and cobalt and a few other ingredients, was selected due to its
ability to carry high flux levels and minimize actuator size and mass. Figures 9 and 10 proved to be
interesting showing that the B-field at the tabs, and thus the torque generated by the tabs, rises rapidly as
the air gap reduces below 5 µm (0.0002 in).

Magnetic Flux Density vs. Air Gap Torque versus Air Gap

2 150
Magnetic Flux B

Torque (in-oz)

1.5
100
(Tesla)

1
50
0.5
0 0
0 0.0005 0.001 0.0015 0 0.0005 0.001 0.0015

Air Gap (inches) Air Gap (inches)

Figure 9. Analytic solution showing Figure 10. Analytic solution showing

relationship of flux with air gap relationship of torque with air gap

NASA/CP—2002-211506 20
Laboratory measurements indicated that providing a 2.5 mA reverse current, which allows the “stuck”
rotor to release, could essentially degauss a lapped unit. Providing too much reverse current resulted in
re-magnetizing the Hiperco in the reverse direction resulting, again, in a residual magnetic torque.

Key Test Results

The following sets of data will be summarized: the ability of residual magnetism to exhibit torques in
excess of the restoring torque of the spring, the ability to reduce the residual magnetism to a minimum by
applying a small current of the reverse polarity to the actuator, and the ability to measure the change in
residual magnetism due to this reverse polarity degauss current by measuring back emf.

Spare Unit Release Current (Pre Lapping)

The release current is the value of current at which the shutter opens due to the spring. The torque due to
friction was found to be negligible due to the design of the shutter mechanism. Measurements of release
current were made during the assembly process. The desired release current was 1 to 1.5 mA. After the
failure occurred, it was determined that the assembly process could have a major impact on the release
current. Machining tolerances were also shown to have a significant effect on residual torques. Better
alignment and tab contact between the rotor and stator lowers the release current and, therefore,
decreases the torque margin available to open the shutter.

Initial testing of the release currents on the spare units

showed lower but acceptable release current values in
some cases. None of the spare shutter assemblies or
actuator components exhibited enough residual magnetism
Intermittent
to overcome the restoring torque of the spring. However, a sticking
spare unit was then disassembled and reassembled. It was
noted that the assembly process could impact the release
current value. After one particular reassembly process, the
assembly did fail to open with zero current on rare
occasions. Figure 11 shows release currents for Side A and
B for the spare unit. It shows the Side A actuator had
intimate contact and better alignment between the rotor and
stator than Side B. The more “normal” release currents on
Side B indicate less tab contact and poorer alignment
between the rotor and stator in the closed position.

These measurements were made using a power supply at

room temperature. Approximately 60 mA was applied to the
actuator causing the shutter to close. The current was then
decreased until the shutter opened. The value of the release
current was then recorded. It is important to note that on the
rare occasions that a mechanism stuck, it always opened
when a slight reverse polarity current was applied. These
reverse polarity currents reduced the residual magnetism
torque to a level below the restoring torque of the spring
thus allowing the mechanism to open. Figure 11. Spare shutter release
currents
Actuator torque characteristics and optimal degauss currents (Bounding the Problem)
After the sensitivity to the nature of the contact between the tab of the rotor to the stator was fully
appreciated, we decided to improve this contact by lapping. Lapping caused the spare unit to fail to open
at zero current consistently. However, the mechanism still reliably opened upon application of a slight
reverse polarity current. We isolated the actuator components in a test set-up to simplify measurement of
actuator torque. Measurements of torque generated verses applied current were then made and are
shown in Figure 12. These measurements were performed both pre and post lapping. Prior to lapping of
the actuator components, intimacy of contact between the rotor and stator was not good enough to
generate significant residual magnetism torques. After lapping, measurements showed significant
residual magnetism torques. Values of two to three times the restoring torque of the spring were

NASA/CP—2002-211506 21
 Non-Lapped Torque vs. Current Lapped Torque vs. Current

7
30
6
25

Torque (in-oz)
Torque (in-oz)

5
20
4
15
3
10
2
5
1
0
0
-4 -2 0 2
-3 -2 -1 0 1 2
Release Current (mA) Release Current (mA)

Figure 12. Torque vs. current for non-lapped (left) and lapped (right) actuators
demonstrated. It is believed that the nature of the contact on the flight unit is somewhere in between the
pre and post lapped cases and that these two cases bound the problem. Therefore, the important concept
to note is that in the shutter mechanism assembly, the problem is bounded: i.e. higher residual
magnetism can still be cancelled with slight reverse polarity currents. The magnetic analysis confirmed
this bounding. There is no condition in which the mechanism can remain closed after sufficient reverse
polarity current is applied.

The efficiency of the actuator in the closed position increases as the contact between rotor and stator
becomes more intimate. Since the gap is smaller with improved contact, a larger percentage of the
possible residual flux due to saturation (~ 1.5 tesla) can be retained. At the same time, since the actuator
is more efficient, it takes a comparably less Ni of reverse polarity to cancel the residual flux in the
magnetic circuit. Thus, the problem is bounded.

Back EMF Measurements and Demonstration of Like Polarity Release Currents

There still remained the task of devising a test to prove that the failure of the flight mechanism to open
with zero current was due to residual magnetism and not friction. Friction was eliminated as a plausible
source of resistive torque through tolerance analysis and observations. Multiple degauss schemes were
devised to show that the residual magnetism could be affected by reverse polarity currents. In an actuator
with relatively poor rotor to stator alignment (i.e. a mechanism that does not stick), the effect of reverse
currents has minimal effect on release currents since the amount of residual flux is small. In addition,
since a relatively small value of current saturates and re-gausses the tabs, it is difficult to degauss the
actuator within the constraints of the flight mechanism. In fact, when constrained to access of one coil at a

Figure 13. Back EMF measurements before (left) and after (right)
applying a reverse polarity current

NASA/CP—2002-211506 22
time as in the flight set-up, it is virtually impossible. However, if both Side A and B are energized at the
same time it is possible to show that residual magnetism exists in a mechanism. It is also possible to
show that this residual can be removed by a slight reverse polarity current. Figure 13 shows the voltage
induced on the shutter coil during opening, or Back EMF, before and after a small reverse polarity
degauss current is applied. The figure shows that ~ 85% of the magnetism is removed. In fact tests at
cryogenic temperature showed that the tabs could easily be degaussed if the shutter was held closed
during degaussing. However, closing the shutter would re-gauss the tabs and generate the same residual
magnetism. Unfortunately, due to SIRTF project management concerns over risk and schedule we were
not able to duplicate these tests on the flight unit.

Proposed Solutions and Recommendations for the Flight System

Once the problem had been assessed, we considered several solution remedies to guarantee recovery
from all plausible failure modes. Opening the dewar and modifying the electronics were both deemed
unacceptable since the flight hardware was in its final delivered state. This imposed condition limited
options, but a simple harness modification solution was designed and tested. This "soft cross-strapping"
concept simply involved adding a resistor between the drive outputs of the Side A and Side B boards
(Figure 14). This resistor would be significantly larger than the coil resistance, but would allow a small
transient current to flow in the unused coil. Therefore, if a failure caused a shutdown and the shutter
remained closed on Side A, simply actuating Side B would provide enough current to Side A to erase the
residual magnetism. Since the cross-strapped current is a fraction of the primary drive current, it could be
controlled by command as well. Models were created and tests on an engineering unit were run to
determine an appropriate range of values for the resistor as well as effects on the un-powered electronics
board. The soft cross-strap approach successfully closed and opened the engineering unit shutter which
had been lapped to have maximum residuals on both Side A and Side B. Effects on the un-powered
board were negligible. Implementing the fix with a combination of series and parallel resistors, this
solution promised a fault tolerant, robust, and simple fix to the Flight Shutter problem. The only fault that
could not be corrected was if the Side A coil opened or shorted during a closing cycle using the Side A
electronics and the residual magnetism was high enough to keep the shutter closed. A thorough study of
the materials and fabrication techniques of the coil and harnessing indicated this scenario was extremely
implausible, but not absolutely impossible. However, the final decision of SIRTF management not was to
run further tests or implement the recommended fix on the Flight Unit.

CROSS-STRAP R CROSS-STRAP R

CROSS-STRAP R CROSS-STRAP R

+12V CROSS-STRAP R CROSS-STRAP R +12V

7

Solid State Relay Solid State Relay

R Harness R Harness
3 6 6 3

Voltage Buffer D D Voltage Buffer

18V R 18V
10K L L R
4

D 10K D
-12V 18V A Coil B Coil 18V -12V

R Harness R Harness
10 10
+ +
Current Sensing 4 R R 4 Current Sensing
OUT OUT
1 Ohm Inside Cold Assembly 1 Ohm
9 9
- -

Figure 14. Schematic showing soft cross-strapping

configuration

NASA/CP—2002-211506 23
 Lessons Learned

During the post failure investigation, several notable lessons in the design, fabrication, assembly and
testing of a space flight mechanism were learned. Errors during each phase of the evolution of a space
flight mechanism can go undetected. We are fortunate that this characteristic of the IRAC shutter
mechanism was discovered during ground testing. Problems and solutions that eluded detection during
each major phase of the IRAC shutter mechanism development will now be discussed.

Design
Designing a defined fixed gap into a closed magnetic circuit is a basic concept that should have been
implemented on the IRAC shutter mechanism rotary actuators. In fact, on a previous mechanism, the
DIRBE shutter (also designed by GSFC), this was accomplished by gold plating of the contacting
surfaces of the magnetic circuit. We understood that theoretically the design was sensitive to changes in
gap in the closed position. However, since testing of engineering units never revealed this sensitivity, this
possibility was largely ignored. Provisions were taken in the electronics to counteract this problem should
it occur. However, the implications of discovering this problem later in the program were not fully
appreciated during the early design phase.

In addition, material selection could have minimized the potential for this problem to occur. We chose
Hiperco 50A because of its stable properties from room temperature to 2K and for its high flux saturation
levels we thought we would need. Hiperco does have a relatively low coercive force when heat treated
properly. However, other materials, such as pure iron, can exhibit an order of magnitude lower coercive
force and also have stable properties over our temperature range.

Fabrication
Be careful what you ask for; you might get it. During fabrication of developmental units it is sometimes
tempting to use parts which may have been discrepant or have been made to tolerances that will later be
tightened for the flight mechanism. In this case, where a design is so sensitive to small variations this
could be a critical error in thinking. The parameters established by testing of early prototypes often
become a benchmark for the mechanism developed for flight. Therefore, tightening tolerances to improve
fit or function may invalidate previous qualification efforts and move the design to an unproven state.

Assembly
During the assembly of the flight unit, care was taken to align all of the contact surfaces in both the A and
B Side actuators. The A Side was aligned first and the B Side was then allowed to conform to the A Side
alignment. Measurements of release currents were made to ensure the desired value is obtained during
the assembly process. Release currents on the A Side are typically lower than the B Side. Just prior to
integration into the IRAC cold assembly, the flight shutter mechanism was disassembled for cleaning.
Upon reassembly, it underwent a characterization at 4K and was then integrated into the cold assembly.
Not until after the failure occurred was it appreciated how much of an impact the assembly process could
have on the release currents and, therefore, the reliability of the mechanism. It is theorized that the final
assembly of the flight mechanism and subtle shifts during operations and vibrations caused the A Side
actuator to become marginally “sticky” and detection of its behavior eluded us. Therefore, it is critical to
understand the impact of the assembly process on the performance of a space flight mechanism.

Testing
Even if extensive test data indicates no problem with a mechanism, if flaws exist in the test program or if a
design error is being masked, there may still be a problem waiting to surface. In this case, continued
system level characterization of the shutter mechanism was limited due to conflicts with other sub-system
demands such as software checkout, optical alignment, etc. It is believed that the failure to open with zero
current on the A side actuator was present but not detected throughout system level testing at GSFC. The
lesson learned here for a mechanism engineer is not to assume your job is done when a mechanism is
delivered to the system level. Do not bow to schedule concerns. Make sure that the mechanism is
functioning properly at the system level. Make sure that the proper check-outs can be implemented at the
system level.

NASA/CP—2002-211506 24
The influence of maintaining a schedule, minimizing costs and accepting some level of risk are always
present during a project. Many of the lessons learned could have been realized if the development effort
had not yielded to these pressures. Although it is inevitable that these factors will always influence the
development process, the mechanism designer must ensure that every aspect of their development
process is validated by theory, analysis, and test.

Conclusions

Disposition of Shutter Failure Investigation

After the extensive test effort, a SIRTF project peer review was held where the following conclusions and
recommendation were presented by GSFC to a panel of independent experts:

• Failure to open at zero current is due to residual magnetism in the shutter mechanism actuator
• The cause of the residual magnetism is a combination of two factors:
o An improvement in rotor to stator tab alignment when the shutter was reassembled just prior
to installation
o The revised shutter open sequence in software v3.20 augmenting previously unseen
characteristics
• The sensitivity of shutter release current to its mechanical alignment was not previously well
understood and unintended magnetic effects were revealed following the software change from v3.13
to v3.20.
• Shutter characterization tests since May 15th have indicated stable performance – no changes in
performance were observed pre and post CTA vibration testing.
• Residual magnetic forces are highly sensitive to rotor to stator alignment and gaps
• Residual magnetism can readily be offset by a small current flow in reverse polarity to the direction of
actuation
• The Flight Mechanism has no “defect” and is operating nominally with current understanding of
mechanism behavior and residual magnetism.
• It has been shown that Side A shutter can require a slight reverse current to open. Side B has always
opened at zero current.
• The shutter can be reliably closed and opened with v3.21 commands.
• In the event of a failure of the primary drive electronics, a simple resistor “soft cross-strap” can
provide a reliable secondary current source to open the shutter from the redundant drive electronics.
• A simple means to provide a small reverse current flow, even in the event of primary electronics
failure has been developed
• The case of an open coil winding has been studied and determined to be non-credible
• Flight shutter Side B has been shown to be completely failsafe in all test and operational scenarios

Recommendations
• Continue use of the flight shutter on both Sides A and B where B is the primary side and A is the
secondary side
• Implement the Soft Cross-Strap Circuit to provide a backup release means for both actuators

The independent review panel concurred with all of the conclusions and recommendations that GSFC
proposed. They agreed with the proposed testing and added that a mini life test be performed to ensure
the reliability of the mechanism. GSFC was prepared to follow through with the proposed cross-strapping
modification and necessary testing. Unfortunately, due to programmatic concerns, SIRTF project
management deemed there would be no further shutter operations and the proposed soft cross strapping
modification would not be implemented. An operational scheme will be used which will allow calibration of
the detectors without using the shutter mechanism.

The IRAC shutter is a prime example of how secondary effects or subtle changes in the design,
assembly, and testing of a flight mechanism can mask or reveal undesired behavior. As stated above, the
ultimate conclusion of the investigative team is that the shutter is behaving as one should expect the
design to behave. It is behaving nominally as built, but not as intended. Its “anomalous” behavior is, in

NASA/CP—2002-211506 25
reality, an unfortunate byproduct of improvements in command sequences, tightly toleranced machining,
precision alignment, and skilled assembly. At first glance, these are all desirable traits for any mechanism.
However, inconsistencies in any one of these areas during the qualification phase of a development can
foster a false sense of security and result in surprises during a phase when more rigorous control is
applied. Fortunately, due to the commitment of a talented group of individuals, a hidden problem for this
mechanism was found and a potential mission ending failure avoided.

References

1. Schwinger, D. Scott and Claef F. Hakun. “A Low Power Cryogenic Shutter Mechanism for Use on
th
Infrared Imagers.” Goddard Space Flight Center, Greenbelt, MD. 34 Aerospace Mechanisms
Symposium, Greenbelt, MD, May, 2000

Acknowledgments

Ted Ackerson/GSFC – Quality assurance

Marty Brown/OSC – I&T
Bill Burmeister/Ball Aerospace – Ball SIRTF Lead I&T
Giovanni Fazio/SAO – IRAC PI
Rainer Fettig/Raytheon – Magnetics and mechanics
Jose Florez/GSFC – Electronics
Leslie Frazier/SAO – Quality assurance
Sid Johnson/GSFC – Mechanical design, assembly, test
Robert Kichak/GSFC – Failure investigation team chair
Ed Lander/ Raytheon – I&T
Norm Lee/GSFC – Quality assurance
Steve Mann/Hammers – Software
Ahmean Mekhail/OSC – I&T
Shirley Paul/OSC – Quality assurance
Lois Workman/GSFC – GSFC IRAC project manager

NASA/CP—2002-211506 26
 A Cryogenic Half-Wave Plate Module to Measure Polarization at Multiple FIR Passbands
+
Timothy S. Rennick*, John E. Vaillancourt**, Roger H. Hildebrand , and Stephen J. Heimsath*

Abstract

One of the key components in a far-infrared polarimeter that is being designed at The University of
Chicago is a locally-powered half-wave plate module. This compact, lightweight, and reliable module will
operate at cryogenic temperatures, rotating a half-wave plate about its axis within the optical path. By
doing so, polarization measurements can be made. Further, by utilizing multiple half-wave plate modules
within the polarimeter, multiple wavelengths or passbands can be studied. In this paper, we describe the
design and performance of a relatively inexpensive prototype module that was assembled and tested
successfully, outline the difficulties that had to be overcome, and recommend improvements to future
modules. This effort now lays some of the groundwork for a next-generation polarimeter for far-infrared
astronomy.

Introduction

Instruments designed for far-infrared (FIR) and submillimeter astronomy generally operate with the
detectors and most of the optical elements at cryogenic temperature. When the design calls for inserting
any of several lenses or filters into the optical path, there must be suitable actuators, bearings, and
sensors for the moving parts. The actuators have usually consisted of external, room temperature, motors
connected to the moving parts by insulating shafts passing through vacuum seals and working their way
through tight spaces at times via right angle gearing to the cold parts. This approach has been workable
for relatively simple instruments, but has become increasingly difficult as the instruments have become
more complex.

Figure 1. Exploded view of HWP module

*
The University of Chicago, University of Chicago Engineering Center, Chicago, IL
**
The University of Chicago, Dept. of Astronomy & Astrophysics, Chicago, IL; now with the University of
Wisconsin, Dept. of Physics, Madison, WI
+
The University of Chicago, Dept. of Physics, Dept. of Astronomy & Astrophysics, Enrico Fermi Institute,
Chicago, IL
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 27
At The University of Chicago, we are designing a far-infrared polarimeter to cover a wide range of
wavelengths. The telescope that this polarimeter is intended to work in conjunction with is the
Stratospheric Observatory for Infrared Astronomy (SOFIA) (Ref. 1). At the heart of the polarimeter, a
rotating carousel-style stage will be used to put any one of five polarizing half-wave plates into the optical
path. It must then be possible to rotate the selected half-wave plate (HWP) about its own axis to conduct
the polarization studies. It does not appear practical to produce the required HWP rotation on a carousel-
style stage using shafts from external motors. We have therefore carried out investigations on a prototype
module in which a cryogenic motor is located immediately adjacent to the rotating HWP, eliminating the
need for protruding shafts and external motors. This prototype module is composed of the HWP holder
with its bearing, motor, gearing, and housing (shown in Figure 1) and is designed to hold a lens, a
spectral filter, and a pupil stop attached to the housing. The module, with its accompanying optical
elements, is designed to be easily replaceable, allowing different options for wavelengths and
magnification.

The final instrument will have five such modules on a carousel-style stage. Such an arrangement
envisioned for measuring multiple FIR passbands is shown in Figure 2. For a sense of scale, each HWP
has a clear aperture of approximately 75 mm diameter.

Figure 2. Carousel-style stage supporting multiple HWP modules

In this paper, we describe the design and performance of the prototype HWP module, outline the
difficulties that had to be overcome, and recommend improvements to future modules.

Module Design

Specific objectives of the design effort were to produce a prototype HWP module that was compact,
lightweight, relatively inexpensive, and reliable, particularly since it would not be accessible during
operation. The module had to operate at both room temperature and near the temperature of liquid
helium (near 4 K). Additionally, a means of sensing the rotational position of the half-wave plate had to be
included. It was desired that such position information be independent of counting step commands sent to
the cryogenic motor. During normal operation, the HWP will be stepped once a minute in 30-degree
increments that take approximately 5 seconds, corresponding to a duty cycle of approximately 8%.

NASA/CP—2002-211506 28
Another requirement was that the motor heat, at the normal duty cycle, should not increase the boil-off
rate of the liquid helium reservoir significantly and thereby impact observation time.

Cryogenic Motor
Price and recommendations were major considerations in the selection of the cryogenic motor. In
addition, the particular motor selected had to be small, lightweight, run on relatively low power, but yet
have sufficient torque, and be easily controllable. A general design envelope defined the extent of the
motor size. Simple computations were performed to assess the torque requirements based upon HWP
inertia. These computations accounted for accelerating the small rotating mass of the HWP and its
holder. The results indicated very low values of motor torque were needed. It was anticipated, though,
that friction in the gearing and bearing, which could not be reliably predicted, would be the major motor
load. To boost the output of the motor, an integral planetary gear unit, with a ratio of 25:1, was selected.
Both the motor and planetary gearing came as one unit, fully wired, internally lubricated, and ready to run.

HWP Bearing
A low profile bearing was chosen to support the HWP in the module because of space limitations.
Further, the eventual implementation of this HWP module would be on board an aircraft, exposed to a
flight environment. To minimize jitter of the HWP and possible microphonic effects picked up in the
instrument’s detectors, a precision X-bearing (i.e. four point contact) was selected. With a bearing of this
type in a cryogenic application, it was critical that the material of the races match that of the ball bearings.
Stainless steel alloy 440C was selected for both.

Gearing
Since the HWP was at the center of the module and required a clear aperture, the motor that drove it had
to be offset. This required gearing, besides the planetary gear unit, to transfer the motion. A worm gear
arrangement was chosen, even though this type of gearing suffers from large amounts of friction,
because it was conducive to the space available and had a relatively high gear ratio to further boost the
output torque and step count. For example, the stepper motor is 200 steps/rev; the planetary gear ratio is
25:1; the worm gear ratio is 25:1. Thus, the theoretical step count for a complete revolution is 200 x 25 x
25 = 125,000 steps. The rotational accuracy desired for the polarization studies was ±1 degree or better,
and this was thought to be easily achievable.

Magnetoresistive Position Sensor

To independently sense the HWP position, small cryogenic sensors were investigated. A
magnetoresistive sensor from Infineon Technologies was recommended and samples were provided. The
sensor was a differential magnetoresistive sensor (type FP 212 L 100) and consisted of two series
coupled InSb/NiSb semiconductor magnetoresistors, whose resistance could be magnetically controlled,
mounted on an insulated ferrite substrate (Ref. 2). A permanent magnet, which supplied a biasing
magnetic field, was fixed to the base of the sensor. The entire sensor was contained in a plastic housing
and had three connecting terminals that were subsequently soldered to a small circuit board for mounting
and wiring. The basic room temperature resistance of the sensor was 2 x 100 Ω. These particular sensors
were selected and had been tested successfully on a related project to a temperature of 12 K, which gave
us some confidence that the sensors would operate at the temperature that we desired. Not only could
these sensors detect position, but because there were two magnetoresistors, the device as a whole could
sense direction; although, we did not choose to use this capability in our prototype testing. The sensor
was quite small at about 6 mm x 3 mm and 0.35g, and it was relatively easy to implement. A 5 V potential
was continuously applied, and output signals were on the order of tens to hundreds of millivolts. A small
electronic circuit was created to amplify the output signal, and a commercial off-the-shelf digital counter
was used to record movement of the HWP. Since the precise HWP location information was not
necessary for the prototype testing, sampling of the HWP position was conducted only once per
revolution by attaching a small ferromagnetic chip on the rotating HWP holder to actuate the stationary
sensor (by altering the magnetic field of the permanent biasing magnet and eliciting a response from the
magnetoresistors) each time it passed. Future implementation of this sensor will likely incorporate more
exact information on HWP position, including any incremental amount of rotation and possibly a sense of
direction.

NASA/CP—2002-211506 29
Aluminum Housing
With many dissimilar materials used in the module, there were obvious concerns over interference and
binding due to shrinkage as the module was cooled to its eventual operating temperature. After
considering several designs, it was decided that the best approach was to accommodate the shrinkage
within the assembly tolerances, particularly since the dimensions involved were not all that great and
shrinkage was mostly concentric about the HWP. Taking this approach had the disadvantage of creating
a “sloppier” mechanism than desired at room temperature. However, by surrounding the stainless steel
HWP bearing with an elastomeric o-ring to center it and fill the purposely-open gap, the module was able
to operate satisfactorily at both room temperature and at cryogenic temperatures. The size of the o-ring
(nominal 1.6-mm diameter) and depth of the groove in the housing were selected so that the o-ring just
contacted the outer bearing race at room temperature and then experienced roughly 10% diametral
squeeze upon cooling.

A listing of all components within the assembled module is shown in Table 1. As can be seen, many of
the components were commercially available, generally reducing cost associated with the prototype. The
total mass of the assembled module was 851g, an amount that is probably more than desired for the
future flight design.

Table 1. Components of HWP module

Component Supplier Part Number Approx. Cost Material
($US)
Stepper Motor Phytron, Inc. VSS 19.200.0.6 4400 Stainless steel housing
Planetary Gearing Phytron, Inc. VGPL 22/25-UHVC-X a Stainless steel housing
HWP Bearing Kaydon Corporation Reali-Slim SA025XSO 380 Stainless 440C
Precision Worm W. M. Berg, Inc. W32S-3F 66 Stainless 303
Precision Worm Gear W. M. Berg, Inc. W32B30-F100 50 Bronze Alloy 464
Magnetoresistive Sensor Infineon Technologies FP 212L-100 b (580) Plastic housing
Housing Univ. of Chicago NA ~ 5000 Aluminum 2024
Shaft Univ. of Chicago NA c Stainless 303
Shaft Bearings W. M. Berg, Inc. B1-26-Q3 ABEC3 18 Stainless 440C
Shaft Coupling McMaster-Carr 59925K88 90 Nickel
Flange Adapter Univ. of Chicago NA c Stainless 303
HWP Retainer Univ. of Chicago NA c Aluminum 2024
HWP Holder Univ. of Chicago NA c Stainless 440C
Gear Retainer Univ. of Chicago NA c Stainless 440C
a – included in cost of motor
b – sensor donated by Infineon Technologies, cost in parenthesis associated with circuit board and digital counter
c – included in cost of housing

Figure 3 shows a front and back view of the prototype module and sensor once assembled and attached
to a bracket for testing.

NASA/CP—2002-211506 30
 Figure 3. Front and back view of prototype module with magnetoresistive sensor

In addition, the module was tested first without and then with a modified tungsten disulfide dry lubricant,
®
Dicronite DL-5, applied to the worm gearing in hopes of reducing gear friction and lessening the motor
load and, thus, heat produced by the module. This lubricant was selected primarily because of its low
cost, availability in the Chicago area, and its advertised characteristics that appeared to match our needs.
Because of time and cost concerns, no testing or extensive investigation was conducted to screen
various lubricants and determine the most effective. This would be worthwhile, though, in the future.

Module Testing

Testing of the prototype HWP module was conducted in a cryostat that was originally designed as a far-
infrared polarimeter for use on the Kuiper Airborne Observatory (Ref. 3). The cryostat has a dual stage
3 3
He refrigerator system capable of reaching temperatures as low as 300 mK. However, the He system
was not used in our tests, and only cooling from the liquid helium (LHe) reservoir was used to obtain
temperatures near 4 K. The cryostat has two radiation shields at LHe temperature, one at liquid nitrogen
(LN2) temperature (77 K), and an outer vacuum shell at room temperature.

The magnetoresistive sensor was wired using low thermal conductivity stainless-steel wire inside the
cryostat. The high current needs of the cryogenic motor (specified at 0.6 A per phase), however,
precluded the use of low conductivity wire. Therefore, copper wire was used inside the cryostat to supply
power to the motor. By observing liquid helium boil-off rates, we estimated the extra heat conducted from
room temperature to the 4 K surface along these 4 copper motor wires (2 per phase) nearly doubled the
heat reaching the 4 K surface. This was an initial concern when the idea to relocate the motor inside the
cryostat was conceived. As such, heat loads on the 4 K surface were a major focus of the testing. In the
future, more attention will be focused on heat-sinking the motor leads to the radiation shields and possibly
taking advantage of a vapor-cooling effect by routing the leads in the vicinity of venting helium gas.

Goals of HWP Module Testing

The goals of the module tests were two-fold. First, and most importantly, would the concept of a locally-
powered half-wave plate function at cryogenic temperatures and would it do so repeatedly and reliably?
Second, what amount of power/heat would be dissipated at the 4 K stage while the module was
operating?

Our astronomy application requires rotating the HWP by 30 degrees in about 5 seconds, the equivalent of
1 rpm. Given that one revolution of the HWP is approximately 125,000 motor steps this would necessitate

NASA/CP—2002-211506 31
a nominal stepping frequency of about 2 kHz. However, accounting for the additional time to start and
stop the motor, the 2 rpm (4 kHz) stepping rate seemed to result in a time closer to the required 5
seconds for a 30-degree increment.

An unexpected obstacle was encountered following initial assembly and during the module’s first stages
of testing. This obstacle required lengthy troubleshooting to overcome. The module would operate
properly at room temperature, but not at cryogenic temperatures even though the design had been
carefully reviewed. Through a process of elimination, possible culprits such as the stepper motor,
clearances on the aluminum housing, and even the magnetoresistive sensor were eliminated. It was
determined that the HWP bearing was seizing due to either warpage of the bearing races or slight
material variations in the stainless steel used throughout the bearing. The bearing supplier worked with us
to incrementally change the bearing's diametral clearance by reducing the ball size until unrestricted
rotation was observed while submerged in a bath of liquid nitrogen. Apparently, the bearing’s level of
precision was too close to withstand the large change in temperature. Once this obstacle was overcome,
further testing proceeded smoothly.

Testing in the bath of liquid nitrogen proved to be a quick, low cost approach to checking cryogenic
operation of the mechanism, particularly when troubleshooting the seized bearing. Fiberglass (G10) rods
were used to manipulate the components in the bath, and the temperature of the component could be
inferred by monitoring the boiling rate. In fact, at one point, the entire module (minus motor and sensor)
was completely submerged in a bath of liquid nitrogen and verified to spin freely.

Further testing was performed once the motor was attached to the gearing and aluminum housing.

Minimum Motor Current

The motor specifications call for running the motor at a nominal current of 0.6 A per phase. At each
temperature (room, nitrogen, and helium) we measured the minimum current required to turn the HWP
module. At room temperature, the system spun freely in both directions and at both the 1 and 2 rpm
speeds with a minimum current of 0.2 A. At liquid nitrogen and liquid helium temperatures the system
required 0.2 A to run at 1 rpm, and 0.3 A to run at 2 rpm. Fundamentally this makes sense: More power is
required to cover an equal distance in a shorter period of time. With more accurate instrumentation, we
may have seen this effect at room temperature as well.

At room temperature, the resistance across the motor coils was measured to be ~3 Ω; when cooled with
liquid nitrogen the resistance was 1 Ω; and at liquid helium it was 0.6 Ω. Therefore, given the minimum
2
currents found previously, the motor should theoretically be able to operate at 2 rpm dissipating (0.3 A) x
0.6 Ω = 54 mW of power per phase (108 mW total). At 1 rpm, this value would drop to 48 mW total. A
large fraction of this power would be dissipated through heating of the motor coils.

Endurance Testing
Successful endurance tests were run at both liquid nitrogen and liquid helium temperatures. These tests
were conducted with a motor current of 0.3 A at 2 rpm. The module was rotated through 720 30-degree
steps of the HWP with five seconds of rest between each step. Additionally, tests were conducted with
the module continually operating. These particular tests were conducted at liquid helium temperature and
dissipated a significant amount of heat into the reservoir of liquid helium. As a result, the flow rate out of
the venting helium gas that was liberated increased significantly. Additionally, the motor coil resistance
gradually increased from 0.6 Ω to 0.8 Ω, indicating that current was warming the motor. This type of test
was conducted more as a measure of the module reliability, because such operation would not be
normally experienced.

Dry Lubricant
®
In an attempt to lower the heat dissipated by the motor a dry lubricant, Dicronite DL-5, was applied to the
worm and worm gear. The minimum current required to turn the HWP at liquid helium temperature was
again found to be 0.3 A at 2 rpm and 0.2 A at 1 rpm. The amount of heat reaching the helium surface due
to the motor and HWP module was measured by examining the boil-off rate of the liquid helium both
before and after applying the dry lubricant. The motor was allowed to run continuously at 2 rpm and 0.3 A

NASA/CP—2002-211506 32
per phase for 20 minutes. The measured change in helium boil-off rate as a function of time is shown in
Figure 4. (The heat is also estimated in mW, using 2.6 J/ml for the heat of vaporization of helium.) As can
be seen, there appears to be little difference. However, had the module run longer until equilibrium was
achieved, a more significant difference might have been apparent. For the application at hand, though,
®
with a relatively low duty cycle under normal operation, the Dicronite DL-5 lubricant will not be of great
benefit.

0.35 250

0.3

200
power off
Change in boil-off rate of liquid helium (l/hr)

0.25

Estimated heat change in mW

no lubricant
150
0.2 tungsten disulfide

0.15
100

0.1

50

0.05

0 0
0:00 0:15 0:30 0:45 1:00 1:15 1:30 1:45 2:00

Time (hh:mm)

®
Figure 4. Effect of Dicronite DL-5 lubricant on worm gear

Duty Cycle Variations

With the lubricant applied to the worm gearing, another series of heat load tests were performed to
determine the expected heat load from continuous rotation of the HWP, from stepped rotation of the
HWP, and from the magnetoresistive sensor alone. For each scenario, the boil-off rate of liquid helium
was allowed to come to equilibrium. Figure 5 shows the extra heat on the 4 K surface for each scenario in
terms of boil-off rate and estimated milliwatts. In Table 2, the measured equilibrium values are
summarized.

As can be seen, the cryogenic motor when run continuously produces more heat than would be predicted
from input electrical power (recall the 108 mW total previously computed). Obviously, from Figure 5, the
measured heat load from the motor is nearly twice that expected. Since input power should theoretically
equal the motor’s output plus losses, this surplus heating is not clearly identifiable. Creating a more
efficient mechanism could certainly reduce this heating some. However, utilizing a reduced duty cycle
seems to be the most effective means of limiting heat loads from the cryogenic motor and rotation of the
HWP. Fortunately, for the intended polarimeter, this is possible.

NASA/CP—2002-211506 33
 0.35 250
power off (cont.)

0.3

200
Change in boil-off rate of liquid helium (l/hr)

0.25 sensor only

Estimated heat change in mW

5 sec every 55 sec
5 sec every 25 sec 150
0.2 motor continuous

power off (sensor)

0.15
100

0.1

power off (25 sec) 50

power off (55 sec)
0.05

0 0
0:00 0:15 0:30 0:45 1:00 1:15 1:30 1:45 2:00 2:15 2:30 2:45 3:00

Time (hh:mm)

Figure 5. Effect of variations in duty cycle and sensor heat load

Table 2. Equilibrium heat loads

LHe boil-off Estimated heat % increase
(l/hr) (mW) above baseline
No electronics baseline 0.37 267 -
Magnetoresistive sensor alone 0.14 101 38%
Continuous motor w/o sensor 0.32 231 86%
Motor 5 sec every 55 sec w/o sensor 0.03 22 8%
Motor 5 sec every 25 sec w/o sensor 0.05 36 14%

Sensor Heat Load

In Figure 5, it can also be seen that the magnetoresistive sensor adds a large amount of heat to the 4 K
surface relative to normal operation of the cryogenic motor. The magnetoresistive sensor consists of two
100 Ω resistors in series at 5 V potential. It should therefore dissipate about 125 mW of power at room
temperature; using a measured resistance of 0.34 Ω at LHe temperature, it should dissipate 73 mW of
power. This correlates reasonably well with the results shown in the graph and table, giving a sense of
the accuracy of the measurements made. Because of the relatively high heat dissipation by the sensor,
future module design and sensor implementation may require either a different sensing device or
necessitate a duty cycle consistent with the motor.

NASA/CP—2002-211506 34
 Conclusion

Some conclusions that can be drawn from this work are:

• Testing of the prototype HWP module was successful, exhibiting confidence in a new approach to
measure polarization at multiple FIR passbands.
• Selection of a bearing for cryogenic application is a critical step. A precision X-bearing apparently
over constrained and bound the mechanism when significant cooling occurred, necessitating
looser diametral clearance between the balls and races.
• Centering the HWP bearing with an elastomeric o-ring allowed both room temperature and
cryogenic operation when dissimilar materials were present and tolerances factored in to account
for shrinkage.
• For the application at hand, with a relatively low duty cycle under normal operation, the
Dicronite® DL-5 lubricant applied to the worm gearing will not be of great benefit.
• The continuously energized Infineon magnetoresistive sensor heat load was higher than desired,
indicating a need to investigate other sensing means or employ a low duty cycle on the sensor,
similar to the motor, to reduce this heat load.
• Future work will be focused upon reducing the module mass and lowering the heat load further.

The success of this small project now lays some of the groundwork for a next-generation polarimeter for
far-infrared astronomy.

Acknowledgements

We would like to thank G. Kelderhouse and H. Krebs of The University of Chicago, A. Kepley of Case
Western Reserve University, W. Lahmadi and J. Lagdam of Phytron, Inc., W. Tenbrink of Kaydon
Corporation, and E. Shankland of Infineon Technologies. This work has been supported by NSF grant
9987441. J. Vaillancourt was supported by NASA Graduate Student Research Program grant NGT 5-63.
A. Kepley was supported by NSF funding through the Center for Astrophysical Research in Antarctica
(CARA) Research Experience for Undergraduates (REU) program.

References

1. Becklin, E. E. “Stratospheric Observatory for Infrared Astronomy (SOFIA).” Proceedings of the ESA
Symposium The Far Infrared And Submillimetre Universe 15-17 April 1997, Grenoble, France, ESA
SP-401 (August 1997), 201-206.
2. Differential Magnetoresistive Sensor FP 212 L 100, Data Sheet 1999-04-01, Infineon Technologies.
3. Platt, S. R., R. H. Hildebrand, R. J. Pernic, J. A. Davidson and G. Novak. “100-µm Array Polarimeter
from the Kuiper Airborne Observatory: Instrumentation, Techniques, and First Results.” Publications
of the Astronomical Society of the Pacific, 103 (November 1991), 1193-1210.

NASA/CP—2002-211506 35
 Tape-Spring Rolling Hinges

Alan M. Watt∗ and Sergio Pellegrino*

Abstract

This paper presents a new design of a low cost, unlubricated, self-deploying, self-locking hinge whose
properties can be easily modified to meet different requirements. A particular implementation is
considered, providing a hinge that is 135-mm long, 30-mm high and 45-mm wide, with a deployment
moment varying between 0.1 and 0.3 Nm and a locking moment of 13 Nm. Stiffness tests have been
carried out on the hinge in the deployed configuration, and it is shown that the six stiffness coefficients
can be estimated using simple analytical models. The moment versus rotation profile of the hinge is
shown along with the results found from a finite-element simulation. The results of deployment testing,
including shock imparted upon latching of the hinge, are presented for hinges with a variety of damping
mechanisms.

Introduction

A number of space-based deployable structures that are being developed at present, such as membrane
synthetic aperture radars and ultra-high power solar arrays, involve the use of stiff members connected by
self-locking hinges. The design of such frames can be simplified and their cost greatly reduced by the use
of continuous, elastic connections instead of standard mechanical hinges.

This paper presents a new design for a Tape-Spring Rolling (TSR) hinge (Figure 1), which is a
combination of steel tape springs, mainly providing the deployment and locking moments, and a rolling
hinge consisting of two sets of ''wheels'' held together by wires wrapped around them. These wires are
held within grooves cut into the body of the wheels and are free to pass from one wheel to the other when
the wheels are rotated. This very low friction arrangement is sometimes known as a ''rolamite hinge''
(Chironis and Sclater, 1996), and has well-defined kinematic properties. Note that there is no sliding, only
rolling contact, and that the kinematic behavior of the hinge is determined only by those parts of the
rolamite wheels which come into contact; the rest of the wheel can have an arbitrary shape.

(a) (b)
Figure 1. TSR Hinge

The proposed design can be readily tailored to meet different applications and, in particular, is more
compact and lighter than earlier designs. The stiffness, moment-rotation properties, and damping
behavior have been characterized and are presented in this paper.

∗
Department of Engineering, University of Cambridge. Cambridge, UK
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 37
 (a) (b)

Figure 2. Configurations of tape-spring hinges

Review of Related Hinge Designs

Simple, self-actuating, self-locking hinges have been developed for a number of years for use as
deployment mechanisms for solar array panels, synthetic aperture radars (SARs), booms, radiators and
the like.

Several successful designs have made use of curved elastic elements (tape springs). Tape-spring hinges
offer a number of benefits over standard hinges involving relative motion between rigid mechanical parts
that make them particularly suitable for use in deployable space structures:

• Zero friction and backlash.

• Elastic latching into locked position, giving highly repeatable and accurate positioning.

• No moving parts to jam or bind due to long-term storage or adverse environmental conditions.

• Simple manufacture and low cost.

Vyvyan (1968), Figure 2(a), patented a method of increasing the locking moment produced by a tape-
spring hinge by placing the tape-springs in a parallel configuration, hence putting them in overall
compression and tension rather than bending. An alternate arrangement is shown in Figure 2(b)
(Chiappetta et al., 1993) where, unlike Figure 2(a), the tape-springs come into contact upon folding.

A major disadvantage inherent in the use of tape-spring hinges is that they have very low stiffness in the
folded configuration. This can cause uncertainties in deployment and makes gravity compensation during
ground testing problematic.

Aerospatiale (Auternaud et al., 1992) proposed a solution to this drawback by attaching a rolling hinge to
the tape-spring, as shown in Figure 3. A functionally similar design has been manufactured by TRW
Astro. Both designs provide better deployment control than the previous hinges, however both of these
hinges were developed primarily for solar array panels, and hence are too wide for many other
applications. Also, the Aerospatiale hinge is heavy (1.17 kg) and complex, whereas the locking stiffness
of the TRW Astro hinge is expected to be quite limited.

NASA/CP—2002-211506 38
 Figure 3. Aerospatiale hinge

A rolling hinge was developed by Hilberry et al. (1976), Figure 4, and functions by holding two rolling
surfaces in contact by means of tensioned bands. As the hinge rotates, the bands pass from one rolling
surface to the other and their curvatures change signs. There is therefore no sliding contact within this
hinge, thus making friction very low and removing the need for lubrication.

Figure 4. Rolling hinge

Hilberry et al. (1976) also describe how the performance of a rolling hinge can be improved by changing
the profile of the rolling surfaces or of the surfaces supporting the tension bands. For instance, the force
pulling the two halves of the hinge together is increased if the bands run on a surface with a smaller
radius than the rolling surface, Figure 5(a). Alternatively by making some of the tension band surfaces
smaller than the others, as in Figure 5(b), a restoring moment is created upon rolling the hinge in the
direction shown.

New Hinge Design

The new hinge design, first presented in Pellegrino et al. (2000), introduces two fundamental design
changes:

• A double tape-spring is used. This produces a higher deployed bending stiffness and a much
higher locking moment, both of which can be tailored to any particular application by varying the
spacing of the tapes. Thus the complex locking mechanism of the Aerospatiale hinge is no longer
required.

NASA/CP—2002-211506 39
 (a) (b)

Figure 5. Alternative forms of rolling hinge

• Steel bands in the rolling hinge are replaced by wires. This reduces the overall width of the
mechanism and hence its weight, and allows a much simpler and more efficient wire tightening
mechanism to be utilised.

The main geometric characteristics of the hinge are defined in Figure 1 and are as follows:

• The length, L, of the tape-springs, measured between the ends of the clamps.
• The separation distance, s, between the tapes, i.e. the distance between the neutral axes of the
two springs.
• The offset distance, d, between the center-line through the tape-springs and the center-line
through the rolamite wheels.
• The radius, r, of the rolamite wheels.

The particular TSR hinge that is investigated in this paper has L = 88 mm, s = 12.5 mm, d = 11mm and
r = 28.1 mm giving overall dimensions of 135 mm by 45 mm by 30 mm. It weighs 0.11 kg including all
connections. The dimensions of the hinge are constrained by the requirement that the tape springs should
not be damaged during folding/unfolding. The smallest hinge that has been constructed with the same
tape springs has dimensions of 106 mm by 44 mm by 21 mm.

The new TSR hinge can be seen in Figure 6. The main parts of the rolling hinge are manufactured from 8-
mm thick Delrin (a space-qualified Acetyl Resin) plate. Aluminium-alloy blocks connect the rolling parts to
the tape springs. Nylon coated, 1-mm diameter stainless steel wire terminated by crimped aluminium
tubes is used to hold together the rolling parts; tension adjustment is provided at one end by the wire
passing through a screw with a lock-nut on the end. The pre-stress in the wires is sufficiently large that
compressive contact between the rolamite wheels is always maintained. The tape springs are cut from
25.4-mm wide, 0.1-mm thick, transverse radius 15 mm, ''Contractor Grade'' steel tape measure, supplied
by Sears Roebuck and Co.

TSR hinges with two or even three stacked tape springs also work well, and both the deployment and
locking moments increase roughly proportionally, but note that the hinge described in this paper is made
from a single pair of tape springs.

Deployed Stiffness

In order to estimate the natural frequency and stiffness of a structure utilizing TSR hinges, the linear
stiffnesses (Kxx, Kyy and Kzz) and the rotational stiffnesses (Txx, Tyy and Tzz) of the hinge are required; the
directions of x, y, and z are defined in Figure 6. The stiffnesses of the component parts of a hinge were
both measured and predicted with various linear elastic analytical and finite element models. A summary
of all the test results and analytical predictions is given in Table 1. The test results for the overall stiffness
of the hinge in each component direction can be seen in Figure 7.

NASA/CP—2002-211506 40
 Figure 6. TSR hinge

Table 1. Summary of Stiffness Results

Experiments Predictions
Direction Rolamite Tape Total Rolamite Tape Total Units
Kxx 1768 5400 7223 1272 12172 13442 N/mm
Kyy 31.9 236 221 47 702 749 N/mm
Kzz 115 10 134 72 33 106 N/mm
Txx 40 29 75 45 54 99 Nm/rad
a
Tyy 0 228 480 0 475 614 Nm/rad
Tzz 360 480 782 779 453 1232 Nm/rad
___________________________
a
This particular value was measured on a hinge with d =4.5 mm.

Analytical predictions for the stiffnesses were obtained using the expressions given below. Note that each
equation contains two separate terms, which correspond to the rolamite and tape-spring stiffness
contributions, respectively. The definition and numerical value given to each term in these expressions
are listed in Table 2. The finite element predictions were slightly more accurate, but are not presented
here because they are much harder to obtain and cannot be readily transferred to other hinge designs.

The axial stiffness, Kxx, was predicted by modelling the rolamite (subscript r) part of the hinge with two
equivalent beams whose axial stiffness takes into account the compliance of the hertzian contact between
the pairs of wheels (Johnson, 1987); the tape-springs (subscript t) were also modelled as two beams:
−1
 L 4  4π r w E r*  2 At E t
K xx= +  ln − 1 +
 
 r r r π w E r
2b d E * Q L
 r 
Here Qr is the contact force, whose value does not significantly affect the results in the range considered.

NASA/CP—2002-211506 41
 Figure 7: Hinge response in six stiffness tests.

NASA/CP—2002-211506 42
The in-plane shear stiffness, Kyy, and the out of plane shear stiffness, Kzz, were predicted by considering
the rolamite and tape-spring parts of the hinge as two separate built-in beams. Hence:
E b d3 12E t l zz
K yy = 2 r 3r r + 2
L L3
E b 3d 12E t l yy
K zz = 2 r 3r r + 2
L L3
The torsional stiffness, Txx, was predicted by considering the end deflections, in the z-direction for the
rolamite and the y-direction for the tapes, induced by a unit torsional rotation of the hinge. Each deflection
causes associated shear forces in the equivalent beams defined above, from which the required twisting
moment is:
E b 3d h 2 12h 2Et l zz
Txx = 2 r r 3r +2
2L 2L3
The in-plane bending stiffness, Tyy, was predicted by considering that a unit rotation about y causes
extension and compression of the tape-springs. The rolamite has zero stiffness on its own, as this is the
direction of free rotation, but it provides an elastic constraint for the tape springs acting as a single beam,
which leads to an additional stiffness contribution. Hence Tyy is:
−1
 L 4  4π r w E r*   2 At E t  −1  At E t s 2
Tyy =  + ln − 1+   d 2
+
 2br d r E r π w E r*  Qr   L  
 
2L
The out-of-plane bending stiffness, Tzz, was predicted by considering that a unit rotation about z puts a
pair of rolamite wheels into tension and the other pair into compression, whilst the tape-spring was
considered as a built-in beam subject to an end rotation. The resulting expression is:
−1
d2  L 4  4π r w E r*  E t l zz
Tzz = r  +  ln − 1 +2
2 π E r*  
 r r r
2b d E w  Qr  L

Moment-Rotation Properties

The moment-rotation relationship of TSR hinges is required in order to model the dynamic deployment of
any system utilizing these hinges. Seffen and Pellegrino (1999) have developed analytical expressions for
the key parameters determining the moment-rotation behavior of tape-springs. From these expressions,
the ''steady-state'' deployment moment of a TSR hinge with a single pair of tape springs is given by:
E t t 3α
M =
6 (1 − ν 2 )
Where t and ν are the thickness and Poisson's ratio, respectively, of a tape spring and α the angle
subtended by its cross-section, in radians. Note that this expression does not include the effects of
contact between the two tape springs and of the constraint imposed by the rolamite wheels; hence the
actual deployment moment is usually larger and also non-uniform.

Greater accuracy requires a fully non-linear numerical formulation to be adopted. Hence, a quasi-static
simulation of the folding of a TSR hinge was made with the finite-element package Abaqus (Hibbit et al.
2000). The purpose of the FE modelling was to accurately simulate the snap-through deformation of the
tape springs and to derive the full bending moment-rotation relationship of the hinge, including the
buckling moment.

NASA/CP—2002-211506 43
 Table 2: Terms used in Analytical Expressions.

Term Definition Value Units

At Cross-sectional area of tape 2.5 mm
br Width of equivalent rolamite beam (z direction) 10 mm
d Center-to-center distance between rolamite wheels and tapes 11 mm
dr Depth of equivalent rolamite beam (y direction) 8 mm
2
Er Young’s modulus of rolamite wheels (Delrin) 3.1 kN/mm
Contact modulus = Er / 2(1 − ν 2 )
2
Er * 1.6 kN/mm
2
Et Young’s modulus of tape (steel) 210 kN/mm
h Width of rolamite hinge 35 mm
nd 4
I yy 2 moment of area for one tape (y-axis) 95 mm
nd 4
I zz 2 moment of area for one tape (z-axis) 95 mm
4
L Length of tape and equivalent rolamite beam 88 mm
Qr Contact force between rolamite wheels 20-800 N
r Radius of rolamite wheels 28.1 mm
s Separation of tape neutral axes 12.5 mm
w Width of rolamite contact area 10 mm

Therefore, a full 3D model was set up, Figure 8. The tape springs were modelled using 50 x 12, 4-node
doubly curved general-purpose shell elements (s4) for each spring. These elements were generated with
logarithmic bias along the tape length so that the finer mesh is concentrated in the middle of the tapes,
where most of the deformation before the snap and the contact between the tapes take place.

Figure 8. Finite element model of TSR hinge

The rolling part of the TSR hinge was modelled as a set of two rigid arms, using rigid beam elements. The
arms connect the centers of rotation, B and C, to nodes A and D, and multi-point constraints were defined
between these nodes and the nodes at the end of the tapes. Nodes B and C are fixed in all directions and
can only rotate around an axis parallel to y. In order to simulate the hinge deformation, equal clockwise
and anti-clockwise rotations of up to 90 deg were applied to nodes B and C, respectively. Contact
between the tapes was modelled as a surface to surface contact.

The Riks solution procedure was initially chosen, in order to trace the complete equilibrium path of the
hinge, including unstable parts. However, there were problems with convergence after the first snap-
through point, hence a rotation controlled solution procedure had to be adopted instead. The ''Stabilize''
function available in Abaqus was used, which automatically switches to a pseudo-dynamic simulation
when an instability is detected. This method gave the desired convergence, but it should be noted that
unstable parts of the response cannot be predicted by this method.

The finite element predictions were validated against experimental results obtained from an ESH Torsion
Machine. Figure 9(a) shows the testing arrangement; the head of the testing machine applies a rotation to
the center of one of the rolamite wheels, whilst measuring the moment, and the center of the other wheel
is held in a bearing coaxial with the wheel. The measured response of the hinge during folding and
unfolding has been plotted in Figure 9(b) along with the FE predictions. The measured peak buckling

NASA/CP—2002-211506 44
moment ---not shown in the figure--- was 13 Nm, which compares with a prediction of 19 Nm. However,
during deployment the maximum moment was 1 Nm, much lower than the 12 Nm predicted by the FE
simulation.

(a) (b)

Figure 9. Moment-rotation test set-up and results

Damping

Damping can have two functions in a hinge; to slow the time required for deployment and to reduce the
shock arising from locking. In order to test the effectiveness of different damping methods, a mock-up of a
deployable panel system on a satellite was made from two identical aluminium honeycomb panels. Panel
1 is attached to a rigid base; panel 2 is connected to panel 1 by two identical TSR hinges. Each panel has
dimensions of 1 m by 0.5 m, and weighs 1.67 kg including hinges and hinge fittings. The whole set up is
shown in Figure 10.

Deployment tests were conducted for three different damping configurations, as shown below:

1. No additional damping.

2. Two brown Oasis foam blocks placed between the two panels, as shown in Figure 11(a), to
absorb energy by crushing the foam during the final approach phase.

3. Single layer of 3M 434 sound damping tape applied to both sides of each tape spring, as shown
in Figure 11(b).

Deployment Tests
The deployment of the panel was recorded using a Kodak EKTAPRO HS 4540 high-speed video camera
and digitized to obtain rotation vs. time graphs. The results can be seen in Figure 12 for case 1
(undamped) and case 3. Case 2 gives exactly the same results as case 1, as the foam does not affect the
large-scale rotation behavior. It can be seen that the damping layers make a negligible difference to the
rotation vs. time response.

A model of the deployment of the panel was made using Pro/Mechanica Motion (Parametric Technology
Corp., 2001), a rigid-body dynamic analysis program. The loads applied to this model were those found
from the Abaqus model of the hinge, taking into account the direction of motion and the state ---i.e., fully
unfolded or not--- of the hinge.

NASA/CP—2002-211506 45
 Figure 10. Deployable panel

(a) (b)

Figure 11. Damping foam and damping tape

Figure 12. Deployment of panel with damped and undamped hinges

NASA/CP—2002-211506 46
The rotation vs. time results from this model can be seen in Figure 13 along with the results from the
undamped test. The deployment is modelled with good accuracy up to the point of locking, but
subsequently the simulation shows the hinge unlocking and rotating back to a rotation of around 20 deg
and then oscillating backwards and forwards a number of times, before finally locking. Note that the
current hinge design can turn only in one direction, and hence in the opposite direction it is able to resist a
very large moment.

Figure 13. Comparison of rotation vs. time response to Pro/Mechanica simulation

This difference in the post-locking behavior could be due to a number of factors, such as:

• Damping within hinge. At locking there could be significant damping present due to stretching of
the wires, compression of the hinge wheels, stretching of the tapes or slipping of the connections.

• Energy loss due to buckling.

• Incorrect modelling of the moment-rotation properties of the hinge where the moment-rotation
properties of the hinge vary from those predicted by the Abaqus analysis.

• Resistance during panel deployment. This resistance could arise due to friction within the hinge
or air resistance on the panel.

Shock Measurements
In addition to measuring the rotation of panel 2, four accelerometers were attached to the apparatus, two
to each panel, in the positions shown in Figure 14. The distance from the hinge attachment to the
accelerometers was minimised to measure the peak shock levels. The accelerometer outputs were
logged at 5000 Hz using a PC with an analogue to digital converter board and a program written in
LabView (National Instruments, 1998).

The shock resulting from deployment of a panel with hinges without additional damping can be seen in
Figure 15. The hinges do not lock fully on first deployment but re-buckle twice; hence a total of three
acceleration peaks can be seen in the plots, each corresponding to the tape springs snapping into the
2
deployed configuration. The maximum acceleration is approximately 1500 m/s (150 g) and there is little
difference between accelerations in different directions.

The shock resulting from deployment of a panel with two 12-mm long pieces of Oasis foam can be seen
in Figure 16. The size of the foam blocks was determined by setting the kinetic energy of the system on
latch-up (equal to the strain energy in the hinges, given by the area under the moment-rotation graph)
equal to the energy required to crush the foam to a size that allows full deployment of the panel. The
panel now locks the first time, without re-buckling the tape-spring hinges, and the maximum acceleration

NASA/CP—2002-211506 47
 2
is reduced to approximately 600 m/s (60 g). Successful deployment of the panel was found to be very
sensitive to the length of the foam blocks used. For example, when the length was increased to 12.5 mm
the panel did not lock; ideally a larger piece of less dense foam with a lower crushing stress should be
used.

Figure 14: Accelerometer positioning.

The shock resulting from the deployment of a panel with 3M 434 sound damping tape attached to both
2
sides of each tape-spring can be seen in Figure 17. The maximum shock is now around 250 m/s (25 g)
in all directions.

Conclusions

The hinge presented in this paper is significantly lighter and smaller than previous designs. It also
provides predictable moment-rotation and stiffness properties. It has been used as the deployment
mechanism in full-size verification models of SARs and Solar Panels.

The deployed stiffness properties of the hinge can be predicted analytically with good accuracy, and the
equations that have been presented can be used to tailor the stiffness properties of the hinge to a given
set of requirements.

Modelling the full moment-rotation relationship of the hinge with good accuracy has to take into account
the effect of contact between the tapes, which was achieved with a finite element analysis. A relationship
obtained thus has been used to successfully predict the deployment dynamics of a structure containing
TSR hinges, however the prediction of the amount of energy dissipated within the hinge during latching is
an area where further work is needed. From a practical viewpoint, it has been found that the addition of
damping tape is more effective at reducing shock than damping foam. Damping tape gives a six-fold
reduction in the imparted shock over an undamped hinge. None of the damping systems tested have any
effect on the large-scale rotation-time properties.

Acknowledgments

The work presented in this paper was funded by the British National Space Centre and technically
supported by ASTRIUM UK Ltd. We thank the technical monitor, Mark Roe, for advice and many valuable
suggestions. Financial support from the Science and Engineering Research Council, in the form of a
studentship for A.M.W. is gratefully acknowledged.

NASA/CP—2002-211506 48
 Figure 15. Locking shock for panel with no damping

Figure 16. Locking shock for panel with 12-mm Oasis foam

NASA/CP—2002-211506 49
 Figure 17. Locking shock for panel with 3M 434 sound damping tape

References

Auternaud, J. et al. (1992). Self motorized antifriction joint and an articulated assembly, such as a satellite
solar panel equipped with such joints, U.S. Patent 5,086,541. February 11th, 1992.
Chiappetta F.R., Frame C.L., Johnson K.L. (1993). Hinge element and deployable structure including
hinge element, U.S. Patent 5,239,793. August 31, 1993.
Chironis, N.P. and Sclater, N. (1996). Mechanisms and mechanical devices sourcebook. Second Edition,
McGraw-Hill, New York.
Hibbit, Karlsson and Sorensen (2000). ABAQUS Version 6.1.0. Pawtucket, RI 02860.
Hilberry B.M. and Hall A.S. (1976). Rolling contact prosthetic knee joint, U.S. Patent 3,945,053. March
23rd, 1976.
Johnson, K.L. (1987). Contact Mechanics, Cambridge University Press, Cambridge.
National Instruments (1998). Labview Version 5.0.1.
Parametric Technology Corporation (2001). Pro/Mechanica 2001, 140 Kendrick St, Needham, MA 02494.
Pellegrino, S., Green, C., Guest, S.D. and Watt A. (2000). SAR Advanced Deployable Structure.
CUED/D-STRUCT/TR191, Department of Engineering, University of Cambridge.
Seffen, K. A. and Pellegrino, S. (1999). Deployment dynamics of tape springs. Proceedings of the Royal
Society of London, series A, 455, 1003-1048.
Vyvyan, W.W. (1968). Self-actuating, self-locking hinge, U.S. Patent 3,386,128. June 4th 1968.

NASA/CP—2002-211506 50
 Material Property Effects on Coaxial Cable Mechanical Failure
* ** ***
R. B. Pan , J.B. Chang*, C.C. Wan*, Y. R. Takeuchi*, R. McVey , and I. Chen

Abstract

A spaceflight coaxial cable experienced mechanical failure at 0.1xlifetime during the qualification life test.
Destructive physical analysis (DPA) performed on the failed cable revealed signs of massive cold welding
between the silver plated copper wire braid and the silvered plated copper foil wrap. Also observed was the
foil-to-foil cold welding in the foil wrap overlap regions. Hardness measurements indicated low hardness
values in all copper elements that included the wire braid, foil wrap and center conductors. This paper
presents the DPA results and postulates a failure mechanism that relates low yield strength associated with
the low hardness values to the cold welding of similar material at the interface, rapid initiation of multiple
crack sites and fast crack propagation.

Introduction

A 4.8-mm (0.19-inch) diameter coaxial cable, fabricated by Manufacturer A and identified as A190A,
failed at 0.1xlifetime during the qualification life test conducted in late 1999. Figure 1 shows the helically
coiled configuration of the coax harness. A bundle of direct current (DC) twisted shielded pairs, seen
behind the coax harness, is installed through the center of the helix. This entire harness assembly is
mounted inside the shaft of a direct drive torque mechanism, one end attached to the shaft and the other
attached to the housing. The life test motion profile was designed to evenly distribute direction reversals
and degrees of motion throughout the entire mechanism travel range of ± 150 degrees. One life consists
of 120-million degrees of travel and 1.5-million direction reversals. The requirement for a flight coaxial
cable in the qualification life test is successful completion of 2 operational lives.

Figure 2 shows the test article placed inside the thermal vacuum chamber. A successful test runs
-6
continuously for 2 months, with pressure maintained at less than 1x10 torr and temperature of the coax
o o
harness controlled between 40 and 120 F. Throughout the life test, critical data such as radio frequency
(RF) performance, hysteresis torque plots, and DC continuity are recorded on a daily basis.

This type of coaxial cable harness is used in space for high frequency, low-loss applications. The cable
consists of five elements: a fluorinated ethylene propylene (FEP) outer jacket, a silver-plated copper wire
braid, a silver-plated copper foil, a polytetrafluoroethylene (PTFE) tape-wrapped dielectric core, and a
strand of twisted silver-plated copper center conductor. Figure 3 shows the cross section view of the
A190A coaxial cable.

A literature survey uncovered a similar coaxial cable failure reported by Chiu (Reference 1). Cold welding
apparently caused some damage to the outer jacket of the coaxial cables and resulted in unacceptable
RF performance.

*
The Aerospace Corporation, El Segundo, CA
**
Boeing Satellite Systems, El Segundo, CA
***
Raytheon Systems, El Segundo, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 51
 Figure 1. Test Harness Removed from Mechanism, Post Life Test

Figure 2. Mechanism/Harness Assembly in Thermal-Vacuum Chamber

NASA/CP—2002-211506 52
 Figure 3. Cross Section View of A190A Coaxial Cable

Failure Investigation

Inspection of the failed A190A coil revealed that the wire braid, the foil wrap and the FEP jacket were all
broken at a common location (Figure 4). Signs of massive cold welding between components were
observed in the broken region. Destructive physical analysis (DPA) was performed in order to seek the
root cause of the failure. Samples cut from both the coil section (CS) and the straight section (SS) of
A190A were examined with low power optical microscopy, fine-focus x-ray and scanning electron
microscopy (SEM) techniques. Signs of massive cold welding were found at the wire braid to foil wrap
interface in A190A/CS (Figure 5). In a few places, wires from the wire braid bonded so strongly to the foil
wrap that they could not be separated from the foil wrap without damaging the foil. An example of wire
braid-to-foil type cold welding was illustrated in Figure 6. After the wire braids were removed from the
coaxial cable sample, multiple circumferential cracks were observed along the overlap boundary of the
wrapped foil (Figure 7). It was found that the foil was difficult to separate during subsequent foil unwrap
attempt. In some cases the foil was torn during the unwrap operation. SEM analysis on unwrapped foil
surface clearly indicated that a foil-to-foil type cold welding had taken place in the edges of the foil overlap
region. However, in the non-flexing section of the coil, i. e., test sample A190/SS, no sign of either wire-
to-foil or foil-to-foil type cold welding was found.

NASA/CP—2002-211506 53
 Figure 4. Failed A190A Coil Showing Fracture pf PTFE Jacket, Wire Braid and Foil Wrap

Figure 5. Sign of Massive of Cold Welding Shown between Wire Braid and Foil
after Removal of Wire Braid

NASA/CP—2002-211506 54
Helically formed, 3-mm (0.120-inch) diameter cables, fabricated by Manufacturer A, were also obtained
for this investigation. These cables had passed 2.0 lives in the life test and were identified as samples
A120. Cracks along the edge of foil overlap were again found in the A120/CS test samples although the
crack length and the number of cracks appeared to be shorter and less in frequency than those found in
sample A190A/CS. However, there was no sign of wire braid-to-foil cold welding in the A120/CS sample.
Like in the case of A190A/SS sample, neither wires braid-to-foil nor foil-to-foil cold welding was found in
A120/SS sample. Table 1 summarizes the inspection results.

Microhardness measurement using an applied load of 25 grams was conducted on all copper elements, i.
e., wire braids, foil wrap and center conductor, of the A190A and A120 samples. Hardness measurement
was also conducted on samples from two other 4.8-mm (0.190-inch) diameter cables manufactured at
earlier dates for comparison. These samples were identified as A190-’97 Lot and A190-’80 Lot. As a
reference, hardness measurement was done on samples, identified as B197, of coaxial cables fabricated
by Manufacturer B. Table 2 summarized the hardness test results from this investigation. Here, the
microhardness values were converted to standard Rockwell hardness B (RB) values for convenience.
Results showed that, except for samples from the early production A190-’80 Lot, the RB values of cables
supplied by Manufacturer A are generally lower than the average RB values from cables supplied by
Manufacturer B. The most noticeable difference is in the foil wrap of the failed A190A/CS sample.

Figure 6. Braid Wire Still Bonded to the Foil after Braid Removal

Failure Mechanisms Postulation

For the samples taken from the straight section and the coil section of the failed cable, and other tested
cables, the following are observed:

(1) Straight section samples do not show any sign of cold welding, possibly because the foil overlap
regions are relatively free to move.
(2) Coil section samples always show signs of cold welding in the foil overlap regions after life test.
(3) The coil section of the failed cable experiences massive cold welding between the wire braid and foil
interface.

NASA/CP—2002-211506 55
(4) Hardness values of wire braid, foil and center conductor from the failed cable are low, especially
those values measured in the foil.

Table 1. Inspection Results

SAMPLE/ TEST HISTORY INSPECTION RESULTS

SECTION

Failed at 0.1x life • Massive cold welding was found at wire

A190-A/CS braid/foil and foil/foil region
• SEMs taken from foils revealed many cracks

A190-A/SS Failed at 0.1x life No sign of cold welding

A120/CS Passed 2.0x life Signs of cold welding and tears were found along
the edge of foil overlaps
A120/SS Passed 2.0x life No sign of cold welding

Table 2. Rockwell Hardness B (RB) Value Comparisons

MEASURED WIRE BRAID FOIL CENTER

SAMPLES CONDUCTOR
1
A190A/CS 45.6 ± 2.2 22.7 ± 2.6 46.5 ± 2.0
A190-’97 Lot 44.8 ± 2.1 46.9 ± 1.7 53.6 ± 1.8
3
A190-’80 Lot 66.2 ± 1.4 47.3 ± 1.6 74.9 ± 0.8
2
A120 47.0 47.0 42.0
B Cables 59.6 49.5 62.8
(Ave of 8 Lots)
B197 58.7 49.6 62.7
Note: (Reference 2)
1. Yield Strength ≈ 69 MPa
2. Yield Strength ≈ 275 MPa
3. Yield Strength ≈ 345 MPa

Figure 7. Multiple Crack Sites Develop in the Foil

NASA/CP—2002-211506 56
Based on the observations, a failure mechanism has been postulated. It starts with the cold welding of foil
wrap in the overlap region during coil flexing. As explained in Reference 1, the cold welding is a
molecular-level bonding between two atomically clean smooth surfaces when they are in close contact.
Silver is highly susceptible to cold welding in a vacuum environment because silver oxide is unstable and
tends to break down in such an environment. As a result, there is no oxide barrier between the two
contact surfaces in the foil overlap region. Due to appreciable contact stress during coil flexure and dither
in a vacuum, cold welding is thought to occur in the foil overlap region. This effectively increases the foil
thickness in the overlap region and thus creates a stress concentration along the edge of foil overlap
during coil flexure. With continued cycles of flexure, cracks will initiate. These cracks will then propagate
along the edge of foil overlap, which is essentially in the direction normal to the principal stress. Torsional
stresses may play a minor role here also. When the local bending stress is relatively small, it will take
many flexure cycles to cause foil failure. This failure mechanism is believed typical for high-cycle fatigue.

When massive cold welding forms at the wire braid-to-foil interfaces, these elements will bond together in
a number of places. Thus there will be regions where the local bending stiffness increases. The classical
strength of material analysis methodology as described in Appendix A was used to estimate the effect of
wire bonding on the effective stiffness and bending moment of the cable. Since coil flexing is a
displacement controlled motion at the support points, the increase of local stiffness will cause the foil in
adjacent regions to be highly strained. Figure 8 shows a plot of the number of wire braids in a layer of the
190A cable vs. the bending moment predicted by the analysis. The bending moment needed to bend the
cable to an increased radius of curvature results in higher bending stress, which may also contribute to an
increase in crack propagation rate in the foil. The wire braid can experience some failures without
significant RF impacts, but the foil must retain its dimensional integrity.

From linear elastic fracture mechanics (LEFM) theory, for a given crack size, “a”, its fatigue crack growth
rate can be expressed by Paris Law (Reference 3):
n n
da/dN = C(∆K) = C(∆σ √πa )

where “a” is the crack dimension; “N” is the number of fatigue cycles; “∆σ” is the remotely applied tensile
stress range; “C” and “n” are the Paris crack growth rate coefficient and exponent respectively.

For copper material, the exponent n is 3.2 (Reference 4). Thus if there are 20 wires per layer (28% of
total wires per layer) bonded to the foil, there will be a 40% increase of applied bending moment (bending
3.2
stress) which will cause crack propagation rate to be increased by a factor of (1.4) ≈ 3, for a crack with
identical initial size. If there are 50 wires per layer (70% of total wires per layer) bonded to the foil, an
increase of applied bending moment (bending stress) will be a factor of two that could cause the crack
3.2
growth rate to increase of (2) ≈ 9.

When a predominant crack propagates to its critical size and causes breakage of the foil, the load
induced by the flexing movement will be redistributed and thus possibly cause breakage of the wire braid
and eventually the outer jacket as seen in the A190A failure. The sequence of failure is not well
understood.

The low hardness values associated with the foil wrap and the wire braid of A190A/CS are thought to play
a very important role in its early failure. Because the hardness of the wire braid and especially the foil are
low, the yield strength of these elements is also low (Reference 2). Perhaps the low yield strength
increases the as-built conformity of the wire braid and the foil, thus increasing the contact area and the
chance for cold welding to develop. Once the wire braid and the foil are bonded together by cold welding,
the cracks that have initiated from the foil stress concentration region may propagate very fast, as seen in
specimen A190A. To further the understanding of these phenomena, consideration is being given to
measuring certain mechanical properties of the foil during the manufacturing process.

NASA/CP—2002-211506 57
 Conclusions

From the results of the failure investigation, the following conclusions are drawn:

• Cold welding or RF failure of the elements made of silver plated copper is likely to occur for the
current A190A coaxial cable design in a space application.
• Cold welding as seen in the foil region did not lead to major damage and RF failure of the A120
coaxial cable in the desired life span. The smaller coax cable size and increased foil hardness are
apparently important factors in reducing cable stress.
• Low coax cable foil hardness may be a contributing cause of serious cold welding leading to early RF
failure. Lot-to-lot hardness measurements are highly recommended.

3 .0
R e la t iv e B e n d in g M o

2 .5

2 .0

1 .5

1 .0
0 20 40 60 80

N u m b e r o f w ir e s p e r la y e r

Figure 8. Relative Bending Moment vs. Number of Wire Bonded per Layer

References
th
1. M. Chiu, “Coaxial Cable Failure in a Spacecraft Mechanism”, Proceedings of the 34 Aerospace
Mechanisms Symposium, Goddard Space Flight Center, May 10-12, 2000.

2. Simon, N.J., Drexier, E.S. and Reed, R.P., Properties of Copper and Copper Alloys at Cryogenic
Temperatures, NIST Monograph 177, February 1992.

3. P.C. Paris, The Fracture Mechanics Approach to Fatigue. Fracture-An Interdisciplinary Approach,
Syracuse University Press, 1964.

4. El Haddad, M.H., Smith, K.N., and Topper, T.H., “A Strain Based Intensity Factor Solution for Short
Cracks Initiating from Notches”, ASTM STP 677, C.W. Smith Ed., American Society for Testing and
Materials, 1979.

NASA/CP—2002-211506 58
 Appendix A

The moment needed to bend the cable a radius of curvature, ρ, was calculated using analytical
methodology based on classic strength of materials.

As shown in Figure A-1, the coaxial cables consist of the central conductors, dielectric material, flat
braids, round wire braids, and a jacket. There are a total of 72 round wire braids in two layers. Over
multiple cycles a progressive number of round wire braids bond with the flat braids, causing the cable to
become stiffer. Consequently, the moment needed to bend the cable a given radius, ρ, increases. The
round wire braids in contact with the flat braids is called layer 1 and the outer wires, layer 2.

To start, the moment needed to bend a cable with no bonded wires is:

n
Ei I i
M =∑ Eqn 1
i =1 ρi

Where n is the total number of components, including center conductors, dielectric, flat braids, round
wire braids, and jacket.
Ei is the Young’s Modulus of the component, i.
Ii is the area moment of inertia of the component, i.
ρi is the radius of curvature of the component, i.

However, the radius of curvature is large compared to the diameter of the wire, so

ρi ≅ ρ Eqn 2

then
n

∑E I i i Eqn 3
M = i =1
ρ
Equation 3 represents the moment needed to bend a cable with no bonded components to a radius of
curvature, ρ. To calculate the change in moment needed to bend the cable to the same radius of
curvature after cold welding, a few assumptions were made. In all cases, the flat braid was assumed to
be cold welded, therefore, modeled as a tube. Furthermore, cold welding was assumed to occur in both
layers of the round wire braids. If, for instance, five wires were cold welded to the flat braids in layer 1,
then five wires were bonded to the wire mesh in layer 2. Next, the bonded wires were assumed to be
equally spaced apart. Finally the wires were assumed to be straight.

For the case where no wires are bonded, the bending of each component (and therefore its neutral axis)
is about its own centroid. Once the wires are bonded, however, the neutral axis is no longer at the
component’s centroid. Instead, the neutral axis is about the bonded structure’s centroid (in this case, the
cable), resulting in higher stresses in the outer wires. The effect this has on the bending moment can be
calculated by the following equations:

Eqn 4
M total = M u + M b

Where Mu is the moment due to the unbonded wires and other components and Mb is the moment due to
the flat braids and bonded wires, which can be expressed by the following equations.

NASA/CP—2002-211506 59
 m n

∑ Ei I i + ∑ E j I j
i =1 j =1
Eqn 5
Mu =
ρ

where ρ is the radius of curvature to the center of the cable.

m is the number of unbonded wires.
n is the total number of other unbonded components including the central conductors, dielectric,
and jacket.
Ei is the Young’s Modulus of the unbonded wires.
Ii is the area moment of inertia of the unbonded wires.
Ej is the Young’s Modulus of the other unbonded components including the central conductors,
dielectric, and jacket.
Ij is the area moment of inertia of the other unbonded components including the central
conductors, dielectric, and jacket.

The additional moment needed to bend the cable due to the bonded wires is:
q

∑E I k k
Eqn 6
Mb = k =1
ρ

where ρ is the radius of curvature to the center of the cable.

q is the number of bonded wires.
Ek is the Young’s Modulus of the bonded wires.
Ik is the area moment of inertia of the bonded wires.

The key difference between the bonded and unbonded wires is the calculation of the moment of inertia.
While the area moment of inertia in the unbonded wire is calculated about its centroid, the Ik of each
bonded wire is calculated using the parallel axis theorem. For layer 1 and layer 2, the area moment of
inertia is

I k 1 = I x + A ⋅ r1 Eqn 7

I k 2 = I x + A ⋅ r2 Eqn 8

where Ix is the area moment of inertia of the wire about it’s centroid.
A is the area of the wire.
r1 and r2 is the distance from the center of the cable to the centroid of the round wire braid.

The distance r1 and r2 can be calculated by the following equations:

Df D π π Eqn 9
r1 = ( + ) ⋅ sin( + 2 ⋅ (i − 1) ⋅ )
2 2 m m

Df D π π Eqn 10
r2 = ( +D+ ) ⋅ sin( + 2 ⋅ (i − 1) ⋅ )
2 2 m m

where Df is the outer diameter of the flat braids.

D is the diameter of the wires.

NASA/CP—2002-211506 60
 r1 is the distance from the center of the cable to the center of the wire in layer 1.
r2 is the distance from the center of the cable to the center of the wire in layer 2.

Using the above equations, one can calculate Mtotal as a function of the number of bonded wires.

Jacket
Flat Braid

Central conductors

Dielectric

Round Braid Wires in Layer 1

Round Braid Wires in Layer 2

Df/2
r1
r2

Figure A-1. Components in the A190A Coaxial Cable

NASA/CP—2002-211506 61
 Design and Testing of the CRISP Tracking Mirror Cover and Release Mechanism
* *
Jeffrey Lees and Ed Schaefer

Abstract

This paper will describe the design and qualification testing of a deployable cover and release mechanism
for the Combined Remote Imaging Spectrometer (CRISP) instrument to be flown on NASA’s COmet
Nucleus TOUR (CONTOUR) mission scheduled to be launched in July 2002. The release mechanism
utilizes redundant HOP (High Output Paraffin) actuators to rotate a cam that allows 2 latches to
simultaneously release a spring-loaded cover. The cover is a thin shell constructed of a composite fabric
with a bonded titanium hinge that is used to cover the tracking mirror during integration/test and launch.
The hinge is designed to allow the cover to rotate trough a limited angle and fly away from the spacecraft
along a linear trajectory. The extensive test program will also be described including high-speed video,
1000 Hz, and a test flight on NASA’s KC-135 Reduced Gravity Experiment plane. Lessons learned will
include an explanation of the actual angle of deployment and why it varied from the “as-designed” angle.

Figure 1. CRISP Instrument

*
Johns Hopkins University Applied Physics Laboratory, Laurel, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 63
 CRISP and the CONTOUR Mission

The CONTOUR spacecraft will fly within 100 Km of three comet nuclei: Encke in 2003, Schwassmann-
Wachmann-3 in 2006, and d’Arrest in 2008. The primary science goals of CONTOUR are:
1. To asses the diversity of comets
2. To study the process by which comet nuclei work in unprecedented detail
3. To asses for the first time the differences between Kuiper Belt and Oort Cloud comets
Short period comets are remnants of solar system formation that originated in the Kuiper Belt of icy
asteroids beyond the orbit of Neptune. Long period comets come from the Oort Cloud that extends up to
one light year away and may be less primitive. CONTOUR’s science payload will consist of four
instruments performing the following functions: High resolution imaging, spectral mapping, impact dust
analysis, neutral gas and ion spectrometry. CONTOUR expects to produce the best-ever images of comet
nuclei with a minimum resolution of 4m, 25× better than Giottos famous view of Comet Halley’s nucleus,
and the best-ever compositional analysis of comet dust and gas in the vicinity of the nucleus. CRISP,
Figure 1, is intended to track the nucleus of the comet as CONTOUR flies by producing high resolution
color images and spectroscopy to determine composition of the nucleus and processes that shape the
surface.

Deployment Environment

The cover is expected to be deployed approximately one to two months after launch. The spacecraft will
reduce its spin-rate for cover deployment and instrument functional check. The nominal operating
temperature of the tracking mirror mechanism, cover and release mechanism is expected to be between
-40° and -60°C.

Lubrication

A dry film lubricant was desired for all moving surfaces because of the low temperatures. A survey of the
literature lead us to believe that sputtered molybdenum disulfide (MoS2) was the ideal choice. However,
we also wanted to compare tungsten disulfide (WS2), as sold by Dicronite and ion plated lead. Samples
were acquired on several test specimens that were similar to our mechanism components. It became
immediately obvious that the Dicronite was vastly superior for our release mechanism application.
Sputtered MoS2 exhibited significantly more stiction and higher coefficient of friction at ambient conditions
where we expected to do the vast majority of our testing. The only deployment tests that were actually
conducted in vacuum were to simulate deployment at -60°C. We also chose the ion plated lead over
sputtered MoS2 for the tracking mirror bearings. This was mainly due to a lack of information available in
the literature for WS2 or Dicronite in this application and that ion plated lead had previously been used
with good results at JHU/APL in an identical motor/encoder.

Material Selection

The CRISP instrument is fairly large, 683H × 1267W × 502D mm and 22.6 kg, and was very mass
constrained. The upper and lower housings forming the primary structure of the instrument, the tracking
mirror assembly base, and the star camera brackets, Figure 1, were all fabricated from solid billets of
magnesium ZK60A-T5. The two-sided tracking mirror, Figure 14, itself was fabricated from aluminum
6061-T6. Titanium 6Al4V was used for the motor/encoder housing. Analysis indicated that all of the
tracking mirror assembly brackets, bearing housings and diaphragm were required to be fabricated from
Ti6Al4V in order to match the motor/encoder housing. Interfacing materials with drastically different
coefficients of thermal expansion (CTEs) required significant amounts of finite element analysis and
attention to detail. The -60°C environment dictated the use of titanium fasteners to mate to titanium parts
so that fastener preload would not be lost. Dissimilar materials were used to advantage in the release
mechanism, where the larger CTE material, stainless, running inside the lower CTE material, titanium,
such that clearance tolerances would never be lost. Ambient temperatures, where most testing occurred,
was our worst hot case.

NASA/CP—2002-211506 64
 Release Mechanism

The most convenient location to place the release mechanism was in a pocket, directly under the motor.
The size limitations of the pocket imposed a challenge to package a release mechanism with redundant
actuators in a 120(l) × 48(h) × 38(d) mm volume. In addition, it was determined that a single point of
release would not be practical. A philosophy was adopted that both latches should be released by a
common component directly coupled to the actuators. With the requirements defined, a release
mechanism was developed that utilized identical redundant actuators that could drive a single cam
releasing twin latches simultaneously and fit within the pocket.

2× LATCH

2× HOP EXTENSION
SENSOR

2× HOP

LATCH
SPRING
PISTON

ROLLER

TANGENT BAND
V
CAM
Figure 2. Release Mechanism Cut-Away

The heart of the release mechanism, shown in Figure 2 and Figure 13, is a cylindrical, Ø25.4mm, cam
with two 180º opposed V grooves. The cam is driven by the actuator’s tangent band, simultaneously or
independently, rotating it against the torsion spring. Once the cam rotates 15º, the cam followers are
driven into the V grooves, providing a quick release. The tension springs inside the cam followers provide
a constant force against the cam. Therefore, all the forces acting on the release mechanism, the single
torsion spring and two tension springs, are all internal to the mechanism, and the mechanism itself has no
external forces acting on it.

Release Mechanism Actuators

Two EH-3525 high output paraffin (HOP) actuators configured with an elgiloy tangent band developed by
Starsys Research Inc. drive the mechanism, as shown in Figure 3. The tangent band was employed for
two reasons:
1. To provide a purely linear force to the actuator
2. To allow the actuators to operate independently
The tangent band converts uni-axial HOP motion to rotary cam motion without exerting transverse loads
on the HOP. The HOP tangent bands are also flexible enough to allow the HOPs to be actuated
independently as well as accommodate differential extension of the HOPs when driven simultaneously.
The EH-3525 HOP actuator was specifically designed with stringent contamination control requirements.

NASA/CP—2002-211506 65
The welded steel bellows provides a hermetically sealed wax chamber and return force as well as
incorporating an internal hard stop. Advantages of this HOP actuator are:

• Shock free actuation • Low power 5W @ 28V

• Self resetting • >500 cycle capacity
• Hermetically sealed • High output force 155.7N
• Low mass (35g) • Integral internal hard stop

Figure 3. HOP Actuators and Tangent Bands

Cover Design

The cover, Figure 4, is constrained to rotate about a fixed point, which is not at its center of mass. The
thin shell cover was designed using two layers of M55J fabric and YLA’s RS-12 resin system. Two
titanium arches were bonded to the composite cover, one at each end using Hysol 9394. The arches
provide the primary interface between the scan mirror housing and the cover. These arches also serve to
stiffen the thin shell structure. The arch at one end of the cover includes part of the hinge assembly, while
the arch at the opposite end provides a rigid seat for the kick-off springs.

Hinge Design

The hinge design is simple, highly reliable, and it employs all passive components. The hinge consists of
two parts, one located on the scan mirror housing and the other on the cover. The fixed part of the hinge
consists of a bracket mounted on the scan mirror housing and contains the hinge pin and cam follower.
The cam follower is fabricated using a titanium sleeve, which rotates on a stainless steel shaft. The
stationary hinge pin is also mounted on this bracket. All of the hinge parts were coated with Dicronite,
(tungsten disulfide, WS2) to minimize friction and wear. This dry film lubricant has a coefficient of friction
of 0.065 in vacuum. The rotating bracket, which is an integral part of the cover, contains the grooved cam
that has been machined into the bracket. The hinge parts are illustrated in Figure 5.

NASA/CP—2002-211506 66
 Figure 4. Cover

Hinge Hinge Track

Pin Roller

Dicronite (WS2)
Dry Lubrication

Cover Rotation
Stop Hinge Rotation
Rotation Stop
Hinge

Stationary Hinge Half Fly-Away Hinge Half

Figure 5. Hinge

Spring Plunger Design

As shown in Figure 6, four spring-loaded plungers, two per side, were used to deploy the cover. Each set
of two spring-loaded plungers was designed to have a high margin of safety on the spring energy. The
four spring-loaded plungers mounted in parallel exert a maximum total force of 103.6 N when
compressed 15.75 mm. The plungers were fabricated from titanium and were also vented to reduce air
dampening while testing at ambient conditions. The plungers ran in a stainless steel sleeve with generous
clearances. The spring was full captured in the plunger when compressed to prevent it from buckling. The
stainless steel sleeves were used to prevent the titanium plungers from scratching or galling the softer
magnesium housing, allowing for hundreds of deployments. Additionally, all surfaces were coated with
Dicronite dry lubricant. The springs and plungers were also designed to have zero preload, which allowed
hundreds of deployments to be made without having to worry about the fasteners pounding out or
loosening in the magnesium housing. CRES or A286 fasteners and locking phosphor bronze Helicals
were generally used throughout, except for some titanium fasteners.

NASA/CP—2002-211506 67
 DICRONITE COATED
TITANIUM PLUNGER

Mg HOUSING

DICRONITE COATED
STAINLESS SLEEVE

Figure 6. Spring-Loaded Plungers

Intended Deployment

Initially, we believed the cover would rotate 30° about the fixed hinge pin. The cover stop would then
come into contact with the inclined rotation stop built into the stationary hinge. This would convert some of
the covers rotational energy to translational energy. The center of mass of the cover would also move
along an arc centered at the fixed hinge pin 30° and then translate along a tangent line as the cover
transitioned into centroidal rotation about it’s center of mass. However, during low gravity testing, this
clearly was not what happened.

Low Gravity Testing

One requirement was that the cover would deploy without hitting hitting any of the surrounding solar cells
on the spacecraft. The best way to verify the design was to test it aboard NASA’s KC-135 Reduced
1
Gravity Experiment aircraft . Low gravity deployment tests were conducted over two days aboard the KC-
135. The HOP actuators were removed and a hand operated lever actuator was installed. There were no
significant issues regarding HOPs actuating the mechanism in zero gravity. However, we wanted to be
able to deploy the cover the instant the crew indicated minimal gravity conditions were present. The
typical time for the HOP actuators to deploy the cover was greater than the duration of the low gravity
conditions. Therefore, to eliminate guess work and erroneous deployments, the hand-actuated lever was
utilized. The only issues with the hand actuated lever were the physical condition of the operators. As a
result, we were unable to achieve a deployment on every parabola the first day as two of the
experimentalists succumbed to motion sickness.

In all, 57 deployments, all successful, were completed during a total of 80 parabolas. Included in the 57
deployments, were several deployments with two spring-loaded plungers de-activated. The deployment
trajectories were identical except the velocities, rotational and translational, were slower.

Figure 7 shows Jim Bell, from Cornell, deploying the cover and the location of the video camera used.

NASA/CP—2002-211506 68
 VIDEO CAMERA

Figure 7. Cover Deployment Aboard the KC-135

Equations of Motion

1 1
T= mVg2 + I g ω o2 (1)
2 2

Vg = rg ω o ( 2)

I o = I g + mrg2 ( parallel axis theorem) (3)

a n = rω 2 ( 4)

a t = rα (5)

1 1
T = ω o2 ( mrg2 + I g ) = ω o2 I o (6)
2 2

NASA/CP—2002-211506 69
 Non Centroidal Rotation

The cover is a rigid body constrained to rotate about a fixed hinge point (O´) that is not at the cover’s
center of mass (g), Figure 8. The cover’s center of mass, g, will move along an arc of radius (r) centered
at the fixed hinge, O´, until it flies away. At that instant, the cover will translate along a straight line
tangent to the arc and simultaneously rotate about its center of mass, g2.

As the latches open, the cover begins to rotate about its fixed hinge accelerating until the spring energy is
released and the plungers reach the end of their travel. The cover continues to rotate at a constant
angular velocity until it separates and flies away. Thus, terms containing α, Equation 5, reduce to zero.

CENTER OF MASS
TRAJECTORY

58.5°

g (CENTER OF MASS)

rg

24.8°

O´ (CENTER OF ROTATION)
Figure 8. Cover Geometry

man

Rx

Ry

Figure 9. Hinge Free Body Diagram

NASA/CP—2002-211506 70
 Point of Separation

Once the cover rotates 30°, it transitions from rotating about a fixed hinge pin to a cylinder rotating on an
inclined plane, point O´. The cover continues rotating about this point until the reaction force parallel to
the base reduces to zero, Figure 9. The centrifugal force holds the cover on the point of rotation until its
center of mass makes an angle of 58.5° with respect to the tracking mirror base. Separation will only
occur when the cover’s center of mass and the point of rotation form a line normal to the base. That is to
say that as long as the cover’s center of mass is not directly over the point of rotation, the centrifugal
force is the only force holding the cover on. Once the cover’s center of mass passes the point of rotation,
the hinge no longer supports a reaction force, and the cover releases along a linear trajectory. In
hindsight, it is probably fortunate that the plane was inclined. Otherwise, as the parallel, or friction,
component of the reaction forces reduced, the cover may have slid towards the hinge pin and adversely
affected the fly-away separation.

Analysis Approach

The analysis of the cover’s release was done assuming rigid body dynamics. The inertial coordinate
system was selected, which uses the normal and tangential directions. The equations of motion
pertaining to the rotation of a body about a fixed point were developed. These equations were developed
using the dynamic characteristics of the spring plungers and the cover’s geometry, as follows: the applied
torque at any point in time is equal to the spring force times the distance between the hinge and plungers.
The torsional spring constant is equal to the applied torque divided by the angle of rotation at any point in
time. Using the principal of conservation of energy and equating the initial potential and final kinetic
energy the maximum angular velocity was obtained. The maximum acceleration is obtained by taking the
maximum torque divided by the cover’s inertia.

Assuming the acceleration is linear and decreasing, the area under the acceleration curve is the velocity
from which the release time is estimated. Using the basic relationships the angular velocity and
displacement of the cover were calculated by integrating the area under the appropriate curves. The
centrifugal and tangential forces were also calculated at the hinge point used to develop the loads and
friction in the hinge. The following two cases were analyzed:

Case 1. zero g, environment: no air resistance, or bearing friction.

Case 2. one g, environment: air resistance including both friction and

form drag, and bearing torque friction.

Analysis and Test Results

The cover release analysis and test results are illustrated in Figure 10. In the 0g case, the analysis
estimated a cover release time of 80 ms with an average angular velocity of 14.4 radians per second. The
test results shows that the cover release took 100 ms with an average angular velocity of 10.47 radians
per second. The analytical prediction agreed to within approximately 20% and 27% respectively.

In the 1g case, the analysis estimated a cover release time of 104 ms with an average velocity of 8.2
radians per second. The test results show that the cover release took 120 ms with an average velocity of
7.52 radians per second. The analytical prediction agreed to within approximately 15% and 9%
respectively. These errors are attributed to friction in both the plunger and hinge assemblies, and basic
modeling assumptions. This modeling was done assuming plane motion, i.e., a two-dimensional
representation verses a three-dimensional problem.

NASA/CP—2002-211506 71
 CRISP Cover Release

120 0g Analysis
0g Test
Angular Displacement (Degrees)

100
1g Analysis
1g Test

80

60

40

20

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

Time (Seconds)

Figure 10. Cover Angular Velocity Analysis

Measurements made from the CAD model, flight hardware, and videos were used in equations 1-6 as
follows:

m = .325kg measured
rg = .136m CAD model
k = .0998m CAD model
ω o = 10.472 rad High speed video
s
ω g = 10.76 rad s KC − 135 video

v g = 1.44 m KC − 135 video

s
I g = .003239kgm 2 Eq. (3)
I o = .009250kgm 2 Eq. (3)
2
To = .5072 kgm Eq. (6)
s2
2
Tf = .5226 kgm Eq. (1)
s2

To − T f
× 100% = −3.0%
To

NASA/CP—2002-211506 72
A composite picture of three frames of video was created to show the cover in a similar orientation at
three different times, Figure 11. The second composite picture, Figure 12, show a similar result, but with
two of the four spring plungers de-activated. The angle of deployment remains unchanged, verifying, that
it is independent of spring energy. Linear and angular velocities and deployment angles were measured
from the videos taken on board the KC-135. The angle of deployment was measured at 60° vs. 58.5°
calculated. Conservation of energy was also verified to within 3% from the measurements and equations
of motion. The error is believed to be attributable to inaccuracies in measurements made from the video
frames.

CUT LINES

Figure 11. Cover Deployment Composite

NASA/CP—2002-211506 73
 CUT LINE

Figure 12. Cover Deployment 2 of 4 Spring Plungers De-Activated Composite

Conclusions and Lessons Learned

Low gravity testing proved to be beneficial in that it not only demonstrated that the cover is unlikely to
damage any surrounding solar cells upon deployment, but it also allowed us to learn how the cover
actually deployed in a manner somewhat simpler than what we had originally conceived.

We also would be interested to see more detailed studies of the performance of WS2 dry film lubricant as
compared to MoS2 for bearings and other mechanisms. Although it may not have the lowest possible
coefficient of friction under ideal circumstances in vacuum, our samples showed it to be vastly superior to
MoS2 during ambient testing conditions there doesn’t appear to be very much WS2 information available
in the literature that we were able to find.

NASA/CP—2002-211506 74
 Figure 13. Flight Release Mechanism

Figure 14. CRISP Flight Tracking Mirror Mechanism

NASA/CP—2002-211506 75
1
Operations User’s Guide JSC Reduced Gravity Program User’s Guide, Rev D, October 2000, National
Aeronautics and Space Administration.
2
Beer, Ferdinand P., Johnston, Russell Jr. Vector Mechanics for Engineers, McGraw-Hill Book Company,
©1984, 734-735, 755-761.

NASA/CP—2002-211506 76
 Zero Deadband, Multiple Strut Synchronized Hinge for Deployable Structures
* **
Matthew Botke*, David Murphy*, Thomas Murphey , Peter Sorensen

Abstract

Many configurations exist in the literature regarding the synchronization of hinged elements in deployable
structures. This paper details the design and manufacturing aspects of a tape drive mechanism that uses
a central pulley to drive one or more hinged elements or struts. The mechanism functions as a
synchronizer and can be powered or unpowered. An advantage to this synchronization approach over
others is its implementation of simple mechanical items in a statically determinant arrangement that
facilitates a straightforward analysis approach. The design also lends itself to manufacture, assembly, test
as well as low cost. This design approach possesses wide applicability to a broad range of deployable
structures.

Introduction

The need for a new approach to multiple strut synchronization arose in the development of a novel solar
array system (patent pending). The array is designed to deploy a large area of thin film or crystalline
2
photovoltaic devices (upwards of 150 m or more per wing). The deployment kinematics require hinge
elements in the array to synchronize each connected member. The array power to mass efficiency is
designed to be in the range of 200 to 250 W/kg. In attempting to achieve this jump in mass efficiency from
the state of practice, the mass of the array structure must be highly optimized. The design challenge is to
develop a structure that is robust, has both high stowed and deployed packing factors, is extremely mass
efficient and provides the functionality required to reliably deploy the array. The subject hinge would be a
repeating element in the array designed to reduce unique mechanism count. The design process for this
hinge, lessons learned so far, as well as the current state of the design are presented here in order to
illustrate a possible solution to a broad range of deployable structures requiring zero-deadband
synchronized and efficient motion of two or more struts.

Background

The deploy sequence of the modular array is depicted in Figure 1. Refer to Figure 2 for structure
nomenclature of a bay. Deployment is initiated by releasing the tiedown mechanism (not depicted). The
array structure and yoke are deployed simultaneously and latched. Each corner hinge element or node
synchronizes the deployment angle of each connected strut. The yoke is synchronized similarly. By
synchronizing each strut at the nodes, each bay and hence the entire array deployment kinematics are
defined. Deploy motive energy is input via motors at selected nodes. Stepper motors are used to control
rate and preclude possible binding modes that could be caused by unbalanced torques. Un-motorized
nodes incorporate helper springs in order to just overcome local sources of parasitic torque. The
deployed structure forms a number of tiled picture frame elements or ‘bays’. Considering the kinematics
of a single bay, the single strut sides of the bay remain parallel during deployment. Lanyards are paid out
during the structure deployment between these sides and terminate at the stowed photovoltaic blanket
packaged against the inboard single strut side. These blankets remain stowed until the structure is fully
deployed and latched upon which time the lanyards are reeled back in to deploy the blanket.

*
AEC-Able Engineering, Goleta, CA
**
Kollabra, Palo Alto, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 77
 Array shown stowed
against large bus.

First, the array

structure deploys
synchronously under
motor control.

The structure is driven

to the fully deployed
and latched state.

1 bay constitutes a repeating

element that can be tiled in various
configurations to construct a wing.

Lastly, photovoltaic blankets

are pulled across bays using
same motors that deployed
the structure.

2
Figure 1. Deployment sequence of a 300 m array

NASA/CP—2002-211506 78
In this embodiment, the node must accomplish many functions in a limited volume. It must synchronize
from two to four struts through 90° of rotation, be very mass efficient, and house blanket management
mechanisms. Since a majority of the nodes do not incorporate a motor and are rather distant from motive
input, the node must not introduce significant retarding forces during deployment. Parasitic moments
would lead to loss of kinematic timing given strut flexibility and could lead to binding. Also, in order to
maintain some stiffness during deployment, the node should have minimal deadband in the synchronizing
mechanism.

node (shown here with

motor and gearhead)

double strut side

single strut side

mid-strut hinge

Figure 2. Bay nomenclature

Requirements

As the array concept matured, the node requirements continued to be refined. Several designs were
competed based on these requirements. Among these requirements, several emerged as most important
aside from low mass, reliability, tolerance to thermal extremes and manufacturability.

1. Function as a powered or unpowered synchronizer.

The ability to function as a synchronizer is important for two reasons. In order to keep mass, cost and
complexity to a minimum, the ability of the node design to synchronize struts in an unpowered mode
affords the ability to reduce motor count. Also, in the event of a motor loss, the affected node must not
inhibit the deployment of the array. This is realized by linking the motor in a ‘pusher’ configuration that
does not retard structure motion.

NASA/CP—2002-211506 79
2. Maintain non-zero stiffness during deployment (no deadband)

A desirable feature of the node was the ability to maintain some reasonable level of stiffness in order to
reduce the likelihood of bay distortion and thus the potential for binding during deployment. Also, this is
typically a desirable characteristic in any deployable structure to facilitate attitude control schemes and
latchout.

3. Possess a high degree of applicability to peripheral nodes.

The node design must accommodate from two to four struts in order to be applicable to the peripheral
nodes as well as interior nodes. In the array configuration depicted in Figure 1, 16 of the 21 nodes in the
wing are located on the periphery. Only five nodes are required to synchronize the full complement of
struts.

4. Possess the ability to package additional mechanism for blanket management.

Given the desire to use the structure deployment motors in order to deploy the blankets, additional
mechanism would have to be packaged at the powered nodes. This dual use approach allows the motor
count to be minimized.

The ability to meet all of these requirements has a direct advantageous effect on the reliability, mass and
complexity of the array system by reducing the number of unique mechanisms. The goal in developing an
array of this nature is to maintain the ability to tile bays in various configurations to suit particular mission
power, stowed and deployed envelope, and deployed frequency requirements without significant
mechanism redesign.

Design Concept

A tape drive mechanism was the chosen concept based on its ability to meet all of the above
requirements in a very simple manner. Central to the node is a cylindrical drum with an axis of rotation
that is rotated 90° from the axes of rotation of each strut end. A given strut clevis is fitted with two arc
sectors each laterally displaced to the tangent of the drum as shown in Figure 3. The arc sectors act as
portions of a pulley and are sized according to the required rotation of the strut. In this case, 90° of strut
rotation corresponds to full deployment. A pair of tapes required for each strut to be synchronized is
mutually preloaded to eliminate deadband. As the drum rotates, one tape is wound around the drum and
is unwound from one pulley sector of the strut clevis. The opposite tape is unwound from the drum and
wound on the pulley sector on the opposite side of the strut clevis.

This arrangement can be used to drive the deployment of one or more struts with a single motor that fits
well in the volume between the stowed struts. Synchronizing additional struts is accomplished in the
same manner. Figure 4 illustrates the tape arrangement for 4 struts. In this application, the pulley
diameter ratio is ~1.1 and so 82.5° of drum rotation causes the required 90° of strut rotation. Tapes are
generally thin compared to the pulley radii and can thus overlap with little error in the angular relationship
between the drum and strut clevis.

NASA/CP—2002-211506 80
 Pulley sectors grounded
to strut (both sectors are
Drum pulley grounded on the same
clevis)

Tapes (2x) are grounded

at each pulley
Input motor torque

Figure 3. Synchronization arrangement for one strut

Synchronizer
drum with 4 sets
of tapes Typical strut clevis
with integral pulley
sectors

Figure 4. Tape arrangement for 4 struts (one strut clevis shown)

Tape preload results in a zero deadband mechanism with a stiffness largely driven the tapes. This
preload also drives hinge efficiency due to frictional losses at each pin axis. A node arrangement with two
orthogonal struts results in the worst-case bearing load (and therefore lowest efficiency) on the drum to
node structure bore. A balanced node (two opposite struts or four struts in this case) with equal tape
preload results in a zero bearing on the drum.

In the arrangements shown in Figures 3 and 4, both rotating elements (the drum and the clevis) rotate on
floating pins (for redundancy) in the node structure. This arrangement can be analyzed for stiffness using
methods illustrated in the following section.

Synchronization Stiffness Analysis

The node structure, shown in Figure 5 illustrates the scale of the node. Node plates locate the drum.
Node lugs locate each strut clevis. The node housing is optimized to provide stiffness between latched
struts.

NASA/CP—2002-211506 81
 Node plate

Node lug

Figure 5. Node housing

The synchronization stiffness of the mechanism can be described by the sensitivity in relative rotational
position of each clevis and the drum with respect to a statically balanced set of moment loads. This
parameter is applicable to the analysis of a larger structure employing the mechanism for investigating
global structure shape and loads as well as elastic stability (buckling) and elastic binding modes. It is also
useful in dynamic analysis for calculation of structure vibration frequencies and deployment dynamics.

If the mechanism drum, housing and clevis are assumed to be rigid, mechanism stiffness is largely due to
tape stiffness. This assumption is generally valid since the tapes are relatively thin and narrow due to the
pulley radii and the 90° twist respectively. A further assumption is that the tapes are preloaded beyond
the stiffness knee due to the twist. Figure 6 shows a simplified stiffness model of the mechanism where
the eight tapes are the only elastic elements. Each clevis has been rotated 90° to allow a two-dimensional
visualization. Each tape is represented by a linear spring of stiffness k. Static analysis of the mechanism
yields five equilibrium equations,
M 1 = 2rF1 , M 2 = 2rF2 , M 3 = 2rF3 , M 4 = 2rF4 ,
r . (1)
M1 + M 2 + M 3 + M 4 + Md = 0
rd
The spring forces are assumed positive in tension and yield four spring constitutive equations,
( ) ( ) ( )
F1 = k ∆1 − ∆ d , F2 = k ∆ 2 − ∆ d , F3 = k ∆ 3 − ∆ d , F4 = k ∆ 4 − ∆ d . ( ) (2)
Five kinematic equations relate element rotations to spring stretch (small rotations are assumed),
∆1 = rθ1 , ∆ 2 = rθ 2 , ∆ 3 = rθ 3 , ∆ 4 = rθ 4 , ∆ d = rdθ d . (3)
Relations between the applied moments and the element angles are derived by substituting the kinematic
and constitutive equations into the equilibrium equations,
M1 M2 M3 M4 Md
2kr 2 = = = = = 2
. (4)
rd rd rd rd rd rd
θ1 − θ d θ 2 − θ d θ 3 − θ d θ 4 − θ d 4 2 θ d − (θ1 + θ 2 + θ 3 + θ 4 )
r r r r r r
Equation (4) reveals the rotational stiffness ( M θ ) of any clevis relative the drum ( θ d = 0 ) is 2kr .
2

Similarly, the stiffness of the drum relative to the four limbs ( θ 1 = θ 2 = θ 3 = θ 4 = 0 ) is 8krd2 . The relations
are also applicable to general cases of arbitrarily imposed loads or angles. However, the mechanism has
one rigid body degree of freedom; it is under-constrained. Specifying one of the element angles
sufficiently constrains the problem to allow solution of the equations (4). For example, with θ 1 = 0 , the
remaining element angles as a function of the applied loads are,

NASA/CP—2002-211506 82
 M 2 − M1 M −M M −M − M1
θ2 = 2
, θ3 = 3 2 1 , θ4 = 4 2 1 , θd = , (6)
2r k 2r k 2r k 2rrd k
where,
r
M1 = − M d − (M 2 + M 3 + M 4 ) . (7)
rd

A stiffness reduction factor due to the twist in each tape is sometimes needed in evaluating the above
equations. The theoretical stiffness of a straight tape (no twist) is given by,
AE
k= , (8)
l
where A is the cross sectional area, E is Young’s modulus and l is the active tape length. Introduction of a
twist modifies the tape load-path so that it becomes curved toward the edges of the tape. With certain
geometries (generally, wider and thinner tapes) the twist can also cause tape buckling patterns. A finite
element analysis (nonlinear transient dynamic) was used to investigate tape stiffness and buckling issues.
An illustrative buckling pattern is shown in Figure 7. Effective extensional stiffness analysis results for a
relatively thin (0.025-mm thick) tape are shown in Figure 8. At low tensions, a tape showing buckling
patterns exhibits a drastically reduced effective extensional stiffness, but at higher preloads, the buckling
patterns can be removed and nearly full tape stiffness is achieved. Even after buckling patterns are
removed, however, only 88% of the full stiffness is achieved. The physical models built for this study
employed tape dimensions (0.132 x 2.54 mm x 135.6 mm long) such that buckling never occurred and
the effective extensional stiffness was not significantly reduced due to the twist (less than 2% reduction).

For these dimensions, the effective spring stiffness of an individual tape is 1900 N/mm, the rotational
stiffness of any clevis (r = 16.5 mm) relative the drum is 1034 N-m/rad and the rotational stiffness of the
drum (rd = 19.0 mm) relative to the four limbs is 5487 N-m/rad.

θ2, M2

∆2

F2 F2
∆1 F1 ∆d F3

rd r
θ1, θd, θ3,
M1 Md M3

F1 F3 ∆3 l
F4 θ4, M4 F4

∆4

Figure 6. Mechanism stiffness model

NASA/CP—2002-211506 83
 Figure 7. Finite element model of tape buckling pattern

0.9
Effective Modulus / Bulk Material Modulus

2.54 x 0.0254 mm tape

0.8 E = 206.8 GPa, Nu = 0.3

0.7
88.6% of bulk material
0.6 stiffness achieved with
removal of all buckling
0.5

0.4
Irregularities due to
removal of buckling
0.3 patterns

0.2

0.1

0
0 10 20 30 40 50
Stress (MPa)

Figure 8. Effective modulus of thin tape with 90° twist and buckling at low stress levels

NASA/CP—2002-211506 84
 Design and Manufacturing Lessons Learned

In the effort to configure the tape drive in a cost effective and reliable embodiment, several design and
manufacturing lessons were learned. Most of these lessons involved the methods of manufacturing a
tape assembly that would possess the strength required as well as terminate easily on each pulley with a
simple method of length and preload adjustment.

Wire rope cables were originally baselined due to the straightforward termination methods, inherent
redundancy of multiple strands as well as the need to wrap on two orthogonal pulley surfaces. As cables
are wrapped around the radius of a pulley, the wire strands tend to broom and the cable assumes an
elliptical shape. This action of repeatedly re-nesting the strands that make up the cable would result in
cable length variations that are not predictable. The error increases as the ratio of the diameter of the
cable to pulley radius increases. Slight errors in synchronization as a result of slight length changes would
introduce end of deployment errors that must be accommodated for in the latching window. Use of
latching window budget is unfavorable as a design baseline and if possible should be saved for
unanticipated sources of error. Also, cables have a low stiffness at low preload. In order to operate in the
higher stiffness region of the cables, larger shear loads on the drum and strut clevis would be required
resulting in higher frictional losses and lower overall efficiency of the node. Given these reasons, cables
were replaced with tapes.

Tape

Drum tape anchor

Clevis tape anchor
showing adjustment
screw location

Figure 9. Tape assembly

Tapes need to twist 90° in order to wrap on each pulley surface. The width of the tape is on the order of
10 times the free length. A buckling pattern in the center of a thin tape can be observed when twisted.
This introduces a low stiffness regime due to the length difference between the outside and center of the
tape. Such tapes must be preloaded sufficiently to strain the center of the tape to eliminate this buckling.
Finite element modeling shows that stiffness reductions can be as high as 15% when operating in the
buckled regime. Cables typically require preloading in the region of 50 to 60% of breaking strength in
order to operate in a linear stiffness regime. The final tape geometry avoided this buckling mode by
optimization of the width and thickness. Stiffness reduction due to twist of the final tape geometry was on
the order of 2%.

Due to certain loading conditions during the deployment of the array structure, the tape design required a
material with a high ultimate tensile strength. Elgiloy was chosen based on its strength and availability in
various thicknesses. The yield strength varies with tape thickness. In the thickness selected, the yield
stress was approximately 2.0 GPa. The ratio of preload to ultimate strength for the assembly is 15% and
is set to ensure preload is maintained under worst-case deployment loads and thermal conditions.

Finding a robust method for tape termination proved to be an enabling technology for the construction of
such a small mechanism (the drum diameter is 38 mm). Tape termination using conventional pinning and
clamping required the use of fasteners that are too large to be practical or too small to be reliable. Several
different methods of tape termination were investigated including brazing and tapes with integral ears.
Traditional welding techniques were not investigated initially due to potentially large residual stress fields
in the heat affected zones and the effect on the temper of the tape material. A brazing material was found
that could be processed at temperatures below the annealing temperature of the tape and this method
became the leading candidate. Although material mismatch would not present a problem as it would with
welding, uncertainties about shear strength consistency, the ability to position the tape during the
process, and possible outgassing remained.

NASA/CP—2002-211506 85
These methods did not possess all the characteristics required for a reliable, simple, and cost-effective
solution and after further searching, laser welding proved to be the perfect solution for this application.
Unlike conventional welding, laser welding produces a very small heat affected zone, is extremely
controllable and repeatable, allows for precise fixturing, and turned out to be extremely cost effective.
Also, using this process it was possible to weld the Elgiloy tapes to almost any common stainless steel
except 303 or 316. Tape anchors were welded to each end of the tape and tested in tension. The welded
tape assembly strength proved to be very repeatable at approximately 70% of the yield strength of the
tape. Figure 10 shows a typical laser welded tape anchor.

Tape preload and length adjustment was accomplished via an adjustment screw located on the strut
clevis. Access and tune-ability proved to be very easy. Accurately measuring the preload in the tapes
proved to be difficult and for expediency was done by feel. During the first iteration, tape preload and strut
angular positioning was accomplished by locking the drum to the node housing using a pin while keeping
the struts in a latched and therefore known position. The pin was able to spin freely in the drum and
housing. Tapes were then preloaded one clevis at a time. By preloading opposing tapes a pair at a time,
equal preload could be verified by the locking pin’s ability to spin. One tape would be preloaded until the
desired value was reached, then the opposing tape would be loaded until the pin was free to spin again.
This procedure was then repeated for the adjacent strut in the node. This method produces a
progressively higher preload in each subsequent strut due to the added stiffness of the preloaded tape
sets of prior struts. An alternative to this approach might be to preload all strut tapes of the same function
(for example those that are used to move the struts in the same direction) at once against the locking pin
then the remaining tapes in the opposing direction would be preloaded.

Second generation hardware may involve bonding a strain gauge on each tape or possibly a removable
extensometer and preloading the entire node on the bench prior to integration into the structure. Each
tape and strain gauge assembly would be calibrated prior to installation. Strain gauge placement (if used)
would likely be on a portion of the tape running over the drum or clevis segment in order to eliminate the
torsional strains due to tape twist. Strain due to bending over the pulley would have to be considered. All
tapes would be taken from zero-load to the required preload at that time. The node would then be cycled
in order to exercise the tapes over the entire range of motion and then checked again. The ability to
exercise a node is important to ensure uniform strain along the entire tape. Exercising a node while
integrated within a potentially very large array would require cycling the entire array and therefore be time
consuming, risky and could potentially degrade system reliability.

Figure 10. Laser-welded tape anchor

NASA/CP—2002-211506 86
 Figure 11. Two-strut node with tapes integrated

NASA/CP—2002-211506 87
 Design Progress

Hardware has been manufactured and integrated into a demonstration single bay of the array structure
measuring approximately 2.5 x 5 m (both sides of one node is shown in Figure 11). The bay has been
successfully cycled upwards of 50 times with very promising results. Second generation node hardware
design is underway and progress has been made incorporating additional functionality. The tape drive
portion of the mechanism has remained essentially unchanged from the first generation hardware. The
ability to package additional mechanism can be a significant driver in competing designs that must be
implemented in limited volume. The top and bottom node plates as well as the volume within the drum
were used to mount or house additional mechanism. The bottom node plate, around which the struts are
stowed, provided the interface for either a motor or spring. Latches and hardstops were mounted to the
upper node plate. All the blanket management mechanism is housed completely within a modified drum.
The same motors used to deploy the structure are used to deploy the blankets. By reversing the direction
of the motor, the motor shaft disengages the drum and engages a multi-track cable take-up reel that
deploys the blankets in two bays. Since the drum only rotates approximately 90°, windows located in the
wall of the drum provide radial access for the blanket cables. Clutch design and cable management are
beyond the scope of this paper, but Figure 12 shows the second-generation node with the additional
planned drum mechanism installed.

Figure 12. Drum with blanket mechanism and drum integrated into node

Conclusions

The tape-driven node met the design requirements and fit very well within the limited design space. The
mechanism proved to be very efficient both in terms of energy transfer and minimal mass. The use of
tapes and pulley segments allowed simple analysis methods. The hardware function is visually clear and
so facilitates inspection and testing.

Accurate determination of tape preload can be difficult although it may be accomplished using strain
gauges permanently fixed or extensometers temporarily fixed to each tape. Other methods may prove
more practical. Tape termination using laser welding proved to be very strong, reliable and economical.
Tape anchors can be routed and sized to adapt to the requirements and volume limitations of various
embodiments.

Tape drives such as this possess wide applicability due to scalability and efficiency. Roller or sleeve
bearings can be incorporated to further increase overall efficiency. Node arrangements with alternate
numbers of struts, non-orthogonal axes, and/or different rotation angles are possible. The above
discussion is meant to provide the reader with some design and manufacturing lessons learned in
implementing such a mechanism with real-world design requirements.

NASA/CP—2002-211506 88
 DC Motor Selection for Space Mechanisms
*
David B. Marks and Richard A. Fink*

Abstract

“DC Motor Selection for Space Mechanisms” presents major types of DC motors, their design differences,
and application guidelines, so that their selection and specification for use in space applications will be
enhanced. DC motors that are examined include two and three phase brushless, brushed, conventional
and hybrid steppers, and specialty brushless for low torque ripple and high speed spinning applications.
Motor design and application concepts are presented that are essential for proper motor selection to point
antennas, drive solar arrays, latch, and to perform other space mechanism functions. A tradeoff matrix
compares motor types for specific applications. Lessons learned from a variety of space flight motor
applications show selections that worked well and some that did not.

Introduction

D.C. motors play an integral and critical role in the operation of space flight hardware from the Space
Station to communication satellites. Motors provide motion drive for solar array deployment, antenna
pointing, boom deployment, cargo latching, aperture movement, inertial spinning for control moment
gyros, and many other applications. Without a proper understanding of how to select motor components
and what has been done previously, it is likely that mechanism performance anomalies will result or re-
design will be required.

Given this diversity of applications, proper selection of motor types and specification of the motor
performance parameters for space flight mechanisms is essential to the success of future space
missions. While the subject of motor design is beyond the scope of any single paper or book, the
fundamentals common to DC motors and their application can be examined. Technical essentials for
understanding and specifying DC motors, types of DC motor designs for space mechanisms, and DC
motors in space applications are presented; a selection matrix integrates key motor design and
application factors with lessons learned to guide future selection. If DC motor selection properly takes into
account motor design differences and their effects on specific space applications, then the resulting
space mechanism will exhibit superior performance and reliability.

In addition to understanding motor design and tradeoff concerns, a close review of space motor heritage
to determine what has been used on previous similar space applications is important for assessing what
to use on new applications and how to specify the unique requirements. Design considerations related to
motor selection are gear reduction, motor drive electronics, bearings, lubricants, and system dynamics.
These associated considerations can significantly affect the success of a motorized mechanism and will
often come up in a review of heritage mechanisms and lessons learned. While a thorough review of these
associated concerns is beyond the scope of this paper, they are indicated as important related factors.

Technical Overview

In order to illustrate the variety of motor selection, Table 1 “Motor Design Overview” shows a list of the
more common DC motor types. Because more than one type of motor may be used for any given
application there are tradeoffs to consider. Additionally, power consumption, envelope, torque
characteristics, life, and drive electronics complexity usually weigh in as major concerns.

*
Moog Inc., Technical Office, Research Triangle Park, NC
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 89
 Table 1. DC Motor Design Overview
DC Motors Design Characteristics

Brushed DC Motors Brushes control power distribution in the windings (internally commutated.)
Simple DC supply operation. Brush life limited.
Brushless DC Motors Two or three phases. No brushes. External drive electronics and position
sensing for commutation are required. Versatile over broad speed-torque
ranges.
Conventional DC Stepper Motors Motor cogging and damping are required. Rotates in discrete steps without
external commutation sensing. Drive electronics required. Performance
analysis is complex.
Hybrid DC Stepper Motors Unique lamination pattern permits a small step angle (typically 1.8 degrees) in
a small diameter. Requires drive electronics but no commutation feedback is
required.
Specialty Brushless DC Motors Custom variations of BDCM that provide special performance such as for
extremely low torque ripple or low motor losses at high speeds. High motor
constant is sacrificed.
Toroidal Brushless DC Motors Magnetic circuit iron ring is wrapped with windings. No cogging. High motor
constant is sacrificed for smooth rotation or large limited angle of rotation.
Difficult to wind.
Ironless Brushless DC Motors Non-metallic stator slots, large air gap, and smooth rotation. Magnetic circuit
iron ring lies outside coil windings. High motor constant is sacrificed for
smooth torque and low losses.

Understanding and Specifying DC Motors

Characteristics of DC Motors
While it is commonly understood that electric motors convert electrical energy into a rotating mechanical
output, it is less commonly understood how this is accomplished with DC motors. Not unlike their AC
counterparts, the DC motors require an intricate assembly of magnetically permeable materials, electrical
windings, and conductive media. DC motors, however, utilize permanent magnet fields to set the stage
for dynamic interaction when direct current electricity is applied to the motor windings. Energized
windings produce opposing magnetic fields that create a rotating force or torque. Power must be correctly
sequenced in the motor winding phases to perpetuate the rotation through the process of commutation.
One of the two members, either the wound member or the permanent magnet member, freely rotates and
transmits torque to the external load. DC motors generally share common design, specification, and
operation terminology to describe how this occurs.

Examination of the various types of DC motors will clarify which designs have wound stators with
magnetized rotors and which do not. Generally, the brushed DC motor is the exception, with its wound
member rotating inside of the magnetized structure. For the majority of DC motors, the laminated and
wound member is commonly referred to as the “stator”, and it is usually the stationary non-rotating
member surrounding the rotating member. The permanent magnets are located on the inner, rotating
member referred to as the “rotor.” Additional circuitry may be required to assist with the commutation or
distribution of power into the motor phases in the form of a copper commutator, Hall effect sensors, or a
resolver.

Stators (or an armature if it is a brushed motor) rely on a bonded stack of magnetically permeable metal
laminations that serve two functions. First, they are slotted to permit the installation and retention of the
electric motor windings. Second, lamination stacks are magnetically permeable having a material that is
well suited for carrying magnetic flux produced by the magnets. If the motor lamination is designed
correctly, it will properly conduct and channel the magnetic flux while maintaining sufficient clearance for
the windings. The stack is laminated rather than made as a solid to restrict the circulation of electrical
currents that are induced by rotation of the magnet fields. These speed-dependent, “eddy currents”
decrease the net output torque and wastefully consume electrical power. By laminating the stack, the
insulation between the laminations serves as barriers to restrict eddy current flow and associated losses.
By adjusting the lamination thickness and material, the losses can be further reduced depending on the
criticality of the application.

NASA/CP—2002-211506 90
DC motor windings vary considerably depending on the type of motor. Windings usually have a symmetry
that is based around the number of motor poles and desired number of motor phases. In the case of
stepper and brushless DC motors, windings may be two, three, or other numbers of phases, but two or
three is typical. Each phase consists of a number of coils that is based on the number of motor magnets.
Winding coil span, turn count, and wire size play critical roles in providing the desired performance.
Torque capability, speed range, and winding inductance are directly related to the number of turns per
coil and coil placement. Figure 1 shows two common phase and connection diagrams, “wye” and “delta”,
used for three phase brushless DC motors and stepper motors.

Figure 1. DC Motor Winding Phase Diagrams

Motor rotors usually contain the permanent magnets. An even number of magnets is always placed on
the circumference of the rotor alternating between North and South poles. Older designs use Alnico
magnet materials polarized axially along the rotor circumference. Newer motor designs use magnets
made from rare earth materials such as samarium cobalt, and they are usually polarized perpendicular to
the rotor circumference. The perpendicularly oriented or “radial” magnets force magnetic flux directly into
the space that separates the rotor outer diameter and the stator inner diameter for rotational clearance.
The clearance air gap between the rotor and stator is a critical feature that affects practically every aspect
of the motor design.

Under the broad category of DC motors, it should be mentioned that some motors are characteristically
DC but in fact operate from alternating sinusoidal drivers. This apparent conflict is reconciled if the issue
of motor commutation is examined. While most DC motor types rely on commutation that brings rotation
and torque through discrete phase switching, certain types do not. Sine / cosine driven brushless and
stepper motors are designed with permanent magnets to produce torque when continuously alternating
sine and cosine currents are applied to two or three phases. With the proper continuous position
feedback from an integral resolver, these motors are capable of producing very smooth, almost ripple free
torque output.

The major issue for DC motors for space applications is to understand that there are tradeoffs between
types of motors and their features. Certain types of designs lend themselves more readily to meeting
certain types of requirements, which is explained further. And, for new motor designs, there are tradeoffs
that can be made within the realm of slot-pole combinations, materials, air gap, and windings to optimize
the design for a specific application. Before examining specific motor types and their features, it is
necessary to understand the general terminology used to specify their electrical, magnetic, and
mechanical characteristics.

Electromechanical DC Motor Specifications

It is important to understand key motor design differences and how the primary requirements should be
specified for every application. As a foundation for this, there are common design factors associated with
all DC motors regardless of their unique design features. The common factors may be derived from the
fact that DC motors use permanent magnets, magnetic iron to establish flux paths, and copper windings
with multiple phases. Hence, the common terminology for DC motors is shown in Table 3. While there is
some variation within the industry in terms of symbols and units, the constants apply universally.
Manufacturer’s literature is readily available that explains motor terminology in more detail.

NASA/CP—2002-211506 91
 Table 2. DC Motor Terminology
Description Symbol Definition
Motor Constant Km Km defines the motor’s output stall torque capability in terms of power
consumption. Generally, it is a factor allowing comparison among
motors. Km is calculated as:
0.5
Km = Stall Torque / (Power Input)
Or
0.5
Km = Kt / (R)
which are equivalent expressions.
Torque Sensitivity Kt Kt defines the motor’s sensitivity to input current or ability to produce
torque as a function of input current:
Constant
Kt = Torque Output / Current Input

Kt is calculated during the motor design analysis based on the motor’s

number of poles, magnetic circuit flux, and total winding conductors.
Testing of the motor allows confirmation of the calculation.
Back EMF Constant Kb Kb defines the motor’s theoretical voltage requirement to achieve speed.
Depending on the units of analysis used, Kb is equal to Kt or differs from
Kt only by a constant.

Kb = VDC / unit of speed.

Winding Resistance R R defines the resistance of the motor winding(s) across which operating
power is applied. In some cases, phase windings are tied together for
excitation, so motor R is this combined value.

These basic constants help define how the motor will interface from both a power input and speed-torque
output standpoint. The application of these factors into a meaningful equation is shown in Equation (1):

Vapplied = (τ / Kt ) • R + (ω • Kb) Eq.(1)

where,

Vapplied is the voltage applied to the motor windings across R

τ is the gross motor torque in units consistent with Kt
Kt is the torque sensitivity constant for the motor
ω is the motor speed in radians per second
Kb is the back emf constant of the motor
R is the motor winding resistance in ohms.

Equation (1) expresses the required motor voltage in terms of two components, the voltage due to torque
production and that due to achieving speed. In the first term, (τ / Kt ), is an expression of the required
current to produce the torque τ . When multiplied by resistance, the voltage required for torque production
is expressed. The second term, ω • Kb, is the voltage required to achieve the operating speed, ω.
Generally, the sum of these terms indicates the minimum current and voltage required for operating at
torque τ and speed ω. Equation (1) is useful when starting with a known speed, torque, and power to
determine the approximate motor winding. Or, if the motor winding is known, then the equation may be
used to estimate the required voltage and current to achieve a certain speed-torque. Note that once Kt
and R are known, the motor constant, Km, may be obtained for comparison of motors.

Equation (1) may generally be used in evaluation of all DC motor types; however, there are modifications
required depending on the motor type and its application. The torque τ shown in fact must include the
combination of output torque, internal drag torque, low-speed cogging torque, and speed related viscous
torques. R is taken as the winding resistance without any temperature effects. But in practice, if there is
thermal rise, R must be increased (or decreased if cold) to account for winding temperature changes.
Equation (2) indicates how R changes with temperature:

NASA/CP—2002-211506 92
R2 = R1 • (234.5 + T2) / (234.5 + T1) Eq.(2)

where

T1 is the ambient temperature of the motor windings in degrees C, typically taken as 25 degrees C
T2 is the modified temperature of the motor windings in degrees C taking into account external
temperature and internal motor heating.
R1 is the motor resistance at ambient temperature in degrees C
R2 is the motor resistance at the modified temperature in degrees C taking into account external
temperature and internal motor heating.

Eq. (1) for Brushed DC Motors

The basic performance Equation (1) applies, but there is a special set of concerns due to mechanical
brush action. Both minimum and maximum supply voltages must be observed. It is important to assure
sufficient voltage is present for adequate contact and rotation. The brushes require a small voltage
associated with their electrical contact, called “Brush Drop” that subtracts from available motor
performance voltage. In the other case, the maximum voltage can be detrimental during commutation
depending on the commutator bar design, condition of the copper commutator, and the ambient air
pressure. The result of excessive voltage can be damage to the motor or total motor failure. Because the
brushes mechanically contact the motor copper commutator surface, if the action is not smooth, the
brushes can “bounce”. Brush bounce, which is observable on an oscilloscope, will reduce the effective
voltage applied to the motor causing a reduction in speed.

Eq. (1) for Brushless DC Motors

Equation (1) must be modified to include running losses due to high-speed effects. Using either empirical
data or a viscous loss constant, the torque τ would be increased by the amount of the viscous torque at
the high speed. Usually, this is in result of hysteresis and eddy current losses due to changing magnetic
fields in the motor stator lamination iron. To minimize these losses, it becomes a design issue of
lamination material selection and the magnetic flux density permitted within the laminations. Rotor
windage loss torque may need to be added at very high speeds. In some cases of high-speed operation,
the motor inductance may require additional voltage.

Eq. (1) for Stepper Motors

Equation (1) must be modified similarly to the brushless DC motor, but there are additional concerns. The
stepper motor and specific application system dynamics must be included in evaluation of a stepper
motor. A stepper motor’s response is not as independently predictable as a brushless DC motor -
assessment of the system mechanical and dynamic performance must be made. Several techniques may
be used, sometimes together, to assess stepper motor performance in the system application.
Essentially, since the stepper motor rotor position is changed by short power inputs that leave the rotor to
settle into its new position on its own, the combined motor and system dynamic effects are an important
issue. The torque output required to move the load cannot simply be predicted using Kt • I. Additional
torque margin is required to address the specific application concerned.

DC Motor Classifications and Design Features

Brushed DC Motors
Brushed DC motors, as the name implies, contain graphite based electrically conductive brushes to
conduct the DC supply power to the motor windings on the inner, rotating member. Not only do the
brushes provide the means of contact from the stationary portion of the motor to the inside rotor with
windings, but the manner in which contact is made serves to distribute or commutate the motor power.
This is accomplished because the motor brushes “ride” on a system of electrically isolated copper bars
that form the motor commutator. The bars lay side by side and are machined to provide a smooth, round
surface for the brushes. Each bar is joined to a different portion of the motor winding, so that as the
brushes change bars during motor rotating, the DC supply is carefully switched across different bars. This
means of commutation required no external switching of power to maintain rotation. Only a simple
application of a DC source is required across two motor leads.

NASA/CP—2002-211506 93
Brushed DC motors have magnets placed in a stationary ring on the outside of the rotor assembly.
Magnets typically have a radial orientation with flux going directly into the air gap towards the laminated
rotor stack. It is necessary for the magnets, windings, and commutator to be “clocked” with respect to
each other. Hence, the magnet ring must be adjusted by design or after installation for optimal position.
This may also be accomplished by adjusting the brush location, but that is usually impossible with housed
motors. The adjustment fundamentally assures that the phasing between the armature, magnets, and
commutator is optimized for performance.

Windings fill the rotor slots, which open on the rotor outer diameter for insertion. The winding style is
typically simplex wave-wound, with identical coils uniformly placed in motor slots. Inter-connections
among the coils are made at the commutator bars. The end result is that all coils are connected together
but differences in excitation polarity occur depending on where the brushes contact the bars. The spacing
of brushes, winding contacts, and coil spans must be correct for acceptable operation.

So, for space applications, there are several over-riding concerns that must be addressed when using
brushed motors:
(1) Is the motor design conservative enough to provide acceptable brush wear?
(2) Is the motor brush life expectancy adequate to meet the application life requirement?
(3) Will the power and performance be met with the brush motor under consideration?

It is prudent to verify these (and other) concerns both by analysis and by test demonstration. Chief of
these concerns for a space application is whether or not the brush life exceeds the application
requirement, since it will be impossible to replace worn brushes in space.

Brushless DC Motors
Unlike the brushed motors, brushless motors are designed to eliminate the brushes with their wear,
speed, and life concerns. Brushless DC (BDC) motors have become a very prominent choice for space
applications, since these concerns are eliminated. However, brushless motors require a considerable
amount of engineering to provide the commutation and drive circuitry external to the motor. Instead of
employing brushes and a copper commutator, BDC motors require device(s) to feed motor rotor position
information to the drive electronics (see Figure 2.). Hall effects, encoders, resolvers, or other means may
be used to supply this information, but the feedback devices must be an integral part of the rotating motor
shaft assembly.

Figure 2. Two-Phase, BDC Motor with Integral Dual Resolvers for Commutation

The issue of electronic commutation is of primary importance, since it is the rotor position information that
is used to control which motor phases are powered at any given time. Note that this feedback is critical to
achieve smooth motion for closed loop system control. Without a well-regulated position feedback and

NASA/CP—2002-211506 94
commutation scheme, the BDC motor will behave irregularly. Two approaches for BDC commutation are
common:

(1) Six-step per electrical cycle method using three Hall effect devices for three phase motors (Hall
effects are solid state magnetic field sensors used to indicate the rotor position. When three Hall
effects are used together and properly located to sense the rotor magnets, they can indicate rotor
position based on their combined state information. Hall effects usually are latching devices that
change between high or low states depending on the magnetic field present. Absolute rotor position is
not required for commutation.)

(2) Continuous commutation method using a resolver for two and three phase devices. (The continuous
sine / cosine output from a resolver is usually converted by analog to digital converters into a data
stream they gives constant rotor position data necessary for commutation. With the digital data, the
number of motor phases allowable is not fixed to two. Before more recent A to D converters, the
resolver electrical “speed” had to be matched to the motor pole count. An absolute position single
speed resolver is not required.)

If a BDC motor is back-driven, using an external motor, then it becomes a generator. Examination of the
back-driven motor voltage reveals several key pieces of information about how well the motor will
produce torque, achieve speed, and commutate. The voltage produced is called the back EMF
(electromotive force), and if viewed on an oscilloscope, generally approximates one of two styles of
waveform: trapezoidal or sinusoidal. The back EMF voltage waveform can be measured on an
oscilloscope to indicate what the back EMF constant “Kb” is (volts required to attain rotational speed) and
what the motor torque sensitivity “Kt” is (torque per unit current.) By examining the waveform, the shape
may be described as flat-topped (trapezoidal) or sinusoidal. This shape is important, because the
trapezoidal back EMF lends itself to commutate with a six-step, three-phase brushless dc motor driver.
The more sinusoidal motor back EMF lends itself to function with a brushless DC sine / cosine driver. The
motor magnet and winding designs play important roles in shaping the back EMF.

The brushless motor generally has magnets mounted on the rotor assembly that lies within the wound
stator assembly. Since BDC motors come in many sizes and design schemes, it must be understood that
quite a bit of variation can exist in rotor constructions. The magnets typically used in newer designs are
radially oriented rare earth cobalt and do not require any “keepering” like alnico motor designs. The rare
earth cobalt magnet material is generally resistant to temperature changes and provides a stable
magnetic field for widely varying space temperatures. BDC rotor assemblies may have bands over the
magnets due to containment concerns or magnet chipping concerns associated with installation.

A BDC stator is the outer, stationary member that contains copper windings inserted into lamination
stacks. The lamination material, thickness, and number of slots are fundamental to the stator design.
Certain lamination materials, such as the silicon steels, are better adapted to handling higher magnetic
flux densities and so are suitable for high motor constant, small air gap designs to obtain the most torque
with minimal envelope and power. Laminations for low-loss applications generally do not handle high
magnetic flux densities as well, but are necessary for high speed, low-loss motors. Thinner laminations
also will reduce motor losses. The number of lamination slots is a function of the number of poles, but it
factors into the magnetic design significantly. The larger the lamination slot area is to accommodate wire
conductors, the less the available metal to carry flux at a reasonable magnetic flux density. Hence, the
amount of copper volume must be constrained as well.

A BDC motor for high torque output with a high motor constant will generally be commutated with the six-
step method and Hall effect devices. It is expected that high-energy product rare earth cobalt magnets
and a high pole count will be used. Further, the lamination material will likely not be a low loss grade but a
grade suitable for handling higher levels of magnetic flux density produced by the high-energy product
magnets. The pole count will drive the number of lamination slots to the practical limitation that the motor
diameter will allow and still leave significant volume for winding copper. Windings will use coils that span
the distance of adjacent pole centerlines (full pitched coils.) The air gap will also tend to be as small as

NASA/CP—2002-211506 95
possible for the size motor chosen. This approach is fairly standard for a low speed range, high motor
constant and output torque BDC motor.

Figure 3. Extremely Low 0.2% Torque Ripple, Two-Phase, 24 Pole BDC

Torque Motor for Hubble Telescope Fine Pointing Mechanism

A bit more design finesse is required when other requirements come into play. Two common specialized
requirements are for low-loss, high-speed BDC motors and low torque ripple disturbance BDC motors
(see Figure 3.) To attain high levels of specialized performance, the motor constant cannot be the primary
requirement. Hence, while envelope and power concerns are present, motor design features must
change to meet the specialized application. The magnetic circuit must change to permit use of lower loss
materials at lower magnetic flux densities. The air gap will be increased, and unique tooth and pole
features may apply that will further impact the motor constant.

For high speed, low-loss designs, the six-step method with Hall effect devices is the better commutation
choice. Hall effects are suitable for high-speed mechanisms of 50,000 rpm or more if the pole count is two
or four poles. Intense design effort will be focused on minimizing lamination losses, which go up both with
increases in speed and the number of poles. It is entirely possible that extremely thin laminations or no
laminations at all will be used, depending on the application. Circuit back iron may surround the windings
but rotate with the magnets to keep losses low. The obvious tradeoff loss is the decrease in the motor
constant.

At the other end of the speed-torque spectrum, low rpm BDC motors with very low torque disturbance
characteristics (see Figure 3) are harder to optimize than either high Km or high speed motors. These
applications require more consideration about the method of commutation and the drive electronics. Two
and three phase BDC motors have been used successfully. Once the method of commutation has been
chosen, then the motor design details diverge. With three-phase Hall effect commutation, magnet and
winding design features are different than those chosen for the BDC two-phase, sinusoidal method,
primarily due to concerns about the motor back EMF waveform. In general, the commutation method is
likely to be driven by the system requirements and concerns rather than giving the commutation method
consideration first. Therefore, the motor design for a low torque ripple disturbance application will require
optimization based on:

(1) System level concerns and requirements (system level of acceptable torque ripple and disturbances
will be established)
(2) Commutation and drive electronics scheme chosen (two or three phases, six step or sinusoidal will
usually be directed requirements from the system level)
(3) Trading off low magnetic cogging versus no cogging (should the motor have laminations or go
“ironless”)

NASA/CP—2002-211506 96
(4) Upper limit of motor torque ripple permissible (depending on percentage the design will undergo
further changes…typically motor torque ripple less than 1.0% requires an extraordinary design.)
(5) Manufacturing concerns (the more specialized the design, the more critical the manufacturing
procedures will be driving yield and cost)

It is important to understand that being overly conservative in specifying very low levels of torque ripple
and disturbance will result in significant technical challenges and much higher manufacturing complexities
and cost. The temptation to apply significant margin to torque ripple requirements should be resisted.
System options should be considered carefully, since they tend to provide the most direction for the motor
design.

Stepper Motors
While there are many design similarities between stepper motors and BDC motors, there are some
profound differences. Stepper motors really fall into a category by themselves, at least from an application
standpoint. Unlike BDC or brushed motors, steppers are designed without a positive feedback
commutation method. While the motor windings are commutated by the drive electronics, there is no rotor
position feedback to assure that the steppers are following the commands to step and rotate. To the
extent that the stepper motor and its load follow the commands, the system is said to be “synchronous.” If
the motor or load break out of the command sequence, then the stepping is “non-synchronous.” A major
portion of the stepper design should focus on determining how likely the system is to remain
synchronous. Stepper motor design must include consideration for the system mechanical dynamics and
the electronic driver. To this end, selection of a stepper motor for an application needs to include a
system level analysis.

Stepper motors are designed to provide a repeatable angular rotation for every input power pulse
command. While the two, three, or four phase stepper motor excitation must follow a certain sequence,
no external rotor position sensing is required to instruct the electronic driver when to go to the next
sequence. This feature simplifies the system by eliminating rotor position feedback and related
electronics. However, the control electronics must be designed to command the stepper within a pre-
determined range of speeds so that the stepper does not “lose step” with the commands. The process of
analyzing the system dynamics to establish the permissible command limits can be complex. If the motor
and system are designed and analyzed correctly, simply counting the number of motor step commands
given and multiplying it by the motor step angle will provide the output position.

Conventional stepper motors are designed and constructed much like BDC motors with wound stators
and magnetic rotors. Because stepper motors run open loop, the presence of magnetic cogging and
internal motor damping are important in stabilizing the stepping action for each step. So, while BDC
motors try to minimize cogging and damping, stepper designs will use smaller air gaps to assure
magnetic cogging is present and intentionally permit running losses to create damping effects. Other
design concerns peculiar to steppers exist such as assuring the positioning during powered steps
remains the same when power is removed.

Step size usually indicates a range in the stator diameter, because a smaller step angle requires larger
diameter motors. Step angle is an inverse function of the number of phases and the pole count. So, to
obtain a small step angle, a typical two-phase stepper motor will require a larger pole count and hence a
larger diameter. For example, a motor with fifteen degrees per step can require a 75-mm outer diameter.
Conversely, large angle steppers (45 or 90 degrees per step) are fairly small with 25-mm diameters.
Because of the need for achieving small step angles within a reasonable diameter, alternate design
approaches have been developed, such as the hybrid stepper motor.

The hybrid stepper motor can provide a 1.8-degree step angle within a 50-mm diameter. It is hybrid in the
sense that it has features of both permanent magnet and variable reluctance motor types. Some hybrids
are larger and some smaller depending on the torque needs. Hybrids have been used successfully in
space mechanisms. Their design approach relies on “tricking” the motor into believing it has many more
poles than it actually does. A two-pole disc magnet is oriented axially and sandwiched between two steel
pole faces directing flux radially towards many stator teeth. Each rotor pole face is machined with

NASA/CP—2002-211506 97
(typically) 50 teeth, causing the motor to function as if it had 100 poles. This apparent increase in the
number of poles is a key feature in obtaining the small 1.8 degree step for each powered input pulse.

Other non-conventional methods exist for providing small angle steppers in envelopes of 50 mm or less.

A thorough discussion of stepper motor design and application issues is complex. However, there are
several key design and application issues that need to be addressed:

(1) Simulating the stepper motor in the system; algebraic analysis is not adequate
(2) Understanding the types of margin used and how much margin is needed for the application
(3) Power supply options and restrictions including current limiting, pulse width limiting, voltage, and
power
(4) Operating speed and torque ranges with attention to need for a start-up ramp
(5) Special system requirements such as unpowered holding torque or ratchet torque limitations of
harmonic drive gearing
(6) Step size small enough to assure mechanism resolution is met
(7) Damping, magnetic cogging range, drag torque, bearing torque variations over temperature.

DC Motors in Space Applications

A thorough review of spacecraft mechanisms and their special requirements is beyond the scope of this
paper. Instead, the characteristics of the most common systems are presented. Similar systems use
similar drive motors for very good reasons. The most common applications of the various motor types are
presented along with cautions and lessons learned.

Solar Array and Boom Deployment

Surprisingly, these large and relatively heavy systems consume very little power during deployment. For
various reasons, deployment rate is slow. This results in a mechanical power requirement as small as one
watt, even on large systems. A key requirement is to have substantial torque margin in the drive system.
Unexpected events such as dirty bearings or a panel that snags part of a solar blanket can prevent
deployment unless there is sufficient power to overcome these anomalies. Some systems use spring
deployment with damper control. If a motor is selected, it is generally a brush or BDC motor. The brush
motor is often acceptable due to the short life requirements. Having special control capabilities, brushless
motors are successfully used by many systems manufacturers. Also, brush motors were a significant
problem in the early days of space operation. Advances over the years make them viable choices as long
as the potential problems are understood and past lessons learned are not repeated. Life in a vacuum
can be very short, typically from 10 to 400 hours. This is due to accelerated brush wear.

Why not use a stepper motor? This is certainly possible but there can be unexpected problems. When a
stepper motor drive is loaded with a high inertia, there is the possibility of instability causing erratic
operation. Also, a stepper motor can “run away” with aiding torque loads, typically at speeds several
times faster than the commanded rate.

In most applications, a drive motor runs at a moderate speed and is geared down to produce high torque
at a slow speed. Figure 4 shows a BDC motor/brake driving through planetary gearing. Note that the size
is dominated by the gearing and astronaut over-ride mechanism. Even so, the total package is much
smaller than would be required for a direct drive torque motor. Power requirements would be many times
higher using direct drive.

Very often, a solar array or boom will have self-deployment characteristics. If some self-deployment
occurs, the motor must actually act as a generator under these conditions to restrain and control the rate.
Allowances must be made in the control electronics to permit reverse current flow. Otherwise the motor
could reach a high rate of speed. Sometimes a Zener diode is used to provide a current path. When the
motor acts as a generator, it controls deployment speed much better than a damper based system.

NASA/CP—2002-211506 98
 Figure 4. Deployment Actuator for Space Station Photovoltaic Radiators

Solar Array Drives

Pointing a solar array is a complex system dynamics problem. Factors such as the large inertia, structural
compliance, low resonant frequencies, and effects of attitude control call for special measures. Accuracy
is generally not very important, a few degrees is often adequate. Since the solar array must rotate or
cycle throughout the life of the satellite, long life with low power consumption is important.

Almost universally, the stepper motor is the drive of choice. It is typically two, three, or four phase
depending on the system electronics drive selected. It can be powered in short bursts to advance the
array, and then be dormant until movement is again required. It can also be powered at very slow step
rates but the actual application of power can be on the order of 50 milliseconds for each step. In this way,
the average power is minimal and the motor is at rest most of the time, contributing to the long life
needed. It also has a high un-powered holding (detent) torque which restrains the array.

The stepper motor needs gearing to be effective. Backlash is a serious system problem so anti-backlash
or zero-backlash gearing is used. The most common gearing is the harmonic drive type. It has clear
advantages in several areas. In addition to its zero-backlash characteristics, it has relatively low
compliance. The stepper motor can “wind up” the flexible spline (cup) like a spring, taking several steps
before the array begins to move. If this were not the case, the stepper could not accelerate the mass in
only one step and maintain synchronism with the applied pulse train. Stepper motor characteristics cause
the system analysis to be critical as instability is a constant concern.

A brush motor is not under consideration due to the short life. Brushless motors have other limitations and
no real advantages over a stepper other than smoothness (desirable where extreme spacecraft stability is
needed). Detent can be designed into a brushless motor to provide un-powered holding torque. The
control electronics is more complex, however.

Antenna Pointing Mechanisms

Antenna drive systems have characteristics similar to solar array drives. They must be more accurate,
however, and usually require two-axis drives. Basic comments on solar array drives apply to antenna
pointing mechanisms also.

Again, the stepper motor is dominant for its simplicity and un-powered holding characteristics. As can be
seen in Figure 5, two harmonic drive actuators can be attached to a suitable bracket for the two-axis drive
capability. Advantages include the use of similar (or the same) actuators for both solar array drive and
antenna pointing mechanisms. Identical electronics can be used. This certainly can reduce total cost.

NASA/CP—2002-211506 99
 Figure 5. Stepper Motors in Two-Axis Antenna Pointing Mechanism

Fans and Pumps

Most fans and pumps operate at a relatively constant speed and torque, used to circulate gasses or
liquids. The Space Station has various applications in the life support system as well as general cooling
and refrigeration needs. Continuous operation for years is often needed along with good efficiency. Power
consumption is much higher than needed for pointing mechanisms due to the usual system demands.

Here, the BDC motor is king. Typically commutated with Hall cells, it is usually a three-phase design with
low resistance for high efficiency. It can be relatively quiet but good design and precise rotor balancing is
necessary. Electronic commutator capabilities can include speed or torque control as well as current
limiting on start-up.

The extremely low efficiency of stepper motors rule them out, and brush motors do not have the life
capability. The motor shown in Figure 6 operates immersed in ammonia for refrigeration. Brushless motor
design permits the rotor to be hermetically sealed. The motor and commutator magnets are enclosed in a
welded housing of stainless steel. Similarly, the stator and Hall commutator assembly is enclosed for
protection against the ammonia.

Figure 6. Three-phase BDC Motor for Space Station Ammonia Coolant Pump

NASA/CP—2002-211506 100
 Space Mechanism Motor / Mechanism Requirements Motor Types Commutation Electronic
Application from Heritage Devices Req’d Drivers

Solar Array Short life with high motor constant; gear reduction for high Brushed None None
Deployment torque output. Ability to survive hard-stop impact.
BDC HED 3-Phase
Boom Deployment High motor constant, often with redundant motors and Brushed None None
geared output.
BDC HED 3-Phase

NASA/CP—2002-211506
Caging Release High torque output and high motor constant often with Brushed None None
gear reduction and integral brakes to hold position.

Rocket Nozzle Rugged construction, short usage life. Brushed None None
Extender

Linear Drive for Speed and torque range vary quite a bit. Precise linear BDC HED / Resolver 2,3-Phase
Precise Positioning motion capability with minimal torque disturbances.
Stepper None 3,4-Phase
Inertia Wheel Smooth torque action, low disturbance. Large in diameter BDC custom for HED / Resolver 2,3-Phase
Gimbal Drive for high torque, direct drive output. low torque ripple

101
Telescope Optical Smooth torque action, low disturbance. BDC custom for Resolver 2-Phase
Pointing low torque ripple

Inertia / Momentum High-speed capability with minimal motor losses. High BDC custom for HED 3-Phase
Wheel Spinning number of life cycles. Usually supplied frameless. high speeds

Cooling Fan Demands low audible output and vibration to minimize BDC custom for HED 3-Phase
Motors noise in living areas. High efficiency, long life. high speeds

Pump Motors High speed, low motor losses. High efficiency, long life. BDC custom for HED 3-Phase
high speeds
Table 3. DC Motor Selection Guide for Space Mechanisms

Solar Array Drives Long life design, high motor constant with gear reduction Stepper None 3,4-Phase
for high torque output.

Antenna Pointing Precise angular motion with high motor constant. Gear Stepper None 3,4-Phase
Mechanisms reduction for high torque output. Unpowered holding
torque
Shutter or Cover Minimize envelope and power, often with redundant Stepper None 3-Phase
Open / Close motors. Step action or limited angle rotation are common.
BDC Ltd. Angle None None
Inertia/Momentum Wheel Drive
Typically, a momentum wheel is used for attitude control and/or energy storage. It consists of a flywheel
spinning at high RPM in an evacuated housing. Low losses and high efficiencies are the dominant
factors, since the wheel will often operate continuously for up to 15 years or longer. Normal speeds are
6,000 to 10,000 RPM. Cogging and ripple torque are secondary factors since the large inertia and high
speed tend to smooth any effect on the spacecraft.

The brushless motor has proven to be the drive of choice. In earlier years, AC motors were used but
advances in brushless technology have made them more practical and efficient. Spin motor design is very
different from the drive motors used in deployment mechanisms or fans and pumps. The magnetic air gap
is large. Weight and size are compromised in favor of low losses and high efficiency. Stator lamination
material is selected for low losses (increasing cost) and is operated at low flux densities. Alternatively, the
stator may only consist of a winding on a non-magnetic coil form. This will eliminate any losses due to
eddy currents or hysteresis in the stator laminations (there are none). While this is the best choice for
minimum losses, the motor must be larger in size and weight.

Ultra-smooth Applications
The most common spacecraft are for communications, but special purpose satellites often require ultra-
smooth operation of mechanisms. A telescope, earth resources satellite, or special experiment may need
extreme attitude stability, minimum perturbations and/or smooth operation as the spacecraft scans an
area. The motors required for this application must be extremely smooth, with low ripple and cogging.
Examples are the gimbal torquers on control moment gyros. Also, the Hubble Space Telescope needs
smooth operation for acquisition and fine pointing (see Figure 3.)

Again, the BDC motor is the design of choice, but it is not in the same design category as the spin motor
or deployment drive. It is very specialized for a particular low ripple application. Size and weight are
secondary to smoothness. The motor will be a large diameter, thin section design. Every step of the
design process must consider the main goal. The air gap will be large. Magnet and tooth proportions are
selected for low cogging and ripple. The winding is unique, with varying turns of wire in each slot. This is
to produce a smooth sinusoidal waveform. Commutation is also sinusoidal to match the torque sensitivity
waveform. A resolver or encoder is used for position information as discussed earlier (see Figure 2.)

Miscellaneous Mechanisms
Most other spacecraft mechanisms fall into the category of occasional actuation and the type of drive
motor is often not the key to proper operation. Applications may be opening a door, driving a latch,
moving a sun shield over a camera, or un-caging a system. Selection can now be based upon minimizing
cost, complexity, size, weight, or the utilization of existing electronics. Often, the design can
accommodate existing motor drives that have been designed for other applications. A simple
stepper/gearhead may do the job at minimum cost and complexity. Brush motors have a place in this
category also.
Conclusion

Given the variety of DC motor types that have been used in space mechanisms, it is apparent that the
application requirements, motor design features, and relevant heritage motor lessons learned must be
understood to assure a new motor mechanism will function acceptably. Table 3, “DC Motor Selection
Guide for Space Mechanisms,” provides an overview of the types of DC motor mechanisms and their
special requirements as a guide, but there are sure to be future space mechanism cases that will not be
clear-cut. Careful consideration and coordination from the systems level through the motor mechanism
design level should be sought as early as possible in such cases to define the optimum design approach.

For further study, the following texts are suggested:

(1.) NASA Space Mechanisms Handbook. Robert L. Fusaro, NASA/TP-1999-206988. Refer to Part III, Chapter 12 on
motor types.
(2.) Space Vehicle Mechanisms, Elements of Successful Design. Peter L. Conley, Wiley, New York, 1998. Refer to
Chapter 13 on motor types and testing.

NASA/CP—2002-211506 102
 Design and Manufacturing of Extremely Low Mass Flight Systems
*
Michael R. Johnson

Abstract

Extremely small flight systems pose some unusual design and manufacturing challenges. The small size
of the components that make up the system generally must be built with extremely tight tolerances to
maintain the functionality of the assembled item. Additionally, the total mass of the system is extremely
sensitive to what would be considered small perturbations in a larger flight system. The MUSES C
mission, designed, built, and operated by Japan, has a small rover provided by NASA that falls into this
small flight system category. This NASA-provided rover is used as a case study of an extremely small
flight system design. The issues that were encountered with the rover portion of the MUSES C program
are discussed and conclusions about the recommended mass margins at different stages of a small flight
system project are presented.

Introduction

The MUSES C Nanorover Mission

The MUSES C mission is conducted by the Japan Institute of Space and Astronautical Science (ISAS).
The MUSES C spacecraft will navigate to a rendezvous with an asteroid and then drop markers onto the
asteroid surface for targeting. The spacecraft will then descend to the asteroid surface for a momentary
touchdown, at which time a projectile will be fired into the asteroid surface. The debris generated from the
projectile will be captured in a cone-shaped collector and guided to a sample container. The spacecraft
will fire its engines to rise approximately 20 kilometers from the surface and maintain that position while
obtaining remote science data on the asteroid.

NASA is contributing to the MUSES C mission in several ways. One of the many aspects of the NASA
contribution consists of a small rover, called the Nanorover, which will be dropped off from the MUSES C
spacecraft while the spacecraft is still at an altitude of 20 to 30 meters above the asteroid and
descending. The Nanorover will then free-fall and impact the surface of the asteroid, landing at a large
enough distance from the spacecraft to prevent the plume produced by the spacecraft engines on ascent
from blowing the Nanorover off the asteroid surface. Once the MUSES C spacecraft has obtained a
sample of the asteroid and risen to its parking orbit, the Nanorover will begin its mission. The first part of
this mission involves a fully autonomous self-righting of the Nanorover vehicle combined with determining
its location using a star map.
1
The Nanorover was built by NASA’s Jet Propulsion Laboratory in Pasadena, California . The Nanorover is
significantly smaller than the Sojourner rover that was deployed on the surface of Mars in 1997. A side-
by-side comparison of the Nanorover and the Sojourner rover sitting on one of the side petals of the Mars
Pathfinder lander is shown in Figure 1. The mass and volume of the Nanorover are both approximately
one-tenth of the Sojourner rover’s mass and volume. The Nanorover is made to operate in a 20 micro-g
environment, which requires additional mobility functions that the Sojourner rover did not have.
Additionally, the Nanorover carries one more instrument than the Sojourner rover, has greater computing
capabilities, and has a higher performance camera and hazard avoidance system. The need to fit this
increased functionality into the small volume of the Nanorover meant that the component parts would all
be very small.

*
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California
1
The Nanorover portion of the MUSES C mission was cancelled and will not fly with the MUSES C
spacecraft. The Nanorover was completed through component fabrication, assembly, and some testing.
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 103
 Mars Pathfinder
Lander Side Petal

Sojourner Rover
Solar Panel

Nanorover

Figure 1. Comparison of Nanorover to the Mars Pathfinder Sojourner Rover on a Mars Pathfinder
Lander Side Petal (Demonstration Models)

Nanorover Configuration
The Nanorover configuration with four wheels and the internal components is shown in Figure 2. The
body dimensions are 140 X 140 X 85 millimeters. Power for the Nanorover is obtained from solar cells
with a switching power supply to generate the required regulation and additional voltages for the complete
system. The electronics subsystem consists of a computer; analog interface circuitry for the instruments,
mobility, and engineering measurements; brushless dc motor driver electronics for operation of the ten
motor driven functions on the Nanorover; and a radio for command and data communications directly to
the MUSES C spacecraft.

IR Spectrometer

Alpha/X-Ray
Spectrometer

Gimballed
Mirror

Strut
Loupe
Mechanism

Shoulder
Camera, Assembly
Filters, Detector Wheel

Figure 2. Nanorover Configuration

NASA/CP—2002-211506 104
The Nanorover instruments are composed of an imaging camera with three focal lengths and nine filters
with a clear position, an Infrared spectrometer for determining the molecular composition of the asteroid,
and an Alpha/X-ray spectrometer for determining the elemental composition of the surface. Additional
items in the Nanorover are sun sensors, a laser range detector for measuring distances, a source for in-
situ calibration of the IR spectrometer, surface contact sensing in the wheel rims, and unidirectional
treads on the wheel surface.

Some of the mobility capabilities of the Nanorover consist of the ability to turn itself over if it ever ends up
on its back, turn in place steering, driving velocity control from 0.04 to 200 mm/sec, raising, lowering, and
positioning of the body relative to the surface by closing and opening the struts that support the wheels
(see Figure 3), attitude determination using star scanning techniques, and microgravity “hopping” to travel
large distances quickly. The hopping maneuver is accomplished by the rapid driving of the struts together
(going from Figure 3A to Figure 3B) to achieve a vertical velocity of the Nanorover’s center of mass. The
wheels are driven forward at the same time as the struts are driven together, producing a controlled
horizontal velocity. The relative magnitude of the two velocities determines how far the rover will travel
during the hop.

(A) (B)
Figure 3. Two views showing the Mobility System’s Flexibility to Position the Nanorover Body in
Different Poses

Constraints on the Nanorover Design

Several resource constraints were imposed by the spacecraft system on the Nanorover, the usual ones
being power, mass, and volume. Additional constraints were derived from the radio communication
requirements to the spacecraft from the Nanorover, the launch vibration environment, and the free-fall
and impact method for getting the Nanorover onto the surface of the asteroid. Along with the system
constraints, the thermal environment of the asteroid posed a significant challenge for the Nanorover
hardware. The thermal environment, coupled with the mass requirements, were the major driving
requirements for the Nanorover design. These two requirement constraints, together forced the use of
some unusual solutions in the design of the Nanorover components.

Mass and Volume

Due to the significant mass constraints on the MUSE C mission, the Nanorover and all of its launch
retention, deployment, communications, and processing support equipment on the spacecraft was given
a mass requirement of no greater than 3.5 kilograms. This mass was allocated as 1.7 kilograms for the
Nanorover and 1.8 kilograms for the Orbiter Mounted Rover Equipment (OMRE). These mass
requirements for the mission led to the optimization of every component of the Nanorover and OMRE for
mass.

NASA/CP—2002-211506 105
The volume requirements on the spacecraft were limiting in all directions. The spacecraft internal
components and external electronics assemblies limited the Nanorover and OMRE in five directions. The
spacecraft’s stowed solar panels limited the Nanorover in the sixth direction and included a large dynamic
envelope for the spacecraft’s solar panel motion during launch. The zero gravity release and separation
of the Nanorover from the MUSES C spacecraft were the source of additional functional requirements.
The most notable requirement was the velocity of the Nanorover at separation had to be within a tight
tolerance region. Too slow and the Nanorover would not be far enough from the spacecraft to be safe
from the engine plumes. Too fast and the Nanorover could skip off the surface of the asteroid and go into
orbit. These requirements had a significant impact on the volume of the final OMRE and Nanorover
designs.

The Temperature Environment

The asteroid for the MUSES C mission rotates at a period of about 19 hours and experiences
temperature variations at the surface from -160°C to +110°C. The large temperature swings at the
surface are a function of the amount of dust covers the surface. It does not take much dust to insulate the
body from the external environment. The dust also has very little heat capacity, resulting in the full
temperature variation in a very short time frame when the surface goes from day to night or night to day.
The quick temperature change and the large temperature range are very similar for any small body with a
layer of dust.

The Vibration Environment

The vibration environment for low mass items can be very harsh. Since a low mass component likely has
a high natural frequency in its assembly, the response to a random vibration environment can result in
very high accelerations. Force limiting during vibration testing, a technique that utilizes the characteristic
that items will not respond to high frequency accelerations because their mass filters out the high
frequency components, is not applicable for low mass items. The transmitted accelerations that many of
the Nanorover components responded to, and had to be strong enough to handle, were above 200 g’s.
The magnitude of the Nanorover random vibration test environment was 32 g’s RMS. Small mass
systems respond to the high frequency components of the vibration environment as well as the acoustic
portion of the environment. Both sources of vibration generate significant loads on the small components
that make up a small mass system. The vibration and acoustic induced loading drove all of the Nanorover
component designs.

Electronic board assemblies contain electronic components on the circuit boards that are small and low
mass. Electronic assemblies also generally contain several energy dissipative materials in their design to
meet the needs of the electrical design. This situation results in the electronic assemblies usually not
responding significantly to the high accelerations from vibration or acoustic sources. Most moving
mechanical assemblies and structural components do not inherently contain significant amounts of
energy dissipative materials. As a result, the components and structure are exposed to very high
accelerations in a random vibration environment.

The Nanorover Design

Thermal Control for the Nanorover

The thermal control system for most spacecraft uses various methods to control the temperature and the
thermal variations of regions of the vehicle where thermally sensitive components are mounted. The
electronics bays, as an example, are one area where the control of the temperature is extremely
important for the functional life of the electronic components as well as the thermal cycling life of the
electronic packaging. Typical thermal control methods employ the use of thermal blankets, heat sources,
heat absorption devices, thermal conductive paths to get the heat to the rejection surfaces, and radiators
to reject the heat to the external environment.

NASA/CP—2002-211506 106
The Nanorover presented a significant challenge in the thermal control design. The vehicle has very little
external surface area due to its small dimensions and it is solar powered. Since it is possible for the
Nanorover to be in any orientation with respect to the sun (including upside down), the Nanorover needs
solar cells on all of its sides that could possibly face the sun in order to maintain power to the computer
and actuators. This resulted in the entire available area on the top, bottom, front, and back to be
populated with solar cells, as can be seen by the blue surfaces in Figure 1. Since the solar cells used
nearly all of the external area available, there was very little space left over that could be used for heat-
dissipating radiators. Another thermal control problem that a rover faces is the intimate thermal “contact”
with the surface of the planet, comet, or asteroid. This condition leads to additional thermal control
system design constraints.

The small internal volume of the Nanorover and the tight mass constraints led to a completely passive
thermal design. There was no space in the vehicle to place batteries, heater units, or thermal switches
and the free volume needed for thermal blankets was not available. The only location that a radiator could
be placed on the vehicle was the two side panels that support the wheels, struts, and shoulder
assemblies. The close proximity of the asteroid surface to these side panel radiators resulted in a heat
input from the asteroid to the Nanorover for most of the daytime functional period. A balance was struck
between improving the heat rejection to slow the temperature rise-time during the daytime operational
period and reducing the heat rejection to keep the lowest nighttime temperature within a reasonable
bound.

This balance was accomplished with several internal passive features of the rover design. To minimize
the heat input from the solar panels during the daytime, the panels are thermally isolated from the interior
of the Nanorover. The electronics, mechanisms, and instruments are all thermally tied to the optical
bench to maintain alignment and minimize thermal gradients. To handle the thermal load of the
Nanorover’s internal components, the optical bench uses the wheel struts and the side panels for heat
rejection radiators.

In order to obtain an operational period that was long enough for the required science return, the
instrument detectors had to be kept cool for a longer period of time than the optical bench. Since the
instrument detectors are mounted directly to the optical bench, some form of additional heat capacity was
necessary to slow down the temperature rise of the detectors as the Nanorover heated up during the day.
The camera and IR spectrometer instrument detectors are mounted onto an assembly that uses a
material phase change from solid to liquid to slow down the temperature rise of the detector. These phase
change assemblies effectively increased the heat capacity of the detector by a hundred times. The final
thermal design of the Nanorover resulted in most of the components being exposed to a temperature
cycle from -170°C to +110°C every 20 hours. Adding ±15°C margin for testing, the design temperature
was -185°C to +125°C. A plot of the temperature of various components of the Nanorover during a
complete day/night cycle of the asteroid is shown in Figure 4.

NASA/CP—2002-211506 107
 Instrument
Detectors
Temperature (K)

Optical Bench

Solar Panel

Time (hours)
Figure 4. Nanorover Temperature Profile for One Day/Night cycle of 20 Hours

Mechanical Components
Nearly all of the mechanical components for the Nanorover are small, with the largest piece having
dimensions about 20 millimeters in diameter. The camera barrel, which is athermalized over a 200°C
temperature change, is shown in Figure 5 with a quarter for scale. The camera has a focus set to six
meters in order to get appropriate blurring of stars to determine the location of the star’s centroid. To
image the surface of the asteroid, a two-meter focus is required. For close-up science data taking, a 70-
millimeter focus is needed. To obtain these diverse focus requirements, a mechanism that moves a pair
of lenses into and out of the optical path was developed. The mechanism, called the Loupe Mechanism,
has three positions (lens1, lens2, no lens) and is shown in Figure 6 with a one-cent piece for scale. The
largest components on the Loupe mechanism are about 10 millimeters in diameter and the smallest are
less than one millimeter in diameter.

Figure 5. Camera Barrel with Optics Installed Figure 6. Loupe Mechanism on Vibration
Fixture

NASA/CP—2002-211506 108
The actuators developed for the Nanorover are 10 millimeter, brushless dc motors with a planetary
gearbox. The actuators are capable of lasting over 100 million revolutions in a cryogenic environment.
Rotor position sensing was not initially included in the motor design because back-EMF sensing for
commutation was considered. Failure modes of the system led to the requirement for unpowered, static
rotor position sensing of the motors and a hall sensor assembly was added to the motors. All of the
motorized mechanisms (10 on the Nanorover and two on the OMRE) used this same actuator.

Nanorover Electronics
The electronics for the Nanorover are significantly more capable than the Mars Pathfinder Sojourner
rover. While the Sojourner rover had batteries to keep the memory alive and the electronics warm
throughout the night, the Nanorover has no space for such luxuries. When the sun goes down on the
Nanorover, all functions cease. The Nanorover uses Electrically Erasable Programmable Read Only
Memory (EEPROM) to store all of its code and overnight variables. Any data collected during the day
must be downloaded to the MUSES C spacecraft or lost. The electronics also contains a large gate array
that performs many of the hardwired functions. These functions include communications decoding, motor
commutation and control, data routing and switching. The electronics system also contains an analog
signal chain that provides the Analog-to-Digital conversion of the instrument data, temperature data, and
engineering telemetry. Additional functions of the electronics system include power conditioning, power
distribution, and motor drive amplifiers.

Since the Nanorover starts functioning as soon as the sunlight on the solar panels is sufficient to initialize
the computer, the electronics assembly has to operate at the temperature extremes of the Nanorover
components. The selection of electronic components that will function properly at the temperature
extremes is very limited. The cold extreme is the defining temperature. None of the suppliers of the
electronic components would rate their devices at the low temperature, so a large amount of testing was
required to identify electronic parts that would properly function at -185°C.

The electronics system was allocated a very small volume in the Nanorover on one side of the optical
bench. The large range of functions combined with the small available volume led to an electronic
packaging design based on chip-on-board technology. The circuit boards are made from standard
polyimide-glass material. The silicon is mounted directly to the polyimide boards and bond wires are
routed from the pads on the silicon chips to pads on the boards. This method eliminates the additional
part package normally used in standard electronic packaging designs. The major benefit for the
Nanorover was a significantly reduced volume for electronic components and, because less board area
was required, significantly reduced mass for the assembly. The most common method of protecting the
chip-on-board assembly is to cover all of the integrated circuits and their wires to the circuit board with a
polymer covering. The polymer has the same mass as the board material (sans copper) and represented
a significant mass increase. The final design used the chip-on-board assembly without the polymeric
covering, leaving the 0.1-millimeter diameter bond wires exposed. This approach required the use of
handling fixtures due to the fragile nature of the assembly prior to completion of the final assembly.

The chip-on-board assembly packaging technique also met the large temperature range requirements of
the Nanorover component assemblies. In addition to surviving the temperature extremes, the packaging
design also survived the thermal cycling requirement of 100 cycles of the full temperature range.

NASA/CP—2002-211506 109
 The Nanorover System Mass and Design Maturity

The MUSES C mission, like most space missions, is very tight on mass margin. From the beginning of the
Nanorover project, it was understood that mass would be very precious. The Nanorover project utilized
the standard margin policies as outlined in AIAA specification number G-020-1992, titled “Guide for
Estimating and Budgeting Weight and Power Contingencies for Spacecraft Systems”. This specification
recognizes that the smaller the mass of the finished system, the higher the mass margins need to be at
various stages of the project. A plot of the AIAA recommendation for mass margin as a function of the
stage of the project completion is shown in Figure 7. The recommendations are divided into four groups
2
based on the system mass. The figure shows the values for Class 1 system designs only.

50
Percent Mass Contingency

50-500 kg
40
0-50 kg

30

20 500-2500 kg

10
> 2500 kg

0
Bid CoDR PDR CDR PRR
Project Stage
3
Figure 7. Mass Contingency Recommendations versus Project Stage
from AIAA G-020-1992 for Class 1 System Designs Only

Every detail of the Nanorover design was optimized to minimize the mass of every component. The
motors have a mass of 3 grams and the gearboxes have a mass of 7 grams. The wheel assemblies with
their proximity sensing capability weigh in at 20 grams. The camera barrel and optics have a mass of 137
grams. The gimballed mirror for the optical path was designed utilizing a Helmholtz coil arrangement, a
[1]
beryllium mirror, and jeweled bearings to minimize its mass of 24 grams . The Infrared Spectrometer‘s
4
mass is 90 grams and the Alpha/X-Ray Spectrometer has a mass of 95 grams.

As the design matured, the mass grew but the relative magnitude of the growth was much greater than
expected. Figure 8 shows the Nanorover mass history as the design progressed from the Preliminary
Design Review (PDR) to the Critical Design Review (CDR) for various components. In nearly every case,
the assembly masses would increase dramatically in the beginning of the design as all of the functions
were incorporated into the detail parts. The next phase of the design was a mass reduction effort that

2
The AIAA specification defines a Class 1 system as “A new design which is one-of-a-kind or a first
generation device”. The AIAA specification has recommendations for systems of different levels of
maturity. This paper only addresses the Class 1 category.
3
Project stage definitions are: Bid = Proposal, CoDR = Conceptual Design Review, PDR = Preliminary
Design Review, CDR = Critical Design Review, PRR = Preshipment Readiness Review.
4
This is a partial listing of the Nanorover components only.

NASA/CP—2002-211506 110
maintained the functionality of the assembly while reducing the mass. Typically, the final assembly mass
would then rise a small amount as the components were manufactured and measured for mass.

450
400 Mobility
Components
350
Mass (grams)

300 Main
Electronics
250
200
150
100
Optical
50 Components
0
Time from PDR to CDR

Figure 8. Selected Nanorover Component Mass History

The electronics assembly had a different history than most of the other components. The design mass
dropped after the initial estimates as some of the circuitry was reduced and consolidated into a gate
array. The next phase of the electronics design determined what components could be used in the
thermal environment of the Nanorover. This resulted in several of the components being removed and
replaced, often with more components or additional circuitry. Once all of the components were verified to
operate at the thermal extremes, the final circuit and packaging design was completed. The large growth
at the approach to CDR is due to several chip-on-board packaging details that were not accounted for in
the original packaging plan. Figure 9 shows the mass history of the entire Nanorover assembly from PDR
to CDR. The effect of the electronics assembly growth near the CDR is evident in the assembly history,
even though other component groups were going down in mass.

NASA/CP—2002-211506 111
 2000

1800 Complete
Nanorover
1600 Assembly
1400
Mass (grams)

1200

1000

800
Mobility
600 Components
(for reference)
400

200

0
Time from PDR to CDR

Figure 9. Complete Nanorover Assembly Mass History

(Mobility Components History included for reference only to Figure 8)

Lessons Learned

Upon a detailed inspection of the trends, the normal mass margins that were planned for at the outset
were very inadequate. The expected growth for a small mass system needs to be much greater than the
AIAA specification suggests. The lowest mass range of the specification is from zero to fifty kilograms
(see Figure 7). This range needs to be broken into smaller groups, with the lowest range being from zero
to five kilograms. Note that five kilograms for the high end of the range represents the mass of the
completed system with contingency. Several examples of the reasons for the large percentage mass
growth on the Nanorover are listed below:

1. The motor assemblies were completed at a mass of three grams. When the need for the unpowered
and static rotor position sensing was identified, a Hall sensor assembly was added to the motors. The
sensor assembly mass was also three grams, or a 100% mass growth for the motors. This condition
would never be the case for a typical sized motor around 25 millimeters in diameter. The motor mass
would be about 100 grams and the addition of a sensor assembly would add around 5% to the motor
mass.
2. The addition of the rotor position sensing on the motors increased the cabling to the motors by 200%.
This was a mass increase of 90 grams to the Nanorover system (or the equivalent of an entire
instrument).
3. The gearbox assemblies had an early mass of 4.5 grams. The mobility system design matured to
require the gearbox output shaft to carry a moment load of 1.2 N•m. This change required an
additional output bearing to carry the moment load. The mass increase was 2.5 grams, or a 55%
increase.

NASA/CP—2002-211506 112
4. The wheel assemblies with the proximity sensing capability built into the structure of the wheel had a
mass of 20 grams. As the electronic design matured, three additional wheel-mounted electronic
components were required to improve the resonant performance of the driving circuit. The three
component masses are 1.2 grams in total, but the required addition of a small circuit board, supports,
and capacitive coupling across the rotating interface to the wheel added 18 grams. This was a 90%
increase in the wheel mass and it is multiplied times four for the Nanorover system.
5. The electronic circuit design was very sensitive to the components that could be found that would
operate at the extreme cold temperature. Some of the components required additional support parts
to perform the needed function. As electronic parts were identified that would function as required at
the temperature extremes, the component count grew.
6. The quantity of memory required for a system is a function of the amount of software code required to
perform the system functions. As the system design matured, the memory quantity requirement grew
accordingly. This was very significant because the memory chips, even in silicon only, are very large.
The radiation hard memory silicon is 1/3 the size of the Central Processing Unit. When another bank
of memory chips were added, eight additional chips were required.
7. The biggest vulnerability to mass increase is in the electronics system. While any individual
component is extremely small mass (especially with silicon chip components), the additional circuit
board area required to mount the component and route the circuitry is the driving mass increase. As
the electronic circuit design matured for the reasons stated in numbers 5 and 6, the required board
area grew significantly. The use of chip-on-board packaging technology reduced the electronic
assembly mass from 1.4 grams per square centimeter for standard packaging to 0.5 grams per
square centimeter. The increase in area of the electronic assembly due to design maturity resulted in
a 108% increase in its mass.
8. Another factor affecting the final system mass is the machining tolerances on the various
components. While a typical component mass will match the CAD models by about one to two
percent, the tolerances used on extremely small parts cannot expect the same result. The tolerances
on the Loupe mechanism components are 0.05 millimeters. This tolerance represents 10% of the
basic dimension. On typical parts, the tolerances are less than 1% of the basic dimension. This
situation results in a greater variation of the final machined component from the design value. In one
particular case on the Nanorover, the material used is Beryllium. Due to the high cost of the material,
its brittle nature, and to minimize the risk of scrapping the part, the machinist made the parts at the
maximum material condition of the tolerances on the entire part. The resulting component mass
arrived 20% higher than the design value.
9. The use of fasteners must be considered early enough to obtain the desired size. Often, the required
size for a fastener from design load considerations is not a standard size in the small fasteners. The
use of larger than needed fasteners is very taxing on the system mass for small assemblies. As an
example, the use of a #2 fastener (the Nanorover used English fasteners) over a #0 fastener is a
mass increase of 105%. Going to the next standard sizes of a #4 over a #2, the mass increase is
70%. At the larger scales, a #10 fastener has a mass that is 34% greater than a #8 fastener, for the
same length. The metric group of fasteners exhibits a similar trend. Fastener changes as design
maturity occurs in a small system design can cause significant mass increases over the same types
of changes in larger hardware.

Conclusions

The design of small spacecraft systems or components requires additional attention to issues that
normally do not significantly affect larger devices and assemblies. The maturity of a flight design almost
always results in mass increases for any system. Small systems are more sensitive to the maturity
changes that occur in the normal design cycle than larger systems.

To begin with, a certain percentage of a very small mass is not very much in absolute value. Additionally,
the growth of component masses as the detail design progresses is often a significant percentage of the
initial value.

NASA/CP—2002-211506 113
Machining tolerances for extremely small components cannot be maintained at the same relative
magnitude as they can for larger parts. This can result in the final machined components having a large
variation from the CAD system design mass.

The thermal environment and thermal control design may often lead to a large temperature swing in a
small system. The internal heat capacity of a small system is low because there is little mass. The internal
volume will often not allow the use of standard thermal control devices. This situation may be mitigated in
the future as smaller thermal control components become available for spacecraft use.

The vibration environment for small components can drive the design significantly due to the component’s
tendency to respond to higher frequencies in the random vibration spectrum. The vibration response of
the components can lead to large displacements that often cannot be tolerated. The addition of features
to reduce displacement magnitudes or to maintain material stresses within allowable ranges will require
additional mass. This additional mass in a small system can be disproportionately higher than a larger
system because the accelerations and loading are markedly higher.

The electronic system mass in a small system is very sensitive to changes in the circuit design. Small
changes in the component count add significant mass due to the additional needed circuit board area. If
the thermal environment is as extreme as the Nanorover requirements, additional time and mass margin
need to be added to allow for the identification of the electronic components that will function at the
temperature extremes of the assembly. The determination of the components that will meet the functional
needs of the electronics subsystem must occur very early in the project to prevent significant late mass
growth that will affect the entire system.

The mass contingency plans on projects with a small system mass need to be greater in the early stages
to account for the many items that will cause a large percentage growth of the system mass. Figure 10
shows a recommendation from this author for a modification to the AIAA specification and Table 1 lists
the same information in tabular form. The smallest mass grouping for estimating should be zero to five
kilograms. Future data from programs with masses in the range above five kilograms may indicate that
there needs to be additional mass groupings between five and fifty kilograms. The design of small
spacecraft systems or components requires additional attention to the issues listed above that do not
significantly affect larger devices and assemblies. Small systems are more sensitive to the maturity
changes that occur in the normal design cycle than larger systems. The recommended specification
modification in Figure 10 provides the mass growth capability needed for small system designs.

300
Percent Mass Contingency

250 Proposed
0-5 kg
200

150 Current
0-50 kg
100

50

0
Bid CoDR PDR CDR PRR
Project Stage

Figure 10. Recommended Modification to the Low Mass Group

For Class 1 System Designs of AIAA Specification G-020-1992

NASA/CP—2002-211506 114
 Table 1. Recommended Modification to the Low Mass Group
For Class 1 System Designs of AIAA Specification G-020-1992

Project Stage Percent Mass

Contingency
Bid 250
CoDR 100
PDR 60
CDR 25
PRR 0

References

1. Boz Sharif, Ed Joscelyn, Brian Wilcox, and Michael R. Johnson “Development of a Miniature, Two-
th
Axis, Triple-Helmholtz-Driven Gimbal” 34 Aerospace Mechanisms Symposium, (May 10-12, 2000),
pp. 189-198.

Acknowledgements

This work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under a
contract with the National Aeronautics and Space Administration. Reference herein to any specific
commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not
constitute or imply its endorsement by the United States Government or the Jet Propulsion Laboratory,
Pasadena, California.

NASA/CP—2002-211506 115
 Design of a Robotic Wrist and Tool-Exchange Mechanism for Satellite Servicing

Russell D. Howard∗

Abstract

This paper describes some of the technical challenges encountered in designing the Ranger telerobot
wrist and tool-exchange mechanism, and the design approaches used to overcome them. A wrist
geometry with four intersecting and independent, but not orthogonal, axes was developed. It is shown to
have considerable advantages over previous designs, particularly in extending the wrist’s angular
workspace and reducing the tool-to-wrist center distance. Also, a compact five-state mechanism for tool
exchange is described. This device ensures secure tool transfer between the manipulator and a storage
facility, while tolerating significant initial misalignment.

Introduction

The Space Systems Laboratory (SSL) at the University of Maryland has long been interested in the
operational aspects of performing work in space, such as satellite assembly and servicing. Initial studies
indicated that the productivity of a human during extravehicular activity (EVA) could be considerably
enhanced using mechanical devices to position the human, carry tools and replacement parts, provide
auxiliary camera views, etc. The SSL has built a variety of such devices and evaluated them in simulated
space operations, primarily using the neutral-buoyancy environment.

Further benefits could accrue with a multi-degree-of-freedom remotely-controlled manipulator system, i.e.
a telerobot. Such a system has the potential of reducing the number of humans required on site, and
ultimately operating by itself at sites inaccessible to human crewmembers. Steps in that direction have
already been taken by NASA with the Shuttle and Station Remote Manipulator Systems (RMS) and the
Special Purpose Dexterous Manipulator (SPDM). However, these systems have very specific functions
and do not overlap the capabilities of a human in EVA to a great extent.

As a result, in 1992 the SSL began development of a general-purpose manipulator system called Ranger.
Its candidate task set includes servicing operations from the Hubble Space Telescope and the
International Space Station, worksite preparation tasks for human EVA, as well as more generic
manipulation tasks. The Ranger program was planned to result in flight-qualified hardware capable of
performing serious work in space, rather than a series of benchtop experiments or laboratory
demonstrations. The first full set of hardware has now been built and is under test. The purpose of this
paper is to describe some of the more challenging design goals for Ranger and how they have been
accomplished.

Ranger Design Objectives

The Ranger design philosophy is based on lessons learned from a variety of sources: video records of
RMS operations, the German ROTEX Spacelab experiment and the Japanese satellite ETS-7, as well as
SSL personnel’s experience using telerobotic testbeds at NASA Marshall Space Flight Center, Johnson
Space Center, the Jet Propulsion Laboratory, the Massachusetts Institute of Technology’s Man-Machine
Laboratory, and Flight Telerobotic Servicer simulators at Goddard Space Flight Center. The SSL has also
1,2
performed several paper studies investigating the utility of various levels of satellite servicing capability.
The resulting ideas have been tested and refined with computer simulations and a series of telerobotic
3
neutral-buoyancy vehicles including the Beam-Assembly Teleoperator (operational 1984-1997), the
Multi-mode Proximity Operations Device (operational 1986-1999), the Secondary Camera Platform
(operational 1992-present), the first-generation Ranger Neutral-Buoyancy Vehicle (RNBV-1, operational

∗
University of Maryland, College Park, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 117
1995-present), and the second-generation Ranger Neutral-Buoyancy Vehicle (RNBV-2, operational
2002).

The resulting design philosophy for a satellite servicing telerobot includes the following tenets:
• Two human-scale dexterous arms mounted on a narrow base with intersecting workspaces
provide a capable platform. The arms should be capable of working in confined areas with the
same access requirements imposed for human EVA servicing.
• Work is done with tools. A rigid, repeatable tool grip is essential. A secure means of stowing and
retrieving tools from a storage facility is necessary.
• Multiple camera views are essential.
• Active compliance is desirable to avoid inadvertent damage to the worksite, particularly when
there is significant time delay in the control loop between the worksite and the human operator. A
local active compliance controller can react to (and limit) contact forces much quicker than the
remote operator.
• Simple kinematics are important. A frustrating feature commonly encountered using teleoperated
manipulators is frequent, unexpected violation of singularity constraints or joint workspace
boundaries, resulting in uncommanded motion by the manipulator or a refusal to move. Simple
robot kinematics with few (or no) offsets between actuation axes can help, along with kinematic
redundancy and a transparent scheme of managing it.
• Packaging requires careful attention. Control electronics should be embedded in the manipulator
where possible to minimize wiring length, protect it, reduce volume, and render the arms modular,
self-contained, and potentially reconfigurable.

Of course a telerobot should also be precise, reliable and capable of functioning in the space environment
as well as neutral-buoyancy simulation, i.e. underwater. The lack of existing or proposed manipulator
systems meeting these objectives prompted the SSL to develop the Ranger design.

Specific Design Problems Addressed

Some of the most challenging design problems encountered in the Ranger program occurred at the ends
of the arms: the wrist and tool-exchange mechanisms. The desire for a wrist with simple kinematics and
no objectionable singularities conflicts with the desires for large angular workspace and compact
packaging. Similarly, it is difficult to obtain a rigid, repeatable, and strong tool grip with a simple, compact
mechanism, while ensuring secure handoff to/from a tool storage device.
4,5
Initial designs were implemented in RNBV-1 in 1995 . Over the course of many simulated satellite
servicing operations, it became clear that the wrist and Interchangeable End-Effector Mechanism (IEEM)
design should be targets for improvement in the second design iteration. This iteration was completed in
2001, and resulted in the arm design described here. It has been implemented on a new neutral-
buoyancy vehicle RNBV-2 (operational mid-2002) and a parallel set of space flight-qualifiable hardware.

Design Methods
A number of candidate wrist geometries were evaluated. Each was refined to give a balance of large
angular workspace and short tool-to-wrist center distance, taking into account dimensions of the required
actuators and sensors. Kinematic analyses were performed of the leading candidates to identify
workspace boundaries and singularities, and the consequences of these on manipulator control were
investigated.

The final wrist design (Figure 1) incorporates four independent degrees of freedom (DOF) plus two tool
drive actuators, all of whose axes intersect at a single point. It can be described as a skew roll-pitch-yaw-
roll (sRPYR) wrist; “skew” because the angle between the initial roll and pitch axes is 45 deg instead of
the conventional 90 deg. The initial roll DOF rotates the entire wrist about the axis of the forearm. The
pitch DOF rotates an armature or yoke, which surrounds a spherical inner wrist. The yaw DOF rotates this
sphere within the yoke, and the final roll DOF rotates the manipulator’s output flange and any attached
tools about its axis, which extends radially outward from the inner sphere. The two tool drive axes are

NASA/CP—2002-211506 118
concentric with the final roll DOF. All six actuation axes intersect in a single point at the center of the
sphere.

Figure 1. Ranger Wrist

The approach taken for tool-exchange was to design a standardized IEEM (Figure 2) as a basis for all
Ranger tools. Each tool is bolted permanently to an IEEM, which handles the couplings between the tool
4
and the manipulator and between the tool and the tool storage device. Initial designs used on RNBV-1
were based on ideas from machine-tool tapered collets and pneumatic couplings. For the second design
iteration the goals were to reduce bulk (particularly diameter), improve tolerance to misalignment, and
allow for larger tools by changing the tool storage geometry.

Figure 2. Tool Post and IEEM

NASA/CP—2002-211506 119
These goals were accomplished with a mechanism consisting of a 7.3-cm (2.875-in) diameter cylinder 4.8
cm (1.875 in) in length, with three concentric mounting surfaces on the distal face (Figure 2). The outer
ring supports the structure of the installed tool, the middle ring couples to the manipulator’s high-speed,
low-torque tool drive, and the inner ring couples to the low-speed, high-torque tool drive. The opposite
face connects rigidly to the wrist output flange with a mechanism inspired by a camera lens bayonet
mount. When not installed on the wrist, each IEEM (and tool) is rigidly attached to one of several “tool
posts” used for tool storage. The end of the tool post resembles a mushroom, and is captured by a
rotating sleeve on the IEEM as the manipulator disconnects from the tool.

The following sections present the functional and environmental requirements for the Ranger wrist and
IEEM, followed by detailed discussion of the designs.

Requirements and Design Drivers

Design objectives fall into two categories. In the terminology used here, a “requirement” specifies a strict
threshold value for some parameter (e.g., temperature range) which the design must meet, but no benefit
accrues from improvement beyond that value. A “design driver” specifies a figure of merit; a parameter to
be maximized or minimized subject to a balance of all the other design constraints. An example of the
latter would be to minimize the distance between the tool tip and the wrist center, with a smaller value
allowing work in tighter spaces. Only the requirements and design drivers with a major impact on
configuration will be discussed here. Many others such as launch and landing vibration loads, corrosion
resistance, etc. were accounted for but had minor effect.

Environmental Requirements
The Ranger arms and tools must be able to function in the laboratory, in vacuum, and neutral-buoyancy
simulation (underwater, maximum depth 12.2 m (40 ft)). For underwater operation the arms are
pressurized with air to around 20 kPa (a few psi) above the ambient water pressure.

Analysis of likely orbit scenarios set an operating temperature range of –40° C to +125° C for most of the
mechanism, with values up to 155° C possible adjacent to motor windings. In neutral buoyancy
simulation, the typical ambient water temperature is 32° C.

Safety Requirements
The decision was made in 1996 to design Ranger to comply with the NASA safety standards applicable to
6
Shuttle payload-bay experiments . In 1999 the project passed its Phase 2 Payload Safety Review. This
effort involved identifying and classifying potential hazards associated with operating the telerobot in the
Shuttle payload bay. Each level of hazard (critical, catastrophic, etc.) then requires appropriate
safeguards. The Ranger project was the first to conform to the new standards for “Computer Control of
7
Hazardous Payloads” . Details of the overall Ranger safety approach are described in reference (8).

The adopted safety plan requires each DOF to have two independent means of measuring relative joint
angle, to ensure that actuator runaway conditions can be detected. In addition, each joint (excluding tool
drives) must have a means of initializing the relative angle measurements to known positions. This was
essential for the boundary-management scheme, designed to prevent the robot from performing
hazardous actions such as releasing an untethered orbital replacement unit (ORU) into free space, or
driving the tool into a non-contact zone of the shuttle.

For the tool-exchange mechanism, loss of a tool into space is a potential catastrophic hazard. Thus
secure transfer from tool storage to the manipulator arm and back is required, and the mechanism must
9
satisfy provisions of the appropriate NASA safety standard .

Electrical safety considerations also influenced the choice of actuators. The maximum voltage levels
considered safe for neutral-buoyancy operations was set at 32 V, and 48 V for laboratory operations.

NASA/CP—2002-211506 120
Functional Requirements
All joint axes of the wrist are required to intersect at a single point. This enables the use of a control
scheme which partitions the kinematics into two independent groups: the arm, four DOF (one is
redundant) used to position the wrist center and the wrist, at least three DOF which determine the attitude
10
of the tool . If wrist axes do not intersect at a common point, tool position and attitude are interrelated
and cannot be solved for separately.

In addition to at least three independent DOF for attitude, the wrist must include two tool drives: a low-
speed, high-torque drive providing at least 33 N•m (25 ft•lbf) at 45 deg/s, and a high-speed, low-torque
drive providing at least 0.66 N•m (0.5 ft•lbf) at 7200 deg/s. Additional gearing can be provided within each
tool as needed. One example of a tool using both drives is a gripper with a central socket drive (required
by several existing ORUs).

A six DOF force/torque sensor is required at the wrist in order to implement an admittance controller. This
allows the arm’s local electronics to control the effective compliance of the tool-tip, and react immediately
to limit contact forces rather than depending on the (remote) human operator. For accurate readings, all
forces and torques between the arm and wrist must be structurally routed through this sensor.

Since most space servicing tasks to date have been designed with astronauts in mind, human capabilities
are reflected in the task specifications. The Ranger arms (at the tool tip) are thus required to equal or
11
exceed the 110 N (25 lbf) and 33 N•m (25 ft•lbf) capabilities considered standard for a human in EVA .

Actuator choice was influenced by packaging requirements for the drive electronics (not detailed here),
which placed a limit of 10 A on actuator current.

The tool-exchange mechanism must couple the wrist’s final DOF to the structure of the tool, and
simultaneously engage the wrist’s two tool drives. Since it must operate in the neutral-buoyancy
environment, the IEEM cannot use any electrical connections or additional powered actuators; it must be
a passive device. For reliability and simplicity, no additional actuators on the wrist may be added for tool-
exchange. All tool-exchange actuation must be provided by motion of the arm endpoint.

The design of the tool storage facility must not restrict the size of tools. This was a drawback of the
4
previous design , which required every tool to pass thru a hole of fixed diameter.

Design Drivers
These parameters have been identified as having a strong effect on productivity and ease of operation of
telerobots.

The wrist kinematics should provide a large singularity-free workspace. Every type of wrist geometry has
singularities, but the angular workspace that can be reached without crossing one depends strongly on
the angular limits of each DOF and their arrangement.

The combined wrist/IEEM should be compact in all dimensions to allow operation in confined spaces. For
instance, the distance from each tool to the back of the wrist should preferably be less than that of a
human (in EVA glove) holding a comparable tool. The tool-tip to wrist-center distance is a particularly
important parameter to minimize. If the operator adjusts tool tip attitude, an arc of that radius back from
the tool tip must be clear of obstacles.

Wrist angular speed capabilities were targeted at 45 deg/s at 33 N•m (25 ft•lbf) torque output in any
direction throughout the workspace. Near a singularity, this may require much higher speeds from
individual actuators.

The primary mode of Ranger operation is with a human directly in control. Accuracy and repeatability are
not as critical in a teleoperated mode of robot control as in a fully autonomous one. However, productivity
would benefit from the provision of some simple autonomous routines such as tool changeout. Hence,

NASA/CP—2002-211506 121
that task was chosen to set target values for accuracy and repeatability. There is a tradeoff with IEEM
design; one that can tolerate large misalignment would require less accuracy of the arm and vice-versa.
Thus both an accurate wrist and a tolerant IEEM were design drivers.

Positioning aids can make teleoperated tasks much easier. One example is a wrist-mounted video
camera. Although the primary video source for the Ranger operator is the stereo pair of cameras on a
separate video arm, several different camera angles are typically required during execution of a task.
Auxiliary views could reduce the need to reposition the video arm, simplifying and speeding up
operations. Also, some servicing tasks designed for NASA’s SPDM and its end-effector, the ORU Tool
Change-out Mechanism (OTCM), incorporate targeting crosshairs which aid in visually aligning the OTCM
with the task equipment. These are offset 7.62 cm (3 in) from the task centerline to line up with the OTCM
wrist camera. If a similarly-offset wrist camera were provided on Ranger, the operator could take
advantage of those alignment aids.

Wrist Kinematics

The purpose of a robotic wrist is to follow an arbitrary commanded trajectory in tool attitude, specified for
12
instance by Euler angles (φ,θ,ψ) with respect to the forearm. In practice, wrists fail to follow commanded
trajectories when they encounter joint limits or singularities. Joint limits result from structural interference
or constraints inherent in the actuator design. Singularities occur when the degrees of freedom lose
independence, e.g. when two axes align or three axes fall into the same plane. Singularities often consist
of single points in joint space, but a neighborhood around each singular configuration must be avoided
due to the excessive joint velocities generated. At the singularity itself, joint velocities become infinite for
motion in the constrained direction.

This sort of trajectory failure can be very frustrating to the operator if it occurs frequently in the course of a
task. Violation of these constraints is hard to predict, since it depends on overall arm configuration. Even
when commanded tool attitude is held constant, a wrist joint limit can be exceeded due to motion of the
forearm. Less restrictive joint limits ameliorate this problem, and hence are a major design goal for robotic
wrists. Since singularities must also be avoided, their location within the workspace requires attention as
well.

Wrist geometry can have a strong effect on successful trajectory following. For example, consider a
trajectory of increasing φ with constant θ, i.e. circular motion around the straight-out position. With a roll-
pitch wrist, the roll angle is φ and the pitch angle is θ, so any limit on the roll angle will be exceeded at
some point, halting the trajectory. With a pitch-yaw wrist, however, this trajectory is just an oscillation in
both pitch and yaw, and can be maintained indefinitely without exceeding joint limits.

Redundancy is another means of avoiding joint limits and singularities. If an extra DOF is added, there is
now a range of joint angle combinations for each tool orientation. Control algorithms can take advantage
of this by favoring combinations which avoid the boundaries (Ranger uses a generalized inverse
10
method ). Adding a joint to a wrist creates more potential alignments between axes, but these are now of
two types. Alignments in which three DOFs of tool attitude are maintained but two joints lose their
independence are manageable. Some redundancy is lost, but trajectory following is not impaired.
Singularities where a DOF of tool orientation is lost must still be avoided.

Ranger wrist technology evolved to its present state through a series of designs over several years. To
illustrate the improvements made, the current sRPYR wrist will be compared to two earlier versions: a roll-
pitch-roll (RPR) wrist and an orthogonal RPYR (oRPYR) wrist. Each design meets the requirement for all
wrist axes to intersect at a common point, and incorporates two tool drives.

NASA/CP—2002-211506 122
RPR Wrist
This is the design implemented on the dexterous arms of RNBV-1 in 1995, which is still operational. A
variant of this wrist is used on the video arms of RNBV-1 and RNBV-2. Figure 3 shows a cross-section of
the forearm and wrist.

Electronics Bay Tool Drives

Roll Joint #1 Pitch Joint Roll Joint #2

Figure 3. Cross-Section of RPR Wrist

The drive hardware for the initial roll joint can be located anywhere along its axis. As shown, it has been
placed as close to the elbow as possible to reduce the angular inertia of the wrist about the elbow. This
feature is common to all three wrist designs discussed here.

The chief advantage of a RPR wrist is simplicity. Nearly all commercial robots have RPR wrists. The joint
drives are structurally serial and can use many parts in common. In fact, the RNBV-1 wrists share most
basic actuator parts with the elbows. The kinematics are simple; joint angles are just the Euler angles.
Ranges of the roll joints are limited by the allowable twist in the wiring routed through their axes rather
than a mechanical hard stop. The pitch joint range is limited by contact between the wrist housing and the
forearm.

Orthogonal RPYR Wrist

This design was intended to address several shortcomings of the RPR wrist. A force/torque sensor was
included, the wrist-center to tool distance was decreased, and a fourth (redundant) DOF was added. The
RPYR configuration also lacks the RPR wrist’s singularity in the straight-out position (pictured in Figure
4). The geometry is similar to that of a gyroscope gimbal.

Yaw Drive
Force/Torque Sensor

Electronics Bay

Roll Joint #2
Tool Drives

Roll Joint #1 Pitch Drive

Figure 4. Cross-Section of Orthogonal RPYR Wrist

NASA/CP—2002-211506 123
Note that the yaw axis (not visible) is oriented out of the page from the center of the wrist sphere. The
pitch and yaw actuators are offset toward the elbow using cable drives. This system causes the yaw drive
to operate differentially, i.e. the yaw angle achieved is the difference of the yaw and pitch actuator
positions. The limitations on pitch and yaw range are due to contact between housings. The final roll joint
range is unlimited since no wiring passes through.

Compared to the RPR wrist, the oRPYR wrist is much more complex and integrated. The part count
increased considerably. Wrist-to-tool distance was greatly improved due to the drive layout, although
forearm length increased somewhat as a consequence. Inclusion of the force/torque sensor also
contributed to forearm growth.

The orthogonal RPYR wrist was never built. As the detailed design progressed, its bulk increased and
angular workspace decreased, largely due to the complications associated with cable drives. Ultimately it
was shelved in favor of the skew RPYR design.

Skew RPYR Wrist

Changing the angle between the first roll axis and the pitch axis from 90 deg to 45 deg brought two main
advantages to the skew wrist (Figure 5). Interference of the tool with the forearm was much reduced,
increasing the range of pitch travel. Also, the 45 deg “layback” angle of the actuator decreased the frontal
area of the wrist significantly.

Pitch Joint
Yaw Joint

Electronics Bay

Roll Joint #2
Tool Drives

Roll Joint #1 Force/Torque Sensor

Figure 5. Cross-Section of Skew RPYR Wrist

Note that a more typical working position of the pitch joint would orient the yaw axis out of the page (e.g.
Figure 1); it has been depicted rotated 90 deg to show the yaw actuator in cross-section. The wrist yaw
drive is direct in this wrist, not differential, and the cable drives have been eliminated. The pitch joint
range is limited only by wiring twist, while the yaw range is set by housing contact. Both values are
increased substantially over the orthogonal RPYR wrist. The wrist is smaller laterally and lengthwise,
reducing forearm length by over 20%. The force/torque sensor is closer to the wrist center by a similar
amount, improving its ability to discriminate contact forces. Mechanically, the skew wrist is simpler than
the orthogonal design, though still complex compared to the RPR wrist.

Table 1 illustrates the advantages of the skew wrist design in terms of large angular joint range and
compact packaging.

NASA/CP—2002-211506 124
 Table 1. Wrist Parameters

Maximum Wrist-center to Roll Joint Pitch Yaw Joint Roll Joint

lateral Tool Base #1 range Joint range #2 range
dimension distance range

RPR 18 cm (7in) 38 cm (15 in) 576 deg 220 deg N/A 576 deg

oRPYR 27 cm (11 in) 12 cm (4.8 in) 576 deg 270 deg 90 deg ∞

sRPYR 24 cm (9.5 in) 12 cm (4.8 in) 576 deg 540 deg 220 deg ∞

Comparison of Joint Workspace and Singularities

In these wrists, singularities are not a function of the roll angles: neither roll #1 nor roll #2 changes the
internal geometry of joint axes. Thus all singularities can be located using pitch and yaw angles (where
present). Figure 6 maps the singularities and shows the pitch and yaw joint limits (singularities outside
joint limits are not problematic).
180° 180°
RPR Wrist Orthogonal RPYR Wrist
135° 135°

90° 90°

Roll #2, Pitch Roll #2, Pitch

45° 45°

Yaw 0° Yaw 0° Pitch

Pitch, Roll #1, Roll #2
-45° -45°
Yaw
-90° -90°

-135° -135°

-180° -180°
-180° -135° -90° -45° 0° 45° 90° 135° 180° -270° -225° -180° -135° -90° -45° 0° 45° 90°

Pitch Pitch

180°
Skew RPYR Wrist
135°

90°
- Workspace
45° Pitch - Tool aligned
with forearm
Yaw,
Yaw 0°
Roll #1 - Redundancy lost in
one or more axis;
-45° constrained axes are
indicated
-90°
- Singularity: wrist
Yaw drops to 2-DOF
-135° Roll #2 Yaw Roll #2 Roll #2 Yaw Roll #2

-180°
-360° -315° -270° -225° -180° -135° -90° -45° 0° 45° 90° 135° 180°

Pitch

Figure 6. Wrist Workspace Comparison

NASA/CP—2002-211506 125
 12
Coordinate systems for each wrist have been chosen using the Denavit-Hartenberg convention .
Alignment of the tool with the forearm (“straight-out” attitude) is indicated by an “X.” The “notches” in the
sRPYR workspace are due to structural interference between the tool and forearm.

The RPR wrist has a singularity at the straight-out position, dividing the workspace into two branches.
Since a region of roughly 5 deg radius around the singularity cannot be traversed, the usable RPR joint
workspace is effectively cut to 105 deg from the mechanically-allowed 220 deg. Singularities in the
oRPYR wrist are outside the workspace and can be neglected. Those in the sRPYR workspace occur in
the configuration of Figure 5, when all axes lie in a single plane. However, due to redundancy the same
tool attitudes can be obtained with non-singular configurations.

Figure 6 illustrates the gains in joint workspace obtained by advances in wrist mechanical layout. The
next question is how effectively the improvements made in joint space translate to increased range of
motion in tool coordinates, i.e., Euler angles (φ,θ,ψ).

Comparison of Tool Workspace

The RPR wrist has no redundancy; its joint angles map directly to the Euler angles. The limits in tool
attitude are just the joint limits: φ range = 576 deg, θ range = 5 deg to 110 deg, ψ range = 576 deg.

For the other two wrists, the second roll joint is aligned directly with ψ and its range is unlimited; hence
the range of ψ is unlimited. θ is determined only by the pitch and yaw angles, and φ depends upon roll #1,
pitch and yaw. Due to redundancy, for every commanded angle θ there is a range of possible
combinations of pitch and yaw. Those joints give a contribution to φ as well. The controller can select
which combination to use, so for every commanded θ there is a range of φ which can be added to the
contribution from roll #1. Thus the φ range for the wrist as a whole is the sum of the roll #1 joint range and
the range of φ available from pitch and yaw for the current commanded θ.

Figure 7 shows the θ,φ workspaces for the pitch and yaw joints of the oRPYR and sRPYR wrists. The
shaded areas indicate achievable angles. These graphs should be interpreted like a latitude/longitude
map of the globe; the right edge can be imagined wrapping around behind and connecting with the left
edge. The top and bottom edges correspond to single points at the poles.

Orthogonal RPYR Wrist Skew RPYR Wrist

0 0

20 20
40 40

60 60
θ 80 80
(deg)100
100
120 120
140 140
160 160
180 180
-180-160-140-120-100 -80 -60 -40 -20 0 20 40 60 80 100 120 140 160 180 -180-160-140-120-100-80 -60 -40 -20 0 20 40 60 80 100 120 140 160 180
φ (deg) φ (deg)
Figure 7. Wrist Tool Workspaces

For the orthogonal wrist, when θ ≤ 45 deg any angle φ can be chosen. Thus for small θ the range of φ for
the wrist as a whole is unlimited, regardless of the joint limits on roll #1. This is an advantage of PY wrist
geometry over RP geometry. However, for θ > 45 deg the controller must choose between two branches,
one centered around φ = 0 deg and the other around φ = 180 deg. The range of φ available in each
branch hovers near 120 deg as θ increases to 120 deg, then rapidly tapers to zero at θ = 135 deg. Values
of θ > 135 deg are inaccessible. For the skew wrist, the range of φ is unlimited for θ ≤ 65 deg. Above that,
the range of φ decreases smoothly from 360 deg to 120 deg at θ = 142 deg. Values of θ > 142 deg are
inaccessible.

NASA/CP—2002-211506 126
When the φ range of the roll #1 joint is added, the total range of possible tool angles for each wrist can be
seen (Figure 8). The top of the graph is truncated; the values reaching the top edge are actually
unlimited.

1080

720
sRPYR
φ Range oRPYR
RPR

360

0
0 45 90 135 180
θ

Figure 8. Tool Workspace Comparison

It is clear from the above that each Ranger wrist design iteration has significantly improved wrist
workspace in tool attitude coordinates.

Wrist Design Details

The internal layout of the Ranger wrist is illustrated in Figure 9. Roll actuator #1 is not shown; it is
structurally separate. The pitch, yaw, roll #2, and slow tool actuators are Inland Motor BMS-3302
brushless motors coupled to HDF-25 pancake harmonic drives (ratio 120:1) from Harmonic Drive
Technologies. Position sensing is provided by reflective optical encoders (Agilent HEDR-8100) with disks
mounted on the motor drive shafts. The disks are 303 stainless steel with machined slits, optically
polished and plated with aluminum and silicon monoxide. Resolution is 1920 pulses per revolution.
Auxiliary position sensing is obtained using the signals from the three commutation sensors (Hall effect)
installed in the motor, giving 42 pulses per revolution. Absolute position indexing is given by MGC-10 10-
bit absolute optical encoders from BEI Motion Systems for pitch and yaw, and single-point reflective
optical position sensors (Honeywell HLC-1395) for roll #2 and the slow tool drive. The fast tool drive is
direct, using an Inland Motors RBE-01512 brushless motor installed in the center of the wrist. Fast tool
position sensing is provided by its commutation signals. Thermistors installed in the motor windings allow
their temperatures to be monitored. The bearings and harmonic drives have phenolic ball separators
vacuum-impregnated with Braycote™ 815Z oil and are 40% filled with Braycote™ 601E grease.

All of the wrist actuators are capable of roughly twice the required torque and twice the required joint
velocities, using the permitted levels of current and voltage.

The force/torque sensor is a model 45E15A-E24ES from JR3 Inc., rated to 667 N (150 lbf) in the x- and y-
axes, 1334 N (300 lbf) in the z-axis and 76.2 N•m (675 in•lbf) about each moment axis. Wiring passes
through the centers of the yaw and pitch actuators to minimize the variation in wiring path length as the
wrist moves.

NASA/CP—2002-211506 127
There are four access covers attached with wire-in-groove devices. The mating pieces are machined with
the inner half and outer half, respectively, of a circumferential groove. When assembled, a 1.1 mm (.045
in) diameter wire is fed into the groove through a tangential hole, locking the pieces together. All openings
are sealed with Viton™ O-rings (for static seals) or Teflon™ rotary seals from Bal Seal.

Figure 9. Ranger Wrist Cross-Section

Positioning aids incorporated in the wrist include a color video camera (Watec LCL-145A), with a white
LED providing proximity lighting. The camera axis is offset 7.62 cm (3 in) from the roll #2 axis. A low-
power red laser (Quarton VLM-670-01L) is bore-sighted down the roll #2 axis to provide a visible red dot
on the worksite at the point of aim for the wrist. Ranger tools are designed with axial clearance holes for
the laser light. The camera, LED, and laser are protected by sapphire windows.

Housings throughout the Ranger arms are made of aluminum, chosen for its light weight, corrosion
resistance, and good heat conduction characteristics. However, the thin-wall ball bearings used to support
the joints employ steel races for high strength and surface hardness. The resulting mismatch in thermal
expansion coefficients means that over the required temperature range, there is a risk of bearings seizing
(at the cold end) or spinning in their housings (at the hot end). This problem was dealt with by interposing
a thin sleeve of Invar™ between the bearing and the housing. The sleeve is dimensioned to be a light
press fit in the aluminum at the maximum temperature expected. The very low thermal coefficient of
Invar™ supplies a force opposing the contraction of the housing as the temperature falls. For a correctly-
chosen sleeve thickness, the result is essentially constant pressure on the outer race of the bearing with
no slippage over a wide temperature range. The Invar™ sleeves in the yaw drive are indicated in Figure
9.

NASA/CP—2002-211506 128
 Tool-Exchange Mechanism Design Details

Tools are attached to IEEMs stored on “tool posts,” which are bolted to convenient surfaces within the
manipulator’s reach. The tool exchange sequence is illustrated in Figure 10.

Tool Post

Slotted Ring

Blocking
Latch

Manipulator
Ball-lock
teeth engaging
Ball-Locks (Released)
Slotted Ring

State 1: IEEM Locked to Tool Post State 2: Manipulator Engages IEEM

Blocking
Latch
(rotated)

60 deg

Three Manipulator
flanges engage
IEEM flanges

States 3,4: Manipulator Rotates State 5: Manipulator and IEEM

60 deg in Roll Disengage Tool Post

Figure 10. Tool Exchange Sequence

There are five states of the mechanism. In State 1 (tool stored on post) a slotted ring retains the
mushroom-shaped end of the tool post. Two ball-locks prevent rotation of the slotted ring. As the
manipulator approaches the IEEM (State 2), three teeth on the output flange engage notches in the
slotted ring, simultaneously depressing plungers which release the ball locks. The ring is now free to
rotate with respect to the tool post, but is rotationally locked to the manipulator. As the manipulator begins
to rotate about the wrist roll #2 axis (State 3), three bayonet-type flanges engage corresponding flanges
on the IEEM, locking the manipulator axially to the IEEM. At the end of 60 deg of rotation (State 4), the
mushroom end of the tool post is aligned with a large slot in the ring, and is free to disengage. As the
manipulator/IEEM translates away from the tool post (State 5), a lip on the tool post rotates a blocking
latch in the IEEM, preventing back-rotation of the slotted ring until the tool post is re-engaged to reverse
the sequence and store the tool once again. A spring-plunger detent prevents the blocking latch from
vibrating out of position while the tool is in use. This approach ensures a secure handoff; at no point is the
tool detached from both the tool post and manipulator. The tool is locked to both for a significant overlap
phase between States 3 and 4.

The mating surfaces of the manipulator output flange and the IEEM (Figures 1 and 2) resemble a probe
and drogue mechanism, which can tolerate significant initial misalignment. The interior of the output
flange is conical and the central extension of the IEEM is cylindrical. As they are pushed together they
self-center, reducing misalignment to a small value. Approximately 1 cm (0.4 in) lateral and 5 deg angular
misalignment is acceptable at initial contact. The tool drives adjust themselves to mesh rotationally at the
same time. The fast tool drive in the manipulator and the slow tool drive in the tool (if present) are
backdriveable to allow this adjustment. To successfully self-center, the manipulator should be in a
compliant control mode. Otherwise, very careful inputs from the human operator are required.

NASA/CP—2002-211506 129
 Summary

A “skew” roll-pitch-yaw-roll wrist geometry was discussed. Combined with careful packaging to maximize
joint limits and minimize bulk, this allowed the Ranger wrist to meet all of its requirements and significantly
outperform previous designs. The angular workspace has the following properties:
• φ range is unlimited for θ ≤ 65 deg; it declines from 936 deg at θ = 65 deg to 689 deg at
θ = 142 deg
• θ range is 0 to 142 deg
• ψ range is unlimited
• All singularities can be avoided
The result is a strong and relatively compact robotic wrist which should rarely fail to track commanded
trajectories. A compact 5-state tool-exchange mechanism was also developed, meeting requirements of
passive actuation, secure tool transfer, tolerance to misalignment, and arbitrary tool size.

Acknowledgements

The author would like to thank David Akin (Principal Investigator for the Ranger project), David Lavery
(Program Executive), and Gardell Gefke (Project Manager) for their enthusiasm and support.

References

1. Smith, D. B. S., Howard, R. D., et al., “Space Applications of Automation, Robotics, and Machine
Intelligence Systems (ARAMIS),” final report of NASA contract NAS8-34381 (Phase I), 1982.
2. Sullivan, B., “Technical And Economic Feasibility Of Telerobotic On-Orbit Satellite Servicing,” Ph.D.
Dissertation, Aerospace Engineering Dept., University of Maryland, 2002.
3. Akin, D. L., Lane, J. C., Roberts, B. J., Weisman, S. R., “Robotic Capabilities for Complex Space
Operations,” AIAA-2001-4538, AIAA Space 2001 Conference, Albuquerque, NM, Aug 2001.
4. Cohen, R., Akin, D. L., “Development of an Interchangeable End Effector Mechanism for the Ranger
th
Telerobotic Vehicle,” 28 Aerospace Mechanisms Symposium, Cleveland, OH, May 1994.
5. Howard, R. D., Akin, D. L., "Manipulator Design and Development for the Ranger Satellite Servicing
th
Vehicle" 26 Aerospace Mechanisms Symposium, Greenbelt, MD, May 1992.
6. “Safety Policy and Requirements for Payloads using the Space Transportation System,” NSTS
1700.7b, Space Shuttle Program, May 1996.
7. “Computer Control of Hazardous Payloads,” MA2-97-083, Space Shuttle Program Integration, 1997.
8. Roderick, S., Churchill, P., Gefke, G., Akin, D., “Design and Certification of a Total Computer-Based
Hazard Control System for a Space Shuttle Payload,” AIAA-2000-5078, AIAA Space 2000
Conference, Long Beach, CA, September 2000.
9. “Mechanical Systems Safety,” MA2-00-057, Space Shuttle Program Integration, 2000.
10. Carignan, C. R., Howard, R. D., “A Partitioned Redundancy Management Scheme for an Eight-Joint
Revolute Manipulator,” Journal of Robotic Systems, vol. 17, series 9, January 2000.
11. “Extravehicular Activity Hardware Generic Design Requirements Document,” JSC-26626A, 1995.
12. Paul, R., “Robot Manipulators: Mathematics, Programming, and Control,” MIT Press, 1981.

NASA/CP—2002-211506 130
 Operational Improvements of a Pyrotechnic Ultra Low Shock Separation Nut
*
Alexander Luna

Abstract

The design requirements and improvements in satellite platform construction have outpaced the design
evolution of pyrotechnic separation devices. Demands of the aerospace industry have quickly moved
toward smaller separation devices that can carry larger preloads and produce shock output levels that
can be tolerated by sensitive electronic and optical payloads. Hi-Shear Technology Corporation’s product
enhancement efforts have successfully brought its SN9400 series Low Shock separation nut technology
inline with industry demands by developing the SN9500 series Ultra Low Shock (ULS) separation nut.
This product enhancement has yielded a new generation of separation nuts with shock outputs reduced
75-85% below the SN9400 series separation nuts.

Introduction

In addition to the requirements of high load carrying capacity and low shock output, the aerospace
industry strives to retain the use of the highly reliable design heritage associated with standard pyro
initiator systems, such as system level use of five ampere twenty-five volt pulsed power. These simple
power allocations have been the most widely accepted for actuation of separation devices.

Two methods can be employed in meeting the demands for reduced shock separation devices, 1)
developing alternative methods of actuation and separation, 2) Improving the operation characteristics of
current pyrotechnic actuated separation devices. Hi-Shear product enhancement efforts used the latter in
developing the Ultra Low Shock separation nut. This paper will review the design of the SN9400 series
separation nut focusing on the areas that were most fruitful for reducing shock output and the successful
results that were achieved.

SN9400 Series Low Shock Separation Nut

Basic Design Parameters

The basic mechanism construction for the SN9400 is shown in Figure 1. The design parameters are as
follows:
2 2
• Shock Output: 29,430 m/s (3,000 g) – 98,100 m/s (10,000 g)
-4
• Leakage: qualified to 1 x 10 cc/second through the internal seals
• Release Mechanism: piston/cylinder
• Actuation Time: less than 10 milliseconds
• Initiation Device: NASA Standard Initiator, single or dual cartridge
• Power Requirements: 5 ampere, 25 volts pulsed power
• Mounting: flange mount
• Mating Fastener: bolt or stud

Theory of Operation
1. Application of a preload to the separation nut via a stud or bolt produces a radial load between three
threaded segments and the stud/bolt.
2. The three threaded segments and the radial load are contained by a locking ring, which is fastened to
a cylinder. The three threaded segments are mounted over a hard-mounted keyed base. These keys
lock the segments and prevent rotation.

*
Hi-Shear Technology Corporation, Torrance, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 131
3. Initiator and pressure cartridge produce high-pressure gasses in the combustion chamber formed by
the piston, cylinder and housing.
4. A displacement between the piston and cylinder occurs as a result of the high-pressure gasses.
5. The displacement of the cylinder pulls the locking ring off the threaded segments while the piston
pushes down on the segments. With the threaded segments free of the locking ring, the segments
are forced out away from the stud under the action of the radial load produced by the preload and the
downward force of the piston.
6. With the three threaded segments free from the mating threads of the bolt, the bolt will be free to eject
under the strain energy produced from preloading. It is this instantaneous release of bolt strain
energy that produces the greatest component of shock output.

SN9500 Series Ultra Low Shock Separation Nut

Basic Design Parameters

The basic mechanism construction for the SN9500 ULS is shown in Figure 2. The design parameters are
as follows:
2
• Shock Output: less than 19,620 m/s (2000 g)
-4
• Leakage: 1 x 10 cc/second through the internal seals
• Release Mechanism: piston/cylinder
• Actuation Time: less than 10 milliseconds
• Initiation Device: NASA Standard Initiator, single or dual initiators
• Power Requirements: 5 ampere, 25 volts pulsed power
• Mounting: flange mount
• Mating Fastener: bolt or stud

Theory of Operation
The theory of operation for the SN9500 ULS is nearly identical to that of the SN9400
1. Application of a preload to the separation nut via a stud or bolt produces a radial load between three
threaded segments and the stud/bolt.
2. The three threaded segments and the radial load are contained by a locking ring, which unlike the
SN9400 is not fastened to the cylinder. The three threaded segments are mounted over a hard-
mounted keyed base, which rides on a rotor cam. These keys along with the ramps of the rotor cam
lock the segments and prevent rotation.
3. Initiator and pressure cartridge produce high-pressure gasses in the combustion chamber formed by
the piston and cylinder.
4. A displacement between the piston and cylinder occurs as a result of the high-pressure gasses.
5. The displacement of the cylinder pulls three pins from the ramps of the rotor cam allowing the rotor
cam to rotate down thus relieving the strain energy of the bolt/stud.
6. With the preload relieved a coiled spring ejects the locking ring from the threaded segments.
7. With the threaded segments free of the locking ring, the segments are forced out away from the stud
under the action of the radial load and the downward force of the separator.
8. With the three threaded segments free from the mating threads of the bolt, the bolt will be free to fall
or be extracted from the separation nut.

Separation Nut Shock Output

Three main sources contribute to shock output during separation nut actuation, 1) Pyrotechnic Shock, 2)
Release of Bolt Strain Energy and 3) Mechanical Impact of internal components. These three sources
occur as three distinct events during actuation of the separation nut. The pyro shock occurs at
approximately 2-3 milliseconds after the application of current to the initiator. This event consists of the
rapid expansion of pyro gasses that do the work of actuating the internal mechanism. The second event,
release of bolt strain energy occurs at 3-5 milliseconds after the application of current. The third event,
mechanical impact of the internal components consists of the piston/cylinder mechanism completing the
stroke and impacting on the separation nut housing. This event occurs at 5-7 milliseconds after the
application of current and makes the largest contribution to shock output in the SN9500 ULS. Table I lists

NASA/CP—2002-211506 132
the percent contribution that each of the three main elements makes to the operation of the SN9400 and
SN9500 series separation nuts.
Development Testing

To gain a better understanding of the three main sources that contribute to the output shock,
development testing was performed on the SN9400 and three iterations of the SN9500 ULS. Various
experimental techniques were employed to isolate the three main sources that contribute shock output.
The data collected from testing each source of shock output was used in optimizing the final design of the
ULS. Allocating a percent contribution that each of the three sources should make during actuation set
the optimized design criteria. The percentage values listed in Table I for the SN9500 ULS were found to
be the optimal arrangement that could be controlled for maximum shock output reduction without
compromising the fastener aspect of the separation nut.

Validation Testing

The optimized design of the SN9500 ULS is shown in Figure 2. Validation testing of the shock output was
performed and compared to that of the SN9400. Testing was performed using a M12-1.75 (0.500-
20UNJF-3B) separation nut of the 9400 and 9500 series, both separation nuts were preloaded to 89 kN
(20,000 lbf). In the absence of any standardized shock output test fixture for separation nuts, a worst case
approach was taken by performing the test on a 61-cm x 61-cm x 1.9-cm (24 inch x 24 inch x 0.75 inch)
aluminum plate. The worst case approach would best serve industry since satellite structures consist of
various types of composite structures that would contribute to an attenuation of the separation device
shock output. By establishing a worst case benchmark for each separation nut, design engineers will be
able to select the appropriate size and type of separation nut for evaluation on particular structural
applications. The test set fixturing and setup used for measuring shock output is shown in Figure 3. The
test article is mounted to a test block and then mounted to the center of the aluminum test plate. The
aluminum test plate is suspended with four elastic cords attached to each corner of the plate. The shock
output was measured for each of the three orthogonal axis via a 2.54-cm x 2.54-cm x 2.54-cm (1.0 inch x
1.0 inch x 1.0 inch) aluminum accelerometer block mounted 15.24 cm from the center of the separation
nut. The shock response for the SN9400 separation nut is shown in Figure 4 with the response in the time
domain shown in Figure 5. The reduced shock output of the SN9500 separation can be seen in Figure 6
with the response in the time domain shown in Figure 7. A comparison of the test results shows a
significant reduction of 75-85% reduction in the frequency range of 1 kHz to 4 kHz. It has long been
accepted that 1 kHz to 4 kHz range is the most vulnerable for electronic and optical payloads.

Summary

The SN9400 series separation nuts have performed as a workhorse for the aerospace industry for more
than twenty years. The only prudent approach to meeting the changing demands for satellite platform
separation nuts was to approach the existing heritage technology with the aim of identifying all design
parameters that could be optimized and produce the highest gains in performance. Hi-Shear’s product
enhancement efforts brought about a new generation of an existing product line while utilizing standard
initiators, non-exotic materials and traditional fabrication methods. Results exceeded the original goal of
reducing output shock by 50% and accomplished reduction levels as much as 75-85%. Continued efforts
to increase performance at the separation device level will yield the greatest benefit to the aerospace
industry.

Table I. Shock Output Contribution from Three Main Sources

Relative
Shock Output Shock Output
Source of Shock SN9400 SN9500 Reduction
SN9400 @ 2kHz SN9500 @ 2kHz
@ 2kHz
2 2
Pyro Shock 10% 2,747 m/s (280 g) 30% 736 m/s (75 g) -73%
2 2
Bolt Strain Energy 60% 16,480 m/s (1680 g) 20% 491 m/s (50 g) -97%
2 2
Mechanical Impact 30% 8,240 m/s (840 g) 50% 1,226 m/s (125 g) -85%
2 2
Total 100% 27,468 m/s (2800 g) 100% 2,453 m/s (250 g) -91%

NASA/CP—2002-211506 133
 Figure 1. SN9400 Construction Figure 2. SN9500 Construction

Figure 3. Shock Response Test Setup

NASA/CP—2002-211506 134
 Figure 4. SN9400 Shock Response Spectrum for a 89 kN (20k lbf) preload

Figure 5. SN9400 Time Domain Shock Response, 89 kN (20k lbf)

NASA/CP—2002-211506 135
 Figure 6. SN9500 ULS Shock Response Spectrum, 89kN (20k lbf) preload

Figure 7. SN9500 ULS Time Domain Shock Response, 89 kN (20k lbf)

NASA/CP—2002-211506 136
 Alternatives to Pyrotechnics - Nitinol Release Mechanisms
*
Andrew Tuszynski

Abstract

Hi Shear Technology Corporation has been producing the SN 9400 series of separation nuts since the
early 1970's. It is one of our most popular products, and is still the lowest shock output pyrotechnically
actuated nut available on the market today. There is a strong demand in the market, especially in Europe,
for a separation nut which produces virtually no shock, produces no gas which may contaminate the
1
payload, and is easily resettable. To achieve these goals, Hi-Shear has developed and patented (pat. no.
5,248,233) a separation nut that releases the bolt preload using an electrically heated Nitinol column.

Introduction

Our standard SN 9400 separation nut is based on three threaded segments held together by a ring. The
preload on the bolt pushes these segments outwards against the ring. Friction between the ring and the
segments holds the components in place. To release the nut, the pyrotechnic initiators on top of the nut
are fired, generating a high-pressure gas. This gas pressure is higher than the frictional force and pulls the
ring off the segments thus releasing the bolt.

The shock, vibration, and thermal cycling loads encountered by the nuts are very severe. The SN 9400
series separation nuts have a long history of success in withstanding these environments and functioning
as required. The new Nitinol separation nut still uses the pedigreed segment and ring arrangement, but
the bolt preload is relieved in a much different way.

The name Nitinol stands for Nickel, Titanium, Naval Ordnance Laboratory. It is known as a Shape Memory
Alloy, and has the unique ability to return to a certain shape when heated. In our application, the Nitinol is
taught to shrink when heated, relieving the tension on the preloaded bolt. The ring is then pushed off the
segments with a spring and the bolt is set free (Figure 1). Since there are no pyrotechnics involved, the
system is very clean and easily resettable and re-useable. Each separation nut can be tested and reset
before shipment, rather than destructively testing 10% of the lot as is the case with pyrotechnic nuts. The
Nitinol nuts can be used over and over by the customer during ground system tests, rather than having to
ship them back to the factory to be rebuilt.

Separation Nut Shock Output

2
The three main sources of shock output during separation nut actuation are 1) Pyrotechnic Shock, 2)
Release of Bolt Strain Energy, and 3) Mechanical Impact of the internal components. The Nitinol SN 9600
series nut has no pyrotechnics and releases the bolt strain relatively slowly, in seconds, not milliseconds.
The mechanical impacts are driven by a spring and are much lower than those driven by pyrotechnic
activation. Since the bolt preload has already been relieved by the shrinking of the Nitinol, the required
force to move the ring is two orders of magnitude lower than what is needed to move the ring in a
standard pyrotechnic device.

The shock output of a standard SN 9400, 9.5-mm (3/8”) Low Shock Sep Nut, with a 30,000 Newton (6,750
lb) preload is approximately 4,500 g’s at about 7,000 Hz (Figure 2). A SN 9600 Nitinol Nut of the same
size, with the same preload, puts out a maximum of 500 g’s (Figure 3). This shock level is roughly
equivalent to the shock obtained by dropping a small bolt onto an aluminum plate from a height of 20 cm
(8”).

• Hi-Shear Technology Corporation, Torrance, CA

th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 137
 Nitinol Training
A column of the Nitinol alloy is at the heart of this separation nut design. Although it is used in the wire
form in the medical field, finding a supply of the size needed in a structural application has been
challenging. There have been many papers written about the various alloys in the academic community,
but useful engineering data for an available product is also very hard to come by.

Most applications of the Shape Memory Alloy require the material to return to its initial shape when heated.
A uni-axial length recovery of approximately 7.5% is easily achievable. In our application, the Nitinol is
trained to both shrink in length when heated, then to grow back when cooled, giving it a “Two Way Shape
Memory” of ± 2%. This is achieved by deforming the bar to form a mosaic of “Stress Based Martensite
3
crystals” . The material returns to its parent shape when heated past its transition temperature, but with
training, remembers these crystals, and returns to the shape it was given when cooled.

The term Transition Temperature describes the temperature at which the Nitinol alloy changes state from
Martensite to Austenite. The transition temperature of the alloy used for spacecraft applications is 90°C.
While the bar is under load, the transition temperature rises up to 125°C. The ability to train the material in
two directions gives us the ability to verify the quality of each column used in each separation nut within
each production lot. It also gives our customers the ability to reuse these separation nuts in ground tests
as needed.

Required Length
A 9.5-mm (3/8 inch) diameter bolt and separation nut system is typically expected to carry a maximum of
45,000 N (10,000 lb). A 75-mm bolt length is expected to stretch approximately 0.23 mm (0.009”) under
that load. The Nitinol column is sized in both length and diameter to shrink and relieve all the preload. In
this case the Nitinol column is 2.54 cm (1.00 inch) long and will give up 0.51 mm (0.02”).

Summary

Engineering design always involves balancing trade offs. The SN 9600 Nitinol Nut does not produce any
gas and has virtually no shock output. It will not contaminate nor jar the optics of a sensitive satellite.
However, the amount of electrical energy required to actuate the Nitinol nut is much higher. Pyrotechnic
nuts typically use 25 watts per cartridge over a 10 millisecond time span. All the energy to release the nut
is stored in the pyrotechnic charge. The Nitinol nut requires 125 watts for 15 to 30 seconds. Since the
actuation time cycle is so long, simultaneity is not achievable. On the other hand, for applications where
low shock output, or cleanliness, is of higher importance than power use and time, this is the separation
nut of choice. At this time, the engineering development is nearing completion and we have customers
interested in qualification and flight of this product as an alternative to pyrotechnics.

References
1. Patent Number 5,248,233, “No-Shock Separation Mechanism”, Richard G. Webster, (Hi-Shear
Technology Corporation) September 28, 1993

2. Alex Luna, “Operational Improvements of a Pyrotechnic Ultra Low Shock Separation Nut” Proceedings
th
of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May, 2002

3. Duerig, T. W., Melton, K. N., Stockel, D., and Wayman, C. M. “Engineering Aspects of Shape Memory
Alloys”, Butterworth-Heinemann, 1990, pp 201

NASA/CP—2002-211506 138
 Threaded Segments

Ring

Nitinol Column

Figure 1. SN 9600 Separation Nut Schematic

NASA/CP—2002-211506 139
 600

Figure 2. SN 9600 Nitinol Separation Nut

Shock Response Spectrum, 30 kN (6,750 lb) Preload

NASA/CP—2002-211506 140
 System Characterization and Motor Step Verification through Rotary Acceleration Signals
*
Scott Starin and Fred Crosno

Abstract

Stepper motor torque output, resonant frequency determination, and pulse step verification have always
been questions of the performance and reliability of stepper motor actuators. Traditionally, stepper motors
have been characterized by open loop step counting. In some instances, load position information (hence
step integrity) has been determined through the use of brush-type potentiometers or through brushless
resolvers. The disadvantage of pots is the concern of reliability and particulate wear of the brushes, and
resolvers requires excitation and additional circuitry to process the information to an analog or digital
format. The ability to use a DC transducer signal to characterize motor step performance and integrity is a
major advantage in permanent magnet stepper motor applications.

Introduction

For background and consistency purposes, we will discuss stepper motor torque characteristics, define
key terms, and establish relationships on load inertia and friction on stepper motor performance.
Additionally, we will present a classical stepper motor position versus time profile, and its corresponding
acceleration versus time. With this background information, we will demonstrate that a DC signal in
proportion to the rotary acceleration may be used to determine a number of parameters such as load
inertia, resonant frequency, overshoot, torque margin, settling time, and change in friction over time. In
addition to the characteristic and performance criteria, we will discuss several of the many options of
using the rotary acceleration information to determine step integrity, as well as direction of rotation of a
stepper motor actuator.

Stepper Motor Dynamic Performance

Figure 1 shows a typical Pulse Rate versus Torque for a permanent magnet stepper motor. Key terms
and performance points are highlighted here. There has been an unfortunate lack of consistency in the
industry for the definition of some key terms. This graphically shows terms we use in our discussions of
stepper motor performance.

Figure 1. Typical Dynamic Stepper Motor Performance under Low Load Inertia

*
CDA InterCorp, Deerfield, FL
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 141
Pull-in and Pull-Out Torque: It is not sufficient to simply specify running torque at a particular pulse rate
for a stepper motor. Pull-In and Pull-Out Torque should be understood and characterized when
designating a stepper motor. Pull-In Torque is a fairly linear relationship that falls on a line from the No
Load Response Rate to the Torque at Low Pulse Rate. The Pull-In Torque is the maximum frictional load
you may “pull-in” from rest, or change direction of rotation without missing pulses. Once synchronous
angular velocity has been achieved, it takes more torque to Pull-Out a motor from dynamic operation. The
Pull-Out torque falls along a line from the Slew Rate to the Torque at Low Pulse Rate.

Load Inertia and Resonant Frequency: The performance of a stepper motor is significantly affected by the
introduction of load inertia. Energy that was going into accelerating and decelerating the motor rotor must
go into accelerating and decelerating the load inertia and friction torque. It is important to establish the
system resonant frequency in these applications, so those motor operational pulse rates may be reviewed
to avoid stepping at a resonant frequency or sympathetic harmonics. As a rule of thumb, it is desirable to
keep the ratio of load inertia to motor inertia down below 5:1. This minimizes the possibilities of
resonance and unstable performance. However, while it is possible to safely operate with significantly
higher inertia factors, it is difficult to predict resonant frequencies and torque performance may be
significantly compromised when trying to drive at or near resonant frequencies. In these applications,
prototypes should be built to characterize system performance at the desired operational pulse rate. We
will discuss inertia and its affect on stepper motor performance later in this paper.

Step Characteristics: Figure 2 shows an actual scope trace of position versus time (top curve) and the
corresponding acceleration versus time (bottom curve) of two steps of an unloaded stepper motor. The
overshoot, settling time, and general step characteristics are clearly illustrated here. With this illustration,
we can determine a few characteristics of the stepper motor. The peak acceleration happens shortly after
the initial “pulse”. This is due to the “rise time” of the current into the motor. Also note how the
acceleration crosses zero at the instant the motor crosses the stable step position of the motor. This is
referred to as the “crossover time”. From this point, a clear indication of the overshoot and bounce region,
and settling time are clearly visible in the acceleration signal. If the power were increased or decreased
for this same motor, fundamental changes would occur in the acceleration profile.

Figure 2. Position and Acceleration versus Time for Unloaded Stepper Motor

Introduction of loads on stepper motor characteristics: When a stepper motor is coupled to an inertia or
frictional load, the characteristic plot shown in Figure 2 will be altered. With the introduction of load inertia,
the peak acceleration will be attenuated because torque is being consumed into accelerating the load
inertia. The number of overshoots and the settling time will increase, due to an increase in angular
momentum.

To compare performance with and without inertial loads, we present the following curves. Figure 3A is a
stepper motor with the following characteristics:
Holding Torque (TH) = 0.115 Nm

NASA/CP—2002-211506 142
 2
Motor + Accelerometer Inertia (JM) = 1.4E-06 kgm
2
Accelerometer Scale Factor (KVα) = 26 mV/krad/sec
Motor Coulomb Friction (FM) = 0.005 Nm

At the rated holding torque, this stepper Figure 3A

2
motor will produce 78 krad/sec
acceleration, which corresponds to 2.0
Vdc peak; very close to the actual
recorded acceleration signal shown in
Figure 3A. The maximum acceleration of
a stepper motor may be calculated by
Equation 1. Note: This equation is valid
only at low pulse rates. As velocity
increases, back emf and viscous frictions
and damping affect the peak
acceleration.

αM = (TH-(FM+FLM))/(JM+JLM)
Eq.1 Figure 3A – Stepper Motor Acceleration under No Load

where FLM is the load friction, reflected to

the motor. Figure 3B shows the
acceleration of a stepper motor with a
2
directly coupled inertia of 2.0E-6 kgm .
This increased inertia reduces the
2
maximum acceleration to 30.6 krad/sec ,
and hence the signal to 0.87 Vdc. Notice
the increase in crossover time and
settling time. It is interesting to evaluate
this characteristic profile when motor
power levels are changed. For instance,
you may experience a resonance in a
high load inertia application that may be
eliminated by decreasing the power
input, because of the reduction of the Figure 3B – Stepper Motor Acceleration under Inertial Load
percentage overshoot and undamped
resonant frequency.

Introduction of friction on stepper motor

characteristics: Friction has a different
affect on stepper motors, when
compared to inertial loads. Increased
friction also attenuates the acceleration
signal, but some energy that was being
used to displace the load is now being
dissipated in the form of heat in the load
friction. The overshoot and settling time
are also reduced, as one would expect. If
the friction increases significantly, the
motor cannot reach the next stable step
point, and cannot continue to drive the
load. Figure 3C shows a scope trace of Figure 3C –Stepper Motor Acceleration under Frictional
the same motor used in Figures 3A, Load
under a pure frictional load, (the author’s

NASA/CP—2002-211506 143
index finger and thumb applying pressure to the motor shaft). The reduced acceleration signal equates to
2
about 54 krad/sec . By rearranging Equation 1, and solving for FLM, we can determine that my finger
frictional torque is approximately 0.034 Nm, by the reduction in acceleration signal from Figure 3A.

System Characterization through Acceleration Analysis

With the background information and

characterization presented, we may expand
on this knowledge to analyze motor
performance in systems. By studying the
acceleration profile of a stepper motor, it is
possible to determine a number of important
aspects of torque margin and resonant
frequencies. It is easiest to accomplish this
goal by first testing the stepper motor actuator
outside the system, at the desired pulse
rates. From here on out we assume the
stepper motor also incorporates gearing
which serves to multiply torque and reduce
reflected load inertia to the motor. Figure 4A – Sample Actuator Acceleration under No
Load
Viewing the Resonant Frequency: Figure 4A
shows a geared stepper motor with
acceleration transducer providing the
acceleration signal shown with no load inertia
or frictional torque, running at a relatively low
pulse rate of 20 pulses per second. Figure 4B
is the same actuator installed in a system with
reflected load inertia approximately two times
the motor inertia. There was also a nominal
frictional torque applied to this system. The
apparent increase in noise on this scope
trace is actually a result of the natural circular
resonant frequency of the system. In some
geared stepper motor applications you may Figure 4B – Sample Actuator Acceleration under
find that the resonant frequency changes very Inertial Load
little with different load inertias. This is due to
gear meshing and hunting tooth ratios that may cause lower resonant frequencies than a straight
calculation of the Resonant Frequency will indicate. However, these frequencies are typically much higher
than stepper motors are driven. In the example shown in figure 4B, we observe a resonant frequency of
approximately 600 Hz.

Torque Margin: It is easy to determine the

operational Torque Margin of a stepper motor
with an acceleration signal. With the stepper
motor installed in a system, monitor the
accelerometer output signal while operating
under desired conditions (nominal or worst
case). Then load the system with frictional
torque (fingers do nicely) until the stepper
motor misses steps, or is unable to pull in
from rest. The torque margin may simply be
calculated as the ratio of the operational
acceleration signal at desired conditions, to
the magnitude of acceleration signal just Figure 4C – Sample Actuator Acceleration under
Maximum Frictional Load

NASA/CP—2002-211506 144
before the actuator is unable to step reliably. Note: It may be necessary to filter the acceleration signal to
filter out the higher resonant frequencies, and the acceleration signal gain may need to be adjusted to
compensate for the required filtering. Figure 4C is the same actuator as in Figure 4B, with an outside
frictional load increased to a magnitude just prior to the point where the actuator pulled out from
operation. By analyzing the ratio of these two acceleration signals, we can determine that we have
approximately a 12:1 torque margin at the desired operating conditions (note different voltage scale). If it
is not practical to load the system down until you miss steps, it may be necessary to test the stepper
motor actuator outside of the system as a component. If this is done, an effort should be made to simulate
the system load inertia while conducting this test.

Pulse Step Verification through Acceleration Signal Processing

There are many practical ways to process an acceleration signal to determine step integrity and direction
of rotation. Mission critical or flight safety critical applications may have a central processing system that
monitors system or mechanism integrity. For such systems, it may be most practical to convert the analog
accelerometer signal to a digital signal and generate an algorithm to analyze the acceleration signal.
From the acceleration signal, a double digital integration routine will provide a position change for each
step. Due to integration constants, this method cannot be used to determine absolute position, however, it
is easy to program a routine to compare step command to step result.

There are many options to verify step integrity with analog techniques. Final configurations should be
reviewed on an engineering model system to determine proper filter and gain circuitry. One generic
analog circuit will not work for all applications because changes in accelerometer voltage gains, system
resonant frequencies, and desired operational pulse rates will require tuning of the electronic circuitry to
sort out fundamental information from system resonant frequencies. These circuits can also be used in
system diagnostic testing to automatically determine if friction levels have increased above a certain
threshold. In other words, you can automatically determine if your motor torque margin had dropped
below acceptable limits.

The circuit shown in Figure 5 provides signal gain and filtering in the first stage of the Operational
Amplifier (Op-Amp #1). The Op Amp #2 is used to bring the signal polarity to its original state (Gain –1).
Each signal goes to a comparitor. Each comparitor is set to determine if the accelerometer signal has
achieved a minimum voltage threshold, indicating a step in either the CW or CCW direction. The filtering
in Op-Amp #1 is set to filter out high frequency resonance and transient acceleration spikes. This filtering
will allow you to separate out a legitimate step versus stepper motor “buzz” which may occur if the motor
could not achieve a reliable step. The positive Bias Voltage (Vb) should be set at a level that provides a
minimum reliable step, but above the maximum overshoot acceleration. This assures you are comparing
the proper step direction, and not falsely detecting overshoot as a pulse. The comparitor outputs will
provide a logic compatible output pulse for each accomplished step. This basic circuitry may alternatively
be used for a minimum torque margin detector by setting the Vb voltage at an appropriate level.

Figure 5. Sample Pulse Verification Circuit

NASA/CP—2002-211506 145
 Figure 6 - Acceleration Signal & Bias Voltage

8.00
Acceleration Signal

6.00

Accelerometer Voltage (Vdc) 4.00

Vb
2.00

0.00
0.200 0.202 0.204 0.206 0.208 0.210 0.212 0.214 0.216 0.218 0.220
-2.00

-4.00

-6.00
Time (sec)

Figure 6 shows the acceleration voltage signal (after some gain) and the Bias Voltage (Vb) to the
comparitor, for a single step of a stepper motor. While the amplified acceleration signal is greater than the
Vb voltage, the comparitor output will latch high, indicating a successful pulse. Since there is a separate
comparitor for each phase of the acceleration signal, direction of rotation information is also provided. The
pulse outputs may go to digital counters or digital logic circuits to determine actual pulse counting. Since
the direction of rotation information is available, if the motor steps opposite to the desired direction of
rotation on the first pulse, the information is maintained.

Conclusion

We have shown many practical applications for the use of rotary accelerometers to analyze permanent
magnet stepper motors. Acceleration information is very useful in determining performance and load
characteristics. We have shown that it is possible to determine the actual torque produced of a stepper
motor while dynamically operated. Additionally, with some known parameters, we can also determine
load inertia, load friction, operational torque margin, and system natural circular resonant frequencies,
simply by analyzing an acceleration signal. Step verification, step counting, and direction of rotation may
also be obtained by simple digital or analog signal processing.

References

1. CDA InterCorp: “Stepper Motor Engineering Reference Data”, 2000

2. CDA InterCorp: “Controllable Drive Actuator and Rotary Transducer Engineering Reference Data”,
2000

NASA/CP—2002-211506 146
 High Resolution Standard Proximity Sensor
*
Roger Blaser

Abstract

The High-Resolution Standard Proximity Sensor (HRSPS) is a true contact-less eddy current
displacement measuring system qualified for high-reliability satellite applications. It is also available in an
883B flight version for micro-gravity and for the International Space Station (ISS). Three different sensor
sizes covering three measuring ranges up to 9 mm (0.35 in) are available. The HRSPS offers either 2
independent single or 1 differential output(s) with a resolution of 1 nanometer. The temperature range is
standard space flight (-55°C to 70°C) but the sensors have provision for operation to -150°C (-238°F) for
missions to Mars. The reliability figures of the HRSPS are 1.7 million hours for ISS applications and 2.5
million hours for the satellite version. It is operating under +28 VDC (with fuse) 2 W or ±15 VDC 1.5 W.
The power consumption in the sensor head is below 15 mW. This paper presents the test results and the
final confirmation of the performances of the HRSPS. The HRSPS should become the standard nano-
measuring system qualified for space mechanisms operating in closed loop control. It is designed in order
to accept the worst launcher environment.

Introduction

ESA identified a world-wide requirement for a Standard High-Resolution Proximity measuring system with
the aim of replacing the multitude of customized sensors used so far in scanning and pointing space
mechanisms or robotics applications. The HRSPS can be used as a measuring system in a closed-loop
controlled mechanism providing the feedback on the position actually realized. This paper presents the
final performance results of the High Resolution Standard Proximity Sensor developed under a GSTP
competitive contract for ESA R&D center ESTEC in the Netherlands. The HRSPS is a general-purpose
displacement measuring system targeted to become the standard displacement measuring system
qualified for space. The HRSPS system, defined as Displacement Measuring System DMS-134, consists
of an electronic conditioner IQS-134 connected to 2 sensor heads TQ-47X operating in either single or
differential arrangement. The HRSPS is an inductive sensor working on the eddy current principle. It can
perform measurements in a single or differential sensor arrangement and can also be used as a proximity
switch.

During the development, emphasis was placed on a design that can easily be adapted to different
measurement tasks and ranges. It is able to meet the highest resolution requirements, to achieve a good
overall accuracy or to provide a high measurement bandwidth. The long-term intention is to offer a new
HRSPS hardware compatible with commercial satellites electronics racks.

Applications

Mechanisms operated in closed loop control generally require sensors to provide a feedback on the
position actually realized. Comparison with the desired position then allows the derivation of an immediate
correction of any mispositioning leading to the superior performance of such mechanisms. The HRSPS
offers displacement measurements in a single or differential arrangement. One of the applications of the
HRSPS will be in the measurement of rotation, although it is more suited for linear measurements. The
HRSPS is also a candidate for the active damping of the ISS Centrifuge rotor. The majority of sensor
designs available industrially and in the launcher field are not readily suitable for high reliability space
(satellite) applications. Typical ground or launcher approaches are very different in terms of material
selection, electronics qualification and reliability status. The HRSPS has been developed from the
beginning for space applications, taking into account the experience of Vibro-Meter in industrial and
space high-pressure cryogenic sensors design for launchers.

*
Vibro-Meter SA, Fribourg, Switzerland
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 147
 cable length max. 2 m
TQ 471

IQS 134

TQ 471

coaxial cable

< 5 mm

Figure 1. High Resolution Standard Proximity Sensor DMS 134

Technology

The choice of the eddy current technology was considered as the best all-around performer and most
suitable and adaptable to the installation requirements. The choice of this technology was also the result
of a questionnaire sent to the space industry worldwide.

An eddy current measuring system operates in an RLC parallel network where the amplitude of the
oscillation is proportional to the displacement to be measured. The passive sensor consists of a coil (L)
and a cable (R, C) fed by a high-frequency excitation generated by a nearby conditioner. The task of the
signal conditioner is to generate the high-frequency oscillation and to convert the resulting signal into a
linear output proportional to the relative distance of the target. The sensitivity of the system is dependent
on the material, mounting constraints, and the relative displacement of the target from the sensor.

Technical description of the HRSPS

Two sensors (type TQ 471) and two electronics conditioning channels located inside the IQS 134
conditioner are required for a differential arrangement measurement (Figure 1). The TQ 471 sensor
includes a coil (position measurement element), a temperature sensing element and a cable of up to 2
meters.

Sensor temperature monitoring permits zero and sensitivity drift compensation independently for each
measuring channel. The electronic conditioner IQS 134 also includes its own temperature monitoring,
allowing independent compensation of temperature gradients up to 220 K between the sensor heads and
the conditioner. Both single and differential outputs are permanently wired on the connector.

Transfer Functions and Measuring ranges with different sensor configurations

The initial requirement was for a maximum range of less than 25 mm (1 in). In order to achieve the low
linearity error required for nano measurement in space mechanisms, the standard measuring ranges
were set as below. The output voltage vs. displacement is shown in Figure 6.

Transfer/ differential Measuring range

V/mm Single Differential
TQ 471 10 1 mm ± 0.5 mm
TQ 472 5 2 mm ± 1 mm
TQ 473 1.66 6 mm ± 3 mm

NASA/CP—2002-211506 148
The standard measuring ranges above do not limit the capability of the eddy current measuring systems.
For instance, Vibro-Meter offers HRSPS measuring systems with +50% measuring range as the ones
indicated above with sensor head diameters of 5 mm for the TQ 471, 8 mm for the TQ 472 and 18 mm for
the TQ 473. See linearity tests results with TQ 471 with 9 mm range on Figure 6. The final HRSPS will be
a compromise between the different parameters defining the final performance of the system.

Advantages over other Technologies

- True contact less technology

- Repeatability
- Small sensor heads in an action-free environment
- Single and differential arrangement
- Sensor and target extreme temperature capabilities (+125 to –150°C)
- Sensor head and conditioner individual temperature compensation
- High vibration survivability (random 30 g rms)
-
8
High radiation survivability (Sensor 10 Rad)
- Operational in harsh environment
- Modular channel configuration (expandable up to 4 sensor heads)

Accuracy
The accuracy is defined as the combined error due to repeatability, non-linearity, hysteresis and the
resolution limits (electrical noise). The HRSPS has theoretically no repeatability error but the temperature
deviation of 100 ppm/°C will deteriorate the practical repeatability of a measurement. We therefore
assume a repeatability error of 1%. The linearity achieved is 0.5% in the differential arrangement and 1%
in the single arrangement (Figure 6). The HRSPS could be supplied with a raw output in form of a
polynome. The HRSPS system has no hysteresis and the resolution error is so low that it becomes
negligible in the calculation of the accuracy. Based on the above, the accuracy is 1.1% in differential
arrangement and 1.4% in single arrangement.

Life and Reliability

The transducers and the conditioner are designed for a total life of 15 years. The calculated reliabilities for
a flight model equipped with two sensors and one conditioner are:
Microgravity and for ISS using 883B components: MTBF > 1,700,000 hours
High Reliability Satellite: MTBF > 2,500,000 hours

Mechanical Design

The sensor casing is made out of a machinable ceramic, which offers reduced thermal expansion. The
sensors are qualified with a mechanical interface represented as a screw (TQ 471: M6 x 0.75, TQ 472:
M10 x 1 and TQ 473: M20 x 1.5) with axial cable output (Figure 2). This simple mounting has eased the
large amount of testing on the existing Vibro-Meter in-house ground support equipment. Different sensor
casings with radial outputs will also be qualified.

Electrical Characteristics

The HRSPS offers the choice of working with ± 15 VDC or + 28 VDC power supply. The HRSPS output is
0 to + 5 VDC in a single arrangement and ± 5 VDC in a differential arrangement for the full measuring
range. On request, the outputs can be increased to 7 VDC. Outputs (differential & single) are always
available on the output connector. There is room for the installation of a fuse in the +28 VDC power
supply line on request. The improved natural filtering eases a further digitalization (sampling). The design
of the anti-aliasing filter, which may be in cascade with the HRSPS in case of digital processing, will only
have to consider the mechanical target frequencies (Figure 5).

NASA/CP—2002-211506 149
 Modular Construction

The electronic conditioner IQS 134 is made of 2 frames (Figure 3):

- One Power Supply frame for either + 28 VDC or ± 15 VDC voltage
- One Conditioner frame connected to two sensor heads.

The power supply has been designed to power two conditioner frames. When more than two sensor
heads are required, it is possible to add a second conditioner frame above the first one. In this case, the 4
sensor heads have to be the same as there is only one oscillator for both frames in order to reduce the
coupling. This configuration allows to use up to 4 single sensors or 2 single and one differential or 2
differentials.

This modular concept offers the following advantages:

- Total mass reduction
- Power consumption reduction
- Cost reduction
- Mixed configuration without calibration change (single and differential output always available on
connector).

Environmental Requirements

The final qualified HRSPS responds to ESA ECCS Standards as well as to MIL-STD-1686, MIL-STD-
6001 and SSP30237/30242/50005/57000 for the ISS. It is qualified for most of the launcher environments
that are flying today.

Conclusions

The HRSPS is a true contact less eddy current measuring system qualified for space applications. It has
a high degree of adaptation to the installation and the environment requirements. It offers an excellent
repeatability and accuracy for application on space mechanisms requiring rugged conditions (30 g rms). It
offers a great flexibility allowing trade-offs with the resolution / linearity versus the measuring range. The
repeatability and the linearity errors practically depend on the temperature deviation, which remains the
limiting criteria of the technology. Practical applications have measuring ranges up to 9 mm (0.35 in) with
a slightly reduced resolution (Figure 4). The HRSPS is flexible enough to offer other compromises.

The HRSPS main performance parameters are:

- Resolution of 1 nm for a measuring range of ±0.5 mm, 1.4 nm in single arrangement for 1-mm range
(100 Hz bandwidth at 1 kHz).
- Resolution of 2 nm for a measuring range of ±1 mm, 2.8 nm in single arrangement for 2-mm range.
- Resolution of 7 nm for a measuring range of ±3 mm, 10 nm in single arrangement for 6-mm range,
- Resolution of 16.5 nanometer for a 9-mm range in single arrangement (Figure 4).
- Frequency response 16.5 kHz (-3 dB) at max. gap (Figure 5).
- Sensor operation to –150°C (-238°F) allows the integration of the HRSPS for missions to Mars.

The HRSPS main achievements are:

- Linearity of 0.5% in differential arrangement and 1% in single arrangement (Figure 6).
- Accuracy better than 1.1% in differential arrangement and 1.4% in single arrangement.
- Flight Model (DMS 134) using the High Reliability version for scientific / commercial satellites.
- Flight Model (DMS 124) in 883B quality for microgravity and for the International Space Station.
- Modular design with a mix of up to 4 sensor heads and one IQS 134.
- Single mode operation allows cost reduction and eases the configuration (mix of single and
differential).
- The HRSPS is the only true contactless displacement measuring system operating within a
temperature range from -150°C to +125°C (-238 to 257°F).

NASA/CP—2002-211506 150
 TQ 471

TQ 472

TQ 473

Figure 2. Typical HRSPS

NASA/CP—2002-211506 151
 Figure 3. HRSPS Electronics (IQS 134)

NASA/CP—2002-211506 152
 TQ 473, IQS 134 in single arrangement at 9 mm. gap
Equivalent RMS I/P noise (FSD/vHz)

9.00E-07
8.00E-07
7.00E-07
6.00E-07
5.00E-07
4.00E-07
3.00E-07
2.00E-07
1.00E-07
0.00E+00
1 10 100 1000 10000
Fréquence/HZ

Figure 4

Frequency Response TQ 473, IQS 134,

single arrangement at max. gap 9 mm.

100 1000 10000 100000

0
-1
-2 -3dB at 16.5 KHz
-3
Uout (dB)

-4
-5
-6
-7
-8
-9
-10

Frequency (Hz)

Figure 5

NASA/CP—2002-211506 153
 Uout=f(gap) TQ 471, IQS 134 in single arrangement

5. 0

4. 5

4. 0

3. 5

3. 0
Uout (V)

2. 5

2. 0

1.5

1.0

0. 5

0. 0
0. 0 0 .2 0. 4 0 .6 0. 8 1.0

Distance Sensor - Target VCL 140 (gap/m m )

Linearity TQ 471, IQS 134 in single arrangement

1.0%

0.8%

0.6%

0.4%
Linearity Error (%)

0.2%

0.0%

-0.2%

-0.4%

-0.6%

-0.8%

-1.0%
0.0 0.2 0.4 0.6 0.8 1.0
Distance Sensor - Target VCL 140 (gap/m m )

Figure 6

NASA/CP—2002-211506 154
 Qualification of the Inflatable Sunshield In Space (ISIS) Mast
*
David J. Rohweller

Abstract

The Inflatable Sunshield In Space (ISIS) mast, Figure 1, is a new telescopic mast that was successfully
qualified for flight on the Space Shuttle. This paper shares some of the successes and failures we
experienced and lessons learned as we went through this process. These included successful
accommodation of customer driven changes, the effect of thermal analysis on design and cost risk, better
ways to design for manufacture and assembly, solving test failures, and the effect of size on overall cost.

Introduction

TRW Astro Aerospace built the ISIS mast for the NASA Goddard Space Flight Center (GSFC). It is
designed to carry a 136-kg (300-lb) payload for launch on a Shuttle Hitchhiker pallet and then deploy it
6.7 m (22 ft) away from the payload bay. The intent was to prevent the experiment, an inflatable
sunshield, from contacting any part of the Orbiter or any payloads in the bay as the experiment deployed.
After completion of the deployment and related experiments, the mast and shield are ejected from the
Shuttle bay.

The stowed ISIS mast is shown in Figure 2 with the various items of equipment mounted on it identified.
The mast outer canister mounts on a Space Shuttle Hitchhiker pallet with a Marman clamp band provided
by the customer. This canister is the primary structure that carries the payload during launch.

The payload is attached to the top of the outer canister via a tip plate and another clamp band that
secures the tip plate to the upper end of the canister. The tip plate also preloads and cages the inner
segments of the mast for launch. To deploy the mast, the upper clamp band (not shown) is opened via a
pyrotechnic cutter, which releases the payload and the inner canister segments for deployment.

After release, the motor is turned on and a Storable Tubular Extendible Mast (STEM) pushes the mast
out. In effect, the mast is pulled out from the tip by pushing from the bottom. The deployment is
sequenced so that the mast deploys from the root with each canister locking into position before the mast
continues its deployment. This interlock system is shown in Figure 3.

The two, 1.9-cm (3/4-inch) diameter electrical payload cables are mounted in a cage inside the innermost
canister as can be seen in Figure 4, which shows the top and inside of the mast. As the mast deploys the
cage uncovers the cable and it feeds itself out and tightens itself around the STEM in a spiral. This mast
is similar to a breadboard design presented at the 28th Aerospace Mechanisms Symposium, except it is
five times larger in diameter and represents the first flight qualification of this mast. Table 1 gives some of
the data for this mast as a reference.

The mast successfully completed all the required objectives and was shipped to GSFC. Unfortunately, the
ISIS program was cancelled and the mast is currently residing in government storage.

*
TRW Astro Aerospace, Carpinteria, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 155
 Table 1. ISIS Mast Data

Description Value

Diameter 61 cm (24 in)

Average Diameter 51.44 cm (20.25 in)
Stowed Length 67.95 cm (26.75 in)
Deployed Length 6.7 m (22 ft)
Number of Segments 12
Time to Deploy ≅ 3 min
6 2 6 2)
Bending Stiffness 1.93 x 10 Nm (4.2 x 10 lb-ft
6 6
Axial Stiffness 50.7 x 10 kN (11.4 x 10 lbf)
Bending Strength 1,853 Nm (1,365 lbf-ft)
Torsional Strength 2,813 Nm (2,080 lbf-ft)
Deployed Natural Frequency with 136 kg Payload 1.7 Hz

Project Events

Design Changes
Design changes after the start of a program are common and this program was no different. Changes to
the ISIS experiment required the mast to increase in size. The design itself is modular and is capable of
accommodating changes such as this easily. However, some of the minor details presented challenges
that were unforeseen.

The biggest change was that the payload increased from 91 to 136 kg (200 to 300 lb). This forced the
outermost canister, which carries this load as a cantilever, to increase from 45.7 to 61.0 cm (18 to 24 in)
in diameter to support the load with the required stowed natural frequency of 50 Hz. In addition to this, the
length of the mast increased from 6.1 to 6.7 m (20 to 22 ft). The practicalities of routing a large cable into
the mast caused the stowed length to increase, which increased the stowed bending moment due to
launch loads. Overall, the stowed stiffness increased by a factor of almost four over the originally
proposed design. These rather significant changes were accommodated with minimal impact to the
program due to the scaleable nature of the mast. The challenges we experienced were caused by tight
tolerances and additional hardware interface requirements as a result of minor design changes.

To accommodate the rather large loads at the interface, a Marman clamp band profile was specified that
required a tolerance of +0.0 and -0.1 degree. This tolerance required the angle of the profile to be held
within ±0.3 µm (±0.00014 inch) over its length. This proved difficult to machine and inspect.

Another change was the addition of handholds and fittings to the structure. These were customer-
furnished equipment (CFE) necessary for safety and mission success. Engineering added these to the
canister in collaboration with the customer fairly quickly . However, the additional hardware significantly
increased the number of customer interfaces that quality assurance had to verify and increased the
chances for assembly problems. When it came time to integrate the CFE we found a subtle difference in
how these parts were made. The CFE had threaded holes that mated with holes on our canister. Their
holes were drilled to make their parts easy to make, that is, in vertical directions rather than radially. Our
canister on the other hand, while dimensioned the same way, was made in the easiest way to make it,
with the holes and countersinks drilled radially. When the time came to mount the hardware, the holes

NASA/CP—2002-211506 156
matched, but the countersunk screws did not seat properly. This can be seen in Figure 5. This required
rework to ensure the heads were below flush on the inside of the canister.

To summarize this, accommodating the larger payload and longer length, which would seem to be major
impacts had minimal impact. Meanwhile the small details created some unforeseen time and effort.
Changes are an inevitable part of design and should be expected. The question is: How could we have
prevented the problems that came about as a result?

On the tight tolerance issue, it is axiomatic that tight tolerances cost money and time. The Marman clamp
band angle profile tolerance was stated; however, we (and our machine shop vendor) didn’t fully
understand the impact of this tolerance. The lesson from this is, be careful with tight angular tolerances.

When estimating the impact of design changes, it helps to consider the impact of verifying additional
interface dimensions and to minimize them as much as possible.

When working customer interfaces on circular mountings, a decision has to be made at the manufacturing
level on how the parts can best be made. Because of the process, it was easier to put the holes in the
canister in the flat pattern and then roll the canister to size. This makes the holes radial. The brackets
were more easily made with holes parallel to the radial direction. In this case, the easiest way to make the
design work is to put radial holes in the brackets rather than in the canister, but this cost the customer
providing the parts more. If overall cost is the final arbiter, the brackets should be made with the holes in
the radial direction. Having manufacturing look at the changes prior to their acceptance might have
spotted the best way to go.

Material Problems
During the program, we used our heritage as much as possible. The original breadboard design used a
Teflon-impregnated nickel coating on all rubbing surfaces on the tubes. We decided early in the program
to stick with this proven configuration to reduce risk. However, this resulted in a thermal problem that
became a material problem.

Our thermal analysis showed the component temperatures inside the canister would exceed +60°C due
to the absorbtance/emittance (α /ε = 0.82/0.85) properties of the nickel coating on the inside and outside
of the tubes. This would cause the motor to exceed its 60°C maximum operating temperature range. To
reduce temperatures, we anodized the outside of the tubes and changed the absorbtance/emittance
properties to α /ε = 0.40/0.82. This reduced the motor temperature to 55°C. In keeping with our heritage,
we kept the inside surface of nickel, which created a processing problem. We asked our vendors if they
had done any parts that were both anodized and nickel-plated and they had not. We also asked what
order of processing they would prefer. Each vendor wanted their coating put on first, which was no help.

Our materials research indicated that the anodize should go on first because the anodize would not be
damaged by the chemicals used to prepare for the nickel plating. Additionally, the masking for the
anodize was more likely to withstand the plating chemicals, while the masking for the nickel would not
survive in the acid bath for the anodize.

We proceeded with this order of operation and it worked fine for all of the parts except one. The canister
shell, which was 7075 aluminum alloy, came in with small pits in the structure that caused us to reject it.

Review of the scrapped canister material trace showed that it was marginal in its response to heat
treating as measured by conductivity. The specification allows reheat-treating to meet specification and
this was done. After heat treating, machining, anodizing, and plating, the pits were found but only where
the original mill markings were located. We suspected that the material itself had a problem, so we sent
out coupons of new 7075-canister material and had one set nickel-plated on one side first, then anodized
second. The other set had the anodize first. Results for the samples showed no pitting, and that the best
process order was anodize first, nickel plate next. Our conclusion was that the new canister material
would be acceptable based on these tests so we proceeded with plating and the new canister turned out

NASA/CP—2002-211506 157
correctly. We suspected the cause of the pitting was a problem with the material because of the poor
response to heat treatment, but other factors could have been responsible such as the environment the
canister went through during processing. Since we had a good canister, we stopped looking for the
cause.

The lessons we learned from these events were as follows:

Carefully review the certifications on 7075 T7351 after heat treatment and make sure the conductivity test
results are correct.

In some cases it makes sense to make coupons to verify the process prior to treating the flight
components. This is in addition to in-process coupons that validate how well the process was done.
Coupons should be a prolongation of the same material after heat treating whenever possible.

If possible, it is better to have only one type of plating operation, such as all anodize or all nickel plating.
We could have ran some tests early on to try to eliminate the nickel plating and just used Teflon-
impregnated anodize. There was a strong feeling at the start of the program to stick with our heritage
design, but if we had taken the time to prove the alternate system was satisfactory, it would have saved
overall cost and schedule.

Designing for Assembly

The next problem we ran into involved some design decisions we had made early on. While designing the
unit we determined that the length of the STEM that deploys the mast needed to be adjusted to match
the length of the canister within a tight tolerance range. The hardware arrangement we used to perform
this is the shim system shown in Figure 6. This is a simple system. However, when it came time to adjust
the length, there were some difficulties. First, it was not practical to perform the shim adjustment on the
bench by using dimensions because the tolerance stack-up dimensions were too large. Next, it was
difficult to perform the adjustment horizontally because the STEM deployer would bow with a horizontal
deployment and throw the measurement off. The method we found ourselves using was to deploy the
mast vertically and adjust the STEM length from the top of a scaffold. The mast was set up vertically in
a high bay and we deployed the mast and marked it. Then we cut it to length, adjusted it, then tested and
verified that it worked. We did all this while working 6.7 m in the air, reaching down inside the canister and
bolting and unbolting loose parts. In one sense this worked fine; everything came out acceptably after
spending a day to set up, and a day to adjust it and take it down.

The design would have been better if we had made the top adjustment infinite with just a screw bolt
arrangement so we could adjust it by merely loosening or tightening a bolt. More broadly, it makes sense
to design so that adjustments are easily made, or even better, not made at all. Shims are inexpensive on
the design side, but costly on the assembly side of the process. Shimming is labor intensive and should
be eliminated if possible.

Another issue that came up was the system for deploying and restowing the cable. The cable consisted of
two 1.9-cm (3/4-inch) diameter cables that were stowed for launch inside the canister as shown in Figure
4. This provided a clean package, but testing was delayed by the time required to open up the top of the
canister and restow the cable after each deployment test. A preferred way of doing this would have been
to design the cable to be self retracting, which we had no budget for, or design the system for quick
removal of the tip components to restow the cable. Using engaging features in the parts and one screw
plate to hold everything together could have easily done this. Unfortunately, our original concept of
restowing from the base of the canister did not work and we had to use the alternative method of
removing all the screws each time we needed to retract the mast. This procedure also was performed 6.7
m in the air as shown in Figure 7.

NASA/CP—2002-211506 158
Test Setup Problems and Solutions
Another issue that came up involved the combination launch restraint/payload tip plate, which was to be
customer furnished for the flight hardware. For testing however, we provided a launch restraint/payload
tip plate.

The mast is restrained for launch by the CFE-tip plate that is itself held on by a Marman clamp band at
the top of the outer canister. The underside of the tip plate has provision for rubber preload cushions that
preload the mast for launch. To size these, we made sample rubber shims that we tested in compression
with a bar that had cuts in it to match the clearances between the canisters. We performed these tests on
a SATEC testing machine, which gave the plots shown in Figure 8. We varied the width and thickness of
the samples to obtain the preload characteristics we wanted so the canisters would not rattle during
launch, yet have reserve preload if the rubber developed compression set. These performed properly
during vibration testing with the mast coming through with no evidence of vibration wear.

Having never used rubber cushions before as preload devices, it was thought they would behave in a
non-linear fashion. It was a surprise to discover the fairly linear response to load that the samples gave
(Figure 8).

Lessons From Testing

The test program ran very well for a new design. There were two problems, both of which occurred during
cold testing. The first was the failure of the first segment to latch up properly shortly after deployment.
This was complicated by the thermal enclosure that we used to perform the test, which prevented us from
seeing the mast as it deployed, Figure 9. The enclosure was a large diameter accordion tube surrounded
by insulation that we pulled up into position around the mast. To achieve cold temperatures we used
liquid Nitrogen (LN) controlled by thermocouples to bring the temperature down. After soaking for the
required time, we began the deployment and it stalled at the point where the first segment should have
latched into position.

At this point we shut off the LN and monitored the mast as the temperature rose to study the problem.
Suddenly, we heard the mast click into position and saw the offloading system move. This prompted us to
see if the mast would deploy further, so we turned on the motor and the mast ran out all the way to the
end, with each segment operating properly. The motor has a shutoff switch that is supposed to stop the
deployment at the end. However, this did not operate, so we stopped the motor manually. Again we
waited and monitored the switch continuity as the temperature in the chamber rose. Finally, at ambient
the switch triggered, which indicated a possible thermal problem with the switch.

To solve the switch problem we pulled the deployer out of the mast and performed a cold test on the
STEM unit alone in a chamber where we could see it. At a cold temperature, the switch lever was
clamped hard on the shaft that supported it, and this prevented it from rotating. We modified the switch by
increasing the clearance between the lever and the shaft and performed the test again and this time it
operated properly. We replaced the unit in the mast and set up again for the cold test.

This time we let the mast dwell for a longer time period at the cold temperature and monitored the test
carefully. Again, the mast deployed and stopped at the first segment latch point. We let the temperature
rise and noted that the unit latched up after a slight temperature rise. We continued deployment, still
within thermal specification requirements. This time, the switch operated properly so we had a partial
success.

We next retracted the mast, set it up again, and performed the hot deployment, just to see if other
problems were hiding in our system on the hot side. Thankfully, it operated perfectly.

We were still concerned about the cold deployment, so we moved the unit down to a thermal chamber
that could only do a short deployment, but that would allow us to watch it and repeat the test quickly. We
set up again and let the unit dwell longer and again it performed the same, meaning we did not have the
solution.

NASA/CP—2002-211506 159
We analyzed the actual temperature the unit would see in space after the predicted dwell time and
changed the test configuration to achieve this more closely. Our test plate was providing a massive heat
sink at the base and was keeping the base of the mast from going cold. To fix this, we mounted the
canister with an air gap between the test fixture and the baseplate to allow the bottom end of the canister
to get colder more quickly and decouple the baseplate from the fixture. We also instrumented the
baseplate to monitor the thermal gradient between the outer canister and the inner canister elements at
the base. To our surprise, the temperature differential between the inner and outer canisters was 40°C at
the point we had started the test previously. We increased the dwell time at cold and monitored the
temperature of the baseplate. When it reached equilibrium, the temperature differential was 16°C. At this
point we ran the test and the mast worked properly. What had happened in the test failure was the outer
canister at the top was cold and the bottom of the next inboard canister was warmer and the difference
was enough to affect operation.

The lessons we drew from this were as follows:

Our practice of designing systems so that whenever possible any subsystem can be removed without
teardown of the whole system proved its worth again. We were able to remove the STEM deployer
quickly, test it, fix it, and reinstall it without a major impact to the program.

We could have tested the STEM subassembly in a cold chamber prior to the full test. This would have
highlighted the problem with the switch before the start of flight testing on the integrated canister.

It is important to match flight conditions as closely as possible during testing. Having a large heat sink and
not dwelling at temperature for a long enough time cost us several days of schedule.

Even though we had a deployment anomaly, we went ahead with what we could do and discovered
another problem with the switch. If we had stopped everything after the first problem, it would have added
more time and testing to the schedule. Sometimes it is worthwhile to continue as far as possible to
discover as many problems as possible in a given test setup, as long as you will not damage the
hardware.

A Heavy Lesson
We discovered something else that is obvious in retrospect. Most of our previous designs were small
enough to carry by hand. This mast, fully assembled with all CFE in place, weighed 54 kg (120 lb), which
made handling more complicated. The weight plus the size of the stowed mast required the use of forklifts
and cranes to move it about, along with several employees to position and handle the unit. Tooling for
manufacturing the parts was larger, cradles and Ground Support Equipment (GSE) were larger, and
existing GSE that worked fine on lightweight masts, could not be used. This resulted in an increase in
staffing for the program, which increased cost and schedule.

The lesson here is this: If a product varies considerably in size and weight from other projects of a similar
nature, then some account should be taken of the impact this will have on the project. This applies to
large increases or reductions in size. As mentioned before, this design was nearly five times larger in
diameter than the prototype mast we based the design on.

Conclusions

The ISIS mast qualification program is typical of any new hardware. The customer requested changes
soon after the start. Our own internal design and analysis team required more changes. Some items we
designed worked better than we expected, and some we just had to endure. Manufacturing had problems
with making some parts. Testing did not run as smoothly as we had planned. In the end however, the
design successfully met the requirements of the specification.

NASA/CP—2002-211506 160
 Figure 1. ISIS Telescopic Mast, Fully Deployed

Figure 2. Stowed ISIS Mast

NASA/CP—2002-211506 161
 Figure 3. Canister Sequencing Interlock

NASA/CP—2002-211506 162
 Figure 4. Inside View of Mast

NASA/CP—2002-211506 163
 Figure 5. Countersunk Screw Arrangement

NASA/CP—2002-211506 164
 Figure 6. Tip Plate Shim System

NASA/CP—2002-211506 165
 Figure 7. Operating Conditions for Restowing the Cable

NASA/CP—2002-211506 166
 Figure 8. Silicone Rubber Spring Rate

NASA/CP—2002-211506 167
 Figure 9. Thermal Test Chamber for ISIS Mast

NASA/CP—2002-211506 168
 ISS S/A Deployment – The Highs & Lows of EVA Contingency Capability,
A Designer’s Perspective

*
Bert Haugen , Malcolm Ferry* and Kevin Klein*

Abstract

Each solar array wing for the International Space Station (ISS) is roughly 38 feet wide by 115 feet long
when fully deployed and is the largest deployable (and retractable) space structure built to date. The
deployment and repair of the first two wings on STS-97 in December 2000 highlighted both the
immensely beneficial capabilities provided by the manned presence during a complicated deployment as
well as some of the difficulties in using this capability to its fullest potential. This paper provides a
designer’s perspective of these capabilities in the context of the ISS solar array operations on STS-97 as
well as some of the difficulties in exercising these capabilities to their fullest efficiency during such a fast
paced investigation. Suggestions will be provided to better equip mechanism personnel involved in future
investigations to deal with the programmatic issues as quickly and effectively as the technical issues.

Introduction

The deployment of the ISS solar array includes both planned Extra Vehicular Activity (EVA) operations
and autonomous operations to properly deploy the 164 flexible solar array panels and their associated
structure. The final operations include the autonomous motor driven extension of the deployment mast
and the application of full blanket tension (a.k.a. deployed preload). During the deployment of the first
wing during STS-97, unexpected adhesion between many solar panels caused a series of sticking then
breaking loose of panels and groups of panels. The resulting dynamic motions induced in the panels and
associated tensioning mechanisms were dramatic and ultimately caused two blanket tensioning cables to
jump their pulleys preventing proper application of the deployed preload. Although this anomaly was not
readily detectable by deployment telemetry or to most observers of the deployment video, the
ramifications for achieving the proper deployed wing frequency and strength were very significant.
Fortunately, the resolution of the Orbiter’s “routinely” available video coverage was adequate to allow
easy detection of the anomaly when viewed by individuals familiar with the design.

With the identification of the anomaly the solar array deployment support switched into the
troubleshooting mode, and the benefits and the difficulties of the manned operations capability quickly
began to materialize. The capabilities of the manned operations allowed for a much swifter and more
conclusive determination of the nature of the anomaly than normally experienced by spacecraft designers
accustomed to the limitations of autonomous operations, and the flexibility in the available repair options
allowed a definitive and complete restoration of the wing to its nominal configuration. The capabilities that
proved most useful in the evaluation and correction of the solar array deployment anomaly included live
video downlink capability over portions of the orbit, delayed video downlink coverage of the full
deployment event, variable resolution still photos of requested items, incredibly detailed inspection
capability from a camera on the astronaut’s helmet, a variety of tools available on orbit, the ability of
experienced astronauts to make safety judgments outside the design guidelines, and the availability of
experienced astronauts and EVA support members to be quickly transported to Lockheed Martin to work
as a team with design personnel to determine appropriate repair procedures.

In spite of this impressive array of capabilities, there were a variety of impediments in using these
capabilities to their utmost effectiveness and efficiency when viewed from the perspective of a member of
the design troubleshooting team. Although some of these impediments were technical in nature, more
often they were self-inflicted by communication and/or organizational barriers. The difficulties that were

*
Lockheed Martin Space Systems Company, Sunnyvale, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 169
technical in nature included the limited time available for live video downlink and the obvious limitations in
view angles and zoom of the fixed video and still cameras, neither of which presented significant
obstacles.

The most notable impediments encountered in non-technical areas were (1) the lack of a clear
definition/understanding of the organizational structure and decision making hierarchy of the very broad
and dispersed set of personnel and organizations involved in the investigations and operations, and (2) a
limited and poorly defined downward communication path to the design team and an even more limited
and poorly defined upward communication path to the operations team during critical events. Certainly,
the frantic pace of the activities, the complex nature of the mission and the familiarity of the processes to
those involved regularly in them all contributed to the lack of clarity in these areas for the design team. To
prevent such organizational and communication issues from impeding the progress of an investigation,
the designer should be prepared to be forthright in addressing such issues early and directly when
involved in investigations of this nature.

Although some technical and non-technical limitations in the manned operations capability were
encountered, the benefits of the human presence in the troubleshooting and repair operations allowed for
an extremely swift and fully successful restoration of the ISS solar array to its design capability without
impact to subsequent missions. The rapid collocation of skilled and experience EVA operations personnel
with the hardware designers from multiple companies was instrumental in both the swiftness and
completeness of this success. In addition to familiarizing the reader with many of the impressive
capabilities available for troubleshooting hardware, some of the organizational and communicational
impediments encountered in the troubleshooting will be described in this paper to better equip
mechanism personnel involved in future investigations to deal with the programmatic issues as quickly
and effectively as the technical issues.

Solar Array Design Background – “How it was supposed to work…”

Design Description
The deployment of the ISS solar array includes both planned EVA operations for release and positioning
of the solar array wings and autonomous operations for the release, deployment and tensioning of the
solar array blanket and supporting structure. The final portions of deployment are autonomous and
include the motor driven unlatching of the stowed flexible solar array panels, the motor driven deployment
of the extendable mast, and finally, the motor driven application of the “high” tension (deployed preload)
to the deployed S/A panels. In addition to the active motor driven mechanisms there are several passive
spring-activated mechanisms utilized to provide proper management of the 164 flexible solar array panels
and associated structure during deployment. The specific sequence of deployment is illustrated in Figures
1 and 2 and described in the following bullets:
• Multiple payload module operations are performed to position the solar array wings to clear the
Integrated Equipment Assembly and associated hardware (not shown in figure).
• EVA operations are performed to manually release then rotate the right and left solar array
containment boxes 90 degrees then manually lock them in place (Figure 1).
• Three launch lock retention bolts are released by EVA operations to free the Mast Canister’s Tip
Fitting from the base of the Mast Canister to enable the subsequent mast extension (Figure 2).
• The Astronauts are then cleared from the area and approximately six thousand pounds of preload
required for the launch restraint of the stowed solar array panels is relieved from the panels by the
motor driven Latch/Unlatch Mechanism over approximately fifteen seconds. Linked to this unlatching
motion are the release of seven spring loaded Blanket Restraint Pins and a positioning of the Blanket
Tension Mechanism into a “low” tension mode. This is done sequentially on one Containment Box at a
time due to available power limitations. Ground available telemetry for these events includes motor
current time history, latched & unlatched limit switch signals on the latch mechanisms and retracted
limit switch signals for the seven Blanket Restraint Pins (wired in series due to channel limitations),
although only at a 0.1 Hz sample rate.

NASA/CP—2002-211506 170
• Once proper unlatching is verified by the internal telemetry and visual “inspection” from both EVA and
the available video cameras on the Orbiter, the extension of the deployable mast is performed by the
motor driven Mast Canister. As the mast deploys, the two solar array blankets, each consisting of 82
solar array panels, are unfolded from their stowed position. The mast extension nominally requires
approximately 12 minutes. Three lightly tensioned guidewires per blanket provide management of the
unfolding panels by guiding every other panel hingeline while the outer-most and innermost panels are
attached to their respective Containment Box structures. Spring mechanisms evenly distribute the
tension forces on the panels. Internal telemetry of this operation includes only motor current time
history, fully retracted & fully extended mast limit switch signals and incidental blanket shunt current
readings.
• Once the Mast is fully extended, the final deployment operation requires commanding the
Latch/Unlatch Mechanism back to its “latched” position in order to position the Tension Mechanisms
into their “high” tension mode to fully preload the deployed solar array blankets. This operation is done
one blanket at a time due to available power limitations and is required to provide proper frequency
and strength to the fully deployed solar array blankets. Internal telemetry consists of motor current time
history and latch/unlatch limit switch signals.

Figure 3 shows a fully deployed solar array wing.

Top Rotation direction of left containment box

View to on-orbit ready-to-deploy position

Rotation direction of right containment box

to on-orbit ready-to-deploy position

Rear Right Side

View View

Bottom
View

Figure 1. ISS Solar Array – Stowed Configuration

NASA/CP—2002-211506 171
NASA/CP—2002-211506 172
cycle Tension Mechanism stroke life and concerns for adding binding failure modes) included cable
keepers on the simple pulleys but no additional restraints on the Tension Mechanisms’ spiral pulley
(Figure 4).

460 in.
max

Solar Array Blanket

(82 panels each)

Mast

1382 in.
max

Tension Distribution Bar

Mast Canister

Figure 3. Fully Deployed Solar Array Wing

EVA Design Requirements

In addition to the typical spacecraft design requirements, the solar array program included a variety of
EVA related design requirements. These requirements included various tool and astronaut capabilities for
planned EVA operations, requirements for manual back-up modes for the autonomous operations, and a
variety of safety related requirements. Although EVA design requirements are interesting and complicated
enough for several papers of their own, only the safety related requirements are relevant to this paper. Of
particular interest during the design were safety concerns regarding pinch hazards, sharp edges, relative
motions, etc. that had as one general design solution the inclusion of some form of relatively smooth
cover over the areas of concern. Given the complicated nature of the mechanism functions and the limits
on envelope, weight and assembly access, the design of the mechanisms could not be hazard free and
covers at many locations were discussed at length during the design process. The compromise that
resulted from the design trades was the inclusion of relatively few covers near the planned EVA operation
areas and known back-up mode EVA contingency areas and the designation of the rest of the wing as
“EVA Exclusion Zones” (Figures 4 and 5). This was done with naive confidence that no EVA would be
required near most of the Containment Box mechanisms, and it had the unintended benefits that it

NASA/CP—2002-211506 173
allowed very detailed on-orbit mechanisms inspections and unplanned contingency repair options that
otherwise would have been impossible if covers had been incorporated.

Tension
Cable

Pulley with
partial guide

Spiral Reel

Guidewire
Mechanism
Cover
Figure 4. Tension Mechanisms & Cable Management

Pulley
Cover

Tension
Mechanism

I/B Tension Mechanisms (with cover) O/B Tension Mechanisms (no cover)

Figure 5. Tension Mechanisms & Cable Management – Close-up

NASA/CP—2002-211506 174
 STS-97 Solar Array Deployments – “How it actually worked…”

Deployment Anomaly – First Wing Deployment

During the deployment of the first wing on STS-97, unexpected adhesion between the solar array panels
caused some panels to stick to each other during deployment. When the panels eventually broke free
under the force of the Tension Mechanisms, the resulting dynamic motions induced in the panels and
deployment mechanisms were dramatic, causing substantial shaking of the deploying blankets and
associated structure. This sticking then breaking free occurred multiple times on each blanket, and in
some instances the adhesion forces and low mechanical advantage of the folded panels was such that
the blanket tension distribution bar (which is spring loaded by the Tension Mechanisms through cables
and pulleys as shown in Figures 3 and 6) was pulled off the Containment Box Base for many inches
before the panels would break free. Once the panels freed themselves, the bar would be retracted, or
“reeled in,” by the forces of the Tension Mechanisms until it impacted the Containment Box Base.
Eventually, one of these retractions and impacts caused one of the blanket tensioning cables to jump its
spiral pulley and allow a slack cable. A subsequent sticking and retraction event caused the second
tensioning cable on that bar to similarly de-spool. Although this anomaly was not readily detectable by
internal deployment telemetry or to most observers of the deployment video, its ramification for proper
wing performance was very significant. As discussed in previous paragraphs, the blankets must be fully
tensioned to provide proper solar array wing frequency and strength, and with slacked tension
mechanisms the proper tension could not be applied to one of the blankets.

Tension
Distribution
Springs

Tension Bar

Tension Cable
(1 of 2 per blanket)
Figure 6. Tension Distribution Bar

Fortunately, the resolution of the “routinely” available video coverage was adequate to allow easy
detection of the anomaly when viewed by individuals familiar with the design. The swiftness and
conclusiveness in confirming the presence of the anomaly was just the first of many advantages provided
by the manned deployment operations during the course of the troubleshooting and repair.

NASA/CP—2002-211506 175
Troubleshooting & Repair – The Advantages of Manned Operations
With the observation of the anomalous dynamic motions during deployment, the ground support teams for
the solar array deployment switched into the troubleshooting mode, and the benefits and the difficulties of
the manned operations capability very quickly began to materialize. The capabilities that were found to be
the most helpful, and those found to be the most cumbersome, in this investigation and repair are
discussed in the following paragraphs in the hope that designers accustomed to the limitations of
autonomous operations present in most spacecraft programs might benefit from this experience should
they too become involved in an investigation similar to the STS-97 solar array anomaly.

One limitation of the video coverage of the space station deployment operations was that it was available
over only a portion of the deployment. As luck would have it, the first solar array was deployed during the
video blackout portion of the orbit where only audio down link was available real time. The audio
commentary describing the unusual dynamic motions of the deploying array was sufficient to put the
various facilities standing by for support into a troubleshooting mode even though much of the
commentary after the mast extension portion of the deployment was indicating a successful deployment
of the fully extended solar array blankets. Confirmation of an anomaly was swift and conclusive shortly
after real time video was reestablished and available to the supporting facilities. The various views and
levels of zoom and resolution available from the Orbiter video cameras were used to inspect the solar
array and could clearly distinguish the anomalous cables. If this capability were not available (as is the
case in autonomous operations), it would have taken many days of analysis and deduction from the
internal solar array telemetry to reach even a moderate suspicion that a cable had jumped a pulley. In
contrast, the Orbiter’s routinely available video allowed a conclusive determination of a failed condition
(slack cables) within a few minutes of the event and the determination of the cause (panel sticking) within
a few hours. In addition to video, still photography with even better resolution could be obtained of
specific items (Figure 7). However, the real time capability and the ability for general inspection provided
by the video proved to be more useful than still photography in many instances.

Figure courtesy of NASA JSC

Figure 7. Medium Resolution Still Photography of the Deployed Wing*

(Note that actual photographs clearly resolve down to individual cables and springs)

In addition to real time video which was linked to each supporting facility, delayed downlink video of the
full deployment was made available. Once this video was provided, the design team could clearly identify
the unexpected sticking of the solar array panels and the resulting motion of the blanket tensioning bar
and cables. Multiple views and zoom levels were available from the various cameras and

NASA/CP—2002-211506 176
review/reduction of this data allowed the estimation of the number of sticking events, the locations where
these occurred, the stroke of the tension bar during these events, the presence of the tension bar
retractions and impacts, the exact point at which each of the two tensioning cables became slack, and the
final configuration of the blankets at the completion of the mast extension. All these parameters relating to
the anomaly were determined within hours of the anomaly and with complete confidence and specificity
such that the critically important effort to determine repair procedures to correct the anomaly could begin
almost immediately.

Considerations for the deployment of the second solar array wing were being worked in parallel to the
investigation of the first wing. A modification of the nominal deployment procedure was selected that
included extending the mast with the Tension Mechanisms in “high” tension mode and providing regular
stops to help allow settling of the panels. The “high” tension mode was to aid in restraining possible
tension bar motions during panel sticking while the intent of the frequent stops was that the panel
adhesion might be reduced with the additional time and panel heating that result. The Tension
Mechanisms would be reset to the low tension mode for the final several inches of mast extension to
minimize potential torque margin concerns in this region. The key item that allowed this modified
procedure was the evolution of the understanding of the mast capability and the deployment loads in the
years since the two-level Tension Mechanism was designed. Analyses and tests since then which were
re-visited as a result of the anomaly (and the proposed solution) showed that sufficient mast strength did
exist for high tension deployment loading even in a partially extended condition.

Three critical items needing further investigation to support the troubleshooting and repair procedure
activities were the determination of (1) whether binding of cables in the tensioning mechanisms was
present, (2) whether cables internal to the Containment Box had become loose, and (3) if the retractions
and impacts of the tension bar had damaged small spring loaded mechanisms necessary for proper
retraction and stowage of the wings in the future. A list of these and other specific areas to inspect and
evaluate during an upcoming EVA was created. The ensuing inspection not only conclusively answered
the critical questions but also showcased two incredibly powerful capabilities of the EVA operations.
These capabilities included the ability to manually manipulate the mechanisms (which allowed the
confirmation of the condition of the internal mechanisms) and the use of a camera installed on the helmet
of the astronaut downlinking in real time and in extreme detail the condition of the various external
mechanisms. Figure 8 provides an example of the detail obtained by the EVA inspections using the
helmet cam and local photography. Although one of the two helmetcams had failed early in STS-97, the
versatility and detail of this capability can not be overstated.

With the condition of the failed hardware understood to a relatively large degree, the creation of repair
procedures proceeded rapidly. Two of the key items enabling the quick progress were the ready
availability of qualification and future flight units of the failed mechanisms for “experimentation” and the
availability of experienced design engineers familiar with the mechanisms. As essential as these two
items were, they are not unique to the manned operations of interest to this paper. They also would have
been dramatically less effective without the third key that enabled the rapid and thorough creation of
repair procedures. The third key item was the rapid “deployment” by NASA of an experienced
multidisciplinary team to collocate with the design team to work the repair procedures. Team members
arrived at Lockheed Martin from multiple organizations and destinations within twenty four hours of the
anomaly and included:
• Astronauts with experience spanning multiple EVA’s and four months on the MIR Space Station,
• Personnel experienced in EVA capabilities, tools and procedure development,
• Safety & Hazards representatives, and
• P6 Truss mechanical systems personnel.

NASA/CP—2002-211506 177
 Figure 8. Local View on Approach to Tension Mechanism Inspection*
(Note that actual photographs clearly resolve down to rivet heads and individual wire strands)

Arriving with these personnel were a variety of tools similar to those available on the ISS including
appropriate gloves for use by the astronaut while developing the repair procedures. Collocation of these
skilled individuals with the designers knowledgeable of the mechanisms and fully representative hardware
allowed the creation of a procedure within one day that would return the mechanisms to a completely
nominal condition. Interestingly, this was for a failure that was at first believed to be beyond the
capabilities for effective repair.

One additional note to make regarding the team make-up was that the experience level of the members
was critical when evaluating EVA options. As discussed earlier, safety issues are very important in
determining acceptable EVA operations. They are also very subjective and open to interpretation. Steel
cables and pulleys are two items specifically on the list of dangerous or unacceptable items, and they
were the very items that required handling in order to fully repair the Tension Mechanisms. Having on the
troubleshooting team an astronaut of sufficient experience and respect such that general safety
“requirements” could be tailored real-time allowed the creation of a repair that could fully restore the
mechanisms to its full design capability while still providing for the essential requirement of crew safety.

With a repair procedure now in hand the efforts of the team turned to training the Orbiter crew on the
procedure and creating contingency procedures and contingencies to the contingencies in case there
were difficulties with the baseline repair plan. The first task of training the Orbiter crew turned out to be
easily dispatched with the use, once again, of video capabilities. A video tape was made of the astronaut
on the troubleshooting team walking through the repair procedure on representative hardware using
representative tools and gloves, and the video was then transmitted to the Orbiter. The video and
associated written procedures and telecons allowed full understanding of the activities with little difficulty.
The contingency planning continued throughout the week until the time the actual repair operations were

NASA/CP—2002-211506 178
performed. Much to the satisfaction of all involved, the repair of the Tension Mechanisms was completely
successfully and required only a few minutes each (Figure 9). Although the multiple backup options were
never put to use, their development was essential to ensure the success of the mission.

Figure courtesy of NASA JSC

Figure 9. Nominally Deployed ISS Solar Arrays*

Troubleshooting & Repair – Some Issues in Using EVA Capabilities

In spite of the impressive array of capabilities described in the previous section, there were a variety of
impediments in using these capabilities to their utmost effectiveness and efficiency. Although some
barriers were technical in nature, more often they were “self-inflicted” by organizational or communication
barriers.

The two barriers that were technical in nature included the limited time available for live video downlink
and the limitations in view angles and zoom of the fixed video and still cameras. Both of these fall into the
category that it is hard to have too much of a good thing. The video provided such clear and descriptive
coverage of the critical deployment events and inspections that having such data real time for each
critical event could very well prevent anomalous or unexpected performance from resulting in a specific
failure. It could also prevent investigative inspections from missing critical items or wasting precious time
by requiring re-inspections. Although the timing for available real time video down link and the procedural
limitations in requesting it for specific activities is probably well understood by the “inner circle” of
operations personnel, neither were well communicated to the troubleshooting team at the solar array
design facility. Furthermore, specific recommendations to perform activities such as the second wing
deployment within this “window” were neither performed nor specifically rejected. The frantic pace of the
activities, complex nature of the mission and the need for tight control of the Orbiter’s activities all
contribute to the need for limitations in the exercise of this option. Thus the advice to mechanism
designers participating in future investigations is to be aware of the value of this capability and request
specific instructions or procedures for how its availability is determined or requested. Similar advice is
given for understanding limitations in dealing with other operational constraints that might be useful such
as eclipse times, spacecraft orientations, attitude control modes, etc. It should also be noted that it is
unrealistic to expect to quickly understand the full intricacies of the operation of a vehicle as complex as

NASA/CP—2002-211506 179
the ISS with a docked Shuttle during the heat of such an investigation, and efforts to gain familiarity with
mission constraints prior to the mission are worthwhile.

Noting the limitations in camera view angles and zoom as a technical barrier is somewhat of an
overstatement. The capability that was available with no extraordinary effort on STS-97 certainly
exceeded the expectations of the design engineers involved in the investigation. Nevertheless, there are
only a finite number of views available for any event. It is noted here primarily to advise the designer that
such limitations are present and that forethought as to what views would be most descriptive can be
beneficial. When such considerations might be critical, then clear and concise communication of
preferential views to the appropriate chain of command within the operations community needs to be
accomplished in a timely fashion. Considerations of the availability of the helmetcam may also prove
useful.

Several of the nontechnical impediments encountered in the investigation related to the lack of a clear
understanding of the basic organizational structure of the very broad and dispersed set of personnel and
organizations involved in the investigations and operations. At times there was some confusion on what
was specifically available in the flying tool inventory. At other times there was slow dissemination of
troubleshooting data and/or inspection results to each key organization involved in the investigation. And
during contingency discussions there appeared to be a lack of authority given to the multi-organizational
NASA/Contractor troubleshooting team established at the design facility. In each of these areas it was not
obvious where the buck stopped or to whom to direct the issue. Requesting the equivalent of an
organization chart showing the decision making hierarchy and the specific personnel involved in different
aspects of the investigation at the earliest stages of the investigation and maintaining it throughout would
help greatly in avoiding confusion or frustration in these areas. It would also allow more rapid closure of
actions assigned or requested. This area may be common and obvious to the inner circle of NASA and
operations personnel involved regularly in such events, but the design engineer or contractor should not
be shy in requesting such information when clarity is not initially provided. While requesting such
information, it must also be recognized that during a rapidly evolving anomaly investigation such as the
one described here, the NASA team is evolving equally as rapidly, with the necessary expertise being
brought in as new issues are identified or resolved, and resolution plans evolve and coalesce. The
design team needs to be prepared to deal with a flexible and rapidly changing customer interface and
should properly prepare for such an eventuality by team building with the customer prior to the mission.
While such team building must of course, have customer blessing (and funding), it should be pursued at
every opportunity by the subcontractor design team.

Finally, the most notable impediments encountered in non-technical areas were a limited and poorly
defined downward communication path to the design team, and an even more limited and poorly defined
upward communication path to the operations team during critical events. Examples of the lack of clarity
in downward communications include the previously mentioned issues relating to the lack of definition
provided for the availability of real time video and other orbital constraints and the lack of understanding
provided to the design team of the decision making hierarchy. The deployment of the second solar array
wing during real time video black out and portions of the eclipse against the specific recommendations of
the design team without notice was a further example of limited downward communication as was the
difficulty encountered in obtaining certain delayed down linked inspection data and specific planned
operation times. Going the other direction, one example of ill-defined upward communication channels
would be the final moments of the second wing deployment where the mechanisms were switched into
the low tension mode too early allowing the tension bar to be pulled off the Containment Box by panel
adhesion similar to the first wing deployment. At this point there was no defined process to provide input
to the decision makers and no request for technical input by them through the established telecon link. A
second notable example comes from the close inspection of the mechanisms for damage with the EVA
helmetcam. During this event, where real time video was available, a conclusive view of some small
stowage mechanisms of interest might not have been obtained except for the use of a cell phone carried
by the astronaut working with the solar array design team and the personal access he had to key
individuals. The lack of clarity in these communication channels provided the greatest opportunity for
potentially avoidable mistakes or wasted time. Here, again, the frantic pace of the activities and the
familiarity of the processes to those involved regularly in them might easily allow the definition of the

NASA/CP—2002-211506 180
available processes to all involved to be overlooked. Thus, the designer should be advised to clarify the
availability of such channels for communication early in the process before they find themselves
frantically searching for additional phone numbers and conference lines while the hardware they are
investigating is being operated or discussed on the TV in front of them.

Summary & Conclusions

The manned operation capabilities that proved most useful during the investigation and repair of the solar
array deployment anomaly included:
• Live video downlink capability over portions of the orbit,
• Delayed video downlink coverage of the full deployment event,
• Variable resolution still photos of requested items, incredibly detailed inspection capability from a
camera on the astronaut’s helmet,
• A variety of tools available on orbit,
• The ability of experienced astronauts to make safety judgments outside the design guidelines, and
• The availability of experienced astronauts and EVA support members to be quickly transported to
Lockheed Martin to work as a team with design personnel to determine appropriate repair
procedures.
Knowledge of these capabilities and their efficient use can greatly aid design engineers in quickly
troubleshooting anomalous hardware.

The technical limitations of most interest in the STS-97 solar array investigation were the limited time
available for live video downlink and the limitations in view angles and zoom of the fixed video and still
cameras. The advice to mechanism designers participating in future investigations is to be aware of the
immense value of these capabilities even with their limitations and request specific instructions or
procedures for how the availability of these resources is determined or requested. Similar advice is given
for understanding limitations in dealing with other operational constraints that might be useful such as
eclipse times, spacecraft orientations, control modes, etc. Furthermore, forethought as to what views
would be most descriptive can compensate for many of the limitation in the views. When such
considerations might be critical, then communication of preferential views to the appropriate chain of
command needs to be done. Considerations of the availability of the helmetcam may also prove useful.

The most notable impediments encountered in non-technical areas were:

1) The lack of a clear definition/understanding of the organizational structure and decision making
hierarchy of the very broad and dispersed set of personnel and organizations involved in the
investigations and operations, and
2) A limited and poorly defined downward communication path to the design team and an even more
limited and poorly defined upward communication path to the operations team during critical
events.
Certainly, the frantic pace of the activities, complex nature of the mission and the familiarity of the
processes to those involved regularly in them all contributed to the lack of clarity in these areas for the
design team. To prevent such organizational and communication issues from impeding the progress of an
investigation, the designer should be prepared to address such issues early and directly when involved in
investigations of this nature.

Although some technical and non-technical limitations in the manned operations capability were
encountered, the benefits of the human presence in the troubleshooting and repair operations allowed for
an extremely swift and fully successful restoration of the ISS solar array to its design capability without
impact to subsequent missions (Figure 9). The rapid collocation of skilled and experience EVA operations
personnel with the hardware designers from multiple companies combined with the impressive array of on
orbit capabilities not available in autonomous spacecraft operations was instrumental in this success.

NASA/CP—2002-211506 181
 References

1. Johnson, Mark E., Bert Haugen and Grant Anderson “Space Station Freedom Solar Array
th
Containment Box Mechanisms” 28 Aerospace Mechanisms Symposium, 18-24 May 1994, pp 1-16.
2. Allmon, Curtis and Bert Haugen “Space Station Freedom Solar Array Tension Mechanism
th
Development” 28 Aerospace Mechanisms Symposium, 18-24 May 1994, pp 123-128.

NASA/CP—2002-211506 182
The Lightweight Deployable Antenna for the MARSIS Experiment on the Mars Express Spacecraft
* * **
Geoffrey W. Marks , Michael T. Reilly , Richard L. Huff

Abstract

TRW Astro Aerospace developed and built an antenna subsystem for the MARSIS (Mars Advanced
Radar for Subsurface and Ionospheric Sounding) experiment on behalf of the University of Iowa, who
provides antenna and transmitter to the NASA Jet Propulsion Laboratory (JPL). The antenna flies on the
extremely weight limited Mars Express spacecraft due to reach Mars in 2004. The MARSIS antenna is an
example of a lightweight deployable structure that is designed purely for the space environment. Because
of this, any significant friction or drag that would occur on Earth will prevent its deployment, and its large
dimensions make it impractical to deploy in any test facility on Earth. The verification process has been
developed through the program and dynamic simulation has become the most important verification tool.

This paper will describe the unique design of the MARSIS antenna and provide details of the test and
verification program.

Introduction

The basic design for the foldable tube used in the MARSIS antenna experiment was originally conceived
for use on the weight limited Sounder Antenna for the planned NASA EUROPA Orbiter spacecraft. These
launch weight limits forced the creation of a completely new antenna element design that has been
1
named the Foldable Flattenable Tube (FFT).

The Mars Express spacecraft is also mass limited. It was determined that using the FFT would fit within
the MARSIS weight budget and allow the experiment to fly. The antenna deploys from its stowage box on
the sidewall of this small spacecraft to provide capability for a very low frequency sounding radar. It
deploys in Mars’s orbit as a 40-meter tip-to-tip transmit/receive dipole aligned in the flight vector and a 7-
meter receive-only monopole in the nadir direction. The whole mechanical antenna, excluding the feed
electronics, weighs 7.1 kg.

The goal of the MARSIS experiment, a joint endeavor between the University of Rome in Italy and JPL, is
to find water strata beneath the surface of Mars and to study the Martian ionosphere.

Background

The MARSIS instrument is a low-frequency ground penetrating radar sounder and altimeter, which uses
synthetic aperture techniques and a secondary receiving antenna to isolate unwanted reflections. The
operating altitudes of MARSIS are up to 800 km for subsurface sounding and up to 1200 km for
ionospheric sounding.

As implemented, MARSIS consists of two electronics assemblies and two antennas mounted on the
spacecraft. In operation, the MARSIS control electronics generates a linear frequency modulated chirp,
which is amplified and then radiated by the nadir-facing dipole antenna. Then the MARSIS switches to
receive mode and the return signal from the Martian surface is processed through both the dipole
antenna and the secondary monopole antenna. The monopole antenna, oriented along the nadir axis,

*
TRW Astro Aerospace, Carpinteria, CA
**
University of Iowa, Iowa City, IA
1
US Patent 6343442, Flattenable Foldable Boom Hinge
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 183
receives the off-nadir surface clutter, which will then be subtracted from the primary data in the ground re-
processing. The MARSIS radar operates at frequencies up to about 6 MHz.

The required mechanical configuration for the deployed antenna shown in Figure 1 is a dipole
transmit/receive antenna deployed in the orbit vector with a total tip-to-tip length of 40 meters, plus a 7-
meter monopole receive-only antenna in the nadir direction. Traditional designs such as a STEM antenna
were considered for the application but would not meet the combination of requirements for stiffness,
weight (~8 kg mass budget), thermal stability and cleanliness (no debris generation). The FFT, despite its
early stage of development, met these requirements and its simplicity gave the experiment team sufficient
confidence in a successful outcome to proceed. The program start date was in February 2000. After
overcoming some minor design problems the qualification was completed in October of 2001. At the time
of writing this paper, the flight hardware is assembled and ready for acceptance test.

The antenna structure consists of lightweight S-Glass/Kevlar composite tubes. The dipoles are 38 mm in
diameter and the monopole is 20 mm in diameter. With these dimensions the tubes are sufficiently stiff to
achieve the deployed frequency requirement of 0.05 Hz.

The tubes are folded and then compressed for stowage. They are folded at points along the length where
cutouts in the side of the tube prevent a singularity in the material. This allows the extremely flexible
composite to fold in a manner similar to a “carpenter’s tape hinge”. The 20-meter dipoles fold to a length
of 1.53 meters and the smaller monopole is folded to 1.3 meters. When folded, the tubes are compressed
into a stowage box for launch. The stowage box is a Nomex Honeycomb graphite-skinned construction
with three doors that enclose the compressed elements.

The stowage method is shown in Figure 2 and is shown in process in Figure 3. The three tubes are
stowed in boxes in layers and released sequentially so that there is no danger of entanglement during the
process. The tubes are stowed by folding them at defined points along their length where cutouts, as
shown in Figure 4, relieve the stress singularities at each side and allow the rest of the tube to flex into
the folded shape. The dipoles and the monopole are compressed from 38 mm to 19 mm and from 20 mm
to 10 mm, respectively, and consequently have considerable stowed strain energy. Deployment of the
tubes is initiated sequentially by small pyrotechnic devices that release the door latches. Once triggered,
the tubes push the doors open very rapidly on release and deploy out to their required configuration by
releasing the energy of stowage and the strain energy in the hinges. (This door energy is absorbed by
innovative, progressive friction dampers).

The stowed tubes emerge rapidly from their stowage all heading in the same direction. The root mounting
turns the two dipoles through 90 degrees into the flight vector while the monopole fires straight out in the
nadir direction and stays there. The tubes reach their full extension in less than two seconds because of
the initial release of stowed energy but the dipole tubes take another 30 seconds to rotate into position.
The dipole hinges have a very low torque of 0.2 N-m and therefore accelerate the fairly large inertia
slowly. On reaching full deployment, the tubes oscillate before completely straightening, though the tubes’
construction significantly dampens this oscillation, as explained below. The dipole element deployment
sequence is shown in Figure 5.

The composite tubes are not conductive per se although they are Indium Tin Oxide (ITO) coated to
prevent static build up. The actual antenna structure is a pair of stranded 22-gauge, silver-plated copper
wires that run along the length of the tube. These wires are interconnected ladder-style along their length
to provide redundancy. In addition, they are slightly rippled along the length to prevent differential thermal
motion between the wires and the near zero coefficient thermal expansion (CTE) tubes. The wires are
connected to the feed and sensing electronics, which are mounted close to the antenna roots and
beneath the stowage box as shown in Figure 6.

As discussed below, one of the most controversial aspects of the antenna project has been the test
program. The structural tubes of the antenna (thin-wall S-Glass/Kevlar composite, 38-mm diameter for the
dipoles and 20-mm for the monopole) are unable to sustain their own weight in Earth’s gravity.
Additionally, the deployment kinematics shown in Figure 5 prevent the use of conventional deployment

NASA/CP—2002-211506 184
support rigs while the deployment motion requires a very large area. Various methods for supporting this
motion were conceived but all proved impractical. The normal environmental tests such as stowed
vibration and thermal cycling are straightforward. The challenge is to test the function of the antenna
elements after exposure. The pyro-releases can be fired and the doors will be forced open and the
compressed tubes will eject themselves from the box. But after that point - unless there is a zero-g
environment - the tube hinges do not have enough torque to deploy and any significant friction will
prevent deployment. The agreed method is to verify that the hinge joint will always unfold –there is no
friction effect to stop it – and verify that it will release from the box. The remainder of verification is
provided by analysis. An ADAMS (Automated Dynamic Analysis of Mechanical Systems) model was
constructed and the individual elements of the model were correlated to test hardware. The model verifies
the deployment dynamics and the interaction with the spacecraft. A sample of the output is shown in
Figure 5.

The Design

The Tube Elements

Requirements:

• The tubes are required to meet the minimum deployed frequency of 0.05 Hz.

• The long dipoles must maintain a total tip deflection of 20 cm under all thermal extremes.

• The external finish must dissipate static charge and must not generate debris when deployed.

Construction:

• The tubes are a Kevlar and S-Glass composite that is specifically laminated to provide near zero CTE
in the longitudinal direction and to prevent creep of the matrix in the stowed condition (the tubes
remain compressed for up to two years on the trip to Mars). The tubes are laid up in 3.0-meter
lengths and then ITO coated. They are then sorted for straightness so that the combination of the
sorted tubes, when correctly oriented, will achieve the straightest possible finished element. The
tubes are then machined to create the fold points. The individual tubes are then spliced together to
form the complete element (Figure 7). The splice is designed to flex with the rest of the tube when
compressed into the stowage box. Although the splice effectively doubled the wall thickness, the use
of a flexible adhesive allowed the high degree of strain to be tolerated.

• The tubes were punched with a number of additional holes to provide access to install and support
the conducting elements. The conductive wires were soldered into a ladder frame such that
extensions to the rungs penetrated the tube walls and were staked in place. That way the legs of the
ladder were supported along opposing walls of the tube.

The Stowage Box

Requirements:

• To take advantage of the extreme low weight of the conducting elements a lightweight stowage box
was required. The design of the box was quite challenging. The three tube elements must be stacked
one on top of the other and mounted across the spacecraft wall as shown Figure 1. The monopole is
able to exit the box and fire straight out, so it could be contained in a four-sided box with a back and
lid. The elements of the dipole had to exit in the same direction as the monopole but then needed to
rotate through 90 degrees, each in different directions. The end walls of the box therefore could not
exist, compromising the structure of the box. The box is also required to support the local electronics
and mount in a kinematical manner to the spacecraft sidewall.

NASA/CP—2002-211506 185
Construction:

• The box is constructed of graphite-skinned Nomex honeycomb panels joined by graphite clips. The
box structure is then completed, as much as possible, by the doors keyed at the latch points as
described below. The panel skins are 0.19-mm M46J graphite laminates and the door skins are 0.38-
mm laminates, designed for high strength and stiffness but with a low mass. The external faces of the
box are painted with a conductive white paint for thermal and static conductivity considerations. The
box is mounted to the spacecraft sidewall on raised aluminum brackets designed to flex to allow
relative thermal expansion of the box and spacecraft.

The Mechanisms

Requirements:

• The compressed tubes needed to be stowed by a hinged door and latching mechanism. The doors
needed to resist the deploying force of the tubes through vibration and thermal cycling and still
release upon pyro-activation. Upon release, the doors would accelerate rapidly from the stowed strain
energy of the tubes and would need to be damped rapidly after a 90-degree rotation to clear the tube
deployments. This presented the challenge to allow the doors to rotate fairly free from 0 to 90
degrees to remove them from the deploying path of the tubes and then stop them between 90 and
180 degrees of rotation before hitting the box structure with any significant force. This required a
special type of fast-acting damping device to be integrated in a limited volume. The door latching
mechanism needed to constrain and latch the door at multiple points without interfering with the
tubes. It also needed to keep the door closed despite both thermal and vibration forces and then
release the door from a pyrotechnic bellows actuator at all temperature extremes along with a limited
volume.

Construction:

• The composite door has internally bonded hinge/latch inserts made of Titanium that interface with
Titanium hinge/latch inserts bonded in the box panels. The door has a graphite rib bonded on the
outside for added stiffness. The unique hinges were designed to be multi-functional providing both a
hinge and latch point as well as having a special tapered hinge tang for the damping feature. The
tapered hinges will compress Belleville and composite washer stacks that are mounted on the hinge
pins and latch rods as the door rotates. Clearance is adjusted to allow the washers to be engaged at
the correct rotational point. Energy is absorbed by both frictional forces and spring compression.

• The latching is performed by a high-strength Titanium rod that functions both as a hinge pin and
latching mechanism. The Vitralube-coated rod runs the full length of the door and has machined flats
at the latch points. As the rod is rotated and constrained by the pyro-mechanism, the flats rotate and
compress down on the latch from the door and hold it onto the latch receptacle on the box panel
hinge/latch insert. A pyro-release mechanism holds the rod in place after it is rotated into position. In
the latched position, the rod is twisted torsionally to hold force on the latches. A link arm is pressed
onto one end of the rod to provide the torque. A latch arm runs off of the link arm and is held back by
two disks in the pyro-mechanism. As either bellows’ pyro is activated, a disk will drop and release the
latch arm allowing the latch rod to rotate freely and release the door and tubes. (Shown in Figure 8)

Analysis

Deployment Analysis

The ADAMS dynamic modeling software was the most important verification tool for on-orbit performance
predictions. The model that was used to predict the deployment profile and the on-orbit dynamics
incorporated component test data for the compression preload, for the hinges and for the boom stiffness.

NASA/CP—2002-211506 186
Stiffness correlation measurements were made on the completed flight- and qualification-tube assemblies
as they floated on a water table. The calculations were also checked using a COSMOS finite element
model. The data is used to prepare the spacecraft control system algorithms.

Launch Phase

The stowed antenna assembly was evaluated for the predicted launch sine, random vibration, acoustic
and shock environments using a finite element model generated in COSMOS/M. While no detailed
representation of the flattened antenna tubes was put in this model, their mass was distributed
proportionally through the structural box model. Preload stresses were superimposed on results of the
dynamic analysis for the defined environments. Two percent structural damping was assumed for this
analysis. In fact, test results indicated significantly higher damping, which varied by frequency and load
level.

Transfer Phase

The transfer phase of flight, with the antenna elements still stowed, involved very low mechanical loads,
but significant temperature variations from ambient build temperature. The structure, which incorporates
design features to accommodate the stresses resulting from the dissimilar coefficients of thermal
expansion of the build materials, was again evaluated using a COSMOS/M finite element model.

Deployed Phase

The deployed system was evaluated using another COSMOS/M model. Frequency response and
strength were evaluated for the predicted environment.

Testing

As stated above, the test program was non-conventional because of the difficulty of testing the extremely
lightweight tubes in a representative manner. The philosophy followed was to demonstrate that the
components of the tube were 100 percent reliable such that if the doors opened it could be assumed that
the elements would fully deploy. It was an accepted fact the tubes were only able to fire themselves onto
a flat table as shown in Figure 9. The test program has qualified the hardware as follows.

Element Tests

The tube elements were tested to demonstrate the following properties:

• Hinge actuation reliability. Multiple operations of sample hinges were made under all temperature
extremes including in liquid nitrogen. The tests did demonstrate some problems with the initial cut out
configuration. Design modifications were made, arriving at the shape shown in the figures.

• Hinge torque/rotation angle characteristics in all axes – this data was used as an input to the ADAMS
analysis.

• Tube lateral compression stiffness - also used in the ADAMS analysis.

• Tube and tube splice long-term creep properties under compression. Tube samples have been kept
under compression for upwards of a year and have shown no significant load decrease.

• Tube strength and stiffness have been measured for all root configurations. The tube strength is
reduced at the fold cut-outs. The nominal tube has a bending strength of 17 N-m (150 lb-in) with 13.5
N-m (120 lb-in) with the slots in the neutral axis, and 0.58 N-m (60 lb-in) when rotated 90 degrees.

NASA/CP—2002-211506 187
Structural Model

A model of a containment box for the larger tube element was constructed to prove the various functions.
It was subjected to repeated deployments to demonstrate the robustness of the tube elements and the
wire installation. It proved the function of the release system and the door energy dampers. It was also
subjected to vibration test and thermal testing. All functions were successfully demonstrated.

Radio Frequency (RF) Performance Model

One of the unknowns of the whole instrument design was exactly how the antenna would work. To prove
its performance two of the 20-meter dipoles were delivered to the University of Iowa. They were bold
enough to construct a field experiment in which the two dipoles and a representation of the spacecraft
body were lifted from the ground and suspended vertically beneath a helicopter. The test was carried out
in Colorado where the whole assembly was lifted 3000 m (10,000 ft) above the ground, looking sideways
at the neighboring mountains. The experiment worked well, proving the function, and the whole
experiment was carried out with no significant damage to the fully flight representative hardware. The
event is shown in Figure 10.

Qualification Hardware

A fully flight-like assembly was built as shown in Figure 2. It was subjected to the following tests:

• Functional Test - The release functions and element performance were demonstrated. The elements
were released onto the deployment table in turn. During this test contamination samples were taken
to verify that the tubes did not generate debris contaminants.

• Vibration Test - The assembly was mounted to a vibration fixture and subjected to sine and random
vibration environments.

• Thermal Vacuum - The assembly was thermal cycled in vacuum. The safety straps (the red clamps
shown in Figure 2) were kept on for this test. The doors were released at temperature extremes
during the test and the safety straps prevented deployment of the tubes.

• Final Functional Test - The full performance of the release mechanisms and tube energy release was
again tested. The tubes were again deployed in sequence onto the flat table and again particulates
were monitored. The particulate count was small and acceptable.

Flight Hardware Test

At the time of writing this paper, the flight assembly is almost complete and will be subjected to the same
test sequence as the qualification hardware.

Conclusions

This program required new methods to be developed to validate the design and to assure performance of
flight hardware. A combination of analysis and component test was utilized to accomplish this goal.

NASA/CP—2002-211506 188
 Figure 1. Mars Express Spacecraft with Deployed MARSIS Experiment

NASA/CP—2002-211506 189
 Figure 2. MARSIS Antenna Stowed for Launch with Safety Clamps in Place

NASA/CP—2002-211506 190
 Figure 3. Stowage in Process

Figure 4. Tubes with Cutouts

NASA/CP—2002-211506 191
 Figure 5. Dipole sequence

NASA/CP—2002-211506 192
 Figure 6. Electronics

NASA/CP—2002-211506 193
 Figure 7. Spliced Tubes

Figure 8. Pyro Release

NASA/CP—2002-211506 194
 Figure 9. Deployment Test

NASA/CP—2002-211506 195
 Figure 10. RF Test Equipment Suspended from Helicopter

NASA/CP—2002-211506 196
 MABE: High-Precision Tip/Tilt Mechanism Based on Magnetic Bearing Technology
* **
N. Loix*, J.Ph. Verschueren and L. Scolamiero

Abstract

This paper presents a 3 degrees-of-freedom long-stroke, high-resolution tip/tilt mechanism for space
interferometric mission applications based on magnetic bearing technology. A breadboard model of the
mechanism has been developed and successfully passed environmental and functional tests. The
translational and rotational resolutions are respectively 1 nm (over a 5-mm stroke) and 10 nrad (over the
two 5 mrad rotation ranges). Thanks to the innovative magnetic bearing design, the power dissipation in
the mechanism (under gravity compensated conditions) is less than 1 mW, which makes it a good
candidate for infrared interferometric applications in a cryogenic environment where dissipated power has
to be minimized. An in-orbit test demonstration of a proto-flight mechanism model is currently under
preparation, and is to be flown on board the Space Shuttle GAS canister carrier. The presented activity
has been performed within the ESA GSTP research program.

Introduction

Scientific satellites are increasingly demanding in pointing accuracy and future missions are expected to
require performances one or several orders of magnitude better than the current state of the art. The
success of these missions is closely linked to the development of advanced control techniques for
disturbance rejection and mechatronic components capable of managing large stroke and very high
precision in the nanometer range. Previous studies [1, 2] have shown that magnetic bearing systems can
perform this challenging requirement. During the MABE project, a breadboard model of a long-stroke,
high-resolution tip/tilt mechanism based on magnetic bearing technology has been designed,
manufactured and successfully tested.

The developed mechanism has 3-degrees-of-freedom (DOF), one translation (tip or piston) and 2 rotations
(tilt). It can be used to control the relative motion of optical mirrors for space Interferometers like for the
GAIA initial optical concept, or the DARWIN infrared interferometer (both ESA projects). It can be utilized
for space telescope applications as well, where a high-accuracy alignment / pointing / refocusing
mechanism of segmented primary mirrors or secondary mirrors is needed. The power dissipation in the
mechanism has been subjected to special attention and has been kept to a minimum thanks to an
innovative magnetic layout that includes permanent magnets. Careful selection of materials allows for
possible use in a cryogenic environment. The main characteristics of the mechanism are:

• The payload is a 1-kg dummy mirror (aluminum plate)

• Number of functional DOF: 3 (2 rotations and 1 translation)
• Stroke:
o Piston 5 mm
o Tilt +/- 2.5 mrad
• Resolution:
o Piston 1 nmrms (RMS value of the moving average over 0.75s)
o Tilt 10 nradrms (RMS value of the moving average over 0.75s)
• Bandwidth: 16 Hz
• Perturbation attenuation: 100 dB at 1 mHz
• Guiding: 1 nmrms (RMS value of the moving average over 0.75s)
• Power dissipation in the mechanism: <1 mW

*
Micromega Dynamics, Angleur, Belgium
**
ESA/ESTEC, Noordwijk, The Netherlands
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 197
 Description of the mechanism

In order to achieve very-high resolution over a long stroke in only one stage, the moving part of the
proposed mechanism is magnetically suspended; during operation there is no contact between the moving
and static parts. The system is divided into three subsystems: (i) the guiding system, based on 3 magnetic
bearings constraining the in-plane movements (Tx, Ty, Rz) of the rotor, (ii) the positioning/steering system,
using voice-coil actuators and high-resolution linear scales, which drives the useful DOFs (Tz, Rx, Ry) and
(iii) a locking device. The mechanism is shown in Figures 1 to 3. The required space to operate the
mechanism is a cylinder of φ=320mm and h=80mm. The mechanical interface consists of 3 holes
(φ=6mm) evenly spaced on a 290-mm-diameter circle. Most of the mechanical parts are made of Al-7075.
The magnetic parts are made of VACOFLUX that has good performance at cryogenic temperatures. All
the contact surfaces of the locking device have been hard anodized and treated with PTFE to improve
their tribology properties.

Figure 1. MABE 3 DOF high-precision mechanism

Magnetic bearings
Unlike voice-coil electromagnetic actuators, reluctance force actuators or electro-magnets allow large
displacements in the directions perpendicular to the action of the actuator and hence, are good candidates
for magnetic bearing applications [3]. However, in these reluctance force actuators, the force comes from
the variation of the energy stored in the magnetic circuit, which is proportional to the square of the current
in the coils and inversely proportional to the square of the air-gap thickness. This quadratic relation of the
force to the applied current leads to the need for operating the actuator around a bias current (i=i0+ic). This
bias current is the source of most of the power dissipation in standard magnetic bearings. For this
application, permanent magnets have been used to create the required bias magnetic field, hence
drastically reducing the required electrical power. The permanent magnets are made of Samarium Cobalt
(SmCo5). More powerful magnets can be made of NeFeBo but the literature has reported that this material
experiences instabilities at cryogenic temperature [4, 5, 6]. The magnetic circuit is made of VACOFLUX,
which has also proven to be efficient at cryogenic temperature. The relation of the force to the air-gap
thickness leads to a negative stiffness, resulting in an unstable open-loop system that requires a control
system to operate it. The two quadratic relations also lead to large non-linear effects that can be
minimized by a proper magnetic design [7]. The selected magnetic configuration is shown on Figure 4.

NASA/CP—2002-211506 198
 Figure 2. MABE Stator Figure 3. MABE Rotor

The guiding control architecture is based on three independent magnetic bearings, each consisting of (i) a
reluctance force actuator, (ii) a high-resolution eddy-current position sensor and (iii) a local controller.
Being simple PID, the local controllers have been implemented using analog electronic components. The
bearings are placed tangentially to a circle and separated by 120° (Figure 2). The diameter of the circle is
twice the radius of gyration of the moving payload, leading to the same open-loop frequencies for the
constrained translation and rotation DOFs. The bearings are operated near their magnetic equilibrium
point leading to power dissipation in the bearings less than 1 mW over the entire thermal operating range
(-10°C / +50°C).

Figure 4. Heteropolar reluctance force actuator

Steering
The three steering DOFs (1 translation Tz and 2 rotations Rx & Ry) are driven by means of linear voice-
coils. For the same reasons as mentioned previously, only the actuators based on SmCo magnets have
been considered. The translation and rotation displacements are measured by means of three linear
measurements. The high-stroke to resolution ratio (>22 bits) imposes the use of digital measurement
devices, either laser interferometers or optical linear scales.

Because the proposed mechanism will in due course be used in a closed-loop system driven by a specific
optical sensor (e.g. a wave front error), the emphasis was put on the resolution of the actuator rather than
on its precision or accuracy. Linear scales have been preferred to laser interferometer because of their
rugged design, good repeatability, and very high-resolution (1 bit = 0.3 nm). The alignment procedures
have also been simplified by using an optical linear scale with a large standoff (3 mm).

NASA/CP—2002-211506 199
The steering control law is based on a multi-input/multi-output (MIMO) frequency-shaped LQG. Because
of the complexity of MIMO controllers, the control system has been implemented on an ADSP-21020
processor from Analog Device, for which a compatible space-qualified processor has been developed by
TEMIC under an ESA contract. The DSP processor is also used for the monitoring of the magnetic
bearings through 16-bit A/D converters.

Locking device
During the launch and landing periods, all the mobile payload movements should be restrained; therefore
a locking device has been designed. The contact surfaces between the fixed and mobile parts consist of
(i) a cylindrical bumper made of VITON (see Figure 2), (ii) an aluminum/aluminum conical contact, and (iii)
three single point contacts. In order to improve the tribology properties of these contacts and to ensure a
stiction-free release while in-orbit, the VITON bumper contains 10% of PTFE inclusions and the aluminum
contact surfaces have been hardened and treated with PTFE, according to [8].

Currently the preload force is obtained by means of a locking screw; however, during the Phase C/D of the
project, an autonomous locking mechanism will be designed. It should be able to release the mechanism
after the launch and to lock it back after the in-orbit experiment has been achieved. It should also be
locked when the power is off. Preliminary investigations have led to the selection of a paraffin actuator to
drive this mechanism.

Tests

Preliminary tests have shown that the mechanism experiences some Control Flexibility Interaction (CFI)
for both the analog magnetic bearings controllers and the digital steering controller. In both cases, it is the
support plate of the bearings and the voice-coils that is destabilized by the control system. To reduce this
CFI, it has been decided (i) to reduce the control bandwidth down to 16Hz and (ii) to notch the controllers.
In the Phase C/D of the project, the mechanical design will be changed to increase (i) the modal frequency
of this plate (currently 330Hz) and (ii) its structural damping.

Functional and environmental tests have been performed on the mechanisms. These tests consist of:
¥ Performance tests verifying that the mechanism is compliant with the performance specification;
¥ Vibration tests verifying that the mechanism (excluding the electronics) can sustain the vibration loads
from the American Space Shuttle;
¥ Thermal tests verifying that the mechanism (excluding the electronics) can operate within the required
thermal range and that the extreme non-operational temperatures do not damage the mechanism.

Oscilloscope Multimeter

Commands
+
Acquisition
MABE MABE
Mechanism Electronics

Laser beams

PC HOST #1
Acquisition

Metrology
System

ZYGO VME Box

GPIB
Interface METROLOGY PC

Figure 5. Test electronics layout

NASA/CP—2002-211506 200
 Figure 6. Test set-up with gravity compensation device

Performance tests
The performance tests consisted of both static tests to determine the mechanism resolution (noise) and
stroke, and dynamic tests to determine the closed-loop bandwidth and the dynamic perturbation
attenuation. The test setup is represented in Figures 5 and 6. It consists of the 3-DOF mechanism, a 3-
axis laser interferometer and a gravity compensation device made of three soft springs. During the test,
-10
the setup was subjected to a ground acceleration of 80 µgrms or 10 m²/Hz between 0 and 500Hz. Figure
7 shows the PSD of the ground acceleration.

Figure 7. Ground acceleration PSD at Micromega facilities

During the static tests, the piston and tilt of the mechanism were acquired at a sampling rate of 2 kHz, a
moving average over 0.75 second is then performed and the RMS value of the averaged data is
computed. The measurements are repeated at the various extreme positions. The control currents in the
bearing and the voice-coils are also monitored; these are used to compute the power dissipated in the
mechanism and also the perturbation force generated by the mechanism on the support. Table 1 gives the
results of these static tests. In addition to very good positioning/steering performances, very low power
dissipation in the mechanism (bearings and voice-coils) has been observed. This is mainly due to the use
of permanent magnet-based actuators and to the operation of the bearings around their magnetic
equilibrium point. Furthermore, the perturbation generated by the mechanism can be neglected.

NASA/CP—2002-211506 201
 Table 1. Static tests results
TEST TEST CRITERIA Results
Magnetic bearing
Resolution RMS value of the MA over 0.75s < 1nmrms 0.774 nmrms
Power dissipation <100 mW/channel 0.06 mW
-6 -9
Perturbation Quasi-static part of control force PSD < 10 N²/Hz 0.89 10 N²/Hz
Steering DOF
Resolution
Piston RMS value of the MA over 0.75s < 1 nmrms 0.793 nmrms
Tip-tilt RMS value of the MA over 0.75s < 10 nradrms 8.391 nradrms
Stroke
Piston +/-2.5 mm +/-2.5 mm
Tip-tilt +/-1.25 mrad +/-2.5 mrad
Power dissipation <100 mW/channel 0.025 mW

The response of the mechanism to a dynamic position perturbation (i.e., thermal load) has been
established by measuring the error of the system to a positioning/steering command. A harmonic
command was applied successively on the 3 DOF at various frequencies, for various amplitudes (to
identify potential non-linearities) and near the maximum stroke of the mechanism. As the system has
proven to be very linear, only the response of the system to the largest perturbation will be considered, in
order to obtain the best signal to noise ratio. Table 2 gives the experimental results of the dynamic test.
Figure 8 shows the experimental frequency response for the piston DOF, compared with the prediction
taking the gravity compensation device into account and to the prediction of the system in-orbit operation.
One can see that,
(i) The attenuation factor of –100 dB is obtained at a frequency of 0.001 Hz, which is ten times better
-4
than the required 10 Hz, despite the reduced bandwidth;
(ii) Higher attenuation ratios are achievable in a zero-g environment;
(iii) The cross coupling between the DOF on which the positioning/steering command is applied and
the other DOF cannot be neglected.

It can be shown that the observed cross coupling is due to the non-uniformity of the stiffness of the gravity
compensation springs. These stiffness variations have been measured and introduced in the model and
Figure 9 shows the comparison between the predictions, with and without the stiffness non-uniformity, and
the experiment. This confirms that the observed cross coupling is due to the gravity compensation device
and, therefore, better performances are expected for the in-orbit model.

Figure 8. Piston movement attenuation factor

NASA/CP—2002-211506 202
 Table 2. Dynamic tests results
Reference Freq Attenuation (dB)
(Hz) Tz Rx Ry MB#1 MB#2 MB#3
0.001 -104.5 -114.0 -121.2 -123.6 -126.4 -121.5

Tz=1mm
0.01 -84.4 -92.8 -100.9 -104.0 -106.4 -102.4
0.10 -63.5 -72.1 -80.4 -85.0 -86.4 -83.3
1.0 -31.7 -52.6 -61.3 -76.6 -78.5 -76.8
Rx=1mrad 0.001 -158.8 -95.6 -142.3 -137.8 -134.1 -131.6
0.01 -140.1 -75.6 -124.4 -118.8 -114.7 -112.1
0.10 -121.1 -55.4 -120.2 -99.7 -95.7 -93.3
1.0 -103.2 -29.9 -99.3 -91.3 -87.0 -85.4
0.001 -154.9 -145.8 -99.6 -138.0 -128.4 -147.8
Ry=1mrad

0.01 -142.2 -118.7 -79.6 -118.7 -108.8 -129.5

0.10 -129.5 -118.0 -59.1 -99.8 -90.0 -109.5
1.0 -111.6 -97.9 -29.8 -91.1 -81.4 -101.4

Figure 9. Piston Cross-coupling

NASA/CP—2002-211506 203
Vibration Tests
Vibration tests, including sinusoidal identification and random qualification tests (Figure 10), have been
performed on the mechanism. The applied random vibration levels are the ones applicable for payloads to
be flown on board the Space Shuttle GAS canister carrier. Special attention has been paid to the local
vibration modes of the mobile part in the stator air gap (which is only 0.3 mm). In order to verify that no
contact occurs during the vibration test between movable parts and the high-resolution eddy-current, the
signal of the latter has been monitored during the vibration tests. This has shown that (i) the peak
deviation from the steady-state position is 0.22 mm, which is safe enough; (ii) there is a rotational settling
movement of the payload on its bumper at the start of the vibration test sequence. This has led to re re-
design of the locking device.

Frequency PSD Level (g²/Hz) 1

(Hz) Qualification Acceptance

Acceleration (g²/Hz)
20 0.025 0.0125
0.1
20-50 +6dB/oct +6dB/oct

50-600 0.15 0.075

600-2000 -4.5dB/oct -4.5dB/oct

0.01
2000 0.025 0.0125
10 100 1000 10000

Overall 12.9grms 9.1grms Frequency (Hz)

Figure 10. Vibration test levels

Thermal Tests
The most sensitive hardware is the optical head of the linear scales whose operating and storage thermal
range are –10°C/+50°C and –40°C/+50°C, respectively. Therefore, thermal tests have been performed on
the mechanism, with the exception of the electronics, to verify that is it able to face this thermal range.
Figure 11 shows the test setup; the mechanism has been placed in a THERMOTRON thermal chamber
and equipped with thermocouples at adequate locations (rotor, stator, gravity compensation device…).
During the operational test, the mechanism is switched on and remains at rest at its neutral position
(Tz=Rx=Ry=0mm/mrad).

Figure 12 shows the thermal cycle for the test in operating conditions. The output of the magnetic bearing
sensors and the linear scales, as well as the control current flowing in the various actuators, are measured
during the constant temperature plateaus. The system has proven to be fully functional at the prescribed
temperatures. Lower resolution values have been observed during these tests, mainly due to the high-
level of perturbations in the thermal chamber (vibrations from the cooler pumps, air blowing the system…).
Figure 13 shows the variations of the current in the bearings during the test. These variations come from
the thermal expansion of the eddy-current sensor support. Because the closed-loop keeps constant the
distance between the eddy-current sensor and the bearing rotor, this results in a bias between the zero of
the sensor and the magnetic equilibrium point of the bearing, hence the increased current.

A thermal cycling test has also been performed on the locked mechanism (8 cycles between –40°C and
+50°C). No mechanical failure has been reported. Only a slight sticking between the moving payload and
the bumper has been observed, which can be easily overcome by the voice-coils.

NASA/CP—2002-211506 204
 Figure 11. Thermal test setup

Figure 12. Temperature measurements (operating tests)

Conclusions

During this project, a 3-DOFs (1 translation & 2 rotations) long-stroke, high-precision mechanism has been
developed. The mechanism can be used to control optical mirrors translational and rotational motions for
interferometric space missions, or for alignment / pointing / refocusing purposes on primary segmented or
secondary telescope mirrors. It is based on the magnetic bearing technology that experiences contact-free
movements. This allows achieving simultaneously large bandwidth, high resolution and long stroke. The
achieved performances of the mechanism are 1 nmrms over 5 mm for the translation DOF, and 10 nradrms
over +/-2.5 mrad for the two rotation DOFs.

NASA/CP—2002-211506 205
 Figure 13. Bearing power (operating tests)

The power dissipation in the mechanism has been kept very low to allow its use in cryogenic applications.
Thanks to an innovative magnetic design, the power dissipation in the mechanism has been proven to be
less than 1 mW over the entire thermal range.

The mechanism has successfully passed a dedicated test campaign, including random vibration exposure,
functional test at extreme operational temperature, and thermal cycling. A proto-flight model of the
mechanism is currently being developed to undergo in-orbit test on board a Space Shuttle GAS Canister
carrier. This activity has been performed under the ESA GSTP research and technology program.

References

1. H. Takahashi, T. Akiba & F. Kondo, A magnetically-suspended mirror driving mechanism for FTIR,
Fourth Int. Symposium on Magnetic bearings, August 1994, Zurich.
2. C.P. Britcher & N.J. Groom, Current and Future Development of the Annular Suspension and Pointing
System, Fourth Int. Symposium on Magnetic bearings, August 1994, Zurich.
3. “Active Magnetic Bearings”, Gerhard Schweitzer, Hochschulverlag, 1994.
4. D.M. Cannon & M. Brereton, “Design of magnetic bearing for use in SIRTF’s Tertiary Mirror
Assembly”, SPIE Vol. 1340 Cryogenic Optical Systems and Instruments IV, 1990, pp. 373-382.
5. K.H. Müller et al., “The Dip in Demagnetization Curves of Sintered NdFeB Permanent Magnets”, Acta
Physica Polonica, vol. A72, pp. 89-92, 1987.
6. J.R. Cost et al., High Performance Permanent Magnet Material “Radiation effects in Rare-Earth
Permanent Magnets”, pp. 321-326, Materials Research Society, 1987.
7. C.K. Sortore et al., “Permanent Magnet Biased Magnetic Bearings-Design, Construction and Testing”,
nd
2 International symposium on Magnetic Bearing, July 12-14, 1990, Tokio, Pg 175-182.
8. Space Tribology Handbook, European Space Tribology Laboratory, 1997.

NASA/CP—2002-211506 206
 Development of a Fine Steering Mirror Assembly
*
Ruben Nalbandian and Jeff Williams*

Abstract

Inter-satellite and terrestrial laser communications depend on the ability to reliably direct a laser beam,
with extremely high pointing accuracy, low power, and sufficient bandwidth to allow rejection of
environmental disturbances.

To address these demanding requirements, Moog Chatsworth Operations has developed a miniature
Fine Steering Mechanism (FSM), used for space laser communications, that is more efficient than voice-
coil actuated devices and is capable of more travel than piezoelectric actuated devices.

The design challenge was to develop a lightweight, compact, high bandwidth, low power, thermally stable
two-axis Fine Steering mechanism, capable of 62.4 degree angular range.

This paper outlines the design, development, and testing of the two-axis Fine Steering Mechanism
assembly, with particular emphasis placed on the magnetic design and the gimbal flexure development
and testing.

Introduction

In order for the newly emerging laser communications to be feasible for both inter-satellite and land-
based telecommunications, a Fine Steering Mirror-tilting device is needed to direct the laser beam from
the transmitting source to the receiving target. Pointing a beam at a target mirror several kilometers away,
therefore, requires a positional accuracy on the order of one (1) microradian. At the same time, the device
must be capable of moving the mirror as much as 62.6 degrees to provide adequate pointing coverage of
the steered beam. The device also must have sufficient control bandwidth to reject disturbances caused
by environmental factors and machinery vibrations. For small signal disturbances of about 0.01 degrees,
the device bandwidth must be about 200 Hz. Obviously larger disturbances must be rejected at lower
bandwidths.

Two technologies are considered for actuation of fast steering mirrors: voice-coils and piezoelectric
actuators. Voice-coils offer larger stroke for relatively low power; however, they are limited in bandwidth
and stiffness. Piezoelectric actuators on the other hand offer high bandwidth and stiffness, but are limited
in position output range. The Fine Steering Mechanism presented in this paper combine the benefits of
both technologies, resulting in a relatively high bandwidth, low power, relatively large displacement
mechanism of compact dimensions.

Principle of Operation

The principle of operation of this mechanism is electromagnetic. An array of four (4) identical
electromagnetic circuits is positioned with 90 degrees angular separation from each other around the
periphery of a circle. The electromagnetic design differs from that of conventional voice-coil actuated
systems in that four permanent magnets are mounted to the moving mirror platform, while the
corresponding coils and the coil cores, which constitute the more massive parts, reside within a fixed
housing. Due to the magnetic pull force of the permanent magnets on the ferromagnetic coil cores, the
mirror platform is continuously under a pull force. A magnetic return structure surrounding the coils
creates a closed magnetic circuit. The magnetic flux in the circuit is the resultant of that due to the
permanent magnet and that due to current. By changing the current and polarity in one coil, a magnetic

*
Moog Inc., Chatsworth Operations, Chatsworth, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 207
field can be generated that either opposes or aids the magnetic flux in that coil circuit. Simultaneously
changing the current in the opposing coil generates a torque in the plane of the two coils. The resulting
torque tilts the mirror platform about its center. Similar operation of the other two coils produces motion on
the orthogonal axis, so that the mirror normal can be positioned anywhere within an optical cone. The
resulting motion is due to the differential forces produced in individual coils, and resisting spring forces.
The mirror platform is mounted on a flexible beryllium copper diaphragm, which provides support and
restoring torque in any direction in reaction to the electromagnetic torques.

In addition to the tip and tilt motions produced differentially, piston motion of the mirror may also be
generated. Current in all four coils may be varied simultaneously, positioning the platform linearly along
its centerline, thus providing the additional degree of freedom. This motion can be utilized in focusing
applications.

Operation of the FSM is fundamentally different from that of other designs using voice-coils. Torque is
produced by varying the magnetic fields inside the attractive air gaps, rather than inducing Lorentz forces
on moving coils. The design is also physically different in that it uses small magnets and large coils, while
the voice coil designs use relatively large magnets and thin coils. Because of this, the extra number of
turns and larger wire size reduces power consumption, while small magnets offer lower cost and easier
handling. One added feature is the stationary mounting of the coils, resulting in higher reliability, as a no
moving electrical connections are required. No mechanical clearance problems associated with moving
parts inside and around the coils are present, thus making the design easier to assemble and reducing
manufacturing cost.

The Fine Steering Mirror assembly is more limited in angular range than voice-coil operated devices, due
to a significant and limiting decrease in magnetic efficiency with increasing air gap. However, the FSM is
significantly less limited in this regard than is the piezoelectric actuator. Piezoelectric actuators may
require 100 Volts to achieve displacements on the order of 0.5 degrees, whereas the FSM can move 2.5
degrees with only 15 Volts applied. Drift and linearity, major concerns in piezoelectric actuators, are non-
existent with the FSM.

The first development unit, shown in Figure 1, used non-contacting inductive position sensors mounted
between the coils to provide closed-loop feedback control. The inductive sensors presented cross
coupling effects due to the sensors not being in line with coils. These sensors are also expensive. The
inductive sensors were replaced by optical sensors, which are relatively small and can be positioned in
line with the coils. They require calibration and software to compensate for non-linear effects inherent in
these sensors. Their working range is in the range of 1 to 4.5 mm.

The FSM has only one moving part (the mirror platform) mounted on a frictionless flexure spring;
therefore, the position accuracy is primarily a function of the sensor error. Thermal distortion effects of the
platform are filtered out by using two sensors per axis. The two coils for each axis are physically wired in
series. In this way, the problem of controlling four channels has been reduced to a simple matter of
controlling two coil pairs, each for one axis of
motion.

Figure 1. Prototype Fine Steering Mirror

NASA/CP—2002-211506 208
 Analysis

Magnetic design and analysis of the FSM was performed using the Ansoft Magnet FE program. Two- and
three-dimensional finite element models were created to check the magnetic saturation. The net magnetic
flux in the circuit, which is a combination of the flux due to the permanent magnets and the coil fluxes,
determines the position of the mirror platform. Mirror platform motion is generated by increasing the flux
density in one air gap while simultaneously decreasing it in the opposite air gap. The magnetic model was
used to find the optimum circuit configuration that would allow flux to be both increased and decreased
efficiently.

The change in air gap due to mirror platform tilting drastically affects the flux in the air gap, resulting in
non-linear behavior which is very hard to control and not power efficient. After the first developmental unit,
the magnetic design was changed to one having a relatively constant air gap. Figure 2 depicts the
magnetic model of the improved coil/pole-piece design.

As shown in the Figure, the air gap between the

pole-piece and the backiron remains constant for
the full range of motion, thus exhibiting relatively
linear magnetic force (torque) behavior.

Figure 3 shows the generated force as a function

of coil current and the tilt angle.

One of the most important reasons for closing a

position loop around the FSM is to reject
disturbances caused by the target. If the mirror is
to stay focused on the target several kilometers
away, even the smallest vibrations could be
significant. Smaller vibrations are generally high in
frequency, whereas larger amplitude disturbances
are lower in frequency. The FSM was designed to Figure 2. Magnetic Model of Improved design
have a bandwidth of 1 Hz for motion of 20,000 micro-radians, a bandwidth of 10 Hz for motion of 2,000
micro-radians and a minimum of 100 Hz bandwidth for motions smaller than 200 micro-radians.

Force vs current and tilt

(flat yoke,cylindrical housing design, NdFe48 3dfl90 )

2 1-2
1 0-1
0
-1 -1-0
-2 -2--1
-3 Force, N
-4 -3--2
-5 -4--3
-6
-7 -5--4
-8
-6--5
-1
-2 -0.5
-1 -0.1
tilt, deg 0
0.1
0 -7--6
1 0.5 current, amp
2 1
-8--7

Figure 3. Force vs Current vs Tilt angle

NASA/CP—2002-211506 209
The moving parts of the FSM, i.e. the mirror, the mirror platform, and the flexure diaphragm were modeled
using the ANSYS Finite Element program to establish the structural resonant modes. The FE model is
shown in Figure 4. Results of the modal analysis for 1 mm and 1.5 mm mirror mount thickness are
summarized below:

Structural Modes - Baseline Configuration

7.2 Hertz - Axis of Rotation Mode

Mirror Mount Thickness - 1mm Thick

557 Hertz Plunging Mode

1024 Hertz Tilt Mode

Mirror Mount Thickness – 1.5mm thick

712 Hertz Plunging Mode
1468 Hertz Tilt Mode

In order to increase the span between the

operating frequency and the structural
resonance frequencies, the mirror mount
thickness of 1.5 mm was selected. Figure 4. Finite Element Model for dynamic analysis

Lessons Learned

The prototype Fine Steering Mirror assembly, as shown in Figure 1, was tested extensively by MCO and
JPL. During software and algorithm development, a frequency shift was observed. The first fundamental
frequency dropped from approximately 150 Hz to approximately 80 Hz. Careful inspection of the FSM
revealed that the Flexure Disk diaphragm had a sustained structural failure. Fatigue cracks surrounding
the center post of the diaphragm, as shown in Figure 5, shows the structural failure. Further investigation
also revealed degradation of magnet pole pieces due to contact and de-bonding of the magnets from pole
pieces (Figures 6 & 7).

Figure 5. Fatigue crack around center post Figure 6. Debonding of magnet from pole piece

NASA/CP—2002-211506 210
 Figure 7. Magnet degradation

The flex disk diaphragm was analyzed with the ANSYS finite element program. A non-linear large
displacement analysis was performed, results of which indicated that the maximum stress at the area of
failure is 324 MPa (47 ksi) (Fig. 8 & 9). Brush-Wellman was contacted for material properties of the
C17200 SN-TD01 solution-treated + cold-worked beryllium copper alloy. Some typical properties:
Endurance limit: 230 MPa (33.350 ksi)
8
Allowable stress: 214 – 248 MPa (31 - 36 ksi), R=-1, 10 Cycles Reverse Bending
Figure 10 is the modified Goodman diagram for the beryllium copper alloy. Fatigue strength for copper
alloys is usually defined as the stress sustainable without failure for 1000 million cycles. For the average
(50%) curves, this actually means that 50% will fail at this stress level.

Maximum loading
Max Principal
Stress
~ 47.6 KSI

Figure 8. Non-Linear FEM Analysis Results-Principal Stress vs Applied Force - Max. Displacement

NASA/CP—2002-211506 211
 FSM Flexure Disk
Maximum Stress

45000

40000

35000

30000

(ps 25000
i)vo
nM
ise SAG
σ
s
20000
LOAD
15000

10000

5000

0
0 2 4 6 8 10 12 14 16 18 20
Load Step

Figure 9. Stress versus Applied Load

FSM Flex Flexure Dia
Diaphragm
hragm Fatigue
Modified Goodman Diagram
80

Sy 70
Stress Amplitude - KSI

60

50

40

Se 1 DOF Deflection
@ 1.25 Deg Deflection
30
Sa 2 DOF Deflection
@ 1.25 Deg Deflection
Sa'
20
Modified Goodman Line
50% Mat Properties
10

0
0 10 20 30 40 50 60 70 80 90
Sm Sm' Syt Sut
Mean Stress - KPSI JTW/18/14/01

Figure 10. Modified Goodman Diagram for BeCu

NASA/CP—2002-211506 212
 Conclusions

Analysis results indicated that in spite of its simplicity, lowest cost, and minimum parts count, the flexure
disk diaphragm presents the following performance limitations:

1- Finite fatigue life

2- Mirror surface figure compromised due to bending. This was caused by placing the mirror in the load
path.
3- Excessive cross coupling proved to be detrimental to servo control
4- High power consumption resulted in excessive
heating.

Several design alternatives were considered. The

following is a summary of design trade-off studies:

Design Alternative 1 – Dual-Axis gimbal using flex-

pivot bearings (Figure 11)

This approach offers:

1- A true gimbal design
2- Rotation axes are coplanar and completely de-
coupled
3- Infinite fatigue life for 63º range of motion Figure 11. Dual axis gimbal with flex pivots
4- Stable mirror platform
5- Minimal power consumption

Design Alternative 2 – Metal flexure design (Figure 12)

1- Design offers modular approach

2- Can be sized for displacement and life
3- Stable mirror platform
4- However this approach requires axial volume, it is
more expensive, it may present modal cross
coupling, and may present high power
consumption.

Figure 12. Metal Flexure Design

Design Alternative 3 – Lord® Miniature Mount

MAA001 Series (Figure 13)

1- Simple
2- Rugged
3- Easy assembly
4- Inexpensive
5- Stable mirror platform
6- May create damping effects, finite life for
elastomer, finite shelf life, temperature affected
properties - stiffness, life, outgassing (mirror
clouding)

Figure 13. Lord Miniature Mount

NASA/CP—2002-211506 213
Based on the foregoing trade-off studies, it was decided that the true flex-pivot gimbal is the best follow-
on candidate for the FSM, since it offers a more stable mirror platform in absence of modal and cross-axis
coupling, and for its low power consumption. The elastomeric flexure base concepts are less attractive
due to life degradation issues, modal and cross-axis coupling, stiffness variability and outgassing issues.

The design was modified to incorporate the flex-pivot gimbal design. Flex-pivot bearings are frictionless,
stiction-free bearings uniquely suited for limited angle travel applications, such as in the case of the FSM.
Flex-pivots require no lubrication and provide infinite life for limited travel range. They provide high radial
and axial stiffness, low hysteresis, exceptional repeatability, predictable performance and electrical
continuity. The lack of any need for lubrication provides for operation in hard vacuum without the danger
of space welding, and problems associated with very small angular dithering movements inherent in
conventional ball bearings.

The magnetic design was modified for relatively constant air gap and the design was optimized for
materials, geometry and thickness.

An accelerated life test has shown that the design of the FSM is rugged and structurally sound for its
intended life expectancy. The FSM was exposed to 64,000 cycles at full range travel of 6 2.0 degrees. In
its intended use, the FSM would only experience a fraction of this full range cycle over its intended
lifetime of 15 years.

The optical mount and the mirror attachment were optimized to maintain the optical figure better than 2.5λ
@ 632.8 nm wavelength for temperatures of 0 to 60 degrees C. The mirror is bonded at three locations to
the mirror mount.

The drift problems inherent in inductive sensors were solved by replacing them with optical sensors. The
accuracy and repeatability of the FSM is within the 1 microradian specification, as demonstrated above.

Future Plans

As is the case with most electromagnetic designs, there is very little passive damping in the open-loop
device. More than any other factor, this makes it difficult to achieve high bandwidth without sacrificing
stability. Future FPM designs will involve a mirror platform with higher passive damping and consequently
higher bandwidth and better stability margins.

The current power requirement of the mechanism for operation at the maximum angular range of 2.6
degree is 0.250 A. The mechanism dissipates
less than 1.0 watt of power to move 2.6 degrees
in either the x- or y-axis, not including the
electronics. The goal is a complete system power
dissipation of less than one watt for operation at
an amplitude of 2.6 degrees.

Ultimately, the FSM ability to survive launch

vibrations and operating in the space environment
must be demonstrated. As the electromagnetic
design is of the same basic technology as many
other mechanisms used in space, no significant
problems are expected. Although the structural
analysis performed on the FSM indicates that the
current design is capable of surviving the
spaceflight launch environment and subsequently
operating, full qualification testing is planned.
Figure 14. Modified FSM unit

NASA/CP—2002-211506 214
 A Description and Performance of Mechanisms used in the
High Resolution Ionospheric Thermospheric Spectrograph
* ** + +
Phillip Kalmanson , Hsiung Chou , Russell Starks*, Kenneth Dymond , and Stefan Thonnard

Abstract

The High Resolution Ionospheric Thermospheric Spectrograph (HITS), built by the Thermospheric and
Ionospheric Physics Section of the Naval Research Laboratory, was designed for flight aboard the
Advanced Research and Global Observation Satellite (ARGOS). HITS was one instrument in an
aeronomy three sensor suite known as the High Resolution Airglow and Aurora Spectroscopy (HIRAAS)
experiment. The development of the HIRAAS instruments was a low cost, rapid prototyping effort.
ARGOS was successfully launched into polar orbit from Vandenberg Air Force Base on February 23,
1999 and continues in its successful operation at the time of this writing.

The HITS instrument is a one axis gimbaled spectrometer operating in the far and extreme ultraviolet.
HITS uses five separate mechanisms in its design. Each of these mechanisms was designed in order to
address specific issues affecting the HITS instrument ranging from optical system operation, data
acquisition, launch, and contamination control. What follows in this paper is a description of each
mechanism and the requirements they were meant to fulfill, the problems encountered during mechanism
construction and their solutions, and the overall cost and effort invested in each mechanism. HIRAAS is a
good example of evaluating the trade space for risk and performance to minimize development cost and
time.

Instrument Overview

The High resolution Ionospheric Thermospheric Spectrograph is an aeronomy instrument designed to

observe the ionosphere by viewing the earth’s limb from low earth orbit. The overall design of the
instrument did not allow for redundancy, however all possible single point failures were evaluated against
the impact to mission objectives. When feasible, alternative or backup systems were employed. Since
weight was not a critical design parameter, a reduction in cost and development time was achieved by
using 6061-T6 aluminum for most parts. The HITS instrument is shown in Figure 1 without is thermal
blankets. The five separate mechanisms integral to the design of HITS, are; the dust cover door assembly
(DCDA), the gimbal caging mechanism (GCM), the one-axis gimbal, the grating drive mechanism (GDM),
and the detector door mechanism (DDM). The location of each mechanism is shown in Figure 2. The
optical design and the object of interest which is the earth’s limb from LEO, are what drive the need for
having two of the five mechanisms in HITS, these mechanisms are the grating drive and the one axis
gimbal. Both the one-axis gimbal and the grating drive were essential for successful operation and were
designed to operate continuously for the one-year lifetime. The DCDA and the DDM will not be discussed
in this paper however the GCM will be mentioned as it relates to the gimbal structure design.

∗
Praxis Inc. Alexandria, Virginia
**
Atlantic Online - Formerly with Raytheon STX, Lanham MD at time of HITS Design
+
Naval Research Laboratory, Washington D.C.
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn research Center, May 15-17, 2002

NASA/CP—2002-211506 215
 Figure 1. HITS Instrument

Gimbal
Encoder Pillar
DCDA

DDM

Gimbal
Drive Pillar

GCM GDM
Figure 2. HITS Mechanical Component Layout

The Grating Drive Mechanism

The optical design of the HITS instrument is described as an f/5.4 1-meter Rowland Circle Spectrograph
1
with an in-plane Eagle mount. Attached to the spectrograph is 1/8 meter off-axis parabolic telescope .
The HITS instrument is designed to have a passband in the far and extreme ultraviolet (FUV and EUV
respectively) of 500-1500 angstroms with a resolution of > 0.5 angstroms. An understanding of the
Rowland circle spectrograph principles is needed to define the satisfactory performance of the HITS
GDM.

A Rowland circle spectrograph operates under the principle that for a concave grating, minimal
aberrations occur when the entrance slit of the spectrograph and the detector lie on a circle. The Rowland
circle is defined as a circle of a diameter equal to the tangential radius of the grating. At different locations
along this circle, the image of the various wavelengths is focused. HITS as an in-plane Eagle mount

NASA/CP—2002-211506 216
design means that the entrance slit and the detector locations are fixed. In order to position different
wavelengths on the detector, the grating must be rotated. While doing so, the grating must also be
translated such that the entrance slit and the detector both remain on the Rowland circle as can be seen
2
in Figure 3 .

Figure 3. Operation of an Eagle mount Rowland circle Spectrograph

Grating Drive Mechanism Design

The Grating Drive Mechanism (GDM) design directly affects the image quality with regard to focus and
coma aberrations. Therefore the design of the GDM must be individually tailored to the optical system in
which it used, specifically the grating determines the performance requirements and are as follows:

• Provide two degrees of freedom for the diffraction grating, one rotational and one translational.
• Provide 100 mm of total translation with 0.10-mm accuracy.
• Provide 25 degree total rotation with 0.2 degree accuracy

The original design of the GDM shown in Figures 4 and 5 involved the use of a ball screw actuated
translation stage driven by a 14 mN-m (2 oz-in) stepper motor with 100:1 harmonic drive gear reducer.
The translated portion of the stage contained a separate but similar motor and harmonic drive to form the
rotation stage on to which the grating housing was mounted. Supporting the rotation stage were two
guide rails composed of a shaft mounted on a T-bracket. The rotation stage was attached to the guide
rails by four three-quarter encircling linear ball bearings, two on each rail. The design of the rotation stage
allowed the grating to move 5 degrees beyond the required 25 degrees.

Since the relationship between the rotation and translation of the grating determines the image quality, a
Slot and Follower Cam was ruled out due to lack of in flight flexibility. Additionally concerns about
changing performance due to wear and thermal variation emphasized the need for in flight corrections. By
decoupling the individual degrees of freedom in the GDM, an image focusing capability is gained by
translating the grating along the optical path. This was a good decision since the original grating had to be
replaced due to poor optical performance. The original grating ruling density was 5000 lines per mm with
a blaze angle of 834 Angstroms. Due to schedule constraints, a replacement grating could not be
manufactured in time. As a result, a grating with a ruling density of 3600 lines per mm and 1200
Angstroms blaze was installed for flight. If a slot/follower or cam mechanism has been used the
replacement grating would have required modifications.

NASA/CP—2002-211506 217
 Restraint
Lead Screw Notch & Pin Antibacklash
Nut
Stepper
Motor

Limit
Limit Switches
Switch
Grating Mount /
Rotation Stage

Figure 4. GDM Translation Stage Layout Top View

Harmonic
Drive

Stepper
Motor
Figure 5. GDM Rotation Stage Layout Front View

Testing, Failures, and Redesigns

Testing done on the original GDM exposed three flaws that required a redesign. The flaws that were
discovered involved the translation drive assembly, the rotation stage support rails, and the restraint
mechanism. In solving each flaw a solution that involved the minimal amount of reengineering of the
original method of operation was necessary due to schedule constraints. All of these flaws were solved
through the substitution of a modified part in place of the old one.

NASA/CP—2002-211506 218
The original translation stage drive assembly had two issues that were deemed unacceptable during
functional testing. Of these one stemmed from torque margins and the other was due to actuation time.
The original 14 mN-m (2 in-oz) stepper motor with 100:1 harmonic drive gear reducer was easily stalled
and was thought that it might not be able to drive the stage along the rails under less than ideal
circumstances. Even with the high gear reduction through the harmonic drive and the torque multiplication
of the lead screw most of the motor’s output torque was consumed in the operation of the harmonic drive.
The high gear reduction also caused the time to actuate the grating through its limits to be unacceptably
long – 45 minutes for ten centimeters of travel. The reasons for using the original drive mechanism was
due to the cost constraints and that fact that this motor was available as a spare from a similar motor
used on another payload. The solution to the problems mentioned was to replace the stepper / harmonic
drive assembly with a Rapidsyn 282 mN-m (40 in-oz) Motor directly coupled to the lead screw. This motor
was readily available as it was also the flight spare for HITS gimbal drive. Even with the removal of the
harmonic drive the new motor was still able to provide the necessary accuracy due to the small step size
of 1.8 degrees per step further reduced by the 10 thread per inch lead screw.

Vibration tests of the GDM exposed another problem in the rotation stage support rails. In the original
design, the rotation stage was connected to the rails by semi-encircling linear bearing. During vibration
testing in certain orientations the rotation stage exhibited excessive rattling. This was due in part to the
looser than expected tolerances in the bearing and also from the bearing not being able to fully encircle
the shaft they rode on. The amount of play in the system leading to the rattle was only apparent in the
vibration tests. This lead to the conclusion that during the tests the linear bearings may have been slightly
expanding and thus allowing the rotation stage to lift off from the rails by a small amount on one cycle of
the vibration and then back again on the other cycle. To solve the problem, the shaft-mounted rails were
replaced with cylindrical shafts of slightly greater thickness supported only at the ends. The semi-
encircling linear bearings were replaced by fully encircling linear bearings. Originally this idea of
supporting only the ends of the shaft was considered but concerns about vibrations caused it be
dismissed. It was thought that there would be excessive shaking enough to damage the GDM if the
rotation drive moved to the unsupported middle of the guide shaft during launch. Without the center
supports a caging system had to be developed to hold the rotation stage at a supported end of the guide
rail during launch.

Cost reduction was a driver in the GDM concept of operations both during launch and on orbit. To reduce
the complexity and cost of the system the caging scheme was built into the GDM. Rather than use a more
traditional paraffin actuator based system the GDM utilized the extra rotation of 5° beyond the required
25°. To restrain the grating during launch the grating was driven to the limit farthest from the ball screw
drive assembly. Two stainless steel pins in the GDM housing were then extended and the translation
stage was back driven against them. After the pin was engaged the grating was then driven along the
translation stage again to the point where the grating would lock in place. After launch the grating was
moved in the reverse sequence to and the pins would drop down and uncage the grating.

Grating Rotation Drive On-Orbit Performance

During operation the GDM would have to both rotate and translate to enable HITS to acquire the
wavelengths bands of interest as well as focus the image. Rather than use a closed loop control system
requiring the use of encoders or resolvers the GDM relied on an open loop system. Before each actuation
the GDM rotation and translation stages were driven to one limit that was designated as the starting
position. The starting position was also defined through the use of mechanical microswitches and optical
switches as fiducials. From the fiducials, steps were counted as they were sent to the stepper motors.
This scheme has so far proven to be effective in that any errors that have arisen would not have been
prevented by a more detailed knowledge position.

The error that does affect the image quality of HITS is due to cogging in the rotation stage stepper motor.
Most often the positions that are entered into the GDM normally coincide with the motor rotation
increments. However, at certain positions the motor tends to “slip” to the next increment or step. This
difference results in an angular error on the order of a tenth of a degree. The angular error then results in
a shift of the image along the detector on the order of one millimeter. Fortunately, this error is easily
corrected by translating the image back to the correct location during analysis on the ground.

NASA/CP—2002-211506 219
 HITS One-Axis Gimbal

The use of the one-axis gimbal was driven by the vertical size of the region of interest in the Limb, the
desired resolution of the features to be observed, and the field of view of the instrument. The limb region
of interest extends from altitudes of 100-750 km above the earth’s disk. The HITS field of view (FOV) is
0.06 degree in the vertical direction and 4.6 degrees in the horizontal. The FOV translates to spatial
resolution of 3 km X 230 km. To be able to scan across altitudes of 100-750 km a gimbal was needed as
shown in Figure 6. This gimbal rotated the HITS optical bench as well as the Ionospheric Spectroscopy
and Atmospheric Chemistry sensor (ISAAC) shown in Figure 7. ISAAC was another separate, coaxially
aligned instrument mounted onto the side of HITS.

Figure 6. One-axis Gimbal range of motion

Ordinarily a cheaper and simpler approach to scan across a field of regard is to use a scan mirror rather
than scanning a whole instrument. Unfortunately the nature of optics in the FUV and EUV prevent the
addition of another mirror due to the extreme losses each optical element causes. Current loses of 60
percent per reflection are typical in the EUV wavelength ranges of interest. Optimizing the light throughput
for the EUV would reduce the throughput for the FUV to unacceptable levels. The additional third surface
of a scan mirror would result in a 90 percent total attenuation of an already weak signal. Furthermore, by
using a gimbal the overall time for instrument development was reduced in that both the optical
subsystem and gimbal subsystem could proceed independently in both construction and calibration. A
scan mirror would have prevented the instrument from being calibrated until the integration of the scan
mirror. Additionally, more calibration steps would be needed to determine the effect of the scan mirror
angle on the optical performance. Being able to use the HITS gimbaled platform also shortened the
development and cost of the ISAAC instrument. The gimbaled HITS optical bench prevented ISAAC from
needing another mechanism to scan across its field of regard and once again allowed for independent
optical system development and calibration.

NASA/CP—2002-211506 220
 Figure 7. ISAAC Instrument mounted on HITS gimbaled optical bench

This approach of using one scan platform for several instruments was also used in another instrument
built by NRL which was known as the Remote Atmospheric and Ionospheric Detection System (RAIDS).
RAIDS contained eight individual sensors mounted on a common frame attached to a one-axis gimbal.

Gimbal Design
The HITS gimbal has several unique qualities flowing from the constraints upon it. The wide optical path
of the spectrograph negated the use of a shaft mounting the camera to the gimbal. The envelope
constraint on the instrument required the gimbal rotation axis to be offset from the camera center of
gravity. Also, budget constraints made the use the non-flight qualified parts an attractive alternative. In
designing the gimbal the following requirements had to be satisfied:

• Provide a singe axis of rotation for the instrument optical bench

• Provide a 17 degree nominal scan from -10 to -27 degrees with respect to the spacecraft horizon
being equal to 0 degrees
• Provide a 30 degree total scan from -10 to -40 degrees with respect to the spacecraft horizon being
equal to 0 degrees.
• Provide a 0.1 degree scan window.
• Perform a nominal scan of the HITS optical bench in 90 seconds (0.189 degree/sec)
• Return the HITS optical bench assembly to the starting position in 15 seconds (1.13 degrees/sec)

The HITS optical bench and ISAAC had a resultant gimbaled mass of 36.8 kg and an inertia of 7.45 kg-
2
m . The gimbaled mass center of gravity was offset from the rotation axis by 27.3 cm. Minimizing the
transfer of uncompensated angular momentum from the gimbal operation was not an issue in gimbal
design. Communication with the ARGOS spacecraft designers during spacecraft design resulted in the
spacecraft being constructed with momentum wheels sized to counteract the expected transfer of angular
momentum.

The general design of the HITS gimbal involved supporting the optical bench on two sides by two
separate support pillars mounted to a common baseplate. The drive system and position feedback
system were mounted inside opposite pillars and supported independently from the main gimbal bearings
in order to decouple the load path from the drive system.

Drive System
The main gimbal bearings consisted of 125-mm inner diameter crossed roller bearings from the
Messinger Bearing Corporation. Crossed roller bearings were chosen due to the volume constraints that
would have been violated in providing for the preload springs in more traditional ball bearings. However,

NASA/CP—2002-211506 221
using these roller bearings increased the amount of torque delivered by the drive system due to the
higher friction as compared to ball bearings. The large size of the bearings was chosen so that it would be
able to fit around much of the drive system as shown in Figure 8.

Gimbal Roller Motor Harmonic Crossed Roller

Bearing Mount Drive Bearing
Retainers

Input
Shaft
Stepper
Motor

Figure 8. HITS Gimbal Drive Assembly Component Layout

The drive system, shown in exploded view in Figure 9, consisted of a stepper motor the same as the one
used in the GDM. The stepper motor output was coupled to an input shaft by a spur gear. This provided a
first stage of gear reduction by 7:1. The input shaft was supported by its own set of bearings and was
directly coupled to the wave generator of a HD Systems harmonic drive. The harmonic drive had a gear
reduction of 160:1 and was mounted with the circular spline as the stationary component. The flexspline
output was directly coupled to the gimbal output hub by a clamp ring that also contained one of the input
shaft bearings. The gimbal mount was then bolted on to the output hub. A harmonic drive was chosen for
this gimbal due to its simplicity, its higher load rating as compared to a planetary gearbox of the same
physical size, and its feature of zero backlash. Although there is a minor source of backlash in the stepper
motor to input shaft coupling, its affect after the 160:1 reduction is negligible as far as positional
3
accuracy .

NASA/CP—2002-211506 222
 Optical Bench Gimbal Gimbal Bearing
Mount Bearing Retainer HD Circular
Spline
Output
Hub HD
Wavegenerator
Harmonic Drive Motor / HD
(HD) Flexspline Mount

HD Clamp
Ring

Shaft Bearing
Stepper
Motor
Gimbal Support
Cover
Input Shaft
Figure 9. HITS Gimbal Drive to Structure Component Layout

Encoder System
Position sensing in the Gimbal was accomplished by using a vacuum rated encoder from BEI
Technologies. Inc. The encoder was mounted independently from the drive system inside the opposite
support as shown in Figure 10. Motion of the gimbaled optical bench is passed to encoder though a
flexible shaft coupling. This direct coupling negates any errors in positional accuracy from drive system
losses. A source of error that this encoder mounting system does not account for is due to distortions
from thermal expansion.

Gimbal Shaft
Mount Coupling

Encoder
Encoder
Mount

Figure 10. HITS Gimbal Encoder Assembly Component Layout

NASA/CP—2002-211506 223
These errors may arise from differential heating of the separate support pillars and from any distortion
caused by the mounting of the steel roller bearings inside the aluminum supports. Fortunately, these
errors are within the allowable specifications for instrument accuracy.

Structure
The optical path inside the HITS instrument prevented the use of a shaft for the gimbal, as it would have
interfered with light traveling from the grating to the detector. Rather than mount the gimbal to the
instrument cowling, a box design was chosen as the main structural support linking the optical bench to
the gimbal assembly as shown in Figure 11. This gimbal anchor box would provide the structural support
3
necessary while not interfering with the operation of the instrument .

Figure 11. Original HITS Camera Structural Layout

Unfortunately, results from further analysis determined that an anchor box design would still have caused
interference in the optical path. Also being able to attain the stiffness required in the anchor box would
prevent the use of aluminum which was the material used for the optical bench. This mismatch in material
would have caused complications in mating to prevent any thermal distortions from affecting the
instrument. Weight reduction became less of a driving factor as time progressed. This reduced emphasis
on weight reduction allowed for the possibility strengthening of the optical bench to be a load bearing
structure. It was decided through the use of finite element analysis that increasing the density of support
ribs as well as the skin thickness near the gimbal mount area (Figure 12) would be satisfactory for HITS
to survive launch. Mounting the gimbal to the optical bench also served to simplify the design by reducing
the number of parts. It also simplified instrument assembly and integration since the development of the
optical bench assembly was decoupled from the development of the gimbal assembly and both could be
then connected or disconnected when required with minimal effort.

NASA/CP—2002-211506 224
 Gimbal
Mount

Increased Density
of Structural Ribs

Figure 12. Optical Bench Strengthening for Gimbal Mounting

Gimbal Caging Mechanism

Affecting the design of the HITS gimbal structure was the gimbal caging mechanism. The GCM was
needed for several reasons. The additional load paths were created through the GCM allowed the gimbal
to be scaled down in size. In addition, the bending moment created by the distance between the center of
mass and rotation axis, as well as the estimated force vectors, prevented the caging mechanism from
being part of the gimbal support structure as is normally done in a more balanced gimbal assembly. The
caging system was mounted to the instrument base plate so that the CM when superimposed on the base
plate would be in the center of a triangle formed by the gimbal supports on the base side and the caging
mechanism at the vertex.

The caging mechanism shown in Figure 13 was a paraffin actuated locking T-Bar design with manual
reset made by Starsys Inc. Modifications to an already existing design were undertaken to account for the
higher preload. These modifications involved changing the Bellville preload washers as well as
developing a custom tool to allow the T-bar to be extended for reset. Even with the higher preload the
paraffin actuator still operated with a release force margin of 300%. Due to the location of the caging
mechanism, resetting was done in a semi-blind fashion requiring the tool to be applied while the optical
bench was moved into place.

Figure 13. HITS Gimbal Caging Mechanism

NASA/CP—2002-211506 225
Use of Non Space Qualified Parts
As a cost savings measure, non-flight qualified parts were used based on expected lifetime compared,
availability schedule, and the impact to other components in the event of a failure. The non-flight qualified
components that were used in the HITS gimbal were the following: the encoder, the optocoupler LED, the
harmonic drive, and the crossed roller bearings. Each component was determined to be suitable for
different reasons.

As earlier mentioned the encoder was vacuum rated. Outgassing was a primary concern for HITS and the
vacuum rating insured no detrimental outgassing from the encoder. However outgassing is only one
factor in the determination that this encoder would work. The differences between a vacuum rated
encoder and a flight-qualified encoder relate to the parts construction for withstanding vibration and more
importantly for surviving in the radiation environment of space. Construction data from the manufacturer
concerning operational vibrations along with a preliminary analysis of launch vibrations determined that
the encoder would be compatible with the instrument. Expected instrument temperatures were also within
the operational range of the encoder. The main fear in using the encoder was the lifetime of the encoder
LED due to radiation degradation. Once a again a good knowledge of the expected radiation environment
along with manufacturer’s data regarding the LED showed that the encoder would last for the required
one year of operation with a moderate possibility of lasting up to the goal of three years. The survivability
of the encoder was enhanced by being shielded inside the encoder’s steel canister as well as by an
aluminum outer cover. In the event the encoder should fail redundancy was built into the gimbal by
counting the steps input to the motor from fiducials as shown in Figure 14. The fiducials take the form of
microswitches and an optocoupler used as an optical switch. The optocoupler was non-flight rated but it
was a mil-spec part. Just as with the encoder, data from the manufacturer was used to determine its
suitability for use. Cooperation from the manufacturer also aided in integrating the harmonic drive.

Optocoupler

Microswitches

Figure 14. Gimbal Position Sensing Redundancy

The Harmonic Drive Systems Inc. HIUC-25 series harmonic drive, was chosen from its rated step size,
torque multiplication, physical dimensions, and availability. A flight qualified drive differs from a non-flight
qualified drive in the materials used for the wave generator bearing retainer, and flexspline; it also differs
in the recommended method of lubrication. Adapting this component for use in HITS involved redressing
issues in outgassing and lubrication. Outgassing requirements involved replacing the bearing retainer
from one made from nylon to one made from phenolic. Testing was done to determining if the drive would
meet lifetime requirements. Ordinarily, a circulating oil bath is the preferred method for lubrication of this
drive. The impracticalities of providing an oil bath as well as outgassing requirements drove the decision
to use Braycoat grease as a lubricant. Concerns by the vendor regarding the use of Braycoat were
addressed by performing a lifecycle test. The test produced a lifecycle estimate that was greater than the
expected number of cycles for mission life. The difference in flexspline construction between flight and
non-flight drives was also addressed by the lifecycle test.

NASA/CP—2002-211506 226
Expediency and size were the main drivers in the selection of the crossed roller bearing used. These
bearings were originally intended for industrial use and did not have as high a precision as the ABEC 7 or
ABEC 9 bearings normally used for flight. The tolerances within the bearing would still keep the
instrument within its pointing systems budget. Before integration into the gimbal, the bearings were first
cleaned of the residual machining oils with ethyl alcohol in an ultrasonic bath. Braycoat grease was added
to lubricate the bearings during use. The concern about pressure breakdown of Braycoat was negated by
the large size of the bearings for the expected load. In addition, the low angular velocity would not cause
detrimental heating of the bearing area. Unlike the previous components that were justified though
analysis and test, the use of the roller bearing was justified through extremely high margins.

Conclusion

The mechanisms used for HITS were designed to support the operational performance requirements of
the instrument by providing the movement necessary for certain optical components and optical systems.
Cost and expediency became more of a major driver as the instrument developed while the more
traditional drivers as weight, and momentum reduced in importance. Products designed for industrial use
on earth should not be ruled out when cost and expediency is a major driver in instrument development.
Good results can be achieved if the deficiencies in non-space flight qualified products are recognized and
amended. Although various problems appeared in certain HITS mechanisms through the development
process none were related to the use of these non-space flight qualified components. Solving the
problems in these mechanisms did not involve using cutting edge technology as much as using already
existing technology in novel ways.

Acknowledgements

The HITS instrument development was supported by the SERDP program and Navy funding. The authors
would also like to thank Stephen Lockwood for Thermal Systems design, the Aerospace Corporation for
analysis support and Don Woods for design support of the HITS instrument.

References

1. Dymond, K.F., Wolfram K.D., Budzien S.A., Fortna, C.B. and McCoy R.P. “The High Resolution
Ionospheric and Thermospheric Spectrograph (HITS) on the Advanced Research and Global
Observing Satellite (ARGOS): Quick Look Results” Proceedings of SPIE, vol. 3818, 137-148
(1999)

2. Samson, James A., Ederer, David L. Vacuum Ultraviolet Spectroscopy. Academic Press (2000)

3. HIRAAS CDR (NRL and STP internal document)

4. HIRAAS PDR (NRL and STP internal document)

NASA/CP—2002-211506 227
 Design of an EVA-Capable Four-Point Restraint System

Armond Asadurian∗, Jim Hammond∗, Kazuo Hamada∗∗, Hiroshi Irikado∗∗,

Michio Ito+, Miho Endo+

Abstract

Launch Lock Mechanism (LLM) is designed to provide four-point restraint for a large Extendible Mast on
the Japanese Experimental Module Space Environment Data Acquisition- Attached Payload (JEM SEDA-
AP). The mechanism provides misalignment-tolerant spring-loaded latching at all four corners of the
square instrument package, using a single actuation point. Provision is made for override of the
electromechanical driver by means of EVA (Extra Vehicular Activity). The system is equipped with a
Backup Latching Mechanism (BLM) to ensure positive restraint during launch and reentry. Visual and
electrical indicators are integrated throughout the system for latching and unlatching demonstrations and
verification.

Introduction

This paper will present an efficient method of achieving four–point spring-loaded latching from a central
actuation point, while retaining full functionality with EVA operation. The principal restraining force
generated at each latching point is traced back to the main actuation point, where a Moog Type 2 stepper
motor actuator provides drive torque. This actuator has unpowered holding torque sufficient to sustain
each of the four latching points in a secure latched condition, with significant margin.

At a joint safety review board held between NASA, NASDA, NEC TOSHIBA Space Systems, and Moog
Inc., it was determined that the unpowered holding force of the mechanism, which relies on the magnetic
holding torque of the permanent magnet motor, is not considered reliable, despite the large force margin.
This decision was influenced partly by considerations of manned launch vehicle safety. To mitigate this
condition and to enhance overall safety of the mechanism during launch, an additional EVA-operated
Backup Latching Mechanism (BLM) was integrated with the LLM. The BLM is capable independently of
locking the overall mechanism and isolating the actuator from the rest of the mechanism.

Four-Point Restraint System

The four-point latching mechanism was designed to provide the restraint feature for the Japanese space
station module of SEDA-AP Extension Mechanism Assembly. SEDA-AP consists of JEM attached
payload main structure, JEM Payload Interface Unit, Payload Attached Mechanism – Payload Unit,
Grapple fixture for RMS handling, Bus electrical equipment and miscellaneous monitoring sensors
including Newtron monitor sensor (NEM-S) at the edge of the extension mast.

The platform, on which the above-mentioned instruments are mounted, is referred to hereafter as the
NEM-S plate. The NEM-S plate is designed to be latched in four corners using the LLM. On orbit, the
NEM-S plate is extended out away from the SEDA-AP approximately 1 meter on an extendable boom.
The monitoring instruments are then activated. Retraction of the NEM-S plate into the SEDA-AP main
structure is followed by the LLM actuation of the LLM, resulting in a latched condition.

∗
Moog Chatsworth Operations, Chatsworth, CA
∗∗
NEC TOSHIBA Space Systems, Japan
+
NASDA, Japan
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 229
 SEDA-AP

Figure 1. SEDA-AP Module

Figure 2. Launch Lock Mechanism

NASA/CP—2002-211506 230
The LLM is designed to provide synchronized latching of all four corners using a single stepper motor
actuator. The actuator design is based on a standard Moog Inc. Type 2 design, which is a permanent
magnet stepper motor of 2 degree step angle, combined with a 100:1 ratio harmonic drive gear reducer.
The main gearbox located at the output of the actuator is designed with a 1:2 (speed-up) gear ratio. It has
two output shafts, converting the actuator output into two separate rotating outputs. Double universal
joints at the ends of each shaft accommodate potential misalignment between the actuator/gearbox and
the two latch blocks.

The latch blocks are the two main assemblies of the LLM, which engage the NEM. Each is designed to
house a secondary gearbox, spring load assembly, ballscrew, and the latching mechanism.

Actuator and EVA Gearbox Assembly

The actuator is a permanent magnet stepper motor of 2-degree step angle with a 100:1 gear ratio
harmonic drive for the output drive. This actuator is equipped with an EVA drive input as a redundant
operating means in case of actuator power failure. The EVA input is through a standard 11-mm
hexagonal bolt head (standard EVA tool interface) on the exterior surface of the SEDA-AP module.

The EVA hex bolt that interfaces with the EVA tool is on the end of the aping-loaded gearbox input shaft.
Inside the gearbox, a bevel gear is mounted on the shaft. Axial force on the shaft engages the bevel gear
with its mating gear. The mating gear is mounted on the actuator motor shaft. The force required to
engage these two gears is 1.36 kg, which is compliant with the NASA EVA operating requirements. EVA
operation is enabled when sufficient force is applied to the EVA bolt to engage the gears. Operation of the
tool then rotates the input shaft, and the motor shaft. This results in the operation of the actuator without
the need to energize the actuator winding. The 5:1 gear ratio of the EVA gear head produces a rotor
speed that is 5 times higher than the input speed at the EVA tool interface. This allows EVA actuation of
the latching and unlatching function within the 10 minute time limit set by NASA EVA operation
guidelines.

Figure 3. Actuator Assembly and EVA Gears

NASA/CP—2002-211506 231
Main Gearbox
The main gearbox input comes directly from the Type 2 actuator output. The input member is a bevel
gear, which drives two output gears on shafts located at 120° apart. The output shafts are connected to
the latch blocks through the articulated drive shafts. The main gearbox has provisions for input from the
Backup Latching Mechanism (BLM) which will be covered in a separate section. The ratio of the input
gear to the driver gears is 1:2 (speed up). The gear shafts all rotate on ball bearings for maximum
efficiency.

Figure 4. Main Gearbox

Connecting Shafts
The connecting shafts make the drive connection between the main gearbox and the latch blocks. The
connecting shafts are designed with double universal joints, to accommodate manufacturing tolerances
and assembly misalignments.

Figure 5. Connecting Shafts with Universal Joints

NASA/CP—2002-211506 232
Latch Block Assemblies
The LLM includes two latch block assemblies. Each latch block assembly implements two of the four
latching points. Each latch block assembly consists of a secondary gearbox, ballscrew assembly, the
preload spring assembly, locking arms, Hall effect sensors, the visual indicator assembly, and the locking
pins. The locking pins interface with the strike assemblies, which are mounted on the NEMS instrument.

Figure 6. Latch Block Assembly Kinematics

NASA/CP—2002-211506 233
 Figure 7. Latch Block Assembly

Secondary Gearbox:

Each secondary gearbox is driven by one of the connecting shafts. The secondary gearbox provides a
rotational change of direction from the connecting shafts to a ballscrew located orthogonal to the
connecting shaft axis of rotation. The speed ration between these two drive components is 1:1 therefore a
1:1 set of miter gear is used. The gears are mounted on ball bearings for maximum efficiency and
accuracy.

Rotation of the ballscrew by the output of the secondary gearbox produces linear movement of the mating
ball nut assembly, required for actuation of the locking arms and for producing the spring preload needed
for the latching operation.

Ballscrew/Ballnut Assembly:

The ballscrew/ballnut assembly consists of a 15.9-mm diameter shaft with a thread pitch of 3.175 mm and
a mating nut. The screw is made of 440C stainless steel, and is dry lubricated with tungsten disulfide
coating. The nut assembly is a re-circulating ball design, with an internal crossover track to minimize the
profile of the assembly.

Preload Spring Assembly:

The preload spring assembly is attached to the ballnut nut assembly through a set of Belleville springs.
The nut housing is keyed to the preload spring assembly housing, preventing rotation of the nut housing
in the preload spring assembly housing. This translates the rotational movement of the ballscrew into
linear movement of the ballscrew nut within the preload spring housing. The preload spring assembly
consists of 13 Belleville springs assembled in series. The large force-versus-deflection ratio of these

NASA/CP—2002-211506 234
springs provides the necessary spring force within a limited travel range. The preload spring assembly
housing is attached to the two locking arms through two hinge points. The hinges are designed with
Nitronic 60 bushings and A-286 stainless steel pins. These two components are dry lubricated, with the
material selection based on the anti-galling characteristics of these materials under the high contact
stresses existing at the hinge point

Figure 8. Belleville Spring Assembly

Figure 9. Preload Spring Housing Assembly

NASA/CP—2002-211506 235
The Belleville Springs were measured and force versus deflection plots was prepared to establish the
required deflection for a given force. These measurements were taken four separate times to determine
repeatability of the springs. The plot below demonstrates the first and last measurement and the resulting
variation are negligible for this application.

Locking Arms:

The geometry of the locking arms yields an 8.55:1 mechanical advantage from the actuation hinge points
on the preload spring housing assembly to the contact points at the latching locations. Each locking arm
is made of titanium, and its cross section is that of an I-beam, for structural efficiency and in order to
minimize the deflection of this member under load.

Hall Effect Sensors:

Hall effect device (HED) sensors are provided to sense the required preloaded latched and unlatched
positions. The preloaded latched condition is sensed by an HED excited by a magnet mounted directly on
the ballnut assembly. As the mechanism output approaches its predetermined preloaded latching
position, and the lock arms engage the strike assemblies, the magnet activates the HED, an electrical
signal is sent to the actuator, and the power to the actuator is cutoff. The HEDs are adjusted to stop the
rotation of the actuator once the preload is established.

The unlatch HED is also adjusted to switch off the actuator once the ballnut is fully retracted and the
mechanism is fully unlatched.

NASA/CP—2002-211506 236
 Figure 10. Hall effect Sensor

Visual indicators:

Visual indicators are provided for visual indication of the mechanism status in latched and unlatched
positions. The visual indicators are spring loaded to their retracted positions, indicating the unlatched
condition. The same bracket that holds the magnets required for the HEDs is also designed to activate
the visual indicator mechanism. The visual indicators are extended against the spring force as the
mechanism is driven into the latched position. The extension of the visual indicators becomes visible at
the outside of the JEM module, serving as a visual confirmation of the latch/unlatched status of the
mechanism to the EVA astronaut.

Locking Pins:

Locking pins are provided at each of the four corners of the SEDA-AP for each of the latching points.
These pins are recessed inside the latch block assemblies to avoid any possible contact with astronaut
space suits during deployment of the NEMS plate. The pins are made of A-286 material and are dry
lubricated with tungsten disulfide to prevent galling during engagement with the receptacles located on
the strike assembly. The pins are the main load bearing components of the latching mechanism in Y and
Z axes in the launch environment.

Strike Assembly
The strike assemblies are located at the corners of the LLM and are mounted on the NEMS plate. Each
strike has one main roller which is the contact point of the locking arm with the strike assembly and which
provides the constraint of the latching mechanism in X axis. Three other rollers are provided on each
latching assembly. These rollers are designed to correct possible misalignment of the NEMS plate during
the retraction operation. The latch blocks are equipped with large ramps to capture and guide the rollers
into an optimum latching position and to engage the locking pins and the receptacles. The locking arms
are then activated and the main roller is captured and the appropriate preload is applied which is reacted
through the locking pins into the latch blocks.

NASA/CP—2002-211506 237
Adjustment of the receptacle to pin engagement individually is made at the strike assembly by shimming.
Shimming each of the strike assemblies produces a uniform preload on each of the latching points. Strain
gauges are mounted on the strike assemblies to verify preload by material strain at final assembly.

Figure 11. Locking Pin Cross section

Backup Latching Mechanism

The backup latching Mechanism (BLM) was added to the LLM as a result of the safety review meeting
held with NASA, NASDA, NEC and Moog Inc. This device is designed to be an integral part of the LLM.
The BLM provides an additional latching means to ensure that the LLM mechanism will not lose preload
and will not disengage during launch or reentry. In the initial design, the actuator magnetic detent torque
was the primary the only holding force for the LLM. This was deemed unreliable due to its dependency on
the motor magnetic force. Further discussion of this issue will be found in the Lessons Learned section.

The BLM is designed to be operated via EVA and is equipped with a visual indicator external to the JEM
module and visible by astronauts. The BLM is also capable of locking the LLM regardless of the actuator
output position at the time full preload is achieved.

BLM operation is based on the engagement of two ratchet gears. One of the ratchet gears is directly
mounted on the output of the actuator, and is free to rotate with the actuator output. The mating gear is
the locking gear, which is rotationally rigid and is designed to engage with the dynamic gear to produce
the backup latching, thereby isolating the actuator from the load path.

The non-rotating locking ratchet gear is mounted on a mechanism that consists of a rotating cam, rollers,
base plate, wave spring, and actuating arms. The actuating arms of this mechanism are also designed to
operate as the visual indicators. This mechanism is located inside the main gearbox, with the actuating
rods designed to protrude through the SEDA-AP module wall so that they are available for EVA
operation.

There are two actuating arms. One is used to lock the mechanism by being depressed down; and the
other unlocks the mechanism by also being depressed down. These two arms are attached internally so
that when one is depressed down the other rises and vice versa. The state of the BLM mechanism is
therefore continuously indicated.

NASA/CP—2002-211506 238
These two actuation arms are directly attached to a cam which is centrally located in the main gearbox
and is designed to rotate at the center point which is above the center of rotation of the actuator and the
dynamic gear. The linear movement of the actuating arms translates into rotating motion of the cam. The
locking gear, which is equipped with rollers, is directly seated on the cam. The cam is designed with
ramps, and rotation of the cam results in linear movement of the locking ratchet gear as the rollers are
forced up the ramps. This linear movement of the locking cam is used to engage and disengage the
locking gear from the dynamic gear.

Design and sizing of the ratchet gear tooth is based on the worst case unlatching loads. To provide
infinite locking resolution, the dynamic gear is designed with one full tooth of free play. The dynamic gear
is centered to a fixed position by a compression spring. This feature ensures that the engagement
requires no more than one tooth deflection. Therefore the engagement of the two gears can be
accomplished at any position.

NASA/CP—2002-211506 239
 BLM

Actuator

EVA override

Figure 12. Backup Latching Mechanism

NASA/CP—2002-211506 240
 Lessons Learned

• Designing space hardware to be transported to space in manned vehicles such as the Space Shuttle,
and to support manned missions such as the space station requires a new approach. The fracture
criticality of materials is highly scrutinized, and ergonomic factors related to human engineering
become critical in the design of the product. NASA’s strict guidelines, while ensuring that the manned
space activities are safe, efficient, and can easily leave the effect of driving the design.

An offsetting factor is the fact that EVA-related requirements can also offer benefits:

Design of unmanned space mechanisms traditionally has been dictated by performance

requirements, reliability, weight and space availability. Human engineering has not been a driving
factor in such designs. Mechanisms such as the LLM would be adequately defined by similar
guidelines; however, the requirement for the EVA mechanisms allows for some of the redundant
mechanism normally used in unmanned systems to give way to other requirements such as
ergonomics and human factor engineering.

• Safety of the mechanism for manned flight is of utmost importance when the payload is reviewed for
security during launch. An unsafe payload is considered a direct threat to the safety of the launch
vehicle and its crew.

The magnetic holding force of an actuator, regardless of the demonstrated design margin, is not
considered a reliable means of securing payloads within the manned space vehicle. A mechanically
positive engagement is also required as a backup for added safety.

NASA/CP—2002-211506 241
 Figure 13. Final Assembly and Testing of Extension Mechanism Assembly of SEDA-AP

NASA/CP—2002-211506 242
 Intricacies of Using Kevlar Cord and Thermal Knives in a Deployable Release System:
Issues and Solutions
* *
Alphonso C. Stewart and Jason H. Hair

Abstract

The utilization of Kevlar cord and thermal knives in a deployable release system produces a number of
issues that must be addressed in the design of the system. This paper proposes design considerations
that minimize the major issues, thermal knife failure, Kevlar cord relaxation, and the measurement of the
cord tension. Design practices can minimize the potential for thermal knife laminate and element damage
that result in failure of the knife. A process for in-situ inspection of the knife with resistance, rather than
continuity, checks and 10x zoom optical imaging can detect damaged knives. Tests allow the
characterization of the behavior of the particular Kevlar cord in use and the development of specific pre-
stretching techniques and initial tension values needed to meet requirements. A new method can
accurately measure the tension of the Kevlar cord using a guitar tuner, because more conventional
methods do not apply to arimid cords such as Kevlar.

Introduction

The Microwave Anisotropy Probe (MAP) Spacecraft utilized a novel Solar Array Restraint and Release
System (SARRS) design that featured a Kevlar cord and thermal knives as the primary solar panel
restraint and release components. The 7.6-m (300-in) Kevlar cord encircled the spacecraft to secure the
solar panels in their stowed configuration for launch. Once in orbit, one of two redundantly configured
thermal knives severed the Kevlar cord and permitted the panels to deploy.

A number of issues arose during the SARRS development involving the thermal knives, Kevlar cord
behavior, and the measurement of the tension in the Kevlar cord. The issues encountered and their
solutions will be discussed, including a process for examining the thermal knives after each use, a
procedure for characterizing the Kevlar cord behavior in different environments, and a method for
measuring the tension in the cord using a guitar tuner. The solutions are presented in a general manner
such that the information can be applied to other configurations of Kevlar cord and thermal knives. The
discussion is preceded by a brief introduction of the MAP Spacecraft and the SARRS Configuration.

SARRS and MAP Spacecraft Configuration

The purpose of the MAP mission is to perform a full sky scan of the cosmic microwave background in
order to study the origin of the Universe. MAP was designed, fabricated, and tested at NASA’s Goddard
Space Flight Center as part of the Medium Class Explorers program. MAP was launched in to low earth
orbit by a Delta II 7425-10 launch vehicle from the Eastern Range on June 30, 2001. After separation
from the launch vehicle, the solar arrays and sun shield were deployed and the spacecraft continued on
to orbit about the L2 Lagrange point.

Spacecraft Configuration
The MAP spacecraft uses a passively cooled microwave differencing assembly to measure the full sky
cosmic background at a temperature of 2.7 K. Since the microwave instrument must be kept very cold,
the solar arrays have been configured to form part of the sunshield that will always shade the instrument
from the sun. The microwave instrument is mounted on top of a hexagonal spacecraft bus and the
protective solar array and sun shield combination is mounted to the bottom of the bus, as shown in Figure
1.

*
NASA Goddard Space Flight Center, Greenbelt, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 243
 Microwave
Instrument

Electronics
Area
Solar Panel (6)

Sun
Shield

Figure 1. MAP Spacecraft Deployed Configuration

The sun shield deploys to a diameter of 5.1 m (200 in) from a diameter of 2.7 m (108 in) in the stowed
configuration, which is shown in Figure 2. The entire spacecraft has a mass of 840 kg (1,850 lb).

Restrain System Requirements

The primary objective of the SARRS was to secure the solar panels and sun shield in the stowed
configuration for launch and ground transportation, but all of the main design requirements were met:

1. Consume less than 20 watts of power in less than 150 seconds per activation;
2. Allow the spacecraft to remain inside the launch fairing for up to 45 days without servicing;
3. Release the solar panels for deployment within 150 seconds;
4. Design mass less than 4 kg; and,
5. Complete SARRS development within schedule.

SARRS Configuration: Kevlar Cord and Thermal Knife

The MAP SARRS was designed to meet all the requirements and address other issues such as actuator
cost, delivery schedule, ability to test the flight components, access to restraint components during
stowage, and risk of potential solar cell damage during deployment. The restraint portion of the SARRS
1
consists of one 3-mm (0.12-in) diameter Kevlar cord assembly that encircles the six solar panels by
resting on twelve cord standoffs. The release portion is composed of two, one primary and one
2
redundant, thermal knives . The complete SARRS configuration is shown in Figure 2.

1
Type 72, Ashaway Line and Twine Manufacturing, Co.
2
Model R09686-407, Fokker Space, Sa.

NASA/CP—2002-211506 244
 Sun
Shield

Turnbuckle
Location (2)

Cord Standoff
Location (12)

Solar Panel, Thermal

Cell Side Knife with
Kevlar Cord mount
location (2)

Figure 2. SARRS Configuration (Instrument not shown)

Two cord standoffs, shown in detail in Figure 3, are positioned on the outer edges of each solar panel.
The standoffs position the cord 33 mm (1.3 in) above the solar cells in order to provide enough clearance
in front of the solar cells such that the cord will not contact and damage the cells during release.

Solar
Cell Area

Figure 3. Cord Standoff with Kevlar Cord

There are two thermal knives in the SARRS. Each knife is held in a mount, shown in Figure 4, which is
attached to one of two solar panels on opposite sides of the spacecraft. The cord standoffs and mounts
are positioned such that the cord remains in contact with the thermal knife heater element, which is
pressed towards the Kevlar cord by a spring within the thermal knife component.

Thermal Knife
Mount

Kevlar
Heater Cord
Element

Figure 4. Thermal Knife in Mount

To deploy the solar panels, power is applied to either of the two thermal knives, which generate
temperatures in excess of 1000 ºC at the tip of their respective heating elements. The thermal knife then

NASA/CP—2002-211506 245
begins to sever the Kevlar cord by melting through the fibers, degrading the tensile strength of the Kevlar
cord. The degradation continues until cord’s strain energy suddenly breaks the remaining fibers, allowing
the cord to fly free of the spacecraft, releasing the panels.

Kevlar Cord Configuration

The cord assembly is constructed of two Kevlar cord sections, each 3.8 m (150 in) in length, that are
joined by two turnbuckles made from titanium, as shown in Figure 5. The turnbuckles are used to
increase the cord tension during installation, as the Kevlar cord requires a large strain, about 300 mm (12
in) to achieve the proper tension. The Kevlar cord is constructed of woven Kevlar fibers that are at various
angles with respect to the tension load. When tension is applied, the fibers try to align themselves along
the load direction and extend the cord length. However, it was found that pre-stretching the Kevlar cord
would increase the cord stiffness, thus requiring a shorter turnbuckle to achieve the same load. After
assembly, the cord was pre-stretched by cycling the cord with a 157 N (700 lb) load 10 times for 3 min
each time. This process is discussed in more detail in the Kevlar Cord Behavior section below.

The turnbuckles are located between the solar panels in two locations 180 degrees apart and 90 degrees
away from the thermal knives, as shown in Figure 2. The advantages of this configuration are a more
distributed cord load during tensioning and the mitigation of potential solar cell damage during cord
release.

Turnbuckle with
Threaded

Threaded Clevis
Attachment Kevlar

Figure 5. Kevlar Cord with Turnbuckles

In addition to generating large strain, the Kevlar cord weave pattern generates a compressive force
towards the cord center when tension load is applied. This behavior enables the Kevlar cord to be
attached to the turnbuckle with a simple but effective loop that relies entirely on internal frictional forces
from the braided pattern, as shown in Figure 6. This design allows a cord assembly to develop its
maximum breaking strength characteristics by avoiding strength reducing knots.

Compressive forces
generated from tension
loading holds cord end in

Threaded Clevis

Tension Load

Figure 6. Kevlar to Turnbuckle Attachment

Kevlar Cord Release Mechanism: Thermal Knife

The thermal knife was originally developed by Fokker Space and has successfully been used on
numerous flight programs. The thermal knife component was selected as part of the SARRS during the
initial design phase. It was also decided that there would be no modifications made to the thermal knife

NASA/CP—2002-211506 246
because of its successful flight heritage. However, the differences in the knife’s utilization in the SARRS
would require an extensive development and test program.

Thermal Knife Description

The thermal knife (T/K) consists of a ceramic substrate heater element, 8x10x 0.7 mm (0.3x0.4x0.03 in),
with dual electrical resistance trace patterns. The trace patterns heat the substrate to temperatures in
excess of 1000 ºC. Two pins support the heater element and provide the electrical power to the parallel
trace patterns. The pins and heater element are held in a cylindrical housing and a spring inside the
housing preloads the heater element to push it against the Kevlar cord during the cutting process.

Figure 7. T/K Heater Figure 8. Thermal Knife

Element with Trace

T/K Use
The MAP SARRS uses a redundant thermal knife configuration that is different than previous thermal
knife release system designs. This design allows either knife to operate and release the system. In
addition, the operation of one knife will not damage the redundant knife and a failure (electrical or
mechanical) will not interfere with operation of the other knife.

Trace Damage
The thermal knife trace material becomes soft when the substrate is heated and is more susceptible to
damage during this period. It is important to maintain minimum contact between the Kevlar cord and trace
during the cutting process. For the MAP SARRS, minimum contact is achieved by using a minimum
diameter Kevlar cord, large strain value, and a continuous 90-degree angular contact configuration
between the cord and heater element.

The SARRS Kevlar cord is less than 3 mm (0.12 in) in diameter. This relatively small diameter combined
with large strain allows the Kevlar fibers to pull away from the trace during the cutting process, as shown
in Figure 9. On the contrary, a smaller strain value and larger cord diameter causes the severed fiber
ends to adhere to the trace during activation and in some instances pull the trace off the substrate when
the Kevlar cord is completely severed and separates, as shown in Figure 10. In this instance, the Kevlar
cutting process would not be affected during the initial trace damage, however, all subsequent cuts would
be affected by an already damaged trace.

Figure 9. T/K Element and Small Diameter Figure 10. T/K Element and Large Diameter
Kevlar with Large Strain Kevlar (MAP design) Kevlar with Low Strain

NASA/CP—2002-211506 247
In the MAP SARRS design, the Kevlar cord maintains a near 90-degree angle with the heater element
surface, as shown in Figure 11. As the trace element is located close to substrate edge, a less than 90-
degree angle brings the cord closer to the heater element trace, as in Figure 12. With this configuration,
there is an increased risk of damaging the trace as the Kevlar cord travels around the heater element
edge. The damage to the trace occurs once the fibers have been severed, thus the damage would not
become apparent until the subsequent activation.

Figure 11. MAP SARRS T/K and Kevlar Cord Figure 12. T/K with Kevlar Cord Less than
Configuration 90-degree Angle

When inspecting the knife after a cut, a continuity check may not reveal the presences of trace damage in
all cases. A resistance measurement must be made to insure the electrical integrity. In addition, for the
MAP SARRS, a visual inspection (20X magnification) was performed prior to and after final ground
activation to insure trace integrity. The inspection was possible because the SARRS design allows the
thermal knife heater element to be exposed after it severs the cord. A long-range microscope was used to
inspect and record the trace condition.

The MAP SARRS never experienced the potential thermal knife problems cited above or any other
problems, due to its thermal knife configuration. However, there have been some issues with the thermal
knife on other projects. After those projects investigated their problems, solutions were suggested, but the
SARRS was already in compliance with those suggestions. Thus, the MAP SARRS configuration has
been supported by much more information than can be presented under the scope of MAP, and those
data will most likely be presented in the future.

Kevlar Cord Behavior

The SARRS cord tension requirement was bound by minimum and maximum values of 45 N (200 lb) and
110N (490 lb), respectively. The minimum tension was the tension required to secure the solar panels
against their stops such that gapping did not occur during launch. The maximum value was based on the
spacecraft structure’s ability to withstand the compression induced by the cord tension. In addition, the
tension must be maintained above the minimum for 45 days on the launch pad and during the launch
environment. A series of tests were performed to determine the Kevlar cord tension characteristics under
these conditions.

Time, pre-conditioning, humidity, and temperature all affect the SARRS Kevlar tension load and
relaxation rate. A series of tests were performed to investigate the individual and combined effects that
each of these conditions would have on the Kevlar cord tension.

The cord assembly has a total unloaded length of 7.62 m (300 in). It was not feasible to place this entire
length within the available test chamber. Therefore, a test was performed to investigate the feasibility of
testing shorter cord lengths and applying the results to longer lengths. These results proved to be
positive, so it was decided to proceed with testing using cord samples that were shorter than the flight
cords.

Relaxation From Final Stowage To On-orbit Deployment

Early testing showed that the Kevlar cord, as delivered, would lose tension very rapidly with respect to
time. In an effort to reduce the tension loss over time, a pre-conditioning, or pre-stretching, process was

NASA/CP—2002-211506 248
developed and improved during the development of the SARRS. The final process involved pre-stretching
the cord by tensioning it with 157 N (700 lb) 10 times for 3 min each time. Then it was installed on the
spacecraft and tensioned to the initial tension. After at least 24 hours, the cord was re-tensioned to its
initial tension. Variations of this process are evident throughout the characterization process as this final
process was derived.

Several tests were performed that subjected tensioned Kevlar cord samples to a simulated environment
from cord installation to on-orbit deployment. The initial purpose of these tests was to determine the
relaxation rate and final cord tension at the end of a 45-day period. Later, the test was extended to
include the launch environment. The goal was to demonstrate that a pre-stretched Kevlar cord could
maintain a minimum tension of 45 N (200 lb) throughout the required time interval.

A Kevlar test specimen of 41 cm (16 in) was used in the first test. A typical test fixture with Kevlar cord
specimen and load cell is shown in Figure 13. The specimen was pre-stretched by cycling the tension to
112 N (500 lb) ten times at three minutes duration and immediately placed in the test fixture at 62 N (275
lb). Two days after the initial loading, a relaxation plot, Figure 14, projected that the cord would not
maintain the minimum tension for the required 45-day period. The specimen was reloaded to its initial
value and a new 45-day period was started.

Load Cell
Kevlar Test Specimen

Figure 13. Load Relaxation Test Fixture

During the first 10 days of the new cycle it became apparent that the ambient humidity fluctuations were
affecting the load relaxation rate. To determine the magnitude of the humidity effects, the specimen was
placed in a humidity-controlled (“Glove”) box so the tension load could be monitored as a function of
relative humidity. Each time the humidity setting was changed, the cord load readings would change
according to approximately 0.5 N (2.5 lb) per % change in relative humidity (RH). At day 24 the RH was
set to 40% for the remaining duration of the test. The rate of relaxation during this period was constant.

Load Change in Vacuum

An additional test to simulate the launch environment was conducted with two 64 cm (25 in) Kevlar
specimens. The specimens were exposed to full vacuum within 10 minutes (37% RH to 0 % RH) after
being pre-stretched (10 times/ 3 min duration with 112 N) and then loaded to 70 N (310 lb). The total load
loss for both specimens was 17 N (75 lb) after 24 hours, or approximately 0.5 N (2 lb) per percent
humidity. Figure 15 illustrates the rate of tension loss while the cord was under vacuum.

NASA/CP—2002-211506 249
 300
16" Specimen

280 60 % RH
Placed in
Glove Box

260
Load 40 % RH
(lbs)
240 30 % RH 40 % RH

220
Initial Loading
20 % RH
200
0 10 20 30 40 50

Time (days)

Figure 14. Load Over Time with Varying Relative Humidity

320
Vacuum Load Loss Data

300

280
Load
(lbs)
260

240
#12
#13
220
0 5 10 15 20 25 30
Time (hrs)

Figure 15. Kevlar Tension Loss under Vacuum

Relaxation Rate Versus Time with Constant Relative Humidity

A test was designed to determine the SARRS cord relaxation rate and whether the initial tension load
would affect the rate. Four Kevlar cord specimens 64 cm (25 in) were preloaded 10 times each to 157 N
(700 lb) for 3 minutes duration. The cords were tensioned to 74, 80, 85, and 91 N in a humidity
environment of ~ 40% RH. It was noted that prior to placing the specimens in the test chamber the
tension was decreasing rapidly. The tensions were reloaded to their initial values three times within a 50-
minute period. Each time the specimens were reloaded the relaxation rate decreased. After the third
reloading, the specimens were placed in an enclosed environment that maintained 22-24 ºC at 50% RH.
The tension values were continuously monitored for 45 days and the results are shown in Figure 16.

The loads were plotted verses time on a logarithmic scale. The load increase at day 9 resulted from an
error in the test chamber humidity setting. From the plot, the rate of relaxation was independent of the
initial tension and decreased an average of 5% per order of magnitude. Based on the slope, the cord lost
5% of the initial tension within the first day, another 5% by day 10, and another projected 5% by day 100
after the final reloading. Thus, the cord would lose less than 15% of if its initial tension after 45 days.

NASA/CP—2002-211506 250
 Humidity set to 50% RH
400
A-7341 A-4530
A-7336 A-7521
380

360

Load 340
(lbs)
320

300 Load spike

due to error in
280 humidity

260
0.1 1 10 100
Time (days)

Figure 16. 45-Day Cord Relaxation Rate

Temperature Effects
A test was performed to determine the temperature effects on the SARRS cord. A typical load versus
temperature plot was generated from temperature and load profile data, as in Figure 17. Below 40 ºC the
Kevlar tension changed at a rate of 0.2 N (0.9 lb) per degree Celsius and the rate of change above 40 ºC
was less, so it was assumed to be zero.
420
#14

400

380

Load
(lbs) 360

340

320
-60 -40 -20 0 20 40 60 80
Temperature (C)

Figure 17. Load vs. Temperature

Cord Tension Prediction

There was no access to monitor the cord tension after fairing installation and during the potential 45 days
on the launch pad. However, it was possible to estimate the SARRS cord tension at any point from the
last measurement to deployment. The calculations were based on the above test results of relaxation
rates due to time, relative humidity changes, and temperature (Table 1).

NASA/CP—2002-211506 251
 Table 1. Calculating the Resultant Cord Tension Given the Environment
As an example, given: The minimum tension during launch would be:

Installation Tension 76.4 N (340 lb) Installation Tension 76.4 N (340 lb)
Humidity during installation 46% RH Change in % RH (6 x 0.5N) - 3.3 N (15 lb)
Time on launch Pad 10 days Time on Pad (0.1 x 76.4N) - 7.6 N (34 lb)
Humidity during 10 day period 40% RH Vacuum (3.3N / hr) - 5.2 N (23 lb)
Pad Temperature 18 ºC Delta Temperature (17 ºC) + 3.3 N (15 lb)
Orbit Temperature 35 ºC
Deployment in 1.5 hours Minimum Launch Tension 63.6 N (283 lb)

Kevlar Cord Tension Measurement

As described in the Background Section, the MAP SARRS relies on a Kevlar cord to restrain the six solar
panels in their stowed configuration during launch. The cord must maintain a minimum tension to
eliminate vibration impact between the solar panels and the spacecraft bus during launch and to fly away
from the probe when released. In addition, the minimum tension must be maintained for at least 45 days
in case of launch delays. During SARRS development, it was found that the Kevlar cord has a creep
characteristic that causes it to relax and loose tension over time as discussed in the section above. In
order to help characterize the creep characteristic and to ensure that the proper initial tension is applied
to the cord before launch, accurate and reliable tension measurements had to be made.

Issues
It was soon discovered that the original method employed to measure the cord tension, the use of a
three-point tensiometer, the only commonly used method of measuring cable tension, proved to not work
for Kevlar cord.

Three-point tensiometers are used in the standard method to measure the tension in stainless steel
cables. These hand-held devices measure a cable’s tension by bending it around two posts and a plunger
as shown in Figure 18. The device measures the force on the plunger from the bent cable and a chart
calibrated specifically for the type and weave of the cable translates this force into the tension in the
cable. This method works well for stainless steel cables.

Plunger
Cable Cable

Post Post
Figure 18. Three Point Tensiometer Schematic

However, it was found that a Tensitron three-point tensiometer did not work well with the Kevlar cord. Its
measurements varied by 17% along one continuous, section of cord, supported only at the ends, where
the tension is actually the same along the entire section. Also, repeated tension measurements made by
the tensiometer in the same location on the cord varied by up to 10%.

Observation of the cord after a tension measurement by the tensiometer revealed that the tensiometer
was permanently deforming the cord and leaving a crimp, resulting in measurement errors. Also, the
tensiometer attempted to increase the length of the cord during measurement by stretching it, but the
friction at the cord standoffs did not allow for the stretching force to be absorbed by the entire length of
the cord, and, thus, greatly increased the tension in the local section being measured. The crimp and the
friction phenomena led to the belief that the problem was with the method of the three-point tensiometer,
rather than the particular device itself.

NASA/CP—2002-211506 252
Based on these issues, it is not possible for the three-point tensiometer to measure the Kevlar cord
tension sufficiently to meet MAP’s requirements and must be replaced. Another existing measurement
method could not be found for Kevlar, so a new method was developed to meet the system requirements.

Solution
The development of a new tension measurement system began with the following design requirements:
• Measure the tension in the Kevlar cord to within plus or minus 4 percent within the desired 22
to 101 N (100 to 450 lb) range
• Repeatedly measure the tension in the cord within the same section within the given
tolerance range
• Be capable of single hand use that can safely be used to measure the cord within close
proximity to the flight solar panels with minimal risk to the flight hardware

After considering a few other options, a rather simple solution was found, the Musical Pitch Method
(MPM). The MPM uses an innovative process to accurately measure tension in cords made from Kevlar
by combing music theory and physics relations.

The MPM employs an off-the-shelf chromatic tuner with a clip-on pick-up, a type of microphone, shown in
Figure 19, to determine the tension in Kevlar cord. Musicians use these tuners to tune musical
instruments. The chromatic tuner measures the musical note, or pitch, that emanates from a free section
of cord with fixed end conditions when plucked like a guitar string. The tuner’s microphone can measure
the pitch, but the clip-on pick-up eliminates the effect of background noise. Music theory assigns a
frequency to each musical pitch, so, by measuring the pitch, the tuner measures the frequency of the
vibrating cord. A physics formula translates the frequency to the tension based the section’s node length
and the cord’s mass per unit length.

Figure 19. The chromatic tuner with the clip-on pick-up

Frequency Measurement
The chromatic tuner measures musical pitches by providing information about the octave, note, and cent
deviation from a perfect musical note. The musical note information must then be translated into
frequency measurements via music theory definitions.

Each musical note has an assigned frequency. By definition, the A above middle C represents a
frequency of 440 hertz (Hz). Twelve notes form an octave, C, C#, D, D#, E, F, F#, G, G#, A, A#, B, with
the # representing a sharp, so C# is C sharp. Similarly named notes in different octaves represent
frequencies that differ by a multiple of two. For example, the A in the octave below middle C represents a
frequency of 220 Hz. This sequence repeats itself in the next octave. The steps between each of the
twelve notes are known as half steps. Notice that B# or E# do not exist. That is because there is a natural
half step between E, F and B, C. The frequency of each note between the A in one octave and the A in
the next octave are determined by the proper fraction of a multiple of two. For example, the frequency of
(1/12)
the A# above the A with a frequency of 440 Hz is determined by 440∗(2 ) = 466.1. Table 2 shows the
frequency that each note represents for the three octaves within MAP’s Kevlar cord tension range.

NASA/CP—2002-211506 253
 Table 2. Frequencies that each musical note represents in octaves -1,0, and 1
Low Octave (-1) Middle Octave (0) High Octave (+1)

Note Frequency (Hz) Note Frequency (Hz) Note Frequency (Hz)

C 130.8 C 261.6 C 523.2

C# 138.6 C# 277.2 C# 554.3
D 146.8 D 293.6 D 587.3
D# 155.5 D# 311.1 D# 622.2
E 164.8 E 329.6 E 659.2
F 174.6 F 349.2 F 698.4
F# 185.0 F# 370.0 F# 739.9
G 196.0 G 392.0 G 783.9
G# 207.6 G# 415.3 G# 830.5
A 220.0 A 440.0 A 879.9
A# 233.1 A# 466.1 A# 932.2
B 246.9 B 493.8 B 987.7

A measured pitch can deviate from a perfect musical tone. The chromatic tuner gives readings in terms of
cent deviations that range from –50 to +50. The cent scale corresponds to percent, but differs
numerically, as the cent reading equals the percent distance between two notes. The cent size does not
relate to the percent deviation from the frequency of the note measured. The cent size differs depending
on which notes the measurement spans. If the tuner reads C+50 cent, the pitch lies halfway between a C
and a C# and, in turn, the frequency lies halfway between the frequencies for a C and a C#. If the tuner
reads C+30 cent, then the measured tone is 30% of the distance between C and C# from C, and the
frequency equals that of C plus 30% times the difference in the frequencies of C# and C. The pitch and
frequencies for C+50 cent and C# -50 cent are equal. Table 3 shows the frequencies for the cent values
between middle C and C#.

Table 3. Frequencies for the cent values between middle C and C#

Note (cent) Frequency (Hz)

C 261.6
C - (+10) 263.2
C - (+20) 264.7
C - (+30) 266.3
C - (+40) 267.8
C - (+50) 269.4
C# - (-50) 269.4
C# - (-40) 270.9
C# - (-30) 272.5
C# - (-20) 274.0
C# - (-10) 275.6
C# 277.2

Tension Conversion
Now that the frequency of the section of cord is known, Equation 1, which is derived from the wave
equation, can be used to determine the cord tension. In Equation 1, the Tension T is dependent on the
frequency γ, the node length L, and the mass per unit length µ of the section of cord.

( 2γ L ) 2 µ = T (1)

Equation 1 represents the general case and does not take stiffness or friction into account. It is possible
to solve the wave equation for the tension in the string while taking into account these factors, but the
solution is cumbersome and correction factors are still needed because of many variables in the
construction of the cord (i.e., weave, pre-stretch).

In order to calibrate the MPM for the particular properties of the cord used for MAP, tests were performed
by tensioning a sample section of cord to known values and measuring the tension with the MPM. Charts
could then be made that calibrated the chromatic tuner readings with the proper tension in the cord, as

NASA/CP—2002-211506 254
the tests proved that the effects of cord construction on the first harmonic frequency became constant.
The correction factors incorporated into the physics equation allowed the MPM to give accurate and
precise measurements of the cord tension.

The Musical Pitch Method demonstrated that it could measure the tension in the MAP Kevlar cord
accurately, as the conversion tables that used to translate the musical pitch to cord tension were
calibrated to within plus or minus 2%. This measurement technique does not permanently deform the
cord, thus making it more accurate than the three-point tensiometer. The Musical Pitch Method met the
project requirements based on its accuracy and ease of use and was used to tension the Kevlar cord on
MAP before launch.

Conclusions

The MAP SARRS operated successfully after the launch of the spacecraft. The thermal knife severed the
Kevlar cord as demonstrated in the thermal vacuum deployment tests. The SARRS design and extensive
test program were the main reasons for its success. The method for measuring the Kevlar cord tension is
a new approach and was developed at GSFC. This method is applicable to all arimid cord configurations
that are tensioned to level at which an audible sound is made when it is excited (plucked). The thermal
knife can be used successfully as a release system as demonstrated in various missions including MAP.
In utilizing the knife and Kevlar cord combination one should be aware of the potential problems.

The SARRS design offers solutions to some of the potential thermal knife problems. First, keep the cord
diameter to a minimum and provide enough stain to pull the severed fibers away from heater element
during activation. This configuration will minimize the chances of the severed fibers adhering to the trace
prior to final separation and causing damage. Finally, inspect the heater element before and after
actuation. Damage to the trace usually occurs during cord separation, and subsequent activation will be
affected by the existing damage. A Continuity check alone may not reveal a damaged trace. The list of
issues and recommendations cited in this report were based on the features in the SARRS design that
addresses them. GSFC is currently compiling the investigation results of the thermal knife issues and
solutions from other programs. This report will be presented in the near future.

References

Marion, Jerry B., and Hornyak, William F., “Part 1: Physics for Science and Engineering,” CBS College
Publishing (1982) pp. 512-519.

NASA/CP—2002-211506 255
 Release Mechanisms on the Gravity Probe-B Relativity Mission

*
Sean McCully and Dennis St Clair*

Abstract

The Gravity Probe-B Relativity Mission is scheduled for launch in 2002. The space vehicle is comprised of
a dewar mounted into an aluminum space-frame or truss (Figure 1). This paper describes the design and
test of the release mechanisms used to constrain and release the solar arrays and attitude reference
platforms.

Figure 1. The Gravity Probe-B Space vehicle, Solar Arrays Deployed

*
Lockheed Martin Space Systems Company, Sunnyvale, CA
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 257
 Introduction

The GP-B solar arrays are 3.5 meters long, weigh 330 N, and are rotated about a single axis from the
stowed to deployed positions where they remain fixed relative to the spacecraft throughout the remainder
of the two year mission life. It is typical to have solar array release mechanisms, but the GP-B solution is
somewhat unique in that it uses two shape memory alloy rods for actuation.

The spacecraft control gyros and star trackers are mounted to an Attitude Reference Platform (ARP) that
is fixed to a thermally stable post. This post requires additional structural support to survive launch. The
ARP Launch/Restraint Mechanism (ALRM) was introduced to provide the necessary constraint changes
on-orbit. The ALRM is comprised of an upper and lower groove plate held together by a restraint rod. The
release mechanism is used to release this restraint rod.

The release mechanism for the gravity probe-B vehicle is an improved version of a previously flight-
qualified design [1]. The load carrying capability has been increased by a factor of five (to 3500 N). The
mechanisms’ release function is fully redundant both electrically and mechanically, and was easily re-
used over one hundred and twenty times in two proto-qualification test programs.

Release Mechanism Design

The GP-B release mechanism is comprised of two heat-activated shape memory alloy (SMA) rods which
begin to straighten when an external temperature of 70 C is reached (Figure 2). Each rod is held in the
titanium chassis by two pivot bushings. The rod is pinned to one bushing and is free to translate in the
other. As the rods straighten, the sleeve assemblies separate releasing the toggle bolt that is held
between them.

Figure 2. Shape Memory Alloy Release Mechanism (SARM)

A toggle bolt holds the solar panel to the spacecraft (Figure 3). The toggle head is constrained between
the two sleeve bearing assemblies. The preload is carried through the sleeve bearings into the chassis
and then into the spacecraft. The rods do not carry any preload, and actuation of either of the two rods
will release the toggle.

NASA/CP—2002-211506 258
 Figure 3. Solar Array Release Mechanism Launch Configuration

Two needle bearings were added to the heritage design in order to increase the release capability of the
mechanism (Figure 4). The rods are capable of 440 N of force as they straighten. By reducing the rolling
resistance in this interface, the preload could be increased to more than 3500 N. The hardware was
successfully released at a cold temperature of -78 C with a preload of 7000 N (demonstrating a margin of
two) using first one rod and then the other as part of the acceptance test program. The development
program demonstrated release at loads in excess of 10600 N.

Figure 4. Sleeve Assembly

The rod design is identical to the heritage design (Figure 5). The rod is 15 cm long and 0.63 cm in
diameter. A nickel titanium alloy with a transformation temperature of approximately 80 C was used. A
heating element was potted into the center of the rod using the heritage process. The rod was then
conditioned to obtain the desired “memory”. When subjected to heat, the material in the rod transforms
from martensite to austenite. This transformation results in the desired change in curvature.

NASA/CP—2002-211506 259
 Figure 5. Rod design

The heating element resistance is 22 ohms. A nominal bus voltage of 28 V across the heater draws
approximately 1.2 amps (2.4 amps total per mechanism). The mechanism releases in less than 120
seconds worst-case (i.e. single-rod, low-voltage, high preload 3500 N).

Release Mechanism Test

Component level releases were performed in a test fixture that simulated the solar panel interface (Figure
6). Both the acceptance test program and the proto-qualification test program included five releases for
run-in purposes at a preload of 3500 N. The baseline functional included three releases; a nominal
voltage dual-rod release at 3500 N, and two single-rod low-voltage releases at a preload of 7000 N. All
environmental testing was followed by a repeat of the baseline functional test.

Figure 6. Release Mechanism Installed in Test Fixture

Toggle preload, rod temperature, rod voltage, and rod current was recorded during all releases. These
data were used to determine the release time, the rod temperature at release, and the total power being
consumed.

NASA/CP—2002-211506 260
Random vibration was accomplished in the same test fixture (Figure 7). The proto-qualification test
program included a random vibration level of 9.2 Grms. The acceptance test program subjected the
hardware to a random vibration level of 6.5 Grms. Bolt preload was verified before and after exposure to
random vibration.

Figure 7. Release Mechanism Random Vibration Tests

The mechanisms were acceptance tested three at a time in the thermal vacuum chamber (Figure 8).
Thermal vacuum tests resulted in a total of seven (7) releases; a dual rod release at both hot (46C) and
cold (-71C) environments, two single-rod low-voltage (26V) releases cold, one single-rod hot release, a
dual-rod high-voltage (34V) release hot, and one dual-rod release at the predicted on-orbit operating
temperature (-25C).

NASA/CP—2002-211506 261
 Vacuum pump (turbo!)
Pirani 315 vacuum sensor
HP 34970A Data Logger
Pirani 315 vacuum controller
Vishay 2120A
8A02019GSE Timer
HP 6038A, 200W
Heater control console
8A02410GSE Switch Box
Type-T thermocouples
HP Pavilion 6470Z
8A02131GSE Vacuum fixture
Vacuum chamber
Freezer

Figure 8. Release Mechanism Thermal Vacuum Tests

Serial number three (3) was the first unit to be subjected to a proto-qualification test program. After nearly
completing the proto-qualification test program, including over one hundred and twenty releases, rods A
and B failed to release in the final post-test single-rod functional. Rod B eventually failed during repeated
attempts to study the problem (Figure 9).

Figure 9. Rod B failure

NASA/CP—2002-211506 262
X-ray imaging helped confirm that the heating element coils were in close proximity to each other (Figure
10). Insulation normally prevents coil-to-coil contact. However, it was determined that the thermal limits of
the heating element insulation is 218 C. A bare heating element was subjected to 50 volts for over 4
minutes at the vendor. Temperatures in excess of 650 C were achieved. The heating element did not fail,
but the insulation was charred. Continued mechanical cycling was required before contact (and shorting)
between coils occurred. The test program was revised to include a thermal limit of 218 C.

Figure 10. X-ray Image of the Rod

A revision in the sleeve assembly design was required to increase the clearance between the bearings
and the chassis. The fix was confirmed by installing the revised assembly and successfully releasing rod
A of serial number three.

The sleeve assembly design change was implemented. A rod temperature limit of 218 C was
incorporated into the test program, and the redundant mode of operation became part of the basic
functional that followed every environmental test. The entire proto-qualification program was successfully
repeated on serial number two.

After successfully completing their component level acceptance test program, serial numbers 7 and 8
completed the sub-assembly acceptance test programs of the ALRM without any additional problems
(Figure 11).

NASA/CP—2002-211506 263
 Figure 11. ALRM During Random Vibe Test

Conclusions

The GP-B release mechanism has successfully completed a proto-qualification and acceptance test
program. The two release mechanisms that were installed onto the ALRM are now installed on the
spacecraft. The solar array release hardware now awaits installation onto the spacecraft. The mechanism
has proven to be effective in releasing preloads of 3500 N at even the coldest of operating environments.

Acknowledgements

Tom McCloskey provided guidance in design, process, and tooling. Dennis Petrakis provided guidance in
material details and performance expectations. The GP-B program office provided resources and
patience as the hardware made its way through the assembly and test program.

References

[1] McCloskey, Tom, “Non-pyrotechnic Release System,” U.S. Patent # 05192147, 9 March 1993.

NASA/CP—2002-211506 264
 Conceptualization and Design of a Mechanical Docking System
*
Troy Nilson and Mitch Wiens

Abstract

A scaleable Mechanical Docking System (MDS) has been developed and tested to support docking and
servicing unmanned spacecraft. To allow for system use on various spacecraft configurations, the MDS
was developed to general requirements and designed for scaling ease. Information on existing spacecraft
docking systems was combined with a conceptualization of how spacecraft servicing might occur to
develop a simple and cost effective soft docking system. The development, testing, and performance of
this system, under the challenging absence of hard requirements, are described. The Air Force Research
Laboratory (AFRL) funded the project through two Phase I and Phase II Small Business Innovative
Research (SBIR) grants.

Introduction

Over the past decade there has been an aggressive push to develop the
capability of unmanned vehicles to service orbiting spacecraft. The reasons
are compelling:

• To extend the life of spacecraft at the end of life due to propellant depletion.
• To replace components that are obsolete or have failed
• To capture spacecraft and move them to more effective orbits (e.g. a
spacecraft is in an unusable orbit from launch vehicle failure)
• To recover a spacecraft with a failed deployment by manually deploying
• To examine spacecraft to determine cause of failure

Over the past 20 years, a servicing capability has been available through the Space Transportation
System (STS), or Shuttle. However, the high cost of this approach has limited its use to very expensive
systems within the Shuttle’s orbital reach (e.g., Hubble). As a result, the system is generally impractical
for current and emerging spacecraft servicing needs.

Unmanned servicing technology is expected to manifest in the next decade as programs such as Orbital
Express are forwarded. Key to these systems is the capability to capture, mechanically connect, and
make fluid and electrical connections between two spacecraft, an outstanding mechanisms challenge.

Starsys Research Corporation (SRC), as part of two Phase I and Phase II SBIR programs, has been
active in developing this type of system. The challenges associated required that a set of parameters for
a mechanical system capable of soft-docking two independent spacecraft be defined. Design goals
included provisions for the alignment, structure, and coupling forces necessary to join fluid and/or
electrical couplings. This same mechanism was to generate latching/rigidizing forces large enough to
react moment loads induced once the spacecraft joined. Additional challenges resulted as the project
went forward without tangible system requirements (i.e., the hardware was developed and tested in a
“requirements vacuum”). Although the projects were undertaken without complete definition of a servicing
program, there was always the clear intention of creating a system applicable to a variety of programs
and missions. The development approach included an evaluation of current docking technology, a
conceptualization of various mechanism configurations and mission scenarios, and then the
implementation of detailed trade studies intended to identify a viable concept. The resulting system was
designated the MDS (Mechanical Docking System).

*
Starsys Research Corporation, Boulder, CO
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 265
 The Challenge

The project began with a Small Business Innovative Research (SBIR) grant from the Air Force Research
Laboratory (AFRL). SRC successfully proposed a design to dock two small spacecraft using Nitinol shape
memory alloy within the coupling members. AFRL had a project called XSS-11 that included an
experimental mission where two coupled spacecraft were launched into orbit. Upon reaching stable orbit,
the experiment was to separate and re-join the structures. After the spacecraft became re-latched in
place, a transfer of water was to occur, and thus demonstrate the technology for future development.

From early on in the design effort, the group was working with only a general concept in mind. AFRL
awarded the grant to promote development of technology applicable to future docking needs. They
provided bounding conditions to the problem that included an estimated spacecraft size and mass (a
cylinder, 50cm OD x 130cm L, 50 kg) and a requirement to dock two spacecraft and transfer fluid and/or
electrical data between them. Without a specific and concrete design goal, the task of creating a
meaningful mechanism necessitated stepping back and looking at the design from an entirely new point
of view. A definition of the various needs that might arise when servicing obsolete or exhausted
spacecraft had to be formulated. The challenge became one of establishing requirements to meet the
needs of a concept rather than designing to solve problems based on a clearly defined set of
requirements.

A research effort was undertaken to evaluate existing technologies. The result was very discouraging.
Most available technology related to extra-vehicular activity (EVA) interfaces, robotic boom end-effector
configurations, or impact docking mechanisms. EVA technology used on orbital replacement unit (ORU)
mounting interfaces yielded some ideas about alignment and latching. This technology was by its nature
intended for manned intervention. Grappling technology developed as robotic end-effectors used on the
shuttle (e.g. Remote Manipulator System (RMS)), presented other ideas on mechanical interface
configurations. The RMS boom end-effector was designed to grapple payloads and allow astronauts to
move and place objects. This mechanism was very complex and large and it tended to fall outside the
emerging design goals. Other technologies, however, were specifically related to spacecraft-to-spacecraft
docking. Examples of both impact docking and “soft” docking were found. Mechanisms such as Apollo-
Soyuz Universal Docking Interface, Apollo Probe and Drogue Docking Assembly, and Gemini-Agena
Docking System were reviewed. These systems were primarily intended for large spacecraft and impact
docking applications. They were not well suited for the more precise requirements of autonomously soft
docking and aligning fluid/electrical couplings. A few soft-docking interface examples were found where
linkage assemblies provide the means of capture. One example demonstrated a series of linkage/latch
assemblies around a large interface perimeter. These assemblies operated independently to latch
together mating spacecraft. Another, the Japanese Experiment Module (JEM), utilized linkage assemblies
to engage a trefoil interface on the mating spacecraft.

There was also a clear need to define the eventual use and application of this intended mechanical
docking system. Further research did not yield any definitive mission scenarios. However, it did present
the design team with general ideas about how a docking system might be applied. It was believed that by
defining these possible mission scenarios, a docking system could be designed and tailored specifically
for that mission/use. Two mission scenarios were conceived.

• Client (spacecraft in orbit needing service) / MicroSat (spacecraft servicing the Client) - A MicroSat
could be launched on an as-needed basis to directly service an orbiting client (good scenario for
technology upgrades, refueling, and repairs)
• Various client satellites and a MicroSat base station (both in orbit) - The Client calls on the MicroSat
base as needed (good scenario for common repairs or refueling needs).

In either case, the MicroSat might be used in a number of configurations. The MicroSat may be designed
to interface with the client, perform a task, and depart (return to an orbiting depot or to earth). It might also
be designed as a single-use unit that attaches to and stays with the client. Here, the MicroSats might
require an option to be stacked. As a fuel cell or control module becomes outdated, an additional

NASA/CP—2002-211506 266
MicroSat could be added directly onto a previously docked MicroSat without the need for multiple
interface locations on the Client.

In a parallel process, the design team traded many different methods of joining, grappling, and aligning.
There were primarily four different concepts that surfaced. Ideas included:

• “harpoon” configuration
• a telescoping probe
• impact docking with a large conical guide
• “claw-type” linkages interfacing with trefoil

The harpoon configuration, in which a probe is launched at a target, latches on, and is reeled back into
guide features to complete the docking process, seemed inherently unpredictable and presented
complications concerning the alignment of fluid and electrical couplings. A telescoping probe, where a
telescoping pole extends to the mating spacecraft, engages a target feature, and retracts to join the two,
turned into a complicated multi-mechanism apparatus not well suited for the direction and vision taking
form. Impact docking was eliminated as an option because soft docking seemed safer for the two
spacecraft and better suited to precisely align fluid and electrical couplings. In the end, the application of
linkage assemblies interfacing with a trefoil was found to be the simplest and most effective means to
reach out, grasp, pull in, and latch.

Different options were explored for providing the motive forces within the mechanism. Nitinol (a shape-
memory alloy) was considered. Because the program was a research effort and because the application
of Nitinol was the focus of the initial design effort, much time was spent understanding this material. This
material can generate work through a phase change transformation that returns the material to an
unstrained state (about 2-4%). The material, when heated, changes lattice structure and “remembers”
the unstrained configuration, performing work as it returns. The actual stroke is generally very short, and
in the end, because we were moving towards soft docking with the need to create effectively long strokes,
Nitinol proved to be ill suited for the application. The use of wax and paraffin actuators, a core SRC
technology, was also considered. This technology involves heating wax in a contained space and using
the expansion in volume due to a phase change from a solid to a liquid state to produce work (generally
in the form of a High Output Paraffin (HOP) actuator). Here again, this technology in not well suited for
long stroke applications. The application of a gearmotor and lead screw was evaluated as well. Clearly,
high forces and long strokes could be achieved using this technology.

Design Objectives

The next phase in the development process became one of defining a concrete set of design objectives.
Current technology had been explored and possible mission scenarios had been defined. There was now
something tangible to which a set of docking system requirements could be applied. Fundamentally, the
design objective was to provide a simple means of docking, aligning, and latching two spacecraft
1
autonomously . Through a series of brainstorming sessions, the following set of primary objectives was
defined.

• provide interface space for couplings

• keep it simple
• accommodate a universal interface
• maximize cost effectiveness
• accommodate design scalability

1
Assumptions were made that a given control system existed which was sophisticated enough to position
two spacecraft within some known range. The design goal became one of completing the docking task
only. There was no focus or intention on developing or incorporation any sensors or guidance technology
into the design.

NASA/CP—2002-211506 267
The design had to have provisions and space for fluid and electrical couplings. This was the goal from the
beginning and continued as a prime driver throughout.

A simple design would use a limited number of sub-mechanisms that incorporate efficient motive forces
and minimize the number and complexity of parts. Throughout the trades and evaluation of different
design concepts, a highly scrutinized area was the quantity of mechanical systems required to reach the
objective. Seen through earlier trades, options arose which required mechanisms to reach out and grab
(i.e. grapple), mechanisms to latch and rigidize, and mechanisms to complete the mating of fluid and
electrical couplings. Nitinol and wax actuators could be used along with stored spring energy to complete
these separate functions. However, the utilization of a gearmotor and lead screw was preferred simply
because a single motive force could perform all these tasks associated with soft docking. Design
concepts incorporating fewer complex parts also greatly influenced the design direction. A clear method
to minimize the complexity of the system was to develop a completely passive interface component (i.e.
one absent of driven mechanisms or power requirements).

In addition to simplicity, developing a docking system with an industry standard in mind seemed prudent.
A docking system should be designed in a way that spacecraft manufacturers could integrate a common
interface and be assured that, later on, the interface would interact with a standard servicing spacecraft. It
was clear that a very simple passive interface, designed to mount as a standard interface on all
spacecraft, must be developed to work with a more complex docking mechanism used in a more limited
sense on the servicing spacecraft only.

By striving for simplicity in design a low cost solution could be achieved. The active side of the MDS was
complex and would therefore be more expensive. However, the lower quantities of the more complex and
expensive mechanism interfacing with higher quantities of very simple inexpensive passive mechanism
proved less costly than the alternative (i.e. to use higher quantities of more complex and expensive
mechanisms at each interface).

Creating a design simple enough to facilitate scaling became an objective as the project matured. There
was always an intention to accommodate different applications; this played a vital role in the eventual
configuration. One benefit includes the ability to dock various sizes of spacecraft. For example, a baseline
preload increase of 3 times (e.g., from 2224N (500lb) to 6672N (1500lb)) might result in an overall
diameter increase of 2.5 times (e.g., from 25.4cm (10in) to 61.0cm (24in)) allowing larger spacecraft to
dock.

Combining primary design objectives with concepts and ideas gathered during various brainstorming
sessions, a list of more specific design requirements was compiled (Refer to
2
Table 1). Of particular note, a list of guidelines outlined in an AIAA publication was included.

The Scalable Mechanical Docking System

Through the process of research, mission scenario development, and concept trades (e.g., mechanical
configurations and motive forces), a design direction took shape that formed the foundation of the
development hardware. Various trades made clear that the preferred design and docking method was to
incorporate a set of three linkage assemblies, driven by a single motor, to interact with a passive
trefoil/interface. The design concept, in the end, resembled that developed by the Japanese (i.e., JEM).

2
“Guide for the Servicable Spacecraft –Grasping/Berthing/Docking Interfaces”, American Institute of
Aeronautics and Astronautics (AIAA), AIAA G-056-1992, 1993

NASA/CP—2002-211506 268
 Table 1. MDS Design Requirements
Parameter Value
Max Capture Distance*: 6.0 cm (2.36 in)
Angular Capture Misalignment Tolerance*:
Pitch/Yaw ±5 degrees
Roll ±5 degrees
Linear Misalignment Tolerance*: ±25 m (±1.0 in)
Linear Contact Velocity Tolerance*: -.005 m/s (-.016 ft/s)
Preload: >2224 N (>500 lb)
Capture Time: <5 s
Capture and Latch Time: <120 s
Interface Outer Diameter: <30.5 cm (12 in)
Mass: <89 kg (20 lb)
*AIAA: Autonomous “Non-impact” Docking Tolerances

The MDS is comprised of many different features, which when described, create a better understanding
of the system functionality and provide some insight as to how design objectives found there way into the
working prototype. These subgroups include:

• grappling features (drivetrain, linkages)

• system alignment features (passive geometry, pin/cone kinematic interface)
• fluid/electrical couplings

Refer to Figure 1 for a breakdown of system components. Figure 3 has also been included to provide a
sense of the prototype system size.

Grappling Features
There are several components that create the grappling functionality within the MDS. They include a 1.5:1
translation gearbox used to transmit motor power off-axis to the central MDS axis and leadscrew. The
motor is a brushless DC gearmotor with a dual-stage 49:1 planetary gearbox. Included in the motor is an
optical encoder providing speed and position feedback to a motor controller allowing the system to be
operated closed-loop. The housing provides a structure to guide the main moving structure within the
mechanism, the piston (Refer to Figure 2). The leadscrew, Ø9.5mm (0.38in) with a 1.27mm (0.05in) lead
(36% efficient), is axially fixed. By coupling the leadscrew to the piston, motor motion causes the piston to
move in either direction within the housing. Attached to the piston is a set of three linkage assemblies that
are spaced 120 degrees apart around the mechanism. These 4-bar linkage assemblies react against a
guide roller that causes the spring biased linkage assemblies to move axially as well as towards the
center axis of the MDS. This motion allows the linkage assemblies to interact with the passive geometry.
With a single mechanical action, these linkage assemblies provide the capture, alignment, and latching
forces required to mate the spacecraft. These linkages have been designed to optimize docking motion.
Lateral motion (i.e. motion towards the center axis of the MDS) was incorporated to maximize grappling
action and prevent the docking event from pushing mating spacecraft apart.

System Alignment Features

There are two levels of alignment designed into the MDS. Course alignment is achieved by driving the
active side linkage assemblies to mate with the passive interface trefoil. These passive features, or “pie
wedge” sectors, are positioned 120 degrees apart to match the 3 sets of linkage assemblies on the active
side. As the linkage arms swing from the “ready-to-dock” position (i.e., arms fully deployed) to the
“captured” position (i.e., arms fully extended in the axial direction) (Figure 3), spherical ball ends on the
linkage assemblies react on the faces within the passive interface wedge geometry (Figure 4). These
reactions force both interfaces to rotate about the long axis of the MDS into alignment.

NASA/CP—2002-211506 269
 3X Linkage Assy. 3X Retention Lip

3X Shear Cone
Passive Interface

3X Shear Pin

3X Spring Can
(Spring Not Shown)
Active Interface

Gear Box and

1X Fluid Coupling
Gearmotor
2X Electrical Coupling

Leadscrew
3X Reaction Roller

Housing
Piston

Radial/Thrust Ball Screw Nut

Bearing

Figure 1. MDS Cross Section

Fine alignment occurs via the interaction of spring-loaded shear pins (active interface) into 3-point
kinematic load points (passive interface), again spaced 120 degrees apart. Each load point location is
designed to constrain from one to three degrees of freedom. All points react forces in the Z-axis, two
points additionally constrain motion in Y axis, and only one point reacts forces in all three axes (X, Y, & Z,
where X and Y are in the interface plane, and Z is in the axial direction). The geometry of this last point is
a 45-degree conical pin mating with a conical hole. The final alignment is fine enough to accurately
position mating fluid and electrical couplings over one another prior to actually making coupling contact.

The fine alignment pins are also preloaded against compression springs prior to seating rigidly and
transferring a latching preload from the active to passive interface. In the final phase of the docking
process (refer to Phase 4 of Figure 5), this axial compliance accomplishes two functions. First, a parallel
interface plane is established above delicate and sensitive coupling interfaces prior to completing the
docking maneuver. Secondly, the preloaded pins provide the coupling de-mate force (approximately 44N
(10 lb) each pin).

NASA/CP—2002-211506 270
 Motor/Linkage/Piston
Housing/Piston/Leadscrew

Figure 2. Motor/Leadscrew/Linkage/Piston/Housing

“Ready-to-Dock” “Captured” or Fully Extended

Figure 3. MDS System in Various Positions

Fluid/Electrical Couplings
The design incorporates 3 locations for fluid and/or electrical couplings (Figure 4). Because the design
uses the center of the mechanism for actuating the system, coupling locations were placed concentrically,
in three locations, at a radial distance from the mechanism central axis. By using compression springs
within the coupling assemblies, all three points, regardless of the type of coupling installed, provide a
balanced resistive load during the final mating and rigidizing. This prevents any binding from occurring.
Each coupling also has micro-alignment features to provide a final level of mating accuracy at each
coupling point.

NASA/CP—2002-211506 271
 Fluid Coupling

3X Kinematic
Mounts

2X Electrical
Couplings

Active Side (Ready to Dock ) Passive Side

Figure 4. MDS Interface

System Operation
To gain a clear understanding of the MDS operation, consider one possible docking scenario. Two
spacecraft (Client and Microsat) orbit near each other, both with independent guidance systems. The
Microsat is commanded to approach the Client and maneuver its’ active docking interface within some
range over the passive interface of the Client. In parallel the Microsat triggers the active interface to
deploy from a stowed configuration. Once feedback indicates the spacecraft are within docking range, the
Microsat disables its guidance system and activates the docking sequence of the MDS. Within seconds,
the MDS has captured the Client and continues to complete docking. The following is a summary
description of the actual MDS docking sequence (Figure 5).

Phase 1, Approach
The active side linkages are fully deployed in the “ready-to-dock” configuration and both interfaces are
brought within the capture envelope. As soon as attitude control signals verify proper positioning, the
MDS is activated.

Phase 2, Capture
Power is immediately applied to motor at full speed causing the piston to stroke axially causing the
linkage assemblies to rotate toward the passive interface. As the linkages engage the trefoil features in
the passive side, coarse alignment occurs. This phase is complete when the linkage assemblies are fully
extended (Figure 3). At this point, the passive interface is considered captured.

Phase 3, Retraction
Still at motor full speed, the linkages withdraw into the housing pulling the passive interface with it. The
linkage assemblies stroke until the pin/cone alignment features have engaged but the preload springs
under the pins have not yet compressed. At this point, the motor speed is reduced to minimize pitch/yaw
oscillations of the interfaces during the mating cycle of the couplings.

Phase 4, Rigidization
The motor continues at a slow rate drawing the two interfaces together and compressing the shear pin
springs in the process. Power is discontinued when feedback to the control system senses desired
preload at the kinematic shear points (the design includes load sensors at all three points). At this stage,
the system is latched and preloaded. Transfer of power/data or fluid through couplings in now possible
without the need for additional docking mechanisms.

NASA/CP—2002-211506 272
 Phase 1 Phase 2 Phase 3 Phase 4

Figure 5. MSD Operation and Kinematic Motion

Testing

Testing of the MDS was undertaken to evaluate the effectiveness, as well as limitations, of the design. The test
program included three significant components.
• An off-load fixture was designed and manufactured to assist in understanding various dynamics of docking as
well as to prove system ability to dock and transfer cryogenic fluids (LN2)
• A test plan was developed and executed to test the MDS hardware in a micro-gravity environment.
• The design was applied to a software simulation package where further information was gathered on the
effectiveness of the mechanical system, in particular, the linkage system.

Off-loading Approach
In order to gain a general understanding of the docking system dynamics, an initial test bed was
constructed using a simulated inertia mass and cable/pulley system (Figure 6). The passive side of the
MDS is shown mounted to a tip/tilt fixture on an x-y translation stage. The fixture and stage allow
adjustment of the passive half relative to the active half to perform various misalignment tests. This
passive side of the fixture as well as the associated fluid and electrical couplings represents the client
satellite. The large tubular structure suspended from the cable simulates a micro-satellite inertia
approximately equivalent to a 50-kg satellite. The active coupling is mounted to the bottom of this
structure. Adjustment of the cable attachment point (universal joint) and offload weights (shown at the
right) provided a close approximation to a zero-g environment. This test setup was used for the following
tests:
• Mating and de-mating of the interface under various alignment scenarios and capture ranges to verify
docking capability (Test results are summarized in Table 2.)
• Electrical and fluid coupler mating and de-mating verification

All couplings were visually inspected after each de-mate to verify no damage resulted from mate and de-
mate cycles. Continuity checks were performed on the electrical connectors once the docking interface
was mated. The fluid coupler was verified by performing liquid nitrogen transfer tests on the mated
coupler. Three liquid nitrogen transfer tests were demonstrated using the developmental fluid coupler
installed in the MDS. During development over 200 mechanical mate/de-mate cycles were successfully
completed.

NASA/CP—2002-211506 273
 Cable/Pulley
System

Universal
Joint/ Cable
Attachment

Inertia Mass
Simulator

Off-Load
Mass

X-Y Translation
Table

Figure 6. MDS Zero-G / Off-load Test Set-Up

Micro-Gravity Testing
The second phase of MDS testing and evaluation included a flight on NASA’s modified KC-135 aircraft
nicknamed the “Vomit Comet” (flown Ellington Field near Johnson Spaceflight Center, Houston (JSC)).
Tests were performed as the aircraft flies in a series of parabolic maneuvers (3-4 sets of 10-12
parabolas). Approximately 25 seconds of micro-gravity followed by 30 seconds of 2-g at pull-up were
experienced within each parabola. Since the entire docking sequence was approximately 2 minutes long
(from docking initiation to final preload) the experiment focused on the grappling and capture events that
could be accomplished within the 25 seconds zero-g window. The intent of utilizing this resource was to
better understand the dynamics of mechanism docking in a simulated micro-gravity environment. Of
prime importance was to determine whether or not the linkage motion had a tendency to push the passive
interface out of capture range during the docking sequence.

Table 2. MDS Simulated Zero-Gravity Off-load Test Results

Requirement Design Goal Test Results

Capture displacement (x-axis) 6.0 cm (2.36 in) 6.3 cm (2.5 in)
Capture displacement (y & z-axis) 2.54 cm(± 1 in) 2.54 cm(± 1 in)
Capture Roll Angle ± 5° (± 5-10° )
Capture Pitch/Yaw Angle ± 5° (± 5-10° )
Preload 2224 N (500 lb) 2669 N (600 lb)
Capture Time <5s 3-4 s
Capture and Latch Time < 120 s 120 s
Mass 89 N (< 20 lb) 80 N (18 lb)

In order to perform experiments on the KC-135, JSC’s Flight Safety Review Board must evaluate test
plans and fixturing for safety as part of a Test Readiness Review (TRR). Although a detailed test plan had
been submitted for review, the Review Board found many points of concern with the proposed test plan

NASA/CP—2002-211506 274
and fixturing once they could see the hardware and truly understand the intent. The zero-gravity docking
test was postponed a day pending resolution of these issues. The Review Board’s main concern involved
the low probability of returning a heavy free floating mass (>113kg (250 lb)) to a fixture cradle prior to the
2-g part of the parabolic cycle. The solution was to modify the fixtures by simplifying them and by
reducing their weight to <45kg (<100 lb). In the end, all tests were run from the floor of the aircraft where
a very simplified fixture support was created of padded 2x4s and plywood.

The experiment was designed to have two separate free floating simulators representing the relative
inertia of a client and servicing satellite (approximately a 2/1 ratio client to servicing) (Figure 7). Each half
of the MDS interface was mounted to a separate inertia/mass simulator. During the micro-gravity phase
the hardware and related mass/inertia simulators floated up above a stationary fixture cradle. With JSC
crew assistance, the two halves were positioned within capture range. The active docking mechanism
was immediately powered in an attempt to demonstrate “zero” gravity docking. Because of the limited
time in micro-gravity, consecutive phases of the docking sequence were performed throughout a series of
parabolas. It is important to note here that the JSC crew’s assistance was invaluable in executing the
planned tests. Not only were they available to assist in running the tests should team members become ill
(motion sickness), their extensive experience working in a micro-gravity environment made the very
difficult tasks possible.

These micro-gravity tests demonstrated two primary points. First, it was clear that the design had
sufficient compliance in the linkages to prevent a “rebound” effect. There were concerns early on that the
relatively small impact forces associated with docking in “zero” gravity would cause the spacecraft to
move out of capture range as they reacted against one another during the docking phase. This proved to
be of little concern during test, but it did emphasize the importance for compliant linkages. Second, there
are definite limitations to testing free-floating docking mechanisms on the KC-135. Although the MDS was
consistently successful in capturing the passive side when the capture range was within specified
guidelines, certain specific attempts to capture failed. This resulted from either the capture range being
too large (i.e., conditions did not allow proper positioning of the interfaces) or the free-floating inertia
impacted the sides or floor of the aircraft. These two events occurred because of inherent difficulties
working and stabilizing in a micro-gravity environment or because of experiment “drift” resulting from
unfavorable flight patterns. Actual “zero” gravity time was reduced significantly when the test crew had to
adjust and limit travel of the experiment as a result of these conditions.

Figure 7. Micro-Gravity Testing of MDS on NASA’s KC-135 (“Vomit Comet”)

Computer Simulation:
The final phase of testing involved performing stiffness tests on the prototype hardware. Lateral and radial
stiffness was measured in both the deployed and captured position of the linkage assemblies. By

NASA/CP—2002-211506 275
including these measured stiffness values in the dynamic simulation model, more precise results were
possible. Results of the stiffness measurements are presented in Table 3.

Table 3. Linkage Stiffness Test Results

Configuration Compliance cm (in) Stiffness N/cm (lb/in)

Deployed Radial 0.23-0.33 (0.09-0.13) 140-147 (80-84)
Lateral 0.67-0.95 (0.265-0.375) 70-74 (40-42)
Captured Radial No Data 114-145 (65-83)
Lateral 0.46-0.62 (0.18-0.245) 58-63 (33-36)

Dynamic Analysis and Design System (DADDS) software was used to model the capture and retraction
3
features of the MDS . The model included mass properties of two satellites, zero gravity, and contacting
features of the MDS (i.e., ball ends of linkage assemblies). The model accurately simulated the
kinematics of the four-bar linkages, including dynamic reaction loads at all points of interest.

Some hand-selected cases with simple worst case relative misalignments were simulated and examined,
with the focus on identifying general issues of concern. Examination of the results of these analyses
revealed the following issues: a) a likelihood of high loads at certain MDS locations b) possible difficulty in
axial roll alignment during retraction c) possible difficulty in timely removal of pitch/yaw misalignments.

An additional 400 cases were simulated to explore performance issues using a uniform distribution of
initial condition parameters. The analysis concluded that the MDS always positively captures the passive
side given reasonable limits on initial test conditions. A case with combined misalignments demonstrating
high loads (Case 1), a case with pure roll misalignment of 5 degrees and axial offset of 9 cm (no data
provided) (Case 2), and a case of slow pitch/yaw angle removal (Case 3) are represented in Table 4 and
Figure 8.

Table 4. Simulation Test Conditions

High Load (Case 1) Initial Conditions Test Case Bounds

Relative position x= 3.0 cm, y= -2.1 cm, z= 6.4 cm (-5 < x,y < 5 cm), (5 < z < 9 cm)
Relative linear rate xd= 0.164, yd= -0.074, zd= -0.159 (-0.25 < xd,yd,zd < 0.25 cm/s)
Relative angular position ax= -2.6, ay= -2.2, az= -1.5 (-4 < ax, ay, az < 4 deg)
Relative angular rate wx= 0.064, wy= -0.080, wz= -0.097 (-0.1 < wx, wy, wz < 0.1 deg/s)
Slow Pitch/Yaw Angle Initial Conditions Test Case Bounds
Removal (Case 3)
Relative position x= 2.8 cm, y= 2.9 cm, z= 8.2 cm (-5 < x,y < 5 cm), (5 < z < 9 cm)
Relative linear rate xd= -0.215, yd= -0.054, zd= -0.152 (-0.25 < xd,yd,zd < 0.25 cm/s)
Relative angular position ax= -3.5, ay= 3.6, az= -3.8 (-4 < ax, ay, az < 4 deg)
Relative angular rate wx= 0.085, wy= -0.022, wz= -0.090 (-0.1 < wx, wy, wz < 0.1 deg/s)

3
The Boeing Space and Defense Group greatly supported SRC and the development effort by
performing these dynamic simulations.

NASA/CP—2002-211506 276
 Case 1 Case 2 Case 3
Figure 8. Misalignment Cases

Lessons Learned

An evaluation of the design and development process has provided insight into aspects of the process as
well as into potential design improvements.

• Don't “Reinvent the Wheel”: The process of thoroughly reviewing all existing hardware, trading these
against requirements, performing a truly unbiased trade study, and then designing from the output
proved invaluable. The result was optimum and has been validated through the selection of the MDS
by an Orbital Express team.

• Don't force new technologies where they aren't needed: Applying new rather than existing
technology is not always preferred. The technical risk and limitations of advanced technologies often
trade poorly against existing proven approaches. The choice to switch to a traditional motor driven
coupling system, in hindsight, was critical to the success of the system. Nitinol would have worked;
the resulting device would have been clever, innovative, but functionally limited.

• A “requirements vacuum” can be a powerful environment for forwarding technology: It allows the
designer to look ahead and derive requirements to meet future needs. It enables "products" to be
developed that have utility for a variety of users. But, under these conditions, the team must avoid
tendencies to simply design, without vision.

• Utilizing a variety of test methods to validate a design can be valuable: When designing to operate
outside gravity but testing under the influence of gravity, finding an appropriate test method is critical
to design validation. The three methods applied (off-loading, micro-gravity, and computer simulations)
all contributed to an overall understanding of the MDS. However, the application of a highly reliable
computer simulation model proved invaluable.

• Design improvements (Figure 9)

a) Incorporation of a failsafe release mechanism at the Piston/Leadscrew interface
b) Ground tabs at the linkages
c) Lateral compliance specifically designed into each linkage assembly
d) Damping at the shear pins

NASA/CP—2002-211506 277
 Safety Release Mechanism Ground Tabs (3X)

Figure 9. Design Improvements

Conclusion

The MDS prototype has demonstrated a simple, cost effective and reliable design that is scaleable. The
MDS was created in the absence of hard requirements. Although this was often a difficult and frustrating
process, it allowed the blending of technology and vision to create a very simple, yet effective, system. It
was necessary to evaluate all types of potential design solutions from a very high level with a derived
mission objective in mind. As shown through trades and system testing, the application of fewer
mechanisms and the use of common proven technology (i.e., motors, leadscrews, and linkages) can
result in a design with a very high level of functionality. Because of these characteristics, the MDS has the
potential to support a variety of future mission requirements.

NASA/CP—2002-211506 278
 Automating Separation System Reliability Testing
*
Ryan L. Perroy

Abstract

This paper presents the development, execution, and resulting test data of an automated separation
reliability test procedure for the Lightband separation system. The automated procedure was developed to
exploit the “easy-to-test” design of the Lightband and resulted in a turnaround time of ten minutes between
successive separation tests. The automated test procedure was used successfully for several different
Lightband separation systems, including a 45.72-cm (18-in) hexagonal Lightband built for the Nanosat
program and a 66.04-cm (26-in) Lightband built and recently launched for the Starshine-3 satellite
program. A concise description of the Lightband and the test set-up is presented, including discussion of
air-bearing assemblies, angular rate sensors, linear position transducer, satellite mock-up and mass
properties, instrument calibration, LabVIEW program and code, dataflow, electrical circuit hardware, and
electrical signal noise and conditioning issues. Separation reliability test results are presented, along with
a discussion of lessons learned.

Introduction

Traditional methods of determining the reliability of separation systems generally consist of suspending
the engaged system above the floor, triggering a separation event, and allowing the deployed separable
section to fall due to the force of gravity. Instruments to characterize tip-off rates are rarely used, and
measuring accurate rotational and translational separation rates may be impossible as gravitational
torques can dominate system dynamics. Traditional separation tests are generally time-consuming. It may
take several days to set and reset V-bands for separation tests, and require the oversight of several
engineers and technicians. Separation testing may also be dangerous, involving explosives and systems
at levels of tension reaching 8900 N (2000 lb). Mechanical fatigue may also be a factor in separation
reliability testing. Many V-band systems may tolerate as few as ten cycles before potentially becoming
unsafe to operate due to fatigue of the band. All of these factors combine to make traditional separation
reliability testing expensive, difficult, and infrequent. A comparison of methods of determining separation
reliability is shown in Table 1.

Table 1. Comparison of Separation Reliability Determination Methods

Method of Determining Verifies Pitch, Roll, and Yaw Translational Nullification of Statistically
Separation Reliability Separation Angular rates Velocity Gravitational Forces Meaningful Data

Automated Separation Yes Yes Yes Yes Yes

Reliability Test Fixture

Traditional Methods of Yes No No No No

Determining Reliability

The separation reliability test fixture was designed to not only test Lightband separation reliability, through
quickly triggering successive and highly repeatable separation events, but also to characterize the
separation event itself. Air-bearings were used to eliminate friction and gravitational torques, and
additional mass and outriggers were added to the fixture to match the mass properties of the planned

*
Planetary Systems Corporation, Silver Spring, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 279
satellite(s). Rate sensors were employed to measure rates in the pitch, roll, and yaw axes and a linear
position transducer was used to capture the straight-line separation motion of the unit. Data from these
instruments were read into LabVIEW for manipulation and processing. The benefits of using the
automated separation reliability test procedure are substantial when compared to traditional methods of
testing separation reliability.

Background

Lightband Separation System

The Lightband separation system is a recently developed mechanism. Its advances over existing
technology include low weight, low shock, low profile, low cost and versatility. Planetary Systems
Corporation substantially developed the Lightband through funding from a Phase I and ongoing Phase II of
a Small Business Innovative Research grant, awarded by the Air Force Research Lab. The Lightband
successfully released and spun-up the Starshine-3 satellite from the Athena (Lockheed Martin) launch
vehicle in September 2001 and is manifested on another ten missions. A US patent has been awarded
and three derivative US and international patents are pending. Figure 1 illustrates the essential
components of the Lightband separation system.

Separation Electrical Connectors Upper and Lower Rings are the De-tensioner initiates separation by
allow signal and power between mechanical interface to adjoining cutting the retaining line with
adjoining vehicles. They employ vehicles. redundant radiant heaters. They
zero-force pins for low tip-off rates. are non pyro-technic and do not
generate any gas or debris

Precision Hinged 'Leaf'

unlatches and rotates to
release. Grooves for shear
pins eliminate dependence
on friction for joint strength.

Tensioner sets the

Lightband for flight by
tensioning the retaining
line. A built-in load cell
redundantly verifies
readiness. Precision Separation Springs
impart separation velocity.
Springs can be added for more
velocity or to compensate for CG
offsets. Angled springs can
Retaining Line under tension produce spin-up if desired.
joins the Lightband rings.

Figure 1. The 38.1-cm (15-in) Diameter Advanced Lightband Separation System

The procedure for setting and operating the Lightband is straightforward. To mate the two rings, a
retaining line is brought to tension with a torque wrench and latches the hinged leaves on the lower ring
into a groove in the upper ring, as shown in Figure 2. Tension is measured by a load cell and can be

NASA/CP—2002-211506 280
adjusted at any time. To initiate separation, the de-tensioner is activated and cuts the retaining line by
radiantly heating it, as shown in Figure 3. The cut line retracts before it can burn or melt, leaving no debris
and producing no safety hazard. Redundant spring plungers at each leaf unlatch and rotate the hinged
leaves, freeing the compressed separation springs. The separation springs push the halves apart, and the
separation connectors disengage as shown in Figure 4.

Figure 2. Setting the Lightband Figure 3. Radiantly heating the

Retaining Line to Break it

Figure 4. Lightband During Separation

The Lightband’s low weight (one third of equally performing V-bands), low profile (one eighth the volume
of a V-band), low shock (an order of magnitude less than a V-band), and versatility have many benefits.
They allow taller space vehicles, longer solar arrays, larger external components in and around the
separation plane, and custom interfaces with space vehicles. The Lightband is thoroughly tested for
strength, stiffness, separation reliability, thermal vacuum, shock, and vibration, and all separation
subsystems are completely included.

Separation Reliability Fixture

Planetary Systems Corporation designed the separation reliability test fixture specifically for the Lightband
separation system. It consists of a stationary frame (emulating the launch vehicle interface), and a five-
degree of freedom displacing side (emulating the separating satellite), as shown in Figure 5. The
displacing side is composed of three main parts: a moveable arm with a frame interface identical to that
on the stationary side, an air tank, and a glass table. The upper ring of the Lightband is attached to the
displacing side and the lower ring is attached to the stationary side. Air bearings provide frictionless
motion in five axes, four planar air-bearings allowing translational motion along the glass surface in the X
and Y directions, and a spherical air-bearing allowing rotational motion of the arm in the pitch, roll, and
yaw axes. There is currently no freedom of motion in the vertical axis, preventing the fixture from having
freedom of motion in all axes.

NASA/CP—2002-211506 281
 Direction of Angle of Displacement
Separation (Exaggerated to show
Degrees of Freedom)

Separating Arm
Spherical (Lightband Upper Ring)
Air bearing Stationary Side
YAW (Lightband
Lower Ring)
ROLL

PITCH

Air Tank

Planar Air
bearings

Glass Table
X
Y
Figure 5. Separation Reliability Fixture

Separation Reliability Fixture Components

Air Bearing Assemblies

Air bearings are fluid film bearings. A thin film of pressurized air forms a layer between the load and the
contact surface, resulting in frictionless motion and no solid-to-solid contact. Air bearings have no bearing
wear and offer a much higher stiffness than ball bearings. They are also clean; using only pressurized air
rather than oil as the lubricating fluid. These characteristics give air bearings a large advantage over ball
bearings in eliminating gravitational torques. A total of five air bearings were used at two interfaces on the
separation reliability fixture to provide frictionless motion. Four planar air bearings were mounted at the
base of the air tank, as shown in Figure 6, to allow translational motion along the glass surface in the X
and Y directions. A spherical air bearing, composed of an upper hemisphere and a lower spherical air
bearing socket, as shown in Figure 7, was positioned at the tank-arm interface to allow rotational motion
of the arm in the pitch, roll, and yaw axes. The air bearings are hydrostatic, and are driven by pressurized
air at 400-550 kPa from the air tank, which is periodically refilled with a ¾ horsepower compressor.

Planar
Air Bearings

Figure 6. Planar Air Bearings

NASA/CP—2002-211506 282
 Separating
Arm

Upper
Hemisphere

Lower Spherical
Air Bearing

Air Tank

Figure 7. Spherical Air Bearing Allows Pitch, Roll, and Yaw Motion

Angular Rate Sensors

Angular rate sensors were attached to the separation reliability fixture, as shown in Figure 8, and were
oriented to measure pitch, roll, and yaw rates upon separation. Special attention was paid to the wires
connecting the rate sensors to the LabVIEW data acquisition (DAQ) program. The frictionless motion
produced by the air bearings created an environment highly sensitive to torques, both from the separation
event and from unwanted external factors. These unwanted torques included those generated by dangling
wires from the rate sensors. Thin and lightweight wires (32 AWG) were used to minimize any torques of
this type, though the decreased wire size may have contributed to an increase in electrical signal noise.
This wire choice did increase the fragility of the system, as there were accidental wire breaks that would
not have occurred with more robust wire. Increasing the size of the connection wires, and repositioning the
wires to connect vertically to the separation reliability fixture rather than from the side, may lead to better
system performance.

Pitch
Rate Sensor Yaw Sensor
not shown

Roll
Rate Sensor
32 AWG
Wires to Rate
Sensors

Figure 8. Angular Rate Sensors Measure Rates About the Pitch, Roll, and Yaw Axes

Velocity Measurement Using Linear Position Transducer

A high-performance linear position transducer was used to measure translational velocity. It was mounted
on the stationary side of the fixture and oriented along the translational path of separation, as shown in
Figure 9. The linear position transducer was connected to the separating side of the fixture by a piece of

NASA/CP—2002-211506 283
string, as shown in Figure 10. The contribution of friction within the motion of the transducer to the
separation velocity measurement was insignificant. The transducer had a range of 15.24 cm (6 in) of
travel, and was connected to the separating side of the fixture such that this range of motion was not
exceeded. The voltage signal from the transducer was connected directly to LabVIEW.

Linear
Position Separating Side
Transducer of Fixture

Engaged
Stationary Lightband
Side of
Fixture Air Tank

Figure 9. Linear Position Transducer (pre-separation event)

Lightband
Lower Ring
Lightband
Upper Ring
Extended
Linear
Position
Transducer

Figure 10. Linear Position Transducer (post-separation event)

Emulating Mass Properties

To accurately test and characterize the separation event, the mass properties of the displacing side of the
fixture were matched closely to those of the satellite to be flown. A Bi-filar pendulum was used to
experimentally determine the moment of inertia of the fixture, as shown in Figure 11. Changes to the
fixture’s moment of inertia were executed by adding an outrigger to the separating arm and varying mass,
both on the outrigger and on the air tank, as shown in Figure 12.

NASA/CP—2002-211506 284
 Thin
Filament

Pendulum
Rotation

Figure 11. Bifilar Pendulum to Measure Moment of Inertia

Outrigger for
Moment of Inertia
Additional
Mass

Figure 12. Mass Properties of Fixture

Matching the moment of inertia in the roll axis was critical for Starshine-3, as the Lightband not only acted
as a separation system but also imparted a desired orbit spin to the satellite. This spin was achieved
through angled separation springs. Determining the spring configuration necessary to achieve the desired
spin of 5 degrees per second was experimentally possible because of the closely matching moment of
inertia. Early spin-up predictions for Starshine-3, derived from angular rate data acquired when the
displacing side had a moment of inertia different from that of the satellite, were found to be inaccurate.
Simply acquiring angular rate data for a system with a known moment of inertia, and then using the
principle of conservation of angular momentum to calculate the equivalent angular rate for a system with a
different moment of inertia, should be straightforward. For reasons that are beyond the scope of this
paper, this real-world system did not follow the principle of conservation of angular momentum, and it was
necessary to match the moment of inertia of the displacing side of the fixture to Starshine-3 to accurately
predict on-orbit spin.

Balancing the Test Fixture

Because the test fixture is a nearly frictionless system, it is essential that the separating arm be exactly
balanced. Any uneven mass distribution around the pitch, roll, or yaw axes would cause drift and
acceleration due to gravity, and would compromise the experimental test data. The separating arm was
balanced through an iterative process of mass distribution. The initial balancing stage took place once the
upper ring of the Lightband was bolted onto the separating arm and the air tank was filled with
compressed air, allowing the arm to move freely. The arm was placed in a “neutral” position (roughly level
and not in contact with anything other than the spherical air bearing) and held as motionless as possible.
Upon release, the movement of the arm was observed and necessary mass distribution adjustments were

NASA/CP—2002-211506 285
made. Balance adjustments generally began with larger units of mass, such as the subtraction and
addition of steel plates, and evolved into smaller adjustments using smaller units of mass, such as
washers and bolts. Final balance refinements were achieved with Vernier center of mass adjustment
screws, as shown in Figure 13, and continued until the arm remained as motionless as possible upon
release. The arm was also placed in “tilted” starting positions (roughly level but now rolled ~45 degrees to
the right or the left) and also released and observed. Additional adjustments were made as needed.

Vernier Center of
Mass Adjustment
Screws

Figure 13. Vernier Center of Mass Adjustment Screws

Instrument Calibration
Rate sensor calibration was performed before testing. There were two methods of calibration used, a
pendulum-based LabVIEW program and a stopwatch-timed manual procedure. For the pendulum-
LabVIEW program, a potentiometer was attached to the rotation axel of a simple one-degree of freedom
pendulum, and the rate sensor to be calibrated was attached at a lower point along the pendulum as
shown in Figure 14. Data from the potentiometer, which gave the angle of the pendulum with respect to
time, were compared to data from the rate sensor, which gave the angular rate of the pendulum with
respect to time. The rate sensor data were integrated and its calibration factor was then varied to match
the output of the potentiometer. This is shown in Figure 15.

Axis of
Rotation

Pendulum
Arm Potentiometer

Rate Sensor
To be
Calibrated

Figure 14. Rate Sensor Calibration Pendulum

NASA/CP—2002-211506 286
 Rate Sensor Calibration Test, Deg/sec/Volt=25.7

15

10

5
Angle [deg]

0
0 1 2 3 4 5
-5

-10

-15
Potentiometer
Time [sec]
Integrated Rate Sensor

Figure 15. Rate Sensor Calibration Data

The stopwatch-timed procedure involves manually turning the air bearing arm at a known rate of rotation
and comparing that rate to the one recorded by the rate sensor. Both methods were used to ensure
correct rate sensor calibration. The calibration of the linear potentiometer was periodically checked by
extending the shaft at a known velocity and comparing that to the calculated speed.

LabVIEW Program and Code

LAbVIEW programs were written to automate the separation reliability procedure once the Lightband was
physically mated. These functions included verifying the readiness of all electrical connections and
instruments prior to test, initiating the separation event, characterizing the separation event through data
acquisition from the attached sensors, timing the event from initialization of the detensioner to actual
separation, and turning the detensioner off once separation took place. Programming in LabVIEW was
relatively simple, though gaining familiarity with the software and developing and debugging the data
acquisition programs took a significant amount of time. Signals from the rate sensors were directly
connected to the LabVIEW DAQ card, as was the Lightband separation connector. The benefits of using
LabVIEW to automate the separation reliability test procedure include high repeatability, automatic data
logging and analysis, low-cost data acquisition, and flexibility. Test data from LabVIEW was output to the
server and could be immediately scrutinized to verify predictions and determine which additional changes
were necessary to meet separation requirements. A system block diagram of the dataflow is included as
Figure 16.

Electronics
An electronic circuit was built to incorporate the sensors and Lightband components into an automated
data acquisition system operated in LabVIEW. The circuit allowed signals from the various sensors to be
input into LabVIEW and also controlled the separation event itself by applying voltage to the detensioner in
a prescribed sequence. Accurate timing of the separation event was determined by comparing the time
the detensioner was initiated to the time of the separation connectors lost continuity, which was read by
monitoring the separation connector signal. All of the sensors operated at a regulated 5 volts, and 12 volt
batteries were used as the original power source in an effort to reduce noise originating from the sensors’
excitation voltage. This method worked well, though it was necessary to continually check and recharge
the batteries as a weakened excitation voltage could produce false output signal levels. An opto-isolator
chip was used to isolate the power supply used to drive the detensioner from the LabVIEW input signals.
This removed any chance of the power supply damaging or interfering with the rest of the circuitry and
hardware.

NASA/CP—2002-211506 287
 12V Battery

Linear
Potentiometer
Server Database

Pitch Rate Sensor

Test Test Assy
Number Director Number
Roll Rate Sensor 5 RLP 4000340
6 RLP 4000340
LabVIEW
7 RLP 4000340
8 RLP 4000340
Yaw Rate Sensor . . .
. . .
. . .
Separation
Connector

Opto-isolator

Detensioner Power Supply

Figure 16. Separation Reliability System Block Diagram

Separation Reliability Test Results

Turnaround time between successive tests was eventually trimmed to ten minutes, resulting in the ability
to perform large numbers of tests within a relatively short timeframe. This greatly reduced the cost-per-test
and guaranteed a robust separation reliability testing regimen for the Lightband units. These factors
allowed a statistically significant number of separation reliability tests to be performed for each Lightband.
The Lightband used for the successful Starshine-3 mission was tested over 50 times on the separation
reliability fixture over the course of a month. The entire test procedure can be performed by one technician
working alone, freeing up engineers for other work. Figure 17 shows the Starshine-3 satellite and
Lightband in preparation for launch.

Data from the tests were read into a LabVIEW program and output into a comprehensive database. The
information was then scrutinized to characterize the Lightband and predict on-orbit performance. A sample
of the data collected from the Lightband used for the Starshine-3 mission is included in Table 2.
Maximum, minimum, average values, and standard deviations were calculated for each of the measured
characteristics, and were used to make adjustments to meet separation requirements and predict on-orbit
performance.

Analytical predictions of system dynamics differed considerably from test data. This is because losses in
separation energy and angular momentum due to friction and other factors were found to be substantial.
Without the test data it would have been impossible to accurately characterize the performance of the
Starshine-3 Lightband. Therefore, relying on analytical means alone for predicting on-orbit performance
would have resulted in an underestimation of the spring energy necessary to produce the required
separation velocity and angular roll rate for Starshine-3.

NASA/CP—2002-211506 288
 The Lightband for Starshine-3 (shown being mated to
LMA’s Athena) successfully spun up the Starshine 3
space vehicle. On-orbit spin rate closely matched above
test data.

Figure 17. The Lightband for Starshine-3

Table 2. Sample Data Set For Starshine 3 Lightband

On Orbit Performance Prediction
Roll Rate [deg/sec] Velocity [cm/sec]
Lightband 001 5.976 15.246
Lightband 002 5.257 14.804
Required Rate 5 (+/- 1.00) 15.24 (+/- TBD)

26 inch Starshine 3 Lightband Serial Number 001

Test Number of Initial Pitch Roll Yaw Velocity Time to

Separate
Number Spring Assemblies Tension [N] [deg/sec] [deg/sec] [deg/sec] [cm/sec] [sec]

41 10 1779 0.211 6.228 0.206 16.492 14.5

42 10 1655 0.252 5.525 -0.446 13.459 15.44
43 10 1632 -0.567 6.151 0.059 15.245 19.89
45 10 1846 0.178 5.372 0.126 14.000 18.02
46 10 1753 -0.067 6.222 -0.059 16.627 18.4
47 10 1779 -0.266 6.129 0.208 15.250 18.51
48 10 1721 -0.427 5.894 0.155 15.984 17.35
49 10 1779 -0.199 6.002 0.198 14.920 17.09
50 10 1761 -0.079 6.258 -0.22 15.237 16.26

Average 1745 -0.107 5.976 0.025 15.246 17.27

Maximum 1846 0.252 6.258 0.208 16.627 19.89
Minimum 1632 -0.567 5.372 -0.446 13.459 14.50
Standard Deviation 67 0.288 0.323 0.227 1.052 1.67

Actual on-orbit spin for the Starshine-3 satellite was measured to be 4.4 degrees per second, well within
the range of acceptable spin values (5 degrees per second [+/-1 degree]). It is also lower than the average
spin value derived from separation reliability tests (5.257 degrees per second) for the Lightband used for
flight. This discrepancy may have been caused by atmospheric differences between separation reliability
testing conditions and launch conditions. Though the lubricating quality of the atmosphere allows the air
bearings to produce a frictionless environment in which to test, this same quality may also impact

NASA/CP—2002-211506 289
separation motion in a way that is absent in a vacuum. The separation reliability fixture cannot be used in
a vacuum in its current configuration, and this limitation should be acknowledged in the design of future
test fixtures.

Discussion of Lessons Learned

There were numerous lessons learned in the process of developing the automated test procedure.
They include:
• Analytical means alone are insufficient to accurately predict on-orbit performance or determine
standard deviation. Variations in performance and hidden losses make repeated and thorough testing
the only sure method of predicting on-orbit performance.
• Mass properties of the satellite mock-up should be matched very closely to the satellite to be flown.
Equivalent mass properties allow a more accurate simulation of actual conditions than calculated results
based on the principle of conservation of angular momentum.
• Rate sensor calibration is an essential part of the testing procedure, and quick and well-understood
calibration methods are useful to make this necessary task uncomplicated.
• 12-volt batteries are a relatively noise-free power source for driving instruments like rate sensors, but
it is essential to continually check on their state of charge as low batteries may cause the sensors to
produce false and misleading signals.
• Sensor signal wires must be thin enough to reduce test interference caused by torques from dangling
wires mounted on the separating arm of the test fixture, and thick enough to prevent excessive signal
noise and fragility caused by very light and thin wires. Slightly thicker wire (26-28 AWG) will be used for
future tests.
• Basic programming principles of good documentation and easy-to-debug code have real value.
Debugging software can become a substantial drain on time and productivity and should be made as
easy as possible through well-organized and well-documented source code.

The status of the automated separation reliability test procedure is functional, though valuable
improvements abound. The LabVIEW code can be improved upon, air pressure for the air bearings
should be automatically regulated, and the instrument wiring should be repositioned on the test set-up and
generally streamlined to minimize impact and breakage. A thorough reassessment of the test fixture is
underway, and possible changes include suspending the entire test unit to allow separation reliability
testing in a vacuum and an additional degree of freedom in the vertical axis.

Conclusion

The automated separation reliability test fixture is a valuable new tool. It confirms separation, monitors and
records rotational and translational separation rates, eliminates gravitational forces and torques, and
allows a significant number of tests to be performed as a result of its short turnaround time and ease of
use. This thorough verification of a mission critical event is lacking in traditional methods of determining
separation reliability. The automated separation reliability test fixture has been used successfully for
multiple Lightband separation systems, and was used to determine the spring configuration necessary to
produce the required on-orbit spin-up rate for the Starshine-3 mission.

NASA/CP—2002-211506 290
 Random Vibration Failure Mechanism of a Conrad Bearing using Crown Ball-Retainer

Klaus Engel*, Regina Kwiatkowska**, and Mihai Vladimirescu*

Abstract

Instrument ball bearings utilizing a “snap-over” crown retainer are well known in space applications.
Although there is some anecdotal history of retainer ejection following extreme vibration or shock, such
events are rare and sparsely documented. A case study is presented in which such a bearing failure
occurred during vibration test of a rotary coaxial cable switch designed for use in a telecommunications
satellite payload. The study addresses the failure mechanism, its subsequent analysis, and the practical
resolution. The occurrence is notable because it cannot be predicted using the standard bearing selection
method based on static and dynamic load-carrying capacity.

Introduction

The paper deals with a rare type of failure specific to the bearings using crown type ball-retainer. The
crown ball-retainer is widely used in miniature bearings (deep groove-Conrad, angular-contact or
self-aligning) for space applications, because of its simplicity and low drag properties. The capabilities and
behaviors of these bearings are well understood and selection criteria are broadly accepted and
documented. Despite this familiarity we have, in a seemingly mundane application, a failure that defies
prediction.

Immediately following, we describe the switch mechanism, its supported load, the dynamic boundary
conditions that define the vibration exposure, and the special test regime adopted to validate switches for
this environment. Subsequent sections focus on the events of the test failure, evidence obtained, analysis
performed and conclusions. The analyses explore the problem in terms of conventional bearing criteria
and also examine possible failure explanations related to dynamic modes. The outcome provides some
practical advice and opportunities for further examination by bearing specialists.

Application Description
The COM DEV high power T-Switch is a discrete electro-mechanical device used to route radio frequency
(RF) signals in coaxial cable networks. The “T” designator indicates a four port device capable of
selectively connecting any port to any other port. The six possible connections are engaged two at a time,
giving the device three distinct mechanical states. The “high power” designator identifies a scaled-up
variant (approximately twice standard size) that has large internal clearances to avoid RF induced voltage
breakdown. This version has an eight year space heritage with several thousand devices operating in
orbit, including many in configurations similar to the present case.

The T-Switch has a modular design based on a unique, patented actuation mechanism (US Patent
5,499,006/Mar. 12, 1996). Figure 1 shows the operating concept and Figure 2 presents a cross-section
through the switch. Six electrical conductors in the RF module, each fitted with a small magnet, are free to
reciprocate between open circuit and closed circuit positions. The actuator module supports a rotating disc
with six magnets (magnetic cam) oriented to toggle, by attraction or repulsion according to the disc
orientation, conductors in the six possible paths. The rotary actuator resembles a brushless DC machine
with three (60° angular separation) phases. Switching is accomplished by supplying the selected motor
phase with a discrete DC command pulse. After the pulse, having acquired the correct position, the switch
remains magnetically latched in its selected state without the application of any electrical current.

* Corporate R&D, COM DEV International, Cambridge, Canada.

**COM DEV Space Group, Cambridge, Canada.
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002.

NASA/CP—2002-211506 291
 Repelling Main Bearing
Magnets
Floating
Bearing

Telemetry/Command
RF Module
Interface PCB

Rotary
Actuator

Magnet
Attracting Disc
Magnets
Figure 1: Mechanism Operation Schematics

RF Module

Actuator Stator Core

Rotating Package

Floating
Bearing

X
Z

Axial-Stroke
Y
Adjusting Shim

Main Bearing &

Snap Ring

Figure 2: Cross-Section Through the High Power T-Switch

Description Of The Supported Package

The rotating package (Figure 3) is a temperature compensated bearing system with rotating mass of
-3
42x10 kg. The permanent motor-magnet is axially offset outside the surrounding stator core to create a
net axial magnetic force of required intensity, preloading the main bearing against the snap-ring. The
magnetic axial force is proportional to the offset. This way the pre-load of the bearing is basically constant
during operation (the samarium cobalt permanent magnets, used for the rotor, are very stable over all the
useful temperature range). The direction of the axial bearing load (for the main bearing) is always from the
snap-ring toward the trailer bearing (reaction of the snap-ring). In the opposite direction only the inertia
(very small) of the balls and outer-race of the main bearing act. The floating (trailer) Conrad bearing has a
sliding fit inside the housing, which eliminates the axial stresses due to mismatched CTE’s between the
shaft material and the bearing. This design solution avoids any axial stresses due to temperature and also
maintains constant bearing pre-load.

NASA/CP—2002-211506 292
 Main Bearing Aluminum
Shaft

Floating Bearing

Crown
Retainer
Motor-Magnet

Figure 3. Rotating Package & Bearings Figure 4. Exploded View of the Main Bearing

The maximum possible axial motion of the mobile package is adjusted using axial stroke adjusting shims
interposed between the trailer bearing and its shoulder in the housing.

The main bearing is a miniature Conrad thin-section ball bearing with an outer-diameter of 33.3375 mm,
(1.3125 in), a bore of 26.9875 mm, (1.0625 in) and width of 3.96748 mm (0.1562 in). The bearing uses a
one-piece phenolic crown ball-retainer riding the outside race (Figure 4) and 14 balls. This main bearing is
situated near the center of mass facing the snap-ring with one shield (opposite side to the retainer).

Description Of The Mounting Panels

T-Switches are mainly used in redundancy ring applications. In a communications payload having multiple
transponders, the switch ring permits replacing any of a set of channels with one from a smaller set of
spare signal paths. Figure 1 shows the three switch state (A, B, and C) and a simple redundancy ring
having two spares for six channels.

3 4
2
1
Pos A
SPARE

3 4
2 SPARE
1
3 4 Pos C
1 2 3 4 5 6
1
Pos B
Figure 5. Switch Positions and Redundancy Application

Telecommunications payloads trend toward increasing size and complexity. As available mounting real
estate becomes scarce, sub-panels raised on brackets (mezzanine structures) are increasingly used. In
turn, coping with launch vibration becomes increasingly challenging. In the present case, multiple RF T-
Switches plus additional equipment are mounted on raised aluminum honeycomb flat panel structures.
Figures 6 shows a typical configuration of such a panel. The switches (as a redundancy ring) are along the
near edge of the panel. Both faces of the panel are similarly populated. Figures 7 and 8 show the
corresponding modal analysis of the whole structure. The significant natural frequencies and mode
shapes for vibration of the equipment1 contain a major mode at 100 Hz and three of four panels have a
secondary mode near 600 Hz.

1 Analysis done for frequencies up to 2000 Hz using Pro/Mechanica Structure modal analysis based on
Lanczos method of eigenvalue extraction.

NASA/CP—2002-211506 293
 Redundancy High Power T-Switch

Y
X

Figure 6. Panel Layout

Figure 7. Panel Major

Mode Z-Axes
Figure 8. Panel
Secondary Mode 590 Hz

Description of the Test Regime

The major mode (for all axes) on all honeycomb panels (around 100 Hz) coincides with the natural
frequency of the free-body motion of the electrical contacts. Due to concerns over potential fretting of
electric contacts during vibration, special vibration screening tests were devised to prove units prior to
integration. Testing was performed on a springboard fixture designed to replicate the envelope of
worst-case transmissions of the four raised panel structures in use. The springboard was therefore tuned
to achieve two peaks (around 100 Hz and 600 Hz). Figure 9 shows the computed desired springboard
output for out-of-plane axis2 (Z-axis) and figure 10 shows a typical measured response as input into the
switches under test. Similar approaches were used to define tests in the two in-plane axes. During
vibration test the rotary package was magnetically latched in one of its standard positions and after
vibration in each axis the position was verified electrically.

2 Coordinate System shown in Figure 6.

NASA/CP—2002-211506 294
 100.0 100.0

PSD
10.0 [g2/Hz] PSD
10.0 [g2/Hz]

1.0
1.0

0.1 0.1
Overall 50.2
Overall 46.5 grms
0.01
0.01

0.001 0.001
10 100 1000 2000 10 100 1000 2000
Frequency [Hz] Frequency [Hz]

Figure 9. Umbrella Switch Vibration Spec Figure 10. Resonant Fixture Response
Out-of-Plane

Bearing Failure History

Initial Fault
As the special springboard tests progressed, three switches failed to actuate after exposure. The failure
occurred, each time, after random vibration in the out-of-plane axis (Z). Disassembly of affected switches
revealed that the main bearing had separated leaving the rotating load unsupported with attendant
damage to surrounding parts. All ferrous parts were heavily encrusted with chips of samarium cobalt from
the actuator rotor magnet. Balls were distributed throughout the package and typically were found clinging
to the magnetic materials of the actuator. All balls were found and none appeared significantly deformed.
Both races experienced substantial impact damage in the grooves, however, the damage was
concentrated at the lips. Little marking was noted in the ball tracking region. Most of the race damage was
therefore attributed to hammering after some initial fault. The retainers were found on the shaft behind the
bearing plane and showed no overt sign of damage. Figure 11 shows the condition of a failed bearing.

Bearing Inner Race Crown Ball Retainer

Bearing Outer Race

Figure 11. Damaged bearing

Subsequent Faults
Additional monitoring and functional checks were added to the vibration test sequence to permit early
detection of the fault. After test resumption, two additional units showed anomalies. These units were X-
rayed and bearing faults confirmed. Figure 12 shows X-ray photographs of a normal assembly (Figure
12a) and one after fault detection (Figure 12b). In the faulted case, all balls are concentrated in an arc
segment on one side of the bearing. Although not apparent in the reproduced X-ray image, the crown
retainer is discernible as being fully removed from the bearing assembly. These two units did not incur
extended fault propagation damage and accurately represented the part condition at the time of fault.
Races, balls, and retainer were in pristine condition (see Figure 13).

NASA/CP—2002-211506 295
 Figure 12: Normal Bearing (a) Bearing After Vibration (b)

Bearing Inner Race Crown Ball Retainer

Bearing Outer Race

Figure 13. Faulted bearing with no propagation damage

It is clear that the fault initiates with the disassembly of the bearing in reverse sequence of the normal
bearing assembly process. As it may be seen in Figure 14, the failure mechanism consists of the following
three steps:
1. The bearing ball-retainer is ejected out of the bearing due to vibration.
2. The balls are therefore free to move and accumulate on one side of the race.
3. The bearing disassembles due to the free play created.

START STEP 1 STEP 2 STEP 3

Figure 14. Bearing Failure Mechanism

The investigation into the causes of this failure dealt with the following questions:
1. Is this failure related to the particular batch of bearings?
2. Is the design of the ball-retainer unsatisfactory?
3. Did the vibration-induced loads on the bearing justify the failure, and further, is the free body
resonance of the rotating package a factor?

NASA/CP—2002-211506 296
 Failure Investigation

Question 1: Is The Problem Batch Related?

All failed switches plus other non-faulted units were characterized dimensionally. All installation fits were
normal and correct. Races and balls from the two undamaged failed units were compliant to design.
Retainers from all of the failed units were examined and compared with new retainers from the same and
from different manufacturing lots. All retainers conformed to design. Particular attention was paid to the
ball pocket openings as these dimensions influence the snap-over force of the assembly. All samples
were similar in size (albeit near the wide end of the tolerance band). The ball pocket openings of the
faulted retainers showed no visible wear or deterioration and were indistinguishable from new material
even under high magnification. No defect or batch dependant issues were found. No indication of
progressive fault as might result from wear or fatigue was found.

Question 2: Is Retainer Design Deficient?

Several experiments were executed to determine the force necessary to dislodge the ball-retainer and to
assess the characteristics of a partially disassembled bearing. The force necessary to fully remove the
retainer ranged from 5 to 7 N. Given the retainer mass of 171 mg, an acceleration of 3,000 g’s would be
required to simply eject the retainer which is much higher that any calculated vibration induced condition.

In terms of partial disassembly, it was also observed that the first (lowest energy) quasi-stable dislocated
state of the retainer entailed one ball being fully disengaged from its cage pocket. Such a ball would not
likely return to its original position in dynamic conditions, but rather would tend to advance on an adjacent
ball and further disengage the cage from the bearing. In this respect small rotary oscillations around the
equilibrium position may favor the failure. The force necessary to attain this partial ejection is between 1
and 2 N and, while lower than that required for straight ejection, it is a moment force application. Such a
condition is equally as remote as the straight ejection.

Question 3: Can Vibration Induced Loads Justify The Fault?

As seen in Question 2 above, accelerations simply resulting from the high vibration inputs cannot
adequately explain the cage ejection. Regardless, there is a vibration test failure having five occurrences
in an application seemingly differing little from prior successful applications. To deal effectively with
vibration induced loads, detailed structural and bearing analyses follow.

The structural analysis uses analytical and measured data to resolve effects of cross axis coupling and
hammering thereby determining, as accurately as possible, accelerations and bearing loads. The bearing
analysis evaluates bearing deflections based on these load conditions. Together, these analyses may be
applied to describe the complex motions involved and to identify resonant conditions that may exacerbate
the fault condition.

Structural Analysis

For analysis purposes the mobile package displacements are separated into displacements due to the
free-body motion and displacements due to structural deformation.

Bearing Axial Loads

The free-body motion of the rotary package consists in a longitudinal translation and a rotation around the
bearing axis. In the case of the longitudinal translation, the simplified system describing the motion is a
single-degree-of-freedom (SDOF) mass supported on an ideal spring. The stiffness of this spring (the
magnetic axial preload from the actuator) is determined by measurement to be 192.3 N/m. Low-level sine
vibration measurements of the total package reveal that the response is dominated by the housing
response and has a minimum natural frequency of 767Hz on Y-axis (coordinate system presented in
Figure 2) and natural frequencies of 1700Hz for the other two axes. This is due to the much higher
housing stiffness and mass when compared to the mobile-package & magnetic-spring system.

NASA/CP—2002-211506 297
 Based on a well-known result when dealing with SDOF systems the response of the housing is almost
equal with the base input for low frequencies3 (around the natural frequency of the mobile package) and is
independent of the housing mounting interface damping. Therefore the input for SDOF mass-system can
be approximated by the profiles in Figures 8 and 9. For the response to Z, Y and X-axis vibrations the
coupling steady-state transmissibility function was determined based on experimental data. A special test
2
was done using a constant PSD random vibration level (0.1g /Hz) in Z, Y, and X-axes and the response
acceleration was measured directly on the mobile package in X-axis. The coupling steady-state
transmissibility function is given by [1], [2]:

 PSD X ( f )
TZX ( f ) =
 PSD Z ( f )

 PSD X ( f )
TYX ( f ) = (1)
 PSDY ( f )

 PSD Xout ( f )
T XX ( f ) = PSD Xin ( f )


The three coupling steady-state transmissibility functions are presented4 in Figure 15 and the vibration
response spectrum of SDOF-system in Figure 16 for the input on Z-axis. Similar functions were derived
for X and Y axis inputs.
10.00
[Gout/Gin] TZX 10 PSD
1 [g2 /Hz]
TYX

1.00 TXX 0.1

0.01

0.10 0.001 Overall 17.6 grms

0.0001

0.01 0.00001
10 100 1000 10,000 10 100 1000 10,000
Frequency [Hz] Frequency [Hz]

Figure 15. Coupling Steady-State Figure 16. Axial Response of

Transmissibility Function Mobile Package in Free-Body Motion

Table 1 presents the random vibration induced axial forces calculated using the peak accelerations (3σ)
corresponding to the overall grms responses from figure 16. As can be seen, all peak forces are higher
than the magnetic-spring pre-load and as a result the main bearing will unload and then will impact on the
return stroke. The impact forces can be calculated using a simplified method based on the assumption
that the shock duration is negligible when compared with the period of oscillation of the elastic bodies.
With this assumption, the impact force is calculated as follows [4]:

3 At least one octave lower than the natural frequency of the housing [3].
4 For a constant bandwidth of 4Hz.

NASA/CP—2002-211506 298
 Fimpact = ψ ⋅ F

 E0 (2)
ψ = 1 + 1 + W
 ST

Where: F=vibration induced force; Ψ=dynamic impact coefficient; E0=kinetic energy of the mobile
package; W ST=deformation energy of the snap-ring considering that F is static applied.

Integrating the PSD responses of the three axes yields overall free-body displacement responses (3σ) of
0.2050034 mm for on Z-axis input, 0.2393188 mm for Y-axis, and 3.7084 mm for X-axis input. All these
values are larger than the maximum axial stroke 0.127 mm. The calculation of the dynamic impact
coefficient is therefore adjusted using the maximum axial stroke. Considering KSR the stiffness of the
snap-ring, from equation (2) the dynamic impact coefficient is:

2 ⋅ K SR ⋅ L
ψ = 1+ 1+ (3)
F

Where: L=the maximum axial stroke;

Table 1. Random Vibration Induced Axial Forces on the Main Bearing

Input Axis Vibration Induced Peak Force [N] Impact Force [N]
X 67.27 397.2
Y 14.94 167.9
Z 21.71 206.5

The stiffness of the snap-ring is calculated [5] assuming the ring is supported at the outer edge under a
uniform load on a concentric circular-ring of radius corresponding to the mean radius of the main bearing
outer race. The calculation of KSR is presented in table 2.

Table 2. KSR Calculation5

µ t [m] a [m] b [m] r0 [m] D [N⋅⋅m] T1 [m/N] T2 [m/N] KSR [N/m]
0.25 6.350x10-4 1.739x10-2 1.515x10-2 1.630x10-2 4.547 4.006x10-7 2.368x10-7 6.106x106

The rotation around the bearing axis does not produce any load on the bearing and will not be analyzed in
detail here. It worth mentioned however that these small oscillations6 might help the retainer ejection as
pointed out in Failure Investigation section.

Structural Response Of The Supported Package

Table 3 presents the mass properties of the High RF Power T-switch mobile package (see Figure 17 for
the coordinate system).

5 In accordance with section B9 of [5]

6 They do not go beyond the unstable equilibrium point in the latching torque characteristic. This is proved
by the fact that the switch does not toggle positions during vibrations.

NASA/CP—2002-211506 299
 Table 3. Mass Properties of the T-switch Mobile Package7

X Center of Gravity X [m] Y [m] Z [m]

Z -6 -3 -6
-9.3x10 -1.4x10 2.6x10
Magnetic Spring Stiffness [N/m] 25192
-3
Mass [kg] 42 x 10
Y
Maximum Axial Force on the
397.2
Main Bearing [N]
Main Bearing Axial Preload [N] 1.76
Main Bearing Stiffnessα [N/m] Axial
6
Radial
6
35.9 x 10 13.6 x 10
α Axial Radial
Floating Bearing Stiffness 6 6
10.6 x 10 5.5 x 10
[N/m]
Figure 17. Coordinate Inertia [kg.m2] 5.123 x 10
6

System used for Modal

Analysis

Figure 18 presents the result of the modal analysis8. The first two mode shapes correspond to frequencies
(411.4 Hz and 414.5 Hz) inside the range of interest (between 20 and 2000 Hz) and are in longitudinal
direction. Due to large diameters, the first transversal mode occurs only over 2,000 Hz.

Major Mode X-axis Secondary Mode Major Z-axis

411.4 Hz X-axis > 2,000 Hz

Figure 18. Modal Analysis Results for the Structural Response of the Mobile Package

Experimental measurements show the major modes at 331.5 Hz on X-axis, 695.3 Hz on Z-axis and
respectively 873.3 Hz on Y-axis (using the coordinate system presented in figure 17). The difference in the
transversal and lateral values of the calculated and measured natural frequencies is determined by
bearing stiffness. It may be concluded from the above that the axial free-body motion of the mobile
package generates the largest axial load on the bearing and the stiffness of the bearings determines the
loads in transversal and lateral directions9.

Bearing Radial Loads

Radial loads are calculated according to the method used for axial loads. Experimentally10 derived
steady-state transmissibility functions are superimposed to determine for lateral and transverse vibration
responses. The peak radial forces (3σ) are determined to be 66.3 N with Z-axis input and 56.3 N from the
Y-axis.

7 Calculated using Pro-Engineer software.

α Calculated with ESTL CABARET around peak axial and radial loads (X-axis).
8 Assuming infinite rigidity of the bearings
9 The shaft package behaves as totally rigid.
10 Using a similar equation to equation (1).

NASA/CP—2002-211506 300
Table 4 summarizes the bearing loads.

Table 4. Calculated Loads on the Bearings

Load Direction/ X-axis Y-axis Z-axis
Vibration axis Vibrations Vibrations Vibrations
AXIALβ [N] 397.20 167.90 206.50
RADIAL Main 13.02 48.54 70.69
[N] Floating 2.08 7.76 11.31

Bearings Analysis

The Bearing Analysis section is based on the calculation flow described by [7] and other industry
practical recommendations. For clarity all parts not relevant to the problem at hand11 (i.e., life analysis,
cage rotation instability, friction and lubrication, temperature effects, etc.) were omitted.

Static And Dynamic Load-Carrying Capacity12

The basic static load-carrying capacity is given by:

C 0 = S 0 ⋅ (X 0 ⋅ FRADIAL + Y0 ⋅ FAXIAL ) (4)

Where: C0=static load-carrying capacity; S0=static safety-factor; X0=static radial-factor; Y0=static

axial-factor; FRADIAL=static radial-load on the bearing; FAXIAL=static axial-load on the bearing.

The dynamic load-carrying capacity can be calculated with a formula similar to (4) in which the terms in
brackets (equivalent static load) are replaced by the equivalent dynamic load-rating. A catalog bearing
similar with the main bearing has rated loads: FRADIAL=858.50 N, FAXIAL=2166.28 N and a dynamic
load-carrying capacity of C=814.02 N. C0CATALOG=1598.24 N and worst-case main bearing based on data
in table 4 (assuming a safety factor13 of 2) C0X-AXIS= 412.82 N or a static safety margin of +287.15%. For
the dynamic load-carrying capacity assuming that the same loads will act on the bearing during operation,
the worst-case dynamic load-carrying capacity is CX-AXIS= 507.76 N or a safety margin of +60.31%. It
results that the bearing has large safety margin versus the loads.

Advanced Bearing Analysis

The results of the advanced bearing analysis (performed using CABARET and A. B. Jones software) are
summarized in table 614. It is assumed that the entire load applied by the rotor is distributed evenly among
the ball contacts of the main bearing. The trailer bearing is considered to be unloaded. Moment loads will
be insignificant since the trailer bearing limits rotation of the shaft about the main bearing center. The
software for analysis of the bearings calculates the number of balls that carry significant loads. The most
heavily loaded ball is considered for the worst-case Hertzian stresses. The analysis was performed for the
worst-case measured impact load of 397.2 N in longitudinal direction, and for 206.5 axial-load
corresponding the axis (Z-axis) on which the failure occurred.

β Axial forces for the main bearing. Radial forces calculated as vector sum of the two
corresponding forces on Y-axis and respectively Z-axis.
11 The failure was under vibration and in latched position.
12 This type of selection calculation is specific for the case where bearings are selected from a catalog.
In our specific case the bearings are custom build, nevertheless we present a calculation based on
similar catalog bearings to show that this is not capable of catching this type of failure.
13 Ball, quiet running, normal loading [7].
14 For A.B. Jones analysis both bearings were considered.

NASA/CP—2002-211506 301
 Table 5. Advanced Bearing Analysis Results

Load Condition Load Max. Mean Max.

Applied Hertzian Stress Peakµ Hertzian Stress
[N] [MPa] [MPa]
CABARET A. B. CABARET A. B. Jones
Jones
X-axis Radial 13.02
2235.34 1135.57 3353 1703.35
Axial 397.2
Z-axisα Radial 70.69
2048.00 1794.91 3072 2692.4
Axial 206.5

All of the cases studied predict stresses below the conservative guideline for quiet operation15, with
margin.

Failure Discussion and Practical Solution of the Problem

Facts
The bearings failed (according to the mechanism described) during vibrations on the out-of-plane axis (Z-
axis). The calculated loads on the bearings are lower than the load-carrying capacity of the bearings and
the maximum mean stresses are within the specified limits by MIL-A-83577B, Para.3.2.3.1.2. The
calculated natural frequency of the free-body motion of the mobile package is 123 Hz far away from the
600 Hz input component. However the structural analysis showed that Z-axis is the second more loaded
vibration case and is the case with the maximum radial loads. The bearings advanced analysis showed
that the calculated axial deflections are 21.71 x 10-3 mm for X-axis and 14.81 x 10-3 mm for Z-axis input
and the calculated radial deflections are 0.96 x 10-3 mm for X-axis and 6.96 x 10-3 mm for the Z-axis. This
is an increase of more than 7 times of the radial deflections for the Z-axis case. This is because a
transversal resonance is located between 600 and 700 Hz and in addition, the coupling steady-state
transmissibility function has a peak in the same region. As a result the axial response on Z-axis vibrations
(figure 16) also shows higher overall grms driven by a higher PSD in the 600 Hz region, higher even than
the PSD corresponding to the 100Hz region (the free-body motion natural frequency region).

Discussion
It seems that the failure is therefore related to a combination of high radial deflections and high overall
vibration levels due to the 600 Hz modes on Z-axis. A structural analysis of the bearing itself was carried
out to determine the structural deformations modes shapes of the bearing. The model consisted in a solid
inner-race (this is trough because of the increase in the stiffness of the inner-race due to the shaft) with a
centered protrusion that ensured that the overall center of mass coincides with the of the center of mass of
the assembly mobile package. The supplementary mass idealization element added the necessary mass
so the model behaves similar to the assembly mobile package. The outer-race was considered supported
on the face that in reality touches the snap-ring. Due to the high stiffness of the snap-ring this was
considered totally rigid. The entire compliance of the bearing was located in the balls, which were modeled
as springs oriented on all three axes. Each of the spring was located between two points one on the
inner-race groove and the other on the outer race-groove with the stiffness derived from table 3. The
outer-race was not constrained on the outer-diameter in order to simulate the assembly clearance. A
standard modal analysis was done to determine all modes inside the frequency interval from 20 Hz to
2000 Hz. The analysis found 6 modes inside this interval: 451.2 Hz, 622.9 Hz, 736.0 Hz, 1082.0 Hz,
1164.5 Hz and 1721.7 Hz. There is good correlation between these modes and the experimental data

µ Max. Peak Hertzian Stress is considered 1.5 Max. Mean Hertzian Stress – elliptical stress distribution
assumption.
α Calculated using A.B. Jones Bearing Analysis Software Version. 5.4.3
15 2309 MPa (335 ksi) for mean Hertzian stresses based on the specification MIL-A-83577B,
Para.3.2.3.1.2 for mechanisms requiring quiet operation and low torque ripple.

NASA/CP—2002-211506 302
used for the structural analysis. Figure 19 shows these results for the first 4 modes of vibration considered
important for the range of input frequencies and a combination of all the four vibration modes.

622.9 Hz
451.2 Hz 736.0 Hz

1082.0 Hz Combined
Figure 19. Major Vibration Modes of the Bearing

The implication of the vibration analyses, with respect to the bearing fault (question 1 in the Failure
Analysis section above), is that the free-body natural frequency (123 Hz) has no effect on the failure. The
failure is considered the result of a combination of modes resultant of a peak in the steady-state
transmissibility function for the out-of-plane axis near 600Hz. Importantly, these modes are determined by
the bearing system itself.

Practical Solution
The important fact for the program at hand was the identification of the correlation between the failure and
the 600 Hz components. Recall that the test regime was a special test at device level against an artificial
spec and was further exacerbated by the high Q of the springboard fixture. The findings allowed further
investigation at payload level and eventually it was possible to avoid the 600 Hz component entirely. Once
the component was no longer over-excited the failure disappeared allowing the program to be finished in
time.

It is possible that the bearing behavior may be improved by increasing the number of balls that will have
as result a change of the natural frequencies depicted in Figure 19 towards higher values. It is also
possible that a different retainer pocket design (such as a keyhole profile) will prove more robust to this
kind of failure. This, however, is purely intuitive and requires research. An important observation is that the
ejection was purely mechanical, without wear or deterioration and would have been contained had there
been shoulder or shield structures to impede the retainer.

NASA/CP—2002-211506 303
 Conclusion and Lessons Learned

1. Ejection of crown retainers under vibration is a potential failure mode of bearings.

2. Classical bearing selection methods based on static and dynamic load carrying capacity and standard
data provided by bearing suppliers do not predict this type of failure, nor is there substantial literature
on the subject.

3. Caution must be exercised when such bearings are used in high vibration environments. The correct
vibration profile is important in proving the reliable operation of the mechanism.

4. In analyzing for high vibration environments, bearing properties may be significant.

5. The risk of retainer ejection can be mitigated in many cases by extending shaft/housing shoulders
beyond the bearing land. In most designs, stable dislocation or irreversible ball migration require the
retainer to displace beyond the frame end face.

6. There is merit in future investigation by bearing specialists to identify robust retainer designs and
application criteria.

References

1. Meirovich, L. Analytical Methods in Vibrations. Macmillan: New York, ©1967.

nd
2. Thomson, W. Theory of Vibration with Applications 2 Edition. Prentice Hall: New Jersey, ©1981.
3. Steinberg, David S. Vibration Analysis for Electronic Equipment. Wiley-Interscience: New York, ©1988.
4. Föppl, A. Vorlesungen über techniche Mechanik. Band 1-3, Oldenburg Verlag: München, ©1949-1951.
5. NASA TMX-73306 Astronautic Structures Manual. Vol. 2, George C. Marshall Space Fight Center,
Structures and Propulsion Laboratory, 1975, section B9.
6. MIL-HDBK-5H Metallic Materials and Elements for Aerospace Vehicle Structures. DOD: Washington,
1998.
7. Design Guide Series, Rolling Bearings, Version 1.0 from Multi-media Handbook for Engineering Design,
Design Information Group, University of Bristol, UK.
8. Tedric A. Harris, Rolling Bearing Analysis. John Wiley & Sons: New York, ©1984, First and Second
Edition.
9. E. G. Gargiulo jr., E. I. DuPont de Nemours and Co. Wilmington DE., Machine Design, July 24, 1980.

NASA/CP—2002-211506 304
 Comparison of Several Different Sputtered Molybdenum Disulfide Coatings
for Use in Space Applications

Robert L. Fusaro∗ and Mark Siebert∗∗

Abstract

Tribology experiments on different types of sputtered molybdenum disulfide (MoS2) coatings (obtained
from different vendors) using accelerated testing techniques were conducted. The purpose was to
determine which would be the best coating for use with auxiliary journal bearings for spacecraft energy
storage flywheels. Experiments were conducted in moist air (50% relative humidity) and in dry air (<100
PPM water vapor content) on a Pin-on-Disk Tribometer to determine how well the coatings would perform
in air. Experiments were also conducted on a Block-on-Ring Tribometer in dry nitrogen (<100 PPM water
vapor) to simulate how well the coatings would perform in vacuum. Friction, counterface wear, coating
wear, endurance life and surface morphology were investigated.

Introduction

NASA Glenn is currently developing magnetic bearings to be used for levitating energy storage flywheels
for the International Space Station and for satellites. To insure safety (if magnetic bearings should fail)
and to prevent damage from “bumps,” mechanical auxiliary bearings must also be developed for this
application. Several different types of mechanical bearings are being considered as well as several
different lubrication systems. If solid lubricants are selected, the one with the longest endurance life with
reasonable friction and wear properties in a vacuum environment is the most desirable. However, many
of the MoS2 based lubricants being considered do not work well in ambient air. It is possible that exposure
or mishandling in air might reduce the life or performance solid lubricant used for an auxiliary bearing.
Thus it is desirable to choose a solid lubricant that works well under all environmental conditions.

Sputtered MoS2 coatings were chosen for this study because they have been shown in many previous
studies to be excellent lubricants in a vacuum environment [1-7]. The problem is that MoS2 oxidizes in air
[8-15] and can lead to damage to the coatings before they even get into space. Recently, new sputtered
MoS2 coatings have become available and have been tested [16-20] that are co-sputtered with various
materials that improve their performance in air.

In order to help determine which of these coatings might be the best for this application, an accelerated
testing program was developed to evaluate their tribological properties of these coatings under different
environmental conditions. This paper deals with tribological accelerated tests on several different
sputtered molybdenum disulfide (MoS2) coatings in a 50 percent relative humidity (∼10,000 PPM moisture
content) air atmosphere and in a very dry air atmosphere (<100 PPM moisture content) using a Pin-on-
Disk Tribometer. In addition, the same coatings were also evaluated in a dry nitrogen atmosphere (<100
PPM moisture content) using a Block-on-Ring Tribometer to simulate a vacuum condition.

Materials

Six different sputtered MoS2 coatings were evaluated that were supplied from 5 different vendors. Table 1
lists the vendors and the additives in the films. The coatings were applied to the disks of the Pin-on-Disk
Tribometer and to the rings of the Block-on-Ring Tribometer. A few Block-on-Rings tests were also
conducted with blocks that were coated with the CSEM-Ti or CSEM-Al coating. The disks, blocks and
rings used in this study were made of 440C stainless steel with a Rockwell hardness of C-57 to C-59. The
disks were lapped and polished to a surface finish of 0.040 ± 0.015 µm centerline average (CLA). Instead
of using pins in the Pin-on-Disk Tribometer, the pin holder was modified to hold and constrain from rolling

∗
NASA Glenn Research Center, Cleveland, OH
∗∗
University of Toledo, Toledo, OH
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 305
a 0.476 cm radius (3/8 inch diameter) commercial grade 10, 440C ball that had the same surface
roughness as the disks. The hardness of the balls was Rockwell C-60.

Table 1. Types of MoS2 Coatings Evaluated in this Study.

Vendor’s Name Designation Major Additive Coating Thickness(µm)

Movic Movic None 0.6
Surftech Surftech None 0.3
Hohman Plating Hohman Antimony Trioxide 1.2
Teer Teer-Ti Titanium 1.2
CSEM CSEM-Ti Titanium 2.4
CSEM CSEM-Al Aluminum 3.5

Testing Apparatus

Pin-on-Disk Tribometer
The pin-on-disk Tribometer used in this study (Figure 1) has been described in detail in Reference 21.
The specimens consisted of a flat rotating disk (6.3-cm diameter) in sliding contact with a stationary ball
(0.476-cm radius) that was securely fastened in a holder. The ball slid on disk tracks that ranged from 6.0
cm to 4.4 cm in diameter. The rotational speed of the disk was controlled at 200 rpm giving linear sliding
speeds of 0.63 to 0.46 m/s. The test specimens were encased in a plastic box to control the atmospheric
moisture content. The load of 9.8 N (1 kg) was applied to the ball by a dead weight using a lever arm
system. A strain gage was used to monitor and measure the frictional force.

Wear volume of the ball was determined by measuring the change in diameter of a wear scar on the ball
and then calculating the volume of material removed. Wear volume of the disk was determined by
measuring the wear track cross-sectional area using a surface profilometer and then calculating the
volume of material removed. A discussion on how to evaluate solid lubricants in a pin-on-disk tester is
given in the next section and in the reference by Fusaro [21].

Figure 1. Schematic of Pin-on-Disk Tribometer

Block-on-Ring Tribometer
A schematic of the block-on-ring test elements is shown in Figure 2. As shown in the figure, the device
consists of a rectangular block (0.6 cm wide x 2 cm long x 1 cm high) pressed against the periphery of a
ring (1 cm wide x 5 cm diameter). The block can be flat (line contact) or it can be conforming (area
contact). In this study only line contact was used.

NASA/CP—2002-211506 306
The block and the ring used in this study were made from 440C stainless steel. The block was stationary
and loaded with a dead weight against the ring. The ring was attached to a rotating shaft that can rotate in
one direction. A probe attached to the block holder contacts a load transducer and measures frictional
force between the block and the rotating ring. A thermocouple is imbedded near the contact area of the
block to measure temperature.

The coating was applied to the contact area around the outside diameter of the ring and to the (0.6 x 1.5
cm) face of the block. In all cases, the surface roughness of the block is very important and can influence
the results. To most closely reproduce the end-use application, the roughness should closely match that
-6
value. In this case, the surfaces were very smooth (0.05 to 0.10 X 10 m Ra). The sliding conditions for
o
the block-on-ring test were as follows: sliding speed, 500 rpm, load, 225 newtons, temperature, 25 C.

Figure 2. Schematic of Block-on-Ring Tribometer Specimens

Procedure

Surface Preparation and Cleaning Procedure

The cleaning procedure for the specimens before they were sent to the vendor was as follows:
1) Scrub surface under running water with a bottle-brush to remove abrasive particles.
2) Clean surfaces with pure ethyl alcohol using a lint-free cloth.
3) Scrub surface with a water paste of levigated alumina. Clean until water wets the surface readily.
4) Rinse the surface under running water to remove the levigated alumina (using the brush to facilitate
removal).
5) Rinse the surface in distilled water.
6) Dry surfaces using dry compressed air. (Surfaces not dried quickly have a tendency to oxidize.)

Pin-on-Disk Testing Procedure

The procedure for conducting the pin-on-disk experiments was as follows: a pin (ball) and a disk (with
applied sputtered MoS2 coating) were inserted into the Tribometer test chamber (Figure 1). The test
chamber was sealed, and dry air (<100-ppm H2O), or moist air (~10,000-ppm H2O) was purged through
the chamber for 10 minutes before starting the test and then continuously throughout the test. When the
purge was completed, the disk was set into rotation at 200 rpm. A 1-kilogram (9.8 N load was then
applied to the disk as it rotated.

Some preliminary experiments were conducted to determine the friction characteristics of unlubricated
440C stainless steel sliding on itself. From those results, it was decided to make the criterion for failure for
these tests to be a friction coefficient of 0.30, much less than the friction coefficient of unlubricated 440C
stainless steel (>0.60). An automatic cutoff system was used to shut down the apparatus when the friction
coefficient reached 0.30.

NASA/CP—2002-211506 307
Two types of friction and wear testing procedures were followed: (1) the “continuous testing method” and
(2) the “interval testing method.” In the continuous testing method, the test was run continuously until a
friction coefficient of 0.30 occurred. The specimens were removed from the Tribometer and the wear
scars were measured using an optical microscope and the coating wear was measured using a surface
profilometer. The visual microscope was also used to evaluate the morphology of the sliding surfaces at
magnifications to 3000X. The number of revolutions to reach this value of friction coefficient was defined
as the endurance life of the coating. In the “interval testing method,” the specimens were removed from
the test chamber at predetermined intervals of sliding and the specimen contact areas were evaluated as
in the continuous testing method. The specimens were then placed back into the chamber and the
previous test procedure is repeated. Sliding continued until a friction coefficient of 0.30 was obtained. The
advantage of the interval method is that wear as a function of sliding distance can be determined and that
the type of wear occurring on the surfaces before failure can be studied. In continuous testing, only wear
at the end of a test can be determined and run-in wear cannot be separated from steady-state wear. One
caveat on the “interval testing method” is that care must be taken to replace the specimens with the same
orientation and alignment that they had before they were removed. Lives from both methods were nearly
identical in air, although a slight friction “run-in” occurs at the start of each interval test.

Block-on-Ring Test Procedure

The specimens were inserted into the apparatus and the chamber sealed. The chamber was then purged
with the nitrogen before starting the test for 10 minutes and then continuously throughout the test. This
procedure was repeated each time a test was stopped until the test was completed. Both the continuous
testing method and the interval testing method (as described above) were used for these experiments but
they were not stopped as many times as for the pin-on-disk tests. This was done because of concern that
the surfaces may have experienced oxidation degradation during the time they were removed.

Results and Discussion

Pin-on-Disk Friction Coefficient

Friction coefficient for each pin-on-disk experiment was constantly monitored though out the tests. Table
2 and Figure 3 present the steady-state friction coefficients for all coatings tested in both moist air and dry
air. The lowest friction coefficient of 0.04 was obtained in dry air for 4 of the films, the Movic, the Surftech,
the Hohman and the Teer-Ti coatings. The CSEM-Ti coating produced a friction coefficient of 0.06 and
the CSEM-Al coating produced a friction coefficient of 0.08.

0.16
Dry Air
Friction Coeffient

0.14
Moist Air
0.12
0.1
0.08
0.06
0.04
0.02
0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al
Type of Sputtered MoS2 Coating
Figure 3. Average Steady State Friction Coefficient for different sputtered MoS2 films (obtained
from various vendors) and tested in moist air and dry air on a pin-on-disk Tribometer.

The Surftech coating also produced a steady-state friction coefficient of 0.04 in moist air, but all the other
coatings produced much higher friction coefficients in moist air than dry air. In moist air, Movic produced
the highest friction coefficient of 0.15, the value for the Hohman and the CSEM-Al coatings was 0.12, the

NASA/CP—2002-211506 308
CSEM-Ti coating was 0.11 and the Teer-Ti coating was 0.09. For vacuum applications, the friction
coefficients in dry or moist air are not really relevant except for the fact that higher friction coefficients
usually mean higher wear and also shorter lives. It may also be useful information for designing
components that might have to operate both in air and in vacuum. With MoS2 coatings in general, friction
coefficient is much more affected by water vapor than oxygen, thus friction in dry air more closely
approximates the friction that would be obtained in vacuum.

Table 2. Pin-on-Disk Test Results

MoS2 Steady-State Endurance Life Counterface Coating
Coating Friction (Kilocycles) Wear Rate Wear Rate
3 -18 3 -16
Coefficient (m /m x 10 ) (m /m x 10 )
Moist Dry Moist Dry Moist Dry Moist Dry
Air Air Air Air Air Air Air Air
Movic 0.15 0.04 7+3 120 + 17 700 + 300 4+2 --- ---
Surftech 0.04 0.04 39 + 6 41 + 7 15 + 12 3+1 --- ---
Hohman 0.12 0.04 76 + 46 100 + 26 145 +135 24 + 6 --- ---
Teer-Ti 0.09 0.04 55 + 10 361 + 15 60 + 20 3+1 107 + 35 16 + 8
CSEM-Ti 0.11 0.06 625 + 100 1440 + 440 30 + 10 3+1 44 + 6 11 + 5
CSEM-Al 0.12 0.08 400 + 52 1248 + 205 300 +120 3+2 113 + 55 22 + 8

Pin-on-Disk Endurance Lives

The average endurance life and variation for each coating in dry air and in moist air is given in Table 2,
and the average is shown in bar graph form in Figure 4.

1600
1400 Dry Air
Endurance Life, kc

Moist Air
1200
1000
800
600
400
200
0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al
Type of Sputtered MoS2 Coating
Figure 4. Average Endurance Life for different sputtered MoS2 films (obtained from various
vendors) and tested in moist air and dry air on a pin-on-disk Tribometer.

In general, most of the coatings had longer endurance lives in dry air than in moist air. The Surftech
coating produced equivalent lives in both atmospheres, but the lives in both were very short (~40 kc). The
Movic coating had the shortest life in moist air (7 kc) but a life of 120 kc in dry air. The Homan coating had
an average life of 76 kc in moist air and 100 kc in dry air, which was equivalent to the Movic coating in dry
air. The Teer-Ti coating had an average life of 55 kc in moist air but an average life of 361 kc in dry air.
The CSEM-Ti and CSEM-AL coatings gave longer lives in moist air that any of the other coatings in dry
air and were even longer in dry air. From an endurance point of view the CSEM-Ti and CSEM-Al coatings
were far superior to the others in either moist or dry air, but they were also the thickest.

Some discussion of the Teer-Ti and CSEM-Ti coatings is appropriate at this point. The process for
producing sputtered MoS2 coatings that contain titanium was invented by Teer Coatings Limited and they

NASA/CP—2002-211506 309
have the patent rights to the coating. CSEM has obtained a license to also make the coating using the
Teer process; therefore the two coatings are basically very similar. As far as could be ascertained by the
authors, the basic difference between the two coatings is that the CSEM-Ti coating contains more
titanium than the Teer-Ti coating. In addition, the CSEM-Ti coating is twice as thick.

Pin-on-Disk Counterface Wear

The variation of wear to the counterfaces (440C balls) sliding against the various coatings is given in
Table 2 and the average in Figure 5. In general, much lower wear to the counterfaces occurred in dry air
than in moist air. The exception was the Surftech coating that produced low wear in both atmospheres,
but part of that was due to the fact that lives were very short and not much actual sliding occurred. Except
for the Hohman coating, the counterface wear of all coatings was equivalent in dry air. The overall best
combination of low counterface wear in both dry and moist air occurred with the CSEM-Ti coating.
800
Ball Wear Rate M3/M x 10-18

700
600
Dry Air
500 Moist Air

400
300
200
100
0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al
Type of MoS2 Coating
Figure 5. Average 440C steel counterface wear for different sputtered MoS2 films (obtained
from various vendors) and tested in moist air and dry air on a pin-on-disk Tribometer.

Pin-on-Disk Coating Wear

The Teer-Ti, CSEM-Ti and CSEM-Al coatings were strong enough to support the load and wore gradually
away until the substrate was reached. Coating wear rate was not measurable for the Movic, Surftech and

4000
Cross Sectional Area,

3500
Dry Air #1
3000 Dry Air #2
(m2 x 10-12)

Moist Air #1
2500 Moist Air #2
2000
1500
1000
500
0
25

50

00

00
10

20

30

40

45

60

90

10

11

Sliding Distance (kilocycles)

Figure 6. Coating Cross-Sectional-Area as a function of sliding distance for 440C steel sliding
against sputtered CSEM-Al MoS2 coatings in moist air and dry air on a pin-on-disk Tribometer.

NASA/CP—2002-211506 310
Hohman coatings, since these coatings were either too soft or too thin to support the sliding 440C ball
counterface. These coatings were worn away very quickly to a thin film that provided the lubrication. See
Fusaro [21, 22] for a discussion on the mechanisms of solid coating lubrication. Figure 6 presents Cross
Sectional Area of the coating wear as a function of sliding distance in kilocycles for 440C steel sliding
against sputtered CSEM-Al MoS2 coatings and tested in moist air and dry air. Two tests are shown for
each condition. The figure shows that the wear rate is fairly reproducible and that the wear rate is nearly
constant in dry air. In moist air, between 50 and 100 kc, the rate increased, but then was nearly constant.

The variation of the coating wear rates is given in Table 2 and the averages in Figure 7 for the three
coatings that were able to support the load. Again the dry air atmosphere provided the lowest wear rates.
-15 3
The CSEM-Ti had the lowest average wear rate of 7 x 10 m /m, the next lowest was the Teer-Ti coating
-15 3 -15 3
(10 x 10 m /m) followed by the CSEM-Al coating (14 x 10 m /m). It thus appears that the primary
reason for the increased endurance life of the CSEM-Ti coating compared to the Teer-Ti coating was due
to the fact that the CSEM coating was much thicker and it took longer to wear through.
80
Dry Air
70
Film Wear Rate,

Moist Air
60
m3/m x 10-16

50
40
30
20
10
0
Teer-Ti CSEM-Ti CSEM-Al
Type of MoS2 Coating
Figure 7. Average coating wear rate of different sputtered MoS2 coatings (obtained
from various vendors) and tested in moist air and dry air on a pin-on-disk Tribometer.

Like friction coefficient, endurance life and counterface wear, the coating wear rate was greater in moist
-15 3
air than in dry air. The lowest wear in moist air was the CSEM-Ti coating, which was 28 x 10 m /m, but
was about 4-times higher than what was obtained in dry air. The other two coatings gave 2 to 3 times
higher wear rates than the CSEM-Ti coating in moist air.

Pin-on-Disk Coating Morphology

In accelerated testing, statistical analysis of test data can provide numerical comparisons between
coatings. But in order to obtain a more complete understanding of which coating would be best for your
application, a simple technique like optical microscopy can be used to help in the evaluation. To do that,
surfaces must be evaluated before failure. Therefore another reason for interval testing is to observe the
rubbing surfaces with an optical microscope to magnifications as high as 3000X.

Figure 8 shows photomicrographs of the Movic coating wear tracks produced in dry air after 23 kc of
sliding and in moist air after 2 kc of sliding. Dark blisters can be observed on the dry air wear track and
dark powdery third body material (third body material is either wear material or decomposed wear
material that remains on the wear surfaces) can be found in the center of the moist air track. Similar
surfaces were found with the Surftech coating. These results are very similar to the results of previous
studies that were conducted by Fusaro [13] on burnished MoS2 films that showed that the MoS2 burnished
film oxidized to form MoO3 in air and that the water vapor in the atmosphere accelerated this process.

Figure 9 gives photomicrographs of the wear tracks on CSEM-Ti sputtered MoS2 coatings tested in dry air
after 660 kc of sliding and in moist air after 160 kc of sliding. Both surfaces are very smooth with no
indication of decomposition in either atmosphere unlike results from the Movic and Surftech tests. Fine

NASA/CP—2002-211506 311
powdery debris can be seen outside of the wear track areas. In the moist air tests, there was some back-
transferred third body material that may have been the cause of the higher wear in moist air.
Figure 10 shows a high magnification photomicrograph of the sputtered CSEM-TI MoS2 coating after 900
kc of sliding showing an area that has been worn through to the substrate. A very thin secondary film has
formed from third body material in this area that has prevented metal-to-metal contact.

Figure 8. Photomicrographs of Movic wear tracks after 23 kc of sliding in dry air and 2 Kc of
sliding in moist air on a Pin-on-disk Tribometer.

Figure 9. Photomicrographs of CSEM-Ti MoS2 wear tracks after 660 kc of sliding in dry air
and 160 kc of sliding in moist air on a Pin-on-Disk Tribometer.

Figure 11 shows photomicrographs after 650 kc of sliding in dry air and after 350 kc of sliding in
moist air of the sputtered CSEM-Al MoS2 coating wear tracks. In dry air, as shown in the figure,
the track is very smooth; however there are areas where brittle fracture has occurred. In moist
air, small surface pits can be seen on the track; and in addition, brittle fracture has also
occurred but is not seen in the area shown on this figure.

Pin-on-Disk Counterface Morphology

Transfer films to the 440C ball counterfaces occurred for most of the sputtered MoS2 coatings
that were evaluated in this study. They were generally characterized by a buildup of material in
the inlet area, thin flowing transfer across the scar flat and then powdery debris in the exit area.
Figure 12 shows a photomicrographs of the transfer to a 440C stainless steel ball that slid
against a CSEM-Al sputtered MoS2 film. This type of transfer is typical of most of the tests in
dry air and moist air on all the films. There was one exception and that was sputtered CSEM-Al
in moist air. Figure 12 also shows a photomicrograph of the transfer to the 440C ball that slid

NASA/CP—2002-211506 312
against the CSEM-Al MoS2 coating in moist air after 50 kc of sliding. In this case, there is a
small buildup of material in the inlet area but a minimal amount of transfer to the scar itself.

Figure 10. Photomicrograph of CSEM-Ti MoS2 wear tracks after 900 kc of sliding in dry air on a
Pin-on-Disk Tribometer showing a thin area where original coating has worn away.

Figure 11. Photomicrographs of CSEM-Al MoS2 wear tracks after 650 kc of sliding in dry air
and 350 kc of sliding in moist air on a Pin-on-Disk Tribometer.

Figure 12. Photomicrographs of the wear scars on the 440C stainless steel ball
counterfaces against CSEM-Al sputtered MoS2 coatings on a Pin-on-Disk Tribometer.

NASA/CP—2002-211506 313
Block on Ring Friction Coefficient
The average steady-state friction coefficients for the tests conducted in dry nitrogen on the Block-on-Ring
Tribometer tests are given in Table 3 and also in Figure 13. In addition to non-coated blocks, a few tests
were also conducted with coated blocks. In general friction coefficients were lower in dry nitrogen on the
Block-on-Ring Tribometer than those obtained in dry or moist air with the Pin-on-Disk Tribometer. The
Movic and Hohman coatings gave the lowest friction coefficients of 0.01. The values found for the other
coatings were: CSEM-Ti, 0.02, Teer-Ti, 0.03, Surftech, 0.04 and CSEM-Al, 0.07. The coated blocks
sliding on the coated rings did not change the steady-state value of the friction coefficients obtained as
compared to the non-coated blocks.

Block-on Ring Endurance lives

The variation of coating endurance lives for the tests conducted on the Block-on-Ring Tribometer are
given in Table 3 and the averages are given in Figure 14. The longest endurance life for the Block-on-
Ring Tribometer tests were obtained with the Hohman MoS2 coating, although there was considerable
variation. The longest average life was obtained for the CSEM-Ti coating tested against a CSEM-Ti
coated block. A coated block versus an uncoated block increased the average life of the CSEM-Ti coating
from 323 kc to 6132 kc. The Teer-Ti sputtered MoS2 coating gave an average life of 303 kc which was
very similar to that found for the CSEM-Ti coating against the uncoated block. The Movic coating sliding
against a non-coated block gave a life of 1165 kc, but the life sliding against a coated CSEM-Ti block
decreased to 360 kc. The CSEM-Al coating lubricated very poorly in nitrogen. The uncoated block gave a
life of 0.3 kc and when it was slid against the block coated with CSEM-Al, it failed immediately. The
Surftech coating also lubricated poorly in nitrogen and failed after 0.6 kc of sliding.

Table 3: Block-on-Ring Test Results

MoS2 Average Friction Endurance Life Block Wear Rate
3 -18
Coating Coefficient (Kilocycles) (m /m x 10 )
Non-Coated Coated Non-Coated Coated Non-Coated Coated
Block Block Block Block Block Block
Movic 0.01 0.01 1166 + 45 360 + 70 2 + 0.5 190 + 160
Surftech 0.04 --- 0.6 + 0.4 --- --- ---
Hohman 0.01 --- 5556 + 3570 --- 1.2 + 1.0 ---
Teer-Ti 0.03 --- 303 + 110 --- 9.5 + 7.0 ---
CSEM-Ti 0.02 0.02 323 + 321 6132 + 950 24,000 + 23,955 0.7 + 0.3
CSEM-Al 0.07 Failed 0.3 + 0.1 Failed --- Failed

0.08
Friction Coefficient

0.07 Noncoated Block

0.06 Coated Block

0.05

0.04

0.03

0.02

0.01

0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al

Type of Sputtered MoS2 Coating

Figure 13. Average friction coefficient for different sputtered MoS2 films (obtained from
various vendors) and tested in dry nitrogen on a block-on-ring Tribometer.

NASA/CP—2002-211506 314
 7000
Non-coated Block
6000 Coated Block
Endurance Life, Kc

5000

4000

3000

2000

1000

0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al
Type of Sputtered MoS2 Coating
Figure 14. Average endurance life for different sputtered MoS2 films (obtained from various
vendors and tested in dry nitrogen) on a block-on-ring Tribometer.

Block-on Ring Block Wear Rate

The variation of block wear rates for the Block-on-Ring tests are given in Table 3 and the averages in
Figure 15. The lowest wear rate was obtained with the CSEM-TI Block sliding against the CSEM-Ti
–18 3
coating (0.7+ 0.3 X 10 m /m). The second lowest was with the non-coated block sliding against the
–18 3
Hohman coating (1.2+1.0 X 10 m /m) and the next lowest was obtained with the non-coated
–18 3
blocksliding against the Movic coating (2.0+1.0 X 10 m /m). The coated CSEM-Ti block sliding against
the Movic coated ring increased the block wear rate almost 2 orders of magnitude compared to the non-
coated block. The non-coated block sliding against the CSEM-Ti coating also had a very high wear rate.

Block-on Ring Surface Morphology

The Movic coatings produced very thin, continuous transfer films on the blocks and very thin, flowing
layers of material on the rings. Figure 16 gives photomicrographs of the transfer films on the non-coated
block and the film remaining on the ring after 1200 kc of sliding. The wear process on the ring appears to
be by very thin layer delamination. Figure 17 gives photomicrographs of the Hohman test specimens after
3700 kc of sliding. The block shows very thin, continuous transfer and the material on the ring appears to
be thicker than that found with the Movic coating (Figure 16) but it also seems to be wearing by
delamination.

45
Block Wear Rate (m3/m x 10-18)

40
Non-coated Block
35
30 Coated Block

25
20
15
10
5
0
Movic Surftech Hohman Teer-Ti CSEM-Ti CSEM-Al

Type of Sputtered MoS2 Coating

Figure 15. Block wear rate for 440C stainless steel blocks sliding on different sputtered MoS2
films (obtained from various vendors) and tested in dry nitrogen on a block-on-ring Tribometer.

NASA/CP—2002-211506 315
 Figure 16. Photomicrographs of the wear surfaces on a 440C block and a Movic coated ring
after 1200 kc of sliding in a dry nitrogen atmosphere on a Block-on-Ring Tribometer.

Figure 17. Photomicrographs of the wear surfaces of a 440C uncoated block and a Hohman
coated ring after 3700 kc in a dry nitrogen on a Block-on-Ring Tribometer.

Figure 18. Photomicrographs of the wear

surfaces on a 440C CSEM-Ti coated block and a Figure 19. Photomicrograph of the wear
CSEM-Ti coated ring after 3150 kc of sliding in a surface on a CSEM-Ti coated ring after 0.2
dry nitrogen atmosphere on a Block-on-Ring kc of sliding in a dry nitrogen atmosphere
Tribometer. against and uncoated block.

NASA/CP—2002-211506 316
 Concluding Remarks

The results of this study showed that the CSEM-Ti and the CSEM-Al coatings gave much longer
endurance lives in air than the other coatings that were evaluated. As stated earlier, Teer Coatings Ltd
has the patent rights to the MoS2-Ti coating and the technology has been licensed to CSEM. The most
probable reason why the Teer-Ti coating gave shorter endurance lives as compared to the CSEM-Ti
coating was that the Teer coating was formulated with less titanium for vacuum use, thus it would not be
expected to work as well in air. Even so, the coating wear rate was nearly equal between the two coatings
in dry air; so if the same thickness of the Teer-Ti coating had been applied, the life probably would have
been equivalent. But in moist air, the CSEM-Ti coating wear rate was much less than the Teer-Ti coating,
which indicates that in humid air, more titanium is necessary. In dry nitrogen, on the Block-on-Ring
Tribometer, the endurance lives for the two films were nearly equivalent, but the wear to the blocks that
slid on the Teer-Ti coating was much less than found with the CSEM-Ti coating. The results indicate that
there should be a more detailed study to determine how much Ti should be added to prevent degradation
and reasonable life in ambient air while giving optimal life and performance in vacuum.

Increased thickness of the coatings which contained additives (and were also able to support the load of
the sliding 440C ball counterface) tended to give increased life. It is not believed that increased thickness
of the non-additive Movic or Surftech coatings in air would have improved the life considerably because
they failed primarily due to degradation of the coatings.

Tests were conducted in nitrogen using the Block-on-Ring Tribometer because this geometry simulates a
journal bearing more closely that does a Pin-on-Disk configuration; a journal bearing contact is closer to
line contact than point contact. There were several drawbacks with using this Tribometer, however.
Misalignment of the block with the ring and wobble or out of roundness of the ring are a couple; thus with
this Tribometer, it is very hard to perfectly align the surfaces. Misalignment can cause high contact stress
that can prematurely cause failure of a solid lubricant coating. It is felt that the CSEM-Ti coated on both
surfaces, helped mitigate this misalignment. Also thicker coatings could be helpful in mitigating
misalignment providing that the coating would not experience brittle fracture during the “run-in.” The
CSEM-Al coating was very brittle thus there was no advantage of sliding it against itself. Sliding a block
coated with CSEM-Ti against the Movic coating was not advantages since the CSEM-Ti coating is
somewhat rough and very hard, thus it promoted more rapid wear of the Movic coating.

Considering all the results, the Hohman coating, which was developed by the Air Force [18] and licensed
to Hohman Plating, functioned overall as the best coating for our application under the conditions of all
the experiments. The lowest friction coefficients and longest endurance lives were obtained with this
coating. The CSEM-Ti gave exceptional results in air and also exceptional results when both the ring and
the block were coated. It may be that the Teer-Ti coating or the other coatings might have worked as well
in nitrogen if both blocks and rings were coated with the same material, but in this study those conditions
were not evaluated.

The next planned stage for this program is to take the best coatings determined by this investigation and
test them in vacuum in a journal bearing at 50,000 rpm under the conditions that they may encounter in a
flywheel system touchdown event to determine which will perform the best in the actual end use
application. The poor endurance life results obtained in dry nitrogen with the Surftech and CSEM-Al
coatings do not make them candidates for the next phase of testing. The short endurance life of the Movic
coating in moist air and degradation of the coating on the disk surface in dry air also discourages their
use for this application.

References

1. Winer, W.O., “Molybdenum Disulfide as a Lubricant: A Review of the Fundamental Knowledge,”

Wear, 10, (1967), 422-452.
2. Johnson, R. L., “A Review of the Early Use of Molybdenum Disulfide as a Lubricant,” NASA TMX
52343 (1967).

NASA/CP—2002-211506 317
 3. Farr, J.P.G., “Molybdenum Disulfide as a Lubricant: A Review of the Fundamental Knowledge,”
Wear, 10, (1967) 442-452.
4. Spalvins, T., “Lubrication with Sputtered MoS2 Films, NASA TMX 67832 (1971)
5. Spalvins, T., “Deposition of MoS2 Films by Physical Sputtering and Their Lubrication Properties in
Vacuum," ASLE Trans., 12 (1969), 36-43.
6. Loewenthal, S.H., Chou, R.G., Hopple, G.B. and Wenger, W.L., “Evaluation of Ion-Sputtered
Molydenum Disulfide Bearings for Spacecraft Gimbals,” Tribology Trans., 37, 3 (1994), 505-515.
7. Hopple, G.B. and Loewenthal, S.H., “Development, Testing and Characterization of MoS2 Film
Bearings,” Surface and Coatings Technology, 68/69 (1994), 398-406.
8. De Gee, A.W.J., Salomon, G., and Zaat, J.H., “Mechano-Chemical Factors in MoS2 Film
Lubrication,” Wear, 7 (1964), 87-101.
9. De Gee, A.W.J., Begelinger, A. and Salomon, G., “Influence of the Atmosphere on the Endurance
of Some Solid Lubricants Compared at Constant Layer Thickness,” Proc. Inst. Mech. Engr., 183,
Pt 3, I, (1968), 18-27.
10. Gansheimer, J., “Review of Chemical Reactions of Solid Lubricants during Friction,” ASLE Trans.,
15, 4 (1972), 244-251.
11. Pritchard, C., and Midgley, J.W., “The Effect of Humidity on the Friction and Life of Unbonded
Molybdenum Disulfide Films,” Wear, 13, (1969), 39-50.
12. Peterson, M.B., and Johnson, R.L., “Friction and Wear Investigation of Molybdenum Disulfide. I-
Effect of Moisture,” NACA TN 3055 (1953).
13. Fusaro, R.L., “Lubrication and Failure Mechanisms of Molybdenum Disulfide Films I – Effect of
Atmosphere,” NASA TP-1343 (1978).
14. Singer, I.L., Fayeulle, S., and Ehni, P.D., “Wear Behavior of Triode-Sputtered MoS2 Coatings in
Dry Sliding Contact with Steel and Ceramics,” Wear, 195, (1996), 7-20.
15. Miyoshi, K., and Pepper, S.V., “Properties Data for Opening the Galileo’s Partially Unfurled Main
Antenna,” NASA TM 105355 (1992).
16. Stupp, B.C., “Synergistic Effects of Metals Co-Sputtered with MoS2,” Thin Solid Films, 84 (1981),
257-266.
17. Hilton, M.R., Bauer, B., Didziulis, S.V., Dugger, M.T. Keem, J.M. and Scholhamer, J., “Structural
and Tribological Studies of MoS2 Solid Lubricant Films Having Tailored Metal-Multilayer
Nanostructures,” Surf. and Coat. Tech, 53 (1993), 13-23.
18. Zabinski, J.S., Donley, M.S., Walck, S.D., Schneider, T.R., and McDevitt, N.T., “The Effects of
Dopants on the Chemistry and Tribology of Sputter-Deposited MoS2 Films,” Tribology Trans., 38,
4, (1995), 894-904.
19. Teer, D.G., Hampshire, J., Fox, V., and Bellido-Gonzalez, V., “The Tribological Properties
MoS2/Metal Composite Coatings Deposited by Closed Field Magnetron Sputtering,” Surface and
Coatings Technology, 94-5, (1997), 572-577.
20. Aharonov, R. and LoBiondo, N., “The Influence of Temperature and Humidity on Wear of MoS2
and MoST Films,” Proceedings of the International Conference of Metallurgical Coatings and Thin
Films, (April 2001).
21. Fusaro, R.L., “How to Evaluate Solid Lubricant Films Using a Pin-on-Disk Tribometer,” Lubr.
Eng., 43, 5, (1986), 330-338.
22. Fusaro, Robert L., “Mechanisms of Lubrication and Wear of a Bonded Solid-Lubricant Film,”
ASLE Trans., 24, 2 (1981) 191-204.

NASA/CP—2002-211506 318
 Space Ball Bearing Load Capacity with Dry Lubrication
*
A.Borrien , J.B. Mondier* and E. Conde*

Abstract

Stainless steel ball bearings are widely used in space mechanisms either with a rigid or a soft axial
preload. The dynamic behavior of these components during launch vibrations is highly dependent on the
non-linear stiffness of the ball-races contact and the modal response of the suspended mass. Therefore,
the prediction of the contact stress in the bearings requires a careful analysis of all the structural parts
around the bearings. This analysis must be done over a large frequency bandwidth because the random
vibration spectrum induced by the launcher environment generates mechanical disturbances up to 2000
Hz. The calculated stress is then compared to the theoretical limit given by the ISO 76 standard, taking
into account a safety margin.
The aim of this study was to determine the maximum allowable stress in the MoS2 dry coated ball
bearings to prevent any degradation that could be evidenced through the friction torque behavior or the
binocular inspection of the coated pieces.
Twenty rigidly preloaded ball bearing pairs were tested to investigate this point. In order to rely on the
experimental results, we decided not to use a shaker fitted with an accelerometer sensor that would have
required a double integration for the displacement estimation. We preferred a pulling machine with a
redundant load cell and an inductive displacement sensor. By doing this, we got a direct measurement of
the axial stroke of the bearing rings, a direct measurement of the load applied to the bearings, and we
avoid any dynamic resonance in the test bench thanks to a 20-Hz low frequency load sine cycle. The
effect of cumulated surface fatigue was taken into account by doing the same number of cycles as in 800
Hz oscillations during two minutes.
Analysis of the bearing contact stress was performed through the CNES software named RBS2. The
effect of the initial preload on the gapping amplitude in the bearing was also computed as well as the ball-
races friction torque.
The first lesson learned is that the Hertzian theory used in RBS2 gives incorrect values of the contact
stress in the case of a very small difference between the ball and the race radii of curvature. A threshold
was clearly pointed out to ensure the validity of the results.
The second lesson is related to the geometrical internal shape of the bearing that could limit its load
capacity to a value that is lower than the allowable stress because the contact area is not entirely located
in the raceway.
The third lesson is related to the scattering of the bearing manufactured radial gap. We found differences
between the theoretical axial displacement under load and the measured one. The measurement of this
radial gap is thus essential to compare the predictions to the test results.
The paper will present the experimental stress limit that is observed previous to the race indentation
onset, and the effect of these indentations on the friction torque. The effect of the gapping amplitude on
the post-vibration friction torque is also discussed. Inspected hardware shows the MoS2 coating aspects
in the contact areas.

*
CNES, Toulouse, France
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 319
 Introduction

The mechanism designer often needs to solve the problem of ball bearing arrangement and preload to
sustain the launch load without degradation while ensuring a low friction torque to achieve the lubricant
lifetime under vacuum. The higher the preload is, the higher the friction torque will be and the shorter the
lifetime will also be. On the contrary, the lower the preload is, the higher the gapping amplitude is. This
creates internal shocks under dynamic loads often called “hammering” and could lead to an early bearing
lubrication failure.
In Europe, people often consider that for MoS2 dry-lubricated mechanisms, the bearing working contact
pressure should be below 1200 MPa to ensure a significant life time and then, the preload has to be
defined as low as possible, while ensuring a gapping amplitude not greater than about 30µm at
qualification level. These rules are more or less derived from experience but justification has never been
provided to demonstrate the validity of this design criterion.
During vibration, for safety reasons it was generally assessed that the Hertzian pressure will remain
below 2000 MPa. Several tests made at CNES show that even with a dynamic pressure exceeding 3000
MPa, a high life time can be reached if a transient increase of the friction torque after vibration is
considered acceptable.
More recently, the European standard (E.C.S.S) stipulated, whatever the lubrication technique is, that a
1.25 factor of safety is mandatory with respect to the maximum allowable stress, which is defined as 4200
MPa in the ISO 76 standard. Therefore, the surface hardness has to be greater than or equal to 60 Rc
(700 Hv) to ensure a permanent deformation smaller than 1/10000 of the ball diameter for 4200 MPa. In
this case, the E.C.S.S. allowable stress including the safety margin can be taken at 3360 MPa.
The need to improve the design rules for dry lubricated MoS2 bearings is obvious and moreover a better
understanding of the pressure distribution in the ball/races contacts as a function of the bearing internal
geometry is also considered very useful for the designers
The aim of this paper is then to answer the following questions:
- Is the maximum contact pressure calculation based on the Hertzian theory always valid to predict a
ball bearing internal mark or indentation?
- Is there a contact pressure limitation for MoS2 dry coated bearings and what is the associated value?
- What is the effect of gapping on the subsequent bearing surface degradation and friction torque
evolution?
To answer to these three questions, CNES prepared a test plan with 20 ball bearing pairs and designed a
dedicated bench allowing the preload and alternative dynamic load tuning.

The Test Bench Description

The bearing pair is hard preloaded by two ground struts and a screw nut device. It is mounted on a shaft
inside a steel cylindrical housing. An inductive displacement sensor is located between the end of the
shaft and the housing. This sensor measures the relative displacement between the internal and external
rings of the bearing pair. At each extremity of the bench, the shaft and the housing can be put in the jaws
of a pulling machine. A load cell is placed on the pulling machine interface. The MTS 810 pulling machine
applies the load through hydraulic jacks. The vibration equivalence is based on an 800 Hz resonance
mode applied during two minutes. The pulling machine is able to work very well at 20 Hz, which means a
1 hour and twenty minute test to reach the 96,000 equivalent cycles.

This kind of test bench allows a direct access to the bearing load and internal displacement and there are
no parasitic effects due to mechanical resonance or transverse load generally induced by a shaker.
Moreover, there is no need of a double integration of the acceleration measurements to evaluate the
bearing displacement. The load sensor also provides the capability to check the applied preload value,
which is dependent on the struts and bearings thickness differences and on the bearing’s radial gap. In
fact, for a 15° contact angle, the radial gap should be equal to 43.8µm. The radial gap measurements of
the various bearing pairs pointed out a variation from 22µm to 46µm.

NASA/CP—2002-211506 320
Figures 1 and 2 show a photo and a cross section of the test bench: Once the bearing pair was attached
to the pulling machine, a first test was performed to measure the preload. At the beginning of motion, the
bearing stiffness was related to the preloaded pair until the load reached 2.7 times the preload. After that
time, the slope of the stiffness curve is changed because one of the bearings was offloaded. Plotting the
slope change on the curve gives the offloading effort and then the preload value. All the measurements
done for a theoretical 600N preload showed a preload value between 530N and 660N.

Figure 1. Test Bench view Figure 2. Test Bench cross section

The effect of internal geometry (first campaign)

A first set of ball bearings was procured for the tests after some preliminary calculations. The bearings
had to be large enough to make the rings and balls observation easier, and not too large to avoid any
stiffness limitation of the test bench itself. A 37-mm external diameter bearing from the ADR Company
was chosen. For the internal geometry, it was decided to achieve different conformities for the inner and
outer race in order to investigate the effect of the Hertzian ellipse size on the contact pressure evaluation.
A microscope observation of the races would confirm the presence of marks in case of plastic
deformation. The main characteristics of the bearing pair are described in Table 1.

NASA/CP—2002-211506 321
 Table 1. Bearing characteristics
ADR ref W61904H
Outside diameter ( mm) 37
Bore diameter (mm) 20
Ball diameter (mm) 4.762
Contact angle (°) 15°
Nominal Preload (N) 600 (± 100 )
Outer race conformity 1.12
Inner race conformity 1.06
Number of balls 14
Bearing width (mm) 9
Lubrication MoS2 ( ADR process )

The RBS2 software was used to estimate the relative axial displacement between the internal and
external ring of the bearing under a given axial load. These values were compared with the computation
of the Palmgren formula (Table 2).
Table 2: Bearing axial displacement computation
Axial load (N) Axial displacement I.R./O.R. Axial displacement I.R./O.R.
RBS2 Software (µm) Palmgren (µm)
4350 48 44
8400 75 78
9600 81 87
11720 91 100
14820 111 120

The values are very similar and this allows us to be confident in the predicted displacement calculation.
The maximum axial load applied to the bearing arrangement to reach 4200 MPa on the two ball/races
contacts was also calculated using the RBS2 software. The result is about 8400N, which induces 4199
MPa on the inner ring contact and 4256 MPa on the outer ring contact. The steel damage on both rings
would normally occur for a load exceeding 8400 N.
For this first campaign, a wide range of axial loads was applied to the bearings to damage them and
make easier the observation of the rings plastic deformation. Table 3 summarizes the main results.

Table 3. Bearing plastic deformation versus axial load.

Axial load (N) Computed Max Computed Measured axial Depth of Depth of
Pressure (MPa) Axial displacement deformation deformation
displacement (µm) (µm) (µm)
(µm) Harris observation
2170 2880 27 22 0 0
8400 4200 75 60 1,7 0
11720 4600 91 90 3,3 0
19760 5300 111 120 9,4 0

NASA/CP—2002-211506 322
The data indicates that even with a 5300 MPa axial limit load, that represents more than two times the
theoretical, there is no evidence of any damage on the raceways. The photo of the inner and outer ring of
test n°6 (19760N) shows succession of MoS2 accumulation spots without any mark on the steel itself.
Figure 3 shows the sensor displacement obtained at low frequency (4 Hz) after the calibration sequence.
The curve is not a pure sine due to the hysteretic behaviour of the contact. The microscope inspection of
the different bearing elements shows the presence of MoS2 on each ball together with drawn lines that
indicates a rotation of the balls during the gapping period. Some MoS2 transfer occurred also between the
balls and the cage that is visible on the cage pockets aspect. Figure 4 shows the aspect of the ball and
cage after vibration.
It was then decided to apply a very high load to one bearing pair to make a ball indentation on the
raceways and to inspect the lubricant presence after test. The chosen load was 80,000N. In that case, the
calculation leads to a very wide mark on the rings. The observation confirmed these marks and clearly
indicated that the contact area has moved to the edge of the ring raceway, getting a truncated ellipse.

déplacements (miillimètres)

1.28
1.26
1.24
1.22
1.2
1.18
1.16
1.14
1.12

0 200 400 600 800 1000

Figure 3: Inductive sensor displacement (mm)

Figure 4: Inner and outer ring aspect after 19760 N axial load

NASA/CP—2002-211506 323
 Figure 5: Ball and cage aspects after test
Why don’t we see any marks at 5300MPa? The steel hardness has been checked and confirms the
60HRc to 61HRc value with a very good homogeneity from one bearing to another. So, the absence of
indentation is obviously linked to a false estimation of the maximum pressure. Everyone knows that when
the contact area becomes large compared to the contact pieces diameter, as it is the case for example in
a ball and socket joint, there is more than one contact point and the pressure distribution is highly
dependant on the micro-geometrical deviations of the surfaces. In this case, the Hertz calculation is no
longer realistic. At this stage, we looked into the contact ellipse size compared to the ball size. This ratio
between the ellipse length and the ball radius is well described by the angle β.

The sine of the maximum half contact angle β/2 between the ball and the race is defined by half the
length of the major axis of the contact ellipse A divided by the ball radius Rb (Figure 6)

Rb

β/2
2A

Figure 6. Ellipse angle β definition

For a given major axis length, that means for a given ellipse angle β between a ball and the races, a
computation made with the RBS2 software gives the values in Table 4 for the Hertzian pressure for a
given axial load as a function of the ball/ring conformity.

NASA/CP—2002-211506 324
 Table 4: Effect of the ellipse size
β = 40° β = 35°
Inner Ring Outer Ring Inner Ring Outer Ring
Conform. Axial load Hertz. Axial load Hertz. Axial load Hertz. Axial load Hertz.
( N) Press. (N) Press (N) Press (N) Press.
(MPa) (MPa) (MPa) (MPa)
1.06 4185 3335 4480 2980
1.08 5640 3940 6240 3245
1.10 7610 4500 8485 4065 4720 3985 5175 3585
1.12 9330 5010 10360 4535 5910 4430 6580 4005
1.14 11480 5480 12745 4975 7090 4850 7910 4390
1.16 13170 5925 14730 5390 8340 5240 9330 4755
1.20 17020 6730 17680 6140 10790 5960 12145 5430
1.22 12015 6290 13585 5740
1.24 13245 6605 14975 6035

These calculations show that for a 1.12 conformity, the damage of the outer ring should occur at 8500 N.
The same conformity leads to a lower value for the inner ring (5900N). This is the reason why the
manufacturers often choose a lower conformity for the inner ring to better balance the bearing resistance.
With a conformity of 1.06 on the inner ring, we observe that the ellipse length becomes quickly very high
as the load increases. For 8500N, the β angle is greater than 40° on the two rings. Figure 7 shows the
evolution of the contact pressure and the ellipse angle as a function of the bearing axial load:
ADR W61904 ball bearing
inner curvature ratio=0,53
outer curvature ratio=0,56
6000 54

49
5000

IR Hertz stress
44
4000

OR Hertz stress
39
MPa

d°

3000
IR contact extent
34

2000 OR contact extent

29

1000
24

0 19
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000

Axial load (N)

Figure 7. Maximum contact pressure and ellipse angle evolution

NASA/CP—2002-211506 325
The ellipse angle curve (β) indicates that for a pressure of 4200MPa, the inner and outer angles are
greater than 35°. This leads us to be careful about the pressure calculation for such a large angle.

Modification of the internal geometry (second campaign)

In order to improve our understanding on the contact pressure behavior, we decided to change the inner
ring conformity to get a significant increase of the contact pressure for a given load. The new conformity
is chosen at 1.15 for the inner ring while the outer one is unchanged. The bearing manufacturer sent us a
new set of bearings with this geometry. The contact pressure computation with RBS2 gives the axial load
corresponding to 4200 MPa for the two contacts. For the inner ring contact, the value is about 4000N and
for the outer ring contact, the value becomes 7400N instead of 8500N. The loss of load capacity is very
important on the inner ring contact, but it is also significant on the outer ring contact. This is due to the
change in the bearing contact angle. The new values of the β angle are 28° for the inner contact and 35°
for the outer contact. The contact pressure and ellipse angle evolution as a function of the axial load are
depicted on the Figure 8. A new test campaign was done with the pulling machine. The main results are
presented in Table 5:
These results are very interesting because they show a very good correlation with the theoretical
pressure limit. Figure 9, related to the bearing pair n°18, points out the steel marks on the inner ring and
the MoS2 accumulation around the contact points on the outer ring.

ADR W61904 ball bearing

inner curvature ratio=0,575
outer curvature ratio=0,56
6000 54

49
5000

IR Hertz stress
44
4000

OR Hertz stress
39

3000
IR contact extent
34

2000 OR contact extent

29

1000
24

0 19
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000

Axial load (N)

Figure 8: max. contact pressure and ellipse angle evolution after modification

NASA/CP—2002-211506 326
 Table 5. Bearings marks after modification

Bearing pair Axial load (N) Inner contact Outer contact Inner contact
number pressure (MPa) pressure (MPa) marks
9 2050 3500 2900 None
20 3900 4200 3500 None
10 4250 4300 3650 Yes
11 4250 4300 3650 Only one
7 5000 4500 3800 Yes
18 5400 4600 3900 Yes
13 6000 4700 4000 Yes + Outer Ring
Note: “Yes” means there is a steel mark at most of ball raceway contacts.

Inner Ring Outer Ring

Figure 9. Bearing n°18, inner and outer rings aspect

The optical aspect of the bearings rings, even if it looks quite unambiguous is not enough to be confident
on the steel indentation. It was then decided to check the raceways surface through a roundness
“Talyrond” measurement. This measurement is not affected by the MoS2 presence and it allows sensing
the raceways on several parallel tracks to quantify the depth and the width of the steel marks. For this
test, the bearing n° 13 was chosen and the results for both inner and outer rings are depicted on the
Figure 10. The ball number is easily visible on this curve and it is also clear that the depth marks are not
similar all around the rings, which indicates a difference in the load repartition between the balls. In the
case of this bearing, the indents depth on the measured ring is between 1 µm and 1.5 µm (see Fig 9).
Compared to the ball diameter, this is greater than Db/10000, that confirms the application of a local
pressure greater than 4200MPa.

NASA/CP—2002-211506 327
 Figure 10. Bearing n° 13, “Talyrond” measurements

The Friction Torque Behavior

The aim of this part was to quantify the friction torque instabilities after the high contact stresses induced
by the alternative axial load. The torque of each bearing pair was measured at low speed under the 600-
N axial preload before and after the pulling machine test. The external rings are driven by a motor while
the inner ones are coupled to a torquemeter. The friction torque average value before test is about 20
mNm with some variation from one bearing to another.

In all cases, the post-test friction torque is noisier. This is due to the MoS2 particles pushed outside the
contact zone and lying on the raceways. When the contact pressure becomes high, several friction torque
spikes appear. These spikes are due to a ring of particle accumulation around each contact zone. There
are 14 balls in the bearings that would normally induce 14 spike disturbances on the raceway revolution.
In fact, the torque measurement is made over one revolution of the external ring and the number of
spikes corresponds to the total angle covered by one ball divided by the angular distance between two
balls.

The calculation gives a value of 7.6 for the inner ring and 6.4 for the outer ring. The picture n° 11
corresponding to 4200MPa contact, shows clearly about 6 spikes for one CW and CCW revolution. The
distinction between the internal and external ring spikes is not easy, but sometimes, two spikes are very
close to each other, indicates that the phenomenon does affect the two rings.

Figure 11. Friction torque measurement before test

NASA/CP—2002-211506 328
 Figure 12. Friction torque measurement after test (4200MPa)

A running-in test made on one bearing pair shows that the amplitude of the spikes decreases a lot after
several thousands of revolutions. This running- in phase is necessary to clean the raceways from all
particles and to smooth the rings accumulation of MoS2.

Some more tests should be done on the damaged bearings to check the influence of possible
interference between the steel marks and the ball raceways under the nominal preload. Unfortunately, we
did not have enough time to perform these tests in 2001.

The gapping influence (Third Campaign)

A lot of things have been said on the influence of gapping amplitude on the bearing internal shocks and
the associated degradation of solid lubricant coatings and in some cases, the steel surfaces.

The related investigation was performed with the same type of bearing by using different strut lengths to
tune the bearing pair preload. Doing that, the preload range was varying from 80N to 1200N, and the
bearing gapping becomes smaller as the preload increases. The different tests performed are depicted on
the Table 6.

Table 6. Pulling machine results for different preload values

Bearing n° Pre-load (N) Contact Axial load (N) Gapping Steel marks
pressure (MPa) amplitude (µm) observation

12 80 4200 3900 85 Yes

14 300 3500 2050 46 None
17 300 4200 3900 72 Several
15 300 4300 4250 76 Yes
19 300 4500 5000 84 Yes
20 600 4200 3900 59 None
7 600 4500 5000 71 Yes
16 1200 4300 4250 48 Yes

NASA/CP—2002-211506 329
The first information given by these results is that the steel marks are observed when the pressure
reaches 4300 MPa whatever the preload is. When the contact pressure is close to 4200 MPa, it seems
that the higher the preload is, the weaker the marks are.

Finally, the last information is given by the bearing n° 14. When the pressure is close to the allowable
ECSS value, there are no visible marks, even with a gapping amplitude of 46 µm. This bearing 14 has
been tested with the torque-meter and no spike has been found over one whole revolution.

Conclusion

This work dedicated on MoS2 dry lubricated bearing resistance to dynamic stresses induced by a satellite
launch leads us to the following conclusions that probably need to be cross checked with other tests
results on different bearings geometry and size.

About the influence of the geometry (first question of the introduction):

- The maximum contact pressure given by a software-based on the Hertzian theory becomes
pessimistic when the contact ellipse angle is greater than 35°. This is probably the reason why
the bearings with a conformity smaller than 1.05 exhibit no steel marks even with a very high
predicted stress.

- The designer has to take care of the groove edge that could be reached during the dynamic
vibrations leading to a truncated ellipse and to a risk of an early bearing degradation. This risk is
not always taken into account by the software computation.

- When the ellipse angle is not greater than 35° and for a bearing hardness of 61 HRc, there is a
very good correlation between the 4200 MPa theoretical stress and the first appearance of steel
indentation on the bearing raceways.

About the pressure limitation due to MoS2 (second question of the introduction):

- Although the friction torque is always noisier after a vibration test due to some MoS2 particles
releasing, there are no spikes in the friction torque when the contact pressure is not greater than
3500 MPa. The MoS2 coating is then able to resist to the maximum ECSS allowable contact
pressure.

- When the pressure reaches 4200 MPa, there are spikes in the torque measurement and some
indentations in the steel. A run-in test makes the spikes much lower, showing that the MoS2 is still
able to lubricate the bearing, and that the indents are outside the ball tracks.

About the gapping amplitude influence (third question of the introduction):

- A gapping amplitude greater than 30µm does not imply a steel degradation due to “hammering”.
The steel indent is much more governed by the maximum contact pressure than by the gapping
amplitude. However, when plastic deformation occurs, it seems that a high gapping amplitude
worsens the indents.

References

1 Harris, T.A. Rolling Bearing Analysis

2 Lewis, S.D. – Observations on the influence of launch vibration on bearing torque and lubricant
th
performance – 9 European Space Mechanisms and Tribology Symposium. Liège Belgium –
September 2001

NASA/CP—2002-211506 330
 Tribological Properties of a Pennzane-Based Liquid Lubricant (Disubstituted Alkylated
Cyclopentane) for Low Temperature Space Applications
* ** + + ++
Clifford Venier , Edward W. Casserly , William R. Jones, Jr. , Mario Marchetti , Mark J. Jansen , Roamer
+++
E. Predmore

Abstract
®
The tribological properties of a disubstituted alkylated cyclopentane, Pennzane Synthesized
Hydrocarbon Fluid X-1000, are presented. This compound is a lower molecular weight version of the
®
commonly used multiply alkylated cyclopentane, Pennzane X-2000, currently used in many space
mechanisms. New, lower temperature applications will require liquid lubricants with lower viscosities and
pour points and acceptable vapor pressures. Properties reported include: friction and wear studies and
lubricated lifetime in vacuum; additionally, typical physical properties (i.e., viscosity-temperature, pour
point, flash and fire point, specific gravity, refractive index, thermal properties, volatility and vapor
pressure) are reported.

Introduction

All spacecraft utilize mechanisms contain moving mechanical assemblies (MMAs) that require some form
of lubrication to function properly [Jones and Jansen, 2000]. Lubricants include liquids, greases, and
solids. These materials normally operate in ultrahigh vacuum. Therefore they must possess extremely low
vapor pressure [Nguyen et al, 2001]. Most spacecraft are thermally compensated so that these lubricants
only experience temperatures in the narrow range of 0 to +60°C. This allows conventional liquid and
grease lubricants to be used. However, new spacecraft mechanisms are being designed that will operate
at much lower temperatures (i.e.,–50 to 0°C). Most standard liquid lubricants possess very high
viscosities in this temperature range, which would necessitate the use of large motors with high power
requirements to maintain design torque margins. Therefore, lower viscosity liquid lubricants that still
possess low volatility are needed.
® ®
Pennzane Synthesized Hydrocarbon Fluid X-2000 (usually called Pennzane fluid) is a multiply alkylated
cyclopentane hydrocarbon that has desirable tribological properties for many space applications [Venier
and Casserly, 1991; Venier, et al., 1992; Carré et al., 1995, Casserly and Venier, 1999]. A six year life
®
test of a CERES bearing assembly using Pennzane fluid has yielded excellent results [Brown et al.,
1999]. Accelerated and operational life tests on bearings for the MODIS instrument have also been
®
completed using a Pennzane fluid formulation [VanDyk et al., 2001] and showed excellent results. Full
®
scale bearing tests conducted at Lockheed Martin compared the performance of a formulated Pennzane
®
fluid X-2000 to Bray 815Z, a standard space lubricant, and showed Pennzane fluid to have at least a 7
times life advantage over 815Z [Loewenthal et al., 1999]. Relative lifetime tests [Jansen et al., 2001;
Jones et al., 2000] using the Spiral Orbit Tribometer (SOT) have also shown that unformulated
®
Pennzane fluid (X-2000) yielded the greatest relative lifetime compared to a series of space lubricants.
Despite the excellent performance of Pennzane X-2000, its high viscosity at low temperature (i.e.,
80,000 cP at – 40°C) precludes its use under these conditions.
®
Pennzane X-2000 is a member of the chemical class of multiply alkylated cyclopentanes or MACs
[Venier and Casserly, 1991; Casserly and Venier, 1999]. It is predominately the tri-alkylated cyclopentane
product prepared from 2-octyldodecanol, namely, tri-2-octyldodecyl cyclopentane. The disubstituted
*
Pennzoil-Quaker State Company, The Woodlands, TX
**
Penreco Company, The Woodlands, TX
+
NASA Glenn Research Center, Cleveland, OH
++
Sest, Inc., Middleburgh Hts., OH
+++
NASA Goddard Space Flight Center, Greenbelt, MD
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 331
product (commercially called Pennzane Synthesized Hydrocarbon Fluid X-1000, or simply X-1000) has
now been produced. It has a more useful low temperature viscosity (29,000 cP at -40°C) and yet retains
relatively low vapor pressure at room temperature.

The objective of this work is to compile some of the physical and tribological properties of interest to
space mechanism designers for this disubstituted product (X-1000). Properties reported include: wear
rates and relative lifetime in ultrahigh vacuum and typical physical properties (i.e., viscosity-temperature,
pour point, flash and fire point, specific gravity, refractive index, thermal properties, vapor pressure and
volatility).

Physical Properties
®
The measured physical properties of bis(2-octyldodecyl)cyclopentane, Pennzane X-1000, are given in
®
Table 1 along with the corresponding properties of tris(2-octyldodecyl)cyclopentane, Pennzane X-2000,
and a linear perfluropolyalkylether (PFPAE). For the hydrocarbons, the properties reported were obtained
by ASTM test procedures at the Pennzoil-Quaker State Technology Center, The Woodlands, TX, except
that thermal conductivity, specific heat, and autoignition temperature were obtained from Phoenix
Laboratories, Chicago, IL, and outgassing data from Ball Aerospace, Boulder, CO. The data for a linear
PFPAE are from the literature.

Vapor Pressure
®
The vapor pressure of Pennzane SHF X-1000 was calculated from the rate of effusion through a small
-6
orifice in a Knudsen cell. At 125° C, the vapor pressure is 6 x 10 Torr. Knudsen cell measurements
[Nguyen and Jones, 2001] on a previous lot of X-1000 yielded a somewhat higher vapor pressure of 2.5 x
-5 -9
10 Torr at 100°C. However, extrapolated to 25°C, the vapor pressure is only about 3 x 10 Torr.
®
The very low vapor pressure and pour point (-52° C) of bis(2-octyldodecyl)cyclopentane, Pennzane X-
®
1000, are the same combination of properties that have made Pennzane X-2000 so useful for lubrication
in spacecraft. Taken together they define a wide, useful temperature range for X-1000 for vacuum
applications.

Viscosity
®
The viscosity index of Pennzane X-1000 is high (131) and the kinematic viscosity of 9.4 cSt at 100° C is
less than might be expected from such a non-volatile fluid. In line with the low vapor pressure, bis(2-
octyldodecyl)cyclopentane has a high flash point of 290° C (550° F).

Thermal Properties
The specific heat of bis(2-octyldodecyl)cyclopentane is 0.46 cal/gm at 30° C, about 10% lower than that of
tris(2-octyldodecyl)cyclopentane. For temperatures between 30° C and 100° C, the specific heat is nicely
linear with temperature, following the equation,
-3
Specific heat = 0.453 + (0.155(° C) x 10 )

The thermal conductivity of bis(2-octyldodecyl)cyclopentane at 30° C is 0.165 W/(m)(° K), very nearly the
same as that of tris(2-octyldodecyl)cyclopentane, and about twice that of perfluoropolyether. The thermal
conductivity of bis(2-octyldodecyl)cyclopentane at 100° C is 0.144 W/(m)(° K),

Optical Properties
The infrared spectrum of bis(2-octyldodecyl) cyclopentane is typical of a saturated hydrocarbon (Figure
1). It is colorless in the visible region and virtually transparent down to 300 nm in the ultraviolet region.

NASA/CP—2002-211506 332
 Figure 1. Infrared Spectra of (a) bis(2-octyldodecyl)cyclopentane, and (b) perfluoropolyether

Tribological Properties

Vacuum Four Ball Tribometer

Figure 2 shows a tribometer developed at NASA Glenn Research Center based upon a four-ball
configuration. It is designed to test the ability of liquid lubricants to reduce wear under pure sliding
-6
conditions at room temperature under a vacuum of at least 10 Torr. The system consists of a rotating
ball sliding against three stationary balls that are immersed in a lubricant. The system is loaded though a
pneumatic cylinder, which pushes the lubricant cup and stationary balls against the rotating ball. The
lubricant cup is held in position by a flex pivot.

Figure 2. Detail of four-ball apparatus

NASA/CP—2002-211506 333
Typical tests are run with 9.82 mm (3/8 in) diameter balls, 100 RPM, room temperature, and an initial
Hertzian stress of 3.5 GPa. This stress drops as a function of test time as a wear scar develops on the
stationary balls. The test is stopped every hour and the wear scar diameters measured. A special
platform allows for the measurement of the wear scars without removing them from the cup [Masuko et
al., 1994]. This allows the test to resume exactly where it was stopped. A full test takes 4 hours. Upon
completion, the wear volume is plotted as a function of sliding distance and the wear rate is calculated
from the slope of the line. This rig provides quick information about the wear characteristics of lubricants
and additive packages to reduce metallic wear.

Spiral Orbit Tribometer

The spiral orbit tribometer (SOT), first introduced by Kingsbury [1989], is essentially a thrust bearing with
flat races (plates) and a single ball (Figure 3). The tribometer simulates rolling, pivoting, and sliding as
seen in an actual angular contact bearing. Accelerated tests are achieved by only using microgram
quantities of lubricant on the ball. During the test, the lubricant is completely consumed by tribochemical
attack, resulting in short test durations. The advantage of this type of acceleration is that operational test
parameters, such as contact stress, speed, and temperature can mimic those in the actual application.
Long lubricated lifetimes imply low lubricant consumption or low tribochemical degradation rates.

Figure 3. The Spiral Orbit Tribometer

The tribological elements of the system appear in more detail in Figure 4. The lower plate is stationary
while the top plate can rotate at speeds up to ~200 RPM. The top plate rotation drives the ball in a spiral
orbit. During every orbit, the ball contacts the vertical guide plate, which returns the ball to the original
orbit radius. The straight-line region where the ball contacts the guide plate is denoted as the “scrub”. The
force the ball exerts on the guide plate during the scrub is measured and the coefficient of friction can be
calculated. After leaving the scrub, the ball’s spiral orbit begins again. The spiral orbit and scrub constitute
a track that is stable, repeatable, and is traversed thousands of times by the ball. A detailed description of
the tribometer and analysis of ball kinematics appear in References by Kingsbury, 1989 and Jones et al.,
2000.

NASA/CP—2002-211506 334
 Figure 4. Detail of the SOT

(

(
ZHDUUDWH PPPP

(

(

( 
; 0$& $ 0$&  3)3($& 3)3(=

Figure 5. Four-Ball Wear Rates for Several Lubricants

NASA/CP—2002-211506 335
Table 1 - Comparative properties of Pennzane® SHF X-1000, Pennzane® SHF X-2000, and PFPAE

Bis(2- Tris(2-
(a)
octyldodecyl)- octyldodecyl)- PFPAE
cyclopentane (X- cyclopentane 815Z
1000) (X-2000)
Specific Gravity 0.85 0.85 1.85
Pour point (° C) -52 -45 -72
Refractive Index (20°C) 1.4682 1.4671 1.294
Vapor Pressure
-6 -7 -8
125° C (Torr) 6 x 10 4 x 10 8 x 10
-12 -12
40° C (Torr) 1 x 10 3 x 10
Vacuum Outgassing (125° C, 24
-5
hr, 10 Torr)
Total Wt. Loss (%) <0.4% <0.2%
Condensables (%) 0.2% <0.1%
Flash Point (° C) 290 315 None
Viscosity
2
100° C (mm /sec) 9.4 15.0 45 @ 99°C
2
40° C (mm /sec) 60 110 148 @ 38°C
-20° C (mPa•sec) 3000 6200 1900
-30° C (mPa•sec) 8000
2
-40° C (mPa•sec) 29,000 80,000 6500 (mm /sec)
Viscosity Index 131 135 350
Autoignition Temp
(ASTM E659)
Hot-flame (AIT) 750° F
Cool-flame (CFT) 610° F
Reaction threshold temp 485° F
Thermal Conductivity
W/(m)(° K)@ 30° C 0.165 0.16 0.08
W/(m)(° K)@ 100° C 0.144
Specific Heat (cal/gm)
@ 30° C 0.458 0.52 0.20
@ 100° C 0.469

(a) Data from Castrol Industrial North America, Downers Grove, IL

Results

Vacuum Four Ball Tribometer

The wear rate obtained with the unformulated X-1000 fluid appears in Figure 5 along with reference data
for unformulated and formulated X-2000, as well as two commonly used unformulated
perfluoropolyalkylethers (Krytox 143AC and Fomblin Z-25). The Z-25 is chemically identical to Brayco
815Z. The X-1000 has a similar wear rate compared to the X-2000 and significantly lower than the two
perfluoropolyalkylether fluids.

Spiral Orbit Tribometer

The normalized lifetime for X-1000 at a mean Hertzian stress of 1.5 GPa appears in Figure 6. For
comparison, results obtained previously with 440C steel specimens [Jones et al., 2001] are shown for
unformulated and formulated X-2000, Krytox 143AC, and Brayco 815Z As can be seen, the X-1000 fluid
performs better than the unformulated X-2000 fluid and is comparable to the formulated X-2000. In

NASA/CP—2002-211506 336
addition, its normalized lifetime is two orders of magnitude greater than 815Z and 30 times greater than
Krytox 143AC.


QRUPDOL]HGOLIHWLPH RUELWVµ J








; 0$& $ 0$&  $& =
Figure 6. Normalized lifetimes (orbits/µg) from the SOT

Conclusion

Pennzane Synthesized Hydrocarbon Fluid X-1000, a lower molecular weight relative of Pennzane SHF
X-2000, has the expected lower viscosity and lower pour point. Although the volatility is greater, it still
exhibits vacuum outgassing of less than 0.5%. In addition, it performs equally as well as X-2000 as a
boundary lubricant in tribological testing (i.e., low metallic wear rates in pure sliding and low degradation
rates in rolling contact).

References

Jones, W.R., Jr. and Jansen, M.J.,”Space Tribology”, Chapter 31 in Modern Tribology Handbook, Vol. 2,
pp1159-1181, CRC Press, Boca Raton, FL (2001)

Nguyen, Q.N., Jones, W.R., Jr., “Volatility and Wear Characteristics of a Variety of Liquid Lubricants for
Space Applications”, Trib. Trans., 44, pp 671-677, (2001)

Venier, C.G. and Casserly, E.W.,”Multiply-Alkylated Cyclopentanes (MACs): A New Class of Synthesized
Hydrocarbon Fluids”, Lubr. Engr., 47, 7, pp 568-591, (1991)

Venier, C. G., Casserly, E. W., and Gunsel, S., “Tris(2-octyldodecyl) Cyclopentane, a Low Volatility, Wide
Liquid Range, Hydrocarbon Fluid,” J. Syn. Lubr., 9, pp 237-252 (1992).

NASA/CP—2002-211506 337
Carré, D.G., Kalogeras, C.G., Didziulis, S.V., Fleishauer, P.D., Bauer, R., “Recent Experience with
Synthetic Hydrocarbon Lubricants for Spacecraft Applications”, Aerospace Report TR-95 (5935)-3,
(1995).

Casserly, E. W., and C. G. Venier (Pennzoil-Quaker State), “Cycloaliphatics”, in Synthetic Lubricants and
High-Performance Functional Fluids, in Rudnick, L. R., and R. L. Shubkin, eds. Marcel Dekker, NY
(1999).

Brown, P.L., Miller, J.B., Jones, W.R., Jr., Rasmussen, K., Wheeler, D.R., Rana, M., Peri, F., “The Clouds
rd
and the Earth’s Radiant Energy System Elevation Bearing Assembly Life Test”, 33 Aerospace
Mechanisms Symp., Pasadena, California, NASA/CP-1999-209259, pp 197-212, (1999).

VanDyk, S.G., Dietz, B.J., Street, K.W., Jones, W.R., Jr., Jansen, M.J., Dube, M.J., Sharma, R.K. and
Predmore, R.E., “The Role of Bearing and Scar Mechanism Life Testing in Flight Qualification of the
MODIS Instrument”, NASA TM-2001-210896, April, (2001).

Jansen, M.J., Jones, W.R., Jr., Predmore, R.E., and Loewenthal, S.L.,”Relative Lifetimes of Several
Space Liquid Lubricants Using a Vacuum Spiral Orbit Tribometer (SOT)”, NASA TM 2001-210966, June,
(2001).

Jones, W.R., Jr., Poslowski, A.K., Shogrin, B.A., Herrera-Fierro, and Jansen, M.J., “Evaluation of Several
Space Lubricants Using a Vacuum Four-Ball Tribometer,” Trib. Trans., 42, 2, pp. 317-323, (1999).

Kingsbury, E., “Tribology in Slow Rolling Bearings”, Mater. Res. Soc. Sym. Proc., 140, 437-442, 1989.

Jones, W.R., Jr., Pepper, S.V., Jansen, M.J., Nguyen, Q.N., Kingsbury, E.P., Loewenthal, S., Predmore,
R.E., “A New Apparatus to Evaluate Lubricants for Space Applications – The Spiral Orbit Tribometer
(SOT)”, SAE Int. Spring Fuels & Lub. Meet. & Expos., paper 2000-01-1828, June (2000).

Masuko, M., Jones, W., Jansen, R., Ebihara, B., Pepper, S., Helmick, L., “A Vacuum Four-Ball Tribometer
to Evaluate Liquid Lubricants for Space Applications”, NASA TM-106264, (1994).

NASA/CP—2002-211506 338
 ISS SGANT Group Level Offloading Test Mechanism
*
Xilin Zhang

Abstract

The International Space Station (ISS) Space-to-Ground Antenna (SGANT) is used for ISS communication
with earth through the Tracking and Data Relay Satellite (TDRSS). Due to the different speeds of travel
between earth, ISS and TDRSS, a steerable SGANT was required on the ISS.

The mechanical design of SGANT is an unbalanced mechanism with insufficient strength and driving
torque to support and drive itself in a 1G environment. For ground testing, a specially designed offloading
mechanism is required. Basically, the test mechanism must offload the SGANT in a two-axis operation,
allowing the SGANT to move within a specific range, speed and acceleration; therefore the SGANT can
move from elevation 0° to 90° and be tested at both the 0° and 90° positions. The load introduced by the
test equipment should be less than 10.17N-m (7.5 ft-lbf). The on-ground group level tracking test is quite
challenging due to the unbalanced antenna mechanical design and tough specification requirements.

This paper describes the detailed design, fabrication and calibration of the test mechanism, and how the
above requirements are met. The overall antenna is simplified to a mass model in order to facilitate the
offloading mechanism design and analysis. An actual SGANT mass dummy was made to calibrate the
system. This paper brings together the theoretical analysis and the industrial experience that were relied
upon to meet the above-mentioned requirements for the ground test. The lessons learned during the
calibration phase are extremely important for future double or multiple offloading system designs. The ISS
SGANT QM and FM units passed their ground test and the SGANT/Boom fit check successfully, and the
Flight Model (FM) was delivered to SSPF in April 1998. It is now installed on ISS and functioning well, as
shown in Figure 1.

SGANT

Figure 1. SGANT Antenna on ISS

* EMS Technologies Canada, Ltd., Space and Technology Group, Ste-Anne-de-Bellevue, Canada
th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 339
 Introduction

In the space industry, it is common, during on-ground testing to provide a “zero G environment” for
gravity-sensitive hardware. The SGANT is an orbit replaceable communication and tracking antenna. It
provides the Ku-band signal transmitting and receiving capability between ISS and TDRSS. In orbit, it is
mounted on a rigid mast on the space station with the Space-to-Ground Transmitter and Receiver-
Controller (SGTRC) mounted close to the base of the mast. The SGANT is used for communication with
the TDRSS satellites which are in geo-synchronous orbit. The space station on which the SGANT is
mounted is in a 90-minute Low Earth Orbit and maintains a constant attitude with respect to the ground
below. The space station therefore performs one complete rotation with respect to TDRSS about every 90
minutes. This rotation is imparted to the base of the SGANT which must therefore be steered to
compensate. To simulate the steering operation and verify the auto tracking function of the antenna, an
on-ground tracking test is definitely a must.

The on-ground group level tracking test is a challenge due to the unbalanced antenna mechanical design
and tough specification requirement. The specification requirements are summarized below (Ref. 1).

1. Balance and/or offload the SGANT to allow two-axis operation in a 1G environment.

2. Let the SGANT move over a range of at least ± 5° in each axis at any angular speed up to
2
3°/second, and at any angular acceleration up to 5°/second . The additional torque imposed not
exceeding 10.17N-m (7.5 ft-lbf) and 27.12 N-m (20 ft-lbf) maximum, to either of the gimbal axes,
respectively.
3. Limit the static torque on each gimbal axis to less than 10.17N-m (7.5 ft-lbf) when the SGANT is
stationary at the center of this ±5° angular range in each axis, to prevent unwanted rotation when
the SGANT is turned off.
4. Provide ±5° motion for two SGANT positions, namely EL=0°/XEL=0°, and EL=90°/XEL=0°.
5. Provide a facility by which the elevation axis of the SGANT can be rotated to its 90° position using
its own power without imposing an additional torque requirement of more than 27.12 N-m (20 ft-lbf)
to the elevation gimbal axis.
6. The fixture shall be sufficiently light and rigid for the first resonant frequency around either of the
SGANT rotation axes to exceed 2.5 Hz.
7. Avoid applying loads to the SGANT during any phase of the ground test, which corresponds to
factors of safety less than 100%.
8. Be dynamically simple.
9. Be attached to the SGANT at an existing attach point.
10. Define the exact angles involved for software limiting.

During the Preliminary Design Review, a counter balancing weight mechanism was proposed. However,
once the detailed design and analysis started, problems were encountered. The balance weight
mechanism creates a lot of unnecessary load and friction on the flight gimbals. Some of the load
requirements are very difficulty to meet. As a result, the balancing weight mechanism approach was
finally abandoned. In its place, many alternative proposals were considered, the most successful one
being the spring motor offloading test mechanism. To design the offloading mechanism, the following
steps defined herein were taken. Ultimately, the project was successful.

Simplification of the Antenna Structure to a Mass Model

Simplifying the overall SGANT antenna structure to a mass model eliminated much unnecessary analysis work
during the design phase. All SGANT antenna components were simplified to mass points and, using a weightless
bar, all mass points were connected together to form the model. The process of simplifying the SGANT antenna

NASA/CP—2002-211506 340
to a mass model is summarized in Figure 2. Information was based on the antenna mass report from the
CAD design model.

Figure 2. Simplification of the antenna system to a mass model

One can see that the model includes all the necessary mass and C/G location information related to
further work. In the following sections, this model will be frequently used for the offloading system
analysis and design.

Location of the best offloading points

A theoretical analysis of the simplified model helped find the balanced offloading point, which is a key for
the overall offloading system design. Ideal offloading points are those that can balance the overall system
in order to eliminate the torque introduced by the SGANT weight and the offloading system in both the EL
(elevation) and XEL (Cross elevation) gimbals.

From detailed analysis and a series of offloading tests on a mass dummy (the mass dummy design is
discussed later), it was found that for a two-axis gimbaled system like the SGANT the best way to offload
the overall system was to theoretically split the system in two, then use two offloading mechanism
systems to take care of each portion at the subsystem balancing point. This approach can reduce the
load on both gimbal axes. In practice, it was also realized that in the SGANT on-ground test configuration,
the mass load on the EL gimbal will be carried by its bearings, and the margin of load capacity on these
bearings is sufficient to carry that load. Therefore, the following calculations take into account the load
introduced by the offloading mechanism only, and the mass load of the SGANT weight introduced into the
EL gimbal driving mechanism train is neglected.

NASA/CP—2002-211506 341
The system was split into front and rear systems, as shown in Figure 3. Since the rear system is relatively
simple, a lifting test from mass dummy defined its offloading point. This point was selected to be at the
existing Motor Drive Amplifier lifting interface. The final application shows that this point is very close to
the rear system C/G. This is why, in both Figure 2 and Figure 3, the component mass and C/G of the
SGANT rear system are not shown. Should the rear system be more complicated, the calculation could
be done in the same way as that used to define the front system offloading point.

Figure 3 SGANT simplified mass model

Since the MGSE (Mechanical Ground Support Equipment) interface was supplied at the RIRS front
surface (another existing lifting point in the SGANT), the front system lifting point was defined from the
RIRS surface as follows (refer to Figure 3):
n
Since∑ M = 0
1

M =W ∗L
WT • LT − W1 • L1 − W2 • L2 − ... − Wn • Ln = 0
n

∑M
LT = 1

WT

NASA/CP—2002-211506 342
Where:
WT − Total Mass of the Front Subsystem
L T − Distance of the total mass C/G to RIRS front surface
Wn − Subassembly Mass
L n − Distance of the Subassembly mass C/G to RIRS Front Surface

The calculation results on the front subsystem shows that the theoretically balanced lifting point should be
26 mm (1.03”) in front of the RIRS front surface.

Analysis of the load introduced into the gimbals by the offloading system

As per specification requirement No. 1, the load introduced by the offloading mechanism is calculated as
follows. The internal friction of the spring motor offloader is omitted from the following calculation. The
worst case to be considered is when both the EL and XEL gimbals are at the 5° position.

Torque load introduced into EL Gimbal

In this calculation, the side load contribution is the major concern due to the large distance between the
loading point and the EL gimbal center. Refer to Figure 4.
TLel = FLel • D
Where :
TLel − − − Maximum torque load on EL gimbal
FLel = W1 • tanα − − − Side load from the offloadingmechanism
D − − − Distancefrom the lifting point to EL gimbalcenter
tan b
α = tan −1 • ( D • )
H
H − − − Offloading cable length
α − − − Worst case offloading cable angle
F − − − Load on offloading cable
After applying all the actual data into the above-mentioned formula, the maximum moment is 6.24 N-m
(55.24 in–lbf), which is about 34% less than the target load limit. The fabrication and mass analysis report
error is considered less than 10% based on EMS experience. The total margin left is about 24%,
therefore, the offloading geometry arrangement is acceptable.

Load introduced into the XEL gimbal

Since the distance from the loading point to the XEL gimbal center is 330.2 mm (13”) less than for the EL
gimbal, the side load is not a concern in the XEL case. Only the acceleration torque required from the
XEL gimbal needs to be considered. Refer to Figure 5.
TLXel = FLXel • R
Where :
TLXel − − − acceleration driving torque required from the Xel gimbal
W1
FLXel = • a − − − acceleration force required
g
a = α •R
α − − − max.angular acceleration
R − − − lifting point rotation radius

NASA/CP—2002-211506 343
 Figure 4. Load introduced into EL gimbal

Figure 5. Load introduced into XEL gimbal

NASA/CP—2002-211506 344
After applying all the actual data into the above equations, the maximum driving torque required is found
to be only 2.41 lb-in, which is far from the target load limit, and again the geometric arrangement is
acceptable.

Calculation of software limits

Software limits are the second layer of safety device, it is used to avoid extra load caused by over-travel
of the gimbal. During the SGANT group level tracking test, if for any reason the gimbals should reach the
software limit point, the power would turn off or the driving mechanism would be disabled to protect the
SGANT.

In requirement 2, the target offloading mechanism design load is less than 10.17 N-m (7.5 ft-lbf), which
should be the number used to determine the allowed gimbal over-travel angles. The calculation method is
as follows:

Calculation of the EL gimbal software limits

With reference to Figure 4:

β S lim = tan −1 d / D
where :
β S lim − − − Software Stop angle
d = H • tanα
α = tan −1 Flim / W1
Flim − − − Maximum load (Introduction requirement 2)
Flim = Tlim / d
W1 − − − Subsystem weight

Applying all the actual data to the formula, the calculated software limit is 7.525°. 7.5° is the limit selected.

Definition of the XEL gimbal software limits

This calculation is omitted because the torque load introduced by the offloading system in the XEL case is
relatively small. The software limit was set at 7.5° to make it the same as the EL gimbal. It is clear that the
gimbal will not encounter problems within that limit.

Mechanical design of the test system

After all of the above-mentioned offloading points were defined, the mechanical offloading system was
designed as follows. It includes a main structure to support the system on the compact antenna range
turntable, the constant force spring motor offloading system, the SGANT transfer and installation system,
the hard mechanical stops and an electrical stop system. The offloading methods are the main topic
covered in this paper.

During brainstorming sessions, many possible methods were listed for the offloading test. The cable
pulley and weight system and the constant force spring motor offloading system received the highest
tradeoff scores in all respects. Figures 6 and 7 show these two different offloaded test system designs.

NASA/CP—2002-211506 345
The cable pulley and weight system was the first one designed into a CAD model, since it was less costly
and easier to build. But moving the weight system required a lot of space and the detailed design work
would have taken a lot more hours than the spring motor offloading system. Once the main structure and
90° turn system design was completed, the spring motor offloading system came to our attention and the
design was quickly switched to Figure 7.

Figure 6. Cable pulley and weight offloading system Figure 7. Spring motor offloading system

The main component of the spring motor offloading system is a constant-force spring motor. The design
of a constant-force spring motor reflects the basic principles shown in Figure 8. Here, one can see that
the so-called constant-force spring has minimum loading capacity variance when used between points A
and B in the graph. Then a cable wound onto a tapered drum is used to compensate the loading capacity
variance in order to achieve constant force output on the loading cable (Ref 2.).

NASA/CP—2002-211506 346
 Figure 8. Constant-force spring motor

In addition to the offloading system, the overall mechanical design includes a main support structure on a
Scientific Atlanta turntable, to support the overall system in the compact antenna range at EMS. The
system includes an SGTRC support structure and wave-guides. The SGANT installation system includes
a lifting setup and a transfer support structure and, of course, safety devices of various kinds.

Offloading system calibration and lessons learned

The calibration of the test system setup is shown in Figure 9.

Using a mass dummy to calibrate a complicated flight system test setup can protect the flight hardware; it
allows the setup to be verified before the flight hardware is installed into it. The mass dummy must be
representative of the mass of the flight system. In the case of the SGANT, at the time that the mass
dummy was being designed, the SGANT CAD model design was unfinished and full information was not
available. Therefore a mass C/G tuning system needed to be added to the system. A partial Design
Verification Test Model (DVTM) gimbal was finally used to represent the gimbal movement and most
importantly, torque sensors were installed into both the EL and XEL gimbal axes to measure the overall
system load introduced into the flight SGANT by the offloading system (Figure 9).

Mass dummy design

The mass dummy design is shown in Figure 10. The C/G tuning weights need to be sufficient to
compensate any design change and manufacturing error. So, the tuning masses have to be relatively
large and have a large range of motion.

The purpose of the torque measurement assembly is to qualify and monitor the overall test process from
beginning to end. The design is shown in Figure 10, in which torque sensors are installed on the DVTM
gimbal. One side is installed on the moving portion of the gimbal and the other is mounted on the
stationary side. A Micro B reading instrument is connected to the sensor. At a stress-free condition, the
instrument reading should be set to zero.

NASA/CP—2002-211506 347
 Figure 9. Calibration setup Figure 10. Mass dummy design

SGAG Test Fixture Calibration Report

1. The table below lists the maximum torque load on the DVTM gimbal during horizontal lifting
operation:

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) -2.711 N-m (-24 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 1.8 N-m (16 lbf-in)

2. Maximum torque load on the DVTM gimbals during transfer between the horizontal lifting
configuration and rest on the temporary support structure:

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) 1.8 N-m (16 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 30.05 N-m (266 lbf-in)**

NASA/CP—2002-211506 348
3. Maximum torque load on the DVTM gimbal when the dummy mass is installed on the SGAG test
fixture (when EL Gimbal is at the 0° test configuration):

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) 8.7 N-m (77 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 10.06 N-m (89 lbf-in)

4. Maximum torque required to drive the gimbal within ±5° when the EL gimbal is at the 0° test
configuration:

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) 9.38 N-m (83 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 7.63 N-m (67.5 lbf-in)

5. Maximum torque load on the DVTM gimbal when transferring the dummy mass from the 0° to the
90° configuration:

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) 0.8 N-m (7 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 0.34 N-m (3 lbf-in)

6. Maximum torque required to drive the gimbal within ±5° when the EL gimbal is at the 90° test
configuration:

Max. load allowance Max. load recorded*

XEL Gimbal 27.11 N-m (240 lbf-in) 9.38 N-m (83 lbf-in)

EL Gimbal 27.11 N-m (240 lbf-in) 7.64 N-m (67.5 lbf-in)

* Maximum loads determined by dummy load measurement.

** Locking pin engaged, load acceptable.

Based on the calibration report, one can observe that during the hoist and transfer operation, a large
unexpected torque load was introduced into the flight gimbals, which is over the maximum load limit. It
was considered a potentially fatal load to the flight gimbals. These operations were repeated several
times to find out the reason. The operator was instructed to reduce the hoist speed to the minimum.
However the recorded loads were unpredictable, between 22.5 N-m (200 in-lb) and 56.5 N-m (500 in-lb).
Finally, two gimbal lock pins 9.53-mm diameter (3/8” dia.) were installed into the flight gimbals, especially
for the hoist and transfer operation. Without the calibration exercise the flight gimbals could have been
destroyed during the test.

NASA/CP—2002-211506 349
 Lessons Learned

For a multiple gimbal mechanism, the gimbals must be locked for transfer and hoist lifting operations.

Define the loading point

Since the front of the RIRS (a part of the SGANT) provides some MGSE holes for a tooling interface, and
the calculation showed that they are very close to the ideal offloading point, it was decided to directly
hook up from the MGSE holes with a bracket, a solution that was in the original design. It is illustrated in
the calibration setup in Figure 9.

The first calibration showed that driving about only 2 degrees off from the starting point, a potentially fatal
load was introduced into the XEL gimbal. A quick analysis on the mass model is shown in Figure 11,
where one can see that force F2 increases rapidly when the XEL gimbal starts to move.

After modification, the lifting point was moved to the position shown in Figure 11. The mechanism used to
move the lifting point is shown in Figure 9. The load introduced into the XEL gimbal is now extremely
small, within its 5 degrees driving range, and is barely detectable.

Figure 11. Lifting point selection mistake

Lesson learned:
To define the offloading point, simply lining up the offloading point to the C/G is not enough. The selection
should be based on the analysis on all 3 dimensional locations to get the best offloading results.

Moving the SGANT between EL=°/XEL=0°, and EL=90°/XEL=0°

In requirement 4 it is stated that the test system must provide a ±5° motion for two SGANT positions,
namely EL=°/XEL=0°, and EL=90°/XEL=0°. Refer to Figure 6 and Figure 7.

NASA/CP—2002-211506 350
To turn the antenna system on the test setup from the 0° to the 90° testing position, logically the telemetry
from the gimbals driving system should be used. Using this telemetry to synchronize the offloading
system drive in order for the gimbal and the offloader to move from one position to another. To do so, a
servo system is required. However, for the SGANT the servo system was eliminated as a cost-saving
measure. It was decided to use an operator to crank the gearbox and synchronize the drive manually.
Let us see the results.

The problem started in the calibration phase. The synchronized drive of the overall system was almost
impossible to control manually. Two technicians had to be trained for a week on the mass dummy to
perform the transfer drive. In addition, a five degrees warning limit switch had to be added to the system.
During the transfer, whenever the flight gimbal and the offloader had an angular difference of 5 degrees
or more, the alarm switch went on and the operator had to adjust the offloader position. In the meantime,
the power went off, and the system had to be restarted. It became the most time-consuming operation of
the SGANT group level test phase. The total cost in hours was much more than a servo driving system.
Since the operator had to climb up to the top of the main structure about 15 meters high, safety also
became an issue.

Lesson learned:
Synchronizing the drive of two mechanical systems manually is a very difficulty task. Depending on the
accuracy required a servomechanism should be used to automate the task and obtain a proper result.
Figure 12 shows the system after final calibration and modification.

Conclusion

The method finally used to offload the SGANT two-axis gimbaled system with two synchronized
offloading systems has been proven to be correct. The torque load introduced into the system is kept well
below acceptable limits. The methods used in this case can be developed and used to design other multi-
gimbal system offloading mechanisms. Calibrating the system using a mass dummy to represent the flight
hardware and installing the necessary instruments to measure torque loading is an efficient method to
help prevent the possibility of damage to flight hardware.

Acknowledgements

The author would like to thank his colleagues Graham Arbery and Peter Richter for their technical advice
on the project and for proofreading this paper to make it ready for publication. Thanks also due to Ms.
Paule Mercier & Catherine Yi Zhang for their sincere effort in editing this paper.

References

1. Arbery, Graham, SGS Control system ground test requirements specification. p12, 1995, EMS
Technologies

2. Caravaggio, Levino, Large item offloading, 31 March 1992 Spar Aerospace Ltd.

3. Zhang, Xilin, ISSA MGSE user’s manual 1998 14 April. EMS Technologies.

NASA/CP—2002-211506 351
 Figure 12. SGANT Group level test set up

NASA/CP—2002-211506 352
 Relationship Between the Difficulty Index and the Evaluation Accuracy of
Ground Deployment Testing
*
Hironori Ishikawa and Akira Meguro

Abstract

An index that well reflects the difficulty of ground deployment testing is the ratio of gravity torque to
deployment torque. The object of this study is to quantitatively clarify the relationship between this index
and the evaluation accuracy of ground deployment testing. We performed ground deployment tests using a
simple planar truss in micro gravity and gravity environments. Ground tests in which the index value of the
planar truss is increased are also performed. A comparison of experimental and analysis results yields an
empirical equation that sets a linear relationship between the difficulty index value and the evaluation error
of the ground testing. We conclude that the modular approach allows large space antennas to be designed
that can be well assessed by ground testing.

Introduction

Large deployable space structures such as solar paddles, deployable antennas and sun shields have been
developed for future satellites. They are designed to have sufficient deployment reliability after launch
environment and shape stability against thermal and vibration disturbances. Deployment reliability is the
most important factor in avoiding the failure of missions. Therefore, deployable space structures should be
evaluated by ground deployment testing before launch to confirm their deployment reliability. When we
perform ground deployment testing, gravity compensation equipment such as suspension systems and
support systems should be used to avoid excessive gravity torque and force. In general, we cannot expect
gravity compensation equipment to well support the moving parts; one reason is that it is impossible to
create a really accurate force and torque profile. As a result, deployment behavior in ground deployment
testing is not the same as in-orbit motion, so an analysis model must be used to evaluate the influence of
the gravity compensation equipment on the deployment characteristics. The gravity force and torque
increase as the space structure becomes larger. This makes it very difficult to evaluate the actual
characteristics of the deployable structure. In these cases, we are actually evaluating the characteristics of
the ground testing equipment, instead of evaluating the characteristics of the deployable structure.

It should be possible to predict the accuracy of deployment testing for a deployable structure of given size.
First, we propose the difficulty index value (DIV) to define the difficulty of ground deployment testing. The
index is the ratio of gravity torque to deployment torque. Figure 1 indicates DIV calculated by Mitsugi’s
formulation [1] for launched and planned space deployable structures [2]-[6]. Here, d represents the stowed
diameter. Next, we quantitatively evaluate the difficulty index and the accuracy with which the deployment
characteristics can be estimated from ground testing. Finally, we introduce an index that indicates the size
limit of deployable structures that still permits accurate ground test evaluations. In a past paper [7], we
clarified the relationship between DIV (1 – 100) and the evaluation error of the ground testing, and obtained
an empirical equation that set a linear relationship between DIV and evaluation error of the ground testing.
However, the DIV of most recent large space structures exceeds 100 so the reliability of the empirical
equation is not valid for these structures. Therefore, we extend the empirical equation to cover DIVs over
100. This paper describes the relationship between DIV and the evaluation error of the ground testing for
DIVs of up to approximately 1000. We revise the empirical equation and consider the evaluation accuracy of
ground tests using the example of the modular mesh antenna developed by NTT [8] -[10].

* NTT Network Innovation Laboratories, NTT Corporation, Kanagawa, Japan

th
Proceedings of the 36 Aerospace Mechanisms Symposium, Glenn Research Center, May 15-17, 2002

NASA/CP—2002-211506 353
 Planar Truss

The deployable structure used in deployment testing was designed and fabricated based on the deployable
truss structure of the modular mesh antenna developed by NTT [9]. The planar truss was selected to avoid
uncertainties in analysis models. Figure 2 shows the planar truss. The planar truss is composed of two rib
structures on both sides of the center beam. Rib structure consists of the upper beam, the diagonal beam,
the lower beam, the side beam and two synchronous beams. Each beam is joined by a rotation hinge. The
right and left diagonal beams and lower beams are linked by a plate. The planar truss is deployed and
stowed by moving plate (A) along the center beam. Teflon forms the friction surface of plate (A) to decrease
friction against the center beam. Two constant force springs and a motor are mounted on plate (A). The
constant force springs provide the deployment force to move the planar truss. Total spring force is 8.2 N.
The planar truss stows when the motor winds in the drive cable, and deploys when the drive cable is
released. When plate (A) rises 0.135m along the center beam, the planar truss is fully stowed. The beams
and plates are made of aluminum. The diameter of the beam is 0.01 m and its thickness is 0.001 m. The
length of the planar truss is 1.2 m (deployed) and 0.1 m (stowed). Its weight is 4.7 kg. One of our intentions
is to examine the effect of the tension in the mesh cable that connects the upper-face of both side beams.
Therefore, a load cell (weight = 0.090 kg) and mesh cable were attached to the planar truss. In this paper,
however, we describe in the case in which the tension of the mesh cable was 0 N.

Deployment Testing

We performed two kinds of deployment tests. One was performed under micro gravity environment, the
other was performed on the ground. We assumed the deployment characteristics measured under micro
gravity environment represented the true deployment characteristics of the planar truss.

Micro gravity deployment testing

Deployment testing under micro gravity environment was performed in the jet-airplane. Figure 3 shows the
deployment testing setup in a jet-plane. The testing time was set to 20 seconds, which was decided by
considering the duration of the micro gravity environment. The planar truss did not stow completely,
however, the planar truss can stow to the configuration shown in Figure 3 within the given testing time. We
measured the tension of the drive cable that was wound in or released by the motor, and the displacement
of plate (A) from the bottom. The deployment drive force can be calculated by adding the tensions of the
drive cable in deployment and stowing motion and dividing by 2 at each position. The friction is cancelled in
this process because of coulomb friction. The displacement of plate (A) indicates the deployment position.
The measurements during deployment and stowing were performed three times under each condition.

Ground deployment testing

Figure 4 shows the deployment testing setup on the ground. The duration of the ground tests equaled that
of the micro gravity tests. The suspension system was applied to offset the gravity force because this test
configuration can be easily evaluated by available analysis tools. In the suspension system, the lower side
of the center beam was fixed to the floor and the upper parts of the side beams were supported by the
suspension cable. The planar truss deployed in the gravity direction. The height of the suspension position
is 1.6 m from the planar truss as shown in Figure 4. Two suspension cables were tensioned by the counter
weights through pulleys. The weights were 0.220 kg (W1) and 0.130 kg (W2) respectively. The weight on
the left side was larger than that of right side because the load cell was installed on the left side. We
attached weights to the stands on both side beams to increase the DIV of the planar truss. When adding the
weight to the stands shown in Figure 2, the counter weights (W1, W2) were correspondingly increased. We
tested the planar truss with large DIV. DIV can be altered by changing the weight of the stands. To minimize
the weight needed to increase DIV, the spring force was changed. The relations among the DIV of the
planar truss, the weight, and the spring force are shown in Table 1. As the planar truss stows, the gravity
torque changes. Therefore DIV changes with the deployment position. Table 1 presents the maximum
values.

NASA/CP—2002-211506 354
 Analysis Model

An analysis was conducted using SPADE (Simple coordinate Partitioning Algorithm based Dynamics of
finite Elements) developed by NTT [11][12]. SPADE can analyze the deployment behavior of elastic
structures. The analysis model of the planar truss was made as shown in Figure 5. We made the analysis
model based on the deployment drive force measured by micro gravity testing. The suspension system and
gravity force were considered to predict the deployment drive force measured by the ground testing.

Experimental Results

Difference between the experimental and analysis results

Figure 6 shows the experimental result obtained by ground testing. The analysis result is also plotted to
show prediction accuracy. There is approximately 10 % difference between the experimental and the
analysis results. The reasons for the difference are thought to be errors in the suspension position, the
counter weight and the center of gravity of the planar truss. In these ground tests, the measurement error of
the suspension position is less than 1 % and the measurement error of the counter weight is less than 0.1 %.
Therefore, we assumed the error of the analysis model for the suspension position and the counter weight
is small compared to the error of the center of gravity of the planar truss. Accordingly, we considered the
accuracy of the experimental result was mainly determined by the error in estimating the center of the
gravity of planar truss [7].

To confirm this hypothesis, we re-measured the position of the pulley and the suspension cable. The pulley
and the suspension cable are shown in Figure 7. Though there were some differences between the
suspension cable and the center of the pulley, they occupied the same position in the analysis model. The
differences were 0.015 m in the gravity direction and 0.012 m in the horizontal direction. We also measured
the tension of the suspension cable, when we pulled or loosened the suspension cable with the counter
weight on one end. These tests were performed for two kinds of counter weights (W1, W2). The friction
between the pulley and the suspension cable was calculated by dividing the difference of the tensions
between pulling and loosening by 2. The friction was 0.002kg in both counter weights. Therefore we
subtracted 0.002 kg from the counter weight in the analysis model of the stowing motion. Figure 8 shows
the result of the analysis model when the suspension position and the counter weight were changed. The
difference between the experimental and analysis results decreased by 0.05N, which is only 0.5 % of the
deployment drive force. It is clarified that the error of the deployment drive force is not strongly influenced by
the accuracy of the analysis model for the suspension position and the counter weight.

Next, we must pay attention to the mass property. They are total weight, the center of gravity, and inertial
moment. Total weight was measured within 0.1 % and inertial moment was assumed to be negligible. The
center of gravity is uncertain. Therefore the center of gravity should be measured. The weight of the rib
structure without the suspension cable was measured to verify the center of gravity of the planar truss. The
measured value was compared to the analysis value as shown in Table 2. The difference of weight in the
experimental and analysis models is approximately 0.02kg even though the weights of the analysis model
corresponded to each beam. The result of the revised analysis model is shown in Figure 9. This model was
revised in terms of not only the center of gravity but also the suspension position and the counter weight. In
Figure 9, the deployment drive force corresponds to the experimental result. This quantitatively proves that
the accuracy of the experimental result is mainly determined by the error of the center of gravity of the
planar truss. The error of the deployment drive force, the difference between the experimental and analysis
values, is approximately 10 % for the simple structure whose DIV=1, but it is approximately 3 % if the mass
properties are calculated in detail. In actual deployment tests, it is very difficult to estimate the true mass
properties of large deployable structures during deployment. Therefore, we used the former analysis model
to clarify the relationship between DIV and the evaluation error of the ground testing.

Evaluation error
The evaluation error of the ground testing is shown in Figure 10. Line (A) shows the analysis based on the
deployment drive force measured by micro gravity testing. Line (B) shows the analysis based on the
deployment drive force measured by ground testing. For line (B), the deployment drive force was updated
by the ground testing result, means that the spring force of the analysis model was changed to minimize the
difference between the experimental and analysis results in the ground testing. The evaluation error of the
ground testing can calculated by dividing the difference between deployment drive force updated by the
ground testing result and true value (deployment drive force measured by micro gravity testing) by true

NASA/CP—2002-211506 355
value. The same process has to be done for other DIVs. The relationship between DIV (1 - 845) and the
evaluation error of the ground testing is shown in Figure 11. Evaluation error of the ground testing is given in
units of percentage. The evaluation error of the ground testing increases as DIV increases as shown in
Figure 11.

Figure 12 shows the deployment drive force for DIV = 160. Here “DDF” denotes the deployment drive force,
“Best Fit” denotes the profile that minimizes the difference between the experimental and analysis results.
The minimized difference value becomes larger when the fluctuation increases. Therefore, we considered
that the fluctuation value of the evaluation error of the ground testing is the ratio of the root mean square of
the difference between DDF experiment and DDF Best Fit to the true value. The relationship between DIV
and the evaluation error of the ground testing including the fluctuation is shown in Figure 13. Evaluation
error of the ground testing is given in units of percentage and o represents the evaluation error of Figure 11.
The evaluation error of the ground testing is minimum when DIV =1 whereas the evaluation error of the
ground testing including fluctuation is maximum.

Empirical equation
By assuming a linear relationship between DIV and the maximum value of the evaluation error of the ground
testing, we obtained the following empirical equation (1).

R= 0.1i+13.8 (1)

Here, i represents DIV and R indicates the evaluation error of the ground testing.
In the empirical equation, the evaluation error of the ground testing is approximately 100% when DIV=800.
Therefore, we can conclude that it is infeasible to use ground testing to evaluate deployable structures
whose difficulty index value exceeds 800.

The evaluation error of the ground testing was assessed using a planar truss and the suspension system.
This model is very simple and it was easy to evaluate by the analysis tool. We considered that the
evaluation error of the ground testing in this model was the smallest among all realistic deployable
structures. Accordingly, this index can be used for any deployable space structure. We will make a model
that consists of three connected planar trusses and calculate the evaluation error of the ground testing. We
will verify the empirical equation and describe the results in a following paper.

Modular Mesh Antenna

One of the methods used to estimate the deployment characteristics of large space structures by ground
testing is to design a large space structure using modular components. The desired size can be created by
combining several basic modules, and the characteristics of the entire structure can be predicted from the
characteristics of the basic module and the effect of combining them [13]. An excellent example of a module
structure is the large modular satellite antenna, called the modular mesh antenna, that NTT has been
developing.
The 18 m X 17 m modular mesh antenna consists of 14 basic modules. Each basic module, which is 4.8 m
in diameter and 0.6m thick, consists of a mesh reflector and a deployable truss structure as a support
structure, as shown in Figure 14 [14][15]. Module construction is shown in Figure 15. The deployable truss
structure is composed of six rib structures that are located around a center beam. The deployable truss
structure has slide and coil springs, which provide the deployment force. Each rib structure consists of the
upper beam, the diagonal beam, the lower beam, the side beam and two synchronous beams. The rib
structure deploys or stows by moving them simultaneously.

We applied the empirical equation to the modular mesh antenna. The characteristics of the entire structure
can be predicted from the characteristics of the basic module and the effect of combining them. Therefore,
it is necessary to test the maximum number of modules combined at one part. The relations among the
number of modules, the DIV of the modular mesh antenna, and the evaluation error of the ground testing
are plotted in Figure 16. It is necessary to test three modules because three modules can be combined at
the same point as shown in Figure 14. In this case, it is thought that the accuracy of the deployment
characteristics is sufficient because the evaluation error of the ground testing is approximately 20 %. If the

NASA/CP—2002-211506 356
evaluation accuracy of the ground testing is insufficient, we can improve the evaluation error of the ground
testing by changing the module size or spring force. Therefore, we consider that this index can be used in
designing module size and spring force margin considering the deployment testing accuracy.

Conclusions

We define an index that predicts the difficulty of ground deployment testing as the ratio of gravity torque to
deployment torque. The index can be used to assess the maximum size of deployment structures for which
ground deployment testing is feasible. A simple planar truss, whose difficult index value was changed from
1 to approximately 1000, was used to estimate the relationship between the index and the evaluation
accuracy of ground testing. Our results can be summarized as follows.

(1) The difference of the deployment drive force between experimental and analytical results is
approximately 10 % for the simple structure whose DIV=1 and approximately 3 % if the mass property
is estimated in detail.
(2) We obtained a new empirical equation that indicates a linear relationship between DIV (up to
approximately 1000) and evaluation error of the ground testing.
(3) We consider that the modular approach allows large space antennas to be designed that can be well
assessed by ground testing.
(4) We believe that this index can be used to design module size and spring force margin considering the
deployment testing accuracy.

Acknowledgements

The authors would like to thank Dr. Hideki Mizuno and Dr. Mitsunobu Watanabe of NTT Network Innovation
Laboratories for their support and encouragement. They are also grateful for the many comments made by
Dr. Jin Mitsugi of NTT Network Innovation Laboratories.

References

[1] Jin Mitsugi and Tetsuo Yasaka, Deployable Modular Mesh Antenna –Concept and Feasibility,
Proceedings of 17th International Symposium on Space Technology and Science, vol. 1, 1990, PP.
599-604.
[2] Deployment of the 15 Meter Diameter Hoop Column Antenna, NASA Contractor Report 4938, Harris
Corporation, 1986.
[3] M.C.Natori, Tadashi Takano, Toshio Inoue, and Takayuki Kitamura, Deployable Mesh Antenna Structure
for MUSES-B Spacecraft, 46th International Astronautical Congress, IAF-95-I.1.04, 1995.
[4] Mark W. Thomson, The Astromesh Deployable Reflector, 6th International Mobile Satellite Conference,
pp. 230-233, 1999.
[5] Hiroshi Sakamoto, Kenji Ueno, and Kohji Horikawa, On-Orbit Test of Multibeam Communication
Equipment on Kiku-6 in Elliptical Orbit 47th International Astronautical Congress, IAF-96-M.2.0.2, 1996.
[6] Akira Meguro, Akio Tsujihata, Naokazu Hamamoto, and Masanori Honma, Technology status of the 13
m aperture deployment antenna reflectors for Engineering Test Satellite VIII, Acta Astronautica Vol. 47, Nos.
2-9, pp. 147-152, 2000.
[7] Hironori Ishikawa and Akira Meguro, High accuracy evaluation method of deployment characteristics for
th
large deployable structures, 9 European Space Mechanisms and Tribology Symposium, 2001. pp.
137-144.
[8] Mitsunobu Watanabe, Kazuhide Ando, Akihiro Miyasaka, and Satoshi Harada, Mechanical Error
Evaluation on Onboard Modular Mesh Antenna, 50th International Astronautical Congress, IAF-99-I.1.01,
1999.
[9] Kazuhide Ando, Akihiro Miyasaka, Hironori Ishikawa and Mitsunobu Watanabe, The Modular Mesh
Reflector developed at NTT, International Mobile Satellite Conference 99, 1999.
[10] Akira Meguro, Satoshi Harada, and Mitsunobu Watanabe, Key Technologies for High Accuracy Large
Mesh Antenna Reflectors, 51st International Astronautical Congress, IAF-00-I.1.0.5, 2000.

NASA/CP—2002-211506 357
[11] Jin Mitsugi, Direct Coordinate Partitioning for Multibody Dynamics Based on Finite Element Method,
Proceedings of 36th Structures Structural Dynamics and Material Conference, AIAA-95-1442-CP, May
1995, pp. 2481-2487.
[12] Jin Mitsugi, Yumi Senbokuya, Dynamic Analysis of Cable-Driven Flexible Multibody Systems and It’s
Experimental Verification, 37th Structures Structural Dynamics and Material Conference, AIAA-96-1484,
May, 1996.
[13] Hironori Ishikawa, Kazuhide Ando, The Effect of the Synchronization Errors in the Deployment Motors
of a Modular Mesh Antenna, Proceedings of 22nd International Symposium on Space Technology and
Science, vol. 2, 2000, PP. 1541-1546.
[14] Mitsunobu Watanabe, Jin Mitsugi, Akihiro Miyasaka, Masashi Shimizu, large Space Antenna Structural
Design Technology Status, 49th International Astronautical Congress, IAF-98-I.1.01, 1998.
[15] Jin Mitsugi, Tetsuo Yasaka, A Modular Approach to Build a Large Space Antenna, Proceedings of 42nd
Congress of the International Astronautical Federation, IAF-91-315, October 5-11, 1991.

NASA/CP—2002-211506 358
 Table 1. Relations among the difficulty index value, the weight, and the spring force

Difficulty index value Weight (kg) Spring force (N)

1 0 8.2
26 0.5 6.3
45 1.0 6.3
84 2.0 6.3
122 3.0 6.3
160 4.0 6.3
415 4.2 2.5
606 6.2 2.5
845 8.6 2.5

Table 2. Weight of the rib structure in experimental and analysis models

Measurement Weight (kg)

point Actual model Analysis model

Left 0.222 0.202

Right 0.135 0.115

1.E+06

1.E+05 d=0.5m ETS- VIII

1.E+04 Unfurlable antenna Modular mesh
Difficulty index

Hoop column
1.E+03
Pe tal VSOP d=1.0m
1.E+02 ETS- VI

1.E+01
d=2.0m

1.E+00
0 5 10 15 20 25
Antenna diamet er(m)

Figure 1. Difficulty index value

NASA/CP—2002-211506 359
 0.6m 0.6m

Load cell Drive cable Mesh cable

0.03m
Synchronous Upper beam
Center beam
Plate(A) beam Side beam 0.15m
Diagonal beam

Plate(B) Lower beam

Constant Stand
Rib structure Motor force spring Rotaional hinge

Figure 2. Planar truss

Stowed configuration Deployed configuration

0.1m
Constant Constant Constant Constant
force spring force spring force spring force spring
Motor Motor

Figure 3. Micro gravity testing setup

Pulley Pulley
1.3m

Counter
Counter
weight(W1)
weight(W2)
1.6m

Planar truss

Load cell

Figure 4. Ground testing setup

NASA/CP—2002-211506 360
 Analysis model
Suspension system

Deployment drive force Deployment drive force

based on true value predicted in ground
testing

Gravity force Comparison

True value Test result

Deployment drive force Deployment drive force
measured by micro measured by ground testing
gravity testing

Figure 5. Analysis model

11
Experiment
10
Deployment drive force

9
Analysis
8

6
0 0.02 0.04 0.06 0.08
Deploy slide length(m) St ow

Figure 6. Difference between experimental and analysis results (DIV =1)

NASA/CP—2002-211506 361
 Pulley

0.015m
Pulley
0.012m
Suspension
cable
Suspension
cable

Gravity direction Horizontal direction

Figure 7. Position of the pulley and the suspension cable

10

New model (revise d suspension position and

Deployment drive force

9.5 count er weight )

Normal mod el
9

8.5

8
0 0.02 0.04 0.06 0.08

Deploy slide length(m) St ow

Figure 8. Influence of the error in suspension position and the counter weight

NASA/CP—2002-211506 362
 11
New model (revised cent er of gravity)
Deployment drive force 10

Experiment
9
Normal mod el
8

6
0 0.02 0.04 0.06 0.08

Deploy slide length(m) St ow

Figure 9. Influence of the error in the center of gravity of the planar truss

Analysis model
Suspension system Line (A)
Suspension position
Line (B)
Deployment drive force Counter weight
based on true value Deployment drive force
predicted in ground
Deployment drive force testing
updated by the ground
testing result
Gravity force
Evaluation error

True value Test result

Deployment drive force Deployment drive force
measured by micro measured by ground testing
gravity testing

Figure 10. Procedure of calculating the evaluation error of the ground testing

NASA/CP—2002-211506 363
 50

Evaluation error 40

30

20

10

0
0 100 200 300 400 500 600 700 800 900
Difficulty index value

Figure 11. Relationship between DIV and the evaluation error of the ground testing

11
DDFexperim et Fluctuation
value
DDFBest Fit
Deployment drive force

9
DDFanalysis

5
0 0.02 0.04 0.06 0.08
Deploy slide length(m) St ow

Figure 12. Deployment drive force obtained by the ground testing and analysis (DIV = 160)

NASA/CP—2002-211506 364
 150

100
Evaluation error

R= 0.1i+ 13.8

50

0
0 200 400 600 800 1000
Difficulty index value

Figure 13. Relationship between DIV and the evaluation error of the ground testing including
fluctuation

Module 4.8m

Mesh surface

18m

0.6 m

17m Deployable truss

structure

Modular mesh antenna Module composition

Figure 14. Modular mesh antenna

NASA/CP—2002-211506 365
 Deployed configuration Stowed
configuration

Rib
structure

Coil spring
Lower beam
Slide
spring

Side beam
Center beam

Upper member
Diagonal beam Synchronous beam

Figure 15. Module construction

200

150
Evaluation error

7module
100

50 3module

1module
0
0 200 400 600 800 1000 1200 1400
Difficulty index value

Figure 16. Evaluation error of the ground testing of modular mesh antenna

NASA/CP—2002-211506 366
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson
Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
April 2002 Conference Publication
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS

Proceedings of the 36th Aerospace Mechanisms Symposium

WU–706–85–00–00
6. AUTHOR(S)

Edward A. Boesiger and Fred B. Oswald, compilers

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION

REPORT NUMBER
National Aeronautics and Space Administration
John H. Glenn Research Center at Lewis Field E–13273
Cleveland, Ohio 44135 – 3191

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING

AGENCY REPORT NUMBER
National Aeronautics and Space Administration
Washington, DC 20546– 0001 NASA CP—2002-211506

11. SUPPLEMENTARY NOTES

Edward A. Boesiger, Lockheed Martin Space Systems Company, Sunnyvale, California 94088; and Fred B. Oswald,
NASA Glenn Research Center. Responsible person, Fred B. Oswald, organization code 5950, 216–433–3957.

12a. DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE

Unclassified - Unlimited
Subject Category: 15 Distribution: Nonstandard
Available electronically at http://gltrs.grc.nasa.gov/GLTRS
This publication is available from the NASA Center for AeroSpace Information, 301–621–0390.
13. ABSTRACT (Maximum 200 words)

The Aerospace Mechanisms Symposium (AMS) provides a unique forum for those active in the design, production, and
use of aerospace mechanisms. A major focus is the reporting of problems and solutions associated with the development
and flight certification of new mechanisms. Organized by the Mechanisms Education Association, the National Aeronau-
tics and Space Administration and Lockheed Martin Space Systems Company (LMSSC) share the responsibility for
hosting the AMS. Now in its 36th year, the AMS continues to be well attended, attracting participants from both the
United States and abroad. The 36th AMS, hosted by the Glenn Research Center (GRC) in Cleveland, Ohio, was held May
15, 16, and 17, 2002. During these 3 days, 32 papers were presented. Topics included deployment mechanisms, tribology,
actuators, pointing and optical mechanisms, International Space Station mechanisms, release mechanisms, and test
equipment. Hardware displays during the supplier exhibit gave attendees an opportunity to meet with developers of
current and future mechanism components.

14. SUBJECT TERMS 15. NUMBER OF PAGES

Spacecraft components; Solar array deployment; Bearings; Tribology; Covers; Latches; 382
16. PRICE CODE
Dampers; Retraction mechanisms
17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT
OF REPORT OF THIS PAGE OF ABSTRACT
Unclassified Unclassified Unclassified
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. Z39-18
298-102