https://ntrs.nasa.gov/search.jsp?

R=20110007070 2017-10-31T12:37:18+00:00Z

19.1 Attitude Determination and Control Systems

Scott R. Starin, NASA Goddard Space Flight Center

John Eterno, Southwest Research Institute

In the year 1900, Galveston, Texas, was a bustling direct hit as Ike came ashore. Almost 200 people in the
community of approximately 40,000 people. The Caribbean and the United States lost their lives; a
former capital of the Republic of Texas remained a tragedy to be sure, but far less deadly than the 1900
trade center for the state and was one of the largest storm. This time, people were prepared, having
cotton ports in the United States. On September 8 of received excellent warning from the GOES satellite
that year, however, a powerful hurricane struck network. The Geostationary Operational Environmental
Galveston island, tearing the Weather Bureau wind Satellites have been a continuous monitor of the
gauge away as the winds exceeded 100 mph and world’s weather since 1975, and they have since been
bringing a storm surge that flooded the entire city. The joined by other Earth-observing satellites. This weather
worst natural disaster in United States’ history—even surveillance to which so many now owe their lives is
today—the hurricane caused the deaths of between possible in part because of the ability to point
6000 and 8000 people. Critical in the events that led to accurately and steadily at the Earth below. The
such a terrible loss of life was the lack of precise importance of accurately pointing spacecraft to our
knowledge of the strength of the storm before it hit. daily lives is pervasive, yet somehow escapes the notice
of most people. But the example of the lives saved from
In 2008, Hurricane Ike, the third costliest hurricane ever Hurricane Ike as compared to the 1900 storm is
to hit the United States’ coast, traveled through the Gulf something no one should ignore. In this section, we will
of Mexico. Ike was gigantic, and the devastation in its summarize the processes and technologies used in
path included the Turk and Caicos Islands, Haiti, and designing and operating spacecraft pointing (i.e.
huge swaths of the coast of the Gulf of Mexico. Once attitude) systems.
again, Galveston, now a city of nearly 60,000, took the

Figure 19-1: Satellite image of Hurricane Ike (NASA image).

Attitude is the three-dimensional orientation of a spacecraft’s rotational velocity; this term is where
vehicle with respect to a specified reference frame. certain other control actuators enter into the dynamics
Attitude systems include the sensors, actuators, as so-called internal torques. The third term shows how
avionics, algorithms, software, and ground support changes in the spacecraft moment of inertia
equipment used to determine and control the attitude of (representing how mass is distributed in the spacecraft),
a vehicle. Attitude systems can have a variety of names, such as by solar array articulation, can affect attitude
such as attitude determination and control system dynamics; in the absence of changes in mass properties,
(ADCS), attitude ground system (AGS), attitude and the third term disappears. The fourth term is called the
orbit control system (AOCS), guidance, navigation and gyroscopic torque, and it shows how the angular
control (GNC), or whatever other term describes the momentum appears to change direction, but not
designers’ focus in achieving the attitude needs of a magnitude, in the spacecraft’s frame of reference when
particular mission. When we use an acronym in this the spacecraft is rotating. All these effects combine to
section, we will use ADCS, but any given specialist determine the rate of change of the angular velocity on
may be more familiar with other terms. the left-hand side.

Spacecraft attitude changes according to the Attitude determination is the process of combining
fundamental equations of motion for rotational available sensor inputs with a knowledge of the above
dynamics, the Euler equations, here expressed in vector spacecraft dynamics to provide an accurate and unique
form: solution for the attitude state as a function of time,
either onboard for immediate use, or after the fact (i.e.
H!=T - ro ×H post-processing). With the powerful microprocessors
now available for spaceflight, most attitude algorithms
This vector equation represents the conservation that used mainly to be performed as post-processing can
equations for the physical vector quantity of a body or now be programmed as onboard calculations.
collection of bodies called angular momentum, which is Therefore, though there are still good engineering
denoted by H. The angular velocity of the spacecraft, reasons for certain processes to be performed only by
omega, is related to H by the equation ground-based attitude systems, it will be sufficient to
focus our attitude determination discussions in this
H=Ir
o +h chapter on the design and implementation of onboard
systems.
where I is the moment of inertia matrix and h is the
The product of attitude determination, the attitude
angular momentum stored by any rotating objects that
estimate or solution, is attained by using sensors to
are part of the spacecraft, such as momentum wheels or
relate information about external references, such as the
gyroscopes. Note that in this form, it is clear that the
stars, the Sun, the Earth, or other celestial bodies, to the
magnitude of angular momentum in a system can only
orientation of the spacecraft. Frequently, any single
be changed by applying external torques, T, because the
sensor has a noise level or other drawback that prevents
change due to the term ro x H can only change the
it from providing a fully satisfactory attitude solution at
direction of H, not the magnitude. So, by the product
all times. Therefore, more than one sensor is often
rule of calculus the Euler equations can be rewritten as
required to meet all mission requirements for a given
a matrix equation:
mission.
I!r o!+h!=T-r
o +Ir o ×H The combination of information from multiple sensors
is a complex field of study. The possibilities for any
or, after moving some terms around: given mission range from simple logical combination of
sensors, depending on mode, to modern information
Ir !!r
o!=T-h-I o -r
o ×H filtering methods, such as Kalman filtering. Many
methods require some projection of an expected attitude
The form of Equation 4 allows us to understand how from current conditions. Because all spacecraft sensors
attitude can change due to a variety of causes. The first must use the spacecraft’s reference frame as a basis for
term on the right-hand side represents external torques’ attitude determination, the development of angular
direct contribution to attitude dynamics; this term momentum according to the spacecraft’s frame of
includes how some actuators can be used to control reference can be important for some attitude
spacecraft attitude by creating external torques. The determination algorithms. This is the reason for the
second term gives the relationship between changes in spacecraft-referenced aspect of the Euler equations.
onboard rotating objects’ speeds and changes in the
Attitude control is the combination of the prediction of affecting total angular momentum. Because
and reaction to a vehicle’s rotational dynamics. environmental disturbances create external torques on
Because spacecraft exist in an environment of small and the spacecraft, they also create angular momentum that
often highly predictable disturbances, they may in must be either stored or removed by the attitude system.
certain cases be passively controlled. That is, a Small external torques that vary over the course of an
spacecraft may be designed in such a way that the orbit but have a mean of zero may be managed just
environmental disturbances cause the spacecraft through storage, but those torques that have a non-zero
attitude to stabilize in the orientation needed to meet mean (secular torques) will cause a gradual increase in
mission goals. Alternately, a spacecraft may include angular momentum, and this momentum build-up must
actuators that can be used to actively control the eventually be removed with actuators that create
spacecraft orientation. These two general types of external torques. Thrusters, magnetic torquers, or even
attitude control are not mutually exclusive. A spacecraft solar tabs can be used to create controlled external
may be mostly or usually passively controlled and yet torques on the spacecraft, thus controlling the total
include actuators to adjust the attitude in small ways or angular momentum.
to make attitude maneuvers (i.e. slews) to meet other
objectives, such as targets of opportunity or Attitude system design is an iterative process. Table 19-
communication needs. 1 lists typical steps in a design process and what inputs
and outputs would be expected for each step. Figure
So, external torques change the total angular 19-2 presents all the processes involved in attitude
momentum, and internal torques exchange momentum systems. The FireSat spacecraft, shown in Figure 19-3,
between different rotating parts of the spacecraft. In this and the Supplemental Communications System (SCS)
way reaction wheels or control moment gyroscopes constellation of spacecraft, shown in Figure 19-4, will
may be used to change spacecraft pointing without be used to illustrate this process.

Table 19-1: Steps in attitude system design. An iterative process is used for designing the ADCS.
Step Inputs Outputs
1 a) Define control modes Mission requirements, mission List of different control modes
1b) Define or derive system-level profile, type of insertion for launch during mission.
requirements by control mode vehicle Requirements and constraints.
2) Quantify disturbance environment Spacecraft geometry, orbit, Values for torques from external
solar/magnetic models, mission and internal sources
profile
3) Select type of spacecraft control by Payload, thermal & power needs Method for stabilization & control:
attitude control mode Orbit, pointing direction three-axis, spinning, gravity
Disturbance environment gradient, etc.
Accuracy requirements
4) Select and size ADCS hardware Spacecraft geometry and mass Sensor suite: Earth, Sun, inertial,
properties, required accuracy, orbit or other sensing devices.
geometry, mission lifetime, space Control actuators: reaction wheels,
environment, orbit geometry, thrusters, magnetic torquers, etc.
pointing direction, slew rates. Data processing avionics, if any, or
processing requirements for other
subsystems or ground computer.
5) Define determination and control Performance considerations Algorithms and parameters for
algorithms (stabilization method(s), attitude each determination and control
knowledge & control accuracy, mode, and logic for changing from
slew rates) balanced against one mode to another.
system-level limitations (power
and thermal needs, lifetime, jitter
sensitivity)
6) Iterate and document All of above Refined mission and subsystem
requirements.
More detailed ADCS design.
Subsystem and component
specifications.
Figure 19-2. Diagram of a Complete Attitude Determination and Control System. Definitive attitude
determination usually occurs in ground processing of telemetry, whereas onboard, real-time determination design
focuses on being extremely reliable and deterministic in its operation.

Figure 19-3. Hypothetical FireSat Spacecraft. We use this simplified example of a low-Earth orbiting satellite to
discuss key concepts throughout the section.

{Insert new figure of SCS here.}

Figure 19-4. Hypothetical Supplemental Communications System (SCS) Constellation. We will also use this
collection of three spacecraft in medium Earth orbit to illustrate attitude system design practices.
19.1.1 Control Modes and Requirements better understanding of the actual needs of the mission
The first step of the attitude system design process is often results from having these modes of controlling the
the definition of guiding requirements based on mission spacecraft well-defined. This iteration takes place in a
goals. Since mission goals often require more than one trade space where a single set of ADCS hardware must
mode of operating a spacecraft, the guiding be used in different ways to meet different sets of
requirements generally begin with a description of the requirements. ADCS will also be dependent on certain
control modes the ADCS is expected to execute to meet other subsystems, such as the power and structural
those goals. Tables 19-2 and 19-3 describe typical subsystems; attitude needs will also impose
spacecraft control modes and requirements. requirements on other subsystems, such as propulsion,
thermal control, and structural stability. Figure 19-5
The final form of ADCS requirements and control shows many of the complex interactions needed to
modes will be the result of iteration; control modes are bring the ADCS design in line with the needs of the
designed to achieve certain sets of requirements, and whole mission.

Table 19-2: Typical attitude control modes. Performance requirements are frequently tailored to these different
control operation modes.
Mode Description
Orbit Insertion Period during and after boost while spacecraft is brought to final orbit. Options include no
spacecraft control, simple spin stabilization of solid rocket motor, and full spacecraft control
using liquid propulsion system. May drive certain aspects of ADCS design.
Acquisition Initial determination of attitude and stabilization of vehicle for communication with ground
and power regeneration. Also may be used to recover from power upsets or emergencies.
Normal Mission, Used for the vast majority of the mission. Requirements for this mode should drive system
On-Station design.
Slew Reorienting the vehicle when required.
Contingency or Used in emergencies if regular mode fails or is disabled. Will generally use less power or
Safe fewer components to meet minimal power and thermal needs.
Special Requirements may be different for special targets or time periods, such as when the satellite
passes through a celestial body’s shadow, or umbra.

Table 19-3: Typical attitude determination and control performance requirements. Requirements need to be
specified for each mode. The following lists the performance criteria frequently specified.
Criterion Definition* Examples/Comments
Accuracy Knowledge of and control over a vehicle’s 0.25 deg, 36, often includes determination errors
attitude relative to a target attitude as defined along with control errors, or there may be separate
relative to an absolute reference requirements for determination & control
Range Range of angular motion over which Any attitude within 30 deg of nadir. Whenever
determination & control performance must rotational rates are less than 2 deg/sec.
be met
Jitter Specified bound on high-frequency angular 0.1 deg over 60 sec, 1 deg/s, 1 to 20 Hz; prevents
motion excessive blurring of sensor data
Drift Limit on slow, low-frequency angular 0.01 deg over 20 min, 0.05 deg max; used when
motion vehicle may drift off target with infrequent command
inputs
Transient Allowed settling time or max attitude 10% max overshoot, decaying to <0.1 deg in 1 min;
Response overshoot when acquiring new targets or may also limit excursions from a set path between
recover from upsets targets
Figure 19-5: The Impact of Mission Requirements and Other Subsystems on the ADCS. Direction of arrows
shows requirements flow from one subsystem to another.

For many spacecraft the ADCS must control vehicle effective. For example, the relatively strong magnetic
attitude during the firing of large liquid or solid rocket fields that occur in low Earth orbit (LEO) can create
motors for orbit insertion or management. Large motors disturbance torques that need to be managed, but they
can create large disturbance torques, which can drive also allow the use of magnetic torquers, a means of
the design to larger actuators than may be needed for attitude control not available at much higher altitudes
the rest of the mission. like geosynchronous orbit (GEO). Here, we will focus
on the torque disturbance environment as the primary
Once the spacecraft is on station, the payload pointing driver for control mode and hardware selection, but the
requirements usually dominate. These may require sensitivity of the ADCS designer must be to more than
planet-relative or inertial attitudes and fixed or spinning just the external torque disturbances of the operational
fields of view. There is usually also a need for attitude orbit. For example, some attitude sensors, such as star
slew maneuvers, and the frequency and speed of those cameras that use charge-coupled devices (CCDs) for
maneuvers must be defined. Reasons for slews can imaging, can be highly sensitive to the intense radiation
include: in the Van Allen belts of the Earth’s magnetosphere;
-Acquiring the desired spacecraft attitude depending on the specific model, the star camera may
initially or after a failure underperform or even provide no information at all
-Repointing the payload’s sensing systems to when the spacecraft occupies these regions.
targets of opportunity or for calibration purposes
-Tracking stationary or moving targets, Only three or four sources of torque matter for the
including communication stations typical Earth-orbiting spacecraft: gravity-gradient
-Directing the vehicle’s strongest motor(s) to effects, magnetic field torques on a magnetized vehicle
the proper direction relative to orbital motion. (as most spacecraft will be), impingement by solar-
radiation, and aerodynamic torques for LEO satellites.
Figure 19-6 summarizes the relative effects of these
19.1.2 Quantify the Disturbance Environment. disturbances for different flight regimes. Chapter 7
The environment in which the spacecraft will operate describes the Earth environment in detail, and Hughes
constrains what types of control methods will be [2004] provides a thorough treatment of disturbances.
Figure 19-6. Effects of major environmental disturbances on spacecraft attitude system design. The diagram
has a roughly logarithmic scale of altitude. The columns represent the four major disturbance sources, with the
intensity of color for each column indicating the strength of that disturbance in the various flight regimes.

Centroids. Some detailed description of the use of along the center of mass, no torques are created. This is
geometrical averaging is useful here, in part because why freely rotating bodies rotate about their centers of
use of computational methods in the estimation of mass.
environmental torques is increasingly common. Anyone
with a technical education will be familiar with the As a practical example, the point that may be regarded
centroid of an area, but it may have been some time as the location of a body for purposes of gravitational
since the reader encountered this concept. The centroid forces is called the center of gravity (cg); i.e. all effects
is the point in an area through which any line drawn in of gravity on the body can be considered to act at the
any direction will evenly divide moments about the line cg. In the essentially uniform gravity that we humans
(or any point along the line). To express it another way, occupy, the center of mass is usually indistinguishable
the sum of all area elements multiplied by their from the center of gravity, but in the free-fall of a space
distances from a line will be zero for any line passing orbit, the absence of direct gravitational forces and
through the centroid. In a sense, it is the average point torques means that the change, or gradient, of gravity
for the area. If a source of pressure were applied evenly over the extent of a body can be important. For
over the area, the solar pressure force could be elongated or flattened objects in orbit, the cm may be
represented as being applied entirely at the centroid for offset from the cg, so that the gravitational force is
the purposes of determining moments, and therefore effectively applied with an offset from the cm, creating
disturbance torques. A solid body can also have a torque—this is the gravity gradient torque. Note that
centroids. The center of mass (cm) is the point (usually the cg is a function of the current attitude of the
inside) the body through which any plane will divide spacecraft, not just its mass configuration, which is
the mass moment evenly. By applying a force at or critical in attitude analysis.
Other environmental effects can be understood in terms Now imagine a spacecraft like FireSat in sunlight.
of offsets between centroids of different effects on a Some parts of the spacecraft stick out further from the
body. When the aerodynamic force centroid, which is at center of mass than others. Some surfaces are more
the centroid of the ram area (the area presented to the reflective than others; solar arrays would absorb more
velocity direction), is not aligned with the cm, a torque light than reflective metallic surfaces would. Also,
is created. Solar radiation pressure is more intense on surfaces that are angled with respect to the Sun would
certain surfaces (reflective) than others (absorptive). have less pressure on them than similar surfaces
The total pressure force over the Sun-pointing surface directly facing the Sun. All this goes to demonstrate
of a spacecraft can be considered to act through a center that accurately predicting SRP torques is very tricky.
of pressure (cp) with an average reflectance, and the That said, a good starting estimate can be gleaned by
offset of that point from the cm results in solar radiation assuming a uniform reflectance and using the following
pressure torque. The location of this cp is a function of equation:
attitude as well as surface properties. Some modern
surfaces can have their reflectance change with a (D
Ts = As ( 1 + q) (cps — cm) coscp
change in applied voltage, usually for thermal reasons,
but which results in a controlled change in cp location.
So, in detailed modeling of spacecraft, the where T s is the SRP torque, (D is the solar constant
determination of the weighted averages of various adjusted for actual distance from the Sun (average
forces is important to a good understanding of the value: 1367 W/m 2 ), c is the speed of light (3 x 10 8 m/s),
torque environment. A s is the sunlit surface area in m2, q is the unitless
reflectance factor (ranging from 0 for perfect absorption
Other external disturbances to the spacecraft are either to 1 for perfect reflection), cp is the angle of incidence
small relative to the four main external disturbances, of the Sun, and cp s and cm are the centers of solar
such as infrared emission pressure, or they are limited radiation pressure and mass.
in time, such as outgassing. Occasionally, what is
normally negligible can become surprisingly large, Atmospheric Drag. In much the same way photons
even exceeding the usual disturbance torque sources, striking a spacecraft can exert pressure, so too can the
but this is one of the reasons for maintenance of healthy rarified atmosphere that clings to Earth (and certain
engineering margins and operational plans that are other planets) at the edge of space. The atmospheric
adaptable to unforeseen events. density is roughly an exponentially decaying function
of altitude, so that generally only spacecraft in low
Modeling Major Disturbances. Now we will present Earth orbit (LEO) encounter enough particles to cause
the equations used to model major disturbances with noticeable disturbances. Those that do experience a
some explanation and demonstration of they can be pressure force known as atmospheric (or aerodynamic)
used to design attitude systems. After the explanations, drag. The atmospheric drag force itself is an important
Table 19-4 will show disturbance calculations for the consideration for orbit planning (Chapter 9) and orbit
FireSat and SCS examples. prediction and tracking (Section 19.2). When the center
of atmospheric pressure, determined by the spacecraft
Solar Radiation Pressure. Sunlight has momentum, and area exposed to the atmosphere in the direction of the
therefore it exerts pressure on those objects it strikes. If orbital velocity (i.e. ram direction), is not aligned with
an object absorbs all the sunlight falling on it, then it the center of mass, a torque results. The atmospheric (or
absorbs all of its momentum and experiences a certain aerodynamic) torque can be estimated as
pressure force because of it. If the sunlight is instead
reflected exactly back along its path, such as by a 2
mirror, the pressure force felt is twice as much. Ta = p Cd Ar V (cpa — cm)
2
If a sunlit flat plate were mirrored on one half and
where Ta is the atmospheric drag torque, p is the
painted black on the other, the pressure distribution
atmospheric density in kg/m3, C d is the drag coefficient
across the plate would be uneven and a torque would
(usually between 2.0 and 2.5 for spacecraft), A r is the
result. Alternately, if the plate were all black, but a
ram area in m2, V is the spacecraft’s orbital velocity in
weight were attached to one end in the plate’s shadow,
m/s, and cp a and cm are the centers of aerodynamic
a torque would also result because the center of
pressure and mass in m. Atmospheric density and
pressure would be in the center of the plate, but the
orbital velocity as functions of altitude are tabulated in
center of mass would be closer to the weighted end.
the Appendices of this text.
These phenomena are called solar radiation pressure
(SRP) torques.
Magnetic Field. The Earth’s liquid core is a dynamo polar orbit will see roughly twice the maximum
that generates a magnetic field powerful to have magnetic torque of an equatorial orbit.
important effects on the space surrounding the planet.
Most spacecraft have some level of residual magnetic Gravity Gradient. As described in the earlier subsection
moment, meaning they have a weak magnetic field of on centroids, gravity gradient torques are caused when
their own. These residual moments can range anywhere a spacecraft’s center of gravity is not aligned with its
from 0.1-20 A • m2 , or even more depending on the center of mass with respect to the local vertical.
spacecraft’s size and whether any onboard Without getting into the math of the matter, the center
compensation is provided. of gravity of a spacecraft in orbit is dependent on its
attitude relative to Earth (or whatever body the
When a spacecraft’s residual moment is not aligned spacecraft is orbiting), and that cg is not, in general, the
with a local magnetic field, it experiences a magnetic same as the center of mass. However, when one of the
torque that attempts to align the magnet to the local spacecraft’s principal axes, as determined by the second
field, much like a compass needle. The Earth’s moment of inertia, I, is aligned with the local vertical,
magnetic field is complex, asymmetric, not aligned the cg is always on that principal axis, and therefore
with the Earth’s spin axis, and varies with both there is no gravity gradient torque. The gravity gradient
geographical movement of the dipole and changes in torque increases with the angle between the local
solar particle flux. However, for use in the ADCS vertical and the spacecraft’s principal axes, always
design process, it is usually sufficient to model the trying to align the minimum principal axis with the
Earth’s magnetic field as a dipole and to determine the local vertical.
maximum possible value of the magnetic torque for a
spacecraft’s altitude. The following equation yields this A simplified expression for the gravity gradient torque
maximum torque: for a spacecraft with the minimum principal axis in its
Z direction is

Tm DB D
3X 3
(R Tg Iz —Iy I sin(24)
R

where Tm is the magnetic torque, D is the spacecraft’s

Where Tg is the gravity gradient torque about the X
residual dipole moment in A • m 2 , and B is the magnetic
principal axis, 3 is the Earth’s gravitational constant
field strength in tesla. The magnetic field strength in
(3.986 x 10 14 m 3 /s 2 ) , R is the distance from the center
turn is calculated from M, the magnetic moment of the
Earth multiplied by the magnetic constant (M = 7.8 x of the Earth in m, 4 is the angle between the local
vertical and the Z principal axis, and I y and I z are the
1015 tesla• m3 ); R, the distance between the spacecraft
moments of inertia about Y and Z in kg • m2 .
and the Earth’s center in m, and X, which is a unitless
function of the magnetic latitude that ranges from 1 at
the magnetic equator to 2 at the magnetic poles. So, a

Table 19-4. Disturbance Torque Summary and Sample Calculations. See text for detailed discussion and
definition of symbols. FireSat is mainly affected by magnetic and aerodynamics torques. SCS satellites are mainly
affected by solar radiation pressure torques.

Disturbance Type FireSat SCS

Solar Cyclic for FireSat is small and Earth-pointing, so SCS is also small and Earth-oriented
radiation Earth- the surface area will be fairly small but (though not Earth-pointing), and has
oriented; the center of pressure may be asymmetric arrays like FireSat. It will
constant for considerably offset with respect to the use more power, hence larger arrays,
Sun-oriented Sun. and may also need to reflect more
sunlight to stay cool.

A s = 2 m x 1.5 m = 3 m2 ; q = 0.6 A s = 2.5 m x 2.0 m = 3 m 2 ; q = 0.7

cp = 0 deg; cp s – cm = 0.3 m cp = 0 deg; cp s – cm = 0.3 m

T s =0.5(1367)(2x1.5)(1+0.6)(0.3)/(3x10 8 ) T s = (1367)(2.5x2.0)(1+0.7)(0.3)/(3x10 8 )
= 3.3x10-6 N• m = 1.2x10-5 N• m
Atmospheric Constant for Similar assumptions as for SRP, except Similar to SRP. The ram face will be
 drag Earth- that being Earth-pointing, the same face the same, but we may have less control
oriented; will be presented to the ram direction all over mass and area placement because
variable for the time, so we can expect more control of the need to fit three satellites
inertially over the cp location. together in the launch vehicle.
oriented
Ar = 3 m2; cp a – cm = 0.2 m; C d = 2.0
For 700 km orbit: Ar = 5 m2; cpa – cm = 0.3 m; C d = 2.0
P = 10-13 kg/m3; V = 7504 m/s For 21,000 km orbit:
P = 10-18 kg/m3; V = 3816 m/s
T a= (0.5)(10 -13 )(2.0)(3)(7504) 2 (0.2)
= 1.7x10-5 N• m T a=(0.5)(10 -18 )(2.0)(5)(3816) 2 (0.3)
= 2.2x10-11 N• m
Magnetic Cyclic Polar orbit; assume 1 A • m2 for a small Equatorial orbit; assume 1 A • m2 for a
field uncompensated vehicle. small uncompensated vehicle.

R = (6,378 + 700) km = 7,078 km R = (6,378 + 21,000) km = 27,378 km

D = 1 A• m2; X = 2 for polar orbit D = 1 A • m2 ; X = 1.2 for equatorial orbit
B = (7.8x10 15 )(2)/(7.078x10 6 ) 3
= 4.4x10-5 N• m T m= (1)(7.8x10 15 )(1.2)/(2.7378x10 7 ) 3
T m = (1)B = 4.4x10-5 N• m = 4.6x10-7 N• m
Gravity Constant for Solar arrays dominate moment of Balanced arrays, but larger for more
gradient Earth- inertia, so Ix = I z > I y . Fairly symmetric power, so the moments of inertia are
oriented; and small, the moment of inertia can be more than FireSat’s.. The ability to
cyclic for balanced very well: We’ll set 0 =1 deg. balance mass may be limited by the
inertially need to fit 3 satellites on the same
oriented. R = 7,078 km launch vehicle, so we’ll assume a
I z = 90 kg• m 2 ; Iy = 60 kg • m2 greater difference between the
geometric and principal axes: 0 = 10
Normal mode: 0 = 1 deg deg.
Tg = (3)(3.986x10 14 )|90-60|sin(2 deg)
(2)(7.078x10 6 ) 3 R = 27,378 km
= 1.8x10-6 N• m Iz = 120 kg • m2 ; I y = 70 kg • m 2

Target-of-opportunity: 0 = 30 deg Tg = (3)(3.986x10 14 )|120-70|sin(20 deg)

Tg = 4.4x10-5 N• m (2)(2.7378x10 7 ) 3
= 5.0x10-7 N• m

Remaining significant disturbances on the control Likewise, momentum wheel friction torques can be
system are internal to the spacecraft. Fortunately, we compensated in either a closed-loop or a compensatory
have some control over them. If we find that one is fashion; some reaction wheels are designed with
much larger than the rest, we can specify tighter values friction compensation included in some commanding
for that item. This change would reduce its significance modes. Liquid slosh and operating machinery torques
but most likely add to its cost or weight. Table 19-5 are of greater concern but depend on specific hardware.
summarizes the common internal disturbances. If a spacecraft component has fluid tanks or rotating
Misalignments in the center of gravity and in thrusters machinery, the system designer should investigate
will show up during thrusting only and are corrected in disturbance effects and ways to compensate for the
a closed-loop control system and through on-orbit disturbance, if required. Standard techniques include
calibration of the thrusters. propellant management devices (e.g. slosh baffles) or
counter-rotating elements.
Table 19-5 Principal internal disturbance torques. Spacecraft designers can minimize internal disturbances
through careful planning and precise manufacturing, which may increase cost.
Disturbances Effect on Vehicle Typical Values
Uncertainty in Center of Unbalanced torques during firing of 1-3 cm
Gravity (cg) couples thrusters
Unwanted torques during translation
thrusting
Thruster Misalignment Same as cg uncertainty 0.1-0.5 deg
Mismatch of Thruster Similar to cg uncertainty +/- 5%
Outputs
Reaction Wheel Friction Resistance that opposes control torque Roughly proportional to wheel speed,
and Electromotive Force effort. These torques are the limiting depending on model. At top speed, 100%
(i.e. back EMF) mechanism for wheels speed. of control torque (i.e. saturation)
Rotating Machinery Torques that perturb both stability and Dependent on spacecraft design; may be
(pumps, filter wheels) accuracy compensated by counter-rotating
elements
Liquid Slosh Torques due to liquid dynamic pressure Dependent on specific design; may be
on tank walls, as well as changes in cg mitigated by bladders or baffles
location.
Dynamics of Flexible Oscillatory resonance at Depends on spacecraft structure; flexible
Bodies bending/twisting frequencies, limiting frequencies within the control bandwidth
control bandwidth must be phase-stabilized, which may be
undesirable.
Thermal Shocks (“snap”) Attitude disturbances when Depends on spacecraft structure. Long
on Flexible Appendages entering/leaving umbra inertia booms and large solar arrays can
cause large disturbances.

19.1.3 Selection of Spacecraft Control Methods the nadir vector caused by other environmental
disturbances. For example, long deployed booms are
Now that we understand the requirements on the control particularly susceptible to thermal shocks when
system and the environment in which it will operate, we entering or leaving umbra. These spacecraft also need a
can select one or more methods of controlling the method of ensuring attitude capture with the correct end
spacecraft. Multiple methods may be indicated when pointed at nadir—the gravity-gradient torques stabilize
different modes of operating the spacecraft have either end of the minimum inertia axis equally.
significantly different requirements or result in different
disturbance profiles (as we will see in our FireSat In the simplest gravity-gradient spacecraft, only two
example). Table 19-6 lists several methods of control, orientation axes are controlled; the orientation around
along with typical characteristics of each. the nadir vector is unconstrained. To control this third
degree of freedom, a small constant-speed momentum
Passive Control Techniques. Gravity-gradient control wheel is sometimes added along the intended pitch axis.
uses the inertial properties of a vehicle to keep it The momentum-biased wheel will be most stable when
pointed toward the Earth. This relies on the fact that an perpendicular to the nadir and velocity vectors, and
elongated object in a gravity field tends to align its therefore parallel to the orbital momentum vector. The
longitudinal axis through the Earth’s center. The stable state of the gravity-gradient plus momentum bias
torques that cause this alignment decrease with the cube wheel establishes the desired attitude through small
of the orbit radius and are symmetric around the nadir energy dissipations onboard without the need for active
vector. Thus, the yaw of a spacecraft around the nadir control.
vector is not controllable by this method. This
technique is used on simple spacecraft in near-Earth A third type of purely passive control uses permanent
orbits without yaw orientation requirements, often with magnets onboard to force alignment along the Earth’s
deployed booms to achieve the desired inertias. magnetic field. This is most effective in near-equatorial
orbits where the North-South field orientation is
Frequently, we add dampers to gravity-gradient reasonably constant for an Earth-referenced satellite.
satellites to reduce libration—small oscillations off of
Spin Control Techniques. Spin stabilization is a less useful for payloads that must be repointed
passive control technique in which the entire spacecraft frequently.
rotates so that its angular momentum vector remains
approximately fixed in inertial space. Spin-stabilized A spinner requires extra fuel to reorient because of the
spacecraft (or spinners), employ the gyroscopic gyroscopic stiffness, which also helps it resist
stability discussed earlier to passively resist disturbance disturbances. In reorienting a spinning body with
torques about two axes. Additionally, spinners are angular momentum, h, a constant torque, T, will
generally designed to be either insensitive to produce an angular velocity, w, perpendicular to the
disturbances around the third axis (the spin axis) or else applied torque and angular momentum vector, of
have active means of correcting these disturbances. magnitude w =T/h. (This follows from the earlier-
introduced Euler equations.) Thus, the greater the
The vehicle is stable (in its minimum energy state) if it stored momentum, the more torque must be applied for
is spinning about the principal axis with the largest a given w. For a maneuver through an angle 0, the
moment of inertia. Energy dissipation mechanisms torque-time product—an indication of fuel required for
onboard, such as propellant slosh, structural damping, the maneuver—is a constant equal to h •0. Alternately,
or electrical harness movement, will cause any vehicle for a vehicle with no initial angular momentum, a small
to progress toward this state if uncontrolled. So, disk- torque can be applied to start it rotating, with an
shaped vehicles are passively stable whereas pencil- opposite small torque to stop it when it has reached its
shaped vehicles are not. Spinners can be simple, survive new target. The fuel used for any angle maneuver can
for long periods without attention, provide a thermally be arbitrarily small if a slow maneuver is acceptable.
benign environment for components (because of even (Note that the spinner can only be maneuvered
heating), and provide a scanning or sweeping motion relatively slowly; a fast slew is usually not an option.)
for sensors. The principal disadvantage of spin
stabilization are that the vehicle mass properties must
be controlled to ensure the desired spin direction and
stability, and that the angular momentum vector that
provides stability also limits maneuverability. More
fuel is required to reorient the vehicle than a vehicle
with no net angular momentum, making this technique

Table 19-6. Attitude control methods and their capabilities. As requirements become tighter, more complex
control systems become necessary.

Type Pointing Options Attitude Typical Accuracy Lifetime Limits

Maneuverability
Gravity-Gradient Earth local vertical Very limited ±5 deg (2 axes) None
only
Gravity-Gradient Earth local vertical Very limited ±5 deg (3 axes) Life of wheel bearings
+ Momentum only
Bias
Passive Magnetic North/South only Very limited ±5 deg (2 axes) None
Rate-Damping + Usually Sun Generally used as ± 5-15 deg (2 axes) None
Target Vector (power) or Earth robust safe mode.
Acquisition (communication)
Pure Spin Inertially fixed Repoint with ± 0.1 deg to ± 1 deg Thruster propellant (if
Stabilization any direction precession in 2 axes applies)*
maneuvers; very (proportional to
slow with torquers, spin rate)
faster with thrusters
Dual-Spin Limited only by Same as above Same as above for Thruster propellant (if
Stabilization articulation on spun section. applies)*
despun platform Despun dictated Despun section bearings
by payload
reference and
pointing
 Bias Momentum Local vertical Fast maneuvers ± 0.1 deg to ± 1 deg Propellant (if applies)*
(1 wheel) pointing or inertial possible around Life of sensor and wheel
targets momentum vector bearings
Repoint of
momentum vector as
with spin stabilized
Active Magnetic Any, but may drift Slow (several orbits ±1 deg to ± 5 deg Life of sensors
with Filtering over short periods to slew); faster at (depends on
lower altitudes sensors)
Zero Momentum No constraints No constraints ± 0.1 deg to 5 deg Propellant
(thruster only) High rates possible
Zero momentum No constraints No constraints ± 0.0001 deg to ± 1 Propellant (if applies)*
(3 wheels) deg (determined Life of sensors and wheel
by sensors and bearing
processor)
Zero Momentum No constraints No constraints ± 0.001 deg to ± 1
Propellant (if applies)*
(CMG) Short CMG life High rates possible deg Life sensors and CMG
may require high bearings
redundancy
• Thrusters may be used for slewing and momentum dumping at all altitudes, but propellant usage may be high.
Magnetic torquers may be used from LEO to GEO.

A useful variation on spin control is called dual-spin body when it is spinning about a principal axis of
stabilization, in which the spacecraft has two sections inertia that is not aligned with a body reference axis or
spinning at different rates about the same axis; this kind axis of symmetry—for example, the intended spin axis.
of spinner is also known as a gyrostat. Normally one Coning looks like motion of the intended spin axis
section, the rotor, spins rapidly to provide gyroscopic around the angular momentum vector at the spin rate.
stiffness, while the second section, the stator or Figure 19-7 shows various natural rotations.
platform, is despun to keep one axis pointed toward the
H
Earth or Sun. By combining inertially fixed and rotating 114i'P

sections on the same vehicle, dual spinners can Z, H. •+.P

A.

accommodate a variety of payloads in a simple vehicle. ^f^r^ ^ rwr ^^ 3 `+► f^^ ^ yLLil1^:^1^
Also, by adding energy dissipation devices to the
platform, a dual spinner can be passively stable
spinning about the axis with the smallest moment of r
y '-^c`ur.i -w
e
inertia, as long as the rotor is spinning about its own
Figure 19-7. Types of Rotational Motion. H = angular
maximum moment of inertia. This permits more pencil-
momentum vector; P = principal axis; w =
shaped spacecraft, which fit better in launch vehicle
instantaneous rotation axis; Z = geometrical axis.
fairings and which would not normally be stable
spinning about their long axes. The disadvantage of
Nutation is the torque-free motion of a simple rigid
dual-spin stabilization is the added complexity of the
body when the angular momentum vector is not
platform bearing and slip rings between the sections.
perfectly aligned with a principal axis of inertia. For
(Slip rings permit power and electrical signals to flow
rod-shaped objects, this motion is a slow rotation
between the two sections.) This complexity can
(compared to spin rate) of the spin axis around the
increase cost and reduce reliability compared to simple
spin stabilization. angular momentum vector. For these objects spinning
about a minimum inertia axis, additional energy
dissipation will cause increased nutation. For disk-
Spinning spacecraft, both simple and dual, exhibit
shaped objects, spinning around a maximum inertia
several distinct types of motion that are often confused.
Precession is the motion of the angular momentum axis, nutation appears as a tumbling rotation faster than
spin rate. Energy dissipation for these objects reduces
vector caused by external torques, including thruster
nutation, resulting eventually in a clean spin. For these
firings used to correct environmental disturbances.
reasons, minimum-axis (or minor-axis) spinners are
Coning (or wobbling) is the apparent motion of the
often concerned with minimizing energy dissipation,
whereas maximum-axis (or major-axis) may actually or aerodynamic control surfaces (e.g. tabs), or magnetic
include mechanisms, such as a passive nutation damper, torquers. Broadly, however, these systems take two
to dissipate energy quickly. forms: one uses momentum bias by placing a
momentum wheel along the pitch axis; the other is
Nutation is caused by disturbances such as thruster called zero momentum and does not use momentum
impulses, and can be seen as varying signals in body- bias at all—any momentum bias effects are generally
mounted inertial and external sensors. Wobble is caused regarded as disturbances. Either option usually needs
by imbalances and appears as constant offsets in body- some method of angular momentum management, such
mounted sensors. Such constant offsets are rarely as thrusters or magnetic torquers, in addition to the
discernable unless multiple sensors are available. primary attitude actuators.

Spin stability normally requires active control, such as In a zero-momentum system, actuators such as reaction
cold-gas thrusters or magnetic torquers, to periodically wheels or thrusters respond to disturbances on the
adjust the spacecraft’s angular momentum vector to vehicle. For example, an attitude error in the vehicle
counteract disturbance torques. In addition, we may results in a control signal that torques the wheel,
need to damp the nutation caused by disturbances or creating a reaction torque in the vehicle that corrects the
precession commands. Aggravating this nutation is the error. The torque on the wheel either speeds it up or
effect of energy dissipating phenomena like structural slows it down; the aggregate effect is that all
flexure and flexible harness or fluid motion, which are disturbance torques are absorbed over time by the
present in any spacecraft to some degree. Once the reaction wheels, sometimes requiring the collected
excitation stops, nutation decreases as these same angular momentum to be removed. This momentum
factors dissipate the kinetic energy added by the control removal—called desaturation, momentum dumping, or
effort. However, this natural damping can take hours. momentum management—can be accomplished by
We can neutralize this error source in minutes with thrusters or magnetic torquers acting automatically or
nutation dampers (see Sec 19.1.5). We can also reduce by command from the ground.
the amount of nutation from these sources by increasing
the spin rate and thus the stiffness of the spinning When high torque is required for large vehicles or fast
vehicle. If the spin rate is 20 rpm and the nutation angle slews, a variation of three-axis control using control
from a given disturbance is 3 deg, then nutation from moment gyros, or CMGs, is available. These devices
the same disturbance would be reduced to 1 deg if the work like momentum wheels on gimbals. The control
spin rate were 60 rpm. We seldom use spin rates above of CMGs is complex and their lifespan is limited, but
90 rpm because of the large centripetal forces their available torque for a given weight and power can
demanded of the structure and the consequent effects on make them attractive.
design and weight. In thrusting and pointing
applications, spin rates under 20 rpm are generally not A very specialized form of zero momentum control,
used as they may allow excessive nutation. However, here called active magnetic control, can be attained
applications unrelated to attitude control, such as from a combination of a magnetometer, a Global
thermal control, are frequently insensitive to nutation Positioning System (GPS) antenna, and
and may employ very low spin rates. computationally intensive software filtering. The GPS
feeds the spacecraft location to the onboard processor,
Three-axis Control Techniques. Spacecraft stabilized which then determines the local magnetic field based on
in all three axes are much more common today than onboard models. The magnetometer data is filtered
those using spin or gravity gradient stabilization. They using the Euler equations to determine the attitude, and
can maneuver relatively easily and can be more stable magnetic torquers make corrections in the two available
and accurate, depending on their sensors and actuators, directions at any given moment—corrections about the
than more passive stabilization techniques. They are, magnetic field vector are not possible. Active magnetic
however, more expensive; they are also often more control can be an inexpensive backup control mode for
complex, but processor and reliability improvements a LEO satellite, or it can be a primary control mode for
have allowed comparable or better total reliability as a satellite in a highly inclined orbit. (The highly
some more passive systems. For critical space inclined orbit has large changes in magnetic field
applications, there is no replacement for thorough risk direction, allowing the filtering algorithm to better
and reliability assessment (see Chapter 24). determine a three-axis attitude solution.) This attitude
knowledge can also be combined with other sensors,
The control torques about the axes of three-axis systems such as Sun sensors, for more accuracy. While this is
come from combinations of momentum wheels, not a common control method, we include it here as an
reaction wheels, control moment gyros, thrusters, solar example of how increased onboard computational
power and the presence of new resources, such as the from the spacecraft, and it can be exploited to control
GPS constellation, can allow completely new methods roll and yaw over a quarter orbit using only a roll (or
of attitude determination and control. only a yaw) sensor, instead of needing one sensor for
each of the roll and yaw axes.
As a final demonstration of zero momentum three-axis Effects of Requirements on Control Type. With the
control, simple all-thruster systems are used for short above knowledge of control techniques, we can proceed
durations when high torque is needed, such as during to select a control type that will best meet mission
orbit insertion maneuvers or other orbit adjustments requirements in the expected operational environment.
(delta-V) from large motors. These thrusters then may Tables 19-7 and 19-8 describe the effects of orbit
be used for different purposes such as momentum insertion and payload slew requirements on the
dumping or provision of small delta-Vs during other selection process. It is also useful here to once again
mission phases. reference Figure 19-6 for information on how altitude
can affect the space environment; certain control types,
Momentum bias systems often have just one wheel with such as gravity-gradient stabilization or active magnetic
its spin axis mounted along the pitch axis, ideally damping, are better in some orbits than in others.
normal to the orbit plane. The wheel is run at a nearly
constant, high speed to provide gyroscopic stiffness to A common control approach during orbit insertion is to
the vehicle, just as in spin stabilization, with similar use the short-term spin stability of the combination of
nutation dynamics. Around the pitch axis, however, the spacecraft and orbit-insertion motor. Once on station
spacecraft can control attitude by torquing the wheel, the motor may be jettisoned, the spacecraft despun
slightly increasing or decreasing its speed. Periodically, using thrusters or a yo-yo device, and a different control
the momentum in the pitch wheel must be managed (i.e. technique used from that point on.
brought back to its nominal speed), as in zero-
momentum systems, using thrusters, magnets, or other Payload pointing will influence the attitude control
means. method, the class of sensors, and the number and kind
of actuation devices. Occasionally, pointing accuracies
The dynamics of nadir-oriented momentum-bias are so stringent that a separate, articulated platform is
vehicles exhibit a phenomenon known as roll-yaw necessary. An articulated platform can perform
coupling. To understand this coupling, consider an scanning operations much easier than the host vehicle
inertially fixed angular momentum vector at some error and with better accuracy and stability. Trade studies on
angle with respect to the orbit plane. If the angle is pointing requirements must consider accuracy in
initially a positive roll error, then a quarter-orbit later it attitude determination and control, and the most
appears as a negative yaw error with no roll component stringent requirements will ultimately drive ADCS
remaining. As the vehicle continues around the orbit, component selection. Table 19-9 summarizes the
the angle goes through negative roll and positive yaw effects of accuracy requirements on the ADCS
before regaining its positive roll character. This approach for the spacecraft. Section 14.5 discusses how
coupling (or commutation) is due to the apparent to develop pointing budgets.
motion of the Earth-fixed coordinate frame as seen

Table 19-7. Orbit Transition Maneuvers and Their Effects. Using thrusters to change orbits creates special
challenges for the ADCS.
Requirement Effect on Spacecraft Effect on ADCS
Large impulse to complete orbit Solid motor or large bipropellant Inertial measurement unit for
insertion (thousands of m/s) stage. accurate reference and velocity
Large thrusters or a gimbaled engine measurement.
or spin stabilization for attitude Different actuators, sensors, and
control during burns. control laws for burn vs. coasting
phases.
Need for navigation or guidance.
On-orbit plane changes to meet Large thrusters needed, but these Separate control law for thrusting.
payload needs or vehicle operations thrusters may be needed for other Actuators sized for thrusting
(hundreds of m/s) reasons also, such as orbit insertion, disturbances (possibly two sizes of
coasting phase, or stationkeeping. thruster).
Onboard attitude reference for
thrusting phase.
 Orbit maintenance/trim maneuvers One set of thrusters Thrusting control law.
(<100 m/s) Onboard attitude reference.

Table 19-8. Slewing Requirements That Affect Control Actuator Selection. Spacecraft slew agility can demand
larger actuators for intermittent use.
Slewing Effect on Spacecraft Effect on ADCS
None or Time- Spacecraft constrained to one - Reaction wheels, if planned, can be
Unconstrained attitude (highly improbable), or smaller
reorientations can take many hours. - If magnetic torquers can dump
momentum, reaction control thrusters may
not be needed
Low Rates Minimal - Depending on spacecraft size, reaction
From 0.05 deg/s wheels may be fully capable for slews
(orbital rate) to 0.5 deg/s - If reaction wheels not capable, thrusters
will be necessary
- Thrusters may be needed for other
reasons; i.e. stationkeeping
High Rates - Structural impact on appendages - Control moment gyros or thrusters needed.
>0.5 deg/s - Weight and cost increase If thrusters needed for other reasons, two
thrust levels may be needed.
Table 19-9. Effects of Control Accuracy Requirements on Sensor Selection and ADCS Design. More accurate
pointing requires better and more expensive sensors and actuators.
Required Effect on Spacecraft Effect on ADCS
Accuracy
(36)
>5 deg • Permits major cost savings Without attitude determination
• Permits gravity-gradient (GG) • No sensors required for GG stabilization
stabilization • Boom motor, GG damper, and a bias momentum wheel are
only required actuators
With attitude determination
• Sun sensors & magnetometer adequate for attitude
determination at >_ 2 deg
• Higher accuracies may require star trackers or horizon
sensors
1 deg to • GG not feasible • Sun sensors and horizon sensors may be adequate for
5 deg • Spin stabilization feasible if stiff, sensors, especially a spinner
inertially fixed attitude is • Accuracy for 3-axis stabilization can be met with RCS
acceptable deadband control but reaction wheels will save propellant
• Payload needs may require despun for long missions
platform on spinner • Thrusters and damper adequate for spinner actuators
• 3-axis stabilization will work • Magnetic torquers (and magnetometer) useful
0.1 deg to • 3-axis and momentum-bias • Need for accurate attitude reference leads to star tracker or
1 deg stabilization feasible horizon sensors & possibly gyros
• Dual-spin stabilization also feasible • Reaction wheels typical with thrusters for momentum
unloading and coarse control
• Magnetic torquers feasible on light vehicles (magnetometer
also required)
< 0.1 deg • 3-axis stabilization is necessary • Same as above for 0.1 deg to 1 deg but needs star sensor
• May require articulated & and better class of gyros
vibration-isolated payload platform • Control laws and computational needs are more complex
with separate sensors • Flexible body performance very important

FireSat Control Selection. For FireSat, we consider two mission orbit. This common option simplifies the
options for orbit insertion control. First, the launch spacecraft design since no special insertion mode is
vehicle may directly inject the spacecraft into its needed. An alternate approach, useful for small
spacecraft such as FireSat, is to use a monopropellant of the available magnetic field. However, the
propulsion system onboard the spacecraft to fly itself up propulsion system has to be included anyway, and
from a low parking orbit to its final altitude. For small constantly running torquers would be an additional
insertion motors, reaction wheel torque or momentum power drain. Since power is a major challenge for this
bias stabilization may be sufficient to control the mission, we will use the thrusters for momentum
vehicle during this burn. For larger motors, delta-V unloading.
thruster modulation or dedicated attitude control
thrusters become attractive.
19.1.4 Selection and Sizing of ADCS Hardware
Once on station the spacecraft must point its sensors at
nadir most of the time and slightly off-nadir for brief We are now ready to evaluate and select the individual
periods. Since the payload needs to be despun and the ADCS components. For all ADCS hardware, we will
spacecraft frequently reoriented, spin stabilization is not determine the minimum performance level needed to
the best choice. Gravity-gradient and passive magnetic meet requirements. Then, standard components
control cannot meet the 0.1-deg pointing requirement or available from manufacturers will be selected if
the 30 deg slews. This leaves three-axis control and possible, sometimes resulting in better performance
momentum-bias stabilization as viable options for the than the minimum required. If standard components are
on-station control. not available, specialized components may be designed
and built, but this can often be prohibitively costly for
Depending on other factors, either approach might most agencies, and so a revision of the requirements is
work, so we will baseline momentum bias control with more likely to be more in line with available hardware.
its simpler hardware requirements. In this case we will
use a single pitch wheel for momentum and
electromagnetic torquers for momentum dumping and Actuators. Options for actuator selection are
roll and yaw control. summarized in Table 19-10. First, we will discuss
momentum-exchange devices, which conserve angular
For the optional off-nadir pointing requirement, three- momentum in the spacecraft: reaction wheels,
axis control with reaction wheels might be more momentum wheels, and control moment gyros. Then,
appropriate. Also, three-axis control often can be we will move on to external torque actuators, which
exploited to simplify the solar array design by using change the angular momentum of the spacecraft when
one of the unconstrained payload axes (yaw, in this they are activated: magnetic torquers & thrusters (cold-
case) to replace a solar array drive axis. Thus, the gas, hot-gas and electric) are the most commonly used
reduced array size possible with 2 degrees of freedom in this category.
can be achieved with one array axis drive and one
spacecraft rotation. Wheel control provides smooth changes in torque,
allowing very accurate pointing of spacecraft. Some
SCS Control Selection. For the Supplemental wheels can cause vibrations, or jitter, at high speeds,
Communication System, we will focus on taking but this can often be mitigated with dampers or changes
advantage of the gentle disturbance environment and 1 in structural design. Reaction wheels are essentially
deg accuracy requirement to design a light, inexpensive torque motors with high-inertia rotors. They can spin in
attitude system that can be installed in all 3 satellites. either direction and provide one axis of control for each
At the SCS altitude of 21,000 km, with the wheel. Momentum wheels are reaction wheels with a
configuration we’ve assumed, there are no good passive nominal spin rate above zero to provide a nearly
stabilization methods available to us. Reaction wheels constant angular momentum. This momentum provides
will be needed to reject the disturbances, which are gyroscopic stiffness to two axes, and the motor torque
dominated by the solar radiation pressure torques (see may be controlled to change pointing around the spin
Table 19-4). So, as long as the reaction wheels are there axis. In sizing wheels we must always consider two
anyway, we might as well use them for three-axis performance quantities: angular momentum capacity,
stabilization. and torque authority.

While the greater part of the SRP torques will be cyclic, To determine the necessary momentum capacity, we
some small part will be secular. Therefore, the must distinguish between cyclic and secular
momentum stored in the reaction wheels will gradually disturbances in the spacecraft’s environment. We
increase and will need to be removed periodically by typically size reaction wheels to be able to store the full
thrusters. The use of 100 A•m2 magnetic torquers for cyclic component of momentum without the need for
momentum removal is feasible from the point of view frequent momentum dumping. Therefore, the average
disturbance torque for 1/4 or '/2 an orbit determines the may be determined by requirements on payload
minimum capacity of the wheels. The secular observation continuity, or it may be the amount of time
component of momentum will also need to be stored for the spacecraft must survive without ground
the amount of time the spacecraft must be operational intervention.
without a momentum dump being performed. This time

Table 19-10. Typical Attitude Actuators. Actuator weight and power usually scale with performance.

Actuator Typical Performance Range Mass (kg) Power

(kg)
Thrusters Thrusters produce force; multiply by moment arm for torque Mass and power vary
Hot Gas (Hydrazine) 0.5 to 9000 N (See Ch. 18 for details on
Cold Gas <5N propulsion systems)
Reaction and Maximum torques: 0.01 to 1 N • m Varies with speed:
2 to 20
Momentum Wheels Practical momentum storage capacity: 0.4 to 3000 N. m• s 10 to 100
Control Moment
Max. torques: 25 to 500 N • m > 10 90 to 150
Gyros (CMG)
Magnetic Torquers 1 to 4,000 A • m2 (See Table 19-4 for torque examples) 0.4 to 50 0.6 to 16
The necessary torque authority of the reaction wheels is actuating). By turning the gimbal axis, we can obtain a
defined either by slew requirements or the need for high-output torque whose size depends on the speed of
control authority to be well above the peak disturbance the rotor and the gimbal rate of rotation. Control
torque. If the peak disturbance is too close to the torque systems with two or more CMGs can produce large
authority available, pointing accuracy may suffer. torques about all three of the spacecraft’s orthogonal
axes, so we most often use them where agile (i.e. high
For three-axis control, at least three wheels with non- angular rate) maneuver are required. Their use requires
coplanar spin axes are required. Often, a fourth wheel is complex control laws and careful momentum
carried in case one of the primary wheels fails. If the management to avoid wheel saturation. Also, because
wheels are not orthogonal (and a redundant wheel never the CMG torque is created by twisting what is
is), additional torque authority or momentum capacity essentially a stiff gyroscope perpendicular to its spin
may be necessary to compensate for non-optimal axis, the bearings of the wheel suffer a great deal of
geometry. However, the redundancy of the fourth wheel wear and tear, causing most CMGs to have shorter
can have additional benefits, such as being able to avoid lifetimes than other actuators. Because CMGs combine
the wheel speed passing through zero (which can cause a short life with high cost, weight, and power needs,
attitude error transients) or even as power storage, as they are generally used only on very large spacecraft
driving a spinning wheel toward zero speed will and only when absolutely necessary to achieve the
provide power to the spacecraft. mission goals.

For spin-stabilized or momentum-bias systems, the Spacecraft also use magnetic torquers as actuation
cyclic torques will cause cyclic variations of the devices. These torquers are magnetic coils
attitude, whereas the secular torques cause gradual (electromagnets) designed to generate magnetic dipole
divergence of the attitude away from the ideal target. moments of commanded magnitude. When three
We typically design the stored angular momentum, orthogonal torquers are mounted to a spacecraft, they
determined by spin rate and inertia of the spinning can create a magnetic field of any direction and
body, to be large enough to keep the cyclic motion magnitude up to the strength of the torquers. Magnetic
within accuracy requirements without active control. torquers can compensate for a spacecraft’s residual
Periodic torquing will still be needed to counteract magnetic field or attitude drift from minor disturbance
secular disturbances, and the cost of this torquing in torques. They also can be used to desaturate
propellant, if performed by thrusters, will constrain the momentum-exchange systems, though they usually
maximum stored momentum. require much more time than thrusters. A magnetic
torquer produces a torque proportional and
For high-torque applications in which fine control is perpendicular to the Earth’s varying magnetic field;
still needed, control moment gyros may be used instead therefore, at any given moment, no torque can be
of reaction wheels. CMGs are single- or double- provided about the Earth’s magnetic field vector.
gimbaled wheels spinning at constant speed (and However, as a spacecraft changes in latitude or altitude
therefore providing momentum bias stiffness when not
while following its orbit, different directions of the field large, instantaneous control torques at any time in the
become available. orbit. The disadvantages of thrusters are that they use
expendable propellant, can disrupt orbit determination
Electromagnets have the advantage of no moving parts, activities, and that the plumes of expelled matter can
requiring only a magnetometer for sensing and wire impinge on the surface of the spacecraft, possibly
coiled around a metallic rod in each axis. (Some heating or contaminating surfaces.
torquers are even internally redundant, as they have two
coils around the same metallic core.) Because they use Attitude functions to which thrusters can be applied
the Earth’s natural magnetic field, and this field reduces include controlling attitude, controlling nutation and
in strength with the cube of distance from the Earth’s spin rate, performing large, rapid slews, and managing
center, magnetic torquers are less effective at higher angular momentum. In all of these functions, large
orbits. We can specify the torquer’s field strength in torque authority can be helpful, but also necessary to
A•m2 and tailor it to any application. Table 19-11 some extent is the ability to change angular momentum
describes sizing rules of thumb for wheels and magnetic by relatively small amounts. So, another consideration
torquers, and works through those rules for the FireSat in selection of thrusters is how fine and how precise the
and SCS examples. SCS has no slewing requirements, delivered torque impulse needs to be. Torque applied
so we size its reaction wheels based only on momentum over a period time creates a change in angular
storage needs. momentum, which has a corresponding change in the
rotational rate of the spacecraft. As attitude changes
Thrusters (i.e. rocket engines) are possibly the most with non-zero rate, attitude control accuracy is directly
frequently flown attitude actuator because of their dual determined by the minimum thruster impulse available.
use in adjusting orbital parameters. Almost every If a 50 N thruster with a 1 meter moment arm must
spacecraft that needs to perform orbital maneuvers will provide a change in angular momentum accurate to
use thrusters to achieve that goal, and in many cases, within 1 N•m•s, then it will be necessary to fire the
some subset of the thrusters used is for attitude control. thruster for a period of time no greater than 20 msec.
Thrusters produce a force on the spacecraft by expelling The thruster valves must be capable of opening and
material, called propellant, at high velocity from their closing that rapidly and precisely, and of propellant
exit nozzles. Hot-gas propulsion systems include actually flowing into the thruster and out the nozzle,
thrusters that chemically alter the propellant to extract which takes a finite amount of time. This need for
the energy needed for rapid mass expulsion. These speed and precision in engine valve control also has
systems may be monopropellant, in which the implications in the design of the avionics that drive the
propellant is catalyzed to break down chemically, or thruster valves; often, special electronics cards must be
bipropellant, in which a fuel is mixed with an oxidizer developed for a given mission to meet thruster-based
to achieve combustion just prior to expulsion. Cold-gas attitude control requirements.
systems include thrusters whose propellant is not
altered chemically during propulsion. In a cold-gas The baseline FireSat spacecraft will use magnetic
system, the energy may come from phase change of the torquers for momentum dumping, so thruster sizing is
propellant, or simply from pre-pressurizing the unnecessary. However, the SCS satellites will need to
propellant in its tank. A third type of thruster is use thrusters for orbit maintenance and momentum
electrical, which, due to the usually small forces dumping. We will want to minimize the number and
involved, is only used for attitude control in special size of thrusters, since the spacecraft are so small.
circumstances. Electrical propulsion is accomplished by However, the lower limit will most likely come from
using magnetic or electrostatic fields to eject plasma or the orbital maneuvering needs, and not the attitude
magnetic fluid to achieve a reaction force on the control needs. We will suppose 4 1-lb thrusters are
spacecraft. needed for orbital maneuvering and steering, and in
Table 19-12, we will present and work through
Thrusters provide torque proportional to their moment procedures and simplified equations for sizing thrusters
arm, which is the amount that their line of force is using SCS as an example. SCS has no stated slew
offset from the vehicle’s center of mass. When thrusters requirement, but for the purposes of the example, a
apply their force along a line that intersects the center slew requirement of 30 deg in 60 seconds will be
of mass of the vehicle, there is no torque. So, while the supposed. A thorough discussion on the topic of
amount of force available from a thruster may be large, propulsion systems, including estimating propellant
the torque available is limited by the physical extent of needs, can be found in Chapter 18.
the vehicle and how the thrusters are mounted on the
vehicle. When mounted to maximize torque authority,
thrusters have the advantage of being able provide
TABLE 19-11. Simplified Equations for Sizing Reaction Wheels, Momentum Wheels, and Magnetic
Torquers. FireSat momentum wheels are sized for the baseline requirements. Reaction wheels are sized for the
optional design with 30-deg slew requirements. SCS reaction wheels are sized for momentum storage capacity.

Parameter Simplified Equations Application to FireSat Example

Torque from Control torque, whether from reaction For the FireSat spacecraft, if the two biggest disturbances
Reaction Wheel wheels or magnetic torquers, must are aligned (this is probably too pessimistic):
for equal worst-case anticipated TD = Tg + Tm = 2 • (4.4 x 10–5 N•m)
Disturbance disturbance torque plus some margin: = 8.8 x 10 -5 N• m
Rejection T C = (TD)(Margin Factor) (Table 19.4). Even this worst-case is manageable by most
reaction wheels and magnetic torquers. Momentum
requirements or slew torque, not disturbance rejection,
will be used for actuator sizing.
Slew Torque for For max-acceleration slews For the 30-deg slews of the 90 kg•m 2 spacecraft
Reaction (1/2 distance in 1/2 time): (Fig. 19-2) in 10 min, this becomes:
Wheels O /2 = 1/2 T/I(t/2)2 T = (4)(30 deg)( n /180 deg)(90 kg• m2 )/(600 sec) 2
T = 4 O I/t 2
= 5.2 x 10-4 N• m (This is also a small value.)
Momentum One approach to estimating wheel For TD = 4.4 x 10 –5 N•m (Table 19-4) and a
Storage in momentum, h, is to integrate the 700-km orbital period of 5926 sec,
Reaction Wheel worst-case disturbance torque, TD, h = (4.4 x 10-5 N• m)(5926 sec)(0.707/4)
over a full orbit, with period P. For = 0.046 N• m• s
gravity gradient, the maximum A small reaction wheel with storage of 0.4 N•m•s would
momentum accumulates in 1/4 of an be sufficient, providing a margin of >9. A good check on
orbit. A simplified expression for such this margin is to add up all four disturbance torques as
a sinusoidal disturbance is: though they were all aligned, which would be TD = 10 -4
h = T D• P • (0.707)/4 N • m, for momentum buildup of = 0.1 N• m• s. We still have
where 0.707 is the rms average of a plenty of margin.
sinusoidal function.
For SCS, T D=1.2x10-5 & P =45083 sec.
SCS storage: h=(1.2x10- 5 )(45083)(0.707/4)=0.1N • m• s
We will size 1.0 N • m• s wheels to minimize thruster burns.
Momentum For momentum-wheel stabilization, The value of h for a 0.1 deg (0.0017 rad) yaw accuracy
Storage in roll and yaw accuracy depend on the would be
Momentum wheel’s momentum and the external h = (4.4x10-5 N• m/0.0017 rad)• (5926/4 sec)
Wheel disturbance torque. A simplified = 37.7 N• m• s
expression for the required momentum For a 1 deg accuracy, we would need only 3.8 N•m•s.
storage is: h = (T/O a) • (P/4) The same results would hold for a spinner.
where O a = allowable motion.
Torque from Magnetic torquers use electrical Table 19-4 estimates the maximum Earth field, B, to be
Magnetic current through the torquer to create a 4.4 x 10 –5 tesla. We calculate the torque rod’s magnetic
Torquers magnetic dipole (D). The Earth’s torquing ability (dipole) to counteract the worst-case
magnetic field, B, acts on D, resulting gravity gradient disturbance, TD , of 4.4 x 10 –5 N•m•s as
in torque (T) on the vehicle: D = (4.4x10-5 N• m)/(4.4x10-5 tesla) = 1 A• m 2
D = T/B which is a small actuator. The Earth’s field is cyclic at
Magnets used for momentum dumping twice orbital frequency; thus, maximum torque is
must equal the peak disturbance + available only twice per orbit. A torquer of 3 to 10 A•m2
margin to compensate for the lack of capability should provide sufficient margin.
complete directional control.
Note: For actuator sizing, the magnitude and direction of the disturbance torques must be considered. In
particular, momentum accumulation in inertial coordinates must be mapped to body-fixed actuator axes.

TABLE 19-12. Simplified Equations for Preliminary Sizing of Thruster Systems. SCS thruster requirements are
small for this low-disturbance, minimal slew application. It is likely that the thrusters needed for orbit maintenance
can also serve for momentum dumping.

Simplified Equations Application to SCS Example

Thruster force level sizing for For the worst case T D of 1.2 x 10 –5 N•m (Table 19-4) and a thruster moment
external disturbances: arm of 0.5 m
F=T D /L F=(1.2x10- 5 N • m)/(0.5m)=6x10- 6 N
This small value indicates orbit maintenance and momentum dumping
F is thruster force, T D is worst-case
requirements, not disturbance torques, will determine thruster size. Also,
disturbance torque, and L is the
using thrusters to fight cyclic disturbances uses precious propellant; it is
thruster’s moment arm
generally better to store the momentum.

Sizing force level to meet slew rates Assume a 30-deg slew in less than 1 min (60 sec), accelerating for 5% of
(optional zero momentum system): that time, coasting for 90%, and decelerating for 5%.
Determine fastest slew rate, w, w = 30 deg / 60 sec = 0.5 deg/sec
required in the mission profile.
To reach 0.5 deg/sec in 5% of 1 min, which is 3 sec, requires an acceleration
Develop a slew profile that accelerates
2
a = w /t = (0.5 deg/sec)/(3 sec) = 0.167 deg/sec = 0.003 rad/sec 2
the vehicle quickly, coasts at that rate,
and then decelerates quickly. The
acceleration required, a, comes from F = Ia /L = (120 kg • m)(0.003
2
rad/sec 2)/(0.5 m) = 0.72 N
equating these two torque definitions: This is small but feasible.
T = F• L = I •a

Sizing force level for momentum For SCS with 1.0 N•m•s wheels and 1-sec burns,
dumping:
F = (1.0 N • m• s)/(0.5 m * 1 sec)
F = h/(Lt) = 2.0 N
where
h = stored wheel momentum This is still well within the range of 1-lb thrusters, which are commonly used
L = thruster moment arm for orbit maintenance on small spacecraft. Reaction wheels with even larger
t = burn time capacity might be desirable if it would further reduce the number of times
the thrusters must be used.

Sensors. We complete this hardware unit by selecting can be used as part of the normal attitude determination
the sensors needed for attitude determination. Consult system, part of the initial acquisition or failure recovery
Table 19-13 for a summary of typical devices as well as system, or part of an independent solar array orientation
their performance and physical characteristics. Note, system. Since most low-Earth orbits include eclipse
however, that sensor technology is changing rapidly, periods, Sun-sensor-based attitude determination
promising ever more accurate and lighter-weight systems must provide some way of tolerating the
sensors for future missions. regular loss of this data without violating pointing
constraints.
Sun sensors are visible-light or infrared detectors that
measure one or two angles between their mounting base Sun sensors can be quite accurate (<0.01 deg), but it is
and incident sunlight. They are popular, accurate, and not always possible to take advantage of that feature.
very reliable, but they require clear fields of view. They We usually mount Sun sensors near the ends of vehicles

to obtain an unobstructed field view, so the Sun sensor respect to the spin axis of the vehicle, and they issue a

accuracy can be limited by structural bending on large pulse correlated to the time the Sun crosses the sensor
spacecraft. Spinning satellites use specially designed to provide spin-phase information. Also popular are

Sun sensors that measure the angle of the Sun with coarse Sun sensors, which are simply small solar cells
Table 19-13. Typical ADCS Sensors. Sensors have co tinued to improve in performance while getting smaller and
sometimes less expensive.

Sensor Typical Performance Range Mass (kg) Power (W)

Gyroscopes Drift Rate = 0.003 deg/hr to 1 deg/hr < 0.1 to 15 < 1 to 200
Drift rate stability varies widely
Sun Sensors Accuracy = 0.005 deg to 3 deg 0.1 to 2 0 to 3
Star Sensors Accuracy = 1 arcsecond to 1 arcminute 2 to 5 5 to 20
(Scanners & Cameras) = 0.0003 deg to 0.01 deg
Horizon Sensors Accuracy:
- Scanner/Pipper 0.05 deg to 1 deg (0.1 deg is best for LEO) 1 to 4 5 to 10
- Fixed Head (Static) < 0.1 deg to 0.25 deg 0.5 to 3.5 0.3 to 5
Magnetometer Accuracy = 0.5 deg to 3 deg 0.3 to 1.2 <1

that issue a current roughly proportional to the cosine of Moon, planets, or even high radiation levels, such as in
the Sun angle. These sensors are so small and the Van Allen belts, which is a disadvantage that must
inexpensive that it is often feasible to put several in be accommodated in their application. Where the
many directions on a spacecraft, and then to estimate mission requires the highest accuracy and justifies a
the Sun direction by solving the linear system equations high cost, we often use a combination of star trackers
that results. Because coarse Sun sensors use no power and gyroscopes. The combination of these sensors is
and almost never fail, they are often used in low-power very effective: the gyros can be used for initial
acquisition and fault recovery modes. stabilization and during periods of inference in the star
trackers, while the trackers can be used to provide a
Star sensors have improved rapidly in the past few high-accuracy external reference unavailable to the
years and represent the most common sensor for high- gyros.
accuracy missions. Star sensors can be scanners or
trackers. Scanners are used on spinning spacecraft. Horizon sensors (also known as Earth sensors) are
Light from different stars passes through multiple slits infrared devices that detect the contrast between the
in the scanner’s field of view. After several star cold of deep space and the heat of the Earth’s
crossings, we can derive the vehicle’s attitude. We use atmosphere (about 40 km above the surface in the
star trackers on three-axis stabilized spacecraft to track sensed band). Simple narrow field-of-view fixed-head
one or more stars to derive two- or three-axis attitude types (called pippers or horizon crossing indicators) are
information. The majority of star trackers used today used on spinning spacecraft to measure Earth phase and
work much like digital cameras (and many of these are chord angles, which, together with orbit and mounting
increasingly called star cameras, rather than trackers), geometry, define two angles to the Earth (nadir) vector.
allowing starlight to fall on a CCD to create an image Scanning horizon sensors use a rotating mirror or lens
of the star field. Then, internal processing determines a to replace (or augment) the spinning spacecraft body.
three-axis attitude based on a star catalog. Many units They are often used in pairs for improved performance
are able to determine a very accurate attitude within and redundancy. Some nadir-pointing spacecraft use
seconds of being turned on. staring sensors, which view the entire Earth disk (from
GEO) or a portion of the limb (from LEO). The sensor
While star sensors excel in accuracy, care is required in fields of view stay fixed with respect to the spacecraft.
their specification and use. The most accurate star This type works best for circular orbits, as they are
cameras are unable to determine attitude at all if the often tuned for a tight range of altitudes.
spacecraft is rotating too fast, and other star sensors
must know roughly where they are pointing to make Horizon sensors provide Earth-relative information
their data useful. Therefore, the vehicle must be directly for Earth-pointing spacecraft, which may
stabilized to some extent before the trackers can operate simplify onboard processing. The scanning types
effectively. This stabilization may require alternate require clear fields of view for their scan cones
sensors, which can increase total system cost. Also, star (typically 45, 60, or 90 deg half-angle). Typical
sensors are susceptible to being blinded by the Sun, accuracies for systems using horizon sensors are 0.1 to
0.25 deg, with some applications approaching 0.03 deg. steadily improve accuracy while reducing size and
For the highest accuracy in low-Earth orbit, it is mass.
necessary to correct the data for Earth oblateness and
seasonal horizon variations. Error models for gyroscopes vary with the technology,
but characterize the deterioration of attitude knowledge
Magnetometers are simple, reliable, lightweight sensors with time. Some examples of model parameters are
that measure both the direction and size of the Earth’s drift bias, which is simply an additional, false rate the
magnetic field. When compared to models of the sensor effectively adds to all rate measurements, and
Earth’s field, their output helps us establish the drift bias stability, which is a measure of how quickly
spacecraft’s attitude, but their accuracy is not as good the drift bias changes. When used with an accurate
as that of star or horizon sensors. The Earth’s magnetic external reference, such as a star tracker, gyros can
field can shift with time and is not known precisely in provide smoothing (filling in the gaps between tracker
the first place. To improve accuracy, we often combine measurements) and higher frequency information (tens
their data with data from Sun or horizon sensors. When to hundreds of hertz), while the tracker provides lower
a vehicle using magnetic torquers passes through frequency, absolutely referenced information whenever
magnetic-field reversals during each orbit, we use a its field of view is clear. Individual gyros provide one
magnetometer to control the polarity of the torquer or two axes of information and are often grouped
output. In earlier spacecraft the torquers usually needed together as an inertial reference unit (IRU) for three
to be turned off while the magnetometer was sampled full axes and, sometimes, full redundancy. IRUs with
to avoid corrupting the measurement. However, accelerometers added for position and velocity sensing
improvements in onboard computing capability mean are called inertial measurement units (IMU).
that coupling matrices can be used to extract the torquer
inputs from the field measurement, allowing constant Sensor Selection. Sensor selection is most directly
sampling even while torquing. influenced by the required orientation of the spacecraft
(e.g. Earth-, Sun- or inertial-pointing) and its accuracy.
GPS receivers are well known as high-accuracy Other influences include redundancy, fault tolerance,
navigation devices, but they can also be used for field of view requirements, and available data rates.
attitude determination. If a spacecraft is large enough to Typically, we identify candidate sensor suites and
place multiple antennas with sufficient separation, conduct a trade study to determine the most cost-
attitude can be determined by employing the effective approach that meets the needs of the mission.
differential signals from the separate antennas. Such In such studies the existence of off-the-shelf
sensors offer the promise of low cost and weight for components and software can strongly affect the
LEO mission, and are being used in low accuracy outcome. In this section we will only briefly describe
applications or as back-up sensors. Development some selection guidelines.
continues to improve their accuracy, which is limited
by the separation of the antennas, the ability to resolve Full three-axis knowledge requires at least two external,
small phase differences, the relatively long wavelength, non-parallel vector measurements, although we use
and multipath effects due to reflections off spacecraft IRUs or spacecraft angular momentum (in spinners or
components. momentum-biased systems) to hold or propagate the
attitude between external measurements. In some cases,
Gyroscopes are inertial sensors that measure the speed if attitude knowledge can be held for a fraction of an
or angle of rotation from an initial reference, but orbit, the external vectors (e.g. Earth or magnetic) will
without any knowledge of an external, absolute have moved enough to provide the necessary
reference. We use gyros in spacecraft for precision information. In three-axis star trackers, each identified
attitude determination when combined with external star acts as a reference vector, which allows a single
references such as star or Sun sensors, or, for brief piece of hardware to generate a full three-axis attitude
periods, for nutation damping or attitude control during solution.
thruster firing. Manufacturers use a variety of physical
phenomena, from simple spinning rotors to ring lasers, For Earth-pointing spacecraft, horizon sensors provide
hemispherical resonating surfaces, and laser fiber optic a direct measurement of pitch and roll axes, but require
bundles. Gyros based on spinning rotors are called augmentation for yaw measurements. Depending on the
mechanical gyros, and they may be large iron gyros accuracy required, we use Sun sensors, magnetometers,
using ball or gas bearings, or may reach very small or momentum-bias control with its roll-yaw coupling
proportions in so-called MEMS gyros. (MEMS stands for the third degree of freedom. For inertially pointing
for microelectromechanical systems.) The gyro spacecraft, star and Sun sensors provide the most direct
manufacturers, driven largely by aircraft markets, measurements, and IRUs are ideally suited. Frequently,
only one measurement is made in the ideal coordinate torquers to rotate the spacecraft so the arrays are lit.
frame (Earth or inertial), and the spacecraft orbit Then, since the attitude relative to nadir will change as
parameters are required to convert a second FireSat follows its polar orbit, we can be sure to get a
measurement or as an input to a magnetic field model. good communication signal at some point, so that we
Either the orbit parameters are uplinked to the can receive telemetry and send commands.
spacecraft from ground tracking and propagated by
onboard processing, or they are obtained from onboard Supplemental Communications System sensors. For
GPS antennas. SCS attitude determination, low-power gyros can
FireSat sensors. The external sensors for FireSat could provide rate information. Accurate gyros can be heavy
consist of any of the types identified. For the 0.1 deg and often use a lot of power; we have neither high
Earth pointing requirement, however, horizon sensors accuracy needs nor an excess of power or mass in our
are the most obvious choice since they directly measure budgets. Therefore, we will use light and inexpensive
the two axes we most need to control. The accuracy MEMS gyros. We need a minimum of 3 MEMS
requirement makes a star sensor a strong candidate as gyros—one for each axis—but by employing 4-6 gyros
well; its information would need to be transformed, we can cross-compare the gyro data and remove the
probably using an onboard orbit ephemeris calculation, larger bias errors that MEMS gyros normally have.
to Earth-relative for our use. The 0.1 deg accuracy is at
the low end of horizon sensors’ typical performance, so With rate information onboard, we only need an
we need to be careful to get the most out of their data. occasional update from an attitude sensor. If the
We assume we also need a yaw sensor capable of 0.1 magnetic field were stronger, we might be able to filter
deg, and this choice is less obvious. Often, it is useful to magnetometer data to get to 1 deg of accuracy, but it is
question a tight yaw requirement. Many payloads, e.g. doubtful at this high altitude. Star cameras are small
antennas, some cameras, and radars, are not sensitive to and very accurate, but they are expensive. One useful
rotations around their pointing axis. For this discussion, rule of thumb is: If at all possible, sense the thing you
we will assume this requirement is firm. We could use need to point at. Sun pointers should have Sun sensors
Sun sensors, but their data needs to be replaced during and Earth pointers should have Earth sensors. Because
eclipses. Magnetometers don’t have the necessary the satellites will have the same direction pointing
accuracy alone, but with our momentum-bias system, toward the Earth throughout their mission, the best
roll-yaw coupling, and some yaw data filtering, a option appears to be an Earth sensor. We would need to
magnetometer-Sun sensor system should work. select a sensor designed for high altitudes. However, we
still have no yaw data, and since the satellites must
At this point we consider the value of an inertial point accurately at each other, yaw accuracy is critical.
reference package. Such packages, although heavy and
expensive for high-accuracy equipment, provide a Because the satellites will be communicating with each
short-term attitude reference that would permit the other, it is conceivable that the communication signal
Earth vector data to be used for full three-axis strength could be used as an attitude determination data
knowledge over an orbit. A gyro package would also source for yaw control. That is, a feedback loop would
reduce the single measurement accuracy required of the close around the communication system’s own measure
horizon sensors, simplifying their selection and of its link margin; maximizing the link margin would
processing. Such packages are also useful to the control provide the attitude goal we want. For this exercise we
system if fast slews are required. Although nice to have, will not assume such an option is available. Instead, we
an inertial package does not seem necessary for FireSat. will choose the star camera after all; a simple onboard
However, a careful trade study between an inexpensive ephemeris calculation will tell the spacecraft where its
MEMS gyro package combined with just the Earth target satellite is in inertial space. There may be clever
sensors, and the need to include Sun sensors and tricks that we could use with ground-based methods to
magnetometers if the gyros are left out might indicate avoid using star cameras, such as combining orbit
the MEMS gyros could give better reliability or lower tracking and attitude data. However, the complexity of
total cost. We would need to do this kind of detailed operating three separate satellites that have to work
trade study in later iterations of the design process. together will likely prove more expensive in software
development and operating costs than just buying three
Finally, we will want a simple, coarse control mode to star cameras. As for our rule of thumb of sensing what
initially point the arrays at the Sun and to protect the you’re pointing at, we now see there can be situations
spacecraft in the event of an anomaly. By using 6 for which this rule cannot be followed. Still, it’s always
coarse Sun sensors pointing along the positive and a good place to start.
negative of each axis, the spacecraft can derive the
location of the Sun from any attitude and use magnetic


We will propose the same plan for an initial acquisition

and safe control mode as for FireSat. However, SCS
will use reaction wheels for control, and will have the
benefit of rate sensors to improve performance. This is
a good thing, since SCS satellites need more power than
FireSat, and so will probably not have as much time to
acquire the Sun (i.e. get the solar arrays lit). Table 19-
14 summarizes our FireSat and SCS hardware
selections.
Table 19-14. FireSat and SCS Spacecraft Control Component Selection.
Mission & Type Components Rationale
FireSat Actuators Momentum Wheel (1) - Pitch axis torquing
- Roll and yaw axis passive stability
Magnetic Torquers (3) - Roll and yaw control
- Pitch wheel desaturation
FireSat Sensors Horizon Sensor (1) - Provide basic pitch and roll reference
- Can meet 0.1 deg accuracy
- Lower mass and cost than star sensors
Sun Sensors (6) - Initially acquire vehicle attitude from unknown
orientation
- Coarse attitude data
- Fine data for yaw
Magnetometer (1) - Coarse yaw data
SCS Actuators Reaction Wheels (3) - Three-axis stabilization
(for each satellite) - Storing momentum from solar radiation torques
Thrusters (4) - Thrusting and steering during orbit maneuvers
- Removing momentum from reaction wheels
SCS Sensors Star camera (1) - Determining absolute attitude
(for each satellite) MEMS gyros (4) - Determining rotational rates; having 4 gyros allows
large biases to be canceled out
- Propagating attitude solution when star sensor data
Sun Sensors (6) unavailable
- Initially acquire vehicle attitude relative to Sun to
get power quickly

Once the hardware selection is complete, it must be Table 19-15. FireSat Spacecraft Control Subsystem
documented for use by other system and subsystem Summary. The baseline ADCS components satisfy all
designers as follows. mission requirements, with thrusters available if
- Specify the power levels and weights required required.
for each assembly Components Type Weight Power Mounting
(kg) (W) Considerations
- Establish the electrical interface to the rest of Momentum Mid-size, < 5 total, 10 to Momentum
the spacecraft Wheel 40 N•m•s with drive 20 vector on pitch
momentum electronics axis
- Describe requirements for mounting, Electromagnets 3, 10 A •m 2 2, 5 to 10 Orthogonal
alignment, or thermal control including configuration
- Determine what telemetry data we must current best to reduce
drive cross-coupling
process electronics
- Document how much software we need to Sun Sensors 6 wide- < 1 total 0.0 Free of
angle viewing
develop or purchase to support onboard coarse Sun obstructions
calculation of attitude solutions sensors and reflections
providing
Specific numbers depend on the vendors selected. A 4 7c
typical list for FireSat might look like Table 19-15, but steradian
the numbers could vary considerably with only slight coverage;
5-10 deg
changes in subsystem accuracies or slewing curac
requirements. Horizon nning 5 total 10 Unobstructed
Sensors view of Earth’s
s (2)
rptylpue horizon
 electronics; Control algorithms are usually implemented in an
0.1 deg
accuracy onboard processor and analyzed with detailed
Optical Hydrazine; Propellant N/A Alignments simulations.
Thrusters 0.5 N force weight and moment
depends on arm to center
mission of gravity are We typically apply linear theory only to preliminary
critical analysis and design. We also maintain engineering
Magnetometer 3-axis <1 5 Need to isolate
magnetometer margin against performance targets when using linear
from theory because, as the design matures, so does our
electromagnets,
either understanding of the nonlinear effects in the system.
physically or Nonlinear effects may be inherent or intentionally
by duty-cycling introduced to improve the system’s performance.
the magnets

Another reason to maintain margin, especially in

19.1.5 Define the Determination and Control actuator sizing, is that while systems engineers always
Algorithms carefully budget mass, they cannot usually track the
Finally, we must tie all of the control components moment of inertia matrix as well. Moment of inertia is
together into a cohesive system. Generally, we begin the most important quantity to ADCS designers, and a
with a single-axis control system design (See Figure 19- given mass may have a wide range of moment of inertia
8). As we refine the design, we add or modify feedback values depending on how the mass is distributed.
loops for rate and attitude data, define gains and
constants, and enhance our representation of the system Feedback control systems are of two kinds based on the
to include three axes of motion (though we may still flow of their control signals. They are continuous-data
treat these as decoupled for early design iterations). To systems when sensor data its electrically transformed
confirm that our design will meet requirements, we directly into continuously flowing, uninterrupted
need good mathematical simulations of the entire control signals to the actuators. By contrast, sampled-
system, including sensor error models and internal and data systems have sensor sampling at set intervals, and
external disturbances. Usually, linear differential control signals are issued or updated at those intervals.
equations with constant coefficients describe the Most modern spacecraft process data through digital
dynamics of a control system well enough to allow us computers and therefore use sampled-data control
to analyze its performance with the highly developed systems.
tools of linear servomechanism theory. With these same
tools, we can easily design linear compensation to Although it is beyond the scope of this book to provide
satisfy specifications for performance. detailed design guidance on feedback control systems,
the system designer should recognize the interacting
effects of attitude control system loop gain, capability
of the attitude control system to compensate for
disturbances, accuracy of attitude control, and control
system bandwidth.

Three-axis stabilization. Different types of active

control systems have different key parameters and
algorithms. For systems in which spacecraft rates will
be kept small, three-axis control can frequently be
decoupled into three independent axes. In this simplest
form, each axis of a spacecraft attitude control system
can be represented as a double-integrator plant (i.e.
1/Is2 in the s-domain) and may be controlled by a
proportional-derivative (PD) controller, where the
control torque T C = KPe E + KD6dotE . The most basic
design parameter in each axis is its proportional gain,
KP . (KP is also often called a position gain.) This gain
represents the amount of control torque we want to
result from a unit of attitude error, and so it has units of
torque divided by angle, e.g. newton-meters per radian.
Figure 19-8. Block diagram of a Typical Attitude The proportional gain is selected by the designer and
Control System with Control along a Single Axis. must be high enough to provide the required control
accuracy in the presence of disturbances, which can be speed of response. The mathematical definition of
guaranteed by setting KP>TD/4E,max, where TD is the bandwidth requires a bit more explanation than we want
peak disturbance torque, and OE,max is the allowable to go into here, but it can be approximated as Con =
attitude error. Note that this error will remain for as (KP/I) /', where I is the spacecraft moment of inertia.
long as the disturbance torque remains; the error that The bandwidth defines the frequency of disturbances
remains in the presence of disturbances is called steady- and of commanded motions at which control authority
state error. begins to diminish. Attitude control and disturbance
rejection are effective from 0 frequency (i.e. constant or
The value of K P also largely determines the attitude d.c. inputs) up to the bandwidth. Speed of response is
control system bandwidth, which is a measure of its

Table 19-16 ADCS Vendors. Typical suppliers for ADCS components. An up-to-date version of this table can be
found at the SMAD web site.
Company Sun Earth Magnet- Star Gyro GPS Mom./ CMG Magnetic Thrusters
Sensors (horizon) ometers Sensors Reaction Torquers
Sensors Wheels
Adcole Corporation X
Aero'et X
Ball Aerospace and X
Technologies Corp.
Billingsley Aerospace &
X
Defense
Bradford Engineering X X X
Comtech AeroAstro X X
EADS Astrium X X X X
EADS SODERN X X
EMS Technologies, Inc. X
Finmeccanica (incl. SELEX
X X X X
Galileo5
ITT Aerospace X
General Dynamics X
Goodrich(incl.Ithaco5 X X X X
Honeywell Space Systems
X X
(incl. Allied5
Jena Jptronik X X
L-3 Space & Navigation X X X
Kearfott Guidance &
X X
Navigation Corp.
Meda X
Micro Aerospace
X X
Solutions
Microcosm, Inc. X
NASA Goddard Space
X X
Flight Center
Northrop Grumman (incl
X X
Litton5
Oersted - DTU X
Optical Energy
X X
Technologies
Rockwell Collins X
Deutschland(incl. Teldix5
Servo Corp. of America X
StarVision Technologies X
Surrey Satellite X X X X X X X X X
Technologies - US LLC
S stron Donner Inertial X
Terma X
Watson Industries, Inc. X X

the correcting control torque to gradually increase over

approximately the reciprocal of bandwidth. Note that time to remove that steady-state error completely. If the
proportional gain is inversely proportional to allowable disturbance torque changes magnitude or direction, the
error, and bandwidth is proportional to the square root PID controller will take some time to adjust, but it will
of the proportional gain. Therefore, high accuracy track with no steady-state error. This improved
implies high KP and high bandwidth. However, there performance generally comes at the price of reduced
are practical limits on bandwidth. If the bandwidth is stability margins, so the stability and performance
high enough that it includes bending frequencies for the requirements must always be carefully balanced.
spacecraft structure, then control-structure interactions
must be carefully considered. If the bandwidth is higher In addressing control-structure interactions, the control
than the maximum sampling rate of critical attitude designer must keep in mind that, if bending causes
sensors, some care must be taken to avoid excessive apparent attitude errors, the control system may
attitude responses to the slow updates. Finally, high aggravate the bending motion, eventually affecting
bandwidth control systems may be more likely to have control stability and perhaps damaging the bending
adverse reactions to inaccurate or lagging system structures. Though there are control design techniques
inputs; this condition is referred to as having to deal with these potential problems, the control
insufficient stability margin. designer will often start by recommending a minimum
frequency for structural bending modes that would
There are design techniques to address these concerns. allow flexible effects to be neglected. This minimum
One such technique is to modify the system response to frequency is generally an order of magnitude greater
increase stability while keeping the same attitude than the bandwidth needed to achieve required
accuracy by increasing attitude rate damping, which is accuracy. Then, structural engineers will attempt to
controlled by the derivative gain, or rate gain, KD . For honor this restriction in their design. For further
a given value of K P , increasing KD will slow down the discussion of how structural flexibility interacts with
response of the control system, but it will not diminish feedback control systems, see section 3.12 of Agrawal
the accuracy in the end, provided the disturbance [1986].
environment or slewing requirements do not require
very high speed of response. This technique is often Spin stabilization and momentum bias. The
used to improve stability margins in systems where the fundamental concept in spin stabilization is the nutation
final accuracy is more important than the amount of frequency of the vehicle. For a spinning body, the
time it takes to achieve that accuracy. Note that, though inertial nutation frequency ( C)ni) is equal to w S I S/I T
there are multiple methods for selecting K P and KD , where wS is the spin frequency, I S is the spin axis of
such as linear-quadratic regulation (LQR) and other inertia, and I T is the transverse axis inertia. For a
state-space methods, the system response generally momentum-bias vehicle with a non-rotating body and a
must still be understood in terms of bandwidth and momentum wheel (or a dual-spin vehicle with a non-
damping to achieve confidence that the control gains rotating platform), the nutation frequency is w ni =h/I T
will provide both stability and adequate performance. where h is the angular momentum of the spinning body
(or bodies). (Note that this is really the same equation
Sometimes the problem facing a designer is that as for a spinning body, for which h = wS I S.) Thus,
stability margin requirements are not difficult to meet, spacecraft with large inertias and small wheels have
but accuracy requirements are very difficult. A common small nutation frequencies (i.e. long periods).
technique for improving on the performance of a PD is
to add an integrator to the feedback loop; this is called a Attempting to control the vehicle with a bandwidth
proportional-integral-derivative controller (PID). faster than the nutation frequency causes it to act more
Recall that in a regular PD controller, a steady-state like a three-axis stabilized vehicle, that is, the stiffness
error of some kind will remain as long as there are of the spin is effectively reduced. In general, we
disturbance torques to create it. The integrator causes
attempt to control near the nutation frequency or
slower, with correspondingly small torques for given The basic algorithms for determination depend on the
attitude errors. Because we generally wish to avoid coordinate frames of interest and the geometry and
changing the spin rate with control torques, the torque parameterization of the measurements. Useful
is applied perpendicularly to the spin. We can then coordinate frames can include the individual sensor
relate the achieved angular rate, co, to the applied torque frames, the local vertical frame, and an inertial frame,
by simplifying Euler’s equations to only the gyroscopic such as Earth-centered inertial (ECI). The geometry of
torque component: T = co x h = co h where h is the the measurements is different for different sensors, as
system angular momentum. A lower limit on control discussed in Sec. 19.1.4, and they are generally
bandwidth is usually provided by the orbit rate coO, parameterized as sequential rotations about the axes of
a frame, called Euler angles (e.g. roll, pitch, yaw), or as
which for a circular orbit is coo = f / r 3 where attitude quaternions, which are unit vectors with four
µ =3.986 x 10
14m3 /s2 for Earth orbits and r is the orbit elements that define a rotational axis in a frame, called
radius. the eigenaxis, and the amount of rotation about the
eigenaxis. Inertial reference units (with their supporting
Example of Magnetic Momentum Management. software) and star sensors are well suited to
Managing the long-term buildup of angular momentum quaternions, whereas Earth-pointing vehicles often use
in the orbiting spacecraft often drives the ADCS a local-vertical set of Euler angles, much like aircraft.
requirements. As would be expected for an element so
central to ADCS design, there are many different Simple spacecraft may use the sensor output directly for
methods for keeping momentum to levels that do not control, whereas more complex vehicles or those with
degrade data or equipment. We present here a simple, higher accuracy requirements employ some form of
common, and relatively inexpensive algorithm for averaging, smoothing or Kalman filtering of the data.
commanding magnetic torquers using only The exact algorithms depend on the vehicle properties,
magnetometer data. It functions as an excellent active orbit, and sensors used.
rate damping controller, and after the spacecraft’s rate
are very low, it keeps the angular momentum at FireSat algorithms. For our momentum-bias FireSat
manageable levels by canceling it out on an orbit- example, control separates into pitch-axis control using
averaged basis. Often called B-dot, this algorithm is torque commands to the momentum wheel and roll-yaw
based on back-differencing magnetometer data to control using current commands to the magnetic
estimate the time derivative of the magnetic field (or torquers. The pitch-wheel desaturation commands must
“BdotÓ). For rate damping purposes, we assume the B- also be fed (at a slower rate) to the torquers. The pitch-
field is changing in the body due only to spacecraft wheel control is straightforward, using PD control and,
rotation, which is true for all but rotation rates on the optionally, integral control, in which commands are
order of orbital rate. Specifically, Bdot = Bk+1 – Bk = - augmented by a small torque proportional to the
co x B. The spacecraft dipole vector, D, is set equal to a integral of the attitude error. The roll-yaw control
gain multiplied by the magnetic field: D = -k Bdot = design starts by using the linearized nutation dynamics
kco x B. Then, when the magnetic field acts on that of the system, and is complicated by the directional
vector, it produces a control torque T c = D x B that is limitations of electromagnetic torque (the achievable
always in a direction that slows down the spacecraft torque is perpendicular to the instantaneous Earth
rotation except when co is aligned with B (in vector magnetic field).
terminology, co- T c < 0). Once the spacecraft is slowed
down, the algorithm will attempt to follow the magnetic The nadir-oriented control system may use an Earth-
field vector, its success determined by the strength of referenced, aircraft-like Euler angle set (i.e. roll, pitch
the gain, k, that is chosen. However, whether it follows & yaw), although quaternions may simplify
the field doesn’t matter: The goal of slowing the calculations during off-nominal pointing; quaternion
spacecraft and then maintaining a slow momentum has calculations generally require fewer arithmetic steps,
been achieved. Flatley, Morgenstern, Reth and Bauer they simplify tracking commands, and they have no
[1997] did not invent the B-dot controller, but their singularities. The horizon sensors directly read two of
paper nicely explains in English and equations how the angles of interest, pitch and roll. Yaw needs to be
something so simple works so well. measured directly from Sun position (during orbit day)
or from the magnetometer readings (using a stored
Algorithms for attitude determination. A full model of the Earth’s field plus an orbit ephemeris), or
discussion of attitude determination algorithms requires inferred from the roll-yaw coupling described earlier.
a dedicated reference such as Wertz [1978]. We will The magnetic field and Sun information require an
highlight only some of the basic concepts. uplinked set of orbit parameters, and increase the
computational requirements of the subsystem. Overall, Flatley, Thomas W., W. Morgenstern, A. Reth, and
meeting the 0.1 deg yaw requirement when the Sun is F.H. Bauer, 1997. “A B-Dot Controller for the
not visible will be the toughest challenge facing the RADARSAT Spacecraft,” NASA GSFC Flight
ADCS designer, and planning to coast through the dark Mechanics/Estimation Theory Symposium, Greenbelt,
periods without direct yaw control may be most MD.
appropriate.
Fortescue, P. and J. Stark. 2003. Spacecraft Systems
Engineering, 3 rd ed. NY: John Wiley & Sons.
SCS Algorithms. A 3-axis PD controller will be
sufficient for our control accuracy needs. In this control Griffin, M. and J.R. French. 2004. Space Vehicle
algorithm, an error angle and error rate is calculated for Design. 2nd ed. Washington, DC: AIAA.
each axis, and a control torque calculated based on
Hughes, Peter C. 2004. Spacecraft Attitude Dynamics,
those. That calculation done for the 3 axes gives a
Mineola, NY: Dover Publications, Inc.
control torque vector, which is then distributed among
the reaction wheels. If one of the wheels is asked to Kaplan, Marshall H. 1976. Modern Spacecraft
issue a control torque that exceeds its capability, then Dynamics and Control. New York: John Wiley and
the spacecraft would have a total torque vector that is Sons.
not in the same direction as the requested vector. For
our mission, that is not too much aof a danger, since we Wertz, James R. ed. 2009. Orbit & Constellation
don’t have any elements that would be damaged by Sun Design & Management. Hawthorne, CA: Microcosm
exposure or the like. Still, it is a simple matter to scale Press and New York, NY: Springer.
all 3 vector components by the same factor that would Wertz, James R. ed. 1978. Spacecraft Attitude
make the largest of them equal to its limit. Note that Determination and Control. Dordrecht, The Nether-
this is safe for a PD controller, but that stability lands: D. Reidel Publishing Company.
problems can result if the same tactic is used for a PID
controller without taking some care to prevent
problems.

The data from the star camera and the rate gyros in each
SCS satellite can be combined in any of a number of
ways. We will simply trust the star camera attitude as
accurate whenever it gives us data it indicates is valid,
and we will use the gyros to provide direct rate
measurements and to propagate the attitude solution by
integrating the rate over the sampling time if the star
camera fails to provide a valid attitude. This method is
not the most accurate, but generally, more accuracy
comes at the expense of more complexity, which then
costs money and time in flight software development
and testing. This is a good lesson to wrap up this
section: Though it can be tempting to always reach for
your best, most expensive tool, it is better engineering
to try to get by with the cheapest system that will meet
your requirements with appropriate margin.

References
Keep all previous references and also include the
following (I have other texts and papers to add to this
list, but I wanted to get this draft out. I’ll send the
reference list later.):

Agrawal, Brij N. 1986. Design of Geosynchronous

Spacecraft. Englewood Cliffs, NF: Prentice-Hall, Inc.
Chobotov, Vladimir A. 2002. Orbital Mechanics, 3rd
ed. Washington, DC: AIAA.