Technical white paper: TE327

Interferometric calibration of rotary axes

by M.A.V. Chapman, A. Holloway, W. Lee, M. May, S. McFadden, D. Wall

Introduction
This paper describes the operational principles behind Renishaw’s XR20 rotary axis calibration system
and how it can be used ‘on axis’ to calibrate axes of rotation. It examines the various error sources that
affect the accuracy of measurement and describes how the system has been designed, and should
be used, in order to minimise such errors. It concludes with a section describing how the XR20 system
can be used ‘off axis’ to calibrate 4th and 5th axes on CNC machines.

Interferometric angular measurement

Figure 1 illustrates the use of a laser
and angular interferometer to measure
small angles of rotation of a rotary
axis. As the axis rotates, the laser
system detects the relative change
between the optical path lengths in
the two “arms” of the interferometer.
As the axis rotates by angle θ, the
laser beam in Arm 1 will get shorter by
S.sin(θ) and the laser beam in Arm 2
will get longer by S.sin(θ) where S is
the separation between the two retro-
reflectors. The total relative change in
the path lengths, between Arms 1 and
2 of the interferometer, is therefore Figure 1
2S.sin(θ). This change in path lengths
is detected by an interference fringe counter/interpolator inside the laser’s detector unit. The resulting
fringe count is converted into a linear distance, ΔL, by multiplying by the laser wavelength/2.

ΔL = Fringe count × laser wavelength / 2

In angular mode the laser system software then converts ΔL into an angular measurement by
calculating arcsin(ΔL/S).

θ = arcsin(ΔL/S)

For a more detailed explanation of angular interferometry, please refer to the Renishaw White
Paper entitled ‘TE326 - Interferometric angle measurement and the hardware options available from
Renishaw’.

Rotary axis calibration

The arrangement shown in Figure 1 is only suitable for checking angular movements over a range of
about ±10° because, at larger angles, the rotation of the angular reflector will cause misalignment of
the returned laser beams and a corresponding loss of signal strength.
 However, this limitation can be
readily overcome by combining
measurements from the laser
interferometer with those from a
high accuracy rotary axis, such as
Renishaw’s XR20. A typical setup
is shown in Figure 2. An angular
reflector is mounted on top of the
XR20, which in turn is mounted
on top of the rotary axis being
calibrated. As the axis under test
is rotated from one target position
to the next, the XR20 is driven in
the opposite direction in order to
maintain alignment of the angular
interferometer. When the axis
under test stops at each target
Figure 2 position, the positioning error is
calculated by comparing the target
position with the arithmetic sum of the angular readings from the laser interferometer and the XR20.
This allows calibration of the axis over a full 360° or even over multiple revolutions.

A key benefit of using an angular interferometer to provide the ‘coupling’ between the counter-rotating
XR20 and a stationary part of the axis under test, is that it’s largely insensitive to small translation
(side to side) movements of the reflector. (This topic is covered in the Operational Principles section of
TE326). This makes system alignment
much easier by largely eliminating
a major potential source of angular
measurement error. For example,
eccentrically mounting an XR20,
1mm from the centre of rotation of the
axis under test, adds less than ±0.5
arcseconds of measurement error. For
comparison a 200mm diameter rotary
encoder disk, with external read-head,
would have to be mounted to within
0.25μm to achieve similar performance.
Even a fully enclosed rotary encoder
with integral bearings and sophisticated
precision shaft coupling has to be
mounted within about 0.05mm (a 40x
tighter tolerance than that required by
XR20). Figure 3

Nevertheless, keeping overall accuracy levels within ±1 arcseconds requires careful design and
attention to detail to ensure all of the possible error sources are similarly controlled. The following
sections give an insight into the internal design of Renishaw’s XR20 followed by an analysis of the
main error sources associated with the system and its use.

2
The XR20 – A closer look
The XR20 is a battery powered, radio
controlled, high accuracy servo controlled
rotary axis with an angular reflector
mounted on a central shaft. Figure 3
shows an external view of the XR20.

The XR20 contains the following items,

as shown in the cutaway view in Figure 4,
and the simplified cross-section in
Figure 5.
• A central shaft on which an angular
reflector, rotary encoder and direct
drive servo motor are all mounted.
• A high accuracy rotary encoder
with two diametrically opposed
read-heads and axial encoder
graduations which are directly
formed on the outer edge of the Figure 4
encoder disk to provide long term
stability. The solid aluminium encoder disk’s high thermal conductivity, low thermal capacity and
thick cross section ensure any temperature changes and resulting expansions/contractions
are evenly distributed around the disk thereby minimising local distortions. The diametrically
opposed read-heads largely eliminate errors due to eccentricity of the encoder disk relative to
the central shaft, or lateral movement due to bearing wander/play. (This is covered in more
detail later).
• A low power, direct drive, servo
controlled motor to rotate the
central shaft. The rotor magnets
are bonded directly to the central
shaft and the coils to the XR20
body. This motor provides a high
resolution, contactless, drive
of the central shaft free from
friction, backlash and play with
minimal power dissipation, thereby
maximising battery life. The
efficient, low power design also
minimises self-heating, and the
rotational symmetry helps ensure
any heat generated is evenly
distributed. Figure 5

• Adjustable spacers to allow the axis of rotation of the central shaft to be set perpendicular to the
mounting surface of the bayonet base.
• Two preloaded, self-centering spherical bearings to minimise eccentricity errors, bearing wander
and end-float.
• Sophisticated electronics including; Encoder graduation interpolation circuitry to give 0.1
arcsecond resolution; Automatic balancing/gain control of the encoder read-head signals in
order to minimise sub-divisional errors (covered in more detail later), and an integrated encoder
error map generated during system manufacture/calibration.
• Home sensor to allow system orientation and error map referencing.

3
 • Bluetooth wireless communications
with the PC providing cable free
operation, simplifying set-up and
avoiding problems with cable drag.
• Interchangeable Lithium Polymer
rechargeable battery allowing 3
hours of typical operation between
charges.
• Bayonet mount with quick-release
clamping to a separate mounting
ring to allow easy fitment and
removal to a wide variety of rotary
Figure 6
axes / tables.
• Semi-reflective target on the rear of the angular reflector to simplify beam alignment (See Figure 6).

Error analysis
The main sources of error may be broadly divided into 3 categories, namely;
• Errors due to angular misalignment of axes
• Inaccuracies in the rotary encoder reading
• Inaccuracies in the laser interferometer reading

These are discussed in sequence below.

Errors due to angular misalignment of axes

There are 3 main angular misalignments to consider, as shown in Figure 7
• Angular Misalignment (α) between the axes of rotation of the axis under test and the XR20.
This typically arises from three sources, the non-perpendicularities of the mounting surfaces
of both the axis under test and the XR20 to their respective axes of rotation, and damage or
debris on these mounting surfaces.
• Angular misalignment (β)
between the “line of sight” of
the laser interferometer and
the plane of rotation of the
axis under test.
• Angular misalignment (ω)
of the centre-line between
the two retro-reflectors in
the angular reflector and
the axis of rotation of XR20.
This centre-line should be
perpendicular to the axis of
rotation.

Some of these misalignments are

tightly controlled by Renishaw
during the XR20 manufacturing
process, others are the
Figure 7
responsibility of the user to control
during system set-up.

4
These angular misalignments give rise to a measuring error, Eθ, at angle θ, of the form;

Eθ = sin(2θ).α²/4 + sin(θ).αβ + cos(θ).αω ………………………………………………… (1)

Note: For simplicity, this paper assumes the misalignment angle α lies in the direction of θ = 0°. In
reality θ should be replaced with θ + η, where η is an arbitrary angular offset. This doesn’t affect the
form and magnitude of the induced measurement error (over 360°), but will vary its phase.

Looking at Equation 1 it can be seen that the measurement

error, Eθ, is the sum of three distinct error terms. An α²/4
based term, an αβ term and finally αω based term, where α,
β and ω are misalignment angles in radians. Each term is now
examined in detail.

α²/4 term - The origin of this term is illustrated by Figures 8,

9 and 10. As the axis under test rotates, the misalignment, α,
between the axes of rotation will cause the axis of rotation of
the XR20 to sweep out a cone, as shown (grossly exaggerated)
in Figure 8. As the test progresses the XR20 will counter-
rotate to ensure the angular reflector remains facing towards
the laser. The combined motions will make the angular reflector
appear to “wobble”. As seen from the laser, the reflector
appears to progress from being tipped slightly forwards (at
position 1), then slightly to the right (at position 2), then slightly
backwards (at position 3) and to the left (at position 4). Note
that, in reality - unlike in Figure 8, the amount of wobble will be
tiny, with the reflector only moving a fraction of a millimetre.

This wobbling motion is illustrated in Figure 9 which shows

how the retro-reflectors wobble relative to the laser due to
misalignment α between the axes of rotation, and when
misalignments β and ω are zero. Figure 8

As seen from the front (upper row of pictures in Figure 9) the reflector appears to tip from side to side
(i.e. roll) by an angle of ±α. At any moment during the test the roll angle is given by α.sinθ, where θ is
the angle of rotation of the axis under test (assuming η = 0°).

As seen from the side (lower

row of pictures in Figure 9)
the reflector appears to tip
(i.e. pitch) back and forth
by an angle of ±α. At any
moment during the test the
current pitch angle is given by
α.cosθ, where θ is the angle
of rotation of the axis under
test (again assuming η = 0°).

The wobbling motion of

the reflector is therefore
a combination of roll of
α.sinθ and a pitch of α.cosθ.
This combination causes
the individual reflectors to
rise and fall relative to one Figure 9
another, as well as moving
towards and away from the laser. The maximum roll occurs when there is no pitch and vice versa. At
positions 2 and 4 the reflectors are at different heights. But, because the reflector is not also tipped
forward or backward, their movements are at 90° to laser beam. Therefore, the laser path lengths
5
 to the reflector are unaffected and there is no measurement error at these positions. At positions 1
and 3 the reflectors are tipped forwards or backwards. But, because they are at the same height, no
measurement error occurs at these positions either.

However, at intermediate positions a measurement error will occur because here there will be some
tipping (pitching) of the reflector whilst the retro-reflectors are also at differing heights.

The difference in retro-reflector height, when the table under test is at an angle θ, is given by S.α.sinθ
and the reflector pitch angle is α.cosθ. This combination will give a change in the relative laser path
lengths in the two arms of the interferometer of S.α.sinθ.α.cosθ or S.α².sinθ.cosθ. Dividing by S and
substituting 2.sinθ.cosθ = sin(2θ) gives an angular measurement error of α².sin(2θ) / 2. This result is
similar to the 1st term in Equation 1, but has twice the amplitude.

To understand why the amplitude is halved, we must also consider the variation in yaw* angle of the
reflector due to the misalignment, α, between the axes. *Yaw is defined as rotation of the reflector
about the axis under test. As the axis under test rotates through an angle θ, the XR20 will counter-
rotate by –θ. However, because the axes are misaligned by an angle α, these rotations will not exactly
cancel producing a small yaw error in the angle of the reflector. This is illustrated in Figure 10.

Imagine the two planes of rotation

are represented by two disks of
unity radius. The axis under test is
shown in blue. The XR20 is shown
in red misaligned by an angle of
α (shown grossly exaggerated in
the figure). Looking from above
(directly along the test axis of
rotation) the test axis (in blue)
appears as a perfect circle. But
the XR20 (in red) appears as an
ellipse. If both axes rotate by 0°,
90°, 180° or 270°. Their angles
of rotation, as seen from above,
appear identical. However, at
intermediate angles this is not
true, as illustrated by triangle ABC. Figure 10
If both axes rotate by an angle θ
then, looking down from directly above, the axis under test (blue) will appear to rotate by θ, but the
XR20 (red) will appear to rotate by θ’. Looking at the plan view and using small angle approximations
it can be seen that length CB ≈ cosθ. α²/2. Looking at triangles ABC and BCD, it can be seen that CD
= CB.sinθ, so CD ≈ sinθ.cosθ. α²/2. Also, the difference in angle θ’ – θ ≈ CD/AC, where AC ≈ 1, so θ’ –
θ ≈ sinθ.cosθ. α²/2. This can be simplified using the trig identity sinθ.cosθ = ½ sin2θ to give the
final result;

Reflector yaw = (θ’ – θ) ≈ sin2θ. α²/4

Inspection shows that the resulting measurement error has the opposite sign to that caused by pitch
and roll of the reflector described earlier. Therefore combining the effects of reflector pitch and roll
(from the optics wobble), with reflector yaw produces a measurement error of:-

α².sin(2θ) / 2 - α².sin(2θ) / 4 = α².sin(2θ) / 4

This explains the 1st term in Equation 1.

6
Figure 11 shows the form of this
measurement error for angular
misalignments of α = 0.1°, 0.25° and
0.5°. Note, although the form of this
error always contains 2 cycles over
360°, the phase may be completely
different depending on η. The graph
indicates the axes of rotation must
be aligned to better than ±0.2° to
reduce the measurement error
contribution, from this term alone,
to below ±1 arcsecond. The XR20
contains spacers (see Figure 5) which
are adjusted during manufacture to Figure 11
ensure the XR20’s axis of rotation
is perpendicular to its mounting face to within ±0.025°. It is also important that the XR20 is mounted
on a surface that is sufficiently perpendicular to the axis of rotation of the axis under test. This can
be verified by indicating the XR20’s mounting ring with a dial gauge whilst rotating the test axis. To
achieve an overall measurement accuracy of ±1 arcsecond, the XR20 manual recommends that the
TIR should be < 0.04mm at 50mm radius (0.025°). Thereby ensuring the total α will be less than
±0.05°.

The origin of the second and third terms, sin(θ).αβ and cos(θ).αω in Equation 1, depends on the
presence of two or more misalignment errors at the same time. There must be a misalignment (α)
between the axes of rotation in combination with some misalignment (β) of the laser and/or some
misalignment (ω) of the reflector optics.

αβ term - Consider what happens when, in addition to a misalignment (α) between the axes of
rotation, the ‘line of sight’ of the laser is also misaligned by an angle β to the plane of rotation of the
axis under test (refer to Figure 7). This additional misalignment alters the effect of the wobble of the
angular reflector, as illustrated in Figure 12.

Previously, when the laser was

correctly aligned (β = 0°), the
side to side tipping (rolling)
of the reflector at positions
2 and 4 didn’t introduce any
measurement error because
the movement of the reflectors
was at 90° to the laser beam.

However, if the laser is

misaligned by an angle β,
then Arms 1 and 2 of the
interferometer will detect a
small resolved component
(ΔL = ±S.sinα.sinβ) of the
differential movement between
Figure 12 the retro-reflectors. The laser
system software will convert
this into an angular measurement by calculating θ = arcsin(ΔL/S), where S is the retro-reflector
separation. Thereby producing an additional variation in the angular reading of ±α.β if α and β are
expressed in radians.

If θ = 0° is defined as being at position 1, (i.e. η = 0), the equation of the induced measurement error is
sin(θ).α.β. This explains the second term in Equation 1 above.

7
 In order to help align the laser and thereby control the value of β, the XR20 contains a reflective target
on the back of the angular reflector which is aligned to the XR20’s axis of rotation to within ±0.33
milliradians. The XR20 manual recommends that, if the laser is placed 1 metre away from the XR20,
the reflected return beam is aligned to within 1mm on the laser shutter. This will ensure the laser is
aligned within ±0.5 milliradians to the reflective target, which in turn (allowing for other errors) will
ensure β is less than 1 milliradian.

Note: The effect of various combinations of misalignment errors on measurement accuracy are
presented in the table in Figure 22, towards the end of this paper.

αω term - Finally, consider what happens if, instead of the laser being inclined at an angle of β, the
angular reflector is twisted by an angle of ω. This is illustrated in Figure 13 (note that Position 1 now
appears on the right).

Previously, when the reflector

wasn’t twisted, (ω = 0°), the
forward and backward pitching
of the reflector at positions
1 and 3 didn’t introduce any
measurement error because
the reflectors were at the same
height.

However, if the reflector is

twisted by ω, this is no longer
true. Comparing positions 1
and 3 it can be seen that the
upper retro-reflector now moves
back and forth in the direction
of the laser relative to the lower Figure 13
retro-reflector. The amount of
differential movement is given by ΔL = ±S.sinω.sinα. The laser system will convert this into an angular
reading by calculating θ = arcsin(ΔL/S), where S is the retro-reflector separation. This will produce
an additional variation in the angular reading of θ ≈ ±sinα.sinω if α and β are expressed in radians.
At positions 2 and 4, although the height of the retro-reflectors still varies (by the same amount as in
Figure 12), this movement is now at right angles to the laser beam, so there’s no measurement error
at these positions.

If θ = 0° is again defined as being at position

1, the equation of the induced measurement
error is cos(θ).α.ω. This explains the 3rd
term in Equation 1 above.

The alignment of the retro-reflectors and

reflector housing are set during XR20
manufacture to be within ±2.5 milliradians.

Combined error – Figure 14 shows an

example of the error profile produced by
combining XR20 manufacturing tolerances
with recommended system alignment
values (giving α = 0.62, β = 1.23 and ω =
2.26 milliradians), and then substituting into
Equation 1.
Figure 14

Note: The effects of various combinations of α and β misalignments on overall measurement accuracy
are presented in the table in Figure 22, towards the end of this paper.

8
Inaccuracies in the rotary encoder reading
There are three main sources of error associated with the rotary encoder reading.
• Sub-divisional errors within the interpolation system
• Bearing wander and eccentricity of the encoder disk
• Inaccuracies in the positions of the graduations formed on the periphery of the encoder disk

These errors and their control are discussed in more detail below.

Sub-divisional errors
Renishaw’s in house rotary encoder manufacturing process produces a total of 7,850 graduations at
20 μm intervals around the periphery of the XR20’s 50mm diameter encoder disk. Giving an effective
angular interval, between graduations, of 165 arcseconds.

In order to increase the resolution, the

XR20 system uses a sophisticated
interpolation system to sub-divide
each angular interval by 2000x to
give an angular resolution of ~0.08
arcseconds. As the encoder disk
rotates, the graduations will move past
the read-head, causing the photo-
detectors inside the read-head to detect
sinusoidal variations in the light levels
falling on them. The photo-detectors
are arranged so that their phases are
offset by 90°, and so produce sine Figure 15
and cosine output signals as shown in
Figure 15. These signals are fed to electronic circuitry which counts the number of cycles of the sine
or cosine signal to give a coarse angular position, to the nearest 165 arcseconds. The electronics also
interpolates within each cycle to give the fine resolution. If the sine and cosine signals are plotted on
the X and Y axes of a graph, a circular “lissajous”
figure appears (see Figure 16). The plotted signals
complete one revolution of the lissajous figure each
time an encoder graduation passes in front of the
read-head, (which occurs every time the XR20
rotates by 165 arcseconds).

Interpolation is carried out by digitising the signals

to give instantaneous values for sine and cosine
and then digitally calculating the phase angle,
φ, by computing φ = arctan(sin/cos). The fine
angle position of the XR20 in arcseconds is then
calculated from φ.165/360, and this is added to the
coarse position to give an overall angular position
Figure 16
value.

This interpolation process works well if the sine and cosine signals are perfect. However, if they are not
perfect the lissajous will be distorted and a sub-divisional error (SDE) will be introduced. The electronic
circuitry within the XR20 is designed to minimise the SDE error as follows;

9
 Firstly, due to optical and mechanical
tolerances between the read-head and encoder
graduations, the phase difference between
the photo-detector signals may not be exactly
90°. This can cause a distorted lissajous, as
shown in Figure 17. This error is removed
during manufacture firstly by mechanical
adjustment and then by electronically mixing a
small proportion of the inverted signal from one
photo-detector with another, to adjust the phase
to 90°.

Figure 17

Secondly, the amplitudes of the photo-

detector signals may not be correct. This can
cause a distorted lissajous as shown in Figure
18. This error is removed electronically as
follows; The XR20 constantly monitors the
amplitudes of the sine and cosine signals.
Errors are corrected by adjusting the intensity
of the LED (which illuminates the encoder
graduations) and by using ABC (automatic
balance control) circuitry, to ensure the both
signals are balanced and of the correct
amplitude.
Figure 18

Finally, there may be a DC offset in one or

more of the signals which can cause a distorted
lissajous as shown in Figure 19. This error is
also removed electronically as follows; The
XR20 constantly monitors the DC offsets of
the sine and cosine signals and uses AOC
(automatic offset control) circuitry to remove
any offsets.

These error corrections ensure that the sub-

divisional error is kept below ±0.25 arcseconds.

Figure 19

10
Bearing wander and encoder mounting eccentricity
There is likely to be some radial wander in the bearings supporting the XR20’s central shaft and
encoder disk. This will allow the encoder disk to move sideways relative to the read-head which could
generate an apparent change in angular position. A rotary encoder system using a small encoder disk
with only one read-head, is extremely sensitive to bearing wander. If an encoder disk of diameter d,
moves a distance t in the direction shown in the Figure 20, due to play in the bearing, the read-head
will detect movement of the graduations and indicate a counter-clockwise movement of t/(d/2) radians.
For example, radial bearing wander of just ±1 micrometres, in combination with a 50mm diameter
encoder disk, would produce a significant error
of ±8.25 arcseconds. Since bearing wander
can have both repeatable and non-repeatable
components, error mapping is not a technique
than can be used to control this error.

To overcome this problem, the XR20 contains two

diametrically opposed read-heads and the final
XR20 position output is based on the average of
the angular positions indicated by each read-head.
Because radial bearing wander causes equal
and opposite errors at each read-head the error
is eliminated. This is illustrated in Figure 21. If the
encoder disk again moves a distance t, read-
head 1 will again indicate a counter-clockwise
Figure 20 movement of t/(d/2) radians. But read-head 2 will
indicate an equal and opposite movement of t/
(d/2) radians in a clockwise direction. When the output from the two read-heads is averaged, the error
is eliminated.

The use of two read-heads also

conveniently removes errors due to
any small off-centre mounting (see
note) of the encoder disk relative
to the axis of rotation of the central
shaft. Note – although the bearings
and encoder disk are designed to be
self centering during assembly, some
eccentricity will still be present due
to manufacturing tolerances.

If the read-heads are not perfectly

aligned (diametrically opposed)
then the cancellation of bearing
wander and eccentricity errors will
Figure 21
not be perfect. However, because
the eccentricity, bearing wander
and read-head alignments within Renishaw’s XR20 are kept within reasonable tolerances by the self-
centering design, these second order errors can be safely ignored.

Note: For more detailed information on bearing wander and eccentricity errors refer to the Renishaw
White Paper entitled ‘The accuracy of angle encoders’.

11
 Graduation Errors
Now consider errors in the positions of the graduations on the encoder disk. Renishaw’s manufacturing
process ensures each graduation on the 50mm diameter disk is positioned within ±0.5 μm, giving an
inherent angular accuracy of just over ±4 arcseconds. Because these graduations are formed directly
on the periphery of the ring they are very dimensionally stable. This, together with the absence of high
frequency components, means this error can be largely eliminated by error mapping. After the XR20
has been assembled, it is calibrated using a laser interferometer system. The measured errors are
stored in an internal error map which is used to correct the XR20’s angular position reading when the
system is in use.

Actual position = Average read-head position ± error map correction

The error mapping process not only reduces inaccuracies due to positional errors of the graduations
on the encoder ring, it also further reduces errors due to any residual misalignment (α) of the central
shaft the mounting surface, misalignment (ω) of the retro-reflectors, and misalignments of the read-
heads mentioned previously.

Note: For more detailed information on bearing wander and eccentricity errors refer to the Renishaw
White Paper entitled ‘The accuracy of angle encoders’.

Inaccuracies in the laser interferometer reading

These error sources have been covered at length in the Renishaw white paper TE326, so a detailed
explanation is not given here, but in summary they are as follows;

Incorrect retro-reflector spacing

The exact centre to centre spacing, between the two retro-reflectors in the angular reflector will vary
depending on manufacturing tolerances and current temperature. To eliminate this error, Renishaw’s
Rotary axis calibration software includes a totally automatic angular optics calibration procedure which
automatically identifies any error in the spacing by comparing the laser reading with the XR20 readout
at 0°, +5° and -5° positions. From these readings the software calculates a correction factor K. This
procedure only takes a few seconds and is always carried out automatically, just before calibration of
the axis under test starts. Subsequent angular readings from the laser are then corrected using the
equation θ = arcsin(ΔL/(K.S).

The benefit of the automated calibration procedure becomes clear by considering a simple example.
Suppose that, due to manufacturing tolerances, the centre to centre spacing between the two retro-
reflectors in the angular reflector is 30.1mm instead of 30.0mm and suppose the interferometer is
measuring an angle of 1°. Without correction, the error in the spacing of the retro-reflectors will cause
an angular measurement error of approximately 0.1mm/30mm x 1° = 0.0033° = 12 arcseconds.
However, after the calibration routine is completed the 0.1mm error in retro-reflector spacing will be
compensated for and the measurement error reduced below 0.1 arcseconds.

Misaligned reflector at datum (i.e. not perpendicular to laser beams)

If the angular reflector is not perpendicular to the laser beam when the system is datumed, a
small measurement error can occur. In addition to identifying any error in the spacing between the
retro-reflectors, the automated angular optics calibration procedure (see above) also identifies any
misalignment of the reflector. The XR20 is then counter-rotated accordingly to bring the reflector into
perfect alignment and the system is then re-datumed before calibration of the axis under test starts,
thereby eliminating this error.

12
Air refractive index compensation errors
The refractive index of air, and hence the laser’s wavelength vary slightly according to the local
weather conditions, but significantly according to the local altitude above sea level. When the
automated angular optics calibration routine identifies any error in the spacing between the retro-
reflectors (see above), the procedure uses laser readings taken under the current atmospheric
conditions. Therefore the value calculated for K automatically includes a correction for air refraction
errors, thereby eliminating them. Note that subsequent changes in air temperature pressure or
humidity will not be compensated for. However, such errors are usually ignored since the path lengths
of arms 1 and 2 of the interferometer are quite similar (and so are affected almost equally) and the
subsequent refractive index change at a fixed altitude is likely to be small. For example a 1°C change
in air temperature, when the interferometer and reflector are misaligned by 2.5°, will only introduce a
change in reading of about 0.05 micro-radians, (0.01 arcseconds).

Thermal expansion of angular interferometer periscope

The Renishaw white paper TE326 shows that the angular interferometer periscope optic typically
has a temperature sensitivity of 13.8 micro-radians/°C (2.8 arcseconds/°C). For this reason it is wise
to ensure the optics have acclimatised before the test starts, to complete the test promptly, and to
minimise local variations in temperature.

Non-parallel beams emerging from the periscope

The beams emerging from the angular periscope are guaranteed to be parallel within ±15 arcseconds.
In the worst case this non-parallelism can introduce a small measurement error of ±0.5 arcseconds
for every millimetre of lateral motion (i.e. at 90° to the laser beam) of the reflector. Such movement
will occur if the axes of rotation of the XR20 and the axis under test are not coincident. It is therefore
recommended that the system is aligned so that the two axes of rotation are coincident within ±0.5mm,
thereby ensuring this error is kept below ±0.25 arcseconds.

Practical error budgeting

It is clear that angular misalignments
have the potential to create significant
measurement errors and the errors
combine in quite a complex manner. It is
therefore important to carefully align the
system and to be aware of the associated
errors. Providing the XR20 is aligned
and operated in accordance with the
recommendations in the manual, the
overall system measurement accuracy is
specified to be within ±1 arcsecond.

If achieving the alignment tolerances Figure 22

recommended is impractical, or if ±1
arcsecond accuracy is not required, it is possible to ‘trade-off’ larger misalignments against decreasing
measurement accuracy levels. In order to simplify this process Figure 22 shows the estimated
measurement accuracies that can be expected for a range of different angular misalignments.

13
 Off axis rotary calibration
This paper so far has concentrated on the
measurement errors associated with calibrating
a rotary axis with an XR20 which is mounted
‘on axis’, i.e. where the axes of rotation of the
two systems are virtually coincident. ‘On axis’
mounting is often straightforward, (for example
when calibrating a rotary table), making it
relatively straightforward to quantify and control
the misalignments α and β.

However there are many rotary axes, especially

on 4 and 5 axis machining centres, where it
is difficult to access the point of rotation and
where there is no convenient mounting surface,
as illustrated by Figure 23. Sometimes it is Figure 23
possible to calibrate such axes by mounting the
XR20 on a custom bracket, as illustrated in Figure 24. But such brackets may be cumbersome, difficult
to align and may flex under the weight of the XR20, producing measurement errors.

In order to overcome this limitation, Renishaw offers an alternative solution involving synchronised
movements of linear and rotary axes as illustrated in Figure 25
which shows a sequence of 3 such synchronised moves. As
the rotary axis moves the linear axis supporting the angular
interferometer is also moved to ensure that laser beams remain
aligned. The advantage of this arrangement is that the bracket
required to support the XR20 can be much more compact and
rigid. The disadvantages are that a more complex part program is
required and the angular data captured maybe contaminated by
angular (pitch or yaw) errors in the movement of the linear axis.

In order to overcome the first disadvantage Renishaw has developed

special software that generates the part program automatically
based on three point setup/alignment procedure.

Figure 24 The second disadvantage can be overcome by also measuring

the angular (pitch or yaw) error in the linear axis over the short
length of travel used
during the rotary axis
calibration. Then,
using special analysis
software developed
by Renishaw, the
angular error recorded
in the motion of the
linear axis can be
subtracted from the Figure 25
rotary calibration
data before it is plotted. Alternatively, if the angular (pitch or yaw) error is checked over a longer length
of the linear axis, it may be possible to select a section of the linear axis where the angular pitch or
yaw error is negligible and position the angular interferometer accordingly.

14
 The same techniques can be applied to the calibration of the
rotary accuracy of tilt and turn trunnion axes as shown in Figure
26. More details of this technique can be found in the ‘Off axis
rotary user guide’ on www.renishaw.com/lasercalsupport.

Figure 26

Error analysis
The errors associated with ‘off axis’ calibration are similar to those described previously for ‘on axis’
calibration, with one addition and a few important caveats.

• The ‘off axis’ procedure relies on correctly

synchronised moves of the linear and rotary axes
to maintain beam alignment. This in turn relies
on a good estimate of the offsets between the
axes of rotation of the XR20 and the axis under
test (see Figure 27). These offsets are estimated
automatically during the three point alignment
setup. If arcsecond accuracy levels are required,
it is important that great care is taken during
this alignment process to ensure that, not only
is adequate signal strength maintained at all 3
points, but also that the alignment of the beams
returned to the laser is accurate. If this is not
checked there is a danger that the reflector will
move laterally (across the laser beam) as the test Figure 27
progresses. As stated earlier, under worst case
conditions (an angular periscope with 15 second beam non-parallelism) a sideways movement
of the reflector of ±0.5mm can add ±0.25 arcseconds of measurement error.

• During ‘off axis’ calibration, the linear separation between the reflector and interferometer
will change during the test (as illustrated in Figure 25 above). Depending on the set-up it
may be difficult to ensure the laser beams, after they are turned 90° by the interferometer,
are accurately aligned. This change in separation combined with poor alignment and the 15
arcsecond non-parallelism of the beams emerging from the interferometer can introduce an
additional measurement error of up to ±0.5 arcseconds. (The origin of this error is described in
detail in sub-section entitled Non-parallel beams from angular interferometer in TE326).

• It is crucial to ensure that the sign convention of the angular data captured during the
movement of the linear axis matches that of the angular data captured during the combined
linear and rotary moves. Otherwise, instead of eliminating the angular errors originating from
the linear axis, the subtraction process will double them!

• It is likely that the alignments achieved for α and β when working “off axis” will not be as good
as those obtained when working “on axis”. As with “on-axis” calibrations, the time spent making
alignment adjustments can be traded against the overall accuracy required using the table
in Figure 22. However, as mentioned above, because “off axis” setups will cause a change in
separation between the reflector and interferometer during the test, it is recommended that an
additional measurement error of ±0.5 arcseconds is added to the values in this table.

15
 Conclusion
This paper has described the operational principles and internal construction of Renishaw’s XR20
Rotary Axis Calibration System. It has shown how the system can be used for both ‘on axis’ and ‘off
axis’ calibration of rotary axes. The main error sources that affect the accuracy of measurement have
been detailed together with how the XR20 system has been designed to minimise them. It has also
provided guidance on the effect of larger alignment set-up errors on the accuracy of measurement.
For further information the reader is referred to the following documents.

References
1. Renishaw White Paper ‘TE326 Interferometric angle measurement and the hardware options
available from Renishaw’.

2. Renishaw White Paper entitled ‘The accuracy of angle encoders’ available on renishaw.com

3. Application note(s) ‘Off axis rotary calibration’

16
This page is intentionally left blank

17
Renishaw plc T +44 (0) 1453 524524
F +44 (0) 1453 524901
New Mills, Wotton-under-Edge,
E [email protected]
Gloucestershire GL12 8JR
United Kingdom www.renishaw.com

About Renishaw
Renishaw is an established world leader in engineering technologies, with a strong history of innovation in product development
and manufacturing. Since its formation in 1973, the company has supplied leading-edge products that increase process
productivity, improve product quality and deliver cost-effective automation solutions.

A worldwide network of subsidiary companies and distributors provides exceptional service and support for its customers.

Products include:
• Additive manufacturing, vacuum casting, and injection moulding technologies for design, prototyping, and production applications
• Advanced material technologies with a variety of applications in multiple fields
• Dental CAD/CAM scanning and milling systems and supply of dental structures
• Encoder systems for high accuracy linear, angle and rotary position feedback
• Fixturing for CMMs (co-ordinate measuring machines) and gauging systems
• Gauging systems for comparative measurement of machined parts
• High speed laser measurement and surveying systems for use in extreme environments
• Laser and ballbar systems for performance measurement and calibration of machines
• Medical devices for neurosurgical applications
• Probe systems and software for job set-up, tool setting and inspection on CNC machine tools
• Raman spectroscopy systems for non-destructive material analysis
• Sensor systems and software for measurement on CMMs
• Styli for CMM and machine tool probe applications

For worldwide contact details, please visit our main website at www.renishaw.com/contact

RENISHAW HAS MADE CONSIDERABLE EFFORTS TO ENSURE THE CONTENT OF THIS DOCUMENT IS CORRECT AT THE DATE OF
PUBLICATION BUT MAKES NO WARRANTIES OR REPRESENTATIONS REGARDING THE CONTENT. RENISHAW EXCLUDES LIABILITY,
HOWSOEVER ARISING, FOR ANY INACCURACIES IN THIS DOCUMENT.

*H-5650-2048-01*
© 2013 Renishaw plc. All rights reserved.
Renishaw reserves the right to change specifications without notice
RENISHAW and the probe symbol used in the RENISHAW logo are registered trade marks of Renishaw plc in the United Kingdom and other countries.
apply innovation and names and designations of other Renishaw products and technologies are trade marks of Renishaw plc or its subsidiaries.
All other brand names and product names used in this document are trade names, trade marks or registered trade marks of their respective owners.
Issued: 0413 Part no. H-5650-2048-01-A