You are on page 1of 17

SPACEWHEEL MICROVIBRATION — SOURCES, APPEARANCE, COUNTERMEASURES

Horst Heimel
Rockwell Collins Deutschland GmbH
TELDIX® Space Wheels
Grenzhoefer Weg 36
D-69123 Heidelberg / Germany
hheimel@rockwellcollins.com

ABSTRACT
Be it in the shape of momentum or reaction wheels – spacewheels are an almost indispensable part of any pre-
sent-day spacecraft that relies on stabilized attitude control. Based on advanced ball-bearing technology, these
devices have proved their usefulness and reliability on satellites and interplanetary space missions over dec-
ades. Inevitably, however, their moving parts are a source of undesirable mechanical vibrations which, though
'micro'-small in absolute size, spread all over the spacecraft structure and have the potential to seriously dis-
turb sensitive instruments.
Earth imaging and extra-terrestric astronomical research are just two examples of space missions today which
make more and more use of high-precision instruments. They continue to boost the need for 'quiet' platforms in
space, which are largely free of vibration noise from whatever source.
As for spacewheels, they turn out to be among the noisiest components aboard a spacecraft. Fighting that noise
needs an understanding of its origin and behaviour first.
The article gives an overview of the major sources of microvibration in spacewheels, explains their various
characteristics and describes proven methods of microvibration measurement and presentation, complemented
by several recent examples of microvibration test results from typical spacewheels. It then focuses on strategies
to mitigate either the intensity of wheel noise itself, or at least the effects of that noise on sensitive components
on board a spacecraft. To this end, appropriate procedures to be followed at the wheel manufacturer's are de-
scribed, and suggestions are made for measures that could be advantageously taken in the responsibility of the
spacecraft builder or operator and, finally, even on the spacecraft under way.

1 INTRODUCTION
Momentum and reaction wheels have continued being the workhorses for spacecraft AOCS through many years.
Recently, however, space programmes with ambitious pointing accuracy requirements like Gaia [6] have be-
come known which do no longer want to rely on wheels for stabilization due to their inevitable microvibration
noise. For a wheel manufacturer, this is reason enough to give a deeper thought to wheel microvibration, and to
think about how a mitigation of that undesired noise can be achieved.

Rockwell Collins Deutschland GmbH (RCD) of Heidelberg, Germany, formerly known as TELDIX GmbH, is
the manufacturer of a broad variety of momentum and reaction wheels which are now known under the trade-
mark of “TELDIX® Space Wheels”. All of these wheels are running on ball bearings, a proven technology with
a high record of applications on successful space missions.
But it is just the ball bearings which have to be counted to the strongest sources of microvibration in a wheel.
Before starting any effort of improving the wheel noise situation, the sources of that noise must be understood in
detail. While some primary sources of wheel microvibration have been known and their generation laws have
been published since years, secondary sources like sidebands etc. have not been in the focus so far, although
they can contribute nearly as much as the primary sources to the whole concert of wheel noise.

With more than 20 years of experience in spacewheel microvibration that have been acquired in this company so
far, recent advances in microvibration measurement and analysis technology have enabled RCD to treat microvi-
bration as a routine task rather than a lab experiment. A number of different graphic representations of wheel noise
can be prepared, but not all of them appear to be useful, though often highly appreciated by the public.

Measurement results from wheel pairs which have experienced different levels of vibration lead to the formula-
tion of a “Golden Rule” of wheel noise mitigation. Based on this rule, a number of recommendations are made
which should, if properly applied, lead to ball bearing spacewheels with reduced noise levels in the near future.

Page 1 of 17
2 BASIC SPACEWHEEL DESIGN
Covering the complete range of angular momentum capacities between 0.04 Nms and 68 Nms, TELDIX®
Space Wheels provide a family of momentum and reaction wheels suiting the needs of spacecraft in size classes
between 30 and 5000 kg. A selection of these wheels is shown in Fig. 1, including wheels driven by external
wheel drive electronics (WDE) like the RDR 68, as well as samples like RSI 12 and RSI 0.1 which have their
WDE integrated into the wheel housing. The operating characteristics of some wheel types are given in Table 1.

The vast majority of all brands of spacewheels built to date are of the ball bearing type. They usually consist of
the following main components:
• Flywheel
• Ball bearings
• Drive motor
• Housing
• Electronics assembly (in case of internal WDE).
For TELDIX® Space Wheels, some distinctive features are to be noted:
• Modular design: Each component built and tested separately
• Self-contained bearing unit with 2 outer-race rotating bearings in solid preload
• Lightweight housing without support function
• A selectable variety of flywheel types, including spoked-wheel and disc-shape designs.

Figure 1: Three representatives of the TELDIX® family of spacewheels:


RDR 68 (left), housing cover removed; RSI 12 (top right), section drawing; RSI 01-5 (bottom right)

Wheel type Ø [mm] Mr [mNm] D0 [Nms] n0 [rpm] Op. range


RDR 68-75/600 345 55 68 6000 0 … n0
RSI 45-75/60 310 55 45 6000 –n0 … n0
RSI 20-215/18 345 200 20 1800 –n0 … n0
RSI 12-75/604 247 55 12 6000 –n0 … n0
RSI 01-5/28i 95 5 0.12 2800 –n0 … n0

Table 1: Operating characteristics of some types of spacewheels (n0= nominal speed,


D0= nominal angular momentum, Mr= min. reaction torque at nominal speed)
3 SOURCES OF WHEEL MICROVIBRATION
Evidently, a variety of wheel types like that presented above will generate a variety of noise signatures under
operation. The sources and general laws of wheel noise, however, are the same for whatever type and brand of
ball-bearing spacewheel. This topic has been presented in detail in various papers such as [1] over the past 15
years. A recent account, providing a generalized formulation of bearing disturbance frequencies (albeit with a
couple of mistakes in the frequency formulas) can be found in [2]. Therefore, we can limit ourselves to a short
summary of the well-known effects here, supplemented by some more detail in a few cases which apparently
have acquired a lower level of familiarity so far.

Induced vibrations from a rotating wheel can generally be attributed to disturbances from three distinct sources:
• Unbalance,
• Motor noise,
• Bearing disturbances.
Each of these sources acts independently from the others, and each of them obeys a different set of laws. What
they have in common is that each of them defines a family of disturbance frequencies which are tied to wheel
speed by characteristic order factors
f
νi = i , (1)
f0
where index i denotes the individual disturbance source, fi the frequency of disturbance i, and f0 the plain fre-
quency of flywheel rotation at a speed of n rpm,
n
f0 = Hz . (2)
60
Except for unbalance each of these disturbances, when acting by itself, will usually not exceed a magnitude of
a few milli-Newtons. Structural resonances, however, make all the difference: Whenever a single disturbance
meets one of these resonances at the proper wheel speed, it may be pushed up to a vibration peak sized a multi-
ple of its original magnitude. So, for a complete picture of wheel noise, it is necessary to describe both, active
disturbances and structural resonances, and their interactions.

3.1 Unbalance
Static unbalance Us occurs when the mass distribution of a perfectly symmetric rotor is disturbed by a small
unbalance mass mu located a distance r from the centre of rotation. Its magnitude is then given by
U s = mu · r . (3)
An alternative formulation, describing the effect of a small eccentricity e at the centre of mass on the complete
rotor of mass mr , is
U s = mr · e . (4)
On a wheel rotating at speed n with angular velocity
ω0 = 2π · f0 , (5)
static unbalance is observed as a radial force Fu with magnitude
Fu= Us · ω0². (6)
Rotating with the wheel at the same angular velocity ω0, the vector of unbalance force can be written as
 cos(ω 0 t + ϕ s )  (7)
Fu (t ) = Fu ⋅   ,
 sin(ω 0 t + ϕ s ) 
where φs is an arbitrary but fixed phase constant.

Similarly, dynamic unbalance Ud occurs when the mass distribution of a rotor is disturbed by a pair of small
unbalance masses mp, each located a distance r from the centre of rotation, and both separated by an angle π and
an axial distance d. Its magnitude is then given by
U d = mp · r · d . (8)
On a wheel rotating at speed n with angular velocity ω0, dynamic unbalance is observed as a transverse mo-
ment Mu with magnitude
Mu= Ud · ω0², (9)
and the rotating vector of unbalance moment can be written as
 cos(ω 0 t + ϕ d )  , (10)
M u (t ) = M u  
 sin(ω 0 t + ϕ d ) 
where φd is an arbitrary but fixed phase constant. The phase difference ∆φ between static and dynamic unbal-
ance is fixed as long as the mass distribution between the rotor and the axis of rotation is not disturbed, but it
can assume any value between 0 and 2π.
Some resulting facts about unbalance as a disturbance element of microvibration are:
• Both types of unbalance appear at order νu = 1, measured in transverse forces Fx,y and
transverse moments Mx,y, respectively.
• While unbalance force Fu can be directly read from measured transverse force Fx at ν = 1,
this does not work with unbalance moment Mu and measured transverse moment Mx. In-
stead, Mx has to be transformed into moment Mx’ first, which is referred to the centre of
mass (COM) of the rotor.
• Correspondingly, the moment transferred onto the spacecraft at ν = 1 cannot be predicted
from a known dynamic unbalance to better than Mu ± Fu · h, where h is the distance be-
tween the wheel’s mounting flange and the COM of its rotor.
• The square law of unbalance force/moment w.r.t. wheel speed presents an idealistic view,
strictly valid for rigid rotors only. In reality, most spacewheel rotors will expand with
speed, and due to the non-ideal symmetry of that expansion, unbalance will change. This
usually makes unbalance force/moment deviate from the square law to either side of the
original balancing speed.
• Unbalance can increase, decrease, or just change its phase relationships through the ac-
tion of external forces, if these are strong enough to make the rotor move through any
pre-existing gaps. Following (4), a “gap” of 2 µm width is sufficient to impose on a rotor
of 5 kg mass an unbalance change of up to ± 1 g·cm under heavy shock or vibration.
• Higher harmonics of unbalance force and moment are occasionally found on ball-bearing
spacewheels, but they are generally much weaker than unbalance force/moment them-
selves, and usually marginal beyond order 4.
• A small impact of unbalance and its harmonics is sometimes detected in axial force Fz
and axial moment Mz as well.

3.2 Motor Noise


Brushless motors with ironless stators, like those employed in all members of the TELDIX® Space Wheel fam-
ily, generally do not contribute much to the overall vibrations induced by a running wheel. Nevertheless, apply-
ing a logarithmic scale on the results of sensitive measuring equipment, some typical low-level contributions
can be detected in axial moment Mz.
The most prominent disturbance generated by the motor is commutation noise, resulting from the ele ctronic
switching between stator phases upon passage of the magnetic poles of the rotor. As such, in a 3-phase motor
with 8 pole pairs, it appears at a fundamental order of ν = 24. Due to the non-sinusoidal nature of the elementary
effect, it is always accompanied by a series of higher harmonics at orders 48, 72, etc., with ever decreasing
magnitudes. The size of the order-24 disturbance is rarely found higher than 10% of the applied motor torque,
with only little speed dependency. This disturbance quickly disappears into high-frequency areas with speed,
leaving the usual 500 Hz range of microvibration measurement at a speed of 1250 rpm.
Magnetic cogging should not play a role in a motor without static iron. Nevertheless, a small cogging-like dis-
turbance can be detected at ν = 8 and its higher harmonics. Usually sized more than an order of magnitude be-
low the commutation disturbance, this component of motor noise can be attributed to secondary eddy current
effects. It is observed even under coast-down with zero motor current and disappearing commutation noise.
Since commutation noise is depending on the delicate timing of phase switching in the WDE, its magnitude can
occasionally be the user’s rather than the wheel manufacturer’s responsibility. This can occur with wheels of
e.g. type RDR, which are sometimes delivered without WDE to be driven by user-supplied electronics.
3.3 Bearing disturbances
Resulting from the rolling of bodies with different diameters on each other, bearing disturbances appear at real-
valued order numbers throughout. All of these order numbers can be derived from a few basic parameters of
bearing geometry. With
d
db = ball diameter, dp = bearing pitch diameter, δ = b = diameter ratio,
dp
N = ball number, α = bearing contact angle, and γ = δ ⋅ cos( α ) , (11)
Table 2 presents the meanwhile well-known order formulas for the 4 principal bearing disturbances in a concise
shape written for outer-ring rotating bearings, along with their order values for the bearing type employed in
most members of the TELDIX® Space Wheel family.
Except for cage disturbance C, all of these bearing disturbances can be explained by assuming a single localized
imperfection (a tiny scar or bump) on the outer or inner race or on a ball, being rolled over multiple times during
every revolution of the wheel. As for disturbance C, the most convenient explanation is that of unbalance in the
ball complement or the cage itself.
BI, the disturbance of ball imperfections, is special in that it appears at twice the underlying ball spin frequency
BSF. This is due to the fact that a single imperfection on a ball will hit on both the inner and the outer race, i.e.
twice, during every ball revolution. Actually, BI is regularly observed on spacewheels, whereas BSF is usually
found much weaker, if not completely vanishing in background noise.
A number of secondary effects associated with the principal disturbances are regularly detected in microvibra-
tion measurements as well. These are listed in Table 3.
Some additional remarks on principal and secondary bearing disturbances are collected below.

Name Alt. name Order formula Order value Remarks


C BTF 1 0.59 Primarily in Fx,y, Mx,y
Cage Ball Train ⋅ (1 + γ ) (12)
disturbance Freq. 2
IRI BPFI N 7.08 Primarily in Fz; also Fx,y , Mx,y
Inner Race Ball Pass Freq. ⋅ (1+ γ ) (13)
Imperfections Inner Race 2
ORI BPFO N 4.92 Primarily in Fz
Outer Race Ball Pass Freq. ⋅ (1 − γ ) (14)
Imperfections Outer Race 2

( )
BI 2×BSF 1 5.00 Primarily in Fz ;
Ball Double Ball ⋅1−γ2 BI appears much stronger (15)
Imperfections Spin Freq. δ than BSF

Table 2: The principal bearing disturbances and their order values in most of the TELDIX® Space Wheels

Name Description Sample order values / Remarks


ORI_k ν_k = k · ν, Higher integer-order ORI_2= 9.84; ORI_3= 14.76; ORI_4= 19.68 ...
harmonics of principal IRI_2= 14.16; IRI_3= 21.24; IRI_4= 28.32 ...
IRI_k k = 1, 2, 3, ...
bearing disturbances
BI_k BI_2= 10.0; BI_3= 15.0; BI_4= 20.0 ...

ORI+, ORI- Upper/lower sidebands due ORI+ = 5.92; ORI– = 3.92


to amplitude modulation: IRI+ = 7.67; IRI– = 6.49
IRI+, IRI- ν+ = ν + m,
m = 1 for ORI,
BI+, BI- ν– = ν – m m = 0.59 for IRI, BI BI+ = 5.59; BI– = 4.41

Beats on C, ORI, IRI, BI Due to unequal but closely - Beat frequencies small, unpredictable
neighboured disturbance
frequencies from paired
bearings

Table 3: Secondary bearing disturbance effects and some of their order values in most of the TELDIX® Space Wheels
• Secondary bearing disturbances usually appear at lower magnitudes than the associated
principal ones. However, they are often strong enough to drive a structural resonance to
considerable height by themselves. A microvibration model that does not include these
effects will be faced with vibration peaks occurring at several unexpected locations in
frequency/speed space.
• Beats on principal or primary disturbances are readily detected on microvibration plots
extending over a range of speeds, but their depths and beat frequencies are hardly predict-
able and will vary from one individual wheel to another. In wheels running on more than
two bearings, these effects will be even more pronounced at less recognizable regularity.
• Waviness of inner or outer races is sometimes introduced as an explanation for higher
harmonics of bearing disturbances (e.g. in [2]). While this concept cannot be excluded as
a potential cause of principal disturbances and their low-order harmonics, it does not
match well with observations made during bearing inspections after measurement. These
regularly reveal singular or grouped marks on balls or races as the causes of disturbances
of conspicuous magnitude. And empirical evidence exists that a single scar, deliberately
applied on an outer race, produced a formidable ORI along with a series of harmonics ex-
tending well beyond the 100th order. So, the concept of singular or multiple imperfections
resulting in non-harmonic elementary effects, each of which gives rise to a series of
higher order Fourier coefficients defining the harmonics, is preferred.
• The speed dependency of bearing disturbance magnitudes has occasionally been de-
scribed as following a square-speed law, like unbalance (see e.g. [4]). But except for the
C disturbance, which can be thought of as resulting, at least partly, from unbalance of the
cage and the ball complement, such behaviour is usually not observed with the other three
disturbances. Actually, most observations on real wheels made so far are compatible with
the assumption of speed-independent disturbances, being gradually amplified as they ap-
proach a broad resonance, and returning to their original magnitude as they move away
from it on the other side. A model relying on a square law instead will substantially over-
estimate the disturbance in question at all frequencies beyond that of the resonance.

3.4 Structural Resonances


Though not belonging to the active sources of wheel vibrations, structural resonances play an important role in
the microvibration concert through their passive but often immense amplification of otherwise hardly percepti-
ble disturbances.
According to the laws of structural dynamics, each wheel type, as a specific dynamic system, comes with a spe-
cific set of eigenmodes associated with presumably fixed frequencies. But since ball bearings exhibit a nonlin-
ear spring characteristic, most of the eigenfrequencies of wheels are found depending on the vigorousness of
excitation. Therefore, mode frequencies observed in vibration tests, under several g of excitation, are never
identical to the frequencies of the same modes when measured under microvibration.
In the following, mode frequencies will always be given as measured under microvibration conditions, i.e. with-
out any external excitation. They will be called ‘static’ when referred to the state with the wheel at rest.

The most important eigenmodes of a wheel are the three lowest ones, in which the heavy flywheel is the most
active part. In the majority of the TELDIX® Space Wheels, these are one translational and two rocking modes,
which are named and described as follows:
• Ax — Axial translational mode of the flywheel,
• Ra — Antiparallel Rocking mode of the flywheel,
• Rp — Parallel Rocking mode of the flywheel.
The naming of these modes is evident from the mode shapes as outlined in Fig. 2: Rp exhibits a rocking motion
of the flywheel that goes with the tilting of the bearing shaft, whereas in Ra the flywheel is rocking opposite to
the bearing shaft. By this, Ra comes close to a transverse-translational mode, but it still includes the rocking mo-
tion of the bearing unit, and in some wheel types even exhibits a true small rocking motion of the flywheel.
The mode frequencies of a number of wheels from the TELDIX® Space Wheel family, as they are observed
under microvibration conditions, are given in Table 4. For virtually all of these wheels, the lowest static mode,
Rp0, is found above 80 Hz.
Figure 2: Outline of three major eigenmodes of TELDIX® Space Wheels, with rest mode at top left

Wheel type Rp0 Ra Ax


RDR 68-75/600 86 Hz 248 Hz 231 Hz
Wheel A 167 Hz 388 Hz 395 Hz
RSI 12-75/604 142 Hz 398 Hz 350 Hz

Table 4: Mode frequencies of some members of the TELDIX® Space Wheel family

Mode Rp is special in that it evokes gyroscopic effects through its rocking as soon as the wheel is rotating. By
this, that mode splits up into two branches, the so-called “whirl modes”, diverging in frequency increasingly
with growing wheel speed. Here, these branches are called
Rpl — the lower branch, or “negative whirl mode”, and
Rph — the higher branch, or “positive whirl mode”.

With f0 = frequency of rotation as defined in (2),


fm = static frequency of rocking mode,
I1 = axial moment of inertia of rotating mass,
I2 = transverse moment of inertia of rotating mass,
the split-up of whirl mode frequencies fh and fl is given by *)
2
 I   I 
f h, l ( f 0 ) = ±  1 ⋅ f 0  +  1 ⋅ f 0  + f m2 . (16)
 2I 2   2I 2 

Campbell Chart
With all the information on sources and amplification of wheel microvibration given so far, a diagram can be
generated that presents the active disturbances and the structural resonances of a wheel together in frequency–
speed space. Such a diagram, often called a “Campbell Chart”, is given in Fig. 3 for a momentum wheel of type
RDR 68-75/600.
The most interesting features in that chart are the intersections between disturbance lines and resonance lines:
These points, at which a disturbance passes through a structural resonance, are potential locations of vibration
peaks. While the chart tells nothing about the magnitude of these peaks, it does predict the wheel speed as well
as the frequency at which each of them will occur. Candidates for the highest peaks are the intersections of the
principal disturbances ORI, BI, and IRI, with any resonance line. Some secondary disturbances like ORI- or
BI2, however, can result in considerable peak heights at their intersections with resonances as well.

*)
When looking up the whirl mode formula (16), please be aware it is often stated with inconsistently mixed plain and angular frequencies.
Disturbances and Structural Resonances in RDR 68-75/600
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500
400 400
<--IRI3
<-BI4 IRI2
350 BI2 350
ORI4
ORI2 ORI+ BI+
IRI4 <---ORI3
300 <----M
300
<-----BI3 IRI ORI
BI Ra -- 248 Hz
250 250
f [ Hz ]

BI- Ax -- 231 Hz
200 Rph 200
ORI- 3
150 150
2
100 Rp0 -- 86 Hz 100
1
50 50
C Rpl
0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500
n [ rpm ]

Figure 3: Computed Campbell Chart for wheel type RDR 68-75/600

4 APPEARANCE OF WHEEL MICROVIBRATION


In the following, two practical aspects of spacewheel microvibrations are discussed: How they are measured,
and how the measurement results can be presented most suitably.

4.1 Measurement
The most widely accepted approach to the measurement of spacewheel microvibration is that of measuring the
forces and moments appearing at the wheel’s mounting flange, its mechanical interface to the spacecraft, in a
standardized environment. The best-suited tool for this task is a multi-channel dynamometer table capable of
capturing 3-axial forces and force pairs at the same time. The devices made by Kistler have become a sort of
industry standard in that field, and ‘Kistler platform’ a sort of generic term for this type of instrument.

The standardized environment consists of having the wheel rigidly mounted on the Kistler platform, which itself
is rigidly mounted on a heavy mass, preferably sitting on a damping structure isolating the assembly from
ground vibrations. At RCD this is achieved with a large and heavy granite table resting on air-bearing supports.
The wheel is usually mounted on the platform with its rotation axis upright, coinciding with the Z axis of the
dynamometer. A close-up photograph of this measurement assembly is shown in Fig. 4.

From its four 3-axial force measurement cells, the Kistler platform provides 8 voltage force signals at the output
socket of the connected multi-channel charge amplifier. Using the specified distances between the elementary
measurement cells, these signals are subsequently computed into the 3 force components Fx, Fy, Fz, 3 moment
components Mx, My, Mz, and finally two special moment components Mx’, My’ referred to the CoM of the
flywheel for the determination of dynamic unbalance.

Following signal acquisition, the multi-channel data have to be pre-processed, stored, analyzed and edited to
arrive at a meaningful presentation. Over the past few years, RCD has initiated and pursued the development of
a powerful computer-based tool for microvibration measurement and analysis, called “The MyVibSystem”
(MVS), by engineering consultants MetaDAQ. Consisting of custom-made software modules for control and
analysis and dedicated acquired digital data acquisition (DAQ) equipment, this tool allows multi-channel data
acquisition and storage under digital control of wheel operation, complemented by a real-time preview display
of time histories as well as FFT-processed spectral data. It includes modules for offline data analysis applying
FFT, order tracking and other techniques, and a report-generating tool. All lines of action, from start of the
wheel to final printout of the report, can be controlled interactively or run in a script-controlled automatic se-
quence. A screen shot showing the DAQ section of MVS is presented in Fig. 5.

Figure 4: A spacewheel mounted on a Kistler platform for microvibration measurements

Figure 5: The MyVibSystem, DAQ section, main control screen and real-time FFT preview (Screen shot courtesy of MetaDAQ)
With this tool, microvibration measurement has turned into a routine task, making it possible to scan space-
wheels routinely for their microvibration characteristics. By this, a data base is being built which, by virtue of
the offline analysis capability of MVS, allows e.g. health checks of wheels after exposure to unscheduled me-
chanical impacts by comparison with unaffected previous data from the same wheel.

As the standard measurement procedure, a run of the wheel through all four quadrants of clockwise (cw) and
counter-clockwise (ccw) operation has been established. The quadrants are labelled Z1= cw-up, Z2= cw-down,
Z3= ccw-up, Z4= ccw-down.
Making use of RCD’s new digital GSE device, the Universal Wheel Interface (UWI), each of the four runs is
performed at a standardized constant acceleration to achieve maximum reproducibility of microvibration data
from a single wheel, and maximum comparability of data between different wheel types.
The standard frequency range of these measurements has been fixed at 5 ... 500 Hz. The lower end of that range
is dictated by the eigenfrequency of the massive isolator assembly, the upper end by the onset of eigenmodes of
the test rig itself between 500 and 1000 Hz, and by the frequency requirements reported by numerous satellite
builders over the years. A recent move towards an extension of the practicable frequency range is described
below. Besides microvibration spectra of all force and moment components, the results of that measurement run
include the precise values of static and dynamic unbalance of the wheel, determined as a reference value at the
balancing speed, or tracked over the complete speed range of the run.

4.2 Presentation
Complex arrays of bulk data as they accumulate in a microvibration characterisation session of a wheel require
dedicated powerful procedures to arrive at a meaningful illustration of the measurement results.
Doubtlessly, the most appealing form of data presentation in the microvibration field is that of the so-called
“Waterfall Plot”. It presents the complete data set of a single force or moment component, recorded during a full
runup or rundown of the wheel, in a 3D view of FFT spectra extending over the speed- frequency plane. Dili-
gently done, this format provides an impressive, if not beautiful view on the whole of microvibration events as
they developed during the run. Fig. 6 gives an example of two waterfall plots of Fx and Mx from the runup of a
sample wheel, enhanced by order track cursors highlighting some of the active disturbances and showing their
frequencyorder numbers.
But there are two drawbacks adhering to the waterfall format: First, while providing a nice overview of mi-
crovibration behaviour of the wheel, it cannot answer precise questions like “What is the highest noise to be
expected at X Hz?” or “At which speeds will the moment noise level be lower than Y Nm?” And second, it has
to be prepared with diligence, otherwise it will often end up in an impermeable mess of lines. Therefore, water-
fall plots are time and cost consuming tasks, well suited to act as a singular signboard for a complete wheel pro-
gram, but quite inappropriate for routine procedures applied to every single wheel on the production line.

Figure 6: Waterfall plots of Fx (left) and Mx (right), runup cw 0 – 6000 rpm of sample wheel A
There are, however, powerful but less demanding alternatives at hand: A measured Campbell Chart, often pre-
pared in the shape of a colour-coded Contour Plot, provides accurate information about where in speed-
frequency space the most prominent microvibration events have occurred. In addition to that, it allows a direct
comparison of measurement results with the predictions contained in the computed Campbell Chart, which is
usually prepared much earlier in the design process. Two Campbell Charts of moment Mx and force Fz, meas-
ured during runup of a wheel of type RDR 68, are shown in Figs. 7 and 8. Together, they are the direct counter-
parts of the computed chart given in Fig. 3. While the latter has both transverse and axial directions combined in
one chart, the measured charts are referred to a single direction each.

Figure 7: Contour Plot (measured Campbell Chart) of transverse moment Mx, RDR 68 in runup 100-6000 rpm

Figure 8: Contour plot (measured Campbell Chart) of axial force Fz , RDR 68 in runup 100-6000 rpm
Evidently, a drawback of the Campbell Chart lies in the fact that it does not provide accurate peak height infor-
mation. Hence, it needs two companions: The Worst-Case Spectrum (WC) and the Noise-vs-Speed Plot (NvS).

The WC Spectrum results from a series of FFT spectra taken quasi-continuously in peak-hold mode during the
run of the wheel, with each frequency bin retaining only the highest amplitude values ever occurring there. That
spectrum can be thought of as representing an edge-on view from the frequency axis side of a waterfall plot. By
this, every highest peak is accurately mapped versus frequency, and the spectrum provides a worst-case enve-
lope over everything that occurred in microvibration in a particular direction of force or moment, over a com-
plete run.
An example of a WC plot of force Fx, again from a wheel of type RDR 68, is given in Fig. 9. In this plot, the
envelope curve of the WC spectrum is accompanied by three different Steady-State Spectra, representing data
recorded at 1000, 3000, and 5000 rpm. These spectra reveal what the WC plot itself tends to hide: There is no
single speed of the wheel, and by this no moment in time, at which the wheel generates microvibration with a
continuous spectrum, like the envelope seems to indicate. It is rather a series of sinusoid peaks, each with a
deterministic cause and at a distinct, predictable frequency, that make up the elementary noise spectrum of a
wheel at a fixed speed. Most of these peaks are far below the envelope, and a few of them at most will ever
touch the envelope at any given speed. Therefore, the WC plot should never be used to obtain parameters for
any single-speed based model. On the other hand, it is best suited to demonstrate compliance of a wheel with a
microvibration level specification.

Complementary to the WC spectrum, the NvS Plot can be seen as an edge-on view on the speed axis side of the
waterfall plot. In this view, all events occurring at a distinct speed are RSS-summed over the frequency range
under consideration. This provides the exact information at which speeds the overall noise level of the wheel
will be high or low. An example of an NvS plot, again taken with a wheel of type RDR 68 during a runup be-
tween 100 and 6000 rpm, is given in Fig. 10.

Together, the WC and the NvS plots, complemented by a measured Campbell chart, provide the information
that is hidden in the waterfall view: What are the magnitudes of the highest peaks, and where are they located in
speed-frequency space.
If, however, more information is needed, .e.g. for the detailed tuning of a wheel microvibration model, then data
reduction in the form of graphical representations is no longer appropriate. In this case it will be necessary to
keep and examine the complete data sets recorded in a microvibration measurement run. RCD’s new measure-
ment, analysis and reporting system MVS paves the way to such demanding tasks in that it actually stores the
complete time histories of all vibrations sensed during a complete wheel run.
RDR 68 Fx, Peak-Hold and Steady-State Results
100
PeakHold Fx, 100-6000 rpm
Fx steady @ 5000 rpm
Fx steady @ 3000 rpm
10
Force Amplitudes Fx [N]

Fx steady @ 1000 rpm

0,1

0,01
0 100 200 300 400 500
Frequency [Hz]

Figure 9: Worst-Case Spectrum of Fx in RDR 68, runup 100-6000 rpm, with 3 Steady-State Spectra embedded
RDR 68 Fx Noise vs. Speed
10

8 Fx Overall_rms (5 - 500 Hz)


Overall_rms force Fx [N_rms]

0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Speed [rpm]

Figure 10: Noise-vs-Speed Plot of Fx in RDR 68, runup 100-6000 rpm, at 5-500 Hz

Some additional remarks on microvibration measurement and data presentation are following:

• Windowing: Microvibration measurements at RCD generally apply the Flat Top Window in short-
time FFT. This is preferable to the widely used Hanning Window in that it provides more accurate peak
amplitudes than the latter, at the cost of a somewhat reduced frequency resolution. The Hanning win-
dow tends to underestimate peak heights.
• Power spectral density (PSD): The microvibration spectrum of a wheel at any constant speed consists
of a number of discrete sinusoidal elements resulting from deterministic elementary events. PSD is in-
appropriate for the description of a spectrum like that, because the measured “densities” of a single
spectrum lines are quantities that depend on the filter bandwidth applied in the FFT.
• Speed selection: NvS plots can be used to select “silent” speeds for continuous operation. However,
final selection of wheel speeds should preferably be based on measurements with the wheel integrated
on the spacecraft, due to coupled modes arising from the flexibility of mounting brackets and other
structural elements which can make dynamic conditions quite different from those of the rigid mount-
ing case. This requires the availability of suitable sensors on board. For a first assessment of the situa-
tion before launch, a coupled dynamic model of the wheel and the spacecraft structure can be used.
• Frequency range: As described above, the frequency range of microvibration measurement was lim-
ited to 5 – 500 Hz so far. Following a recent call for an extension to above 1000 Hz from a European
space programme, a development was started to overcome the higher-end limit set by test rig eigen-
modes between 500 and 1000 Hz. By first determining the frequency response function of the test rig,
then applying it to the measured raw data as a correction to suppress the detrimental test rig modes, the
measurement system MVS is now capable of delivering measured data of true wheel-generated
microvibration up to a frequency limit of 1200 Hz. Investigations into a further extension are ongoing.
• Dominant noise events: As a general rule for most TELDIX® Space Wheels, the most prominent
noise events in wheels running up to 6000 rpm are:
- At 50 – 100 Hz — Unbalance (continuous; largely insignificant at speeds below 3000 rpm)
- At 100 – 500 Hz — Principal or secondary bearing disturbances amplified by resonances
(at certain discrete speeds only).
5 COUNTERMEASURES
After introducing noise sources as well as measurement and presentation techniques, the possibilities of noise
avoidance or mitigation can now be discussed.
The most promising targets for any measures towards noise mitigation are, in order of precedence:
• Avoidance / mitigation of wheel noise at the source,
• Prevention / mitigation of noise propagation onto the spacecraft,
• Prevention / mitigation of noise reception at the locations of sensitive instruments.
The author’s area of competence calls for a concentration on the first point here.

5.1 The Golden Rule


Which is the most prominent cause of wheel noise at its point of generation, in the wheel itself? Microvibration
measurement techniques do detect bearing disturbances and unbalance in any ball bearing wheel, even if it is
pristine, and at rather low levels then. But what makes these disturbances increase up to the point where they are
recognized as detrimental noise? An example, illustrated by Fig. 11, leads the way to the answer:

Two wheels of the same type, built to the same specification, are measured on the Kistler platform. One of these
wheels is a flight model and has been driven through an FM level acceptance test, the other one, a protoflight
model, has experienced an acceptance test at PFM levels. The main difference in the history of these two wheels
is that the PFM went through random vibration testing at an rms level 25% higher than that of the FM (the ratio
is 1.562 at PSD level).
Expecting an increase of wheel noise by 25% or even 56% results in a surprise: The two WC plots of Fx mi-
crovibration over a wheel runup reveal that the peak noise level of the PFM at about 245 Hz has increased by a
factor of 3.6 over that of the FM. Only little lower factors are found at other locations between 5 and 500 Hz,
except for the unbalance trace which shows a rise by 2.2 “only”.

Clearly, simulation of the loads applied as well as visual inspection of the PFM wheel’s bearings after meas-
urement showed no sign of bearing degradation beyond the level of flightworthiness. Nevertheless, a modest
increase of mechanical load has resulted in a far over-proportional increase of noise level. The experience from
this example, which has been confirmed again and again in other wheel programmes, is simply: The reason for
a different microvibration behaviour of the two wheels is their different history of mechanical environments,
which has created some more tiny marks in the bearings of the PFM than of the FM wheel. Apparently,
tiny marks on balls or races below the flightworthiness threshold cannot be seen but heard.
And the only reasonable conclusion on how to avoid such tiny marks in the ball bearings is what we call
“The Golden Rule of Wheel Noise Mitigation at the Source”:
To keep a wheel silent, treat it gently .

RSI 20 FM, Fx in Runup 100-1800 rpm cw RSI 20 PFM, Fx in Runup 100-1800 rpm cw
[ Input-PSD(PFM) = 1.562 · Input-PSD(FM) ]
Figure 9: WC plots of Fx, two wheels with different mechanical environment history.
Of course, this does not mean ball bearing wheels have to be treated like raw eggs at any time. Of course,
flightworthiness of a wheel still means it has passed the specified vibration levels, and the supplier demonstrates
full specified performance and guarantees the specified operating life. But there could be a strategy to minimize
mechanical impacts in the area below the required treatment / below anything hazardous to bearing life to even
lower magnitudes, and thereby keep wheel-induced microvibration at a minimum.
This strategy would have to involve all phases of wheel life, from planning to building, and from mounting over
carrying to orbit till using it on a spacecraft. A number of recommendations on how to establish such a strategy
have been drafted for each of these phases.

5.2 Recommendations
• To System Engineers & Mission Planners:
o Keep mechanical environmental requirements to a minimum.
Avoid stacking load envelopes resulting from different requirements on each other. A trade-off
between maximum safety assurance by testing and noise mitigation can be part of the strategy.
o Avoid crafting mechanical load requirements after the worst-case launcher scenario.
Since there are launchers in the world that are known for greater violence, and others that are more be-
nign, in terms of the vigour of vibrations they impose on their payloads, planning for a mission to ride
on a benign one would reduce vibration test levels and thus help to get a more silent wheel on orbit in
the end.
o Get in contact with the manufacturer early for tailored solutions.
Basic parameters like spacewheel size, resonance positions, operating speed ranges etc. should be dis-
cussed in the light of wheel noise mitigation as early as possible in a wheel programme.

• To Wheel Manufacturers:
o Avoid shocks in handling, transport etc.
Using anti-shock carts instead of standard lab trolleys, and carrying wheels in cushioned containers,
can do a good job in the mitigation of transport shocks.
o Call for sufficient notching in vibration tests.
With noise mitigation in mind, notching should be called for at levels lower than those just satisfying
bearing life criteria.
o Investigate better damping in the wheel.
Improved damping characteristics of wheel components would reduce the effect of mechanical impact
on the bearings.
o Apply special treatment for improved unbalance.
Certain measures can be introduced in the production process to reduce potential unbalance changes
under shock or vibration.
o Build wheels which
do not generate bearing noise,
compensate unbalance by themselves,
are not susceptible to external impacts.
These are the characteristics of the magnetic bearing wheel presently under development at RCD.

• To Spacecraft Builders:
o Avoid shocks in handling, transport etc.
(Same as above: Use anti-shock carts etc.)
o Accept calls for notching below the flightworthiness level.
o Mount the wheels on dampers (active / passive solutions).
Wheel dampers can be configured to mitigate the impact of external vibrations on the wheel, thereby
enabling a reduction of vibration levels in wheel production, or such as to mitigate the propagation of
wheel-induced microvibration onto the spacecraft. Due to the frequency shift of wheel modes with vi-
bration level, it can be difficult to combine both tasks in one element. (Of course, many satellite build-
ers have been working in this field already for years, see e.g. [4], [5].)
o Provide for dampers at the locations of delicate equipment.
• To Spacecraft Operators:
o Avoid shocks in handling, transport etc.
o Carry accelerometers on board to check for noisy wheel speeds.
These may differ from the speeds found under rigid mounting conditions.
o Avoid noisy operation regimes:
- select noise-limited dwell speeds based on on-board accelerometer readings.

• To everyone involved:
o More feedback and more direct contact between wheel maker’s and wheel user’s microvibration ex-
perts would be beneficial for the establishment and control of the proposed strategy towards wheel
noise mitigation.

The above recommendations do not claim completeness, nor do they want to interfere in fields where appropri-
ate actions have been established for years. They are nevertheless deemed helpful to arrive at wheels with a
lower level of self-induced microvibration for space programmes with a dedicated need of low noise. Since they
do not require new technology (with the exception of the call for a magnetic bearing wheel), there might be a
spacewheel on the market in the near future which is even electable for noise-sensitive programs like Gaia [6].

6 CONCLUSION

After the above excursion into the world of spacewheel microvibration, the following main points can be sum-
marized:
• Spacewheel microvibrations are predominantly due to
- Bearing noise,
- Unbalance, and
- Structural resonances as amplifiers,
all of which obey different but well-known rules.
• RCD’s new microvibration measurement system MVS provides a modular approach to routine measure-
ments through fully automated runs including analysis and reporting, and has extended the usable fre-
quency range from 500 to 1200 Hz.
• As a means of data presentation, waterfall plots are beautiful but ineffective. Apart from a few less de-
manding graphic alternatives, direct exchange of gigabyte-sized sets of measured data appears to be the
method of the future.
• The history of mechanical environments experienced by an individual wheel is decisive for the level of
microvibration it produces afterwards, even if these environments have been kept well below the flight-
worthiness threshold.
• The golden rule to the mitigation of spacewheel microvibration at the source is:
“To keep a wheel silent, treat it gently”.

As a result, recommendations towards noise mitigation have been given for all stages of spacewheel planning,
production, and application. Practising these recommendations appears feasible, but would require more
direct communication and feedback between all parties involved in the process.
7 ACKNOWLEDGEMENTS
The author would like to thank W. Kupferschmitt (RCD) for granting the resources needed to prepare this arti-
cle, and Dr. M. Ehinger (RCD) for his continuous encouragement to proceed in microvibration research. Special
thanks are owed to A. Hergesell (MetaDAQ) for providing and sedulously supporting a masterly measurement,
analysis and reporting tool, The MyVibSystem (MVS), which has boosted microvibration measurements from a
lab art to production maturity.

8 REFERENCES
[1] H. Heimel, “The Microvibration Characteristics of Momentum and Reaction Wheels",
Second Space Microdynamics and Accurate Control Symposium (SMACS 2), Toulouse, May 1997
[2] B. Bialke, "Microvibration Disturbance Fundamentals for Rotating Mechanisms",
34th Annual AAS Guidance and Control Conference, Breckenridge, CO, Feb. 2011
[3] R.A. Masterson, D.W.Miller, R.L.Grogan, “Development and Validation of Reaction Wheel
Disturbance Models: Empirical Model”, J. Sound and Vibration (2002) 249(3), 575-598
[4] A. Defendini et al., “Technology Predevelopment for Active Control of Vibration & Very High
Accuracy Pointing Systems”, 4th Spacecraft Guidance, Navigation and Control Systems Conference,
ESA 1999
[5] F. Boquet, F. Malric-Smith, J-P. Lejault, “Active & Passive Microvibration Mitigation System for Earth
Observation and Space Science Missions”, 34th Annual AAS Guidance and Control Conference,
Breckenridge, CO, Feb. 2011
[6] P. Chapman, T. Colegrove, E. Ecale, B. Girouart, “Gaia Attitude Control Design: Milli-Arcsecond
Relative Pointing Performance using Micro-Thrusters and Instrument in the Loop”, GNC 2011—
8th International ESA Conference on Guidance and Navigation Control Systems, Karlovy Vary, Czech
Republic, June 2011

You might also like