106 Philips tech. Rev. 34,106-111,1974, No. 4

The reaction wheels of the Netherlands satellite ANS

The attitude-control system of the ANS satellite includes a number of actuators. Theirpurpose is twofold. First, they have to stop the initial spin of the satellite and point thesolar panels for thepower supply at the Sun. Secondly, they have to point the astronomicalinstruments on board at the objects to be observed. Actuators for use in a satellite mustbe small and of low mass; at the same time, however, they have to meet the highest stan-dards of accuracy. A description of the actuators has already been given in an earliergeneral article on the attitude-control system ofthe satellite. The article below deals withthe reaction wheels in greater detail. They differ from reaction wheels previously used inspace vehicles in that a solid lubricant has been used instead of oil or grease.

J. Crucq

The mechanical actuators of the attitude-controlsystem for the ANS satellite comprise three reactionwheels [IJ. The wheels control the attitude and move-ment of the satellite by exchanging angular momentumwith it, and may be regarded as mechanical 'storagebatteries'. They should not therefore be confused withgyroscopes, as used for navigation in ships and air-craft and also in space vehicles. The axis of rotation ofeach of the three wheels coincides with one of thecoordinate axes of the satellite. Each wheel thus formspart of a control system that regulates the movementsof the satellite about the corresponding axis. As thethree systems are virtually identical, a description ofone of them should be sufficient.After the first, non-recurring operating modes im-

mediately after the launch, the attitude-control systemhas two main functions: performing the slew manoeu-vres of the satellite in scanning for a stellar object andpointing the observation instruments accurately at thatobject. During the star-pointing mode a number ofexternal disturbance torques - usually very small -have to be compensated (Table I).

Table J. Origin and magnitude ofthe external disturbance torquesacting on the ANS satellite during the flight and to be compen-sated by the reaction wheels.

< 10-4 Nm<2xl0-4 Nm

5X 10-4 Nmnegligible

Aerodynamic forcesGravitational gradientEarth's magnetic fieldRadiation pressure

After a discussion of the principle of a reactionwheel the requirements that must be satisfied by thewheels used in the satellite will be given. This will be

Ir J. Crucq is with Philips Research Laboratories, Eindhoven.

followed by a discussion of the mechanical and elec-trical design, with a brief look at the way in which thewheels are incorporated in the attitude-control systemand at the results of the life tests on the wheels.

Principle of a reaction wheel

A reaction wheel contains a body with a certainmoment of inertia, called the inertia wheel, that canbe kept in rotation by an electric motor. Since to everyaction there is an equal and opposite reaction, thesatellite can 'react' against the moment of inertia of thewheel. The resultant torque T is given by the equationof motion for rotations: T = lw, where l is the mo-ment of inertia of the inertia wheel about its axis andw is the angular acceleration of the wheel.The attitude of the satellite can be changed by the

successive acceleration and slowing down of the wheel(optimal switching or 'bang-bang control', see fig. 1).After a change of attitude the satellite returns to thestationary state and the inertia wheel rotates again atthe original speed. Frictional forces in the bearings areinternal forces in the system, and thus have no effecton this process.When a disturbance torque Td acts on the satellite, it

follows from the law ofthe conservation ofmomentumthat:

laatWaat + lrwlWrwl = jTd dt,

where quantities with the subscript sat relate to the satel-lite and those with the subscript rwl to the reactionwheel. By appropriately accelerating and slowing downthe reaction wheel it is possible to keep Waat equal tozero in spite of the presence of disturbance torques. If

Philips tech. Rev. 34, No. 4 ANS REACTION WHEELS 107

the compensation of a disturbance torque constantlyacting in one direction should cause the reaction wheelto rotate too fast, the wheel can be 'discharged' bymeans of the magnetic actuator or 'torquer' [2]. Theinteraction between the torquing coils and the Earth'smagnetic field produces a variable torque that slowsdown the speed of revolution of the wheel. In this waythe torquing coils transfer angular momentum to theEarth.

Design specifications

To perform reasonably rapid slew manoeuvres areaction wheel must be capable of delivering a torqueof 10-2 Nm (even then a slew of 60° takes about 2 or3 minutes). On the other hand the torque must beaccurately controllable for fine pointing. The smallesttorque value that can be generated by a reaction wheelis therefore tx 10-3 Nm.

The attitude-control system of ANS works with a samplingfrequency of 1 s-1, which means that in every second there is adetermination of the torque to be delivered by the reaction wheelduring the next second. A 5-bit control command is used, one bitdetermining the sign of the torque and the other four the magni-tude so that for a maximum torque of 10-2 Nm the minimumtorque is +x 10-3 Nm. Since this division is too coarse there isin addition a 'quarter-second mode', in which the desired torqueis delivered for only a quarter of a second. Averaged over onesecond this corresponds to the torque of t X10-3 Nm men-tioned earlier. To obtain a torque of sufficient magnitude withsuch a fine subdivision, the circuit is designed to permit duplica-tion of the most significant bit of the command. For the fine sub-division this gives a maximum torque of 21 X 10-3 Nm percommand.

w

ol Ir I

r ol ~

al

SJ

~1t

Fig. 1. A change in the angular position Q of one of the axes ofthe ANS satellite effected by the appropriate reaction wheel. Theupper two graphs show the corresponding variation of theangular acceleration cV and the angular velocity w of the reactionwheel. Since the moment of inertia of the satellite is about 104times greater than that of the reaction wheel, the movement ofthe satellite will be 104 times slower than that ofthe wheel.

Computer simulations of the movements to beperformed by the satellite have shown that the reactionwheel in the attitude-control system we have adoptedmust be capable of storing an angular momentum of0.6 Nms. An important factor here is that the transferof angular momentum to the Earth with the magnetictorquer, which normally begins at a speed of only480 rev/ruin, may sometimes be very poor because ofan unfavourable position of the magnetic coil withrespect to the Earth's magnetic field. The speed of awheel can therefore become temporarily fairly high.The maximum speed of a wheel should not be too

high, to ease the load on the bearings and also to limitthe gyroscopic forces exerted by the wheel when tiltingabout its axis, since these forces produce disturbancetorques that have to be compensated by the other twowheels. The maximum speed was therefore set at 200rad/s (about 2000rev/mm). This speed, and the angularmomentum of 0.6 Nms mentioned earlier, fix the mo-ment of inertia of the wheel at 3X 10-3 kgm",An attitude-control system for a satellite using a

reaction wheel as an actuator may operate in one oftwo ways. In the one case there is a direct commandto deliver the torque needed to change the attitude,irrespective of the speed, the direction of rotation orthe sign of the acceleration of the wheel. In the othercase the speed of the wheel is controlled and a torqueis obtained by requesting a change in the set value ofthe speed controller.The first system is the one we hare adopted for ANS.

The wheel is driven by a d.c. motor with a permanent-magnet rotor. If the stator current is held constant thetorque of such a motor is independent of the speed andis proportional to the stator current. To obtain thedesired torque it is therefore necessary to set the statorcurrent to a value corresponding to the torque.

The only complication now is that the frictional torque of thebearings and the hysteresis losses of the motor appear in theattitude-con trol loop in the form of disturbance torques. Ifwe hadchosen the other alternative, the frictional torque would no lon-ger have been a disturbance in the attitude-control system.Instead, however, the speed of the wheel would have had tobe accurately controlled over the whole range from 0 to 2000rev/ruin, which would have been far from simple.

Apart from the requirements mentioned abovethere are various 'interface' requirements that a reactionwheel for a satellite will have to satisfy. These includein the first place the volume and weight requirementsand compatibility with the digital control system of

[1) P. van Otterloo, Attitude control for the Netherlands astrono-mical satellite (ANS), Philips tech. Rev. 33, 162-176, 1973(No.6).

[2) See page 165 of the article of note [1].

108

the satellite. The wheels must of course be capable ofwithstanding the forces and vibrations set up duringthe launch, and the mechanical operation should notbe affected by the temperatures and low pressure« 10-3 Pa or about 10-5 torr) prevailing in the satel-lite during the mission. Finally, and this is no easyrequirement to meet, no materials can be used thatmight cause contamination of the lenses and mirrors ofthe astronomical instruments on board. This meansthat only solid lubricants can be used for the movingparts. If this is impossible or undesirable, all oil- orgrease-lubricated parts must be contained in a her-metically sealed housing. Materials for purposes suchas electrical insulation must be specially proved andreliable plastics, such as PTFE (polytetrafluorethylene,'Teflon'). Alloys containing cadmium or zinc (such asbrass) must on no account be used, because of theirhigh rate of evaporation.

Mechanical and electrical design

The general design of the reaction wheel can be seenfrom the cross-sectional drawing in fig. 2. An alumi-nium frame contains the motor and the printed-circuitboards with the control electronics. The shaft project-ing out of the housing carries the inertia wheel, whichis completely enclosed by a dust cover. These com-ponents, apart from the dust cover, are shown infig. 3.

The rotor consists of a number of magnetized ferrox-dure segments. They are fastened together by adhesiveto produce a four-pole rotor whose reluctance torquehas an amplitude of only 6 X 10-4 Nm ('skewed poles').The reluctance torque is angle-dependent and has aperiod of one-eighth of a revolution; the torque hasa very small effect at low speeds only. The stator cur-rent is commutated electronically. In addition to the. advantage of eliminating brush friction and wear, withthe associated dust contamination, the electronic com-mutation has the advantage of delivering signals thatcan be used to determine the speed of rotation [31. Thecircuit for controlling the reaction wheel, details ofwhich will be discussed presently, is mounted on threeprinted-circuit boards located at the base ofthe housing.

The inertia wheel consists of a spoked aluminiumwheel with a fairly broad stainless-steel rim. The rimis profiled in such a way that the centre of mass of thewheel lies between the bearings, thus minimizing thesensitivity to lateral shock. It was nevertheless foundnecessary to provide the housing with à plastic bufferstrip opposite the lower edge of the wheel, to preventdamage from any tilt of the wheel that might occurduring the launch. (The stiffness of the assembly is atits lowest for such movements: the resonant frequencyis 160 Hz.) Fig. 4 shows the complete reaction wheel.

J. CRUCQ Philips tech. Rev. 34, No. 4

Fig. 2. A cross-section of a reaction wheel. W rotating inertialmass, the inertia wheel. B bearings. F frame housing a brushlessd.c, motor M and printed-circuit boards E carrying the controlelectronics. The components of the electronic commutator areshown shaded. H dust cover. The rim R has 240 perforations and .is used for accurately measuring the speed and acceleration ofthe wheel during tests.

The stiffness of the bearing has to be high, so thatthe resonances fall at fairly high frequencies (350 Hz foraxial movements, 600 Hz for transverse movements),where the vibrational load that can arise during thelaunch is less than 5 g. For the shaft-bearing system wehave therefore chosen a pair of angular-contact ball-bearings with a fixed pre-load (fig. 5). These bearingsare small and very light, readily permitting rotation ineither direction at widely different speeds.

To enable the pre-loading of the bearings to beaccurately adjusted by means of two spaeer bushes ofslightly different lengths, the two bearings are mountedon the same side of the motor. Temperature differencesin the construction, especially between the two spaeerbushes, cause a change in the pre-loading and thus inthe bearing friction as well. It is therefore importantto keep the temperature differences as small as possible.For this reason as many components as possible aremade matt black to give good heat exchange by radia-tion.It did not seem desirable to put the reaction wheel in

a hermetically sealed enclosure, since this would makethe wheel less accessible for measurements before thelaunching and would in addition involve a weightpenalty, quite apart from the sealing problem itself. Wetherefore decided on dry lubrication for the reactionwheel - the first time it has ever been used for thispurpose in a space vehicle.

[3) The design of this motor is due to W. Radziwill and K. W.Steinbusch, who are with Philips ForschungslaboratoriumAachen GmbH, Aachen, The principle of the commutationis discussed by W. Radziwill in Philips tech. Rev. 30, 7, 1969.

Philips tech. Rev. 34, No. 4 ANS REACTION WHEELS 109

Fig. 3. Components of the reaction wheel; from left to right: themotor with bearing, the housing and the inertia wheel.

Fig.4. The reaction wheel. The whole structure is 150 mm inheight and in diameter, and weighs 2795 g. The connector blockhas an extension cord (the 'pin-saver'), which is used during pre-flight testing to protect the connector pins in the wheel frombeing damaged by repeated plugging and unplugging.

Fig. 5. The shaft bearing of the reaction wheel. The angular-contact bearings are given a fixed pre-laad by making the outerspacing bush 3 to 5 fLm longer than the inner one. The pre-loading eliminates play in the bearings and gives the wholestructure great axial and radial stiffness.

Lubrication of the ball-bearings

Although some general information on the dry lubri-cation of ball-bearings can be found in the literature,there are no details relating to the application we hadin mind for the ANS reaction wheels, in which thebearings are required to carry an inertia wheel with aweight of about 1 kg. The characteristic features of ourapplication are that the wheels are required to operateboth in air and in vacuum, and at ternperatures varyingbetween -20 oe and +50 oe. The wheels must alwaysbe capable of rotating in either direction at varyingspeeds and accelerations. The required life in air is1000 hours, followed by at least 2 X 108 revolutions invacuum. This corresponds to six months of normaloperation of a reaction wheel in the ANS satellite.

On the basis of the available data and our ownexperience with the first series of reaction wheels, itwas decided, in consultation with our American ad-visers, to use a composite of MoS2 and PTFE, to givethe same kind of lubrication as used in the horizonsensor. Problems to be solved for this application werecleaning the ball-bearings, the method of applying thelubricant, running in the bearings and the choice of thepre-laad for the angular-contact bearings.

The lubricant was applied by making the bearingcages of 'Duroid' 5813 (70% PTFE, 15% MOS2,reinforced with 15% fibreglass); the design of thebearings is illustrated in .fig. 6. To avoid damage tothe lubricant film, the pre-laad on the bearings iskept relatively low, at a value of 4 N instead of thevalues between 10 and 30 N used for bearings of thissize in normal operation. The frictional torque, whichis directly related to the pre-laad, therefore remainsrelatively low (1-1.5 X 10-4 Nm).

The friction of the balls against the cage causes theballs and the tracks of the bearing to become graduallycoated with a thin film of lubricant. This film reducesthe friction in the bearing and prevents cold weldingbetween the balls and the tracks, which could damagethe bearing surfaces and cause the bearing to seize up.

Fig. 6. Diagram of a ball-bearing for the reaction wheel. Thecage containing the balls is made from 'Duroid' 5813 (70%PTFE, IS % MOS2 and 15% fibreglass for reinforcement); thecage is shown as thinner than it is in reality.

110

The friction in a bearing lubricated in this way isindependent of the speed of rotation, unlike an oil-lubricated bearing, in which friction does depend onspeed because of the viscous nature of the lubricant.The frictional torque of a well run-in dry-lubricatedbearing is only 0.5 to 0.75 X 10-4 Nm, whereas thesame bearing lubricated with a small quantity of low-viscosity oil has a frictional torque of 6 X 10-4 Nm. Thedry-lubricated bearing does however run rather noisily.

A dry-lubricated bearing of this type will run in airor vacuum. One strange effect that we have noticed,but cannot as yet explain, is the following. When air isadmitted to a bearing after it has run satisfactorily forsome time in vacuum, the friction initially increases,fluctuating very strongly in amplitude with time, thengradually returning to the original situation. When thebearing is again evacuated, the whole process repeatsitself, as can be seen in jig. 7.

Electronic control

Control signals for the reaction wheels can comeboth from the attitude-control logic (ACL) and fromthe onboard computer (OBC) [11. The ACL providesa coarse attitude control and can only choose fromthree torque values (maximum positive, maximum neg-ative and zero). This coarse control always takespriority over the fine attitude control, which is effectedthrough the onboard computer.

We shall now look at the operation of the controlelectronics with the aid of the block diagram in jig. 8.The priority logic 1 determines which of the two in-coming commands is to be carried out, and sends tothe digital-to-analog converter 2 a signal specifying themagnitude of the torque to be delivered. At the sametime it sends to the circuit 4 a signal specifying the signof the torque. The analog signal for the requiredmagnitude ofthe torque is compared in the comparatorcircuit 3 with the average stator current i«. If neces-sary, switch 5 is then operated, which ensures that thecurrent has the appropriate mean value. The electroniccommutator of the motor 6 delivers pulses that reportthe times of commutation to the circuit 4, which in itsturn controls the phase of the current pulses sent to themotor by switch 5 in accordance with the required signof the torque to be delivered by the motor.

The commutator pulses also serve to produce adigital speed signal in the circuit 7. From this the coun-ter 8 forms a digital word once every second, which issent as housekeeping information to the ground stationvia the telemetry system of the satellite. The same speedinformation is passed via the discharge logic 9 to thesystem that controls the magnetic torquer, which hasto reduce the speed of the reaction wheel if it becomesexcessive.

J. CRUCQ Philips tech. Rev. 34, No. 4

III

wI5IIIII

.. : J

20 30h

Fig.7. Bearing friction W as a function of time t. During theintervals I and III the bearing has run in vacuum, during intervalII in air.

ACLaBC

T

MCL

TM

Fig. 8. Block diagram of the control electronics. The input signalsmay come either from the attitude-control logic (ACL) or fromthe onboard computer (OBC). 1 priority logic. 2 digital-to-analogconverter. 3 comparator. 4 circuit for controlling the sign of thetorque. 5 circuit for controlling the magnitude of the torque.6 motor. 7 digital speed-signal generator. 8 counter. 9 dischargelogic. ist stator current. T generated torque. MCL output signalthat sets the magnetic torquer in operation. TM output signaltransmitted via the telemetry system as housekeeping data to theground station.

The control-logic system consists of integrated cir-cuits, which take very little power; the whole circuittakes no more than 65 mA at a supply voltage of 5 V.The supply voltage for the motor is 20 V; the currentrequired varies between 7 and 122 mA depending onthe speed and the torque to be delivered.

Life tests

The usefullife ofthe satellite is determined to a largeextent by the life of the reaction wheels, and here thelife of the bearing plays the most important part. Fora useful life of six months the minimum requirementto be met by the bearing is a total of 2 x 108 rev-olutions as stated earlier, some in air but mostly invacuum. The estimated number of reversals in direc-tion of rotation is 6500. The temperature is allowed tovary in the range from -20 to +50 "C,

Life tests on three prototype wheels using dry lubri-cation with MOS2 gave good results. It was therefore

Philips tech. Rev. 34, No. 4 ANS REACTION WHEELS III

finally decided to use this kind of lubrication, althoughit had never previously been used for reaction wheels.The life tests were concluded after 5X lOs revolutions(2i times the required life).Life tests are currently being performed with four

wheels in the definitive design. By the end of April,1974, these wheels had completed more than 109revolutions without showing any defect and withoutthe frictional torque rising to a value that could indicateundue wear.The results of the life tests have given a good impres-

sion of the reliability of the construction and the meth-od of lubrication, and all the indications are that thisdesign of wheel will give satisfactory service in theANS satellite.

Summary. The ANS satellite has three reaction wheels,which areused for changing the attitude of the satellite on commandsreceived from the attitude-control system. Furthermore, tomaintain a particular attitude accurately, external disturbancetorques can be compensated by the exchange of angular momen-tum between the satellite and the wheels. A maximum torque of10-2 Nm can be delivered to ensure that slew manoeuvres areperformed in a reasonably short time. The minimum torqueis iX 1O~3Nm. The angular momentum is kept below themaximum value of 0.6 Nms by the magnetic torquer, whichreduces the angular momentum of the wheel in good time byinteraction with the Earth's magnetic field.The wheel is driven by a d.c. motor with a permanent-magnet

rotor and electronic commutation. The lubricant used is a com-posite of MoS2 and PTFE ('Teflon'). This dry lubrication hasremoved the necessity for hermetic sealing, which is necessarywith conventional oil lubrication to prevent contamination ofthe optical system. The dry lubrication works well both in airand in vacuum, and at temperatures varying between -20 and+50 °C. In life tests on four prototype models a life of at least109 revolutions has been recorded, which is more than five timesthe required life.

112 Philips tech. Rev. 34, No. 4

Recent scientific publicationsThese publications are contributed by staff of laboratories and plants which form part ofor co-operate with enterprises of the Philips group of companies, particularly by staff ofthe following research laboratories:

Philips Research Laboratories, Eindhoven, Netherlands EMullard Research Laboratories, Redhill (Surrey), England MLaboratoires d'Electronique et de Physique Appliquée, 3 avenue Descartes,

94450 Limeil-Brévannes, France LPhilips Forschungslaboratorium Aachen GmbH, WeiJ3hausstraJ3e, 51 Aachen,

Germany APhilips Forschungslaboratorium Hamburg GmbH, Vogt-Kölln-StraJ3e 30,

2000 Hamburg 54, Germany HMBLE Laboratoire de Recherches, 2 avenue Van Becelaere, 1170 Brussels

(Boitsfort), Belgium BPhilips Laboratories, 345 Scarborough Road, Briarcliff Manor, N.Y. 10510,

U.S.A. (by contract with the North American Philips Corp.) N

Reprints of most of these publications will be available in the near future. Requests forreprints should be addressed to the respective laboratories (see the code letter) or to PhilipsResearch Laboratories, Eindhoven, Netherlands.

J. Bloem: Electrical properties of flux-grown rutile(Ti02) crystals.Philips Res. Repts. 28, 596-604, 1973 (No. 6). E

P. M. Boers: Comment on determination ofthe velocityfield characteristic for n-type indium phosphide fromdipole-domain measurements.Electronics Letters 9, 134-135, 1973 (No. 6). E

J. van den Boomgaard: Preparation and some proper-ties ofmonodisperse two-phase in situ composites fromquaternary melts in the Fe-Co-Cr-C system.Philips Res. Repts. 28, 605-617, 1973 (No. 6). E

E. Bruninx: The 85Kr leak test: An improved detectionmethod.Int. J. appl. Rad. Isot. 24, 359-360, 1973 (No. 6). E

J. Cornet & D. Rossier: Properties and structure ofAs-Te glasses, I. Glass-forming ability and relatedproperties, 11. Local order parameters and structuralmodel.J. non-cryst. Solids 12, 61-84 & 85-99, 1973 (No. 1). L

J. P. Deschamps & A. Thayse: Applications of discretefunctions, Part I. Transient analysis of combinationalnetworks.Philips Res. Repts. 28, 497-529, 1973 (No. 6). B

J. Flinn: Piezoelectric ceramics.Electron (GB) No. 32, pp. 59 & 62-64, 12 July 1973. M

M. J. C. van Gemert: High-frequency time-domainmethods in dielectric spectroscopy.P~ilips Res. Repts. 28, 530-572, 1973 (No. 6).

K. H. Härdtl & D. Hennings: Wechselwirkungen zwi-schen Gefüge und Gitterstruktur in der ferroelektri-schen Mischkristallreihe PbTiOa-PbZrOa.Science of Ceramics 6, VII/I-15, 1973.

J.A. Kerr, J.A. G. Slatter & D. Vinton: AnfT anomaly.Electronics Letters 9, 338-339, 1973 (No. 15). M

D. J. Kroon: The national air pollution monitoringnetwork in the Netherlands.La Chimica e l'Industria 55, 49-52, 1973 (No. I). E

G. Le Floch & H. Arnould: Electroluminescence dansune hétérojonction ZnTe-ZnSe.Solid-State Electronics 16, 941-944, 1973 (No. 8). L

A. Milch: On the formation and thermal stability ofBi20a films.Thin Solid Films 17, 231-236, 1973 (No. 2). N

J. G. J. Peelen: Relation between microstructure andoptical properties of polycrystalline alumina.Science of Ceramics 6, XVII/I-13, 1973. E

M. J. Sparnaay, A. J. van Bommel & A. van Tooren:Auger electron spectroscopy as a tool for measuringthe diffusion of foreign atoms in solids near theirsurface.Surface Sci. 39, 251-254, 1973 (No. I). E

W. Tolksdorf & F. Welz: Über die Züchtung von gal-liumsu bstituierten Y ttrium- Eisen-Granat- Einkristallenaus schmelzfl.üssiger Lösung bei konstanter Tempe-ratur.J. Crystal Growth 20, 47-52, 1973 (No. I). H

T. S. te Velde & J. Dieleman: Photovoltaic efficienciesof copper-sulphide phases in the topotaxial hetero-

E junction copper-sulphide - cadmium-sulphide.Philips Res. Repts. 28, 573-595, 1973 (No. 6). E

A

J. A. Weaver: A research worker's view on the futureof automatic reading machines.AGARD Conf. Preprint No. 136, 13/1-8, 1973. M

Volume 34, 1974, No.4 Published 30th August 1974pages 85-112