You are on page 1of 292

Delft University of Technology

Master of Science Thesis

Design, Integration and Verification of the


Delfi-n3Xt Reaction Wheel System

Antonius G.L. Hoevenaars


Space Systems Engineering
Student number: 1215698

First supervisor: G.F. Brouwer


Second Supervisor: J. Bouwmeester

Final Version
February 23, 2012
Acknowledgements

This thesis would not have been possible without the continuous support
of my supervisors Geert Brouwer and Jasper Bouwmeester. I would like to
thank both for their great help and valuable advice throughout the entire
project and for being a great motivation all along. Geert’s extensive experi-
ence has saved me from making many mistakes and he has helped me solve
a whole range of mistakes that I did make. Jasper’s system-level view on
the project has significantly contributed to the cohesion of this work and
ensured a regular reflection on the progress.
Next I would like to thank Steven Engelen for all his helpful explanations
about the electronics that were involved and his constructive support in the
development of this project. Also I would like to thank Paul Beckers for
his help and support throughout the entire project and Chris Verhoeven for
his fresh view in urgent times. I thank Erwin Dekens for all the work he
did on this project before I took over. The final realization of the project
would never have been possible without the support of the Electronic and
Mechanical Support Division, in Dutch “Dienst Elektronische en Mechanis-
che Ontwikkeling” (DEMO) team of the Delft University of Technology, who
have supported me in every phase of the design and manufacturing.
Throughout this project there has also been significant professional sup-
port from outside the Faculty of Aerospace Engineering. In particular the
Faculty of Mechanical, Maritime and Materials Engineering who have kindly
made available their shaker hardware for the preliminary vibration test cam-
paign and ISIS and the NLR for facilitating the final vibration campaign.
Last but not least I would like thank all those around me for supporting
me all along and a big thanks to the Delfi-n3Xt team as a whole for creating
such a fruitful engineering work environment in which this thesis work was
completed.

iii
Contents

Acronyms ix

Nomenclature xi

1 Introduction 1
1.1 General Context . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Delfi Program . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Reaction Wheel System for Delfi-n3Xt . . . . . . . . . . . . . 3
1.4 Thesis Objectives . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Reaction Wheel System Design 7


2.1 Top-level Design . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Requirements and Constraints . . . . . . . . . . . . . . . . . . 8
2.2.1 General Requirements and Constraints . . . . . . . . . 8
2.2.2 Functional Requirements . . . . . . . . . . . . . . . . 8
2.2.3 Performance Requirements . . . . . . . . . . . . . . . 10
2.3 Development Process . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Motor Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Brushless DC motors . . . . . . . . . . . . . . . . . . . 13
2.4.2 Selection Criteria . . . . . . . . . . . . . . . . . . . . . 14
2.4.3 Trade-Off . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.4 Faulhaber 1202 004 BH Specifications . . . . . . . . . 15

3 Hardware Design 19
3.1 Flywheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Drive Electronics . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Motor Adaptor . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 Reaction Wheel Bracket . . . . . . . . . . . . . . . . . . . . . 27
3.5 Protective Caps . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.6 Mechanical Prototype . . . . . . . . . . . . . . . . . . . . . . 28

v
4 Dynamics and Control of the Reaction Wheel System 31
4.1 Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2 Open Loop System Response . . . . . . . . . . . . . . . . . . 33
4.2.1 Validation of Open Loop Transfer Function . . . . . . 33
4.3 Closed-Loop Controller Design . . . . . . . . . . . . . . . . . 35
4.4 System Model Discretization . . . . . . . . . . . . . . . . . . 36
4.5 Model-Based PID Control . . . . . . . . . . . . . . . . . . . . 37
4.5.1 Proportional Control . . . . . . . . . . . . . . . . . . . 37
4.5.2 Derivative Control . . . . . . . . . . . . . . . . . . . . 40
4.5.3 Integral Control . . . . . . . . . . . . . . . . . . . . . 42
4.6 Control Gains . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.6.1 Low Speed Region Gains . . . . . . . . . . . . . . . . 43
4.6.2 High Speed Region Gains . . . . . . . . . . . . . . . . 43
4.7 Simulated Dynamic Responses . . . . . . . . . . . . . . . . . 46
4.7.1 Acceleration from Zero Speed . . . . . . . . . . . . . . 46
4.7.2 Discretization of Input Command . . . . . . . . . . . . 47
4.7.3 Error in Delivered Torque . . . . . . . . . . . . . . . . 48
4.7.4 Zero Crossing . . . . . . . . . . . . . . . . . . . . . . . 49
4.7.5 Effect of Derivative Control . . . . . . . . . . . . . . . 50
4.8 Conclusions on Reaction Wheel Control . . . . . . . . . . . . 51

5 Reaction Wheel Disturbances 53


5.1 Sources of Disturbances . . . . . . . . . . . . . . . . . . . . . 53
5.2 Induced Disturbances . . . . . . . . . . . . . . . . . . . . . . 55
5.2.1 Mathematical Representation . . . . . . . . . . . . . . 55
5.2.2 Disturbance Accelerations . . . . . . . . . . . . . . . . 57
5.2.3 Disturbance Forces . . . . . . . . . . . . . . . . . . . . 60
5.2.4 Disturbance Torques . . . . . . . . . . . . . . . . . . . 60
5.3 Verification of Disturbance Model . . . . . . . . . . . . . . . . 63
5.4 Control Sensitivty to Disturbance . . . . . . . . . . . . . . . . 65
5.5 Relative Impact of Disturbance Sources . . . . . . . . . . . . 67
5.6 Conclusions on Disturbance Torques . . . . . . . . . . . . . . 68

6 Verification of Reaction Wheel System 69


6.1 Verification of General Requirements . . . . . . . . . . . . . . 69
6.1.1 Volume and Mass Budget . . . . . . . . . . . . . . . . 69
6.1.2 Environment . . . . . . . . . . . . . . . . . . . . . . . 71
6.1.3 Power Budget . . . . . . . . . . . . . . . . . . . . . . . 72
6.1.4 Housekeeping Data . . . . . . . . . . . . . . . . . . . . 73
6.1.5 Reliability Standards . . . . . . . . . . . . . . . . . . . 73
6.2 Verification of Functional Requirements . . . . . . . . . . . . 73
6.3 Verification of Performance Requirements . . . . . . . . . . . 75
6.3.1 Motor Performance . . . . . . . . . . . . . . . . . . . . 75
6.3.2 Disturbances . . . . . . . . . . . . . . . . . . . . . . . 77
6.3.3 Control and Determination Accuracy . . . . . . . . . . 77

7 Proposition for Next Generation of Design 79


7.1 Double Bearing . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.2 Simpler flywheel . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.3 Impact on Power and Dynamic Range . . . . . . . . . . . . . 82
7.4 Conclusions on Proposed Design . . . . . . . . . . . . . . . . 84

8 Conclusions 85

9 Recommendations 89

Bibliography 91

A Design Technical Note 95

B Reaction Wheel Assembly Manual 171

C Technical Drawings 187


C.1 Reaction Wheel System Components . . . . . . . . . . . . . . 187
C.2 Reaction Wheel System Assembly Tools . . . . . . . . . . . . 197

D Labview Characterization Manual 205

E Derivative Gain as Function of Damping Factor 217

F Reaction Wheel System Test Document 221

G Analysis of Controller Robustness 273

H Graphs on Relative Impact of Disturbance Sources 275


H.1 Impact of Axle Deflection Angle . . . . . . . . . . . . . . . . 275
H.2 Impact of Manufacturing Errors . . . . . . . . . . . . . . . . . 277

DELFI-N3XT REACTION WHEEL SYSTEM vii


This page intentionally contains only this sentence.
Acronyms

3ME Mechanical, Maritime and Materials Engineering.

AC Alternating Current.
ACS Attitude Control System.
ADCS Attitude Determination & Control System.

BLDC Brushless Direct Current.

COG Center Of Gravity.


COTS Commercial Off-The-Shelf.

DC Direct Current.
DEMO Electronic and Mechanical Support Division, in Dutch “Dienst Elek-
tronische en Mechanische Ontwikkeling”.

EMF ElectroMagnetic Field.

MOSFET Metal-Oxide-Semiconductor Field-Effect Transistor.

NLR National Aerospace Laboratory, in Dutch “Nationaal Lucht- en


Ruimtevaartlaboratorium”.

PCB Printed Circuit Board.


PID Proportional-Integral-Derivative.
PWM Pulse Width Modulation.

RAM Random Access Memory.


RMS Root Mean Square.
RRS Rational Rate Sensor.
RWS Reaction Wheel System.

ix
Acronyms

TRL Technology Readiness Level.

x MSC THESIS A.G.L. HOEVENAARS


Nomenclature

M
α,cog the angular acceleration vector of the flywheel center of gravity in
the body-fixed motor reference frame due to misalignments in the
reaction wheel design.
avg The percentage of error in delivered torque with respect to the com-
manded torque.
axle the maximum angle of the rotor axle with respect to the bearing axis
resulting from the radial play in the rotor axle at the bearing exit.
cog the deflection of the COG-vector with respect to the rotor axis rep-
resenting the axisymmetric error in the flywheel.
ω the rotation speed.
Ω the motor rotation vector as a function of the radial play in the rotor
axle at the bearing exit.
ζrws the reaction wheel damping factor.

M
a,cog the acceleration vector of the flywheel center of gravity in the body-
fixed motor reference frame due to misalignments in the reaction
wheel design.

C0 the motor static friction.


Cv the motor dynamic friction coefficient.

e the rotation speed error.

fadcs the control loop frequency of the central ADCS controller.


frws the control loop frequency of the RWS controller.
Ffd w the disturbance force vector acting on the flywheel center of gravity.

H,0cross The angular momentum step input on the satellite resulting from a
zero-crossing of a reaction wheel.

xi
Nomenclature

I the current.
I the moment of inertia.
IM
fw the vector containing the moments of inertia of flywheel around the
principal axes of the motor reference frame.

kE the motor back-EMF constant.


km the motor torque constant.
Kp the proportional gain.
Kp,hs the proportional gain in the high speed region.
Kp,ls the proportional gain in the low speed region.
Kscale the scaling factor to implement control gains as integers in the mi-
crocontroller.

L the terminal inductance phase to phase.

Madcs the torque commanded by the ADCS central controller.


M0 the torque acting on the body’s COG.
Md the disturbance torque generated by a reaction wheel along an axis.
Mr the friction torque internal to the reaction wheel motor.
Mfd w the disturbance torque vector acting on the motor reference frame
origin.

n the rotation speed.


n0 the rotation speed at the moment an input command is received.
n,d the speed measurement error due to disturbance torques.
nref the reference rotation speed.
nss the steady-state error of the system response.
nss,ls the steady-state rotation speed error of the motor system response
in the low speed region.
n9 ss,hs the steady-state angular acceleration error of the motor system re-
sponse in the high speed region.

R the motor terminal resistance.


rM
,0 the vector from the body-fixed motor reference frame origin to the
origin of the rotor reference frame.
rM
,cog the vector from the origin of the body-fixed motor reference frame
to the flywheel center of mass.
rR
,cog the vector from the motor reference frame origin to the rotor refer-
ence frame origin.
RM {R the rotation matrix to express a vector in the rotor reference frame
in the motor reference frame.
Rωt the time-dependent transformation matrix representing the temporal
evolution of a vector due to the rotation of the rotor.

xii MSC THESIS A.G.L. HOEVENAARS


Nomenclature

tk the discretized time.


Td the derivative time.
Td,hs the derivative time in the high speed region.
Td,ls the derivative time in the low speed region.
Ti the integral time.
Ti,hs the integral time in the high speed region.
Ti,ls the integral time in the low speed region.
Ts the Reaction Wheel System control step size.

U the voltage that is applied over the motor.


UP W M the voltage that is applied over the motor after PWM modulation.

DELFI-N3XT REACTION WHEEL SYSTEM xiii


This page intentionally contains only this sentence.
Chapter 1
Introduction

1.1 General Context


Nanosatellites are satellites in the range 1 kg to 10 kg and have become
an increasingly popular alternative over the traditional more bulky satel-
lites. Their relatively low development costs make them highly popular at
universities and also many players in the space industry have started to
adopt nanosatellite as part of a philosophy to develop more, but simpler
satellites in order to reduce overall cost and increase overall performance.
As a result of their popularity, research and development in nanosatellite
technology has also seen a steep growth and nanosatellite capabilities are
rapidly increasing. One of the desired capabilities is more precise attitude
determination and control, which is also a focus in the development of the
Delfi-n3Xt satellite of the Delft University of Technology which is scheduled
for launch in September 2012.
In attitude control a division can be made between active- and passive
Attitude Control Systems (ACSs). Passive ACS typically provide damping
of an undesired motion whereas active ACS generate a control torque to
create a desired motion. Magnetorquers are a form of active ACS that are
already implemented regularly in nanosatellites. They operate by generat-
ing a magnetic dipole which results in a control torque through interaction
with the Earth’s magnetic field. Torque can be generated continuously, but
control accuracy is limited and magnetorquers for nanosatellites typically
generate torques in the order of 107 Nm, while torques in the order of 105
Nm are required for adequate pointing and tracking. Additionally magne-
torquers are only capable of generating torque perpendicular to the local
Earth magnetic field vector. Therefore there is currently a focus on the
development of an additional actuator type for ACS using reaction wheels.
Reaction wheels are a combination of a motor and a flywheel and rely on
the conservation of spacecraft angular momentum; if a reaction wheel accel-

1
Introduction

erates in one direction, the satellite will accelerate in the opposite direction.
Reaction wheels are capable of delivering torques that are both larger and
more accurate and the delivered torque vector is not limited by external fac-
tors. Their limitation is that they have a maximum rotation speed and can
therefore become saturated. The combination with magnetorquers is an ef-
fective one, because magnetorquers can be employed to desaturate reaction
wheels when needed.

1.2 The Delfi Program


This thesis is performed in the context of the Delfi Program of the Delft Uni-
versity of Technology and this section is largely taken from J. Bouwmeester,
Delfi-n3Xt Mission Definition, Systems Overview and General Design Ap-
proach [1]. The Delfi-n3Xt mission is part of the Delfi Program, a devel-
opment line of nanosatellites for education, technology demonstration and
research & development. The Delfi Program has three general objectives:

• Educational objective. The Delfi program shall provide students


optimal preparation for careers in the space industry. This prepara-
tion includes improvement of skills in systems engineering, teamwork,
scientific writing, and communication and facilitates hands-on experi-
ence with all aspects of the development of a (small) spacecraft.

• Technology demonstration objective. Delfi satellites will perform


technology demonstration of micro-technologies for space applications,
emerging from various developments within the Dutch space sector.
These technologies can be cooperative developments on the spacecraft
bus or experiments which stand alone.

• Nanosatellite bus development objective. The Delfi program will


advance the nano-satellite platform gradually with the aim to make
very small satellites more capable for advanced technology demon-
stration, scientific or commercial purposes. There is a commonality
with the technology demonstration objective, but whereas technology
demonstration can be stand alone systems or experiments, this ob-
jective is to advance the spacecraft bus as a whole on a systematic
and consistent matter. In this way, Delfi can create spin-off for the
Dutch and international space sector for more advanced scientific or
commercial missions. A special emphasis is put on the required devel-
opments for formation flying and constellations of nanosatellites. This
will enable space applications which have not been feasible before.

2 MSC THESIS A.G.L. HOEVENAARS


1.3. Reaction Wheel System for Delfi-n3Xt

1.3 Reaction Wheel System for Delfi-n3Xt


One of the requirements for the Delfi-n3Xt nanosatellite is that it “shall
demonstrate attitude determination and control capabilities in three axes”
[1]. In the preliminary design phase of Delfi-n3Xt it was decided that a
Reaction Wheel System (RWS) would be added as a subsystem for fast
response attitude control and complement the magnetorquers. While the
magnetorquers are used to generate small torques over an extended period
of time, reaction wheels are used to generate relatively high torques over
short periods of time. Together with a comprehensive set of sensors, the
magnetorquers and reaction wheels provide Delfi-n3Xt with the required
full three-axis active control. When successfully launched, the Delfi-n3Xt
Reaction Wheel System will be among the first three-axis reaction wheel
system for nanosatellites in space. The only known existing nanosatellite
with a three-axis reaction wheel system is BeeSat, but the reaction wheels
are added for in-orbit verification and are not applied for three-axis control
[2].
Designed with full three-axis active control, Delfi-n3Xt is a big step to-
wards more advanced nanosatellite-missions in accordance with the objec-
tives described in Section 1.2. More specifically, full three-axis active control
gives Delfi-n3Xt several capabilities that extend the range of possible mis-
sions;

1. Sun-pointing of a solar array. This could yield more power, allow-


ing instruments or subsystems with a higher power consumption on a
nanosatellite.

2. Ground station tracking. Active pointing in combination with a


directional antenna results in higher data-rates than in the omnidirec-
tional case. This enables missions that yield more data.

3. Instrument pointing. When instruments can be pointed, remote


sensing of more specific areas would become possible. This would
enable remote sensing missions using nanosatellites.

4. Pointing of thrust vector. For formation flying the orbit of a satel-


lite has to be controlled which requires a propulsion system. For proper
orbit control the thrust vector must be pointed with sufficient accu-
racy.

1.4 Thesis Objectives


This thesis covers both the preliminary- and detailed design of the Delfi-
n3Xt RWS development. An earlier prototype was developed as described
in Version 1.4 of the Delfi-n3Xt Reaction Wheel System Design Technical

DELFI-N3XT REACTION WHEEL SYSTEM 3


Introduction

Note [3], but after an encountered failure during excessive vibration testing
it was decided that a full re-design was necessary. However, much of the
requirements were already derived and therefore the main objective of this
thesis is:

Design, integrate and verify the Reaction Wheel System for the Delfi-n3Xt
nanosatellite

The main objective can be expanded into the following more detailed
activities:

• Perform a re-design of the RWS leveraging on the experience with the


earlier prototype.

• Develop a new prototype to perform preliminary verification of the


re-design.

• Perform the detailed design of the RWS, including control algorithms


to command the RWS.

• Develop an engineering model based on the detailed design.

• Verify the engineering model.

Because this thesis work concerns the practical development of a subsystem


in the framework of a larger team, factors like component lead times and
developments in other subsystems had to be considered. Planning and com-
munication with parties both internal- and external to the Delfi-n3Xt team
therefore played a crucial role in the execution of this thesis.

1.5 Thesis Overview


Within the Delfi program a well-organized documentation structure exists
to manage the flow of information. For that reason separate documenta-
tion types exists, e.g. budget documents, design reports, test reports or
user manuals. Because the main goal of this thesis is the development of
a physical reaction wheel system within the Delfi program framework, the
emphasis has been on adequate documentation of activities in the format
dictated by the Delfi program document structure. To avoid multiplication
of information, in the body of this thesis report only a summary of the de-
sign, integration and verification activities is presented. All Delfi documents
that have been produced are added as appendices.
In the body of this thesis report the focus is more towards the dynam-
ics and control of the reaction wheels, because that is not well covered in
the Delfi documentation and because it more explicitly demonstrates the
academic skills of the author. Nonetheless it should be kept in mind while

4 MSC THESIS A.G.L. HOEVENAARS


1.5. Thesis Overview

reading this thesis that the practical development of the Reaction Wheel
System was the main goal of the thesis.
Chapter 2 is mostly focused on the systems engineering aspects of the
RWS development. The requirements are introduced together with the de-
velopment process and the trade-off for the motor which largely dictates the
RWS design. Chapter 3 next introduces the various separate components of
the hardware that were developed and their impact on the system. A model
of the reaction wheel system is developed in Chapter 4 together with the
controller to generate the commanded torque. The novelty of this controller
is that it generates the commanded torques using only Hall sensor measure-
ments as inputs and with Pulse Width Modulation (PWM) of the supply
voltage as the only control variable. The controller therefore fully operates
in the digital domain.
During the development of the RWS the negative impact of vibrations
induced by axle play and manufacturing imperfections was quickly realized.
Because of their potential impact on both Reaction Wheel System perfor-
mance and the performance of the satellite as a whole, it was decided to
investigate these disturbances in more detail. Chapter 5 presents a proof
for the hypothesis that manufacturing imperfections and axle play are the
major sources of disturbance.
A summary of the verification activities is presented in Chapter 6, fol-
lowed by a proposition for the next generation Delfi reaction wheel system
in Chapter 7. Conclusions and recommendations can finally be found in
Chapters 8 and 9.

DELFI-N3XT REACTION WHEEL SYSTEM 5


This page intentionally contains only this sentence.
Chapter 2
Reaction Wheel System Design

This chapter presents a summary of the system engineering activities that


were part of the development of the Delfi-n3Xt RWS. More detailed informa-
tion can be found in the Delfi-n3Xt Reaction Wheel System Design Technical
Note V2.2 which is also included in Appendix A. In this chapter first the
top-level design choices are introduced followed by the requirements for the
RWS. The top-level design is discussed first because the requirements are
partly the result of the choices made in the top-level design. Next the de-
velopment process that was applied in the re-design is introduced, followed
by the trade-off and the selection of the Faulhaber 1202 004 BH motor as a
basis for the re-design of the RWS. The chapter ends with a brief description
of the prototype that was developed as part of the design process.

2.1 Top-level Design

It was decided early in the development of Delfi-n3Xt that the reaction


wheels would be designed in-house on the basis of Commercial Off-The-Shelf
(COTS) Direct Current (DC) motors [4]. Only DC motors are considered,
because they are widely available and more compatible with Requirement
C.05 (All satellite systems shall comply with power and data bus interfaces)
than Alternating Current (AC) motors. Fully developed reaction wheels
are also available on the market, but this was not deemed compatible with
the educational objective of Delfi-n3Xt. Developing motors in-house was
another alternative approach, but this was considered too risky and time-
consuming. As such, only COTS DC motors are considered in the design
trade-off.

7
Reaction Wheel System Design

2.2 Requirements and Constraints


This section introduces the requirements and constraints for the Reaction
Wheel System. They have evolved significantly over the course of the
project. This evolution will not be described here, but only the final re-
quirements and constraints as described in the Delfi-n3Xt ADCS Top Level
Design document [4] are discussed. The requirements and constraints are
split up in general-, functional-, and performance requirements and con-
straints. For derivations of the functional- and performance requirements,
see the RWS design document that is added to this report in Appendix A.

2.2.1 General Requirements and Constraints


The general requirements that are listed in Table 2.1 are valid for every
system of Delfi-n3Xt. For derivations and more detailed discussion on these
requirements, see Bouwmeester (2010) [1].

2.2.2 Functional Requirements


Before the functional requirements are introduced, the Delfi-n3Xt RWS is
put in the appropriate context. For the RWS the context is relatively simple
and presented in Figure 2.1. Functional requirements translate functions of
the system within this context into requirements [6]. To be complete, every
link in the context diagram should at least have one requirement. The
functional requirements for the RWS are listed in Table 2.2. The first three
requirements relate to the ADCS, the fourth to satellite operations and the
fifth to the Delfi-n3Xt structural environment.

Attitude
Determination &
Control System

Reaction
Wheel System

Delfi-n3Xt
Satellite
Structural
Operations
Environment

Figure 2.1: The context diagram for the Delfi-n3Xt Reaction Wheel System.

8 MSC THESIS A.G.L. HOEVENAARS


2.2. Requirements and Constraints

Table 2.1: General requirements and constraints for all Delfi-n3Xt systems.
Identifier Description

C.01 All satellite systems shall comply with the mass budget.
120 grams is budgeted for the RWS.

C.02 All satellite systems shall comply with the volume bud-
get. Originally 90 x 90 x 19 mm3 was budgeted for the
RWS including support hardware and electronics. At
a later stage this was changed to 90 x 90 x 46.2 mm3
for the full Attitude Determination & Control System
(ADCS).

C.03 All satellite systems shall comply with the power bud-
get. Originally 400 mW was budgeted for the RWS but
as a result of the analyses made throughout this thesis
work, this requirement was later changed to 710 mW.

C.04 All satellite systems shall comply with the data budget.
This is covered in the top-level ADCS design [4].

C.05 All satellite systems shall comply with the power and
data bus interfaces. The power is supplied as 3.3V DC
and data is exchanged using the I2 C protocol.

C.06 All satellite systems shall be able to withstand the


launch environment as described in the Launch Envi-
ronment Levels document [5].

C.07 All satellite systems shall be compatible with the space


environment.

G.01 All satellite bus systems shall generate housekeeping


data when of interest to satellite operation.

G.02 All satellite bus systems shall adhere to reliability stan-


dards.

DELFI-N3XT REACTION WHEEL SYSTEM 9


Reaction Wheel System Design

Table 2.2: Functional requirements for the Delfi-n3Xt Reaction Wheel Sys-
tem.
Identifier Description

F.01 The Reaction Wheel System shall be able to store an-


gular momentum about all three principal axes.

F.02 The Reaction Wheel System shall be able to generate


torques about all three principal axes instantaneously.

F.03 The Reaction Wheel System shall be able to handle


commands from the central ADCS microprocessor.

F.04 The Reaction Wheel System shall generate reliable in-


formation about the stored angular momentum among
all three principal axes.

F.05 The reaction wheels shall not have a negative impact


on the stability of Delfi-n3Xt.

2.2.3 Performance Requirements


There should be at least one performance requirement for each functional
requirement. The derived performance requirements for the Delfi-n3Xt RWS
are presented in Table 2.3.

10 MSC THESIS A.G.L. HOEVENAARS


2.2. Requirements and Constraints

Table 2.3: Performance requirements for the Delfi-n3Xt Reaction Wheel


System.
Identifier Description

P.01 The reaction wheels shall be able to generate a min-


imum torque of 5.5  106 Nm about each individual
axis.

P.02 The reaction wheels shall be able to hold angular mo-


mentum over a dynamic range spanning 1.56  103 Nms
about each individual axis.

P.03 The reaction wheels shall not generate systemic distur-


bance torques higher than 107 Nm.

P.04 The reaction wheels shall not generate disturbance an-


gular momenta greater than 2  107 Nms.

P.05 The Reaction Wheel System shall be able to control the


angular momentum about each individual axis with a
minimum accuracy of 2  107 Nms.

P.06 The Reaction Wheel System shall be able to determine


the angular momentum about each individual axis with
a minimum accuracy of 2  108 Nms.

DELFI-N3XT REACTION WHEEL SYSTEM 11


Reaction Wheel System Design

2.3 Development Process

This section focuses on the development process to translate the require-


ments of Section 2.2 into a final concept for the Reaction Wheel System.
The development process for the RWS re-design is visualized in a flow dia-
gram in Figure 2.2.

Delfi-n3Xt reaction
wheel requirements

NO

Appropriate
Motor options Motor selection YES alternatives
available?

NO

Motor perfor-
Preliminary design YES mance adequate?

Develop RWS prototype

Preliminary verification
through testing

Compatible with
requirements?
NO

NO
YES

Develop engineering
YES Engineering problem?
model

Verification through
testing

Compatible with
requirements?
NO

YES

Develop flight model

Figure 2.2: Flow diagram of the design process for the Delfi-n3Xt Reaction
Wheel System.

12 MSC THESIS A.G.L. HOEVENAARS


2.4. Motor Selection

2.4 Motor Selection


The motor is the most important element of the RWS. For the RWS only DC
motors are considered because this avoids unnecesary transformation of the
provided DC current. In particular only Brushless Direct Current (BLDC)
motors are considered. This section first briefly introduces the arguments
behind this choice followed by a summary of the selection criteria and the
final trade-off for the motor that is used in the RWS. A more detailed trade-
off can be found in the Delfi-n3Xt Reaction Wheel Design document included
in Appendix A.

2.4.1 Brushless DC motors


BLDC motors were developed in the 60’s because it was considered necessary
to get rid of the carbon brushes in DC motors to permit motor operation
in the vacuum of space [7]. It was already recognized that DC motors
outperformed AC motors, but the required brushes could cause sparks and
increase friction and wear. With the development of the BLDC motor,
the major disadvantages of DC motors were overcome without affecting the
advantages. General characteristics of BLDC motors are:
• High efficiency
• High stall torque
• Complex control
• Fast response
• High ratio of power over mass/size
• Longer lifetime
When integrated circuits became part of everyday life in the 70’s, the cost of
the circuitry used for amplification, commutation, and control logic reduced
drastically. With the development of miniature Hall sensors, the basis for
the modern-day BLDC motors was complete. BLDC motors are believed to
be the optimal choice for reaction wheels thanks to their linear, symmetric
torque response and their efficiency [8]. This linearity breaks down at the ro-
tation speed zero-crossing, where the motor demonstrates highly non-linear
behavior due to the static friction that needs to be overcome to accelerate
the BLDC from zero rotation speed. The friction is mainly due to bearing
friction and is typically divided into various regions using the Stribeck curve
which is presented in Figure 2.3. For the purpose of the Delfi-n3Xt the four
regions are simplified by assuming a low-speed region with speed   500 rpm
(roughly regions I-III in Figure 2.3) and a high-speed region with speeds
¥ 500 rpm (roughly region IV in Figure 2.3) as well be discussed in more
detail in Section 4.5.

DELFI-N3XT REACTION WHEEL SYSTEM 13


Reaction Wheel System Design

Figure 2.3: The Stribeck curve shows the relation between the friction torque
and speed [9].

2.4.2 Selection Criteria


In compliance with the requirements and the top-level design, only motors
are considered with the following characteristics:

• No dimensions larger than 22 mm (Requirement C.02)

• BLDC motor type (Requirement C.05)

• Available in vacuum resistant version (Requirement C.07)

• Compatible with design temperature range (Requirement C.07)

• Equipped with sensor to determine the rotation speed (Requirement


F.04)

The identified motor options are next compared on the basis of a set of
parameters that are calculated using motor characteristics that are typically
provided by manufacturers. The calculations will not be discussed here (they
are included in Appendix A for the interested reader) but a description of
the selection criteria is listed here:

• Maximum useful rotation speed. Using the power budget as in-


put, the theoretical maximum rotation speed can be determined where
the motor is capable of providing the minimum required torque of
5.5  106 Nm. The maximum budgeted power in the preliminary de-
sign phase was 400 mW, corresponding to 120 mW per motor if 10%
is budgeted for the electronics. The maximum useful rotation speed is
a direct indication of the angular momentum dynamic range.

• Flywheel diameter. With the calculated maximum useful rotation


speed the required flywheel diameter is estimated in order to comply

14 MSC THESIS A.G.L. HOEVENAARS


2.4. Motor Selection

with the required angular momentum dynamic range of 1.56  103


Nms. This diameter should not exceed the earlier introduced maxi-
mum dimension of 22 mm.

• Flywheel mass. It was determined that bronze is used for the fly-
wheels because of its high density, non-ferromagnetic behavior and
compatibility with the space environment. The resulting mass of the
flywheel can be estimated on the basis of the estimated flywheel di-
ameter.

• Combined mass of motor and flywheel. The mass budget for


the full Reaction Wheel System is 120 grams and budgeting 60% for
support structure mass, the mass budget for each motor including
flywheel is set at 16 grams.

• Indication of launch load on axis. Because the earlier design had


disintegrated during an excessive vibration campaign, special attention
was given to the expected launch load on the axis. Using the Root
Mean Square (RMS) value of the random vibrations during launch as
an indicator, the loads on the axis as a function of the specified allowed
load are estimated.

2.4.3 Trade-Off
Using the criteria introduced in Section 2.4.2 the various candidate motors
are compared. This trade-off is presented in Table 2.4. In the calculation of
the maximum useful speed for the Faulhaber 2209 T 005 BSC it was taken
into account that it has a built-in controller and the power available for this
motor is therefore 133 mW instead of 120 mW. That still proved insufficient
for this motor to deliver any torque.
From the trade-off presented in Table 2.4 it is concluded that the Faul-
haber 1202 004 BH motor is the best option for the Reaction Wheel Sys-
tem. With the available power it can effectively operate up to a high rotation
speed range and thus requires a relatively small flywheel. The other realistic
option is the Faulhaber 1608 003 BH, which has slightly bigger dimensions
than the Faulhaber 1202 004 BH. Considering the limited volume available
for the RWS and especially in the preliminary design phase, the Faulhaber
1202 004 BH is a more sensible choice. Especially when the fact is taken
into account that also additional hardware and support structure has to be
developed and additional volume will be required.

2.4.4 Faulhaber 1202 004 BH Specifications


The motor is the most important element of a RWS and its characteris-
tics therefore greatly influence the design of the full system. The impact

DELFI-N3XT REACTION WHEEL SYSTEM 15


Reaction Wheel System Design

Table 2.4: Comparison of reaction wheel motor options on the basis of basic
characteristics.
Maximum Flywheel Indication of
Flywheel Combined
Motor useful speed diameter launch load
mass [gram] mass [gram]
[rpm] [mm] [% of allowed load]

Faulhaber
29,300 18.0 6.7 7.8 50%
1202 004 BH

Faulhaber
1307 004 BH 4,550 28.9 17.3 19.4 45%
geared, ratio 6:1

Faulhaber
7,530 25.1 13.1 20.0 8%
1509 006 B

Faulhaber
19,400 19.4 7.8 11.9 4%
1608 003 BH

Faulhaber
0 - - - -
2209 T 005 BSC

Maxon
0 - - - -
EC 9.2 Flat A

Maxon
11,300 23.0 11.0 11.8 13%
EC 10 Flat A

Maxon
0 - - - -
EC 20 Flat A

Portescap
0 - - - -
22BT

[] Value not compliant with requirements.


[] Value outside the formal budget, but not so much to be entirely rejected.

16 MSC THESIS A.G.L. HOEVENAARS


2.4. Motor Selection

of the motor on the characteristics of the resulting Reaction Wheel System


was already assessed in Table 2.4. Table 2.5 introduces the most important
Faulhaber 1202 004 BH motor specifications as provided by the manufac-
turer [10]. These specification will play a key role in the development of the
RWS model and controller described in Chapter 4.
The back-EMF constant is the value of the induced voltage over the
electro coils caused by the rotating permanent magnets of the motor rotor.
This ElectroMagnetic Field (EMF) generated by the rotating stator in the
electro coils has a negative effect on the motor performance.

Table 2.5: Specifications of the Faulhaber 1202 004 BH as provided by the


manufacturer [10].
Parameter Symbol Value Unit

Terminal resistance, phase-phase R 16 Ω

Terminal inductance, phase-phase L 26  106 H

Friction torque, static C0 0.003  106 Nm

Friction torque, dynamic Cv 0.52  109 Nm/rpm

Back-EMF constant kE 0.094  103 V/rpm

Torque constant km 0.902  103 Nm/A

Rotor inertia I motor 1.25  108 kgm2

DELFI-N3XT REACTION WHEEL SYSTEM 17


This page intentionally contains only this sentence.
Chapter 3
Hardware Design

This chapter describes the hardware design of the engineering model of the
Delfi-n3Xt Reaction Wheel System. This hardware was developed in close
collaboration with DEMO. To verify the key performance requirements and
the assembly process, a prototype was developed in the early stage of de-
velopment. The protoype will not be discussed in this chapter, which only
describes the final design. More information regarding the prototype can be
found in Appendix A.
The RWS that is introduced in this chapter is among the most compact
reaction wheel systems ever developed for satellites. On top of the compact
design, the design is assembled an integrated easily thanks to a modular as-
sembly appraoch. The manual for the full assembly and integration process
is discussed in detail in Appendix B. In this chapter the various components
that are designed for the RWS are introduced individually together with
their design rationale. The full engineering model consists of the following
components, which are also indicated in Figure 3.1. The technical drawings
of all components (as well as the technical drawings of the assembly tools
that were developed to facilitate assembly and integration) are added in
Appendix C.

1. BLDC motor

2. Flywheel

3. Drive electronics Printed Circuit Board (PCB)

4. Motor adaptor

5. Reaction Wheel bracket

6. Risers

7. Protective caps

19
Hardware Design

The components introduced in this list will be individually discussed in the


remainder of this chapter. The alignment of the separate reaction wheels in
the RWS with respect to the satellite body axes is presented in Figure 3.2.

pr
otect
i
vecaps

f
lywheel

BLDCmot
or

mot
oradapt
or

dr
iveel
ect
roni
cs
PCB

r
iser
s
br
acket

Figure 3.1: The full engineering model with one of the mounted motor
subassemblies enlarged.

20 MSC THESIS A.G.L. HOEVENAARS


Figure 3.2: The alignment of the RWS with respect to the satellite body
axes.

DELFI-N3XT REACTION WHEEL SYSTEM 21


Hardware Design

3.1 Flywheel
A simple disk-shaped flywheel was designed for the Delfi-n3Xt RWS with
additional material around the axis of rotation for a better connection with
the motor axle. A simple design was chosen to minimize the risk of unbal-
ance in the manufactured flywheel caused by manufacturing imperfections.
Figure 3.3 shows a picture of the flywheel accompanied by its 3D model. The
characteristics that were verified for the produced flywheels are presented in
Table 3.1.

Figure 3.3: A picture of the Reaction Wheel System flywheel accompanied


by its 3D model.

Table 3.1: Main characteristics of the flywheel designed for the Delfi-n3Xt
Reaction Wheel System. In the calculation of the moment of inertia also
the Faulhaber 1202 004 BH rotor inertia is included.
Diameter Mass Inertia
[mm] [kg] [kgm2 ]

20 6 10 3 2.95 10 7

As was indicated in Requirement P.02, each reaction wheel should be able to


store angular momentum over a dynamic range of 1.56 10 3 Nms. Therefore,
taking the moment of inertia of the motor itself into account, the flywheel
moment of inertia around the axis of rotation must be at least:

nmax π
I zz I motor Hreq
30
30 Hreq (3.1)
I zz I motor
nmax π
Where Izz is the required flywheel moment of inertia [kgm2 ], nmax is the

22 MSC THESIS A.G.L. HOEVENAARS


3.2. Drive Electronics

maxium useful speed [rpm], Hreq equals half the required dynamic range
[Nms] and I motor is the rotor moment of inertia [kgm2 ], which for the Faul-
haber 1202 004 BH is 1.25  108 [kgm2 ]. In Table 2.4 the maximum useful
speed was calculated as 29,300 rpm. To account for design uncertainties in
the preliminary design phase, a margin of 10% was taken on the maximum
useful speed and the flywheel is designed for a maximum useful speed of
26,370 rpm. Using Equation 3.1 the required flywheel moment of inertia is:

Izz ¡ 2.7  107N ms (3.2)


The simple disk shaped flywheel that was designed and verified after man-
ufacturing has a moment of inertia Izz  2.83  107 Nms and thus complies
with the requirement on the angular momentum. The total moment of in-
ertia of the flywheel and motor rotor is therefore 2.83  107 1.25  108 
2.95  107 Nms.

3.2 Drive Electronics


BLDC motors were introduced in Section 2.4.1 and basically consist of a
permanent magnet rotor and a set of electro coil stators where the cur-
rent through the electro coils is determined by the configuration of a set of
six Metal-Oxide-Semiconductor Field-Effect Transistors (MOSFETs). Often
BLDC motors are complemented with Hall sensors, which is also the case for
the Faulhaber 1202 004 BH motor that is selected for the Delfi-n3Xt RWS. A
schematic representation of the Faulhaber 1202 004 BH BLDC motor includ-
ing Hall sensors is presented in Figure 3.4. Using the position information
from Hall sensors H1, H2 and H3, the configuration of MOSFETs T1-T6 is
determined such that electromagnets U, V and W drive the permanent mag-
net in the desired direction. Thus, a BLDC motor is controlled by applying
a voltage in combination with an appropriate MOSFET-configuration. The
switching between configurations of the MOSFETs is called commutation.
Separate commutation schemes are required for both rotation directions and
have been developed as part of the RWS. Details about the implementation
of the commutation can be found in Appendix A.
Ir. S. Engelen was responsible for the design of the electronics board that
is shown in Figure 3.5 and has a size of 13.4 x 13.4 x 4.5 mm3 . These drive
electronics includes a microprocessor which, together with the additional
required electronics and interfaces, executes the following set of tasks:

1. Perform commutation.

2. Handle I2 C communication with the ADCS central microprocessor.

3. Control motor torque on the basis of commands from the ADCS central
microprocessor. For that an accurate speed determination is required

DELFI-N3XT REACTION WHEEL SYSTEM 23


MSC THESIS A.G.L. HOEVENAARS
U
T1 T3 T5

electrocoil
H2 H3
N Z
il ele
co ctr
ctro o co
ele il
W V
T2 T4 T6
H1
Hardware Design

Figure 3.4: schematic representation of a the Faulhaber 1202 004 BH BLDC motor.

24
3.2. Drive Electronics

and the ADCS-commands need to be translated into appropriate mo-


tor voltages. The latter is discussed in detail in Chapter 4.

socketf
orFaulhaber I
2Cheader
f
lex-
ri
gidconnect
or

Pr
ogramming
header

Figure 3.5: Picture of the drive electronics PCB develop for the RWS with
the interfaces indicated.

An important part of the electronics design is the microprocessor. The


RWS is equipped with an Atmel ATmega164P microprocessor, which runs
at 8 MHz. Faster microprocessors with comparable size and characteristics
were not identified and it had been decided that this microprocessor should
have the capability to perform all required tasks, although no structured
trade-off was performed. The choice for a compact microprocessor with
relatively low computational power does have a number of impacts on the
design. The first is already mentioned and is advantageous, namely that a
more compact design is possible. Others however are less favourable and are
briefly introduced below:

• Integer calculations. To avoid intensive computing power in the


calculation, only integer calculations are used and preferably bit cal-
culations.

• Maximum 16 bit integers. To further reduce the computational


power only 8- or 16 bit integers are used. This also reduces the required
storage of variables in the limited Random Access Memory (RAM) of
the microprocessor.

• Resolution of PWM counter. The ATmega164P only has an 8-bit


PWM counter, which significantly limits the number of allowed PWM
settings. With a maximum rotation speed of roughly 25,000 rpm the
controller will play a major role in switching between PWM settings
to control the reaction wheels at a specified speed.

• RAM size. Because it is impossible to achieve the required speed


determination using only integer calculations with a bit-depth of 16
bits, look-up tables are required to translate the measured time be-
tween Hall changes into rotation speed. With only 1 kByte of RAM

DELFI-N3XT REACTION WHEEL SYSTEM 25


Hardware Design

available on the ATmega164P, the resolution of this translation varies


between 1 rpm at speeds close to zero and 350 rpm at speeds close to
25,000 rpm.

• Minimum speed measurable. To avoid undesired timer overflows


in the Hall-change timer at low speeds, a prescaler value of 64 is used
for the Hall-timer (thus operating at 125 kHz) and the speed is based
on the time between six Hall changes (corresponding to 90 rotation).
The minimum speed that can be meausered with this setting is 29
rpm.

• Speed averaging. It was introduced that the resolution of speed de-


termination at speeds close to 25,000 rpm is roughly 350 rpm. There-
fore a variable speed averaging scheme is implemented which increases
the resolution at 25,000 rpm from 350 to 4 rpm but also introduces
a delay of 0.1 seconds on the speed measurements. The number of
speed measurements considered in the averaging depends on the ro-
tation speed itself but is always applied over a power-of-two number
of measurements (i.e. 20 , 21 , 22 etc.) and is thus optimized for bit
calculations.

More information on the trade-offs and analyses involved with the micropro-
cessor and microprocessor settings can be found in the Delfi-n3Xt Reaction
Wheel Design document in Appendix A.

3.3 Motor Adaptor


The Faulhaber 1202 004 BH motor is connected to the required drive elec-
tronics that are external to the motor itself by means of a flex-rigid con-
nector. The motor must therefore be integrated with the drive electronics
and the motor adaptor provides a mechanical interface for easy integration.
Additionally it forms a support platform for the thin motor base and a load
path between the motor and the bracket, minimizing the load on the drive
electronics PCB. Figure 3.6 shows the adaptor together with its 3D model.

26 MSC THESIS A.G.L. HOEVENAARS


3.4. Reaction Wheel Bracket

Figure 3.6: A picture of the Reaction Wheel System motor adaptor accom-
panied by its 3D model.

3.4 Reaction Wheel Bracket


The main purpose of the bracket is to provide a well-alligned, rigid platform
for all three reaction wheels that make up the RWS. Figure 3.7 shows a
picture of the bracket together with several risers on which the drive elec-
tronics are mounted. The pillar-shaped extrusions at the corners together
function as a rigid base for the protective caps that are introduced in Section
3.5. The bracket was developed in close collaboration with the DEMO of
the Delft University of Technology in order to effectively design for accurate
manufacturing.

Figure 3.7: A picture of the Reaction Wheel System bracket accompanied


by its 3D model.

3.5 Protective Caps


With the experience of the earlier prototype that had disintegrated during
excessive vibration tests, protective caps were implemented whose purpose

DELFI-N3XT REACTION WHEEL SYSTEM 27


Hardware Design

is twofold:

1. Limit the allowed displacement of the flywheel and therefore also the
motor axle.

2. Contain the flywheel in the case of mechanical failure to avoid sec-


ondary damage in the satellite.

Figure 3.8 shows a picture of the protective caps for reaction wheels that
are mounted respectively parallel- or perpendicularly to the ADCS PCB
together with their 3D models.

Figure 3.8: A picture of the protective cap for reaction wheels mounted
perpendicular to the ADCS-PCB (full square) and for the reaction wheel
mounted parallel to the ADCS-PCB (square with cut off corner) accompa-
nied by their 3D models.

3.6 Mechanical Prototype


The prototype developed to perform early verification of the structural in-
tegrity of the design is shown in Figure 3.9. In the picture the bracket is
mounted on the shaker kindly made available by the Faculty of Mechanical,
Maritime and Materials Engineering (3ME) to perform preliminary verifi-
cation of the design’s structural integrity. Additionally, integration of the
protoype brought several short-comings of the design to light which were
corrected in the design of the engineering model that was discussed in this
chapter.

28 MSC THESIS A.G.L. HOEVENAARS


3.6. Mechanical Prototype

Figure 3.9: The protoype of the Reaction Wheel System developed to verify
the structural integrity and the assembly procedure, mounted on the shaker
kindly made available by the Faculty of 3ME.

DELFI-N3XT REACTION WHEEL SYSTEM 29


This page intentionally contains only this sentence.
Chapter 4
Dynamics and Control of the
Reaction Wheel System

In Chapter 2 it was explained that the motor is the most important ele-
ment of a reaction wheel system and the compact design of the Delfi-n3Xt
RWS discussed in Chapter 3 is largely possible thanks to the use of perma-
nent magnet BLDC motors. This chapter will develop a novel model-based,
discrete Proportional-Integral-Derivative (PID) controller for these COTS
BLDC motors using the rotation speed as the only state variable. BLDC
controllers dealing with the non-linear behavior of BLDC motors around the
zero-crossing have been developed by Sion et.al. (2011) [11] and Wang et.
al. (2003) [9], but are not straightforward to implement in a microcontroller.
Other controllers, like that developed by Rodriguez (2005) [12] are suitable
for microcontroller implementation but require additional hardware. In this
chapter a simple model-based discrete PID controller for COTS BLDC mo-
tors will be developed and verified. A second advantage of the controller
developed in this chapter is that is specifically designed for implementation
in the digital environment of a microcontroller.

4.1 Equation of Motion


There are two governing equations for a BLDC motor, namely Newton’s
Second Law of motion and Kirchhoff’s Voltage Law. Newton’s Second Law
of Motion states;
¸
I ω9  M0 (4.1)

where
°
I is the moment of inertia [kgm2 ], ω is the rotation speed [rad/s]
and M0 is the sum of all moments around the body’s Center Of Gravity
(COG) [Nm]. Kirchoff’s Voltage Law dictates;

31
Dynamics and Control of the Reaction Wheel System

U  RI dI
dt
L (4.2)

where U is the voltage that is applied over the motor [V], R is the motor
terminal resistance [Ω], I is the current [A]
°
and L is the terminal inductance
phase to phase [H]. For a BLDC motor, M0 is expressed by:
¸
M0  kmI  pC0 Cv  nq (4.3)
km is the motor torque constant [Nm/A], C0 is the motor static friction [Nm]
and Cv is the motor dynamic friction coefficient [Nm/rpm]. C0 is caused by
static mechanical friction of the bearings and magnetic hysteris of the stator
and is highly non-linear near the zero-crossing;
$
& C0 n¡0
C0 pnq  C0 n 0 (4.4)
n0
%
0
Cv includes viscous friction of the ball bearings as well as the Foucault
currents in the stator, caused by the rotating magnetic field of the magnet
[13]. Equation 4.1 can now be rewritten in the form;

n9  km I  pC0 Cv  nq
πI
(4.5)
30
where n is the rotation speed [rpm]. The next step is to get rid of I in Equa-
tion 4.5 because it cannot be measured or directly controlled. To achieve
that, Equation 4.2 is the starting point and is first rewritten for the specific
situation of a BLDC motor by adding the motor back-EMF constant kE
[V/rpm]:

U  RI
dI
dt
L
kE n (4.6)

Now Equation 4.6 is reordered to express I as a function of the other vari-


ables:

I  UR  kRE n  RL dI
dt
(4.7)

Next Equation 4.7 can be inserted in Equation 4.5:




n9  km  n  pC0 Cv  nq
πI U kE L dI
(4.8)
30 R R R dt
Finally, the factor depending on the derivative of I is written in an alter-
native form. For that the equation for the rotation speed is rewritten and
subsequently derivated once. The original equation for the rotation speed as
provided by Faulhaber [13] is presented first, followed by its rewritten form
and its derivative:

32 MSC THESIS A.G.L. HOEVENAARS


4.2. Open Loop System Response

n
U  RI
kE

I  U RkE n
dI
dt
  nk
9 E

R
(4.9)

Where the term containing U disappears in the derivation because U is set as


a constant in the Delfi-n3Xt RWS. Equation 4.9 can be inserted in Equation
4.8 which results in the final equation of motion for each individual Delfi-
n3Xt BLDC reaction wheel. Notice that the full system is now described
with only n and U as variables, where n is the state variable and U is the
control variable.

30km kE R 30Cv R2 30R2 C0


n9 
30kM R
30LkE km  πIR2 πIR2  30LkE km
30LkE km  πIR2
n U
(4.10)
Alternatively this equation can be rewritten in the form;



πIR
30km
 LkE
R
n9 kE
Cv R
km
nU  RC
k
0
(4.11)
m

4.2 Open Loop System Response


The open loop transfer function allows validation of the system model, be-
cause it can be compared with measurements of the response of the actual
system. The open loop transfer function of the individual reaction wheels is
derived by taking the Laplace Transform of Equation 4.11;


  
πIR
30km
 LkE
R
s kE
Cv R
km
N psq  U  RC0
km
W psq (4.12)

where W(s) is a step function. The transfer function is then:

N psq U  RC 0

H psq    πIR k
W psq
m
(4.13)
30km  LkRE
s kE Cv R
km

4.2.1 Validation of Open Loop Transfer Function


The reaction wheels are operated at 3.3 Volt and the moment of inertia I
is 2.95  107 kgm2 as described in Section 3.1. The other values can be ob-
tained from the Faulhaber 1202 004 BH Micromotor specification sheet [10].

DELFI-N3XT REACTION WHEEL SYSTEM 33


Dynamics and Control of the Reaction Wheel System

Using a specially designed LabView script, the open loop response of the en-
gineering model described in Chapter 3 can also be measured. Appendix D
gives more details about this LabView script.
Figure 4.1 includes both the simulated open loop response as well as
the measured open loop response to a step input of U  3.3 V. The val-
ues for C0 and R as specified by Faulhaber are not in accordance to the
measured values. Figure 4.1 therefore also includes the simulated response
with adjusted values to match the measured response. The experimentally
determined values for the final RWS are C0  0.026  103 Nm and R  17
Ω, i.e. a higher static friction torque and terminal resistance than specified
by Faulhaber, which specified C0  0.003  103 Nm and R  16 Ω [10]. The
value for the static friction specified by Faulhaber was theoretically incon-
sistent with their value for the friction current and is therefore considered
a miscommunication in their specification sheet. Analysis of the matched
open loop transfer function showed that it has one real pole at 0.18 and is
therefore a stable system.
Now that the system model is confirmed by measurements and the true
values for the motor static friction and the motor terminal resistance have
been established, Section 4.3 will develop an adequate controller for this
system. Although the measured response closely follows the simulated re-
sponse with the adjusted variables, the measured response is clearly less
smooth than the simulated response. These deviations from the simulated
system response are hypothesized to be the effect of imperfections in the
flywheel design and axle play of the motor, which is further investigated in
Chapter 5.

35,000

30,000
Rotation speed (rpm)

25,000

20,000

15,000

10,000
Simulated response (C0 = 0.003⋅10−3 Nm, R = 16 Ω)
5,000 Measured response
Simulated response (C0 = 0.026⋅10−3 Nm, R = 17 Ω)
0
0 5 10 15 20 25 30
Time (seconds)

Figure 4.1: Simulated and measured open loop reaction wheel response to
step input of 3.3 Volts

34 MSC THESIS A.G.L. HOEVENAARS


4.3. Closed-Loop Controller Design

4.3 Closed-Loop Controller Design


As indicated in Section 2.4.1 one of the characteristics of BLDC motors is
that they require more advanced control. PID control is the most general
form of feedback control. In the feedback vector of a BLDC controller often
both the rotation speed and the current are included [14], but since the
highly miniaturized drive electronics of the Delfi-n3Xt RWS leave no room
for components to measure the current, the feedback vector contains only the
rotation speed. This rotation speed is determined using Hall sensors which
are also required to control commutation. Thus BLDC control consists of
two elements:

1. Commutation control. As introduced in Section 3.2 commutation


is actively executed by the drive electronics. Commutation control
is required to ensure that the applied voltage is at all times applied
over the proper set of stator electro coils for effective actuation of the
motor.

2. Control available power. In general it is not desired that a mo-


tor continuously delivers maximum torque. Therefore PWM is imple-
mented over the input voltage to control the power available to the
motor.

The commutation control is not discussed here, but more details can be
found in the Delfi-n3Xt Reaction Wheel Design document added in Ap-
pendix A. This chapter will develop a controller to regulate the available
power, which on its turn controls the torque of the motor. The commonly
used PWM control is implemented so that the supply voltage and thus the
torque can be regulated. Using a typical representation of PID control [15],
the controlled voltage can be expressed as
 »t

deptq
UP W M  Kp eptq 1
Ti
epτ qdτ Td
dt
(4.14)
0

Where UP W M is the voltage that is applied over the motor after PWM
modulation, e is the rotation speed error, Kp is the proportional gain, Ti is
the integral time and Td is the derivative time. e can be expressed in terms
of the instantaneous rotation speed n and a reference rotation speed nref
using the relation

e  nref n (4.15)

The fact that the feedback vector only contains the rotation speed, has
a significant impact on the controller design because the desired control
variable is torque, which is proportional to n.
9 To facilitate torque control, a

DELFI-N3XT REACTION WHEEL SYSTEM 35


Dynamics and Control of the Reaction Wheel System

controller is therefore developed where the reference speed is incrementally


increased or decreased to reflect the desired n;
9

nref
∆n
∆t
 n0
t (4.16)
where n0 is the rotation speed at the moment an input command is received
and ∆n
∆t is the slope of the rotation speed corresponding to torque requested
by the ADCS central microprocessor:

∆n
∆t
 30M adcs
πfadcs I
(4.17)
where fadcs is the control loop frequency of the central ADCS controller and
Madcs is the torque commanded by the ADCS central controller. To design
an effective controller, the values for the respective gains will be expressed
as a function of specified design variables. To accomodate that, the term
in Equation 4.14 including Ti is omitted in first instance. The proportional
and derivative gains Kp and Td can then be calculated on the basis of the
system model represented by Equation 4.11. Ti will be added later to act
as a low-pass filter to reduce steady-sate errors in the system. Replacing U
for UP W M in Equation 4.11 with the term including Ti omitted leads to the
equation of motion for the closed loop system:


πIR
30km
 LkRE Kp Td n9

(4.18)
Kp n  Kp pnref Td n9 ref q 
Cv R RC0
kE
km km

4.4 System Model Discretization


The controller that is developed in this chapter will be implemented in the
microprocessor that is part of the drive electronics (see Section 3.2). Thus,
the controller will be implemented in a discrete environment using integer
calculations only. Additionally, in order to generate a constant torque, nref
will need to be increased/decreased linearly. With the RWS control loop
operating at a finite frequency frws , nref will change incrementally. In order
to effectively design a controller for this discrete environment, the system
model represented by Equation 4.18 is discretized. A simple and effective
approach is to use Euler’s Backward differentiation method [15]. Applied to
Equation 4.18, the result is:


nptk q  nptk1 q
πIR
30km
 LkE
R
Kp Td
Ts

(4.19)
Kp nptk q  U ptk q
Cv R
kE
km

36 MSC THESIS A.G.L. HOEVENAARS


4.5. Model-Based PID Control

1
where Ts is the RWS control loop stepsize with value [s], tk is the
discretized time and U ptk q is epxressed by
f rws

nref ptk q  nref ptk1 q


U ptk q  Kp nref ptk q  RC0
Td (4.20)
Ts k m

Next, the transfer function of Equation 4.19 is established using the Z-


Transform which is the discrete equivalent of the Laplace Transform. The
Z-Transform of Equation 4.19 is:




1 z1
πIR
30km
 LkRE Kp Td
Ts z
kE
Cv R
km
Kp N pz q  U pz q
(4.21)
Where 

Kp Td z  1
U pz q  Nref pz q 
RC0 z
km z  1
Kp (4.22)
Ts z
Finally, also nref expressed by Equation 4.16 can be expressed in the Z-
transform:

Nref pz q  n0
z ∆n z
z1 ∆t pz  1q2
(4.23)

4.5 Model-Based PID Control


This section covers the derivation of equations to determine the proportional-,
integral- and derivative gains for the RWS. Different approaches are taken
in the derivation of gains for the low speed- and the high speed region. This
decision is driven by the non-linear motor behavior around the zero-crossing.
The developed equations express the gains as function of system response
parameters, which allow calculation of the proportional- and derivative gains
based on desired system behavior. Integral gains are determined rather than
calculated through reasoning and trial- and error.

4.5.1 Proportional Control


A commonly used technique to determine proportional gains is the Final
Value Theorem in combination with a requirement on nss , the steady-state
error of the system response. The Final Value Theorem as explained by
Wescott [16] states that a signal nptk q that is zero for all k   0, has a finite
value as tk goes to infinity and has a Z-transform N pz q, the value of nptk q
as tk goes to infinity is equal to:

lim nptk q  lim rpz  1qN pz qs (4.24)


tÑ8 z Ñ1
DELFI-N3XT REACTION WHEEL SYSTEM 37
Dynamics and Control of the Reaction Wheel System

The Final Value Theorem can be applied directly to determine the actual
final value of a system response, but if e.g. a system response settles to a
steady-state velocity instead of a steady-state value, it is also possible to
use the Final Value Theorem on the derivative of the transfer function. In
the case of the Delfi-n3Xt RWS, the Final Value Theorem will be used to
determine two different gains; one for the low speed region (speeds up to 500
rpm), the second for the high speed region (speeds of 500 rpm and higher).
The reason for this division is that BLDC motors demonstrate highly non-
linear behavior near the zero-crossing as was mentioned in Section 2.3. Thus
gain scheduling is applied to design separate controllers for the non-linear
low-speed region and the linear high-speed region.

Proportional Gain for Low speed region

The low speed region is dominated by the motor static friction which needs
to be overcome before the motor can be set to move. This fact is expressed
by a step function containing C0 in the Z-transform expressed in Equation
4.22. Although additional non-linear terms are expected to play a role [9],
these are neglected in this simplified model. The Final Value Theorem is
applied to design an effective controller to deal with this static friction step
input.
Because Nref pz q, is a ramp function expressed by Equation 4.23 it is
not possible to apply the Final Value Theorem directly to Equation 4.20;
nptk q will not go to a stable value (as tk goes to infinity) when Nref pz q does
not go to a stable value (as tk goes to infinity). To determine the required
gains in the low speed region the ramp function for Nref pz q is replaced by
a step function. Therefore in the low speed region Nref pz q in Equation 4.22
is replaced by:

Nref pz q  ∆n
z
z1
(4.25)

where ∆n  ∆nTs , the increase in commanded ∆n with each time-step of


the RWS control. n0 drops out from Equation 4.23 because the focus is on
acceleration from n0  0. Replacing the ramp function by a step function
essentially means that in order to cope with the non-linear behavior in the
low speed region, the controller is designed to be more responsive than
desired. Before we continue to determine the appropriate value for the
proportional gain, Equation 4.21 is expressed in N pz q with Equation 4.22
inserted and Nref substituted by Equation 4.25.



Kp,ls Td z 1
∆n z z 1  RC z
Kp,ls 
0

N pz q  
Ts z km z 1
(4.26)
πIR
 LkR
E
Kp,ls Td 1 z 1 kE Cv R
Kp,ls
30km Ts z km

38 MSC THESIS A.G.L. HOEVENAARS


4.5. Model-Based PID Control

Applying the Final Value Theorem gives:

Kp,ls ∆n  RC0


lim rpz  1qN pz qs  km
 nss,ls (4.27)
z Ñ1 kE Cv R
km Kp,ls
Which can be rewritten to express Kp,ls as a function of nss,ls , the steady-
state rotation speed error of the motor system response in the low speed
region [rpm] and ∆n [rpm]:

Cv R RC0
nss,ls kE
Kp,ls  km
∆n  nss,ls
km
(4.28)

For the Faulhaber 1202 004 BH motor with the parameter values as in-
troduced in Table 2.5 the term in the numerator containing nss,ls will be
negligible with respect to the term containing C0 for all realistic values of
nss,ls . Therefore Equation 4.28 can be simplified to

Kp,ls  k p∆nRC

0
 nss,lsq (4.29)
m

which gives better insight in the impact of the various parameters. In order
for the system response to follow the increasing reference speed, Kp should
be positive and thus ∆n ¡ nss,ls . It is interesting to note that Kp will be
larger for a smaller value of ‘∆n ¡ nss,ls ’. Thus, for a given value of nss,ls ,
the system will be more responsiveness but less precise when a smaller value
of ∆n (but bigger than nss,ls ) is used as input to calculate Kp,ls .

Proportional Gain for High speed region


Once a reaction wheel is spinning and the static friction torque is overcome,
the instantaneous change in required torque will not be near as dramatic as
when the reaction wheel is accelerated from standstill. This allows the ramp
function to be considered again as input to the system.
The proportional gain for the high speed region can be determined by
applying the Final Value Theorem on the first derivative of N pz q. Because
of its direct relation to torque as was expressed in Equation 4.17, the input
reference speed was represented by the ramp function of Equation 4.23.
Therefore the Final Value Theorem should be applied to the first derivative
of the transfer function. The Final Value Theorem for n9 ptk q is:
 
z1
lim n9 ptk q  lim pz  1q N pz q (4.30)
tÑ8 z Ñ1 z
To apply the Final Value Theorem , first Equation 4.21 is rearranged to ex-
press N pz q explicitly. Subsequently inserting Equation 4.22 and expressing
Nref in U pz q by the ramp function of Equation 4.23 gives:

DELFI-N3XT REACTION WHEEL SYSTEM 39


Dynamics and Control of the Reaction Wheel System

 
Kp,hs Td z 1 n0 z z 1 ∆n z z
 RC z
Kp,hs pq  
0

N pz q   nss,hs
Ts z ∆t z 1 2 z 1 k z 1
m
 9
πIR
 LkE
Kp,hs Td 1 z 1  kE Cv R
Kp,hs
30km R Ts z km
(4.31)
Applying the Final Value Theorem:

 
z1 Kp,hs ∆n
lim pz  1q N pz q k ∆t
 nss,hs
9 (4.32)
z Ñ1 z E
Cv R
Kp,hs
km

Which can be rewritten to express Kp,hs as a function of n9 ss,hs , the steady-


state angular acceleration error of the motor system response in the high
speed region [rpm] and ∆n∆t [rpm], the torque expressed as the derivative of
the reference speed:

Cv R
n9 ss,hs kE
Kp,hs  km
(4.33)
∆n
∆t  nss,hs
9

It should be stressed that the resulting value for Kp,hs will be significantly
lower than the value calculated for the low speed region, because the system
can be designed much less responsive once the static friction is overcome.
This allows the system to be designed such that its response is more ‘smooth’.

4.5.2 Derivative Control


Derivative control is generally applied to slow the rate of change in the
system response and therefore acts as a damper. In the Delfi-n3Xt RWS
derivative control is applied in both the low- and the high speed region for
two reasons:

1. Consistency of delivered torque. Ideally the delivered torque


would be constant over the two-second period between ADCS com-
mands. Oscillations in the reaction wheel system response are there-
fore undesired and damping of these oscillations would thus increase
the consistency of the delivered torque.

2. Reliability of delivered torque. The RWS interprets torque com-


mands into a new desired slope of the rotation speed and in this in-
terpretation the instantaneous rotation speed of the reaction wheels is
not considered. The RWS is thus effectively open loop with respect to
the full satellite dynamics. One impact of this approach is that any
error remaining in the rotation speed with respect to the interpreted
reference speed at the moment a new command is received will be dis-
regarded and therefore propagate in the dynamics of the full satellite

40 MSC THESIS A.G.L. HOEVENAARS


4.5. Model-Based PID Control

system. It is therefore desired that the RWS response has stabilized


at the moment a new command is received such that the value of the
disregarded error will be minimal.

In the low speed region the main concern is damping the initial overshoot
in the system response, which is partly due to a delay in the speed measure-
ments at low speeds (see Section 3.2) and partly due to the static friction
that needs to be overcome before the reaction wheel starts to accelerate.
In the high speed region derivative control is applied to damp oscillations
caused by disturbances in the system. Although a BLDC motor demon-
strates highly linear behavior at high speeds (see Figure 2.3), in practice
disturbances will occur. These disturbances will not only be external but
also internal, due to the fact that the system is discrete. It was earlier
demonstrated in Section 4.2 that the modeled open loop system is stable
but to allow additional tuning of the system, derivative control is still ap-
plied. In designing the derivative control it will also be taken into account
that discrete jumps in the measured rotation speed take place due to dis-
cretization effects in the speed measurements discussed in Section 3.2, which
shall not be amplified.
To determine the derivative gain, the characteristic polynomial of the
closed loop system is determined by comparison of the closed loop trans-
fer function with standard transfer functions in the Z-transform. Using the
characteristic polynomial, the derivative gain Kp Td can be determined ac-
cording to the desired damping. First the closed loop transfer function for
the low speed region is defined:

N pz q
H pz q 
Nref pz q
(4.34)

Where N pz q is described by Equation 4.21 and Nref pz q is part of the feed-


back control:


Kp Td z  1
U pz q  Kp Nref pz q (4.35)
Ts z
The term including the friction torque is not included because it is not part
of the closed-loop system. The closed-loop transfer function therefore is:

Kp 
Kp Td z 1
H pz q   Ts z
 (4.36)
πIR
 LkR
E
Kp Td 
1 z 1
kE Cv R
Kp
30km Ts z km

We would like to determine the derivative gain such that the system sta-
bilizes before a new command is received. This ensures that the average
delivered torque over the two second period is close to the desired torque
and therefore minimizes the impact of the non-linear behavior due to the

DELFI-N3XT REACTION WHEEL SYSTEM 41


Dynamics and Control of the Reaction Wheel System

static friction. The next step is to transform the closed loop transfer func-
tion described by Equation 4.36 into the associated characteristic polynomial
such that the exponential term can be identified and related to a defined
damping factor ζrws . This damping factor ζrws is here defined as the ratio
between the rotation speed error after one second and the initial error in
the system response to overcome the non-linear static friction. One second
is choosen because it is half the time between input commands and thus
accounts for a possible delay in the measurement of the initial response to
a step input. If for example one wishes to reduce the initial error in the
system response (introduced by the non-linear static friction torque) by a
factor hundred in the second following the acceleration from zero rotation
speed, ζrws  0.01.
The relation between the derivative time Td and the introduced damping
factor ζrws is derived in Appendix E. In the appendix first the characteristic
polynomial is constructed using the techniques of partial fraction expansion
and identification of the inverse Z-transform. Next the exponential term is
rewritten such that it defines Td as a function of the damping factor ζrws
and general motor characteristics. The final relation is derived as:

{ ?ζ 

 K p1  ?ζTs q  30kπIRK
1 Ts
rws Cv R LkE
Td 1{Ts
kE Kp (E.11)
p rws km RKp m p

4.5.3 Integral Control


Including integral control in a controller allows the elimination of any steady-
state error that often remains when only proportional-derivative control is
applied. As explained in Section 4.5.2 any remaining error at the moment
a new torque command is received will propagate in the global satellite
dynamics. Integral control is implemented to prevent this.
However, integral control can also potentially lead to an oscillatory re-
sponse that could even threaten the stability of the system [17]. In Section
3.2 rotation speed averaging was introduced as a means to improve mea-
surement accuracy. Due to the speed averaging applied, the rotation speed
measurements that form an input to the feedback control are effectively de-
layed by 0.1 seconds. Because of this delay, integral control can cause a
‘resonant feedback signal’ if the integration time is not at least twice the
speed averaging time. Thus, to prevent such feedback resonance the inte-
gration time should be at least 0.2 seconds. To allow some margin, the
minimum integration time is set to 0.4 seconds.

4.6 Control Gains


In Section 4.5 the equations have been developed to calculate control gains
on the basis of a limited set of design parameters. In the derivation of

42 MSC THESIS A.G.L. HOEVENAARS


4.6. Control Gains

the equations it has been taken into account that the system is a discrete
system. However, not all discretization effects as discussed in Section 3.2 are
considered in those equations. Therefore the developed equations alone are
not sufficient to effectively calculate the appropriate gains, but some insight
in the various discretization effects is required. A MATLAB simulator was
therefore built that simulates the behavior of the microcontroller and thereby
allows reliable verification of the designed controller.
The calculated gains are not implemented directly, because for imple-
mentation in the microcontroller, gains can only be integer values. Therefore
a scaling factor Kscale was designed:
7  28
Kscale  2 3.3 (4.37)

This factor takes into account that the 0-3.3 Volt design input range is
covered by a 28 PWM range and that 16-bit signed integers are used as
control variables (thus allowing an additional scaling factor of 27 ).

4.6.1 Low Speed Region Gains


Table 4.1 lists the values of the various gains as they are calculated for
the low speed region. The table also presents the discrete values of the
gains as they are implemented in the controller using the scaling factor as
introduced in Equation 4.37. The argumentation of the determined gains
is also included although the final values did require a degree of manual
fine-tuning.
The developed equation relies on the knowledge of a set of system param-
eters. Most parameters are specified by the manufacturer of the motor and
were introduced in Table 2.5. The actual values for C0 and R were experi-
mentally determined in Section 4.2 as C0  0.026  103 Nm and R  17Ω.
The RWS controller is operated at a frequency of 25 Hz, thus the step size
Ts  215 seconds.

4.6.2 High Speed Region Gains


Table 4.2 lists the values of the various gains as they are calculated for the
high speed region. The table also presents the discrete values of the gains
and the argumentation of the determined gains after manual fine-tuning.

DELFI-N3XT REACTION WHEEL SYSTEM 43


Dynamics and Control of the Reaction Wheel System

Table 4.1: Control gains for the RWS when operated in the low speed region
(<500 rpm).

Control Value Rationale

In the calculation of Kp,ls a discretized step input of


∆n  90 and a steady-sate error nss,ls  1 are used.
The minimum speed that can be measured is 29 rpm
as was introduced in Section 3.2. However, roughly
0.5 seconds will pass before this rotation speed can
be detected. Because the reference speed is incremen-
Kp,ls 54 tally changed with time, the proportional term will
continue to grow in this period with a significant over-
shoot in the system response as a result. Therefore
the proportional controller is designed to respond to
an input of ∆n  90 rpm. The calculated value of
is 5.5  103 V/rpm, which translates to 54 with the
scaling factor applied.

It was reasoned in Section 4.5.3 that the minimum


Kp,ls integral time should be 0.4 seconds. In order to as-
Ti,ls 109
sure good responsiveness a value close to this limit is
chosen, Ti,ls  0.5.

Td,ls is calculated using Equation E.11. The only de-


sign variable in that equation is ζrws ; the damping
factor of the closed loop system over a time period of
Kp Td,ls 1
85 one second. A value of 400 was choosen, which means
Ts
that an overshoot of 400 rpm would be fully damped
within one second. The calculated value of Td,ls is 0.05
s, making pTsd,ls  85
K T

44 MSC THESIS A.G.L. HOEVENAARS


4.6. Control Gains

Table 4.2: Control gains for the RWS when operated in the high speed region
(¥ 500 rpm).

Control Value Rationale

After discretization of the input, the minimum value


for the reference speed slope is ∆n ∆t  16 rpm/s.
n9 ss,hs  14 rpm/s is used because for lower values
the system was found to be not sufficiently respon-
sive at speeds ¡ 15, 000 pm and ‘saturate’ well be-
Kp,hs 7
fore its maximum rotation speed. The resulting value
for Kp,hs calculated using Equation 4.33 is 7.2  104
V/rpm, or 7 after multiplication by Kscale . With a
steady state error of n9 ss,hs  14 rpm/s it is evident
that there is a clear need for integral control.

To compensate for the significant steady-state error


Kp,hs
Ti,hs 7 that is expected from the proportional term, an inte-
gration time of one second is applied.

To damp any oscillatory motion, derivative gain is de-


sired. But a derivative gain that is too large could am-
plify the occasional ‘jumps’ in the speed measurement
due to reduced measurement accuracy at high speeds
Kp Td,hs and cause undesired behavior. An effective trade-
Ts 1
off between damping and undesired amplification is
achieved by setting the damping factor ζrws  0.25.
This indicates that for the controller to operate ef-
fectively at high speeds, oscillations are damped rela-
tively slowly.

DELFI-N3XT REACTION WHEEL SYSTEM 45


Dynamics and Control of the Reaction Wheel System

4.7 Simulated Dynamic Responses


This section will demonstrate some characteristics system responses of the
fully developed Reaction Wheel System in both the low- and high speed
region. The considerations that were made in the controller design will be
demonstrated by means of a number of curves:

1. Ideal response. This curve corresponds to the evolution of the rota-


tion speed perfectly matching the commanded torque over time.

2. Discretized reference speed command. This curve illustrates


how the controller interprets the commanded torque into a linear, but
discrete increase (or decrease) of the reference speed.

3. Digital speed measurements. This curve shows the discrete speed


measurements that are performed in the microcontroller using the Hall
sensor counter and subsequent processing using look-up tables and
averaging. Although speed measurements are performed continuously,
the curve consists only of the values that were know to the controller
when a new PWM-setting was calculated. New PWM-settings are
calculated each time step Ts  215 s.

4. Simulated system response. How the reaction wheel is expected


to respond based on the equation of motion derived in Section 4.1 is
illustrated by this curve.

In the analyses presented in the remainder of this section, the parameter


values as introduced in Section 4.2 are used. However, several values are
expected to change over the course of the Delfi-n3Xt mission. The con-
troller should be sufficiently robust to deal with these variations. From
Faulhaber’s Technical Information document [13] and the thermal verifi-
cation tests described in Appendix F it is known that the values for the
terminal resistance R, the static friction C0 and the back-EMF constant kE
are expected to change. An analysis included in Appendix G shows that
the controller developed in this chapter remains effective within the bounds
of the expected parameter variations and the robustness of the controller is
therefore adequate.

4.7.1 Acceleration from Zero Speed


It was indicated several times that to accelerate from zero speed, the re-
action wheel motor is required to overcome the static friction before any
acceleration will be observed. Figure 4.2 illustrates the resulting closed loop
system response with PID-control implemented using the gains introduced
in Table 4.1. It can be clearly seen that when the input threshold is reached
and the reaction wheel starts to accelerate, first an overshoot occurs due to

46 MSC THESIS A.G.L. HOEVENAARS


4.7. Simulated Dynamic Responses

the delay in the digital speed measurement using Hall sensors. This over-
shoot is damped within one second after which the reaction wheel follows
the ideal response to within several rpm accuracy. Moreover, once in motion
the acceleration of the reaction wheel is nearly constant, corresponding with
a constant torque as desired.

800
Ideal system response
700 Discretized reference speed command
Digital speed measurements
600
Simulated system response
Rotation speed [rpm]

500

400

300

200

100

−100
2 3 4 5 6 7 8
Time [s]

Figure 4.2: Closed loop reaction wheel system response in the low speed
region including acceleration from zero speed

4.7.2 Discretization of Input Command


Because only integers are used, the stepwise increase/decrease of the refer-
ence speed is also an integer-value. Because the torque command is given
as a value for ∆n 5
∆t which is subsequently cut in 2 equal reference speed in-
crements, the 5 least significant bits are discarded in the value for the speed
increment. In order to compensate for this loss in command accuracy, the
controller adds double the value of the 5 least significant bits to the first in-
cremental increase of the reference speed. Double the value is added because
a new torque command is received roughly every two seconds.
The effect of this discretization is particularly evident when ∆n
∆t is small
(roughly below 50). This is vizualized in Figure 4.3, where the reference
speed regularly ‘jumps’ to a new value instead of the linear increase over
the two-second ADCS control step. This means that occasionally relatively
high torque peaks are applied on the satellite instead of the desired constant
torque. This effect is best represented by angular momentum step inputs
on the satellite with an estimated worst-case magnitude of 2.0  106 Nms.

DELFI-N3XT REACTION WHEEL SYSTEM 47


Dynamics and Control of the Reaction Wheel System

400
Ideal system response
Discretized reference speed command
200 Digital speed measurements
Simulated system response
Rotation speed [rpm]

−200

−400

−600
0 2 4 6 8 10 12 14 16 18 20
Time [s]

Figure 4.3: The effect of discretization of the incrementally changed refer-


ence speed command on the closed loop reaction wheel system response

4.7.3 Error in Delivered Torque


Figure 4.4 shows the system response for a realistic campaign where the
reaction wheel is accelerated gradually to approximately its maximum speed.
One thing that is evident is the slow deviation from the ideal response. This
is due to the fact that the reaction wheel controller maintains no history of
previous commands, but simply interprets torque commands into reference
speeds. The controller is designed to deliver the torque that is required at
that instance, irrespective of the torques it was commanded to deliver up to
that moment.
The deviation that is observed when the reaction wheels appraoch 25,000 rpm
therefore corresponds to the error in delivered torque integrated over the ap-
proximately 100 seconds up to that point. This error is roughly 1,000 rpm,
which corresponds to an average error in the delivered torque of:

avg  ∆n
∆t

 100%  4%
The origin of this error lies in the delay that exists between the moment a
certain rotation speed is reached and the moment that it is registered by the
microcontroller. This delay is caused by two factors, one of which is most
prevalent at low speeds while the other at high speeds:

1. Instanteous measurement delay. Speed is measured on the basis


of the time the rotor takes to rotate over an angle representing six Hall
changes (which is 90 [deg]), At low speeds this time is significant and
therefore the instantaneous measured speed is always slightly lower
than the actual value when the reaction wheel is accelerating.

48 MSC THESIS A.G.L. HOEVENAARS


4.7. Simulated Dynamic Responses

30,000
Ideal system response
Simulated system response
25,000
Rotation speed [rpm]

20,000

15,000

10,000

5,000

0
0 50 100 150
Time [s]

Figure 4.4: Closed loop reaction wheel system response for a campaign over
full dynamic range of rotation speeds

2. Delay due to averaging. In order to improve the measurement


accuracy at higher speeds, an increasing level of averaging is applied
as rotation speeds increase. This averaging effectively creates a delay
on the speed measurement when the reaction wheel is accelerating.

Thus, when a new command is received, a new ramp function is calculated


originating from the rotation speed corresponding to the the most recent
determined value of the rotation speed after averaging. Due to this speed
averaging applied, the new ramp function does not originate from the ac-
tual rotation speed at that moment, but from a value that is the average
over the last 0.1 seconds. Therefore the reference speed expressed by the
ramp function experiences a small, but systemic off-set. The combination
of these factors result in an error in the delivered torque of approximately
3  107 Nm.

4.7.4 Zero Crossing


Finally a scenario is considered including a zero-crossing, which are notori-
ously difficult to control. It can be seen from Figure 4.5 that the controller
clearly has some difficulties in handling the zero crossing. After an initial
acceleration at t  27 [s] the reaction wheel under consideration is first
decelerated back to zero before it finally heads off.
This effect is mainly due to the earlier discussed delay in the speed
measurement at low rotation speeds, which can be observed in the evolution
of the digital speed measurements. First the reaction wheel is accelerated to
a speed higher than the desired speed and once the controller takes notice

DELFI-N3XT REACTION WHEEL SYSTEM 49


Dynamics and Control of the Reaction Wheel System

the reaction wheel proportional- and derivative control react, resulting in


the deceleration of the reaction wheel. The reaction wheel then reaches zero
rotation speed before the controller receives a new speed measurement.

System response
1000
Rotation speed [rpm]

Ideal system response


Discretized reference speed command
500 Digital speed measurements
Simulated system response

−500
0 5 10 15 20 25 30 35
Time [s]
PWM setting
0.8
Proportional control
0.6
Control setting [−]

Integral control
0.4 Derivative control
Combined PID control
0.2
0
−0.2
−0.4
0 5 10 15 20 25 30 35
Time [s]

Figure 4.5: Closed loop reaction wheel system response for a campaign
including a zero crossing

The implicit impact of this behavior is that effectively no torque-control


is possible over the range of rotation speeds 150   n   150 rpm. This
corresponds to a jump in angular momentum of

H,0cross  I∆n30
,0cross
 9.3  106N ms
Which is equal to approximately 0.6% of the design total angular mo-
mentum dynamic range of 1.56  103 Nms.

4.7.5 Effect of Derivative Control


The deceleration back to zero at t  27 seconds in Figure 4.5 is of course
undesirable, but inevitable if the controller is designed to be responsive. It
is also interesting to observe the evolution of the PWM setting during the
zero-crossing and near t  27 seconds in the PWM-setting graph of Figure

50 MSC THESIS A.G.L. HOEVENAARS


4.8. Conclusions on Reaction Wheel Control

4.5. Derivative control seems to play a prominent role in the deceleration to


zero and this begs the question whether derivative control is desired at all.
The positive effect of derivative control is demonstrated in Figure 4.6,
which is a close-up of the time around the zero crossing. The figure includes
both graphs for the simulated system response using the developed controller
and an alternative controller with Td,ls  0 s. It can be observed that also
without derivative control the reaction wheel is decelerated back to zero after
an initial acceleration at t  27 seconds and that it is not derivative control
that causes this deceleration to zero. However, when the reaction wheel is
accelerated away from zero speed for the second time, the system reaches a
stable speed faster with derivative control implemented. This indicates that
derivative control is advantageous to the system.

400
Ideal system response
350 Simulated system response with Td,ls=0.04
Simulated system response with Td,ls=0
300

250
Rotation speed [rpm]

200

150

100

50

−50

−100
25 26 27 28 29 30
Time [s]

Figure 4.6: Close up of the zero crossing in Figure 4.5 including graphs of
the system response for both the developed controller and an alternative
controller with Td,ls  0 seconds.

4.8 Conclusions on Reaction Wheel Control


This chapter has introduced the dynamics and control of a single reaction
wheel in the Delfi-n3Xt Reaction Wheel System. The dynamics of the re-
action wheels are independent of each other and the control logic developed

DELFI-N3XT REACTION WHEEL SYSTEM 51


Dynamics and Control of the Reaction Wheel System

in this chapter is thus equally valid for each individual reaction wheel.
First a reaction wheel model was developed leading to an equation of
motion that has only the rotation speed as state variable and the voltage as
control variable. This model was validated with measurements obtained on
the actual hardware which also allowed determination of the actual values
of the non-linear static friction coefficient C0 and the terminal resistance R.
A discrete PID controller was then developed based on the system model
where control over the voltage is achieved using PWM. Because the reaction
wheels are designed to deliver torque while the state variable is rotation
speed, torque is represented by a ramp function of the reference speed.
Different approaches were considered in the determination of the pro-
portional gain for the low speed region (defined as   500 rpm) and the high
speed region (¥ 500 rpm). In the low speed region the proportional gain is
determined with the goal to effectively overcome static friction and in the
high speed region the aim is to achieve a constant acceleration.
Differential gain is included to improve stability of the system and thereby
reliability. The differential gain is determined on the basis of the closed loop
transfer function and the introduced damping factor ζrws , which represents
the fraction of an initial error remaining after one second. Integral con-
trol was included to reduce any steady state error remaining in the reaction
wheel system response.
With the developed expressions and control design principles, gains were
calculated, tuned and verified using a reaction wheel simulator that also
mimics the behavior of the microcontroller and electronics. This simulator
demonstrated that the worst-case accuracy of the delivered torque is ap-
proximated as 3  107 Nm. Next an occasional angular momentum step
inputs is expected on the satellite dynamics with a maximum magnitude of
2.0  106 Nms. During a zero crossing, the expected step input of angular
momentum is 9.3  106 Nms. Finally it has been shown that the controller
is not likely to be affected by the disturbance torques that are the topic of
Chapter 5.

52 MSC THESIS A.G.L. HOEVENAARS


Chapter 5
Reaction Wheel Disturbances

In Section 4.2 the open loop response of a single reaction wheel was discussed
and it was observed that the measured response showed deviations from the
simulated response. It was hypothesized that those observed deviations are
the result of disturbance torques in the system itself. Disturbance torques
and disturbance forces acting on the motor are of special concern because
they potentially have a great impact on other subsystems too. In this section
only periodic disturbance forces and torques from the RWS are analyzed.
In the section on the Engineering Model in Appendix A also the effect of
misalignment of reaction wheels with respect to the satellite’s principal axis
is analyzed. This misalignment does not result in periodic disturbances and
its effect is better described as torque off-set. This chapter thus only dis-
cusses the more severe, periodic disturbances torques. First the considered
causes of periodic disturbances are introduced, followed by a derivation of
their impact and analysis of the obtained results.

5.1 Sources of Disturbances


Two sources of Reaction Wheel System disturbances are investigated, which
are vizualized in Figures 5.1 and 5.2. Both have specific effects on the
dynamics of the system as described below;

1. Axial play in the rotor shaft. In the specification sheet of the


Faulhaber 1202 004 BH motor [10] it is stated that the axial radial
play in the rotor axle ¤ 0.011 mm measured at the bearing exit, which
is located 0.074 mm above the motor origin as presented in Figure 5.1.
This radial axle play results in potential deflection of the rotation
axis which will result in a moment acting on the RWS due to the
gyroscopic effect. In the analysis in this chapter, this axial play is
therefore represented by a constant deflection of the rotor axle with

53
Reaction Wheel Disturbances

respect to the bearing axis over an error angle axle . axle is vizualized
on the left in Figure 5.2 and has a maximum value:


axle,max  tan 0.011


0.74
 0.85 (5.1)

2. Flywheel manufacturing tolerances. There is no such thing as


perfect manufacturing and therefore inevitably axisymmetric errors
will be present in the system. Axisymmetric errors are caused by a
mass division in the flywheel that is not perfectly symmetric around
the rotation axis. In the analysis in this chapter, this error is repre-
sented by a deflection of the cog-vector with respect to the rotor axis
over an error angle cog . cog is vizualized on the right in Figure 5.2

ε cog

ε axle

M
Z
distance to flywheel base: 0.94 mm
distance to bearing exit: 0.74 mm

M
Y

Figure 5.1: Sources of disturbance torques internal to the Reaction Wheel


System

54 MSC THESIS A.G.L. HOEVENAARS


5.2. Induced Disturbances

R
Z M R
Z Z

ω εaxle ε cog

YR YR

rε,cog
rε,0 YM

Figure 5.2: Vizualizations of the error angles representing the effect of axial
play in the rotor shaft and flywheel manufacturing tolerances. Vector r,0
is the vector from the bearing center of rotation to the flywheel reference
frame origin, r,cog is the vector from the flywheel reference frame origin to
its center of gravity

5.2 Induced Disturbances


This section assesses how the error angles introduced in Section 5.1 impact
the RWS dynamics. Special interest goes to the resulting worst-case accel-
erations of the flywheel with respect to the origin of the body-fixed motor
reference frame. It is assumed that this point also acts as pivot point for
the motor axle.

5.2.1 Mathematical Representation


In order to assess the acceleration of the flywheel center of gravity in the
body-fixed reference frame, first the vector to the flywheel center of gravity
is expressed in the body-fixed motor reference frame:

rM
,cog  rM,0 RM {R rR
,cog (5.2)

Where rM ,cog is the vector from the origin of the body-fixed motor reference
frame to the flywheel center of mass, rM ,0 is the vector from the body-fixed
motor reference frame origin to the origin of the rotor reference frame. rR,cog
is the vector from the origin of the body-fixed motor reference frame to
the flywheel center of mass expressed in the rotor reference frame. The
rotation matrix RM {R transforms a vector expressed in the rotor reference
frame in the motor reference frame. Because the rotor reference frame is
rotated with respect to the body-fixed motor reference frame over an angle
axle (here arbitrarily chosen around the motor X-axis), the rotation matrix
RM {R is expressed as:

DELFI-N3XT REACTION WHEEL SYSTEM 55


Reaction Wheel Disturbances

 
1 0 0
RM {R   0 cos axle  sin axle  (5.3)
0 sin axle cos axle
The origin of the rotor axle reference frame is defined as the point of the
flywheel base on the motor axle. The distance along the axis from the motor
reference frame origin to this point is 0.94 mm. This is also vizualized in
Figure 5.2. The flywheel reference frame origin coincides with the rotor
reference frame origin.
The instantaneous velocity of the flywheel center of gravity with respect
to the motor origin is obtained by taking the derivative of Equation 5.2:

M
v,cog  rM,0
9 Ω  rM
,0 RM {R r9 R
,cog Ω  RM {R rR
,cog (5.4)
with Ω expressed as a function of the rotation speed ω [rad/s] and axle the
maximum angle of the rotor axle with respect to the bearing axis resulting
from the radial play in the rotor axle at the bearing exit:
 
0
Ω   ω sin axle  (5.5)
ω cos axle

Because rR,cog is a result of imperfections in the manufacturing of the fly-


wheel, the vector does not change with time. Additionally, for the remainder
of this analysis the situation is assumed where rM ,0 is a fixed vector in the
motor reference frame. This corresponds to the worst-case situation where
the rotation axis is instantly deflected over an angle axle .
In practice the angular deflection of the axle will change over time under
the influence of the generated disturbance torque, but this effect is ignored
in this analysis because the purpose of this analysis is to assess the worst-
case disturbance torques. Thus assuming that rR M
,cog and r,0 are fixed in
time within their respective reference frames, Equation 5.4 can be simplified
to:

M
v,cog  Ω  rM
,0 RM {R rR
,cog (5.6)

Equation 5.6 can be further simplified by inserting Equation 5.2:

M
v,cog  Ω  rM,cog (5.7)
The instantaneous acceleration of the center of gravity in the body-fixed
motor reference frame can then be expressed by taking the first derivative
of Equation 5.7:

M
a,cog  Ω  rM,cog
9 Ω  r9 M
,cog Ω  Ω  rM
,cog (5.8)

56 MSC THESIS A.G.L. HOEVENAARS


5.2. Induced Disturbances

Again, for simplification it is assumed that rM


,cog is fixed with time and
therefore Equation 5.8 can be simplified to:

M
a,cog  Ω  rM,cog
9 Ω  Ω  rM
,cog (5.9)
Now we have an expression for the instantaneous acceleration of the center
of gravity of the flywheel expressed in the motor reference frame for a fixed
orientation of the rotor reference frame with respect to the motor reference
frame. However, the rotor reference frame is rotating around its Z-axis with
speed ω which also affects the temporal evolution of the acceleration vector
expressed in the motor reference frame. This temporal evolution can be
represented by adding a time-dependent transformation matrix Rωt to the
expression of rM,cog in Equation 5.2:

rM
,cog  rM,0 RM {R Rωt rR
,cog (5.10)
Rωt is expressed by Equation 5.11.
 
cospωtq sinpωtq 0
Rωt    sinpωtq cospωtq 0  (5.11)
0 0 1
It should be noted that the addition of this rotation matrix was not included
earlier in the derivation because it was not required to calculate the resulting
acceleration for a given vector. Equation 5.11 is now introduced to include
the evolution of this vector over time. Inserting Equation 5.10 into Equation
5.9 gives the final equation representing disturbance accelerations of the
flywheel center of gravity:


M
a,cog Ω 
9 rM
,0 RM {R Rωt rR
,cog
   (5.12)
Ω  Ω  rM
,0 RM {R Rωt rR
,cog

With Ω expressed by Equation 5.5, RM {R expressed by Equation 5.3 and


Rωt expressed by Equation 5.11.

5.2.2 Disturbance Accelerations


Using Equation 5.12 it is possible to calculate the expected disturbance
accelerations on the flywheel. Figure 5.3 presents the resulting disturbance
accelerations of the flywheel center of gravity for the principal axes of the
motor reference frame over a period of one revolution and for a rotation
speed of 25,000 rpm (approximately the maximum useful rotation speed). In
determining these graphs, the variables and vectors introduced in Table 5.1
were used. It is thus clear that the frequency of the disturbance acceleration
is equal to the reaction wheel rotation speed.

DELFI-N3XT REACTION WHEEL SYSTEM 57


Reaction Wheel Disturbances

Table 5.1: The variables and vectors considered in the assessment of the dis-
turbance torques resulting from imperfections in the flywheel manufacturing
process and the radial axle play.

Identifier Representation Variable value(s)

cog  0.4 deg


Value is estimated on the basis of the con-
cog - centricity and total runout tolerances in
the technical drawing of the flywheel (Ap-
pendix C).

zcog  1.1  103 m


zcog -
Value as calculated by Catia.

The axisymmetric error in the flywheel is


 
0 represented by a deflection of the center
rR
,cog
  sinpcog qzcog  of gravity vector over an angle cog around
cospcog qzcog the X-axis. Ideally this vector only has a
Z-component with the value zcog .
 
0 axle,max  0.85 deg, as calculated in
rM
,0
  sinpaxleqzaxle  Equation 5.1 zaxle  0.94 mm, see Figure
cospaxle qzaxle 5.1 for a vizualization of the angles.

n  25, 000 rpm is roughly the maximum


ω nπ
30 rotation speed, thus ω  2618 rad/s
ω is used in the rotation matrix Rωt

  In theory the motor rotation ω is fully


0 around the motor Z-axis, but to assess the
Ω  ω sinpaxle q  worst-case effect of the axial play a deflec-
ω cospaxle q tion of the rotation vector over an angle
axle around the X-axis is considered.

T  5.5  106 Nm is the maximum re-


quired torque
ω9 T
I I  2.95  107 kgm2 is the combined in-
ertia of wheel and motor rotor
Therefore, ω9  18.6 rad/s.
 
0
9
Ω 
 ω9 sinpaxle q  The same reasoning applies as for Ω .
ω9 cospaxle q

58 MSC THESIS A.G.L. HOEVENAARS


5.2. Induced Disturbances

X−axis Y−axis Z−axis


60 60 1

40 40
0.5
Acceleration [m/s2]

Acceleration [m/s2]

Acceleration [m/s2]
20 20

0 0 0

−20 −20
−0.5
−40 −40

−60 −60 −1
0 1 2 0 1 2 0 1 2
time [ms] time [ms] time [ms]

Figure 5.3: Calculated components of the disturbance acceleration over a


period of one rotation based on the variables and vectors as described in
Table 5.1

Because the disturbance acceleration is periodic with a frequency equal to


the rotation speed, the amplitude of the disturbance acceleration can be
plotted as a function of rotation speed n. This is done in Figure 5.4.

X− and Y− axis Z− axis


60 1
Disturbance acceleration [m/s2]

Disturbance acceleration [m/s2]

50
0.8

40
0.6
30
0.4
20

0.2
10

0 0
0 10,000 20,000 0 10,000 20,000
Rotation speed [rpm] Rotation speed [rpm]

Figure 5.4: Amplitude of the components of the calculated periodic distur-


bance acceleration as a function of the rotation speed based on the variables
and vectors as described in Table 5.1

DELFI-N3XT REACTION WHEEL SYSTEM 59


Reaction Wheel Disturbances

5.2.3 Disturbance Forces


Newton’s Second Law of Motion can be used to calculate the disturbance
force Ffd w acting on the flywheel center of mass:

Ffd w  mf w a,cog
M
(5.13)
The mass of the flywheel is 6 grams and the acceleration of the flywheel was
expressed in Equation 5.12. The resulting components of the disturbance
force as a function of time over a period of one rotation at 25,000 rpm are
presented in Figure 5.5. Figure 5.6 presents the amplitude of the components
of the periodic disturbance force as a function of the rotation speed.

X−axis Y−axis −3
x 10 Z−axis
0.4 0.4 8

0.3 0.3 6

0.2 0.2 4

0.1 0.1 2
Force [N]

Force [N]
Force [N]

0 0 0

−0.1 −0.1 −2

−0.2 −0.2 −4

−0.3 −0.3 −6

−0.4 −0.4 −8
0 1 2 0 1 2 0 1 2
time [ms] time [ms] time [ms]

Figure 5.5: Calculated components of the disturbance force over a period of


one rotation based on the variables and vectors as described in Table 5.1

5.2.4 Disturbance Torques


With the disturbance force on the flywheel center of gravity expressed by
Equation 5.13 and the vector from the origin of the motor reference frame
(which acts as pivot point) expressed by Equation 5.10, the disturbance
torque Mfd w acting on the motor reference frame origin can be calculated:

Mfd w  I Mfw α,cog


M
(5.14)

IM
f w represents the moment of inertia vector of the flywheel around the prin-
cipal axes of the motor reference frame. For simplicity the moment of inertia
M
of the motor rotor is neglected. a,cog was already expressed in Equation 5.12
and again assuming that the origin of the motor reference frame is the pivot
M
point of the motor axle, α,cog is defined as:

60 MSC THESIS A.G.L. HOEVENAARS


5.2. Induced Disturbances

X− and Y− axis x 10
−3 Z− axis
0.4 5

0.35
4
0.3
Amplitude [N]

Amplitude [N]
0.25 3
0.2

0.15 2

0.1
1
0.05

0 0
0 10,000 20,000 0 10,000 20,000
Rotation speed [rpm] Rotation speed [rpm]

Figure 5.6: Amplitude of the components of the calculated periodic distur-


bance force as a function of the rotation speed based on the variables and
vectors as described in Table 5.1

M
α,cog  rM,cog  a,cog
M
(5.15)
The moments of inertia of the flywheel around the principal axes of the
motor reference frame are listed in Table 5.2. The resulting components of
the disturbance torque as a function of time over a period of one rotation at
25,000 rpm are presented in Figure 5.7. Figure 5.8 presents the amplitude
of the components of the disturbance torque as a function of the rotation
speed.

DELFI-N3XT REACTION WHEEL SYSTEM 61


Reaction Wheel Disturbances

−3
x 10 X−axis −3
x 10 Y−axis −5
x 10 Z−axis
1 1
1.5

1
0.5 0.5

0.5
Torque [Nm]

Torque [Nm]

Torque [Nm]
0 0 0

−0.5

−0.5 −0.5
−1

−1.5
−1 −1
0 1 2 0 1 2 0 1 2
time [ms] time [ms] time [ms]

Figure 5.7: Calculated components of the disturbance torque over a period


of one rotation based on the variables and vectors as described in Table 5.1

x 10
−4 X− and Y− axis x 10
−5 Z− axis
8 1

7
0.8
6
Amplitude [Nm]

Amplitude [Nm]

5 0.6
4

3 0.4

2
0.2
1

0 0
0 10,000 20,000 0 10,000 20,000
Rotation speed [rpm] Rotation speed [rpm]

Figure 5.8: Amplitude of the components of the calculated periodic distur-


bance torque as a function of the rotation speed based on the variables and
vectors as described in Table 5.1

62 MSC THESIS A.G.L. HOEVENAARS


5.3. Verification of Disturbance Model

Table 5.2: Mass of the flywheel and its moments of inertia around the prin-
cipal axes of the motor reference frame.
Variable Description Value

m Mass 6  103 kg

1.696  107 kgm2


Moment of inertia around
Ixx
X-axis

1.696  107 kgm2


Moment of inertia around
Iyy
Y-axis

2.83  107 kgm2


Moment of inertia around
Izz
Z-axis

5.3 Verification of Disturbance Model


To verify the calculations in this chapter, the results are compared with
measurements of the RWS. In particular the calculations are compared with
the observed torque deviations in the open loop response as presented in
Section 4.2. The deviations in the open loop response are more clear when
the measured torque is plotted as a function of rotation speed. This is done
in Figure 5.9. In the same figure also the theoretical torque curve is plotted
as well as the expected torque bounds when disturbance torques are taken
into account.
The equation for the theoretical curve is established from a set of well-
known relations, which are presented in Equations 5.16, 5.17, 5.18.

M  Ikm  Mr (5.16)

I  U RkE n (5.17)

Mr  C0 Cv n (5.18)
where Mr is the friction torque internal to the reaction wheel motor. Com-
bining these equations gives the relation between rotation speed n and de-
livered torque M :

M  U RkE n km  pC0 Cv nq (5.19)

where kE is the motor back-EMF constant [V/rpm], km is the motor torque


constant [Nm/A], C0 is the motor static friction [Nm] and Cv is the motor

DELFI-N3XT REACTION WHEEL SYSTEM 63


Reaction Wheel Disturbances

−4
x 10
1.8

1.6 Theoretical torque


Measured torque
1.4 Expected torque bounds
1.2
Torque [Nm]

0.8

0.6

0.4

0.2

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]

Figure 5.9: The measured torque delivered by a single reaction wheel in the
representative engineering model of the Delfi-n3Xt compared to the theo-
retical torque curve as described by Equation 5.19. The expected torque
bounds represent the maximum effect of the disturbance torques in the mo-
tor Z-axis.

dynamic friction coefficient [Nm/rpm]. The actual values for C0 and R had
been experimentally determined in Section 4.2, while the values for kE , km
and Cv were confirmed to be in accordance to the specified values. The
specified values can be found in the Faulhaber 1202 004 BH specification
sheet [10].
The theoretical torque curve expressed by Equation 5.19 only concerns
torque around the Z-axis of the motor reference frame. Now also theoretical
torque bounds can be determined by adding or substracting the amplitude
of the disturbance torques around the Z-axis from the theoretical torque.
In Figure 5.9 this is done for the situation where cog  0.4 [deg] as was
presented in Table 5.1 and axle  0.85 [deg] is considered as was determined
in Equation 5.1. The resulting disturbance torques (aroud the motor Z-axis)
that determine the theoretical torque bounds were determined before and
vizualized in Figure 5.8.
The measured torques in Figure 5.9 are indirectly determined from the
observed change in rotation speed around the motor Z-axis. The rotation
speed is measured by Hall sensors which are positioned in a X-Y plane and
speed measurements therefore always concern rotation speed around the
motor Z-axis, independent of any possible deflection of the actual axis of
rotation due to radial axle play. Direct comparison of the calculated torque
bounds with the measurements is thus allowed because all torques are either
calculated or measured around the motor Z-axis.
It is acknowledged that the torque measurements are not perfectly accu-

64 MSC THESIS A.G.L. HOEVENAARS


5.4. Control Sensitivty to Disturbance

rate, because they are indirectly determined from speed measurements. But
since the observed deviations are significant and improving the measure-
ment accuracy did not result in any reduction in the observed deviations,
it is concluded that the observed deviations are in fact the result of distur-
bance torques in the system itself. More information on the developed test
set-up can be found in the Reaction Wheel Characterization Manual which
is included in Appendix D.

5.4 Control Sensitivty to Disturbance


In Chapter 4 the control logic has been discussed. The disturbances that
were analyzed in this chapter could have a significant impact on the con-
trol performacne. A quick analysis in this section will show that despite
their significant magnitude, these disturbances are not likely to affect the
performance of the controller.
The disturbances have a frequency equal to the instantaneous rotation
speed. The rotation speed is based on the time between six Hall changes
(Section 3.2) which corresponds to a quarter rotation. Speed measurements
are therefore performed with a frequency four times the rotation speed. The
impact of the disturbance on the measured speed will therefore completely
level out if the average of four speed measurements is considered.
Speed averaging is applied as explained in Section 3.2 and the number
of speed measurements that are averaged is always a power of two. Thus,
for rotation speeds where the averaging factor is ¥ 22 , the disturbances are
completely leveled out. Information regarding microcontroller implementa-
tion can be found in the Delfi-n3Xt Design Document added in Appendix A
reveals that this is the case for rotation speed ¡ 2, 690 rpm. This fact is also
vizualized in Figure 5.10, where the average of four speed measurements is
equal to the average value. At speeds lower than 2, 690 part of the distur-
bance will propagate in the speed measurements. Although it is not clearly
visible in Figure 5.7, in speed region   2, 690 rpm, the maximum expected
disturbance torque Md around the motor Z-axis is   1  107 Nm.
To assess the impact of the disturbance torque when it is not leveled out
with speed measurement averaging, this torque is translated in the corre-
sponding maximum speed measurement error. First the disturbance torque
is converted in the corresponding acceleration which is then integrating over
half a period. The integration period is the gray area in Figure 5.10 and
results in a maximum variation n,d of:
»π
n,d  30Md
πI
sinpθqdθ (5.20)
0

With Md as introduced above and I=2.95  107 kgm2 , the expected max-
imum disturbance in the speed measurement n,d  6.5 rpm. At speeds

DELFI-N3XT REACTION WHEEL SYSTEM 65


Reaction Wheel Disturbances

Figure 5.10: The top sketch illustrates the variation in the acceleration
caused by the disturbance torque discussed in Chapter 5, while the bottom
sketch illustrates the resulting variation in rotation speed. With the circles
representing speed measurements, it is clear that with averaging of ¥ 22
measurements the net effect is zero.

66 MSC THESIS A.G.L. HOEVENAARS


5.5. Relative Impact of Disturbance Sources

  2690 rpm there is therefore a potential variations of 6.5 rpm in the speed
measurement. Because the controller is not designed to react fast to such
small inputs, the impact is not expected to be significant.

5.5 Relative Impact of Disturbance Sources


Throughout this chapter both the axial play in the rotor shaft and manufac-
turing errors have been considered as sources of disturbance. The magnitude
of both disturbance sources was estimated in Table 5.1. In this section the
the relative impact of these sources is briefly discussed. Appendix H contains
the graphs that support the arguments in this section.
In the derivation of Equation 5.12 also the acceleration of the flywheel
due to motor actuation was included with the factor ω. 9 A quick inspection
of e.g. Figure 5.8 shows that this acceleration has negligible impact on the
reaction wheel disturbance, because all disturbances virtually go to zero as
the rotation speed approaches zero. This leads to the conclusion that the
rotation speed dictates the magnitude of the periodic disturbances. In other
words, the frequency and the amplitude of the disturbance torques are fully
correlated.
Changing the value for the angle of deflection of the motor axle only
impacts the disturbance in the motor reference frame Z-direction. The cal-
culated angle of deflection of the axis of rotation due to radial play of the
axle is 0.85 deg and thus the deflection is close to zero. This means that
cospaxle q  1 and sinpaxle q  axle and as a result there is an approxi-
mately linear relation between axle and the resulting disturbances. This
is also clearly visible in the graphs included in Appendix H where the dis-
turbances are plotted for a range of deflection angles axle . If axle goes to
zero, also the disturbances in the motor reference frame Z-direction go to
zero. As mentioned before, disturbances in the motor reference frame X-
and Y-directions remain virtually the same as axle is changed.
In the analysis vizualized in the graphs in Appendix H the axle deflection
angle was held constant. In practice the axis of rotation will of course move
around within the margins of radial play, which will result in secondary
disturbances. The impact of this potential motion is however not analyzed
here. It is argued that the additional disturbances resulting from this ran-
dom motion will not be as large as the disturbances analyzed in this chapter
because their frequency will be far less than the disturbances considered in
this chapter (which have frequencies equal to the rotation speed).
Manufacturing errors are the driving factor for disturbance torques. The
graphs in Appendix H demonstrate that disturbances virtually dissappear
when the manufacturing error (and thus the deflection of the flywheel center
of gravity) goes to zero. For the same reason as with axle , there is an
approximately linear relation between cog and the resulting disturbances.

DELFI-N3XT REACTION WHEEL SYSTEM 67


Reaction Wheel Disturbances

There are virtually no remaining disturbances when cog goes to zero,


despite the fact that an acceleration of the reaction wheel is considered and
axle is kept constant at 0.85 deg. The latter makes sense because the center
of gravity will not move in that situation.

5.6 Conclusions on Disturbance Torques


In this chapter a mathematical derivation of the disturbance torques was
presented. It was argued that the two main sources for disturbance torques
are the axial play in the rotor shaft and manufacturing imperfections in
the flywheel. It was discovered that significant accelerations of the flywheel
center of gravity are expected with disturbance forces in the order of tenths
of Newtons and disturbance moments in the order of milliNewton-meters.
These disturbances are significant, especially when we consider that control
torques are in the order of microNewtons.
Estimates were made for the magnitude of the impact of axial play in
the rotor shaft and manufacturing tolerances in the flywheel. These esti-
mates were verified in the motor Z-axis direction and proven to be in close
accordance with the measured deviations in delivered motor torque. A quick
analysis showed that the main contributing source to the disturbances are
manufacturing tolerances, while the radial axle play has as effect that those
disturbances are also observed along the motor Z-axis. It is therefore be-
lieved that the disturbance model developed in this chapter is a good repre-
sentation of the main disturbance torques and supports the argument that
there is especially much to gain in a reduction of manufacturing tolerances.
The effect of disturbances of one reaction wheel on another is not math-
ematically analysed and also depends on the structure connecting separate
motors. It was calculated that the disturbance torques around the motor
X- and Y-axis are approximately two orders of magnitude larger than those
around the motor Z-axis and because either the X- or Y-axis of one motor
always corresponds to the Z-axis of another motor, the impact is likely to
be significant.

68 MSC THESIS A.G.L. HOEVENAARS


Chapter 6
Verification of Reaction Wheel
System

This chapter gives a summary of the RWS verification which is done on


the basis of the requirements that were introduced in Chapter 2. First the
general requirements are verified, followed by the functional requirements
and the performance requirements. The results and observations will be the
input for the conclusions in Chapter 8. This chapter only gives qualitative
discussions on the verification results. The argumentation and conclusions
in this chapter summarize the arguments presented in the Delfi-n3Xt Re-
action Wheel Document included in Appendix A which on their turn are
supported by tests described in the Delfi-n3Xt Test Document included in
Appendix F. This chapter therefore presents a high-level overview of the
verification activities.

6.1 Verification of General Requirements


Table 6.1 lists the requirements that were introduced in Chapter 2, the
specified values and finally the verified values. Next the requirements will
be discussed in more detail.

6.1.1 Volume and Mass Budget


The mass of the final design is well within budget and illustrates a common
situation for nanosatellites where the volume is more stringent than the
mass. The volume of the RWS is also within the latests requirements which
are presented for the full ADCS instead of for the RWS alone. The reason is
that during the design process it was decided to integrate all ADCS hardware
on a single PCB. It is therefore more useful to talk about the total volume
occupied by the ADCS. The volume budget for the full ADCS system is

69
Verification of Reaction Wheel System

Table 6.1: Verification of the general requirements

Req. Short description Target value Actual value

C.01 Mass budget 120 grams 82 grams

90 x 90 x 46.2
C.02 Volume budget mm3 for full 90 x 90 x 34.6 mm3 for full
ADCS ADCS

originally 400
C.03 Power budget mW, changed to 530 - 710 mW
710 mW

C.04 Data budget - Compatible, covered in the


Delfi-n3Xt Top-Level Design
Document [4]

C.05 Power and data bus in- I2 C I2 C


terfaces

C.06 Withstand launch envi- Dnepr launcher Qualified with representative


ronment sine and random vibration
loads

C.07 Withstand space envi- 15C to 45C Compliant in vacuum over


ronment range 0 C to 50 C, but opti-
mum range is 10 C to 30 C

G.01 Generate housekeeping - rotation speed for all three


data when of interest to wheels, but not current
satellite operations

G.02 Adhere to reliability - Mitigation of failure modes


standards

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

70 MSC THESIS A.G.L. HOEVENAARS


6.1. Verification of General Requirements

90 x 90 x 46.2 mm3 . The RWS has a footprint of roughly 40 x 40 mm2 and


a height of 27 mm, where both the flywheel size and the integrated drive
electronics are the limiting factors.
By tuning the dimensions of the RWS with the magnetorquers that were
introduced in Section 1.3, an optimal use of the occupied volume is achieved.
The structures of both the RWS and the magnetorquers have the same
maximum dimensions, do not interfere with each other and leave sufficient
surface area on the PCB for integration of all other ADCS electronics. Figure
6.1 demonstrates the assembled engineering model including both RWS and
magnetorquers.

Figure 6.1: Photo of the engineering model of ADCS with both the Reaction
Wheel System and the magnetorquers integrated on a mock-up PCB.

6.1.2 Environment
Vacuum was demonstrated to have no significant impact on the performance.
Tests described in the Reaction Wheel Test Document (Appendix F showed
that the maximum temperature increase of the RWS during operation in
vacuum is only 2.5 C, which is not significant and motor performance was
similar. Outgassing is of no concern, because suitable materials are used
and a vacuum-rated version of the motors is used which has a lifetime well
over the mission lifetime of three months [1].

DELFI-N3XT REACTION WHEEL SYSTEM 71


Verification of Reaction Wheel System

0.35
°
T = −20
0.3 °
T=0

0.25 T = 20°
T = 50°
Power [W]

0.2

0.15

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]

Figure 6.2: Impact of temperature illustrated by the characteristic perfor-


mance for 20 C and 20 C. The characteristic performance curve is ex-
plained in Appendix F.

The thermal environment on the other hand does have a big influence
on RWS performance. Figure 6.2 illustrates the impact of temperature on
the performance of an individual reaction wheel. Based on these results it is
concluded that the RWS is not designed to operate at temperatures below
0 C. The optimum temperature range was determined as 10 C - 30 C.
Changes in the temperature have an impact on R, kE and C0 . The RWS
is also compatible with the launch environment. Documentation provided
by ISIS states that sine and random vibration qualification tests are required
and that shock tests are not required [5]. The RWS was tested accordingly
in a vibration campaign facilitated by ISIS and National Aerospace Lab-
oratory, in Dutch “Nationaal Lucht- en Ruimtevaartlaboratorium” (NLR)
which proved its compliance with these requirements.

6.1.3 Power Budget


The power consumption of the motors has been proven to be significantly
outside the original budget and depends not only on rotation speed, but
is also affected by temperature and whether other reaction wheels are ac-
tuated. The impact of temperature is illustrated in Figure 6.2. In the
determined compatible temperature range of 0 C to 45 C, the worst-case
power consumption equals 200 mW per reaction wheel.
Additionally it has been observed that when other reaction wheels are ac-
tuated power consumption is further increased as explained in the Delfi-n3Xt
Test Document added in Appendix F. The impact is an additional increase
in maximum power consumption of approximately 10%. The worst-case

72 MSC THESIS A.G.L. HOEVENAARS


6.2. Verification of Functional Requirements

power consumption per motor is therefore determined as 1.1  200 mW which


equals 220 mW. Including 15 mW for each reaction wheel the total worst-
case power consumption is calculated at 710 mW. In Delfi-n3Xt there was
fortunately sufficient margin in the power budget to deal with this increase
and as a result of this analysis the requirement on the power consumption
was adapted.

6.1.4 Housekeeping Data


Full monitoring of each reaction wheel is possible when information is avail-
able on both rotation speed and drawn current. But due to the limited space
on the drive electronics board, no adequate resistor could be integrated to
measure the current. As such there is no direct information on the power
consumption. Of course the power consumption can be derived using the
model that was developed in Chapter 4 but that requires information on a
selection of motor characteristics which have been demonstrated to depend
on e.g. temperature.
The fact that no reliable information about the power consumption of
individual reaction wheels will be available presents a blind spot during
operations, because power consumption has proven to be a critical parameter
of the RWS. With only measurements of the rotation speed the RWS does
not allow continuous monitoring of the system’s health.

6.1.5 Reliability Standards


The most severe failure case that is considered is complete block of the
satellite when it is rotating at maximum speed. Calculations have shown
that this will give a step input on the satellite’s rotation speed of 2.7 /s
which is well below the thresholds that put the satellite in danger. Another
considered failure is mechanical failure. The impact of a mechanical failure
mode where a flywheel breaks loose from a motor is contained by including
protective caps in the design. These caps prevent a flywheel from causing
secondary damage in the satellite because there movement is contained to
the RWS bracket.

6.2 Verification of Functional Requirements


Table 6.2 lists the functional requirements that were introduced in Chapter 2,
the specified values and finally the verified values. All functional require-
ments are met and as such the RWS is an effective design. Only the relia-
bility of the I2 C communication between central ADCS microprocessor and
reaction wheel drive electronics still requires verification.

DELFI-N3XT REACTION WHEEL SYSTEM 73


Verification of Reaction Wheel System

Table 6.2: Verification of the functional requirements

Req. Short description Target value Actual value

about all three


F.01 Store Angular Momen- about all three principal axes
principal axes
tum

about all three


F.02 Generate torque principal axes in- about all three principal axes
stantaneously instantaneously

F.03 Handle commands from - Reliability of command han-


the ADCS microproces- dling by RWS microprocessor
sor open issue

for all three prin-


F.04 Generate reliable infor- for all three principal axes
cipal axes
mation about stored an-
gular momentum

No negative im-
F.05 Stability of Delfi-n3Xt Negligible vibrations
pact

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

74 MSC THESIS A.G.L. HOEVENAARS


6.3. Verification of Performance Requirements

6.3 Verification of Performance Requirements


Table 6.3 lists the performance requirements that were introduced in Chapter 2,
the specified values and finally the verified values.

Table 6.3: Verification of the performance requirements

Req. Short description Target value Actual value

P.01 Minimum torque about each 5.5  106 Nm 5.5  106 Nm


individual axis

P.02 Dynamic range of angular mo- 1.56  103 Nms 1.35  103 Nms
mentum about each individ-
ual axis

P.03 Systemic disturbance torques   107 Nm   1.4  107 Nm


P.04 Disturbance angular momen-   2  107 Nms   1.9  106 Nms
tum

P.05 Control over angular momen- 2  107 Nms 4.4  107 Nms
tum about each individual
axis

P.06 Accuracy of angular momen- 2  108 Nms 1.2  107 Nms


tum determination about
each individual axis

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

6.3.1 Motor Performance


As part of the motor characterization performed in the Delfi-n3Xt Reaction
Wheel Test document in Appendix F, the torque curve for the Faulhaber
1202 004 BH motor is established. From this curve it can be concluded that
the required torque of 5.5  106 Nm is approximately 4% of the maximum
torque the motor can deliver. Torque is therefore not the limiting factor.
More important is the maximum useful speed, which is the maximum speed
where the motor can still deliver this torque, because this value is directly
related to the dynamic range of the reaction wheel.

DELFI-N3XT REACTION WHEEL SYSTEM 75


Verification of Reaction Wheel System

Table 6.4: Impact of temperature and interference from other reaction


wheels on the angular momentum dynamic range.
Within optimum Outside optimum
Case
temperature range temperature range

Nominal 1.61  103 Nms 1.46  103 Nms

Including interference from 1.50  103 Nms 1.35  103 Nms


other reaction wheels

The maximum useful speed is defined as the rotation speed where the
motor is no longer capable to deliver the required torque. The relation
between maximum useful speed and angular momentum dynamic range can
be derived using Equation 3.1 and Table 6.4 shows how these factors impact
the dynamic range.
From Table 6.4 it can be concluded that only in the nominal case within
the optimum temperature range the RWS will meet the requirement for the
angular momentum dynamic range of 1.56  103 Nms. This indicates that
margins throughout the development of the RWS were insufficient to cope
with the analyzed worst-case situations. The original margin was 10% and
at least 25% should have been applied to cover the negative effects observed.
The main impacts on the angular momentum dynamic range were:

1. Higher terminal resistance. Figure 6.2 illustrates that within the


optimum operating temperature range of 10 C to 30 C, the nominal
maximum useful speed is 26,000 rpm. This is lower than anticipated
and is caused by a higher terminal resistance of the motors. The
motors used in the engineering model were found to have a terminal
resistance of 17Ω compared to 16Ω for the motors that were used for
the prototype. The specified value by Faulhaber is 16Ω with a margin
of 12%. This potential variation was insufficiently taken into account.

2. Temperature. Outside the optimum temperature range of 10 C to


30 C, the worst-case maximum useful speed is 24,500. This is the
worst-case maximum useful speed within the determined temperature
range of 0 C to 50 C where the RWS is compliant.

3. Interference from other reaction wheels. It was discovered that


when a second wheel is actuated, this has a negative impact on the
performance of the other reaction wheels and reduces the maximum
useful speed by approximately 10%. This is believed to be caused by
the vibrations induced in the RWS structure.

76 MSC THESIS A.G.L. HOEVENAARS


6.3. Verification of Performance Requirements

Although the requirement was a dynamic range of 1.56  103 Nms, a sensi-
tivity analysis called ‘SmallAct’ performed on the attitude determinination
and control algorithms for the Delfi-n3Xt [18] indicates that no significant
performance degradation of the ADCS is expected as a result of the reduced
angular momentum dynamic range of the reaction wheels.

6.3.2 Disturbances
Two types of disturbances are considered, systemic disturbance torques
and instantaneous disturbance angular momenta. The instantaneous dis-
turbances are expressed in angular momentum because those can be more
directly determined from the measured rotation speed. The conclusions on
both disturbances are:

• Systemic disturbance torques. If the Z-axes of the motor refer-


ence frames are not well-alligned with the satellie body-axes, the com-
manded torque will be delivered with an undesired angular off-set.
The maximum misalignment error was calculated as 1.45 resulting in
a maximum systemic disturbance torque of 1.4  107 Nm.

• Disturbance angular momenta. Figure 5.8 illustrated that the


maximum calculated amplitude of the periodic disturbance torques
is 7  104 Nm. If this torque is integrated over half a period of its
periodic motion, the maximum resulting angular momentum change
is 1.9  106 Nms, which is an order of magnitude higher than the
required 2  107 . However, because this disturbance momentum is
highly periodic it is estimated that the absolute impact on the satellite
attitude will not be higher than 3.8  107 deg.

Moreover these disturbances have resulted in the removal of Rational Rate


Sensors (RRSs) from the Delfi-n3Xt design because they were unable to deal
with the induced vibrations in the satellite.

6.3.3 Control and Determination Accuracy


The required angular momentum determination accuracy of 2  108 Nms
corresponds to a minimum rotation speed determination accuracy of 0.6
rpm. Because only discrete values for the rpm can be determined, the max-
imum angular momentum determination accuracy is 3  108 Nms. At high
rotation speed the angular momentum determination accuracy goes down
to 1.2  107 Nms. Equation 4.7.3 in Section 4.7.3 had demonstrated that
the average error in delivered torque is 4%. With a maximum torque of
5.5  106 Nm this corresponds to an angular momentum control accuracy
of 4.4  107 Nms. Both the angular momentum control and determination
accuracies are therefore lower than required, meaning that the impact of the

DELFI-N3XT REACTION WHEEL SYSTEM 77


Verification of Reaction Wheel System

RWS on the satellite is not negligible with respect to impact of the maximum
external disturbance torques of 106 Nm.
The Delfi-n3Xt RWS was designed with a relatively small flywheel, which
was enabled by the high maximum useful speed of the Faulhaber 1202 004
BH motor in comparison to other motor candidates. This has had a positive
effect on the control accuracy, because a smaller flywheel means that a
given acceleration corresponds to a smaller torque. Because rotation speed
is the state variable, control accuracy will decrease if a larger flywheel is
implemented in the design.

78 MSC THESIS A.G.L. HOEVENAARS


Chapter 7
Proposition for Next Generation of
Design

This chapter proposes a design that could be considered in a follow-up to


the Delfi-n3Xt Reaction Wheel System that was developed in this thesis
work. The proposition is based on the Delfi design philosophy of stepwise
baseline improvement of systems [19] and extends the modular approach in
the design of the Reaction Wheel System discussed in this thesis.
The proposed improvement consists of changes in two parts of the Re-
action Wheel System. One is to incorporate a miniature bearing as are
available from e.g. Dynaroll [20] that is integrated in the protective cap and
a flywheel that allows improved manufacturing tolerances. In this chapter
the proposed design is presented and the expected impacts are discussed.

7.1 Double Bearing


The current design includes a protective cap to limit excitation of the fly-
wheel during launch and protects the satellite in cause of a structural failure
of a reaction wheel. It is here proposed to include a miniature bearing in
the protective cap as vizualized in Figure 7.1, which would press-fit around
the axle when the protective cap is assembled. The goal is to inhibit the
transformation of disturbances into other planes which will result in addi-
tional disturbances by fixing the axle also at its free end. Fixing a reaction
wheel motor axle is an approach seen more often in reaction wheel design,
e.g. in the LeoStar wheel [21].
Additional benefits could be achieved when magnetic bearings are used
instead of ball bearings because they would significantly reduce friction and
especially static friction [22]. Efforts to miniaturize this technology are being
carried out, but the Technology Readiness Level (TRL) currently is level two
[23] and this technology is therefore far from mature. If improvements are

79
Proposition for Next Generation of Design

20 mm

2 mm

current design

22 mm

miniature bearing
incorporated in cap

1.7 mm

proposed next generation design

Figure 7.1: The current design and the proposed next generation design.
The bearing is included in the protective cap to reduce disturbances from
axle play, and a more simple flywheel design is implemented to decrease the
complexity of flywheel manufacturing with the aim to improve axisymmetry.

80 MSC THESIS A.G.L. HOEVENAARS


7.2. Simpler flywheel

to be made in the near future, than ball bearings will still be the logical
choice.
A challenge for the realization of this design will assembly and integration
which will likely become significantly more difficult than is currently the
case. It will be up to future Delfi engineers to determine the best approach.

7.2 Simpler flywheel


The flywheel can be manufactured with greater precision of less operations
are required to manufacture it. It is therefore believed that some improve-
ments in the axisymmetry of the flywheel are possible by removing the dome-
shaped extension around the motor axle. This extension was designed to
ensure good fixation around the axis, but with the second bearing, this is
no longer required.
The inclusion of the second bearing in the design does dictate a reduction
in flywheel thickness to 1.7 mm. The axle of the Faulhaber 1202 004 BH
motor extends only 3 mm above the motor base and the width of the minia-
ture bearing is 1 mm [20]. With a 0.3 mm margin, the flywheel thickness is
therefore set to 1.7 mm.
The reduction in moment of inertia due this downscaling of the flywheel
is compensated by increasing the diameter to 22 mm, which still fits within
the current bracket design (see Section 3.4). A negative side-effect is that
new assembly tools would need to be designed. The current assembly tools
are included in Appendix C covering the Technical Drawings. Assuming
Bronze Rg7 with density 8950 kg/m3 , the moments of inertia of the current
flywheel and the new flywheel are presented in Figure 7.1

DELFI-N3XT REACTION WHEEL SYSTEM 81


Proposition for Next Generation of Design

Table 7.1: Comparison of dimensions, mass and angular momentum of the


currently implemented flywheel and the flywheel for the proposed next gen-
eration design.

Property Current flywheel Proposed flywheel

Dome feature Yes No

Diameter [mm] 20 22

Thickness [mm] 2 1.7

Mass [g] 6.0 5.7

Moment of inertia kgm2 2.83  107 3.44  107

7.3 Impact on Power and Dynamic Range


An inevitable impact of the inclusion of a second bearing will be increased
static- and dynamic friction. Table 7.2 lists the assumptions on the respec-
tive coefficients and the resulting angular momentum dynamic range. The
dynamic range is calculated using:

H  I nmax
30
π
(7.1)

where nmax is read from Figure 7.2 which vizualizes the impact on the
performance of the reaction wheels. The graphs show the required power
to deliver the required 5.5  106 Nm torque as a function of rotation speed.
At a certain speed the motor can no longer deliver the required torque; this
speed is the maximum useful speed. Despite the lower maximum speed, the
angular momentum dynamic range is equal for both designs, although the
power consumption of the proposed next generation design is significantly
higher; some 60%.

82 MSC THESIS A.G.L. HOEVENAARS


7.3. Impact on Power and Dynamic Range

Table 7.2: Estimation of impact of the proposed design on the static friction
C0 , the dynamic friction Cv and the angular momentum dynamic range
compared with the current design.

Current Proposed Rationale

It is estimated that the static friction will


increase with 75% if a second bearing is
C0  0.026  103 C0  0.044  103 included. The static friction is mainly due
to the bearing and the value will therefore
almost double.

It is estimated that the dynamic friction


(in Nm/rpm) will increase with 50% if a
Cv  0.052  109 Cv  0.078  109 second bearing is included. The dynamic
friction is also largely due to Foucault cur-
rents in the stator, which will not be af-
fected by a second bearing.

nmax  26, 100 nmax  21, 700 The estimated reduction in maximum use-
ful speed is roughly 20%.

Despite the lower value for nmax , the pro-


posed improvement to the reaction wheel
∆H  1.6  103 ∆H  1.6  103 design has the same angular momentum
dynamic range due to the higher moment
of inertia of the new flywheel.

DELFI-N3XT REACTION WHEEL SYSTEM 83


Proposition for Next Generation of Design

Current design
0.25 Proposed design

0.2
Power [W]

0.15

0.1

0.05

0
0 0.5 1 1.5 2 2.5 3
Rotation speed [rpm] 4
x 10

Figure 7.2: The required power to deliver the required 5.5  106 Nm torque
as a function of rotation speed for both the current design and the proposed
next generation design.

7.4 Conclusions on Proposed Design


In an attempt to reduce disturbances from the reaction wheel, a follow-up
to the current design is proposed. The proposed next generation design in-
cludes a second bearing as part of the protective cap to remove the axle play.
Additionally the flywheel design is simplified which is likely to increase man-
ufacturing accuracy. The flywheel is also increased in diameter and reduced
in thickness to accomodate the protective cap with miniature bearing.
The result is a proposed design that has an equal angular momentum
dynamic range but will likely reduce the disturbance torques as discussed in
Chapter 5. However due to additional bearing friction, the power consump-
tion is estimated to go up 60%. Whether the proposed improvement will be
beneficial for future Delfi designs will therefore likely be a trade-off between
power and impact of disturbances.

84 MSC THESIS A.G.L. HOEVENAARS


Chapter 8
Conclusions

To improve the pointing capabilities of nanosatellites a fast response atti-


tude control system is required. Reaction wheels are an effective solution
and rely on the simple principle of conservation of angular momentum. Also
they can be developed within the stringent budgets of nanosatellites with
current technologies. The Delfi-n3Xt satellite of the Delft University of
Technology that is scheduled for launch in September 2012 will also include
reaction wheels and this thesis covers the design, integration and verification
of the Delfi-n3Xt Reaction Wheel System (RWS), which will be among the
first three-axis reaction wheel systems for nanosatellites in space.

Conclusion on design. The RWS developed in this thesis project can be


described as effective and balanced. Effective because all functional require-
ments are met and balanced because although a number of performance
requirements are not met, performance degradation is evenly spread over
all performance requirements as was presented in Table 6.3. The original
requirements were very ambitious and the design described in this report
demonstrates a best effort approach to meet those.

Summary of hardware design. The Delfi-n3Xt RWS consists of three


orthogonally placed reaction wheels and is integrated on the Attitude De-
termination & Control System (ADCS) PCB. The motor has the biggest
impact on the design and for the Delfi-n3Xt RWS Commercial Off-The-Shelf
(COTS) Brushless Direct Current (BLDC) motors are used. Based on the
requirements, the Faulhaber 1202 004 BH micromotor was identified as the
best choice, because of its small size and high maximum rotation speed. To
reduce the volume of the system and increase modularity, the drive electron-
ics were also integrated in the RWS. The drive electronics also contains the
microprocessor required for commutation, torque control and command- and
data handling. To ensure the structural integrity during launch, protective

85
Conclusions

caps were succesfully integrated in the design to limit the excitations of the
axle with respect to the motor and mitigate the effects of a structural failure.

Torque control. To support the development of an effective controller,


first the system model was established for a single reaction wheel. After
validation of this model, a discrete controller was developed with rotation
speed as the only state variable and the supply voltage as the control vari-
able using Pulse Width Modulation (PWM). Because the purpose of the
RWS is to deliver torque, which is proportional to the acceleration of the
flywheel, a discretized ramp function was designed to interpret commanded
torque inputs into an incrementally changed reference speed. Gains were
calculated separately for the low speed region and the high speed region
to ensure a good responsive over the full dynamic range. Simulations have
shown that the controller can effectively deliver the commanded torque and
is able to handle zero crossings.

Assembly, integration and testing. Early in the development process a


prototype was developed to verify the structural integrity of the full RWS
design. This prototype also allowed verification of the assembly and inte-
gration process with the result that special procedures and tools were de-
veloped to improve the ease of integration for the final design. To facilitate
motor performance verification a dedicated test was developed that allows
easy comparison between theory and practice. Other test activities included
thermal-vacuum tests, mechanical tests and electromagnetic tests.

Disturbances. During the testing activities it was observed that the RWS
causes significant vibrations in the surrounding structure. A mathemati-
cal model has confirmed that the main contributors to these disturbances
are manufacturing imperfections in the flywheel and radial axle play of the
motor. The disturbances induced by the RWS greatly exceed the required
maximum value and their amplitude is directly related to the reaction wheel
rotation speed. At 25,000 rpm the calculated amplitude of the resulting dis-
turbance torques is in the order of 104  103 Nm, whereas the maximum
external disturbance torques are in the order of 106 Nm. But since the
disturbances originating from the RWS are periodic, the absolute impact on
the satellite attitude is negligible. They do however impact the performance
of surrounding hardware and a Rational Rate Sensor (RRS) has been re-
moved from the Delfi-n3Xt design as a result.

Requirement verification. The design meets all functional requirements


and is compliant with the majority of general requirements and constraints.
The majority of the performance requirements were however not fully met.
The RWS operates well in the vacuum environment of space, but temper-
ature below 0 C result in an unacceptable increase in power consumption.

86 MSC THESIS A.G.L. HOEVENAARS


The requirement on the design temperature range was therefore changed
to 0 C to 45 C (originally 15 C to 45 C) to limit the worst-case power
consumption. Even then the RWS is not compliant with the original power
budget of 400 mW, which is now considered unrealistic. Because sufficient
margin in the power budget was available, the requirement on the power
budget was changed to 710 mW to cover the expected worst-case power
consumption.
The total mass of the RWS is 82 grams and the system is integrated in
a ADCS that has a total volume of 90 x 90 x 34.6 mm3 . However, because
of the small flywheels and because insufficient margin was taken in the pre-
liminary design phase, the requirement on the angular momentum dynamic
range is not met. Under the influence of temperature and disturbances from
other reaction wheels, the angular momentum dynamic range of each reac-
tion wheel is 1.35  103 Nm in the worst-case situation, wheras 1.56  103
Nms is required. Finally, the developed RWS also does not comply with the
determined requirements on torque control and angular momentum determi-
nation. This means that the inaccuracy in the delivered torque is significant
with respect to the external disturbance torques.
The performance requirements were derived such that the RWS does
not limit the overall performance of the satellite. The fact that many per-
formance requirements are not met indicates that this is not the case for
this design. Nonetheless the design is functional as discussed throughout
Chapter 6 and will thereby greatly improve the pointing capabilities of the
Delfi-n3Xt satellite and represents a great step forward to more advanced
nanosatellite missions.

DELFI-N3XT REACTION WHEEL SYSTEM 87


This page intentionally contains only this sentence.
Chapter 9
Recommendations

Because the Delfi-n3Xt RWS is among the first reaction wheel systems for
nanosatellites, there is little practical experience in the development of such
systems for nanosatellites. This chapter therefore firstly contains a number
of recommendations that will be helpful to anyone who is developing reac-
tion wheels for nanosatellites. These recommendations are:

Greater margins in preliminary design phase. In this design 10%


contingency on the maximum useful speed was applied after verification of
the preliminary design, but this did not cover the worst-case situation. 25%
would be a better value for the margin on the maximum useful speed. Also
the original estimate of the power consumption was too optimistic. The de-
sign discussed in this report exceeds the original requirements by 80% due
to negative impacts of the thermal environment and interference between
reaction wheels. It is therefore advised to carefully consider the impact of
environmental variations and parameter margins on the power consumption
in the preliminary design phase.

Electronics in design trade-off. In the preliminary design, the drive


electronics and in particular the microprocessor were not considered to be
a driving factor in the development. However, the drive electronics were
later found to be the limiting factor on the RWS volume. Additionally, dis-
cretization effects in the microprocessor have a clear impact on the control
and determination performance. Taking the drive electronics earlier into
account may therefore benefit the overall design.

Include current measurements. Including a resistor in the drive elec-


tronics design to measure the current will benefit a reaction wheel design
based on BLDC motors in two ways. First it allows more direct torque
control because with information on both the rotation speed and current

89
Recommendations

the delivered torque can be directly determined. In this thesis work control
was implemented with rotation speed as the only state variable and because
torque is proportional to acceleration of the flywheel, it had to be based
on the first derivative of that variable. Secondly it allows better health
monitoring of a reaction wheel system, because if the supply voltage is kept
constant and torque control is applied using PWM, a BLDC motor can be
fully described using information on the rotation speed and the drawn cur-
rent.

Account for disturbances. Throughout the design and verification it


has been realized that disturbances are of major concern to the performance
of the RWS and surrounding hardware. Discussions with professionals of
the industry have confirmed that reduction of disturbance vibrations is of
key importance in the development of future reaction wheel systems for
nanosatellites. Because this thesis has presented a model to predict distur-
bances, future designs could leverage on this model and take disturbances
into account in an earlier phase of development.

Simple flywheel design. Although a simple disk-shaped design is of-


ten sub-optimal from a mass- or volume perspective, it is believed that it is
the best option for nanosatellites. The small manufacturing tolerances that
can be applied to such simple design will greatly reduce disturbances which
have been identified as one of the main concerns for future developments in
reaction wheel systems for reaction wheels.

Above recommendation are valid for nanosatellite reaction wheel develop-


ment in general. Below are also a number of recommendations that are more
specific to future developments based on the specific design introduced in
this thesis.

Proposed next generation design. In this thesis report a proposal was


made for a follow-up to the developed RWS. The proposed design has a dual
aim: first it tries to minimize the amplitude of the disturbances through a
simpler flywheel design. Next it cancels the transformation of the distur-
bance vector by fixing the motor axle using a miniature bearing in the
protective cap. This proposed design would however come at the cost of
roughly 50% increase in power consumption. Including a magnetic bearing
instead of a ball bearing in the protective cap would likely be a better solu-
tion from a power perspective, but the Technology Readiness Level (TRL)
of this technology for nanosatellites is too low to consider this option in the
short term.

Tunable Compliant Vibration Isolators. At the faculty of Mechani-


cal, Maritime and Materials Engineering (3ME) of the Delft University of

90 MSC THESIS A.G.L. HOEVENAARS


Technology work is being done on Tunable Compliant Vibration Isolators
which could be implemented to damp vibrations caused by the reaction
wheels. This technology could be investigated for future designs.

DELFI-N3XT REACTION WHEEL SYSTEM 91


This page intentionally contains only this sentence.
Bibliography

[1] Bouwmeester, J., “DNX-TUD-TN-0263 Delfi-n3Xt Mission Definition,


Systems Overview and General Design Approach,” tech. rep., Delft
University of Technology, 2010.

[2] Briess, K., Baumann, F. and Trowitzsch, S., “Present and Future Pi-
cosatellite Missions at TU Berlin Definition of Picosatellite,” 8th IAA
Symposium Small Satellites for Earth Observation, 2011.

[3] Dekens, E.H., “DNX-TUD-TN-0284 ADCS - Reaction Wheel Design


V1.4,” tech. rep., Delft University of Technology, 2010.

[4] Vrakking, V., “DNX-TUD-TN-0142 Top Level Design ADCS,” tech.


rep., Delft University of Technology, 2011.

[5] Bolhuis, M.F., “DNX-TUD-ISS-1008 Environment Levels Auxiliary


Payloads ISILaunch03,” tech. rep., ISIS, 2012.

[6] Gill, E.K.A., “Space Systems Engineering course notes,” 2010.

[7] Midwest Research Institute, “Brushless DC motors,” Tech. Rep. Jan-


uary, 1975.

[8] Fulcher, R., “A Brushless DC Torquer-driven reaction wheel for space-


craft attitude control,” tech. rep., 1969.

[9] Wang B., Gong K. et. al. , “Fine attitude control by reaction wheels
using variable-structure controller,” Acta Astronautica, vol. 52, pp. 613–
618, Apr. 2003.

[10] Dr. Fritz Faulhaber GmbH, “Brushless Flat DC-Micromotors,” 2010.

[11] Siong T.C., Ismail B., Siraj S.F. et. al. , “Fuzzy Logic Controller for
BLDC Permanent Magnet Motor Drives,” International Journal of
Electrical & Computer Sciences, vol. 11, no. 02, pp. 13–18, 2011.

93
BIBLIOGRAPHY

[12] Rodriguez, F. and Emadi, A., “A novel digital control technique for
brushless DC motor drives: conduction-angle control,” IEEE Inter-
national Conference on Electric Machines and Drives, 2005., vol. 2,
pp. 308–314, 2005.

[13] Dr. Fritz Faulhaber GmbH, “Technical Information,” 2010.

[14] Sidi, M. J., Spacecraft Dynamics and Control. Cambridge Aerospace


Series, Cambridge University Press, 1st ed., 1997.

[15] Haugen, F., “Lecture notes in Models, Estimation and Control,” 2009.

[16] Wescott, T., Applied Control for Embedded Systems. Embedded tech-
nology series, Elsevier Inc., 2006.

[17] Ogata, K., Modern Control Engineering. Pearson Education, 4th ed.,
2002.

[18] Reijneveld, J.P.J., “Design of the Attitude Determination and Control


Algorithms for the Delfi-n3Xt,” 2012.

[19] Bouwmeester, J., Gill, E.K.A. et. al., “DelFFi - a Formation Flying
Demonstration of Two CubeSats within the QB-50 Mission,” 2012.

[20] Dynaroll, “Metric Precision Bearings.”


http://www.dynaroll.com/pag054.asp, Visited January 2012.

[21] Ahronovich, E. and Balling, M., “Reaction Wheel and Drive Electronics
For LeoStar Class Space Vehicles,” 12th Annual/USU Conference on
Small Satellites, 1998.

[22] Rajagopal, K.R. and Sivadasan, K.K., “Low-stiction magnetic bearing


for satellite application,” Journal of Applied Physics, vol. 91, no. 10,
p. 6994, 2002.

[23] Davis K., “Magnetic Bearings for Small Satellite


CMGA’s and other Miniature Spacecraft Mechanisms.”
http://sbir.gsfc.nasa.gov/SBIR/abstracts/10/sbir/phase1/SBIR-
10-1-S3.06-9597.html, 2010.

94 MSC THESIS A.G.L. HOEVENAARS


Appendix A
Design Technical Note

This appendix contains the Reaction Wheel Design document. Its main
purpose is the documentation of design choices leading towards the physical
design that is implemented in the Delfi-n3Xt satellite.

95
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 1 / 74

ADCS – Reaction Wheel System Design

Description: Design of reaction wheel system

COMMS

Launch
MechS

T3μPS
CDHS
ADCS

ITRX

SDM

GSN
GSE
EPS

TCS
STS
Subsystem(s)
involved:

Revision Record and Authorization

approved
Reviewer
checked
Issue

Affected
Date Author / Editor PM Description of change
Section(s)

0.1 20-11-2008 Wouter Edeling All First Draft


Drive electronics, power
1.0 14-01-2009 Wouter Edeling All
consumption, conclusion added
1.1 19-01-2009 Wouter Edeling JB All Comments processed
1.2 27-11-2009 Erwin Dekens Flywheel redesign
1.3 13-10-2010 Erwin Dekens All Complete revision
Corrected erroneous power
1.4 09-11-2010 Erwin Dekens GB JB Appendix A
values
Determination of requirement
1.5 17-12-2010 Teun Hoevenaars
SAT.2.2.2.2.P.04
1.6a 27-3-2011 Teun Hoevenaars 4–6 Update (partly) to new motor
1.7 16-5-2011 Teun Hoevenaars All Full update to new motor
Draft of full document on the
2.0 11-11-2011 Teun Hoevenaars All final Delfi-n3Xt Reaction Wheel
System design.
More accurate performance
mainly tests performed on final design,
2.1 23-01-2012 Teun Hoevenaars
9 – 11 conclusion adapted, action item
list added.
Document fine-tuned, controller
extended. Comments JB and GB
2.2 24-01-2012 Teun Hoevenaars 7, 9 – 11
processed, performance
verification finalized
Outgassing added plus minor
2.3 21-02-2012 Teun Hoevenaars 2.1, all
mistakes corrected

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 2 / 74

Action Items
Applicable
TBW
TBD
TBC
Description of action item
Section(s)

X 7.3 Implementation of motor control


X 9.3 Verification of motor control.
X 7.4 Assign location of the 2 bytes in 7 byte read-out of the reaction wheels by
the central ADCS microprocessor.
List of Used References
SLR code Version Data/Variable
0017 5.0 Original power budget for Reaction Wheel System
0018 4.0 Mass budget for Reaction Wheel System
0142 2.2 Data budget for Reaction Wheel System
0143 1.5 Disturbance torque calculations
0144 1.5 Required control torques
0167 3.33 Delfi-n3Xt General Requirements and Top-Level Reaction Wheel Requirements
0263 2.3 Reliability standards
0287 2.4 Delfi-n3Xt standard system bus interface
0303 2.1 Original volume budget for Reaction Wheel System
2.2 Adjusted volume budget for Reaction Wheel System
0565 2.04 Electrical wiring interface
0771 - Reaction Wheel Test Scripts For LabView
0766 2.0 Reaction Wheels Vibration Test Series at 3ME Report
0854 - Faulhaber 1202 004 BH Brushless Flat DC Micromotor specifications
0861 2.01 ADCS data budget
0868 1.2 Reaction Wheels Vibration Test Series at NLR Report
0872 2.2 Thermal budget
0869 1.0 Reaction Wheel Motor Characterization Manual
0927 1.0 Catia drawings of Reaction Wheel System hardware and assembly hardware
0979 2 Faulhaber motor parameter relations
1007 1.1 Assembly Manual for Reaction Wheel System
1008 0.9 Launch vibration levels

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 3 / 74

Table of Contents
1  INTRODUCTION __________________________________________________________________ 5 

2  REQUIREMENTS __________________________________________________________________ 7 

2.1  General requirements __________________________________________________ 7 

2.2  Functional requirements ________________________________________________ 9 

2.3  Performance requirements ______________________________________________ 9 

3  TOP-LEVEL REACTION WHEEL SYSTEM DESIGN _____________________________________ 11 

3.1  High-level inputs ____________________________________________________ 11 

3.2  Hardware architecture ________________________________________________ 11 

4  REACTION WHEEL MOTORS _______________________________________________________ 12 

4.1  General considerations ________________________________________________ 12 

4.2  Selection criteria ____________________________________________________ 12 

4.3  Motor options _______________________________________________________ 15 

4.4  Motor selection ______________________________________________________ 15 

5  PRELIMINARY DESIGN ___________________________________________________________ 18 

5.1  Flywheel design and sizing _____________________________________________ 18 

5.2  Bracket prototype design ______________________________________________ 19 

5.3  Prototype assembly __________________________________________________ 19 

6  VERIFICATION OF PRELIMINARY DESIGN __________________________________________ 21 

6.1  Verification of manufactured flywheels ___________________________________ 21 

6.2  Motor performance verification _________________________________________ 21 

6.3  Preliminary verification of structural integrity ______________________________ 25 

6.4  Budget compliance ___________________________________________________ 25 

6.5  Conclusion from prototype verification ___________________________________ 26 

7  REACTION WHEEL DRIVE ELECTRONICS ____________________________________________ 27 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 4 / 74

7.1  Commutation _______________________________________________________ 28 

7.2  I2C communication ___________________________________________________ 28 

7.3  Motor control _______________________________________________________ 29 

7.4  Microcontroller implementation _________________________________________ 30 

7.5  Drive electronics PCB _________________________________________________ 35 

8  ENGINEERING MODEL ____________________________________________________________ 36 

8.1  Flywheel ___________________________________________________________ 37 

8.2  Integration of the drive electronics PCB ___________________________________ 38 

8.3  Bracket ____________________________________________________________ 38 

8.4  Protective caps ______________________________________________________ 39 

8.5  Engineering model assembly ___________________________________________ 39 

8.6  Assembly alignment error______________________________________________ 40 

8.7  Axisymmetric error ___________________________________________________ 42 

9  VERIFICATION OF THE FINAL DESIGN______________________________________________ 49 

9.1  Verification of general requirements _____________________________________ 49 

9.2  Verification of functional requirements ___________________________________ 51 

9.3  Verification of performance requirements _________________________________ 52 

10  CONCLUSIONS __________________________________________________________________ 58 

11  NEXT STEPS _____________________________________________________________________ 60 

APPENDIX A.  RAW MEASUREMENTS OF PROTOTYPE MODEL PERFORMANCE _____________ 61 

APPENDIX B.  THE COMMUTATION SCHEMES FOR A FOUR POLE PAIR BLDC MOTOR _______ 64 

APPENDIX C.  SCHEMATIC OVERVIEW OF THE MICROPROCESSOR IMPLEMENTATION OF


REACTION WHEEL CONTROL __________________________________________________________ 66 

APPENDIX D.  DERIVATION OF FLYWHEEL ACCELERATION DUE TO AXISYMMETRIC ERROR 69 

APPENDIX E.  RAW MEASUREMENTS OF ENGINEERING MODEL PERFORMANCE ___________ 72 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 5 / 74

1 Introduction
Reaction wheels act as fast response attitude control actuators to give Delfi-n3Xt direct pointing capabilities.
This document details the design of the Reaction Wheel System (RWS). Figure 1 illustrates the design
process, in which the relevant chapters or references are indicated.

In this design document, first the budget constraints and requirements on the reaction wheel design are
discussed, which is done in Chapter 2. Chapter 3 next explains the design philosophy of the Reaction Wheel
System and the top-level architecture of the reaction wheels. The motors that are used for the reaction
wheels have the biggest impact on the design and it was decided to go for commercially off-the-shelf motors.
The freedom of choice is however limited by availability at the commercial market. For this reason selection of
the motors is the first step in the design process, which is done in Chapter 4. The preliminary design and
assembly of the prototype based on these motors is subsequently discussed in Chapter 5.

With the prototype developed, initial compliance with the requirements is assessed in Chapter 6. After this
preliminary verification, the required drive electronics and control algorithms are developed in Chapters 7 to
complete the full Reaction Wheel System. Chapter 8 introduces the engineering model that was developed to
verify the full Reaction Wheel System.

The Reaction Wheel System described in this document concerns a second design iteration. The first design
that has been developed up to qualification testing of the preliminary design did not survive excessive
vibration tests at NLR. For more details on the previous design, which was based on Faulhaber 2209 T 005
BSC motors, see version 1.4 of this document. The current design builds on the experience from the initial
work by Erwin Dekens and has resulted in a Reaction Wheel Design that is better compliant with the
requirements. Chapter 9 concerns the verification of the engineering model with respect to the requirements,
followed by conclusions and next steps in Chapter 10 and Chapter 11.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 6 / 74

Figure 1: The Reaction Wheel System design process.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 7 / 74

2 Requirements
The requirements listed below were extracted from the Delfi-n3Xt Requirements and Configuration Item List
[SLR0167]. All requirements are briefly discussed and only if the requirement is introduced in this document,
a more elaborate derivation is presented.

2.1 General requirements


Identifier Description
SAT.C.01 All satellite systems shall comply with the mass budget, as given in [SLR 0018]
SAT.C.02 All satellite systems shall comply with the volume budget, as given in [SLR 0303]
SAT.C.03 All satellite systems shall comply with the power budget, as given in [SLR 0017]
SAT.C.04 All satellite systems shall comply with the data budget, as given in [SLR 0861]
All satellite systems shall comply with power and data bus interfaces, as specified in
SAT.C.05
[SLR0287] and [SLR0565]
SAT.C.06 All satellite systems shall be able to withstand the launch environment
SAT.C.07 All satellite systems shall be able to withstand the space environment
All satellite bus systems shall generate housekeeping data when of interest to satellite
SAT.2.F.02
operation
SAT.2.C.01 All satellite bus systems shall adhere to reliability standards, as specified in [SLR 0263]

Interpretation for the RWS of Requirement SAT.C.01


The mass budget for the Reaction Wheel System at the moment of development can be found in the Delfi-
n3Xt Mass Budget document [SLR0018, V4.0]. The available mass for the RWS is maximum 120 grams.

Interpretation for the RWS of Requirement SAT.C.02


The originally budgeted volume for the reaction wheels PCB was roughly 90 x 90 x 32.6 mm3 [SLR0303,
V2.1]. The maximum available height above the Attitude Determination & Control (ADCS) PCB on the side
where the reaction wheels were to be mounted was roughly 19 mm. It was however later decided to combine
both the ADCS PCB and the RWS PCB, resulting in a new volume budget for the full ADCS of 90 x 90 x 46.2
mm3 [SLR0303, V2.2]. If the height of the ADCS PCB itself is subtracted, the actual available volume for
integration of the RWS is 90 x 90 x 38.6 [mm3], which is shared with other ADCS hardware as well.

Figure 2: On the left: the original volume budget, where the ADCS and RWS board were separated. On the right, the new
lay-out where both boards are combined on a single PCB. In both sketches also the distance to the underlying PCB is
presented.

Interpretation for the RWS of Requirement SAT.C.03


The power budget for the Reaction Wheel System can be found in the Delfi-n3Xt Power Budget document
[SLR0017, V5.0]. The originally specified value for the power consumption of the full RWS including drive
electronics was maximum 400 [mW].

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 8 / 74

Interpretation for the RWS of Requirement SAT.C.04


As the Reaction Wheel System is part of the ADCS subsystem, the data budget of the ADCS subsystem also
covers the Reaction Wheel System. For more details see the ADCS Top Level Design of Attitude Control and
Determination Subsystem [SLR0142, V2.2].

Interpretation for the RWS of Requirement SAT.C.05


The RWS shall operate on 3.3V DC input power and exchange data using the I2C bus protocol. The RWS is
connected with the ADCS PCB and has no direct interface with the DSSB.

Interpretation for the RWS of Requirement SAT.C.06


The Reaction Wheel System is a physical system with real chance of damage during launch. Therefore this
requirement is of particular interest to the RWS. Delfi-n3Xt is launched using a Dnepr launch vehicle for which
the launch loads are detailed in the Launch Environmental Levels for the ISILaunch03 document [SLR1008].

Interpretation for the RWS of Requirement SAT.C.07


The Reaction Wheel System is a physical system that is likely to be impacted by the space environment.
Therefore it should be assessed whether the system can withstand worst case scenarios and how the space
environment impacts performance. The space environment consists of vacuum, radiation and a thermal
environment. The design thermal environment is described in the Delfi-n3Xt Thermal Budget [SLR0872, V2.2]
and introduces a design temperature range of -15°C to 45°C. Impacts of radiation are not required to be
assessed. Vacuum also causes outgassing of materials and this should not affect the performance of the
satellite during the design lifetime of three months as introduced in Requirement Sat.2.C.02 of [SLR0167,
V3.33].

Interpretation for the RWS of Requirement SAT.2.F.02


The Reaction Wheel System is commanded by the ADCS processor. This processor will require data to verify
the actions of the RWS and effectively use the system. Additionally, because the Reaction Wheel System is a
technology demonstration subsystem of high interest to future Delfi missions, a complete data set is
invaluable to the assessment of the performance of the RWS.

The behavior of the RWS can be fully monitored when the power consumption and the rotation speed over
time are known. These variables should therefore be included in the housekeeping data. The requirement on
the rotation speed accuracy is inherently included in the Requirement SAT.2.2.2.2.P.06 on the angular
momentum determination accuracy. The power consumption can be monitored by measuring the current,
because the supply voltage is fixed at 3.3 [V]. A resolution of 1 mA is sufficient to determine the power
consumption to within 2% accuracy for a maximum power of 400 mW.

Interpretation for the RWS of Requirement SAT.2.C.01


Because the RWS is not a critical system, but a technology demonstration subsystem, a pragmatic approach
is taken with respect to reliability. This means that the effects of potential failures should be mitigated such
that they do not impact critical systems of the satellite.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 9 / 74

2.2 Functional requirements


Identifier Description
The Reaction Wheel System shall be able to store angular momentum about all three
SAT.2.2.2.2.F.01
principal axes.
The Reaction Wheel System shall be able to generate torques about all three principal
SAT.2.2.2.2.F.02
axes instantaneously.
The Reaction Wheel System shall be able to handle commands from the central ADCS
SAT.2.2.2.2.F.03
microprocessor.
The Reaction Wheel System shall generate reliable information about the stored angular
SAT.2.2.2.2.F.04
momentum among all three principal axes.
SAT.2.2.2.2.F.05 The reaction wheels shall not have a negative impact on the stability of Delfi-n3Xt.

Explanation of Requirement SAT.2.2.2.2.F.01


In order for the Reaction Wheel system to be compliant with Requirement SAT.2.2-F.01; The ADCS shall
demonstrate active attitude determination and control capabilities in three axes [SLR0142, V2.2], the RWS
should be able to store angular momentum about all three axes.

Explanation of Requirement SAT.2.2.2.2.F.02


In addition to Requirement SAT.2.2.2.2.F.01, the Reaction Wheel System should also be able to generate
torques about all three principal satellite axes instantaneously in order to be an effective solution.

Explanation of Requirement SAT.2.2.2.2.F.03


Because the RWS is a subsystem of the ADCS system and its impact on the satellite is controlled by the ADCS
microprocessor, the RWS should operate on the basis of ADCS commands.

Explanation of Requirement SAT.2.2.2.2.F.04


This requirement is an explicit interpretation of Requirement SAT.2.F.02 introduced in Section 2.1 because
reliable information about the stored angular momentum is critical to verify the performance of the RWS.

Explanation of Requirement SAT.2.2.2.2.F.05


Because the Reaction Wheel System is being developed to improve satellite control, its implementation should
not have a negative impact on satellite stability. This requirement concerns e.g. the assessment of generated
disturbance torques and effects of potential failures of the RWS.

2.3 Performance requirements


Identifier Description
The reaction wheels shall be able to generate a minimum torque of 5.5·10-6 Nm about
SAT.2.2.2.2.P.01
each individual axis.
The reaction wheels shall be able to hold angular momentum over a dynamic range
SAT.2.2.2.2.P.02
spanning 1.56·10-3 Nms about each individual axis.
The reaction wheels shall not generate systemic disturbance torques higher than
SAT.2.2.2.2.P.03
10−7 Nm.
The reaction wheels shall not generate disturbance angular momenta greater than
SAT.2.2.2.2.P.04
2·10-7 Nms.
The Reaction Wheel System shall be able to control the angular momentum about each
SAT.2.2.2.2.P.05
individual axis with a minimum accuracy of 2·10-7 Nms.
The Reaction Wheel System shall be able to determine the angular momentum about
SAT.2.2.2.2.P.06
each individual axis with a minimum accuracy of 2·10-8 Nms.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 10 / 74

Derivation of Requirement SAT.2.2.2.2.P.01


The minimum torque that should be generated by the reaction wheels among each axis is determined by
calculating the worst case disturbance torque among any of the three principal axes [SLR0144,V1.5]. The
maximum expected disturbance torque occurs during operation of the T3μPS cold gas micropropulsion system
due to misalignment of the thrust vector with a calculated value of 4.05 · 10−5 [Nm]. This value would have a
big impact on the design, while operation of the T3μPS is a one-time event, the resulting disturbance torque is
based on large thrust misalignments and rejection of this torque is not mission critical. Therefore, the
minimum required torque is prescribed by the second largest required control torque, namely for ground
station tracking; 5.5 · 10−6 [Nm].

Derivation of Requirement SAT.2.2.2.2.P.02


The requirement on the dynamic range of the reaction wheels is based on the angular momentum coinciding
with the maximum accumulated torques necessary for sun-pointing the nanosatellite [SLR0144,V1.5]. This
requirement is derived such that the reaction wheels are not the limiting factor in sun-pointing the satellite.
Therefore the RWS should be able to store a minimum of 7.7 · 10−4 [Nms] in each axial direction. As such,
the dynamic range amounts to 1.56 · 10−3 [Nms] about each axis.

Derivation of Requirement SAT.2.2.2.2.P.03


The reaction wheels are also used for disturbance torque rejection. External disturbance torques typically vary
slowly with time and affect the attitude of Delfi-n3Xt because their impact accumulates over time. In order to
be fully effective, the systemic disturbance torques TD,sys from the reaction wheels themselves should be
lower than these external disturbance torques. For this purpose the maximum systemic disturbance torque
from the reaction wheels is defined as one order of magnitude lower than the maximum external disturbance
torque of 10−6 [Nm] [SLR0143,V1.5]. Therefore the maximum systemic disturbance torques of the reaction
wheels is 10−7 Nm.

Derivation of Requirement SAT.2.2.2.2.P.04


The disturbance momentum from the reaction wheels is defined as the disturbance torques integrated over
one control cycle (i.e. 2 seconds). The disturbance momenta from the reaction wheel shall therefore not be
larger than 2·10−7 [Nms].

Derivation of Requirement SAT.2.2.2.2.P.05


In the ideal situation, the reaction wheels generate a constant torque over the time of actuation. The control
accuracy is therefore defined as the integrated error in the delivered torque. Torque integrated over time is
angular momentum and the control accuracy is therefore given in angular momentum. In accordance with
Requirement P.04, the disturbance momenta shall not be greater than 2·10−7 Nm. Therefore, the control
accuracy shall be better than 2·10−7 [Nms].

Derivation of Requirement SAT.2.2.2.2.P.06


The requirement on the measurement accuracy is derived from SAT.2.2.2.2.P.05. In order to ensure the
derived control accuracy, the measurement accuracy of angular momentum should be at least one order of
magnitude better than the control accuracy; 2·10−8 [Nms]

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 11 / 74

3 Top-level Reaction Wheel System design


The design philosophy that is introduced in this chapter together with the list of requirements introduced in
Chapter 2 form the inputs for the motor selection which has a large influence on the overall design of the
Reaction Wheel System.

3.1 High-level inputs


Internal development versus buying COTS
Three options for the degree of internal development were analyzed in the top-level design of the ADCS
[SLR0142]:

1. Designing, developing and manufacturing of a dedicated reaction wheel.

2. Buying a Commercial Off-The-Shelf (COTS) reaction wheel.

3. Combining a COTS motor with a customized flywheel.

The third option was chosen, because it is considered to offer the best combination of available components,
development effort and cost.

Number of individual reaction wheels


The top-level ADCS design [SLR0142] calls for three individual reaction wheels. This is the minimum number
needed for attitude control about all axes. A higher number would add redundancy, but this was not deemed
necessary due to ADCS not being a mission-critical subsystem [SLR0167].

External interfaces
The Reaction Wheel System is a subsystem of the ADCS and most interfaces are with this subsystem. The
RWS is commanded by the ADCS central microprocessor.

3.2 Hardware architecture


The hardware architecture for the reaction wheels based on the top-level ADCS inputs is depicted in Figure 3.

Figure 3: Reaction wheels hardware architecture

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 12 / 74

4 Reaction wheel motors


This chapter presents the arguments for the choice for the Faulhaber 1202 004 BH motor as a basis for the
Reaction Wheel System. General considerations and selection criteria are presented, after which the motors
that have been considered are introduced. Finally the trade-off and its results are presented.

4.1 General considerations


Motor type
Requirement SAT.C.05 states that the reaction wheels should be compatible with 3.3 V Direct Current (DC)
input power. In order to minimize losses it was decided to apply this voltage directly to the motors because
sufficient choice in motors is available that are compliant with this value. Thus, DC-motors are preferred.
These can be divided into two main types: brushed motors and brushless motors. Brushless DC motors are
the primary choice for space-based reaction wheels and have clear advantages over AC motors. Some of the
main reasons are:

1. Efficiency. DC motors typically have higher efficiency than AC motors and are available at higher
power/mass ratios.

2. Controllability. DC motors show a fast response and speed reversing is easily applied. Control is
generally more complex because it is implemented completely in the digital domain.

3. Wear. Wear in brushed motors will result in contamination of the satellite’s environment by metal
particles and/or graphite from the brushes.

For above reasons Brushless DC-motor (BLDC) are chosen. Additionally, brushless motors offer superior
performance compared to brushed motors. They have longer lifetimes and can achieve higher rotation
speeds. As a downside, they are slightly more expensive and require more complex controllers because
commutation is no longer performed mechanically like in brushed motors, but electronically.

Vacuum-resistance
An important requirement on the motors is the ability to operate in the vacuum of space. Standard COTS DC-
motors are not compatible with vacuum, because they contain liquid lubricant in their bearings. Liquid
lubricant would quickly evaporate in vacuum, resulting in bearing failure soon afterwards. A solution is to
order a special motor version in which the liquid lubricant is replaced with a vacuum-graded version. It is also
possible to replace the lubricants manually or have it done by a specialized company.

4.2 Selection criteria


Below the criteria are listed that influence the choice of Brushless DC Motor. Each criterion is briefly
explained.

Operating temperature range


The temperature range within which the motor is able to function should be compatible with the design
temperature range. Delfi-n3Xt will operate in space and will therefore be subjected to a vacuum. Additionally,
temperatures in space can be highly variable. The expected temperature range at ADCS-level is described in
the Thermal Design document [SLR0872, V2.2]. The design temperature range for the reaction wheels is
-15 °C – 45 °C.

Sensor
The motor should possess a sensor that provides information about the motor’s rotation speed, in order to
satisfy Requirement SAT.2.2.2.2.P.04.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 13 / 74

Controller
Brushless DC motors require a controller that phases electrical power into the motor coils. If the motor does
not have such a controller built-in, one has to be designed which requires additional effort.

Maximum power consumption


The power consumption Pin [W] is calculated using:

Pin  U  I Equation 1

Where U is the motor voltage [V] which can be controlled and where the motor current I [A] depends on
the rotation rate of the motor as stated in Faulhaber’s Technical Information document [SLR0979]:

U  kE n
I Equation 2
R

Where k E is the Back-EMF constant [V/rpm], n is the rotation rate [rpm] and R is the motor terminal
resistance [Ω]. From Requirement SAT.C.05 we have a preferred voltage of 3.3 [V] for the reaction wheel
motors because this would not require a voltage transformer. Additionally from Requirement SAT.C.03 we
know that originally the power budget was 400 mW. With three reaction wheels this translates in a maximum
power consumption of 133 mW. Budgeting 10% of the power consumption for the support electronics,
maximum 120 mW is available for each motor.

Maximum useful speed


The torque that can be generated by any DC motor decreases (approximately linearly) with rotation speed.
The maximum useful speed is reached when the generable torque falls below the minimum required torque
of 5.5 · 10−6 Nm (Requirement SAT.2.2.2.2.P.01). The theoretical maximum useful speed can be calculated
by setting the equation for the generated torque equal to (or bigger than) the minimum torque. Using this
rationale a performance curve was developed which gives information on both the power consumption and
the maximum rotational speed. Figure 4 gives an example of this curve and the theory behind the curve is
explained in more detail in the Reaction Wheel Test document [SLR0766].

0.25

0.2

0.15
Power [W]

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 4: Example of reaction wheel performance curve. It illustrates the required power to generate the required torque
of 5.5·10-6 [Nm] and shows a steep drop at the maximum speed for which the motor is able to deliver this torque.

A high nmax is beneficial, as it decreases the rotor inertia needed to achieve a given angular momentum
range.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 14 / 74

Estimated volume and mass


In order to comply with the original Requirement SAT.C.0.02, no motor dimension was allowed larger than 19
mm. Next to the motor itself, the flywheel that will be mounted on the motor will likely impact the volume
occupied by the Reaction Wheel System. An estimation of the flywheel size can be made using the maximum
useful rotation rate and the required angular momentum dynamic range. When the motor can be rotated in
both directions, the maximum required angular momentum in one direction is:

H req  0.5 1.56 10 3  0.78 103  Nms  Equation 3

An estimate of the required flywheel diameter and mass can be made from the equation for angular
momentum, the equation for mass moment of inertia for a solid cylinder and the equation for the mass of a
solid cylinder; respectively Equation 4, Equation 5 and Equation 6.

n0.9max  
H req   I zz  I r   z   I zz  I r   H req
30
30  H req
I zz   Ir Equation 4
n0.9max  

4
D
   t
2 Equation 5
I zz 
2

m   r 2t Equation 6

Where Ir is the moment of inertia of the rotor of the motor and n0.9·max = 0.9·nmax , with nmax the maximum
useful speed. In the preliminary design of the flywheel 90% of the theoretical maximum speed is considered
due account for the uncertainties at this stage of development.

Assuming a disk-shaped flywheel with thickness t = 0.003 [m] and using bronze Rg7 (with density ρ = 8800
kg/m3), the resulting diameter of the flywheel can be estimated by combining Equation 4, Equation 5 and
Equation 6:

60  H req 2  Ir
D  2 4  Equation 7
n0,9max  t
2
 t

D is the resulting flywheel diameter, which shall not be larger than 19 mm, because else the reaction wheels
for the X- and Y-axis would again be larger than the budgeted volume.

Additionally the formal mass budget is 120 grams for the reaction wheel system. Assuming that the reaction
wheels make up for 40% of the mass, whereas 60% is reserved for support structure and drive electronics,
than the maximum mass for each motor with flywheel is 16 grams.

Indication of the launch load on the axis


The motor should be able to withstand the mechanical loads that are generated during launch. Unfortunately
manufacturers provide little information about mechanical load resistance. Therefore the Root Mean Square
(RMS) value of the random vibrations during launch on the axle in axial direction as percentage of the
allowed load is used as an indicator for mechanical load resistance;

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 15 / 74

9.81 g RMS  m flywheel


RN  100% Equation 8
N shaft ,max

Delfi-n3Xt is launched with a Dnepr and the interface is managed by ISIS, which state an RMS value of 6.5 g.
[SLR1008, V0.9]. The maximum shaft load Nshaft,max is usually provided by motor manufacturers.

4.3 Motor options


In order to determine the most suitable motor, Table 1 lists various motor candidates for the Delfi-n3Xt
Reaction Wheel System and their characteristics of interest. Only BLDC motors are considered that:

1. Have built-in (Hall-)sensors

2. Are compatible with the design temperature range

3. Have no dimensions larger than 22 mm. In fact the original volume budget limits the RWS to 19 mm
height, but because some margin was available on that budget, also motors whose dimensions
slightly exceeded this budget were considered in the trade-off.

4. All identified motor options have the possibility to reverse the rotation direction, which is important
because it doubles the dynamic range of the motors.

Table 1: Brushless DC motors that are considered for the Delfi-n3Xt Reaction Wheel System

No- Maximum Max Specified


Integrated Back-EMF Terminal Torque Current Rotor
load dimensions Mass axial max
Motor control constant resistance constant max inertia
current lxbxh [g] shaft speed
electronics [mV/rpm] [Ω] [mNm/A] [A] [gcm2]
[A] [mm3] load [N] [rpm]

Faulhaber 12 x 12 x 41,740
none 0.094 16 0.902 0.028 0.199 1.1 0.125 1
1202 004 BH 5.44 (at 4V)
Faulhaber
12.5 x 12.5 6272
1307 004 BH none 0.105 16 1.005 0.026 2.05 2.1 0.16 5
x 14.08 (at 4V)
geared, ratio 6:1
Faulhaber 19 x 19 x 14,700
none 0.381 22 3.64 0.0174 0.174 6.9 0.69 15
1509 006 B 13.8 (at 6V)
Faulhaber 16 x 16 x 17,872
none 0.135 18.6 1.289 0.032 0.184 4.1 0.628 20
1608 003 BH 13.12 (at 3V)
Faulhaber
22 x 22 x 19,620
2209 T 005 digital 0.148 40 1.412 0.08 0.09 8.5 1 1
17.5 (at 5.5V)
BSC
Maxon 10 x 10 x 14,700
none 0.185 4.44 1.77 0.0588 0.5 3 0.143 15
EC 9.2 flat A 18.2 (at 3V)
Maxon 10 x 10 x 15,000
none 0.219 46.4 2.09 0.0151 0.111 0.82 0.08 6
EC 10 flat A 6.9 (at 4V)
Maxon 20 x 20 x 8960
none 0.617 6.67 5.88 0.0625 0.9 15 3.84 10
EC 20 flat A 18.5 (at 6V)
Portescap 22 x 22 x 7500
digital 0.59 8.2 5.63 0.071 0.6 32 17.7 50
22BT 21.5 (at 5V)

Indicates that the maximum dimension is outside the formal volume budget, but not enough for obvious rejection.

4.4 Motor selection


Initially the Faulhaber 2209 T 005 BSC had been selected because of its integrated digital speed controller
and a first prototype of the reaction wheel system was made (see V1.4 of this document). Vibration tests
(with loads exceeding the Dnepr launch loads) on that prototype unfortunately resulted in the loss of that
prototype, see the Reaction Wheels Vibration Test Series at NLR Report [SLR0868, V1.2]. It was decided not
to pursue a redesign using the same motor, but to use the Faulhaber 1202 004 BH despite the lack of
integrated control. This section will present the arguments that support this choice.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 16 / 74

Using the selection criteria outlined in Section 4.2 the motors introduced in Table 1 can be compared. To
determine the maximum useful rotation speed, the curves in Figure 5 are used. The torque is plotted based
on a maximum available power of 120 mW (see Section 2.1). The minimum required torque of 5.5 · 10−6 Nm
(see Requirement SAT.2.2.2.2.P.01) is also plotted. The Faulhaber 2209 T 005 BSC, the Maxon EC 9.2 Flat A,
the Maxon EC 20 flat A and the Portescap 22BT do not appear in Figure 5, because they cannot generate any
torque with 120 mW maximum available power.

The maximum useful speed is read as the rotation speed where the torque curve goes below the minimum
required torque. In the calculation of the maximum useful speed for the Faulhaber 2209 T 005 BSC it was
taken into account that it has built-in drive electronics for control. The power available for each motor
including drive electronics is 133 mW, but that also was insufficient for this motor to deliver any torque.

Figure 5: Graphs of available torque as a function of rotational rate, taking a maximum available power of 120 mW per
motor into account. Also the minimum required torque of 5.5 · 10−6 Nm is plotted. As recommended by Faulhaber, as a
first approximation the friction torque is considered constant over the full range of rotation speeds.

The maximum useful speed can be read from Figure 5 for the remaining motors. Also all other parameters
defined in Section 4.2 can now be calculated. The flywheel size is estimated using Equation 7 while the
resulting flywheel mass is calculated using Equation 6 and subsequently an indication of the launch loads on
the axis (as a percentage of the maximum specified shaft load) can be determined using Equation 8. The
results are combined in Table 2. It is clear that the volume requirement is the most stringent requirement,
when the maximum available power and required torque are considered as a given.

From the information on the motors that is available from the respective specification sheets (combined in
Table 1) and the comparison of the available motors in Table 2 it is concluded that the Faulhaber 1202 004
BH is the best option. With the available power budget it can effectively operate over a large speed range
(see Figure 5) resulting in a small diameter for the flywheel. The other realistic option is the Faulhaber 1608
003 BH, but it has two disadvantages over the Faulhaber 1202 004 BH:

1. The Faulhaber 1608 003 BH is designed to be operated at 3V while the comparison in the trade-off is
done for a supply voltage of 3.3V. When operated at its specified voltage of 3V the maximum useful
speed will therefore be below the value presented in Table 2. Additionally this would also require
additional electronics to transform the standard I2C supply voltage to 3V.

2. The Faulhaber 1608 003 BH has bigger dimensions than the Faulhaber 1202 004 BH. Considering the
limited volume in the Delfi-n3Xt, the latter is therefore a more sensible choice. Especially when the
fact is taken into account that also additional hardware and support structure have to be developed.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 17 / 74

Moreover the Faulhaber 1608 003 BH was not yet identified as an option during the initial motor choice, but
for above reasons the Faulhaber 1202 004 BH is still the best choice for the Delfi-n3Xt Reaction Wheel
System.

Table 2: Comparison of the motors on the basis of 120 mW maximum power consumption for the motor, (or in the case
of the Faulhaber 2209 T 005 BSC 133, mW) for the motor including drive electronics.
Mass
Maximum Flywheel Flywheel Indication of
(motor +
Motor useful speed diameter mass launch load on
flywheel)
[rpm] [mm] [gr] axis
[gr]
Requirement for full RWS
90 x 90 x 19 6.5 [g] RMS random
(three reaction wheels including - - 16 grams
[mm3] vibrations
support electronics and structure)
Faulhaber
29,300 18.0 6.7 7.8 50%
1202 004 BH
Faulhaber
1307 004 BH 4,550 28.9 17.3 19.4 25%
geared, ratio 6:1
Faulhaber
7,530 25.1 13.1 20.0 8%
1509 006 B
Faulhaber
19,400 19.4 7.8 11.9 4%
1608 003 BH
Faulhaber
0 - - - -
2209 T 005 BSC
Maxon
0 - - - -
EC 9.2 flat A
Maxon
11,300 23.0 11.0 11.8 13%
EC 10 flat A
Maxon
0 - - - -
EC 20 flat A
Portescap
0 - - - -
22BT

Value not compliant with requirements


Indicates that the value is outside the formal budget, but not enough for full rejection.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 18 / 74

5 Preliminary design
With the choice for the Faulhaber 1202 004 BH motor made, a preliminary design was made and a prototype
was developed to assess the preliminary design. To turn the Faulhaber motor into a reaction wheel, a
flywheel should be press-fitted on its shaft. This chapter will first describe the detailed design of that
flywheel, followed by a description of the prototype bracket. The focus during the prototype development was
on the development of the flywheel and the support bracket to quickly verify the primary requirements and its
capability to survive the launch loads.

For the development of the prototype, non-vacuum rated versions of the Faulhaber 1202 004 BH motors are
used, because lead times are shorter and costs are lower. It is argued that if the non-vacuum rated motors
do not meet the specifications, than the vacuum-rated versions will certainly not meet them because a
vacuum environment imposes special requirements on the motor bearing lubricants which generally degrades
the motor performance. This will be assessed in the engineering model.

5.1 Flywheel design and sizing


The function of the flywheel is to augment the moment of inertia of the motor rotor. At the same time any
other impact on the performance of the motor should be minimized.

Flywheel shape
In order to maximize the moment of inertia for a given flywheel mass, it is desirable to locate mass far from
the axis of rotation. In practice there is often a limit to the maximum size of a flywheel and therefore a design
with upright edges is commonly used. Such design is illustrated in Figure 6 together with a traditional solid
disk-shaped flywheel. For the RWS based on the Faulhaber 1202 004 BH motor a flywheel was developed
based on a solid cylinder for several reasons;

 Limitations on flywheel thickness. In order for the flywheel to keep its structural integrity, the
minimum thickness was set to 2 mm. This value is based on the opinion of several experienced
designers and manufacturers.

 Manufacturing tolerances. Any imbalance in the mass distribution of the flywheel results in
vibrations during operation and reduced efficiency. Because any additional feature on a flywheel will
increase its manufacturing complexity and thereby the risk of imbalance.

Figure 6: Two common flywheel design options. The left shows a simple disk-shaped flywheel, while the right shows an
alternative shape with more mass located away from the rotation axis

Flywheel material
Metals are a good class of materials to choose from as they offer high density, good manufacturability and
high strength. Bronze has been chosen, because it a non-ferromagnetic material that has a high density
compared to other metals. The DEMO workshop at the Faculty of Aerospace Engineering always works with

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 19 / 74

bronze supplied by Salomon’s Metalen B.V. Unfortunately no specific details about the bronze are available,
except that it is Rg7 bronze. The density was later calculated (see Table 3 in Section 6.1) as 8950 kg/m3
instead of the
8800 kg/m3 that was assumed in the initial calculations.

Flywheel size
Based on Requirement SAT.2.2.2.2.P.02, Equation 3 calculated that Hreq = 0.78·10-3 [Nms] in either direction
around the motor’s axis of rotation, which is considered the local Z-axis. The required flywheel diameter can
be estimated using Equation 7 where n0.9·max is 90% of the maximum useful speed as calculated in Table 2.
Therefore n0.9·max = 0.9 · 29,300 = 26,370 [rpm].

In Table 2 an estimate was presented of the required diameter for a simple disk-shaped flywheel with a
thickness of 3 mm. The resulting flywheel diameter was only 18 mm. However, it was later discovered that
the limiting factor for the size of the RWS were the drive electronics to support the motor (see Section 7.5)
and therefore a bigger flywheel could be designed. To optimize the diameter of the disk-shaped flywheel to
the available space, the allowed diameter for the flywheel is 20 mm. Rewriting Equation 7 as a function of the
new design variable (flywheel thickness t), gives:

 60  H req 2  Ir 
t  24  
n
   m  Equation 9
 0,9max   D  D 4
2 4

With diameter 20 mm and the other values as determined, the disk-shaped flywheel can therefore be
designed with a thickness of 2.0 mm, which is also the defined minimum thickness. More details on the
available space for the flywheel are discussed in Section 8.3. The final design of the flywheel is presented in
Section 8.1.

5.2 Bracket prototype design


The prototype bracket was designed to verify the primary requirements and whether the design has sufficient
structural integrity to survive the launch loads. The prototype bracket was designed such that:

 It could support three reaction wheels in three orthogonal directions.

 It could house a drive electronics PCB underneath each individual reaction wheel motor. The
preliminary design of the drive electronics PCB had a surface area of 13.4 x 13.4 [mm2].

 The bracket would allow for easy integration of the full Reaction Wheel System.

 The bracket would allow vibration testing using the shaker that is available at the 3ME Faculty.

5.3 Prototype assembly


As is explained in Section 8.2 in more detail, it was decided to integrate the required drive electronics
underneath the individual reaction wheel motor but when the prototype was developed, these drive
electronics and the additional required hardware were not yet available. Therefore the motors were mounted
on aluminum risers to create a comparable set-up.

Commercial reaction wheels often include a cover to protect the flywheel from contact with other objects and
damage to other objects after mechanical failure. Protective caps were therefore designed to protect and
contain the flywheel. The protective caps were designed with minimal impact on the interfaces with the
motor, such as electric connectors. More details on the final protective cap design are found in Section 8.4.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 20 / 74

Figure 7 shows a photo of the prototype bracket including two reaction wheels mounted on the 3ME shaker
using a specially developed adaptor disk. The bracket itself has a footprint of roughly 40 x 40 [mm2] and a
height of 27 [mm]. More details about the bracket can be found in Section 8.3 on the final bracket design.

Later the prototype was extended with mock-ups of the drive electronics and necessary adaptors to facilitate
the successful integration of the full assembly. A picture of the extended prototype is shown in Figure 8. It
was also realized that assembly takes significant time due to the small dimensions and the tendency of the
permanent motor rotor to attract M1 nuts. It was therefore concluded that additional effort should be put in
the assembly and integration of the engineering model. Details about these efforts and the assembly
procedure of the later developed engineering model can be found in the Reaction Wheel Assembly Manual
[SLR1007].

Figure 7: Photos of the prototype bracket with one two of the reaction wheels mounted some distance above the bracket
using aluminum risers.

Figure 8: Extended prototype with one reaction wheel and mock-up of the drive electronics. In this picture the protective
cap was left out.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 21 / 74

6 Verification of preliminary design


This section presents the results of the verification process of the Reaction Wheel System prototype. First the
performance verification is presented, followed by the assessment of the structural integrity and an analysis
of Reaction Wheel System mass, volume and power budget.

6.1 Verification of manufactured flywheels


Verification of the flywheels is necessary because the moment of inertia of the flywheels determines the
torque delivered to the satellite given a certain acceleration of the flywheel. Using the Mettler Toledo AB204-S
measurement unit available at the Physics Lab of the Faculty of Aerospace Engineering, the mass of a set of
flywheels was measured to verify the inertia of the flywheels. For verification of the inertia the dimensions are
supposed to be a perfect match with the Catia drawings and that mass is the only unknown parameter. With
the volume calculated by Catia as 6.665e-7 [kg] (see Catia part DNX-41-202), and the mass of the five
wheels measured; the expected bronze density can be determined.

The range of expected bronze densities thus represents the manufacturing uncertainties. In practice the
bronze density was constant for all flywheels because they were manufactured from the same piece of
material. However in keeping the volume of the flywheel constant, the manufacturing uncertainties are
expressed in density uncertainties. The measurements of the actual flywheel masses thus allow both
determination of the bronze density (taken as average of calculated values) and give an insight in the
manufacturing errors.

Feeding the expected bronze densities back to Catia, the moments of inertia are verified. From the flywheels
that were weighed the average moment of inertia determined by this process is 2.825·10-7 kg·m2 with a
standard deviation of 0.005 kg·m2.

Table 3: Flywheel inertia verification


Mass Expected bronze density Moment of inertia around Z-axis
Flywheel #
[g] [kg/m3] [kgm2]
1 5.964 8948 2.826·10-7
2 5.966 8951 2.827·10-7
3 5.949 8926 2.819·10-7
4 5.957 8938 2.822·10-7
5 5.980 8972 2.833·10-7

Therefore, assuming the error in the specified motor rotor inertia to be negligible, the combined inertia of the
motor and flywheel is 2.95·10-7 kg·m2 with a standard deviation of 0.005 kg·m2. Because the standard
deviation is only 0.15% of the total inertia, this variation is neglected in the remaining verification activities.

6.2 Motor performance verification


It is common practice to compare motors on the bases of their torque and power curves. In this section these
characteristic curves are established for the standard, non-vacuum rated version of the Faulhaber 1202 004
BH motor and compared to the theoretical curves to verify the motor performance. Additionally Figure 4
introduced a developed motor performance curve that will be used to assess the performance of the
Faulhaber motor. Test drive electronics were developed together with LabView software [SLR0771] to
perform this performance verification. The test drive electronics allow a reaction wheel to be accelerated to
their maximum speed using maximum power. Continuous measuring of both the rotation speed and current
during this spin-up gives all the information required to characterize the motor. This allows the characteristic
motor curves to be plotted. For more information about the LabView characterization software, see

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 22 / 74

[SLR0869]. A photo of the test set-up is included in Figure 9. Due to the short flex-rigid connector it is not
possible to fully characterize the motor when integrated with the reaction wheel bracket.

Figure 10 and Figure 11 show the summarized results of these performance verification tests. More details
about the characterization tests can be found in the Reaction Wheel Test Document [SLR0766]. Both figures
contain three types of data:

1. The specified curve: using the parameter values provided in the Faulhaber 1202 004 BH data sheet
[SLR0854] the theoretical curves are plotted for an input voltage of 3.3V.
2. The measured values: The torque can be calculated from the measured accelerations during the
spin-up.
3. Lower bound curve: Using the theory, the static friction torque and the maximum dynamic friction
coefficient are calculated that account for the lower bound of measured values.

Figure 9: A photo of the test drive electronics board for motor characterization including motor and flywheel. Also notice
the paperclip which was required to ensure good connectivity because the flex-rigid of the motor in this picture was
damaged during vibration testing.

Initially there was a big discrepancy between the theoretical and measured curves. It was discovered that the
permanent magnet rotor induced Eddy currents in the copper layer of the test drive electronics PCB. Eddy
currents generate magnetic fields that counteract the movement that created them. As such, magnetic
friction was greatly increased due to these Eddy currents. Removal of this copper layer solved this issue and
has resulted in the performance as presented in Figure 10 and Figure 11. The measurements that were the
input for these figures are included in Appendix A.

Conclusions on maximum power consumption


The measurements confirm that the non-vacuum rated motor is able to deliver the specified torque with the
specified power consumption. It can also be seen that the delivered torque (and output power) frequently
drops below the specified value, while the input power remains nearly constant. This is likely caused by
instantaneous variations in the experienced friction torque (expressed as friction current) described the
relation:

C0  Cv  n
I0  Equation 10
kM

As a first approximation, manufacturers provide I0 as a constant for the full range of rotation speeds.
Faulhaber specifies a no-load current of 0.028 [A] for the Faulhaber 1202 004 BH [SLR0854]. This matches
with the maximum values of the measured available torque as can be seen in Figure 10. The theoretical
values for the static friction torque C0 and dynamic friction coefficient Cv that correspond with this value of

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 23 / 74

are presented in Table 4. These interpreted values for C0 and Cv do not match with the specified values and
therefore the design will continue using the specified values. The specified value for kM is 0.902·10-3 [Nm/A],
which is fixed.

-4
x 10
1.8
Specified torque
1.6 Measured torque
Lower bound torque
1.4

1.2
Torque [Nm]

0.8

0.6

0.4

0.2

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
Figure 10: The theoretical torque curve and the measured torque for the Faulhaber 1202 004 BH. Measurements taken on
a single non-vacuum motor (representative for all motors as is discussed in SLR[0766]) conducted at room temperature
after vibration testing with 3ME shaker. The motor had not been operated for longer than 20 hours.

0.8
Specified P
in, max
0.7
Measured P
in, max
0.6 Specified P
out, max
Measured Pout, max
0.5
power [Watt]

Lower bound P
out, max
0.4

0.3

0.2

0.1

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
Figure 11: The theoretical power curves and the measured power consumption and –output for the Faulhaber 1202 004
BH. Measurements taken on a single non-vacuum motor (representative for all motors as is discussed in SLR[0766])
conducted at room temperature after vibration testing with 3ME shaker. The motor had not been operated longer than 20
hours.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 24 / 74

To account for the worst case scenario where the minimum measured torque is delivered, an adjusted value
for I0 is determined where the dynamic friction coefficient Cv is non-zero. An overview of the resulting values
to match the theory with the maximum and minimum delivered torque is presented in Table 4.

Table 4: Comparison of specified values for the friction coefficients and the coefficients determined from measurements
Specified by Interpreted Lower bound
Faulhaber value measured torque
Friction torque C0 [Nm] 0.003·10-3 0.026·10-3 0.026·10-3
Dynamic friction coefficient Cv 0.52·10-9 0 1.04·10-9
[Nm/rpm]
No-load current I0 0.028 (Equation 10 gives 0.028 0.028 – 0.062
(over the range 0 – 30,000 rpm) range 0.003 – 0.21)

Although the worst-case values for C0 and Cv introduced in Table 4 allow the worst-case friction situation to
be theoretically represented, this do not explain the observed variation. It was later hypothesized that the
spread in measurements can be accounted for by including manufacturing tolerances and axial play in the
shaft in the theoretical motor model. This hypothesis is verified in Section 8.7.

For the case where the friction coefficients match the measured worst-case lower bound torques, the effect
on the power consumption is assessed. The results are presented in the form of the performance curve
introduced in 4.2. Figure 12 shows the result and it can be seen that the power consumption can be up to
210 [mW]. At high speeds the required power therefore significantly exceeds the budgeted power of 120
mW.

Initially 133 mW was budgeted (see Section 2.1) for each reaction wheel, where 10% was budgeted for the
drive electronics. The drive electronics have been measured to consume around 11 mW, which is indeed close
to 10% of 133 mW. The motor power consumption on the other hand ranges between 120 mW and 210 mW
making the total power consumption range for the Reaction Wheel System 390-660 [mW]. It can therefore be
concluded that the power consumption will significantly exceed the power budget when all reaction wheels
are operated at a high rotation speed and the reaction wheels continuously experience the dynamic friction
coefficient corresponding to the lower limit of the observed performance.

0.25

0.2

0.15
Power [W]

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 12: Assessment of the required power to deliver the minimum torque with the static friction torque C0 = 0.026·10-3
and the dynamic friction coefficient Cv = 1.04·10-9. This matches the worst-case measured values during the preliminary
verification tests. Above roughly 24,600 rpm the motor is unable to deliver the minimum torque.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 25 / 74

Conclusions on maximum useful speed


From Figure 12 it can be concluded that assuming the worst-case observed dynamic friction coefficient, the
maximum rotation speed where the required torque can still be generated is slightly below 25,000 rpm. The
maximum useful speed can also be calculated using

Treq
U  (I0  )R
kM Equation 11
nmax 
kE [rpm]

Where I0 is expressed in Equation 10, Equation 11 can be expressed as a function of n0 as following:

k M U  R  C0  Treq 
nmax  Equation 12
k M k E  RCv [rpm]

With kM = 0.902·10-3 [Nm/A], U = 3,3 [V], R = 16 [Ω], Treq = 5.5 ·10-6 [Nm], kE = 0.094·10-3 [V/rpm] and the
friction coefficients as determined before, the resulting calculated worst-case nmax is 24,536, confirming the
observation in Figure 12.

It is therefore concluded that the non-vacuum rated motor is not able to reach the theoretical maximum
rotation speed of 29,300 rpm as introduced in Table 2. In fact, the maximum measured rotation speed during
the measurements was 28,800 rpm, but the motor was unable to operate stable at that speed. The dynamic
range is therefore calculated for the worst-case situation where the maximum rotation speed is 24,600 rpm.
In Section 6.1 the moment of inertia of the flywheel and rotor was verified as 2.95·10-7 kg·m2. Therefore the
dynamic range of the Reaction Wheel System along each axis is;

nmax  
H  2  I
30
H  1.52 10 3 [Nms]

The required dynamic range is 1.56·10-3 [Nms] and it can therefore be concluded that the current design
does not meet the requirement. But because a redesign would have a bigger impact than to continue with
this value that is slightly below the requirement, it was decided to continue with this design. In parallel the
cause of the observed variations in the delivered torque was investigated and an explanation is presented in
Section 9.3.

6.3 Preliminary verification of structural integrity


The prototype including two motors with flywheels was subjected to a moderate vibration test campaign
using the shaker available at the Faculty of 3ME. No observable performance degradation was observed,
except for a damaged flex-rigid connector on the motor mounted parallel to the bracket base. The flex-rigid
connector was not connected during the test (the drive electronics were not yet developed) which caused it
to go into resonance vibration which likely caused the damage. The preliminary conclusion was therefore that
the flex-rigid is the most vulnerable part of the motor.

6.4 Budget compliance


The prototype allows a good assessment of compliance with the budget. More elaborate discussion on the
mass and volume can be found in Section 9.1 covering the final design. Here the discussion will be limited to
the conclusions that were drawn from the prototype.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 26 / 74

Mass compliance
The prototype including three reaction wheels has a verified mass of 55 grams. With an estimated 50%
additional mass for the drive electronics and other additionally required support hardware, the mass of the
final design would be roughly 85 grams. This is well within the determined mass budget of 120 grams.

Volume compliance
The budgeted volume of the ADCS system is currently 90 mm x 90 mm x 32.6 mm. The height of the
Reaction Wheel System is determined by the bracket, which is 27 mm. However, as the flex-rigid connectors
extend above the Reaction Wheel System, the height is 29 mm. The footprint on the ADCS PCB is now 40
mm x 40 mm. Integration of all ADCS subsystems will be necessary to assess the compliance of this footprint
with the budget. At this point it was decided to combine the two PCBs that were budgeted for the ADCS could
be combined into a single PCB with a budgeted volume of 90 x 90 x 46.2 [mm3].

Power compliance
As was concluded in Section 6.1, the motor has a worst-case power consumption at high rotation speeds that
is more than twice the value specified by the manufacturer. This has a direct effect on the estimated power
consumption. It can be concluded from Figure 12 that up to 220 mW is consumed by a single reaction wheel
when the maximum torque is required and the reaction wheel has a high rotation speed. Of all reaction
wheels would operate in this region, the RWS total power consumption could reach 690 mW, which is far
above the targeted 400 mW.

If the lower observed torque bound of Figure 10 is considered, the worst case power consumption could go
up to 220 mW per wheel. Including the measured 11 mW for the drive electronics of each reaction wheel, this
translates in a worst-case total power consumption of the Reaction Wheel System of 690 mW. As there are
no feasible solutions conceived that would reduce the power consumption of the Reaction Wheel System this
is something that should be investigated in more detail. This will be done in Section 9.1 on the verification of
the final design.

6.5 Conclusion from prototype verification


This section briefly summarizes the conclusions and lessons learned during development and testing of the
prototype:

 The motor shall not be mounted directly above a conducting metallic surface, because else the
rotating permanent magnet stator will induce Eddy currents which greatly decrease motor efficiency.

 Verification of the flywheels is necessary because the moment of inertia of the flywheels determines
the torque delivered to the satellite given a certain acceleration of the flywheel.

 The mechanical design involves very small parts. Therefore significant effort is required to anticipate
on the assembly and integration phase with good design and the development of assembly tools if
required.

 Verification tests on the prototype have demonstrated that the maximum useful speed is below the
expected value. Where the design had accounted for 10% degradation in performance, resulting in a
maximum useful speed of 26,370 rpm the worst-case observed maximum useful speed was roughly
24,600 rpm. On the other hand the inertia of the manufactured flywheels was slightly higher than
calculated. The combined effect is that the worst-case dynamic range of the Reaction Wheel System
is slightly below the required value. It was also decided to investigate the cause of the variation in
dynamic friction investigated to assess its true impact.

 It was concluded from the verification tests that the power consumption will exceed the budgeted
power. This should be accounted for in a higher level of systems engineering as there are no
conceived, feasible solutions that would only affect the Reaction Wheel System.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 27 / 74

7 Reaction wheel drive electronics


A BLDC motor basically consists of a permanent magnet rotor and electromagnet stators where the current
through the electromagnets is determined by the configuration of a set of six MOSFETs. Often BLDC motors
are complimented with Hall sensors, which is also the case for the Faulhaber 1202 004 BH motor that is used
in the Delfi-n3Xt Reaction Wheel System. A schematic representation of a BLDC motor including Hall sensors
is presented in Figure 13. Using the position information from Hall sensors H1-H3, the configuration of
MOSFETs T1-T6 is determined such that electromagnets U-W drive the permanent magnet in the desired
direction. Thus, a BLDC motor is controlled by applying a voltage in combination with an appropriate
MOSFET-configuration.

Figure 13: the schematic representation of the electrical system for the Faulhaber 1202, which has four permanent
magnet pole pairs, three Hall sensors (H1-H3) and three coils (Coil U-W). In the schematic representation MOSFETs 3 and
2 are activated making the current run from Coil V to Coil U. The white dot on the rotor is the set as the reference point
for derivation of the commutation scheme consisting of the six steps shown above.

As was shown in the overview in Figure 3, the interface between the ADCS Electronics and the reaction wheel
motors is formed by the drive electronics, which also includes a microprocessor to manage and execute all
activities. These drive electronics were developed by Steven Engelen. The task of these drive electronics is
three-fold.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 28 / 74

1. Perform commutation

2. Handle I2C communication with the ADCS central microprocessor

3. Control motor torque, which is split up in several subtasks:

a. Perform speed determination

b. Interpret input commands from the ADCS central microprocessor into a useful form.

c. Deliver required torque by means of closed-loop PWM control.

7.1 Commutation
The switching between electro coils drive configurations is called commutation. Commutation is required
every time a Hall sensor switches state. Commutation is performed by means of switching between a set of
MOSFET-configurations which control how power is provided to the stator electro coils. With three electro
coils six different MOSFET-configurations are required to effectively drive the rotor in a specific direction. It is
therefore said that one electrical rotation consists of six commutation steps. Because the rotor of a BLDC
motor can consists of multiple magnet pole pairs, a mechanical rotation can also consist of multiple electrical
rotations. The Faulhaber 1202 004 BH consists of four pole pairs. The commutation scheme for such a four
pole pair BLDC motor is derived and presented in Appendix B for respectively the counterclockwise and
clockwise actuation. The white dot on the rotor in Figure 13 is set as the reference point for derivation of the
commutation schemes. The number of magnetic poles in the rotor also affects the step size and torque ripple
of the motor. More poles provide smaller steps and less torque ripple.

Commutation is implemented digitally in the Delfi-n3Xt reaction wheel system and is triggered by a Hall-
change. It is essential that commutation is performed directly after a Hall change is detected, because
otherwise the motor performance will decay drastically. So-called Hall-interrupts therefore have the highest
priority in the implementation of motor control. Especially at high rotation speeds these interrupts occur with
a very high frequency, because each mechanical rotation contains 24 Hall interrupts.

7.2 I2C communication


The reaction wheel microprocessors communicate with the ADCS microprocessor using the I2C protocol.
Table 6 lists the various commands that are required and how they are communicated using I2C in the format
used by a BusPirate.

The motor address that is required in each instruction listed in Table 6 is different for each of the three
motors. The respective addresses are listed in Table 5.

Table 5: Addresses for the three motors of the Reaction Wheel System
Orientation direction of motor I2C address
motor
X-direction 20
Y-direction 40
Z-direction 60

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 29 / 74

Table 6: I2C commands that are used in the communication with the reaction wheel microprocessor
Command Implementation in I2C
Turn on motor [ 0x 20 space 0x 01 ]
start hexadecimal motor motor on stop bit
bit identifier address
Turn off motor [0x 20 0x 02 ]
motor motor off
address
Set rotation direction [0x 20 0x 03 ]
to direction 1* motor set
address direction
Set rotation direction [0x 20 0x 04 ]
to direction -1* motor set
address direction
Set control setting [0x 20 0x 05 0x 3F 0x FF ]
motor open 1st byte of 2nd byte
address control new setting of new
register setting
Read motor variables [0x 20 0x 07 [ 0x 21 r:7 ]
motor open read motor read 7 bytes
address register address+1
*the resulting direction depends on the implementation in the microprocessor. As a convention, positive acceleration in the direction of
‘1’ corresponds to a positive torque on the satellite.

The 2 bytes of the new setting in the ‘set control setting’-command together form a 16-bit signed integer that
represents the torque through a value ∆n/s. The value for ∆n/s is calculated by the controller in the central
ADCS microprocessor.

The ‘read motor variables’ returns 7 bytes. However, only 2 bytes (together representing a 16-bit signed
integer) are needed to communicate the measured speed (after averaging, see Section 7.4) to the central
ADCS microprocessor. The location of these 2 bytes in the 7 byte read-out is not yet determined.

7.3 Motor control


The reaction wheel microprocessor considers the control command as a 16-bit signed integer. This input
command is defined as ∆n/s [rpm/s], which is the desired torque expressed in rotational speed change per
second. The reason to define the command input in this way is to have an integer input that requires only
limited additional processing in the drive electronics. By defining input as the delta rotation speed per second,
the relation with torque is maintained, which is the change in angular momentum per second.

The RWS control microprocessor needs to convert this control input into a commutation direction and input
power to the reaction wheels such that the desired torque is delivered. Torque is thus not directly controlled,
but by interpretation in a power setting. The power is set by means of Pulse Width Modulation (PWM). By
using a larger pulse width, the resulting (average) current running through the coil will increase, hence
resulting in a larger attractive or repulsive force on the permanent magnet rotor.

The only information which the microprocessor has about the reaction wheel system comes from the Hall
sensors. From the time between two Hall changes the rotation speed can be deduced. This rotation speed is
used to design feedback control loop as is shown in Figure 14.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 30 / 74

I RW
s

Figure 14: A schematic overview of the reaction wheel control loop

nref in the control loop is a variable reflecting the acceleration to generate the required torque. It is expressed
in the discretized version with equation:

 n 
nref  n0    t
*
[rpm] Equation 13
  t  adcs

Where (∆n/∆t)adcs reflects the required torque. n0 is the rotation speed of the reaction wheel at the moment
when a new (∆n/∆t)adcs command is received. ∆t* is the step size of the RWS controller.

The input variable (∆n/∆t)adcs has a sign corresponding to the desired torque. Thus, if a positive torque
around the X-axis of the satellite is desired, a positive value of (∆n/∆t)adcs is communicated to the RWS. In
order to match a positive torque with a positive rot_dir-setting (introduced in Appendix C presenting a
schematic of the microprocessor control), the relation between the acceleration directions that deliver positive
torque are derived from inspection of Figure 24. The results are presented in Table 7.

Table 7: Acceleration directions of the different reaction wheels corresponding with positive torque
Axis corresponding with motor direction of positive
torque on satellite
X-direction CW
Y-direction CW
Z-direction CCW

The microprocessor used in the Reaction Wheel System drive electronics is an 8 MHz Atmel ATmega164P. For
this microcontroller the PWM setting is an 8-bit variable. Because the PWM setting is determined using PID
control, this setting is regularly updated. In other words, modulated PWM is used to control the input power.

7.4 Microcontroller implementation


Commutation, I2C command handling and the interpretation of control commands are all implemented in the
same microcontroller, which has to perform these tasks in parallel to each other. Commutation has the
highest priority; I2C command and timer overflow interrupt handling have medium priority and the control
loop is performed in the remaining time. A high-level overview of the microprocessor implementation is
presented in Figure 15 and the full overview can be found in Appendix C.

Assuming an ADCS control loop with frequency 0.5 Hz, the time that the microprocessor is occupied with the
various tasks is estimated in Table 8. The estimation of commutation time is based on the number of
microprocessor operations per commutation and the maximum number of commutations (Hall changes) per
second. The I2C command handling time is estimated by assuming two I2C interrupts per ADCS control loop
(one read and one write) of maximum 10 ms. The remaining time is available for motor control, although
especially at low rotation speeds, the microcontroller will remain in idle mode for a large portion of that time.
After all, there is only need for a change in control input when a change has occurred, e.g. after a Hall
change interrupt.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 31 / 74

Δn = Δnin/frws;
nref = n + 2· discarded 
Commutate function TWI_flag = TRUE? YES bits (Δin/frws)
step_flag = TRUE;
TWI_flag = FALSE;

NO
HIGH PRIORITY INTERRUPT
(maximally each 60 μs, routine
takes only few clock cycles)
nref = nref + Δn
speed_flag = TRUE
step_flag = TRUE? YES step_flag = FALSE;
reset control loop
clock timer_rws

NO
Hall timer overflow
function

speed determination
MEDIUM PRIORITY INTERRUPT Hall_flag = TRUE? YES
function
(occurs after 0.5s if no Hall
change is detected)
NO

speed_flag =  YES PID function


2 TRUE?
I C function

NO
MEDIUM PRIORITY INTERRUPT
(roughly each 2 s, routine takes
maximally 10 ms) if timer_rws > 1/fstep
step_flag = TRUE;
PWM case function

speed_flag = FALSE;

MAIN LOOP (continuous loop, designed such that variables are updated only when required)
Figure 15: A high-level overview of the microprocessor implementation of reaction wheel operations. Commutation has
the highest priority, followed by I2C command handling. Reaction wheel control is performed in the main loop.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 32 / 74

Table 8: Estimation of time available for the three main tasks performed by the reaction wheel microprocessor
Task Percentage of time
Commutation 0 – 2.5 %
I2C command handling ~1%
Motor control < 96.5%

The Atmel ATmega164P has four hardware counters. They are assigned to the following three tasks:
1. Measure time between a pre-defined amount of Hall changes.
2. Implement PWM on the motor power.
3. Control the reaction wheel control update frequency

Microprocessor settings
Each mechanical rotation consists of 24 Hall changes and the rotation speed is deduced from the time
between a pre-defined number of these Hall changes. Therefore the accuracy of speed determination is
directly related to resolution of the timer that is part of the microprocessor. The microprocessor runs at 8
MHz, but clock dividers can be set to adjust the time steps of the counter. Table 9 presents the available
options that were identified. Options where the lowest measurable speed is larger than 100 rpm were not
considered.

Table 9: Microprocessor setting option and impact on speed determination


at 100 rpm at 25,000 rpm
Speed  Instantaneous   Speed determination 
determinations  Lowest  Speed  speed  accuracy after 0.1 
Clock  per mechanical  measurable  measurement  determination  determination  second averaging 
divider  rotation  rotation speed  time [s]  frequency [Hz]  accuracy [rpm]  [rpm] 
8  24  38  0.03  1.0E+04  250  1.0 
64  4  29  0.15  1.7E+03  350  3.0 
64  24  5  0.03  1.0E+04  2,000  2.0 
256  1  29  0.60  4.2E+02  350  9.0 
256  4  7  0.15  1.7E+03  1,300  8.0 
256  24  1  0.03  1.0E+04  13,000  13.0 
1024  1  7  0.60  4.2E+02  1,300  32.0 
[] Unacceptable
[] Significant performance impact expected
[] Minor performance impact expected

The option with clock divider 64 and 4 measurements per mechanical rotation is selected for various reasons.

 Hall changes between consecutive speed determinations. There are inevitable errors
associated with the misalignment of Hall sensors in the motor and symmetry of permanent magnet
distribution in the motor rotor. The permanent magnet errors have no influence if a speed
measurement is done once per mechanical rotation. The Hall misalignments have no influence when
a speed measurement is done once per electrical revolution, i.e. six Hall steps. Because software
provided by Atmel also uses the setting ‘once per electrical revolution’-setting, it is assumed that
permanent magnet misalignments are no major concern.
 Measurement time. At low speeds measurement times can become quite long. 0.1 seconds is only
5% of the time between input commands, which is acceptable. 0.6 seconds on the other hand is
considered unacceptable to develop a responsive controller.
 Speed determination frequency. Because the motor controller is an interrupt-based controller,
there is no control over the amount of time dedicated to specific tasks. The speed determination
frequency determines the maximum allowed time for the controller to execute all required tasks,

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 33 / 74

because the behavior become significantly more unpredictable when tasks exceed this specified limit.
If the speed is determined between every single Hall change, resulting in a speed determination
frequency of 104 Hz, there is only time for approximately 800 instructions in between. 800
instructions are not sufficient to state with certainty that all possible tasks and interrupt routines
could be executed. With a speed determination frequency of 1.7·103 this is certain and option that
result in this speed determination (or a lower one) are therefore preferred.
 Speed determination accuracy. Requirement SAT.2.2.2.2.P.06 states that angular momentum
should be determined with accuracy 2·10-8 Nms, which corresponds to 0.6 rpm. However, only
integers are allowed, making the maximum theoretical accuracy 1 rpm. Averaging over 0.1 seconds,
where the instantaneous error reduces by a factor 1/N (with N the number of speed determinations
in 0.1 seconds) leads to a maximum achievable accuracy of 1 rpm. Although clock divider 64 and 6
Hall changes between measurements gives an estimated accuracy of 3 rpm, it is expected that the
effects can be sufficiently mitigated with an adequate controller.

Because the effects of magnet misalignment are mitigated, there is a safe margin on the numbers of
instructions possible between speed measurements and speed determination accuracy is manageable, the
option of clock divider 64 and 4 speed determinations per mechanical rotation is decided on.

Speed determination accuracy after averaging


Due to limitations in the microprocessor only integer-calculations are allowed, which affect the accuracy of
the reaction wheel control. The averaging is implemented in the controller as;

SpeedMeasure = SpeedMeasure - (SpeedMeasure>>kAVG) + (speedtemp>>kAVG)

To determine the average speed, the kAVG least significant bits of the current average speed are discarded
and subsequently the kAVG least significant bits of the instantaneous rotation speed are added. This results in
a running average over the previous kAVG measurements. kAVG is designed as a parameter that depends on
rotation speed for two reasons:

1. Maximum integration time. As was introduced in Table 9, the speed determination accuracy after
averaging is limited by the integration time of 0.1 seconds. This value is determined by some trial and
error; smaller integration times would result in a significant decrease of the resulting accuracies,
whereas longer integration times decrease the responsiveness of the system. Furthermore, 0.1
seconds is also the measurement time at 100 rpm, making the integration time more or less constant
over the full speed range.
2. Instantaneous measurement accuracy. Because the kAVG least significant bits of each
instantaneous measurement are discarded, 2 should not be higher than the instantaneous
measurement accuracy.

Taking above considerations into account the resulting optimal values for kAVG are listed in Table 10. The
speed determination accuracy before and after applying the running average over 0.1 seconds is presented in
Figure 16.

Table 10: Values for kavg and their respective speed ranges
Speed range kavg
> 19,132 7
10,901 – 19,132 6
7,716 – 10,901 5
5,419 – 7,716 4
3,858 – 5,419 3
2,690 – 3,858 2
1,929 – 2,690 1

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 34 / 74

500 5

400 4
Accuracy [rpm]

Accuracy [rpm]
300 3

200 2

100 1

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
Rotation speed [rpm] Rotation speed [rpm]
Figure 16: Instantaneous speed determination accuracy (left) and the speed determination accuracy after applying the
running average (right).

Additional effects of the discrete system


Besides the effect on the achievable speed determination accuracy, there are other factors which impact the
design. Table 11 lists the main additional effects and their impact.

Table 11: Sources of reduction in the control accuracy


Description Explanation Impact
Division of control In Equation 13 it was introduced that the input from To compensate, the
input the central ADCS processor is incrementally translated discarded bits are added
into nref. Dividing (∆n/∆t)adcs by the RWS control in the first increment step
frequency results in a certain number of bits discarded. after a command update.
Gains limited by 16- Use of integer number drastically reduces the Original gains with
bit numbers resolution with which gains can be implemented. A maximum resolution of
scaling factor of 27 is applied because starting from a 1/(27).
16-bit unsigned integer this results in 8-bit signed
integer PWM resolution.
Look-up table for The measured counter value between six Hall changes With the available RAM
instantaneous (four measurements per mechanical rotation) relates memory in the
speed measurement to rotation speed according to: 60 f CLK / 64  2  4 .  16
 microprocessor, a look-up
table can be constructed
This calculation cannot be performed by the which requires
microprocessor using 16-bit integers only and interpolation between
therefore a look-up table is used to transform the values and a maximum
counter value into an instantaneous speed inaccuracy of 2 rpm.
determination. This also saves many instructions in the
microprocessor.
Resolution of Because derivative control is Kd·( ref), it Maximum initial resolution
derivative control depends on the achievable resolution of ref. This is in of derivative control is
first instance determined by the control frequency equal to the RWS control
frequency
PWM bit-depth The PWM bit-depth determines the numbers of steps Voltage control resolution
between 0 Volt and 3.3 Volt applied to the motors of 3.3/28 [V].

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 35 / 74

Failure cases
As was introduced in Requirement SAT.2.C.01, a failure in the reaction wheels should not pose a danger to
the satellite as a whole. The operational risk of the reaction wheels is that a large torque is suddenly applied
to the satellite. The following scenarios could reasonably occur:

1. A reaction wheel suddenly blocks electronically. This failure can be mitigated by implementing
a piece of code that puts a reaction wheel in ‘free-spinning mode’ when e.g. the microprocessor runs
into an error and needs to reboot.

2. A command is wrongly interpreted, causing the reaction wheel e.g. to start accelerating rapidly.
This can be mitigated by implementing a maximum acceleration in the microprocessor algorithm.

3. A reaction wheel suddenly blocks mechanically. If damage or a bad design could result in
sudden blocking of a reaction wheel, all angular momentum would instantaneously be transferred to
the satellite. In the design attempts are made to avoid this in any situation, but also a calculation is
made of the effect on the satellite in case this would happen at maximum speed. The satellite has the
smallest inertia around the Z-axis, namely 0.017 kg·m2. The change in rotation rate of the satellite
ωDNX due to a sudden stop of a reaction wheel with inertia 2.95·10-7 kg·m2 at a rotation speed of
25,500 rpm would be:

nmax I RW
 DNX   360  2.7[deg/ s]
60 I zz , satellite

This rotation is well within the limits of what the Delfi-n3Xt detumbling algorithm can handle.
Therefore a blocking reaction wheel will not pose a threat for loss of the full satellite due to sudden
spinning up of the satellite.

7.5 Drive electronics PCB


The functions that were explained in this Chapter are all integrated onto a single drive electronics PCB with a
size of 13.4 x 13.4 [mm2]. Three interfaces are included on the PCB, which are also indicated in Figure 17.
The height above and below the PCB substrate is never above 1.5 mm. The PCB itself has a thickness of 1.65
mm.

Figure 17: Picture of the drive electronics PCB for the Delfi-n3Xt Reaction Wheel Sytem with the various interfaces
indicated.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 36 / 74

8 Engineering model
After the initial verification of the design using the Reaction Wheel System prototype, a full engineering model
was developed including the drive electronics as introduced in Chapter 7. This chapter first introduces the full
design followed by more elaborate descriptions of the separate parts in the assembly. The Catia models and
technical drawings can be found in [SLR0927].

Despite the compact design, the Reaction Wheel Systems allows for easy integration thanks to a stepwise
assembly approach consisting of the following steps:

1. Press-fitting of the flywheels on the Faulhaber 1202 004 BH motors

2. Integrating the motor-flywheel combinations with the drive electronics in separate


subassemblies

3. Mounting of the subassemblies on the reaction wheel bracket

The full assembly procedure is explained in the Reaction Wheel Assembly Manual [SLR1007]. Additionally the
resulting Reaction Wheel System is designed such that it can be integrated on the ADCS PCB using only a nut
and bolt.

The engineering model and all individual parts and components are shown in Figure 18. They will be
discussed individually in the succeeding section; not in the order that they are presented in Figure 18 but in
the order that they were designed.

Figure 18: The Reaction Wheel System engineering model with all components and parts indicated.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 37 / 74

8.1 Flywheel
With the input on the required flywheel thickness and diameter determined in Section 5.1, a model of the
flywheel was developed in Catia. This design is stored under part-number DNX-41-202 [SLR0927].

In order to maximize the contact area between the flywheel and the motor axle and to allow the possibility to
apply some glue on top of the axle, the simple disk-shaped flywheel was complemented with a ‘dome’ as can
be seen in Figure 19. The main characteristics of the flywheel are summarized in Table 12.

Figure 19: The flywheel that was designed for the Delfi-n3Xt RWS together with its Catia model

Table 12: Main characteristics of the Reaction Wheel System flywheels


Parameter Value [unit]
Diameter 20 [mm]
Mass 6 [gr]
Inertia 2.95·10-7 [kg·m2]
Maximum angular momentum 7.88·10-4 [kg·m2/s]
at 25,500 rpm, including the motor
rotor inertia

Fixation to motor axle


The flywheel is fixed to the motor axle using a press-fit. A tool was designed for this assembly process. The
parts for that tool are stored under part number DNX-91-903 to DNX-91-905 and DNX-91-908 [SLR0927].
Figure 20 shows a picture of a flywheel press-fitted on a motor.

Figure 20: A Faulhaber 1202 004 BH motor with a flywheel press-fitted on its axle.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 38 / 74

8.2 Integration of the drive electronics PCB


The drive electronics PCB was already introduced in Section 0 and will not be repeated here. This section
deals with additional support hardware that was developed to integrate the drive electronics PCB. As can also
be seen in Figure 20, each reaction wheel is connected to the required drive electronics that are excluded
from the motor itself using a flex-rigid connector. Because of the limited length of this connector, namely 24
mm [SLR0854], a successful attempt was made to integrate individual drive electronics underneath each
reaction wheel. This also increased the modularity of the design. As is visible in Figure 18, some additional
parts had to be developed to accommodate this, namely:

1. Drive electronics adaptors. The motor could not be mounted directly on the drive electronics PCB,
but some distance above it to accommodate the required components on the drive electronics PCB.
Additionally the drive electronics PCB has to be fixed to the bracket. Therefore an adaptor was
developed to connect the motor to the drive electronics. Additionally the adaptor forms a support
platform for the thin motor base and a load path between the motor and the bracket, minimizing the
load on the drive electronics PCB. The technical drawing for this part can be found under drawing
number DNX-41-205 [SLR0927]. Figure 21 shows a picture of the adaptor together with the 3D
model.

Figure 21: A picture and 3D model of the drive electronics adaptor

2. Risers: Because the drive electronics PCB has components on both sides, this too has to be mounted
some distance above the bracket itself. The height of the risers is 2 mm. The technical drawing for
this part can be found under drawing number DNX-41-204 [SLR0927].

8.3 Bracket
The main purpose of the bracket is to provide a well-aligned, rigid platform for all three reaction wheels that
make up the RWS. Figure 22 shows a picture of the bracket together with some risers on which the drive
electronics are mounted.

The most prominent feature is the series of little pillars on which the protective caps are mounted. They
greatly influence the size of the bracket, because they must be outside the envelope occupied by the drive
electronics. Additionally they have to be large enough to allow a threaded hole to connect the protective caps
to the bracket. Because aluminum is a soft metal, it was decided to go for a helicoil solution instead of simply
threading the hole.

Another limitation to the design is due to a manufacturing limitation, which does not allow sharp, blind
corners. Such corners have to have a minimum radius of 3 mm. This is also the reason that the protective cap
for the reaction wheel that is mounted parallel to the bracket base, is only mounted on three points. This is
not visible in Figure 22, but it can be seen in Figure 18. If the far corner would also require a pillar, the
design would occupy more volume.

The bracket was designed such as to simplify manufacturing as much as possible to decrease the risk of
errors and reduce cost.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 39 / 74

Figure 22: A picture of the Reaction Wheel System bracket accompanied by its 3D model

8.4 Protective caps


Because the earlier reaction wheel design did not survive the heavy vibration loads to which it was exposed,
the effectiveness of a protection cap was successfully investigated. Figure 23 shows a picture of the
protective caps together with their 3D model. The purpose of these protection caps is:

 To limit the displacement of the reaction wheel assembly and therefore the axial loads. Because
reaction wheels are placed in all three axis of the satellite, the launch loads will also be applied in a
direction perpendicular to the motor axis for two of the reaction wheel motors. Nothing is known
about the allowed loads in those directions.

 Contain the separate parts of the reaction wheel in case of disintegration, so that no additional
damage is done to other hardware.

Figure 23: A picture of the protective cap for reaction wheels mounted perpendicular to the ADCS-PCB (full square) and
for the reaction wheel mounted parallel to the ADCS-PCB (square with cut off corner) accompanied by their 3D models.

8.5 Engineering model assembly


Although assembly of the engineering model is not a trivial exercise, it is not discussed in detail in this report.
A dedicated document elaborating on the assembly of the Reaction Wheel System was developed for that
purpose [SLR1007]. Figure 24 demonstrates the alignment of the Reaction Wheel System with respect to
satellite body axes.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 40 / 74

Figure 24: Alignment of the Reaction Wheel System with respect to satellite body axes

Additionally errors in the manufacturing and assembly process could result in reaction wheels with a mass-
distribution that is not axisymmetric around the motor axis of rotation. Such errors could result in periodic
disturbance torques. Both the alignment error and axisymmetric error are analyzed in the remaining sections
of this chapter.

8.6 Assembly alignment error


The Reaction Wheel System is a mechanical system and tolerances on the alignment therefore have to be
considered. From Requirement P.03, the maximum allowed systemic disturbance torque is 10−7 Nm.
Additionally, from SAT.2.2.2.2.P.01 we have that the maximum required torque equals 5.5·10−6 Nm. From
this we can calculate the requirement on the alignment of the reaction wheel motor along the principal axis:

 TD , sys 
 ,max  tan 1    1.0 Equation 14
 max 
T
[deg]

Alignment of each reaction wheel motor shall therefore be better than 1.0 [deg] with respect to each principal
axis. This maximum assembly alignment error translates in design tolerances for the following sources of
misalignment in the full design:

1. Alignment of the ADCS PCB with respect to the principal axes of the body reference frame.

2. Alignment of the base surfaces of the reaction wheel bracket with respect to the ADCS PCB.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 41 / 74

3. Alignment of the three motors with respect to the base surfaces of the reaction wheel bracket

These three sources of alignment errors are visualized in Figure 25 and discussed below. The maximum total
alignment error is αε,max = 0.7 + 0.15 + 0.6 = 1.45 [deg], which is higher than the required
1.0 [deg]. As a result the maximum expected systemic disturbance torque is;

TD , sys ,max  Tmax  tan   ,max  1.4 10 7 [ Nm]

It is not considered realistic to push the tolerances any further and therefore the maximum expected systemic
disturbance torque is higher than the derived requirement.

Figure 25: Visualization of the three sources of the alignment errors

Alignment of the ADCS PCB with respect to the principal axes.


Alignment of the ADCS PCB is determined by the precision with which it is mounted in the Delfi-n3Xt PCB
stack. The ADCS PCB itself is mounted on the threaded rods that extend over the full length of the satellite
and which are considered perfectly aligned with the Z-axis of the body reference frame. It is expected that
the maximum accuracy that can be achieved in positioning a PCB on each rod is 0.5 mm. The maximum
misalignment then occurs when a PCB is mounted 0.5 mm too low in one corner and 0.5 mm too high in an
adjacent corner. Threaded rods are spaced 82 mm apart and the potential misalignment of the ADCS PCB
therefore is:

 2  0.5 
 PCB  tan 1    0.7[deg]
 82 

Alignment of the base surfaces of the reaction wheel bracket


In order for the base surface to be well aligned, design tolerances are applied to the design of the bracket.
The base surface of the bracket could be e.g. not completely parallel to the surface that is mounted on the
PCB. Additionally, in case of the reaction wheels in the X- and Y-direction, the base surface could be not
completely perpendicular to the PCB or not completely flat. With the design tolerances applied to the
technical drawings [SLR0927], which have been determined in collaboration with the DEMO workshop, the
maximum expected misalignment of either of the bracket base surfaces is εbase surface = 0.15°.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 42 / 74

Alignment of the motor with respect to the base surface


The reaction wheel subassembly is aligned on the basis of the measured distance between the motor and the
base surface by means of a customized tool (see [SLR1007] for more details). The maximum allowed
variation of this distance as a result of this manual alignment process is 0.1 mm. The motor base PCB is 10 x
10 mm2 and as such the maximum error in the alignment of the motor with respect to the base surface is:

 0.1 
 motor  tan 1    0.6
 10  [deg]

Figure 26: The axle center of rotation is assumed to be positioned halfway the bearing. The axial play is determined at
the bearing exit, thus located 0.74 [mm] from the assumed center of rotation (adapted from Faulhaber Series 1202
Brushless Flat DC Micromotors, [SLR0854]).

8.7 Axisymmetric error


An axisymmetric error in the motor-flywheel assembly results in periodic forces on the satellite. Such periodic
forces have a negative impact on the performance of the Reaction Wheel System and affect the satellite
performance. In compliance with requirement SAT.2.2.2.2.P.04 the resulting disturbance angular momenta
shall not exceed 2·10-7 Nms. But because angular momentum is torque integrated over time, first the periodic
forces resulting from axisymmetric errors must be analyzed. They are dominated by:

1. A motor axle that is not perfectly aligned with the axis of rotation

2. A flywheel that is not perfectly axisymmetric

These two factors are also indicated in Figure 27.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 43 / 74

Figure 27: The two contributing factors to the axisymmetric error that causes periodic disturbance torques

Disturbance accelerations
To assess the impact of misalignments on the satellite, the accelerations of the flywheel must be projected in
the satellite body reference frame. These accelerations are derived in Appendix D as:

   r M (t )       r M (t ) 
aM,cog (t )     , cog    , cog Equation 15

M
Where r ,cog (t ) is a time-dependent vector which expresses the center of gravity in the body-fixed motor
 are the rotational speed and acceleration of the rotor with respect to the body-
reference frame.   and  
fixed motor reference frame in the presence of an off-set of the axis of rotation over an angle εaxle. Therefore
 
aM,cog (t ) is the time-dependent acceleration of the vector rM,cog as observed in the body-fixed motor reference
frame.
M
Newton’s Second and Third Law of Motion can now be used to assess the force FFW on the motor at the
motor origin, which is the point where the axle enters the motor. It is important to remember that the motor
reference frame is fixed with respect to the satellite, because this means that the forces experienced by the
motor are in fact experienced by the satellite, assumed that the motor is rigidly fixed to the satellite. Thus:
M 
FFW (t )   maM, cog (t ) Equation 16

Where m is the mass of the flywheel. The moment on the motor origin (and therefore the satellite) can be
assessed in a similar way, namely:

 M
M 0M  r M, cog (t )  Fcog (t ) Equation 17

From Equation 35 in Appendix D we know;

  FW  M
rM,cog (t )  R ,axle  R (t ) R ,cog rcog  r0 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 44 / 74

  FW
r0M   0 0 0.74 103 [m] [SLR0854] as derived in Section 8.6. rcog   0 0 1.304 103 [m],
T T
With
which is obtained from the technical drawing for the flywheel DNX-41-202 in [SLR0927] where the base of
the flywheel is positioned a distance 0.2 [mm] away from rotor base, as indicated in Figure 26.

Resulting disturbance moments


Using these equations the forces and moments on the satellite caused by axisymmetric errors in a reaction
wheel can be calculated. The worst case scenario is when maximum torque is applied at the maximum
rotation rate with a maximum deflection εaxle of the rotation axis. Both the torque and rotation are assumed to
   0 0   and 
   0 0  T . The worst-case
T
be perfectly aligned with the motor Z-axis; i.e.
values for  ,  and εaxle are introduced in Table 14

The only remaining variable is the axisymmetric error εcog which is due to imperfections in the manufactured
flywheel. Table 13 lists the main contributing factors to the axisymmetric εcog and the estimated magnitude of
their contribution to εcog. The total estimated value of εcog is therefore 0.40°.

Table 13: Estimate of axisymmetric error εcog


Feature Value of maximum tolerance estimated impact on εaxle
Concentricity 0.01 [mm] 0.26°
Total runout 0.01 [mm] 0.06°
Total: 0.32°
25% uncertainty 0.08°
Estimated εcog 0.40°

In Figure 28 and Figure 29 the resulting disturbance moments on the origin of the body-fixed motor reference
frame are calculated using Equation 17 for the axisymmetric error εcog angle estimated in Table 13 as well as
for errors that are slightly higher and lower.

Table 14: Worst-case values as input for the periodic disturbance torques.
Parameter Worst-case situation Resulting value
n
  [rad/s]
 30   2618 [rad/s]
nmax  25, 000 [rpm]
dH T
T  I    
 dt I   18.6 [rad/s2]
7
I  2.95 10 [kgm ] Tmax  5.5 10 6 [Nm]
2

  axle 
 axle  tan 1  
 r0 
εaxle  axle is the radial axle play on the bearings at  axle  0.85 [deg]
measured at the bearing exit [SLR0979],
 axle  0.011 [mm] [SLR0854]

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 45 / 74

Figure 28: Graphs of the resulting moments on the origin of the body-fixed motor reference frame for a rotation speed of
25,000 rpm, maximum torque applied and  axle =0.85 [deg] for one rotation.

Figure 29: Graphs of the resulting moments on the origin of the body-fixed motor reference frame for a rotation speed of
8,000 rpm, maximum torque applied and  axle =0.85 [deg] for one rotation.

Resulting disturbance momentum


It can be seen from these graphs that the combination of the manufacturing error and radial play in the
rotation axle results in disturbance torques along all three axial directions, although the disturbance torques
around the X- and Y-direction are significantly higher than those around the motor Z-axis. In fact, for the
estimated εcog of 0.4 [deg] the amplitude of the resulting periodic disturbance torque are as high as 8·10-4
[Nm]. However, these disturbance torques are periodic (with a frequency up to 400 Hz) which will limit the
net-effect on the satellite. The periodic disturbance torques vary over time;

M 0M (t )  M 0,max
M
sin t  [ Nm] Equation 18

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 46 / 74

These disturbance torques M 0 (t ) act in the local motor reference frame and ω is the rotation rate [rad/s].
M

In Figure 30 the maximum amplitude of the resulting instantaneous periodic disturbance torques around the
motor X- and Y-axes is plotted as a function of frequency. The frequency is actually the rotational speed of
the reaction wheel. Therefore at any moment in time three periodic disturbance torques are acting on the
satellite.

x 10
-4 X- and Y- axis x 10
-5 Z- axis
7 1

6
0.8
5
Amplitude [Nm]

Amplitude [Nm]
4 0.6

3 0.4

2
0.2
1

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]
Figure 30: The maximum amplitude of the periodic disturbance torques acting on the motor origin induced by a single
reaction wheel. The axes are expressed in the local motor reference frame.

Next, in order to analyze the worst-case scenario, it is assumed that the periodic disturbance torques are fully
transferred to the satellite. Therefore the satellite will also vibrate. The maximum value of the disturbance
momentum is when the disturbance torque is integrated over a half period. The amplitude of the maximum
disturbance torque was calculated as 8·10-4 [Nm] at a rotation speed of 25,500 rpm. The resulting
disturbance momentum H D ,max is then obtained by integrating Equation 18 over a half period;



H D ,max   M 0,max
M
sin t  dt  9.5 107 [ Nms ]
0

In the worst case where two reaction wheels are operated at the same high rotation speed, the local X-
component of the disturbance torque of one reaction wheel could be in phase with the Y-component of the
disturbance torque of a second reaction wheel. In that case the resulting disturbance angular momentum will
be 1.9·10-6 [Nms], which is one order of magnitude higher than the required 2·10-7 [Nms].

Impact on other subsystems


Because the reaction wheel rotation speed is directly related to the frequency of the identified disturbances,
disturbance torques have a frequency up to 400 Hz. Figure 30 introduced the disturbance torques as a
function of frequency. However, the linear disturbance accelerations on the satellite are a better indication of
the impact on other subsystems. Using the relation

mFW
asat   aFW Equation 19
msat

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 47 / 74

aFW is calculated using Equation 15, the mass of the flywheel is 6 grams and the mass of the Delfi-n3Xt
satellite is approximately 3 kg. Figure 16 illustrates the expected disturbance accelerations in the satellite as a
function of the rotation speed of a reaction wheels. These disturbance accelerations are significant and have
resulted in the removal of gyros from the ADCS. These disturbance accelerations (and the resulting torques
and moments) should be carefully considered in future reaction wheel design.

X- and Y- axis x 10
-3 Z- axis
0.12 2

0.1
1.5
Amplitude [m/s ]

Amplitude [m/s ]
0.08
2

2
0.06 1

0.04
0.5
0.02

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]
Figure 31: The maximum amplitude of the periodic linear disturbance accelerations in the satellite induced by a single
reaction wheel. The axes are expressed in the local motor reference frame.

Impact on satellite attitude


In order to assess the impact of this periodic angular momentum change on the satellite attitude, the
resulting angular amplitude of these vibrations is assessed using the relation;

M  I DNX DNX Equation 20

Where DNX is the (resulting) angular velocity of Delfi-n3Xt. Rewriting Equation 20 to express DNX and
integrating the resulting equation once gives an expression for the resulting maximum angular deflection
 DNX ,max ;


 M
M 0,max cos t  180
 DNX ,max    dt
 I DNX 
0 [deg]

This integral can be solved with equation;

2  M 0,max
M
180
 DNX ,max  2 [deg] Equation 21
 I DNX 

Where  DNX ,max [deg] is the maximum angular deflection of Delfi-n3Xt due to the periodic disturbance torque,
[Nm] is the amplitude of the periodic disturbance torque, ω is the rotational speed [rad/s] and I DNX is
M
M 0,max
the smallest moment of inertia of the satellite [kgm2]. It can be deduced from Equation 15 and Equation 17

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 48 / 74

that this maximum angular deflection is actually constant as a function of ω, because M 0,max is roughly
M

proportional to ω2. Using I DNX  0.017 [kgm2],  DNX ,max is calculated as 3.8·10-7 [deg]. This is negligible in
comparison to the noise of the attitude determination sensors.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 49 / 74

9 Verification of the final design


In this chapter the final design in the form of the engineering model is verified with respect to the
requirements introduced in Chapter 2.

9.1 Verification of general requirements


In this section the general and functional requirements are listed and tested against the final design.

Requirement Short description Target Actual


SAT.C.01 Mass budget 120 grams 82 grams
SAT.C.02 Volume budget 90 x 90 x 46.2 mm3 90 x 90 x 34.6 mm3
SAT.C.03 Power budget 400 mW 530 – 710 mW
SAT.C.04 Data budget see top-level ADCS see top-level ADCS
SAT.C.05 Power and data bus interfaces I2C I2C
Withstand the launch Sine and random vibration
SAT.C.06 Dnepr launch loads
environment loads performed
Compliant in vacuum over
Withstand the space -15°C to 45°C in range 0°C to 50°C, but
SAT.C.07
environment vacuum optimum range is 10°C to
30°C
Generate housekeeping data
Rotation speed and rotation speed for all three
SAT.2.F.02 when of interest to satellite
current reaction wheels
operation
not impact critical mitigation of failure mode
SAT.2.C.01 Adhere to reliability standards
systems effects.

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

Mass budget
Table 15 shows the mass-breakdown of the Delfi-n3Xt Reaction Wheel System. The budgeted mass was
roughly 120 grams, and therefore the RWS is compliant with the mass budget.

Table 15: Mass breakdown for the Delfi-n3Xt Reaction Wheel System
Component Amount of units Mass per unit (g) Total mass (g)
Bracket 1 32 32
Motors 3 1.1 3.3
Flywheels 3 6 18
Drive electronics 3 1.5 4.5
Protection caps 3 3 9
Screws, nuts & washers - approx. 15 15
TOTAL 82

Volume budget
The footprint of the Reaction Wheel System is 40 x 40 [mm2] with some additional structure to connect the
RWS with three of the four threaded rods. The height above the PCB is 27 mm. The resulting volume of the
full ADCS is 90 x 90 x 34.6 [mm3]. Originally, the volume budget for the reaction wheels PCB was 90 x 90 x
32.6 mm3 [SLR0303, V2.1]. However, over the course of development it was decided to integrate all ADCS
hardware and electronics on a single PCB. This volume available for this ADCS system PCB is 90 x 90 x 46.2
mm3 [SLR0303, V2.2].

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 50 / 74

Power budget
As is discussed in Section 9.3, the power consumption of the reaction wheels highly depends on the rotation
speed and temperature. In the nominal case, as can be seen from Figure 35, the minimum power
consumption of each motor is 120 mW when operated at a speed below roughly 3,000 rpm. On the other
hand, if a reaction wheel is operated around its maximum rotation speed the power consumption can be as
high as 165 mW in the nominal case.

However if the effects of temperature and interference from other wheels are taken into account as discussed
in the Reaction Wheel Test Document, the absolute worst-case power consumption when all wheels are
operated at maximum speed and deliver maximum torque increases significantly. The drive electronics
consume around 15 mW. All factors are combined in Table 16 and the total power consumption of the full
Reaction Wheel System including drive electronics is thus estimated by the range 515 – 710 mW.

Table 16: Assessment of worst-case minimum and maximum power consumption.


Case Impact Power consumption Power consumption
min [mW] max [mW]
Motor nominal max + 3·120 (min case) 360 495
+ 3·165 (max case)
0°C temperature x 1.2 432 595
Interference other x 1.12 485 665
wheels
Electronics + 3·15 530 710
WORST-CASE 530 mW 710 mW

In practice it is highly unlikely that all three wheels will be operated simultaneously at high rotation speeds,
because the ADCS algorithm starts the reaction wheel unloading procedure well before the maximum useful
speed is reached. It is possible to make an estimate of the average power consumption of each reaction
wheel. With the central ADCS algorithm designed such that it tries to keep the reaction wheels spinning below
roughly 8,000 rpm, a realistic estimate of the nominal average motor power consumption is 130 mW per
reaction wheel (see Figure 35). The worst-case average power consumption would then be three times
130·1.2·1.12+15, which equals 570 mW.

Launch environment
Vibration tests have demonstrated that the Reaction Wheel System can withstand the random- and sine
vibrations during a launch with the Dnepr launch vehicle. This is discussed in more detail in [SLR0766].
Because no shock tests are required according to the ISIS qualification standards [SLR1008], the ability to
survive the launch environment is successfully verified.

Space environment
Thermal-vacuum tests described in the Reaction Wheel Test Document [SLR0766] demonstrate that vacuum
has no impact on the reaction wheel performance and the specified lifetime of the vacuum-rated motors is
roughly 10,000 hours, which is well over three months. Temperature does however directly affect the RWS
performance and in [SLR0766] it is concluded that there is an unacceptable increase in power consumption
below 0°C. Additional, as temperatures increase, the maximum useful speed decreases to just below 25,000
rpm. The optimum temperature is approximately 10°C to 30°C, where power consumption is at a minimum
and the maximum useful speed is 26,000-27,000 rpm. Thus, although the Reaction Wheel System is able to
operate over the full temperature range of -15°C to 45°C, the temperature range where the RWS can
effectively operate is 0°C to 45°C, with a optimum performance between 10°C and 30°C.

Generate housekeeping data


Full monitoring of the reaction wheels is possible with information on both rotation speed and drawn current.
However, due to the limited space on the drive electronics board, no adequate resistor could be integrated to
measure the current and there is thus no direct information on the power consumption. The rotation speed
can however be measured using the Hall sensors but this alone does not allow continued monitoring of the
system’s health.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 51 / 74

9.2 Verification of functional requirements


In this section the functional requirements are listed and tested against the final design.

Requirement Short description Target Actual


about all three principal about all three principal
SAT.2.2.2.2.F.01 Store angular momentum
axes axes
about all three principal about all three principal
SAT.2.2.2.2.F.02 Generate torques
axes instantaneously axes instantaneously
I2C commands verified to
turn motor on or –off,
Handle commands from the command torque, and read
SAT.2.2.2.2.F.03 -
central ADCS microprocessor. rotation speeds. Reliability
of RWS command handling
being assessed.
Generate reliable information
SAT.2.2.2.2.F.04 about the stored angular all three principal axes. all three principal axes
momentum.
Not have a negative Negligible shaking due to
SAT.2.2.2.2.F.05 Stability of Delfi-n3Xt.
impact. periodic torques

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

Handle commands from the central ADCS microprocessor


During implementation of the controller, serious issues were observed in the software-controlled I2C
communication with the Atmel AtTiny861 microprocessor. The processor speed was inadequate to handle
read requests without messing up registers, which eventually caused the controller to crash and resulting in
the uncontrolled spin-down of the reaction wheel in question.

As a solution to this, it has been decided to apply a new approach using the Atmel ATmega164P
microprocessor which has a hardware I2C protocol. All other characteristics of the ATmega164P are nearly
identical to the AtTiny861, although the ATmega164P has the additional disadvantage that the PWM-counter
is only 8-bit instead of 10-bit. The latter microprocessor does have the additional advantage of a third
counter.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 52 / 74

9.3 Verification of performance requirements


In this section the performance requirements are listed and compared with the measured performance.

Requirement Short description Target Actual


Minimum torque about each
SAT.2.2.2.2.P.01 5.5·10-6 [Nm] 5.5·10-6 [Nm]
individual axis.
1.36·10-3 [Nms] outside
dynamic range of angular
optimum temp-range,
SAT.2.2.2.2.P.02 momentum about each individual 1.56·10-3 [Nms]
1.45·10-3 within optimum
axis.
temp-range
SAT.2.2.2.2.P.03 Systemic disturbance torques. < 10−7 [Nm] < 1.4·10−7 [Nm]
SAT.2.2.2.2.P.04 Disturbance angular momentum < 2·10-7 [Nms] < 1.9·10-6 [Nms]
Control over angular momentum
SAT.2.2.2.2.P.05 2·10-7 [Nms] 4.4·10-7 [Nms]
about each individual axis.
Accuracy of angular momentum
SAT.2.2.2.2.P.06 determination about each 2·10-8 [Nms] 1.2·10-7 [Nms]
individual axis

[] Significantly below requirement and potential high impact on satellite


[] Below requirement but medium impact on satellite
[] Compliant with requirement

Performance analysis
Section 6.2 dealt with the motor performance of the preliminary design. Here the same graphs are
established for the final design. Not only have the wheels now been integrated in the full system, but
additionally vacuum-rated motors have been used instead of regular ones. The results are presented in
Figure 32 and Figure 33 whereas the raw measurements can be found in Appendix E. For comparison the
results that were presented earlier in Figure 10 and Figure 11 are including in these figures on the left. The
conclusion is that the nominal performance of the final design is slightly below the nominal performance of
the preliminary design. Nominal performance is determined at room temperature and when no other wheels
are actuated.

In Figure 32 and Figure 33, also the hypothesis is tested that the observed spread in the measurements is
caused by periodic disturbance torques that result from the axial play in the motor as discussed in
Section 8.6. To test this hypothesis, curves are plotted where the disturbance torque is either added or
subtracted from the theoretical curve that best described the measurements. How to determine the
theoretical curves that best described the measurements is described in the Reaction Wheel Test Document
[SLR0766].

Both the measurements obtained during preliminary verification and final design verification match well with
these hypothesized bounds. Some measurements of the verification on the preliminary design fall outside
these bounds, but it is reasonable to assume that these can be attributed to measurement errors due to
limitations in test set-up. Those were better corrected for during verification of the final design and this can
be seen in the slightly reduced spread in measurement.

When applying the same technique on the measurements obtained from verification of the final design, it was
found that in order to have a good match with the measurement, terminal resistance had to be adjusted to
17 Ω instead of the specified 16 Ω that matched with the measurements on the preliminary design, leading to
an increase in power consumption and decrease of maximum useful speed. There is however no reason to
believe that the resistance has changed due to the design changes in the step from preliminary- to final
design, because all hardware (including the drive electronics) have remained nearly identical. The main
change was the integration in the bracket as is shown in Figure 34.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 53 / 74

It is therefore argued that the observed change in terminal resistance of the motor is in fact a true change in
terminal resistance. Faulhaber Technical Information [SLR0979] states that the value for the terminal
resistance has an uncertainty of 12%. It was verified in the Reaction Wheel Test Document that the value of
17 Ω was consistent for two motors of the vacuum-rated motor batch, giving a first indication that variation
may occur between different motor production batches, and that within a single production batch
performance is very similar. Because 17 Ω is 6% more than the specified value, this does not represent the
worst-case scenario as indicated by the margin provided by Faulhaber.

For the remainder of this analysis the values are used that match the measurements done on the engineering
model. The determined value for the friction torque C0 is 0.026·10-3 [Nm], the observed dynamic friction
coefficient is 0.52·10-9 [Nm/rpm] and the terminal resistance is 17 Ω.
-4
x 10 -4
1.8 x 10
Specified torque 1.8
1.6 Measured torque Theoretical torque
1.6
Hypothesized torque bounds Measured torque
1.4 Expected torque bounds
1.4
1.2
1.2
Torque [Nm]

Torque [Nm]
1 1
0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 5,000 10,000 15,000 20,000 25,000 30,000 0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 32: The torque curve as established for the individual reaction wheels in the prototype model, using standard
Faulhaber 1202 004 BH motor (left) and the curve as established for the engineering model, using the vacuum-rated
version (right). In the right picture the theoretical torque curve matching the measurements is plotted and not the
specified torque as on the left picture. The hypothesized torque bounds have also been plotted.

0.8
0.8
Specified P
in, max Specified Pin, max
0.7 0.7
Measured Pin, max Measured Pin, max
0.6 Specified P 0.6 Specified Pout, max
out, max
Measured P Measured Pout, max
out, max
0.5 0.5
power [Watt]

Hypothesized Pout, max bounds


power [Watt]

Hypothesized Pout, max bounds


0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0
0 0 5,000 10,000 15,000 20,000 25,000 30,000
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
rotation speed [rpm]
Figure 33: The power curves as established for the individual reaction wheels in the prototype model, using standard
Faulhaber 1202 004 BH motor (left) and the curve as established for the engineering model, using the vacuum-rated
version (right). In the right picture the theoretical power curves matching the measurements are plotted and not the
specified power as in the left picture. The hypothesized bounds for the delivered power have also been plotted.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 54 / 74

Figure 34: Comparison of test set-up during verification testing on the prototype (left) and verification on the
engineering model. The main differences are that in the engineering model the protective cap is included and
that the prototype is mounted on a large aluminum plate for stability.

Verification of required torque


With the values for the static friction and dynamic friction verified, the performance curve as introduced in
Figure 12 can be established for the final design in the nominal case. The nominal case is when no other
reaction wheels are actuated and the environment is at room temperature.

The value for the friction torque C0 was determined as 0.026·10-3 [Nm] and the observed dynamic friction
coefficient was 0.52·10-9 [Nm/rpm]. The terminal resistance was determined as 17 Ω. The corresponding
characteristic curve is presented in Figure 35. The maximum useful speed can again be calculated using
Equation 12, which is repeated here

k M U  R  C0  Treq 
nmax  Equation 12
k M k E  RCv [rpm]

With the verified, specified values for the friction torque and dynamic friction coefficient, the maximum useful
speed corresponds to roughly 26,700 [rpm].

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 55 / 74

0.25

0.2
Power [W]

0.15

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 35: Assessment of the required power to deliver the minimum torque with the 17 Ω terminal resistance, the static
friction torque C0 = 0.026·10-3 and the dynamic friction coefficient Cv = 0.52·10-9. These values were determination in
during verification tests of the final design.

Verification of dynamic range


A maximum useful speed of 26,000 rpm was determined for the final design in the nominal situation. But
temperature and interference from other reaction wheels greatly impact the performance as was discussed in
the Reaction Wheel Test Document [SLR0766]. The various impacts on the motor performance are
summarized in Table 17:

Table 17: Impact of temperature and interference from other reaction wheels on the maximum useful speed
Case Impact Max useful speed [rpm] Max useful speed [rpm]
Within optimum temp-range Outside optimum temp-range
Motor nominal max - 26,000 24,500
Interference other x 0.90 23,400 22,000
wheels
Worst-case 23,400 22,000

A maximum useful speed of 23,400 rpm coincides with a dynamic range of;

nmax  
H  2   I  1.45 10 3  Nms 
30

When the temperature goes outside the temperature range 0°C < T < 50°C than, the dynamic range is only
1.36·10-3 [Nms].

Verification of systemic disturbance torques


In Section 8.6 the maximum alignment error was calculated as 1.45 [deg]. The maximum resulting systemic
disturbance torque was calculated as 1.4·10-7 [Nm].

Verification of disturbance momentum


As was calculated in Section 8.7, the instantaneous disturbance angular momentum can be up to 1.9·10-6
[Nms], which is an order of magnitude higher than the required 2·10-7. However, because this disturbance
momentum is highly periodic it will not have any significant impact on the satellite attitude as also calculated
in Section 8.7.

Verification of angular momentum determination


The required angular momentum determination of 2·10-8 Nms can be translated in a minimum determination
accuracy of the rotational rate using the relation:

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 56 / 74

n  I
H  [ Nms ] Equation 22
30

Where Hε is the desired angular momentum control accuracy, nε  is the necessary rotation speed
determination accuracy and I is the inertia of the motor including flywheel. Due to the discretization effects
described in Section 7.4 the worst-case calculated speed determination accuracy of is 4 rpm (See Figure 16)
at high rotation speed after implementation of speed averaging. This accuracy coincides with an angular
momentum determination accuracy of 1.2·10-7 Nms, which is approximately six times lower than required.

At low rotation speeds the determination accuracy can be at best 1 rpm, coinciding with 3·10-8 Nms, which is
still below the requirement. Due to discretization of the rotation speed determination a better performance is
not possible.

The requirement on the accuracy of angular momentum determination was derived to assure sufficiently
accurate control. However, with the discrete system that is implemented an accuracy of 2·10-8 Nms is not
possible. This will therefore limit the control accuracy as well.

Verification of control over angular momentum


The requirement for control over angular momentum was defined as the error in the achieve angular
momentum at the end of each control step of 2 seconds. Due to the time delay in the measured speed with
respect to the actual speed that was introduced in Section 7.4, new reference commands are always with
respect to a slightly lower value than the real value. This causes a continuous off-set in the delivered torque
as is also demonstrated in Figure 36.

30,000
Ideal system response
25,000 Simulated system response
Rotational speed [rpm]

20,000

15,000

10,000

5,000

0
0 50 100 150
Time [s]
Figure 36: Vizualization of the torque off-set due to the delay in speed measurement, which effectively results in a slighlty
smaller slope of the wheel speed than desired. Curves where generated using a representative reaction wheel simulator

The torque off-set is estimated by calculating the average slope deviation as percentage of the commanded
average slope:

n 1,000
T  100%  100%  4%
n 25,000

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 57 / 74

With a maximum torque of 5.5·10-6 [Nm] and control step size of two seconds, the estimated angular
momentum control accuracy is:

H ,control  2  5.5 106  0.04  4.4 107 [Nms]

The required control accuracy is 2·10-7 [Nms] and thus the estimated worst-case control accuracy is roughly a
factor two below the requirement. This means that the impact of the RWS on the satellite is not negligible
with respect to the external disturbance torques.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 58 / 74

10 Conclusions
This document has described the design and verification of the Delfi-n3Xt Reaction Wheel System (RWS). The
Reaction Wheel System can be roughly divided in the reaction wheels, the support structure, the drive
electronics and the control algorithm. The reaction wheels are based on Faulhaber 1202 004 BH Brushless DC
Micromotors, which were selected due to their small size and mass, which allows a very compact design. The
full system was successfully developed although some specifications are slightly below their required values.

Flywheels were designed on the basis of the required dynamic range of 1.56·10-3 [Nms] and a margin of
10%. Support structure was developed for easy assembly and integration of the Reaction Wheel System. In
particular protective caps were developed to reduce the maximum excitation of the flywheels during launch.
The Reaction Wheel System was designed to be very modular through the integration of dedicated drive
electronics underneath each reaction wheel. This also results in a very compact design. The full system is also
very easy to integrate in the nanosatellite itself.

Dedicated drive electronics have been developed for each reaction wheel which include a microprocessor to
control motor commutation, to handle commands from the central ADCS processor and execute motor
control. The control algorithm was introduced in this document but still needs verification on the final design.

Early in the development process a prototype was developed to verify the motor performance and give a first
indication of whether the Reaction Wheel System could withstand the launch loads. This was considered a
high priority activity as an earlier design (see V1.4 of this document) had disintegrated during an excessive
vibration campaign. After a successful preliminary vibration campaign, the engineering model was developed
which was successfully tested at NLR with representative random and sine vibration loads.

It was discovered during the development of the Reaction Wheel System that the allocated power budget is
insufficient to also cover the worst-case situation. When all wheels are operated at high speeds, the RWS can
consume over 700 mW instead of the originally budgeted 400 mW. Early in the design there were indications
that the Reaction Wheel System would exceed the power budget and an increase in the allocated power was
already considered. Currently the worst-case calculated 710 mW is budgeted for the RWS and the original
requirement has been adapted. Due to the fully digital control implementation, the electronics of each
reaction wheel consume only 15 mW.

With respect to functional requirements, the design is very adequate where only the reliability of command
handling remains an open issue. The design however struggles in fully meeting the performance
requirements. During verification of the prototype, variations in the generated torque were observed which
were later attributed to highly periodic disturbance torques cause by axisymmetric errors. On the basis of this
conclusion it could be verified that the reaction wheels are compliant with the dynamic range in the nominal
situation but in the worst case situation the dynamic range could drop to 1.39·10-3 [Nms] instead of 1.56·10-3
[Nms]. The required torque of 5.5·10-6 [Nm] is not a problem.

In preparation to the assembly and integration of each Reaction Wheel System model, it will be required to
verify both the flywheels as well as the performance of the motors. It was observed that the terminal
resistance changes between motors, although it was argued that this change is mainly between production
batches. The margin on the terminal resistance stated by Faulhaber is 12% which clearly impacts the
performance.

The maximum systemic disturbance torque that is caused by misalignment of the reaction wheel axes was
calculated as < 1.4·10−7 [Nm] where 1·10−7 [Nm] is required. The expected instantaneous disturbance
torques are an order of magnitude higher than required; 1.9·10-6 [Nms] instead of 2·10-7 [Nms]. But because
of their periodic nature the net effect on the satellite is a periodic disturbance with negligible maximum
amplitude of 3.8·10-7 [deg]. The disturbances do however contribute to the calculated worst-case value of the
angular momentum and have caused rotational rate sensors to be omitted from the ADCS design.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 59 / 74

Disturbance torques are therefore concluded to have the biggest impact on the overall performance of the
Reaction Wheel System.

Finally the accuracy values of the control and determination of the reaction wheel are below the initially
derived requirements. The angular momentum control accuracy is better than 4.4·10-7 [Nms], where 2·10-7
[Nms] was required, while angular momentum determination accuracy is 1.2·10-7 [Nms] instead of the
required 2·10-8 [Nms]. It is difficult to improve these accuracies because they are caused by discretization in
the system and thus considered a negative side-effect of the fully digital control.

Overall it can be concluded that an effective and balanced Reaction Wheel System has been designed.
Effective because it meets all functional requirements and balanced because most performance have not
been met to roughly the same degree. The original requirements were very ambitious and the design
described in this report demonstrates a best effort approach to meet those.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 60 / 74

11 Next steps
Although an engineering model of the Delfi-n3Xt Reaction Wheel System has been developed, it does not yet
represent the full system. In summary, for the current model the final activities to be executed are:

 Implementation of the full control algorithm in the new microprocessor.

 Analysis of the performance of the fully functional Reaction Wheel System

 Assign location of the 2 bytes in 7 byte read-out of the reaction wheels by the central ADCS
microprocessor.

It was also observed in the verification process that disturbances play a key role in the performance and
effectiveness of the Reaction Wheel System. If future designs will be designed, disturbances will likely play an
important role in the design process and therefore requirements should be derived to support that. For future
development it is therefore meaningful to define a requirement on the allowed linear and angular
accelerations (as a function of frequency if required). In the case of Delfi-n3Xt the rotational rate gyros were
omitted because they were unable to cope with the high vibrations induced by the RWS. If there is a
maximum allowed vibration frequency spectrum in which the rotational rate sensors can operate, this could
serve as input to the RWS design.

Additionally, the microprocessor proved to have a big impact on the controller. Using the rationales
introduced in this document, a better assessment of the impact of microcontroller characteristics is possible.
At least it should have a hardware I2C protocol and at least three counters to perform all tasks. Other aspects
to take into consideration are the maximum bit-depth, the use of doubles instead of integers and higher clock
frequencies. These factors all influence the accuracy of speed determination and control that can be
achieved. Finally it could be investigated to implement different operating modes for different speed regions
(with e.g. other clock divider settings) in a single microprocessor. This would allow tuning of the parameters
for more optimum speed determination and control over the full rotation speed dynamic range of the motor.

For future models it is advised to carefully consider the observations and conclusions that were made in this
design document. The fact that the requirement on the angular momentum dynamic range was not met can
be accounted for in future designs with larger flywheels. In this design a margin of 10% was taken on the
calculated maximum useful speed, but that was insufficient. Based on the verification performed and
described in this document, a better value for the initial margin is 25%. For the power consumption, a margin
of 80% on the calculated values is recommended because temperature and interference from other reaction
wheels have proven to have a big impact on the power consumption. On the other hand, a bigger flywheel
would reduce torque control accuracy, because torque accuracy is determined by the combination of rotation
speed control accuracy and flywheel inertia. Torque accuracy versus dynamic range will therefore be a likely
trade-off in future designs.

Additionally it can be considered to e.g. incorporate a second bearing in the protective cap to fix the motor
axle and an even simpler flywheel design to reduce manufacturing errors. Fixing the axle will not reduce the
vibrations but will only contain them within the plane of the flywheel. This will make the vibrations more
predictable and thereby manageable. At the Faculty of 3ME there is work on tunable compliant vibration
isolators which could be worth investigating further.

If you are reading this document as a designer of a Reaction Wheel System for future missions in the Delfi
than feel free to contact Teun Hoevenaars because former Delfi students are generally more than happy to
share their experiences and provide practical advice in exchange for a cold beer ;).

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 61 / 74

Appendix A. Raw measurements of prototype model


performance

2011‐12‐28_A#3_YW_testbaord_alubase3V3_torque
Date 28‐12‐2011
Time (at beginning of the 3:41 PM
Measurement time (s) 40
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)
0 0 0.012 0 0 0.011 0 0 0.011
0.256 0 0.011 0.252 0 0.011 0.252 0 0.011
0.512 0 0.011 0.506 0 0.011 0.509 0 0.011
0.767 0 0.012 0.76 0 0.011 0.841 0 0.012
1.024 0 0.012 1.014 0 0.011 1.095 0 0.011
1.278 0 0.012 1.271 0 0.011 1.357 0 0.011
1.531 0 0.011 1.526 0 0.011 1.612 0 0.011
1.786 0 0.011 1.782 0 0.011 1.882 0 0.011
2.043 0 0.011 2.037 0 0.011 2.133 0 0.011
2.499 919.09 0.609 2.495 892.883 0.61 2.597 718.374 0.609
3.746 6465.518 0.512 3.774 6465.517 0.512 3.779 6048.387 0.518
4.923 10416.667 0.442 4.944 10714.286 0.441 4.959 10416.667 0.445
6.104 13888.889 0.382 6.111 13888.889 0.385 6.142 14409.222 0.387
7.273 17045.455 0.342 7.294 17045.455 0.339 7.31 17045.455 0.345
8.451 17867.025 0.304 8.461 18750 0.305 8.476 19724.182 0.305
9.622 20819.899 0.272 9.645 22074.661 0.274 9.656 20847.458 0.277
10.807 22058.824 0.255 10.829 22043.011 0.248 10.829 22043.011 0.251
11.985 23455.38 0.233 11.996 23437.5 0.233 12.009 23455.38 0.236
13.156 24979.691 0.217 13.179 23437.5 0.219 13.193 24979.691 0.217
14.336 24959.416 0.207 14.361 24979.691 0.209 14.376 25000.001 0.207
15.506 25020.342 0.203 15.545 25020.342 0.195 15.56 25040.718 0.196
16.689 26809.069 0.194 16.728 26739.13 0.195 16.745 26762.402 0.192
17.876 26785.717 0.186 17.91 26785.717 0.183 17.926 26809.069 0.186
19.056 25020.342 0.187 19.094 26785.714 0.18 19.095 26809.069 0.176
20.238 24999.999 0.18 20.26 26785.714 0.178 20.275 26785.717 0.178
21.422 26809.069 0.175 21.444 26809.069 0.172 21.459 28819.12 0.179
22.605 26809.069 0.177 22.628 26762.402 0.171 22.644 26809.069 0.169
23.788 26762.402 0.172 23.811 28819.12 0.166 23.827 26785.714 0.165
24.975 26785.714 0.174 24.995 26809.069 0.171 25.01 26762.402 0.17
26.156 26785.714 0.17 26.178 26785.714 0.171 26.193 26762.402 0.17
27.34 28873.24 0.169 27.361 26785.717 0.169 27.377 26785.717 0.17
28.524 26809.069 0.172 28.546 26762.402 0.165 28.559 26809.069 0.166
29.704 26785.714 0.17 29.73 26784.628 0.168 29.745 26809.069 0.162
30.89 26785.717 0.169 30.91 26809.069 0.167 30.929 26786.803 0.163
32.057 28820.378 0.168 32.096 26785.714 0.164 32.096 26762.402 0.161
33.239 26809.069 0.161 33.28 26854.779 0.164 33.277 28819.12 0.164
34.408 26809.069 0.166 34.462 28819.12 0.167 34.462 28819.12 0.168
35.589 26762.402 0.16 35.644 28846.154 0.164 35.643 26785.714 0.167
36.774 28819.12 0.161 36.829 28846.154 0.161 36.826 26785.714 0.164
37.942 28873.24 0.164 38.012 26832.464 0.165 38.013 26809.069 0.161
39.127 26762.402 0.169 39.196 28846.154 0.166 39.195 28846.154 0.159
40.306 26785.714 0.162 40.38 26809.069 0.162 40.359 28847.413 0.162

spindown 90 s spindown 90 s

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 62 / 74

Continued from previous page

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)
0 0 0.012 0 0 0.011
0.262 0 0.012 0.254 0 0.011
0.518 0 0.012 0.51 0 0.012
0.773 0 0.011 0.764 0 0.012
1.028 0 0.012 1.018 0 0.012
1.284 0 0.012 1.272 0 0.012
1.538 0 0.012 1.527 0 0.012
1.791 0 0.012 1.784 0 0.011
2.045 0 0.012 2.038 0 0.011
2.505 2053.656 0.6 2.495 0 0.601
3.883 7500 0.494 3.498 6000 0.524
5.14 11646.448 0.422 4.739 10563.781 0.445
6.374 15120.968 0.366 5.931 14151.044 0.382
7.599 17281.106 0.325 7.131 16591.935 0.339
8.822 19431.279 0.288 8.308 19034.297 0.306
10.051 21186.44 0.264 9.496 20601.27 0.268
11.27 22864.155 0.24 10.681 22189.349 0.246
12.483 23293.689 0.22 11.864 23445.643 0.226
13.705 24191.646 0.21 13.062 24193.548 0.212
14.917 24834.437 0.197 14.241 25272.647 0.2
16.17 26221.54 0.186 15.429 25002.034 0.186
17.384 26408.452 0.185 16.609 25510.204 0.184
18.604 26226.013 0.174 17.796 26785.714 0.178
19.837 26788.046 0.168 19.011 26041.668 0.173
21.071 26595.745 0.169 20.215 26600.462 0.172
22.305 27171.511 0.17 21.416 27369.827 0.165
23.535 26790.498 0.168 22.609 27578.474 0.164
24.751 27576.002 0.16 23.796 26973.686 0.159
25.97 27372.264 0.16 24.98 27777.779 0.161
27.185 27780.287 0.159 26.161 27173.912 0.155
28.404 27576.002 0.159 27.343 27581.329 0.161
29.617 27573.528 0.154 28.529 27173.912 0.154
30.833 27369.827 0.156 29.709 27985.074 0.157
32.053 27777.779 0.161 30.893 27777.779 0.156
33.269 27173.912 0.157 32.112 27573.528 0.159
34.485 27372.261 0.152 33.296 27775.27 0.151
35.702 27987.621 0.155 34.53 27777.779 0.16
36.917 27176.314 0.152 35.728 27377.137 0.153
38.137 27573.528 0.157 36.912 28192.905 0.151
39.352 27372.264 0.161 38.096 27987.621 0.156
40.57 27372.264 0.158 39.279 27369.827 0.157
40.465 27982.527 0.16

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 63 / 74

Continued from previous page

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)
0 0 0.012 0 0 0.011
0.255 0 0.012 0.254 0 0.011
0.509 0 0.011 0.511 0 0.011
0.763 0 0.011 0.765 0 0.012
1.017 0 0.011 1.02 0 0.012
1.27 0 0.011 1.277 0 0.012
1.524 0 0.011 1.531 0 0.011
1.781 0 0.012 1.785 0 0.011
2.036 0 0.012 2.039 0 0.011
2.498 0 0.6 2.496 0 0.602
3.506 6260.088 0.519 3.514 6167.764 0.518
4.754 10932.517 0.443 4.686 10273.973 0.447
5.997 14044.335 0.38 5.878 14044.944 0.386
7.231 17359.102 0.33 7.08 16592.966 0.335
8.464 19034.355 0.291 8.26 19132.652 0.303
9.696 20833.334 0.265 9.461 20377.667 0.273
10.914 22322.968 0.246 10.647 22189.349 0.247
12.131 24038.463 0.228 11.844 23006.135 0.232
13.413 24350.651 0.208 13.03 24038.463 0.213
14.629 25000.001 0.201 14.227 25171.807 0.202
15.848 25167.784 0.181 15.412 25337.836 0.194
17.063 26041.666 0.18 16.592 25512.321 0.184
18.279 26408.452 0.174 17.812 26595.745 0.179
19.498 27176.314 0.172 18.995 26223.777 0.175
20.731 27374.826 0.169 20.18 27367.392 0.167
21.963 26975.656 0.162 21.38 26973.686 0.17
23.18 27573.528 0.157 22.585 27171.511 0.16
24.397 27775.27 0.159 23.758 27173.912 0.156
25.611 27176.314 0.161 24.941 27982.527 0.157
26.829 27982.397 0.155 26.127 27171.511 0.154
28.048 27985.074 0.155 27.313 27780.286 0.156
29.263 27782.798 0.151 28.495 27369.827 0.157
30.48 27985.074 0.152 29.699 27993.523 0.159
31.694 27570.93 0.156 30.869 27369.827 0.158
32.914 27992.36 0.158 32.044 27985.074 0.155
34.13 27372.264 0.15 33.23 27578.599 0.151
35.35 27583.077 0.153 34.43 27952.852 0.155
36.564 27573.528 0.149 35.692 27985.351 0.151
37.782 27372.264 0.151 36.816 27985.074 0.157
38.996 27576.002 0.148 37.996 27987.621 0.153
40.215 27372.264 0.15 39.18 28198.072 0.154
40.364 27570.93 0.15

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 64 / 74

Appendix B. The commutation schemes for a four pole pair


BLDC motor
Commutation in counterclockwise direction
This is rotation in the positive direction when considering a motor reference frame with the local Z-axis
pointing in the direction of the motor axle.

0° 15° 30° 45° 60° 75°


step 1 step 2 Step 3 Step 4 Step 5 step 6

H1

H2

H3

Coil U

Coil V

Coil W

Active
T3, T2 T3, T6 T1, T6 T1, T4 T5, T4 T5, T2
MOSFETs

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 65 / 74

Commutation in counterclockwise direction


This is rotation in the positive direction when considering a motor reference frame with the local Z-axis
pointing in the direction of the motor axle.

0° 15° 30° 45° 60° 75°


step 6 step 5 step 4 Step 3 Step 2 Step 1

H1

H2

H3

Coil U

Coil V

Coil W

Active
MOSFETs T1, T4 T1, T6 T3, T6 T3, T2 T5, T2 T5, T4

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 66 / 74

Appendix C. Schematic overview of the microprocessor


implementation of reaction wheel control

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 67 / 74

YES Commutate using look-up 
rot_dir = 1?
table 1

timer_hall = timer_hall 
Commutate using  + (read hall timer)
look-up table 2 reset Hall_timer clock
set Hall_flag=TRUE

COMMUTATE FUNCTION

reset Hall_timer clock


timer_hall = 2^15;
speed_flag = TRUE;

HALL TIMER OVERFLOW FUNCTION

Write commanded 
NO
Receive command read command? variables  to 
temporary variables

YES

read bytes TWI_flag = TRUE;

I2C FUNCTION

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 68 / 74

n = rot_dir * rpm;
rpm = f(hall_timer) using 
hall_flag = FALSE; Apply speed averaging
look-up table
timer_hall = 0;

* rot_dir = 1 is defined as the rotation direction such that a 
positive acceleration produces a positive torque on the satellie,
rot_dir = -1 is defined as the rotation direction is such that a 
SPEED DETERMINATION FUNCTION positive acceleration produces a negative torque on the satellie.

if abs(n) < 500 rpm
K_p = 54;
K_d = 109;
K_i = 85;
else Propoprtional control
K_p = 7; PWMp = Kp·e
K_d = 1;
K_i = 7;

n_old=SpeedMeasure
n = SpeedMeasure;
e = nref ‐ naveraged
n_dot = dn - (n–n_old) Derivative control
∑/27
PWMD = KD·n_dot

PWMI  = PWMI + n_int
Integral control
n_int = KI·e/frws if n_int > 2^15
n_int = 2^15;

PID FUNCTION

if PWM > 256 
PWM = 256;
if PWM < -256 
PWM = -256;

if PWM ≥ 0 && nref > 0


PWM = PWM, ref_dir = 1;
if PWM < 0 && nref > 0
PWM = 0, ref_dir = 1;
if PWM < 0 && nref < 0
PWM = abs(PWM), ref_dir = -1;
if PWM ≥ 0 && nref < 0
PWM = 0,ref_dir = -1;

if ref_dir != rot_dir
rot_dir = ref_dir;
Run commutate function

PWM CASE FUNCTION

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 69 / 74

Appendix D. Derivation of flywheel acceleration due to


axisymmetric error
To derive the accelerations of the flywheel in the body-fixed motor reference frame due to an axisymmetric
error, first we simplify the effect of an axisymmetric error. The main two contributors to the axisymmetric
error are:

1. A motor axle that is not perfectly aligned with the axis of rotation

2. A flywheel that is not perfectly axisymmetric

Both errors are visualized in Figure 37. The rotor reference frame is rotating around its local Z-axis with
rotation speed ω, which is tilted with respect to the body-fixed motor reference frame with angle εaxle. In the
rotor reference frame, the flywheel center of gravity (cog) is tilted with respect to the rotor Z-axis with angle
εcog.

Figure 37: The rotor reference frame is rotating around the rotor Z-axis, which is tilted from the body-fixed motor
reference frame around an angle εaxle. The origin of the body-fixed motor reference frame is the center of the motor
bearing.

In order to assess the acceleration of the flywheel center of gravity in the body-fixed reference frame, first
the vector to the flywheel center of gravity is expressed in the body-fixed reference frame:
  
r M,cog  r M,0  R M / R r R,cog Equation 23

 
Where r M,0 is the vector to the origin of the rotor reference frame and R M / R r M, cog is the center of gravity
vector in the rotor reference frame expressed in the motor reference frame. The rotation matrix R is:
M /R

1 0 0 
M /R 
  0 cos  axle  sin  axle 
R Equation 24
 0 sin  axle cos  axle 


The velocity of r M,cog can be determined by taking the first derivative of Equation 23.

    
vM, cog  r M,0     r M,0  R M / R r R, cog     R M / R r R, cog Equation 25

Where Ωε is expressed in the motor reference frame;

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 70 / 74

 0 
    sin  axle 

Equation 26
  cos  axle 

 
Given that both r M,0 and r R,cog are constant over time in their respective reference frames, only the rotation of
the reference frame contributes to the velocity. The velocity of the center of gravity in the body-fixed
reference frame expressed by Equation 25 can therefore be expressed as;

  
vM,cog     r M,0  R M / R r R,cog  Equation 27

Or, if we fill in Equation 23, simply;


 
vM,cog     r M,cog Equation 28

Next the accelerations of the flywheel center of gravity in the body-fixed motor reference frame are
determined by taking the derivative of Equation 28;

   r M    r M       r M 
aM,cog    ,cog   ,cog    ,cog Equation 29


Again, r M, cog  0 so the final equation to express the acceleration of the center of gravity in the body-fixed
motor reference frame is;

   r M       r M 
aM,cog     ,cog    ,cog Equation 30


Where r M,cog is expressed in Equation 23.


Due to the rotation around the Z-axis of the rotor reference frame, r M,cog also rotates with time. Therefore in
order to assess the evolution of the acceleration of the flywheel center of gravity it is helpful to express the

evolution of the vector r M,cog as a function of time. For that we simply expand Equation 23. First, we express

r R,cog in Cartesian coordinates in the rotor reference frame;

  FW
r R,cog  R  rcog Equation 31

  
Where rcog
FW
is the flywheel center of gravity vector in the flywheel reference frame. To get to r R,cog , rcog
R
is
rotated an angle εcog around the rotor axis XR (not that this could just have well been the YR axis). This
rotation represents the slightly displaced center of gravity due to inevitable manufacturing errors. The
 ,cog
rotation matrix R can be expressed using;

1 0 0 
 , cog  
R  0 cos  cog  sin  cog  Equation 32
0 sin  cog cos  cog 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 71 / 74


The next step is to express the vector r R,cog in the rotating rotor reference frame in the body-fixed motor
reference frame. For that we have to account for the rotation of the reference frame and rotate the vector
over a time-dependent angle –ω(t) around the rotor Z-axis;
M 
rcog  R  r R, cog Equation 33

In which

 cos t sin t 0 
R (t )    sin t

cos t 0 
 0 0 1 

Where we have made use of the relations sin   t    sin  t and cos   t   cos  t . Combining Equation
31 and Equation 33 gives now;
M  FW
rcog (t )  R  (t ) R  ,cog rcog Equation 34

M
Now only the vector from the center of the motor bearing to the rotor origin of the rotor reference frame r0
must be added and the sum of both vectors rotated over the angle εaxle. This gives:

  FW  R
rM,cog (t )  R M / R  R (t ) R ,cog rcog  r0  Equation 35

Where R is expressed in Equation 24. Inserting Equation 35 in Equation 30 allows assessment of the
M /R

accelerations of the flywheel center of gravity in the body-fixed motor reference frame as a result of the
axisymmetric errors in the reaction wheel system. As such Equation 30 is better written as;

   r M (t )       r M (t ) 
aM,cog (t )     ,cog    ,cog Equation 15

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 72 / 74

Appendix E. Raw measurements of engineering model


performance
2012‐02‐07_B#1_EM1_3.3_20C_torque
Date 7‐2‐2012
Time (at beginning of the 4:31 PM
Measurement time (s) 40
PWM setting (‐) 1023
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)
0 0 0.011 0 0 0.011
0.255 0 0.011 0.257 0 0.011
0.509 0 0.011 0.512 0 0.011
0.763 0 0.011 0.766 0 0.011
1.016 0 0.011 1.02 0 0.011
1.273 0 0.011 1.274 0 0.011
1.525 0 0.011 1.528 0 0.011
1.778 0 0.011 1.782 0 0.011
2.035 0 0.011 2.036 0 0.011
2.494 1642.576 0.593 2.494 1.485E+39 0.593
3.756 6720.43 0.503 3.713 6611.656 0.501
5.012 10593.585 0.436 4.943 10775.862 0.433
6.227 13586.956 0.384 6.184 13489.209 0.382
7.415 15956.66 0.344 7.415 16447.368 0.339
8.617 18203.882 0.302 8.639 18656.717 0.296
9.8 19738.046 0.278 9.87 20276.664 0.272
10.997 21306.821 0.255 11.101 21544.052 0.247
12.187 22322.968 0.234 12.352 22455.09 0.237
13.382 22865.854 0.224 13.57 23291.927 0.224
14.582 23285.799 0.21 14.786 24036.676 0.208
15.766 24038.459 0.199 16.037 24509.804 0.194
16.961 24682.053 0.189 17.25 25167.784 0.184
18.174 24832.432 0.186 18.468 25510.204 0.185
19.364 25684.931 0.186 19.686 26039.572 0.175
20.58 25860 0.176 20.933 25862.069 0.177
21.78 26041.668 0.177 22.15 26405.947 0.165
22.964 26225.903 0.164 23.367 26037.255 0.162
24.148 26978.417 0.162 24.584 26598.16 0.165
25.364 27171.632 0.159 25.801 27372.264 0.163
26.55 26408.452 0.161 27.019 27372.261 0.162
27.733 27176.31 0.155 28.231 26980.906 0.156
28.913 26978.417 0.155 29.449 27142.454 0.153
30.098 27173.912 0.149 30.664 27573.657 0.155
31.28 26947.203 0.154 31.883 27377.256 0.159
32.462 27775.144 0.154 33.099 27374.819 0.15
33.648 27551.155 0.147 34.317 27576.125 0.155
34.831 27178.835 0.149 35.534 26985.758 0.145
36.014 27987.621 0.145 36.8 27173.677 0.151
37.198 27777.779 0.147 38.017 26976.161 0.157
38.381 27364.957 0.145 39.232 27171.39 0.152
39.564 27576.125 0.143 40.451 27173.912 0.149
40.749 27576.125 0.142

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 73 / 74

Continued on next page

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
Document DNX-TUD-TN-0284
Date 21-02-2012
Technical Note Issue 2.3
Page 74 / 74

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)
0 0 0.011 0 0 0.011
0.256 0 0.011 0.252 0 0.011
0.511 0 0.011 0.506 0 0.011
0.768 0 0.011 0.763 0 0.011
1.022 0 0.011 1.017 0 0.011
1.276 0 0.011 1.271 0 0.011
1.53 0 0.011 1.525 0 0.011
1.782 0 0.011 1.779 0 0.011
2.036 0 0.011 2.033 0 0.011
2.493 1.485E+39 0.594 2.493 1.485E+39 0.594
3.758 7047.3 0.499 3.407 5742.844 0.521
5.022 10997.068 0.428 4.692 10107.817 0.44
6.261 14478.765 0.366 5.93 13636.364 0.38
7.493 16815.269 0.325 7.155 16519.823 0.335
8.709 19319.694 0.285 8.392 18564.355 0.297
9.944 20834.744 0.259 9.624 20161.291 0.264
11.161 22190.947 0.236 10.842 21800.702 0.244
12.376 23437.5 0.228 12.143 23007.772 0.224
13.59 24172.322 0.214 13.374 23883.587 0.211
14.811 24346.582 0.204 14.591 24669.075 0.197
16.025 24668.97 0.191 15.804 25165.823 0.185
17.242 25682.789 0.184 17.024 25684.931 0.179
18.46 25165.829 0.182 18.242 25864.47 0.174
19.673 26041.668 0.173 19.46 26783.498 0.172
20.89 26408.452 0.163 20.673 26593.447 0.169
22.111 26597.93 0.161 21.892 27173.915 0.163
23.343 26221.893 0.15 23.109 27374.579 0.162
24.56 27171.39 0.154 24.327 26973.682 0.157
25.773 26980.785 0.152 25.538 26792.952 0.151
26.99 27372.385 0.155 26.754 27374.819 0.159
28.206 27372.261 0.158 27.97 26785.714 0.149
29.425 27176.435 0.147 29.186 26788.163 0.157
30.64 27173.912 0.159 30.405 27173.912 0.152
31.858 26978.416 0.157 31.621 27372.261 0.155
33.078 26783.264 0.157 32.837 27576.002 0.156
34.287 26980.9 0.144 34.053 26978.416 0.15
35.508 26788.044 0.156 35.273 27573.528 0.15
36.725 27372.385 0.156 36.488 27369.941 0.154
37.94 27372.261 0.148 37.707 27369.831 0.145
39.155 27173.912 0.155 38.923 27985.074 0.152
40.373 26980.9 0.158 40.141 27550.532 0.147

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TN-0284 [2.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, Reaction Wheel System Design.docx
is not permitted, except prior and express written permission of Delft University of Technology.
This page intentionally contains only this sentence.
Appendix B
Reaction Wheel Assembly Manual

Because the reaction wheel system developed in this thesis work has actually
been manufactured and integrated, assembly and integration have played a
clear role in the development process. To document the experience that
was gained throughout this process and to faciliate future assembly and
integration, this manual was constructed.

171
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 1 / 15

ADCS – Reaction Wheel System Assembly Manual

Description: Stepwise explanation on how to assemble the Delfi-n3Xt Reaction Wheel System.

COMMS

Launch
MechS

T3μPS
CDHS
ADCS

ITRX

SDM

GSN
GSE
EPS

TCS
STS
Subsystem(s)
involved:

X X X

Revision Record and Authorization

approved
Reviewer
checked
Issue

Affected

PM
Date Author / Editor Description of change
Section(s)

1.0 23-12-2011 Teun Hoevenaars GB All First full version


1.1 20-01-2012 Teun Hoevenaars type of glue added

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 2 / 15

Table of Contents
1  REQUIRED PARTS AND TOOLS ______________________________________________________ 3 

1.1  Bill of Materials _______________________________________________________ 3 

1.2  Assembly tools _______________________________________________________ 5 

2  ASSEMBLY PROCESS ______________________________________________________________ 7 

2.1  Preparation __________________________________________________________ 8 


2.1.1  Verify quality of all manufactured parts _______________________________________________ 8 
2.1.2  Insert helicoils in the bracket holes for mounting of the protective caps _____________________ 8 
2.1.3  Clean all parts ___________________________________________________________________ 9 

2.2  Subassembly_________________________________________________________ 9 
2.2.1  Press fit flywheel on motor_________________________________________________________ 9 

2.3  Final assembly ______________________________________________________ 13 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 3 / 15

1 Required parts and tools


This chapter lists the required parts and tools for the Reaction Wheel System assembly.

1.1 Bill of Materials


The parts required for assembly are:

Part # Purpose Item Source # required


/ model
P1 Preparation M2 Helicoils Viba 11
P2 Preparation Bracket DNX-41-203 1
P3 Motor subassembly BLDC motor 1202 004 BH 45 Faulhaber 3
P4 Motor subassembly Flywheel DNX-41-202 3
P5 Motor subassembly M1 x 6 ISO 1207 12
P6 Motor subassembly M1 nuts ISO 4032 12
P7 Motor subassembly Drive electronics adaptor, h=2.9 mm DNX-41-205 3
P8 Motor subassembly Drive electronics PCB for X- and Y- axes DNX-41-201 2
[requires two SAMTEC HTMS-104-02-G-S- [Farnell] [4]
RA connectors for each PCB]
P9 Motor subassembly Drive electronics PCB for Z- axes DNX-41-201 1
[requires one SAMTEC HTMS-104-02-G-S- [Farnell] [1 + 1]
RA connector and one SAMTEC HTMS-
104-01-G-S for each PCB]
P10 Final assembly PCB riser DNX-41-204 12
P11 Final assembly M1.6 x 8 sunk head ISO 7046 4
P12 Final assembly M1.6 x 8 rounded head ISO 14583 8
P13 Final assembly M1.6 washer for rounded head ISO 7089 8
P14 Final assembly Protective cap for X- and Y- axes DNX-41-207 1
P15 Final assembly Protective cap for Z- axes DNX-41-206 3
P16 Final assembly M2 x 8 screws ISO 7380-A 11
P17 Final assembly M3x12 ISO 14583 1
P18 Final assembly M3 washer + nut ISO 7089 + 1
ISO 4032
P19 Final assembly ADCS board DNX-41-100 1
[requires three SAMTEC CLP-104-02-F-D
sockets]

Figure 1 shows an overview of most of the parts listed above.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 4 / 15

Figure 1: an overview of a great deal of the required parts for the Reaction Wheel System

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 5 / 15

1.2 Assembly tools


Additionally, the following special assembly tools are required:

Tool # Purpose Item Source

T1 Preparation Helicoil set Cleanroom


T2 Subassembly Press-fitting tool DNX-91-903, -904, -
[consisting of a base, a press-guide and 905 and -908
press-fitting]
T3 Subassembly Hand-press Cleanroom / Geert
Brouwer
T4 Subassembly Drive electronics integrator tool DNX-91-909
T5 Subassembly & Final assembly Screwdrivers Cleanroom
T6 Subassembly & Final assembly Torxs screwdrivers Cleanroom
T7 Subassembly & Final assembly Loctite 406 glue & glue rod Adhesives box in
cleanroom
T8 Final assembly Gaffer tape cleanroom
T9 Final assembly Customized caliper Rob van der List
(DEMO)

Figure 2 shows an overview of most of the tools listed above.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 6 / 15

Figure 2: An overview of many of the tools required to assemble the Reaction Wheel System

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 7 / 15

2 Assembly process
This chapter describes the steps required to assemble the RWS subsystem from the drawings in [SLR0927].
First the overall assembly procedure is described, followed by a more detailed description of each step in
separate sections. The overall assembly procedure is presented in Figure 3.

Insert helicoils in 
Verify quality of all 
1. Preparation manufactured 
the bracket holes 
for mounting of 
parts
the protective caps

Clean all parts

Mount motor on 
Press fit flywheel 
2. Motor subassembly on motor
drive electronics 
with adaptor

PERFORM THREEFOLD Connect flex-rigid 
connector to drive 
for motors in X-, Y-, 
electronics
and Z-axis

3. Final assembly Mount X- or Y- axis  Mount protective 


motors on bracket  cap over X- or Y-
using adaptors axis motor

PERFORM TWICE: for X- and Y-axis

Mount Z- axis  Mount protective 


Mount bracket on 
motors on bracket  cap over Z- axis 
ADCS PCB
using adaptors motor

Figure 3: The assembly process for the Reaction Wheel System

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 8 / 15

2.1 Preparation
The preparation phase consists of three activities which are required to prepare for proper assembly.

2.1.1 Verify quality of all manufactured parts

Once all parts listed in the Bill of Parts of Chapter 1 are available, it is good practice to control the quality of
the parts. This includes, but is not limited to:

 Verifying that all parts are available in sufficient number with some spares. It is easy to lose a 1mm
screw during assembly work.

 Checking whether dimensions and other features of produced and ordered parts are correct. Mistakes
are human or designs of ordered parts may have been adjusted in the meantime.

 Measure the exact mass of the flywheels using the Mettler Toledo AB204-S measurement unit
available at the Physics Lab of the Faculty of Aerospace Engineering. If the dimensions of the
flywheels are correct, the exact density of the material can be deduced from this, which can be used
as input in Catia to determine the moment of inertia of the flywheels with more precision.

 Verifying that no sharp edges etc. are left. Remove these with care. This is to assure a proper fit, as
the assembly requires very high precision of parts dimensions.

 Make sure that the connectors on the drive electronics board are positioned properly. Two types of
drive electronics PCB are required as indicated in the Bill of Parts, but for clarity also sketches of both
are showed below in Figure 4.

Figure 4: Sketches of the position and type of the connectors for the drive electronic. The left
sketch is for the X-, and Y-axis reaction wheels, whereas the right sketch is for the Z-axis reaction
wheel. The microprocessor forms a good visual reference on the actual reaction wheel PCB.

2.1.2 Insert helicoils in the bracket holes for mounting of the protective caps

Helicoils are used for the bracket mounting holes for the protective caps because the bracket is made of
aluminum, which does not make good material for robust threads. The bracket holes that require helicoils are
indicated in Figure 5. Helicoil insertion consists of the following steps, the tools for which can be found in the
toolset T1.

1. Threading the holes. Although they are already threaded, it is good practice to use the dedicated
helicoil tool to verify the holes are threaded deep enough by the workshop. The thread should go at
least 6 mm.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 9 / 15

2. Insertion of the helicoils using the designated tool.

3. Breaking of the base support tip from the helicoils using the designated tool.

Figure 5: Indication of the eleven bracket holes which require insertion of helicoils

2.1.3 Clean all parts

Because it concerns space hardware and the performance of the motors will go down if dirt is accumulated in
the bearings, all parts should be cleaned either in an ultrasonic bath or using appropriate alternative cleaning
equipment. It is also common practice for electronics to be cleaned with flux-remover to remove excess
solder. Communicate this with the person responsible for manufacturing of the drive electronics. The motors
themselves should not be cleaned but come in sealed bags. Keep them sealed as much as possible, also after
integration.

2.2 Subassembly
The subassembly should be manufactured in three-fold, where special care should be taken that the Z-axis
requires a different drive electronics PCB than those for the X- and Y-axis. The Z-axis is the direction
perpendicular to the ADCS PCB plane. Each subassembly step is described in more detail below.

2.2.1 Press-fit flywheel on motor

1. To press-fit the flywheel on the motor, the motor should first be placed on the base of the press-fitting
tool [T2] as is depicted in Figure 6.

Figure 6: Placement of the motor on the base of the press-fitting [T2]

2. The assembly press-fitting [T2] is firmly fixed in the hand-press [T3]. There is also a disassembly press-
fitting available to unmounts the flywheel from the motor in case this is necessary.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 10 / 15

3. The press-guide [T2] is positioned over the base with the motor and positioned under the hand-press as
depicted in the picture below. The press-guide can be positioned in two ways, but only the position where
the press-guide fully rests on the base allows good press-fitting as pointed out in the picture.

Figure 7: Positioning of the press-guide over the base

4. Next the flywheel is carefully placed inside the press-guide and made to rest on the axis of the motor as
showed in Figure 8. The axis should already rest on the base of the shaft of the flywheel, although the
flywheel will likely not be level. This is no problem, because it will be pushed to its proper orientation by
the press-fitting when force is applied.

Figure 8: The flywheel is carefully placed on the axis of the motor

5. Now the flywheel can be press-fitted on the motor by applying a manual force on the hand-press. This is
a delicate task, so take your time. It is possible to take the motor with flywheel from the base to visually
inspect the assembly. This allows you to confirm whether the flywheel is already pressed far enough on
the motor shaft or not. It is some effort to put it back, but better safe than sorry.

NOTE: Pressed far enough means that the shaft is no longer visible, only the shaft base. Figure 9
indicates the shaft base.

Some Loctite 406 glue should be applied inside the shaft. Although it is in principle possible to remove the
flywheel after assembly, for the flight model the flywheel should in no case be removed and put back
after assembly because it is very likely that this will weaken the assembly. For the final assembly glue
should therefore definitely be applied.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 11 / 15

Figure 9: On the left is a picture pointing out the base of the motor. On the right the motor is shown with
the flywheel press-fitted on its shaft

6. With the flywheel firmly in place, it is time to connect the motor to the drive electronics PCB. First M1
nuts should be placed as indicated on the drive electronics integrator tool [T4]. The motor with the
flywheel can now be slid into place. Additionally a drive electronics adaptor [P7], drive electronics PCB
(for the X- and Y-axis [P8] is required, while [P9] should be used for the Z-axis) and four M1 x 6 screws
(where [P11] should be used for the X- and Y-axis and [P12] for the Z-axis) are required.

First the drive electronics adaptor is positioned over the bottom side of the motor. Next the drive
electronics PCB should be positioned on top of that, in the orientation as presented in the picture below.
This assures that the flex-rigid connector is properly positioned. The whole assembly can then be fixed
together by using the M1 x 6 screws. [T4] keeps the nuts in place, allowing for easy fastening.

Figure 10: The placement of the M1 nuts is demonstrated on the left, while an overview of the required
parts and their orientation is given on the right.

7. The end-result for the X- and Y-axis reaction wheels looks like in. The Z-axis reaction wheel will have one
set of connector pins pointed in the local –Z direction of the motor, i.e. away from the flywheel.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 12 / 15

Figure 11: Completed subassembly, pictures are for X- and Y- axes

8. Finally, connect the flex-rigid to the connectors on the drive electronics PCB. This is necessary, because it
will be much more difficult to connect it once the reaction wheel subassemblies are integrated on the
bracket. Also, apply glue at the edges of the M1 nuts if necessary.

Figure 12: Final subassembly with flex-rigid connected.


This is necessary before further integration

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 13 / 15

2.3 Final assembly


The final assembly of the reaction wheel system is relatively straightforward. The steps to follow are
described below.

1. Start with the either the X- or Y-axis, because you won’t be able to access the [P12] screw heads for the
X- and Y-axes reaction wheels once the Z-axis reaction wheel is in place. Here we will start with the Y-
axis. For this axis four M1.6 x 8 with rounded head are required [P12]. It helps to fix them with some
Flashbreaker or Kapton tape as depicted in Figure 13 until the next step.

Then position the reaction wheel subassembly adaptors [P10] over the threads on the opposite side as
indicated in the left picture of Figure 13.

Figure 13: In the left picture the required screws for either the X- or Y-axis or the Z-axis are indicated,
whereas the right picture shows how to temporarily fix the screws.

2. The next step is to position one of the reaction wheel subassemblies designated for the X- and Y-axis
over the screw threads and fix them by fastening the screws. It is good practice to fasten the screws one
by one, removing the earlier attached tape for one screw at a time.

3. With the subassembly in place, it is time to fine-tune its orientation. This is a delicate step due to the high
potential for misalignments. Misalignments could result in contact with the protective caps (increasing
friction and disturbance vibrations) as well as misalignments with the body reference frame. Therefore a
caliper has been adjusted to support this operation. By using this caliper (available via Rob van der List of
DEMO-LR) the distance between the bracket and the M1 nut that was used to fasten the motor can be
measured as shown in Figure 14. By slightly fastening or loosening the torques screws that were used to
mount the subassembly on the bracket the alignment can be adjusted. Of course the aim is that the
distance at each corner is equal, although a maximum variation of 0.1 mm. should be reached.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 14 / 15

Figure 14: using the customized caliper to fine-tune the subassembly mounting

4. Mount one of the protective caps designated for the X- and Y-axis over the motor. This requires some
tweaking and adjusting in order to allow a good fit. A good fit is when the motor is not hindered in its
rotation by the protective cap.

Figure 15: A good fit is when the motor is not hindered in its rotation by the protective cap.

5. Apply Loctite 406 glue at the screws that were used to fix the subassembly and the cap using tool T7.

6. Repeat steps 1 – 4 for the X-axis. The picture below shows your assembly before mounting the protective
cap for the X-axis.

Figure 16: The picture your assembly before mounting the protective cap for the X-axis.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-1007
Date 07-02-2012
Technical Note Issue 1.1
Page 15 / 15

7. Now also repeat steps 1 – 4 for the Z-axis. This time use the M1.6 x 8 screws with sunk head [P11] in
step 1 and the horizontal cap [P15] in step 3. The left picture of Figure 17 shows how the assist the
assembly using Flashbreaker tape, the right shows the assembly after completing steps 1 - 4 for the Z-
axis.

Figure 17: The left picture below shows how the assist the assembly using Flashbreaker tape, the right
shows the assembly after completing steps 1 - 4 for the Z-axis.

8. The final assembly can finally be connected to ADCS PCB if the connectors are placed on the appropriate
place on the ADCS board. Using an M3 bolt, washer and nut the reaction wheel assembly can be
connected at the connection point that is not at an extremity of the ADCS PCB. The final assembly
integrated on a test PCB is shown in where the assembly was not yet fixed using the M3 bolt.

Figure 18: The final assembly on the test PCB with the X- and Y-axis indicated. Note that the assembly will
be assembled ‘upside-down’ in the Delfi-n3Xt stack. Also, the assembly was not yet fixed at the center
corner using the M3 bolt.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-1007 [1.1] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Assembly
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Appendix C
Technical Drawings

This appendix includes the technical drawings of all components and asem-
bly tools that have been developed for the Reaction Wheel System. First
the Technical Drawings of the RWS components are presented, followed by
the various tools that were developed to facilitate assembly and integration.

C.1 Reaction Wheel System Components


The next pages present the technical drawings of the developed RWS com-
ponents.

187
C.2. Reaction Wheel System Assembly Tools

C.2 Reaction Wheel System Assembly Tools


The next pages present the technical drawings of the developed RWS as-
sembly tools.

DELFI-N3XT REACTION WHEEL SYSTEM 197


This page intentionally contains only this sentence.
Appendix D
Labview Characterization Manual

In order to perform the characterization test, a special script was developed


using LabView. This document contains the manual that describes the
necessary steps to set up the characterization test and explains how to use
this software.

205
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 1 / 10

ADCS – Reaction Wheel Test Script Manual

Description: Explanation of Reaction Wheel Test Scripts for LabView [SLR0771].

COMMS

Launch
MechS

T3μPS
CDHS
ADCS

ITRX

SDM

GSN
GSE
EPS

TCS
STS
Subsystem(s)
involved:

Revision Record and Authorization

approved
Reviewer
checked
Issue

Affected

PM
Date Author / Editor Description of change
Section(s)

1.0 23-12-2011 Teun Hoevenaars All First full version

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 2 / 10

Table of Contents
1  REQUIRED EQUIPMENT ____________________________________________________________ 3 

2  TEST SET-UP _____________________________________________________________________ 4 

3  USER INTERFACE _________________________________________________________________ 7 

4  EXAMPLE OF RESULTS OF TORQUE TEST ____________________________________________ 10 

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 3 / 10

1 Required equipment
This chapter lists the required equipment to perform Reaction Wheel characterization using the specially
developed LabView software stored under [SLR0771].

The parts required to use the LabView Reaction Wheel characterization scripts are:

Equipment Purpose

Computer with LabView, four serial ports, two Run the script and communicate with the peripherals
USB ports (typically located in the cleanroom)
Delta Elektronika Power Supply ES 030 - 10 Provide power to reaction wheels. Power supply is also
possible via USB interface, but supply voltage will not be
stable 3.3V
Hewlet Packard 34401A Multimeter Measure current drawn. Currently only possible to measure
current drawn by a single reaction wheel
Tektronix TDS 2014B Oscilloscope Measure speed by interpreting signal from one of the Hall
sensors.
Bus Pirate Communicate with the reaction wheel electronics, Bus
Pirate v3.5 available.
Dedicated connector cables (available in Facilitate connection between reaction wheels and
Reaction Wheel Hardware Box, stored in the BusPirate via serial connection (which also allows
cleanroom) interfacing with reaction wheels via the vacuum oven serial
connectors)

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 4 / 10

2 Test set-up
Before the LabView scripts can be used, all equipment needs to be connected appropriately. The test set-up
is divided in a back-end and front-end test set-up. The bacl-end set-up always needs to be in place, whereas
the front-end set-up depends on the testing that you want to perform.

An overview of the back-end set-up is presented in Figure 1. Pay special attention to the following parts:

 The switches on the back of the power supply are switched to ‘P’, which means that both current and
voltage are operated by the computer. LabView will give an error if this is not the case.

 The connector to interface with the BusPirate is already manufactured and should be available in the
Reaction Wheel hardware box. Pay special attention to the pull-up connection, connecting the 3V3 pin
with the VPU pin, which sometimes comes loose.

Currently there are two options for the front-end set-up:

1. Test configuration 1: Interface with the Reaction Wheel testboard, which allows measurements of
the rotation speed through monitoring of a single Hall-signal using the oscilloscope. The appropriate
connection with the test-board is visualized in Figure 2.

2. Test configuration 2: Interface with the reaction wheel engineering model test PCB. This test set-
up allows all three wheels to be commanded simultaneously.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 5 / 10

Figure 1: Schematic representation of the back-end test set-up to perform reaction wheel characterization.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 6 / 10

Figure 2: Front-end test set-up for test configuration 1, which allows direct speed measurements.

a) b)

c)

Figure 3: The front-end test-setup for test configuration 2 which allows simultaneous commanding of all three reaction
wheels. a) the top-side of the developed test PCB for test configuration 2. b) the connectors developed for test
configuration 2 (cannot be wrongly connected) c) picture of the engineering model mounted to the bottom-side of the
test PCB.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 7 / 10

3 User Interface
When you launch the I2C_Characterization.vi you will see below user interface which consists of an
equipment configuration tab on the left and a test configuration tab on the right.

Figure 4: LabView reaction wheel user interface

In principal all settings should be correct, but if for some reason e.g. the port addresses of one the hardware
is changed, you can change that in one of the configuration tabs before running the LabView script. The
Oscilloscope tab also includes the PWM setting that is used during calibration, which we will get back to in a
moment.

The LabView script is started by pressing the ‘run’-button in the top-left corner.

Figure 5: The 'run'-button starts the LabView script

The initialization process that is then started directly allows verification of the full test set-up, because as part
of the initialization process the oscilloscope is calibrated, which requires the wheels to speed up sufficiently.
You should manually verify on the oscilloscope whether the PWM setting that was set is sufficiently high so
that the oscilloscope calibrates at a gridsize of 2.5ms as below. This is also visualized in Figure 6.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 8 / 10

Figure 6: Overview of the status of the various equipment during initialization. The oscciloscope should calibrate to 2.5 ms
and the poewr supply should indicate 3.3 Volt

Unfortunately there is no indicator that communicates when the initialization process is complete, but you
should have a little patience. Now the program should be ready to start whatever you want to do/test. The
first thing that you should know about your reaction wheel(s) is whether it is in open-loop modus or closed-
loop modus.
Open-loop would mean that every time we speak about a command, we are talking directly about a
PWM-setting. Closed-loop means that the command is a value for ∆n/∆s and thus represents a desired
acceleration. Once a command is given in closed-loop control, the reaction wheel in question will therefore
continue to accelerate with that value.

Once the question on open- or closed-loop is answered we can focus our attention again on the user
interface. As can also be seen in Figure 4, there are four tabs:

1. Test RWs. This is a very simple test to confirm whether all reaction wheels are working (have the
right I2C addresses been entered in the I2C configuration?) and whether they accelerate in the right
direction when the switch is set to CCW (which represents a positive direction). Because the CW/CCW
division might at the current time no longer be an adequate division of rotation directions, it is good
to know that CW relates to command 0x03 and CCW to command 0x04. This should be sufficient
information to allow verification of the positive acceleration direction in all three reaction wheels.
2. Manual. This simply allows you to command each individual reaction wheel with a manual command.
3. Stability test. This is a test script that was once developed to assess reaction wheel stability at a
certain rotation speed. It is however currently of no use and not operational. It is therefore best to
ignore this tab.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 9 / 10

4. Torque test. This is a test developed to continuously monitor both the rotation speed and power
consumption of a reaction wheel connected to the testboard. There are several settings to be
considered:
a. Experiment title. This text is added to the excel sheet that is generated during the test.
b. Measurement time. This sets the running time of individual test runs. 40 seconds is
generally enough when a standard torque test is performed; i.e. in open-loop control modus
with PWM set at a maximum.
c. # of tests. Sets the number of consecutive runs that are automatically performed as part of
one test. After each run, there is a dead time of 100 seconds to allow the reaction wheel to
decelerate back to zero.
d. PWM setting. Sets the control command, which is a PWM setting in the open-loop control
modus. 1023 was the maximum for the AtTiny861 microcontroller. If the Atmega88PA is
implemented, the maximum will be 256.

Finally it should be pointed out again that after each test, the script waits 100 seconds to allow the
reaction wheel to decelerate back to zero. During this time you should not save/change the name of
the generated Excel sheet because that will result in an error. Excel will briefly light up when the
script finishes its routine. This is not clearly indicated in the script, but it is not claimed to be perfect.

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
Document DNX-TUD-MA-0869
Date 03-02-2012
Technical Note Issue 1.0
Page 10 / 10

4 Example of results of torque test


The results from the torque tests should look as shown below. Using e.g. the Backward Euler Method it is
possible to calculate from these measurements the motor torque if the moment of inertia of the motor rotor-
flywheel combination is sufficiently well-known. Next it is also possible to relate these torques to the power
consumption in order to establish the characteristic motor curves as presented in [SLR0284,V2.2].

Table 1: Example of generated Excel sheet by torque test script


2012‐01‐20_B#1_YW_EM_prot‐cap_V3.3_torque_CW
Date 20‐1‐2012
Time (at beginning of the  9:12 AM
Measurement time (s) 40
PWM setting (‐) 1023
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.011 0 0 0.012 0 0 0.012
0.257 0 0.012 0.256 0 0.012 0.255 0 0.011
0.511 0 0.012 0.514 0 0.012 0.51 0 0.012
0.765 0 0.012 0.768 0 0.012 0.764 0 0.012
1.019 0 0.012 1.021 0 0.012 1.02 0 0.012
1.271 0 0.012 1.273 0 0.012 1.272 0 0.012
1.524 0 0.012 1.526 0 0.012 1.525 0 0.012
1.777 0 0.012 1.78 0 0.012 1.779 0 0.012
2.031 0 0.012 2.034 0 0.012 2.033 0 0.012
2.491 1.485E+39 0.59 2.493 1.485E+39 0.59 2.493 1.485E+39 0.59
3.41 5660.114 0.516 3.41 5746.858 0.518 3.43 5769.635 0.516
4.407 9803.121 0.452 4.427 9740.24 0.449 4.428 9868.031 0.452
5.52 13042.061 0.393 5.554 13513.514 0.391 5.54 13392.137 0.393
6.617 15626.986 0.347 6.654 15789.474 0.34 6.64 15788.474 0.342
7.718 18294.06 0.311 7.755 18070.944 0.309 7.74 18515.731 0.305
8.817 19738.407 0.279 8.852 20266.932 0.275 8.84 19460.502 0.273
9.919 21430.508 0.253 9.954 21737.188 0.248 9.941 22056.773 0.247
11.021 22727.324 0.229 11.055 23437.502 0.224 11.04 23085.229 0.223
12.121 23814.135 0.211 12.155 23435.24 0.205 12.139 24203.494 0.209
13.22 24997.491 0.196 13.255 24590.167 0.197 13.243 25000.129 0.194
14.317 25853.95 0.186 14.353 25005.081 0.178 14.34 25862.067 0.178
15.42 26313.01 0.167 15.453 26310.159 0.171 15.44 26315.79 0.17
16.519 26321.422 0.168 16.554 26736.746 0.169 16.541 26785.787 0.167
17.62 26310.293 0.157 17.654 27266.607 0.154 17.64 26321.418 0.154
18.718 27777.779 0.155 18.756 27777.779 0.153 18.741 27269.885 0.155
19.818 27278.776 0.153 19.853 27269.74 0.149 19.841 26791.551 0.142
20.918 27278.216 0.148 20.955 27771.658 0.145 20.941 26785.714 0.148
22.022 27774.752 0.14 22.056 28301.966 0.141 22.043 27783.979 0.148
23.119 28298.672 0.145 23.154 28295.453 0.145 23.14 28311.462 0.139
24.218 28314.613 0.14 24.303 27765.535 0.133 24.24 28305.185 0.141
25.319 27774.606 0.136 25.403 27784.05 0.14 25.342 27777.775 0.139
26.418 28305.103 0.131 26.505 28305.106 0.135 26.44 28305.181 0.139
27.52 28292.001 0.136 27.604 27777.704 0.134 27.542 28305.185 0.138
28.638 28852.758 0.13 28.705 28305.185 0.134 28.66 28846.154 0.137
29.737 28839.391 0.137 29.805 28292.468 0.129 29.758 28301.885 0.135
30.837 28842.812 0.133 30.905 28301.889 0.138 30.859 28842.812 0.13
31.938 28849.495 0.133 32.006 28842.731 0.133 31.959 28839.553 0.135
33.037 28852.919 0.133 33.105 28308.402 0.13 33.057 28295.376 0.135
34.137 28301.889 0.135 34.204 28849.579 0.132 34.158 28881.227 0.132
35.235 28305.185 0.133 35.304 28301.808 0.131 35.257 28853.088 0.136
36.337 28846.238 0.13 36.406 27784.054 0.136 36.36 28839.391 0.13
37.438 28800.449 0.135 37.506 27774.681 0.13 37.458 28305.106 0.133
38.536 28839.553 0.133 38.606 28301.889 0.134 38.559 28301.889 0.13
39.637 28893.86 0.13 39.706 29408.291 0.135 39.657 29401.588 0.128
40.737 27777.929 0.134 40.805 28846.154 0.13 40.757 29404.735 0.129

2011. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-MA-0869 [1.0] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheel Characterization
not permitted, except prior and express written permission of Delft University of Technology. Manual.docx
This page intentionally contains only this sentence.
Appendix E
Derivative Gain as Function of
Damping Factor

In Section 4.3 the closed loopt transfer function of the RWS was derived and
is repeated here for convenience:

Kp 
Kp Td z 1
H pz q   Ts z
 (4.36)
πIR
 LkRE
Kp Td 
1 z 1
kE Cv R
Kp
30km Ts z km

Table E.1: Simplified parameters to assist partial fraction expansion of the


RWS closed loop transfer
Identifier Description

A πIR
30km  LkR E
Kp Td 1
Ts

Cv R
B kE km Kp

Kp Td
C Ts

D Kp

To start the process of partial fraction expansion, this equation is simplified


by introducing the constants of Table E.1, resulting in:

C z z 1
H pz q 
A z z 1
D
B
 A 1 B pC z DqzA  C (E.1)
A B

217
Derivative Gain as Function of Damping Factor

To be able to use partial fraction expansion to identify the inverse Z-transform


functions later on in the process, it is also required to divide both sides of
Equation E.1 by z:

H pz q
 A 1 B pC Dqz  C
(E.2)
z z z A
A B

Equation E.2 is in the form required to perform partial fraction expansion.


More specifically, partial fraction expansion allows Equation E.2 to be writ-
ten in the form:

1 pC Dqz  C
 A 1 B Cz1 C2
z
(E.3)
A B z z A A
A B
A B

To determine coefficient C1 , both sides of Equation E.3 are multiplied by z


and subsequently setting z  0:

  

1 pC Dqz  C 
  C1 C2 


z
z z z
A B z z A  z A 
A B 
z 0
A B 
z 0 (E.4)
C pA B q
A
C1
To determine coefficient C2 , both sides of Equation E.3 are multiplied by
z  A AB and subsequently setting z  A AB :



 

z
A 1

p
C D z C 

q   z
A C1 
A B A B z z A   A B z z  A
A B z  AAB A B




z
A C2 

A B z A 
A B z  AAB
pC DqA  C pA Bq
C2
A
(E.5)

Now replacing C1 and C2 in Equation E.3 with the derivations in Equation


E.4 and Equation E.5 and inserting the result in Equation E.2 results in

 
H pz q  C pA B q pC DqA  C pA Bq 
A 1
 (E.6)
z B Az A z A
A B

218 MSC THESIS A.G.L. HOEVENAARS


Simplifying Equation E.6 and multiplying both sides again by z finally re-
sults in the partial fraction expansion of 4.36 with the simplified parameters
as introduced in Table E.1.


H pz q  
C C D C z
z A
(E.7)
A A B A A B

A comparison between Equation E.7 and common Z-transforms allows iden-


tification of the various components of the characteristic polynomial. Table
E.2 shows the Z-transforms that are identified and their respective time-
domain representations. The resulting polynomial using the simplified pa-
rameters therefore is:


tk
nptk q  dptk q  CA uptk q
C C D A
(E.8)
A A B A B
with the simplified parameters as expressed in Table E.1.

Table E.2: Some common Z-transforms and their time-domain representa-


tion, source: Wescott 2006 [16]
N pz q nptk q Comments
"
dptk q dptk q 
1 tk 0
1
0 tk 0
"
z
ptk q
dtk u d must be a constant, uptk q 
0 tk  0

z d 1 tk ¥0

Using Equation E.8 it is possible to relate the derivative time to the damping
in the system. The factor with exponent tk determines the stability of the
system and should be smaller than zero for a stable system. tk is the discrete
time variable and each increment of tk corresponds to a time-step Ts . We
now introduce a new variable ζrws , which is the damping factor of the closed
loop system over a time period of one second; half the time between input
commands. We would like to determine a derivative time such that the
system stabilizes before a new command is received. This would result in
an average delivered torque over the two second period that is close to the
desired torque and therefore minimize the impact of the non-linear behavior
due to e.g. stiction.
Expressed in the discrete time-variable tk , one second corresponds to
tk  T1s . The factor with exponent tk in Equation E.8 thus relates to the
damping factor ζrws as:

DELFI-N3XT REACTION WHEEL SYSTEM 219


Derivative Gain as Function of Damping Factor

A
 { aζ
1 Ts
(E.9)
rws
A B
Or, using the explicit expressions for A and B:


πIR
 LkR E
Kp Td 1
{ aζ

30km

Ts
 2 Ts
(E.10)
 LkR
rws
πIR 1 Cv R
30km
E
Kp Td Ts kE km Kp

This equation can be rewritten to express Kp Td as a function of ζrws and


the general motor parameters;

{ ?ζ T 

 ?  30k
2 Ts
rws s Cv R LkE πIR
1  2{Ts ζ
Kp Td kE Kp (E.11)
rws km R m

220 MSC THESIS A.G.L. HOEVENAARS


Appendix F
Reaction Wheel System Test
Document

Significant testing has been performed and the characteristic performance


curve that developed is used extensively throughout the development pro-
cess. This document contains all relevant tests that have been performed
and serve as an input in the verification process.

221
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 1 / 51

Reaction Wheels System Test Document

Description: Report of all tests performed regarding the Reaction Wheel System.

COMMS

Launch
MechS

T3μPS
CDHS
ADCS

ITRX

SDM

GSN
GSE
EPS

TCS
STS
Subsystem(s)
involved:

X X

Revision Record and Authorization

approved
Reviewer
checked
Issue

Affected
Date Author / Editor PM Description of change
Section(s)

1.0 13-10-2010 Erwin Dekens First draft


2.0 01-08-2011 Teun Hoevenaars All Update to Faulhaber 1202
motor
3.0 09-12-2011 Teun Hoevenaars All New set-up of document and
merging with SLR[0868]
3.1 08-02-2012 Teun Hoevenaars 3, 6-8 All tests processed
3.2 08-02-2012 Teun Hoevenaars 6-7 Interference between reaction
wheels added.
3.3 21-02-2012 Teun Hoevenaars All Minor errors improved

Action Items
Applicable
TBW
TBD
TBC

Description of action item


Section(s)

x Chapter 7 Degradation testing still has to be performed and documented.


Interference by reaction wheels on reaction wheels could be worth
investigating further.

List of Used References


SLR code Version Data/Variable
SLR0858 -
SLR0766 [1.2]
SLR0284 [2.2] RWS requirements
SLR0869 [2.0] LabView characterization test script manual
SLR0872 [2.4] Temperature range
SLR0979 [1.0] Operational lifetime for Faulhaber brushless DC motors
SLR0989 [1.0] NLR vibration test series report, vibration spectra used
SLR1008 [0.9] Launch loads

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 2 / 51

Table of Contents
1  INTRODUCTION __________________________________________________________________ 4 

2  REQUIRED TESTING FOR THE REACTION WHEEL SYSTEM _____________________________ 5 

2.1  Motor consistency_____________________________________________________ 5 

2.2  Launch loads _________________________________________________________ 5 

2.3  Thermal-vacuum environment ___________________________________________ 5 

2.4  Motor performance variation ____________________________________________ 5 

2.5  Degradation testing ___________________________________________________ 5 

2.6  Interference with surrounding hardware ___________________________________ 5 

3  MOTOR CHARACTERIZATION _______________________________________________________ 6 

3.1  Determine motor variables. _____________________________________________ 6 

3.2  Determine performance curve ___________________________________________ 8 

4  MOTOR CONSISTENCY____________________________________________________________ 10 

5  LAUNCH LOADS __________________________________________________________________ 12 

5.1  Expected launch loads ________________________________________________ 12 

5.2  Vibration test campaign at 3ME on RWS prototype __________________________ 14 


5.2.1  Test set-up ____________________________________________________________________ 14 
5.2.2  Excitations_____________________________________________________________________ 18 
5.2.3  Results _______________________________________________________________________ 18 

5.3  Vibration test campaign at NLR on RWS engineering model ___________________ 20 
5.3.1  Test set-up ____________________________________________________________________ 20 
5.3.2  Excitations_____________________________________________________________________ 21 
5.3.3  Results _______________________________________________________________________ 21 

6  THERMAL-VACUUM ENVIRONMENT ________________________________________________ 24 

6.1  Impact of temperature on performance ___________________________________ 24 

6.2  Heat load on the satellite ______________________________________________ 25 


6.2.1  Heat load of a single reaction wheel. ________________________________________________ 26 

6.3  Impact of vacuum on motor performance. _________________________________ 28 

6.4  Degradation testing __________________________________________________ 29 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 3 / 51

7  INTERFERENCE WITH SURROUNDING HARDWARE __________________________________ 30 

7.1  Interference between reaction wheels ____________________________________ 30 

7.2  Interference from RWS on RRS _________________________________________ 31 

7.3  Interference from RWS on MTM _________________________________________ 31 

8  CONCLUSIONS & RECOMMENDATIONS _____________________________________________ 32 

APPENDIX A.  RAW DATA FOR CONSISTENCY TESTS ___________________________________ 33 

APPENDIX B.  CHARACTERISTIC CURVES OVER DESIGN TEMPERATURE RANGE.___________ 36 

APPENDIX C.  MATLAB CODE TO GENERATE .ARB-FILES FOR EXECUTION OF SINESWEEP AND
RANDOM VIBRATION FILES IN SIGLAB _________________________________________________ 39 

APPENDIX D.  VIBRATION LEVELS DURING TEST CAMPAIGN AT NLR _____________________ 42 

APPENDIX E.  GRAPH OF THE DEVELOPMENT OF TEMPERATURE OVER TIME BEFORE POST-
PROCESSING 51 

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 4 / 51

1 Introduction
The test campaign for the Delfi-n3Xt Reaction Wheel System (RWS), which is part of the Attitude
Determination and Control System (ADCS), that is described in this report, concerns the RWS design which is
based on Faulhaber 1202 004 BH micromotors. An earlier RWS design for Delfi-n3Xt, based on Faulhaber
2209 T 005 B motors, failed the vibration tests as described in version 1.2 of this document. In the resulting
design iteration it was decided to change to Faulhaber 1202 004 BH motors. Test plans and reports for the
earlier iteration can be found in documents DNX-TUD-TP-0194, DNX-TUD-TR-0376, DNX-TUD-TR-0766 and
DNX-TUD-TR-0868.

For the second design iteration, based on the the Faulhaber 1202 004 BH motors, quick vibration testing was
required and therefore vibration testing was done in two steps. The first step was vibration testing of a
prototype using a shaker available at the faculty of Mechanical, Maritime and Materials Engineering (3ME) in
April 2011. The second step in the campaign was executed in October 2011 and concerned vibration testing
of the first engineering model (EM1) of the RWS at the National Aerospace Laboratory (NLR) in Flevoland.

On October 31st, 2011 the Chair of Space Systems Engineering (SSE) of TU Delft signed a launch contract
with ISIS for launch of Delfi-n3Xt in September 2012 using the Dnepr launch vehicle. The vibration loads the
RWS must be able to resist are therefore those dictated by the qualification loads for the Dnepr launch
vehicle.

Additional testing was then initiated to assess the impact of thermal-vacuum on the performance of the RWS.
Also interference tests were carried out to assess the impact of the reaction wheels on other systems. To
facilitate easy comparison of performance a test was designed to characterize the performance of a motor.
This test will be introduced in Chapter 3. Next motor consistency is tested in Chapter 4 and Chapter 5 covers
the tests that were performed to verify the RWS for launch loads. The impact of thermal-vacuum environment
is assessed in Chapter 6 and finally the interference with surrounding hardware is assessed in Chapter 7.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 5 / 51

2 Required testing for the Reaction Wheel System


This chapter describes which tests will be required and what requirements and constraints are the driving
factors for those tests.

2.1 Motor consistency


Six vacuum-rated Faulhaber 1202 004 BH motors where ordered from Faulhaber. All are briefly tested to
verify that they all have similar performances.

2.2 Launch loads


The physical structure of the RWS should be able to withstand the loads that it will encounter before-, and
during the launch with the Dnepr launch vehicle. The requirements for those tests are specified by ISC
Kosmotras, the provider of the Dnepr and are processed by ISIS into qualification levels [SLR1008].

2.3 Thermal-vacuum environment


Delfi-n3Xt will operate in space and will therefore be subjected to a vacuum. Additionally, temperatures in
space can be highly variable. The expected temperature range at ADCS-level is described in [SLR0872]. The
design temperature range for the reaction wheels is -15 °C – 45 °C.

2.4 Motor performance variation


Because three motors are used in the RWS and a new batch will be ordered for the flight model and
additional models it is verified that motor performance is similar for the batch of three vacuum-rated motors
that are ordered for use in the development phase.

2.5 Degradation testing


The motors used in the RWS are vacuum-rated Faulhaber 1202 004 BH motors. Faulhaber’s technical
specifications document [SLR0979] states that the lifetime of BLDC motors “[o]n average […] may exceed
10,000 hours if he motors are operated within the recommended values indicated on the data sheet.”.
However, nothing is said about possible degradation of the motor performance due to wear. Test should
therefore be done to assess the potential of degradation over time.

2.6 Interference with surrounding hardware


The reaction wheels produce an audible amount of vibrations, while also electro-magnetic fields are created
by the electrocoils that are part of the motors. The impact on other subsystems is investigated in this
document.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 6 / 51

3 Motor characterization
Throughout the development of the Reaction Wheel System motor characterization is performed in many
occasions. This chapter will elaborate a bit more on the rationale behind the characterization test developed
and how the results say something about the reaction wheels. How this test should be performed is described
in [SLR0869].

Characterization of the reaction wheels is performed to quickly assess the overall performance of the reaction
wheels. Based on the requirements for the Reaction Wheel System as described in [SLR0284] the motor
performance parameters that we are interested in are:

 Power consumption within limits


 Dynamic range of each reaction wheel 1.56·10-3 Nms.
 minimum torque of 5.5·10-6 Nm

An effective characterization test therefore covers those three aspects. The test approach followed to assess
those three aspects is the following:

1. Run test script to accelerate reaction wheels using maximum input power and establish the
associated torque and power curves. The torque curve presents the torque as a function of rpm and
the power curve presents the associated maximum input power as a function of rpm.

2. Determine associated variables of terminal resistance R, static friction C0 and dynamic friction Cv (and
kE when of interest). These variables are variable with environment and their behaviour is of interest
to the performance of the motor over the full design range.

3. Establish curve that presents the power consumption to deliver the minimum torque as a function of
rotation speed.

These activities will be discussed separately below.

3.1 Determine motor variables.


The LabView characterization script discussed in [SLR0869] applies a step input of maximum PWM-setting to
the reaction wheel motor. Due to the automatic commutation, this step input causes the motor to draw the
maximum amount of power it possibly can to accelerate in a given direction. This acceleration of the reaction
wheel corresponds to the torque that is generated by the motor. During the test campaign, both rotation
speed and current drawn are continuously monitored. This data allows the torque and input power graphs to
be plotted.

The measurements can then be correlated with the theoretical curves to determine the variables describing
the reaction wheel performance. First the measured current is correlated with motor parameter R using the
equation provided by Faulhaber [SLR0979]:

U  kE n
I A  Equation 1
R

kE is the back-EMF constant [V/rpm] which is sensitive to temperature changes, just as R. They can be
determined by matching the measured current curve with the theoretical currents. It is common practice to
use the input power instead of the current to do the matching, which is in principle the same, because V was
determined to be constant. Figure 1 gives an example of the power curve.

Pin  U  I  W  Equation 2

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 7 / 51

0.8

0.7

0.6
power [Watt]
0.5

0.4

0.3

0.2

0.1

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
Figure 1: Example of a input power curve that is correlated with the measured input power during characterization tests.

Next the torque curve is established using the following relations provided by Faulhaber [SLR0979]:

kM I  T  TR Equation 3

Where I(n) is the current drawn as a function of rotation speed n, kM is the torque constant [Nm/A], T [Nm]
denotes the resulting torque and TR is the resistance torque. The resistance torque can also be expressed as:

TR  kM I 0 Equation 4

With I0 the no-load current (or friction current). The no-load current on its turn is a function of the static
friction torque C0 and a viscous damping factor Cv according to the equation:

C0  Cv  n
I0  Equation 5
kM

Combining these equations gives an expression for torque as a function of rpm:


 U  k E n C0  Cv  n 
T    kM [ N ] Equation 6
 R kM 

R and kE was already experimentally determined when the power curve was correlated with measurements.
Next kM is the torque constant [Nm/A], which is also a fixed motor parameter that is verified by the
manufacturer. Of course this, and the variable kE are also verified in-house, as will be discussed in SEC REF.
Therefore, the remaining parameters are the static friction torque C0 and a viscous damping factor Cv, which
signify a constant off-set of the torque and the slope of the torque versus rpm curve. Cv and C0 can be thus
be determined by correlating the theoretical torque curve with the measured torque curve.

Figure 2 gives an example of a theoretical torque curve that is correlated with measurements. Why there is
such a spread in the measurements is discussed in the Reaction Wheel Design Document [SLR0284].

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 8 / 51

-4
x 10

1.5
Torque [Nm]

0.5

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
Figure 2: Example of a torque curve that is correlated with measured torque. Why there is such a spread in the
measurements is discussed in the Reaction Wheel Design Document [SLR0284].

Additionally the measured torque can be used to determine the output power, which relates to the torque in
the form:

 Equation 7
Pout  T  n 
30

Figure 3 presents an example where Figure 1 has been extended with the theoretical output power curve.
The same spread as observed in Figure 2 is evident there.

0.8

0.7 Specified Pin, max


Measured P
in, max
0.6
Specified Pout, max
0.5 Measured Pout, max
power [Watt]

0.4

0.3

0.2

0.1

0
0 5,000 10,000 15,000 20,000 25,000 30,000
rotation speed [rpm]
Figure 3: Example of the power curve together with the output power curve

3.2 Determine performance curve


With the values for the terminal resistance R, the static friction C0 and the dynamic friction Cv determined
using the measurements, a performance curve can be established which shows the power consumption to

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 9 / 51

deliver the minimum torque as a function of rotation speed. The equation that is used to establish this curve
is:

 Treq C0  Cv  n  Equation 8
Preq    U
 km kM 

With the additional constraint that the voltage required by the motor can never be larger than the supply
voltage:

 C  Cv  n Treq 
U  kE n   0  R Equation 9
 kM km 

With above relations and the terminal resistance R, the static friction C0 and the dynamic friction known, a
curve like presented in Figure 4 can be established. From this curve several important characteristics can be
determined:

1. Maximum useful speed. The speed where the sudden drop in the curve occurs corresponds to the
maximum speed where the motor is still able to provide the required torque. This speed corresponds
with half the dynamic range according to the relation:

H n 
 I zz max Equation 10
2 30

2. The maximum power consumption: The maximum value of the curve corresponds to the
maximum expected power consumption to deliver the required torque.
0.25

0.2

0.15
Power [W]

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 4: Example of reaction wheel performance curve. It illustrates the required power to generate the required torque
of 5.5·10-6 [Nm] and also shows the maximum speed for which the motor is able to deliver this torque.

This technique is often applied during the development of the reaction wheels because it gives a good
overview of the main motor performance characteristics.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 10 / 51

4 Motor consistency
Six vacuum-rated Faulhaber 1202 004 BH motors where ordered from Faulhaber. All are briefly tested to
verify that they all have similar performances. Because no flywheel was mounted on the motors, the motors
where accelerated very fast to their maximum speed and not enough measurement points were available to
determine the characteristic curve as introduced in Chapter 3. Table 1 does however list the maximum
rotation speeds observed for the five motors tested and Appendix A provides the raw data.

Although it may seem that motor B4 and B6 perform significantly below the other motors, this is not
considered representative of the other motors. The data in Appendix A shows that speeds above 30,000 our
outliers and all motors have average maximum speeds of approximately 28,850. This inaccuracy is partly
contributed to the state of the test script when these measurements were made, because the test set-up had
not been tuned well at that moment.

Motor B1 was not tested without a flywheel because the test script was not yet ready when the flywheel was
mounted on the motor.

Table 1: Maximum observed rotation speed for different motors with no flywheel mounted
Motor # Maximum observed speed
B2 31,284
B3 28,819
B4 31,277
B5 31,220
B6 28,850

In [SLR0284] it had been observed that the terminal resistance was different for the non-vacuum rated
motors from the batch ordered during the preliminary design and the vacuum-rated motors ordered for the
engineering model. Because the lubricant should not impact the terminal resistance of a motor, it is argued
that this is due to variation between motor batches. Unfortunately no conclusions on the terminal resistance
can be made based on the maximum speed alone. Thus characterization as described in Chapter 3 is
necessary.

This characterization is performed for two motors only, because it requires significant effort to assemble the
motors and although a third one was assembled as well, it did not allow for proper characterization as the
flex-rigid connector was damaged during development (see Section 5.3). The results are demonstrated in
Figure 5 and Figure 6.

In order to match the theoretical graphs with the measured points, the parameters as presented in Table 2
were determined. For both the vacuum-rated motors that were tested, the same parameters for the terminal
resistance, static friction and dynamic friction were determined to match the measurements. This gives a first
indication that motors within the same batch have roughly the same performance, but that performance
between batches may differ. However, the possibility that parameters change also among motors of the same
batch, it can only be said that characterization is a must.

Table 2: Motor parameters matching characteristic curves with measurements


Non-vacuum motor Vacuum motor
Parameter
batch batch
Terminal resistance [Ω] 16 17
Static friction 0.026·10-3 0.026·10-3
Dynamic friction 0 to 1.04∙10‐9 0.052∙10‐9

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 11 / 51

x 10
-4
Motor B#1 x 10
-4
Motor B#2
1.6 1.6
Theoretical torque
1.4 1.4 Measured torque

1.2 1.2
available torque [Nm]

available torque [Nm]


1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 5: Measured torques and theoretical torque curve plotted through the measurement points

Motor power for motor B#1 Motor power for motor B#2

Theoretical Pin, max


0.6 0.6 Measured Pin, max
Theoretical P
0.5 0.5 out, max
Measured Pout, max
power [Watt]

power [Watt]

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 6: Measured consumed- and delivered power together with the theoretical power consumption and delieverd
power to match the measurement points.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 12 / 51

5 Launch loads
The Dnepr User Manual [SLR299] states; “For spacecraft qualification and acceptance, sinusoidal, shock and
random tests are mandatory”. Additionally quasi-static and dynamic accelerations should be accounted for
and any satellite should meet the stiffness criteria. This section describes the required levels which the
spacecraft should be able to withstand for all types of loads.

5.1 Expected launch loads


During launch Delfi-n3Xt will experience a range of launch loads, each of which will be introduced in more
detail in this section.

Harmonic oscillations
The harmonic oscillations encountered during launch with the Dnepr launch vehicle are described in table
Table 3, where the direction are indicated in Figure 8. However, as the orientation in the launcher is not
known sufficiently far ahead, ISIS prescribes the sinusoidal launch load as described in Table 4.

Table 3: Sinusoidal loads for the Dnepr launch vehicle specified by ISC Kosmotras
Longitudinal axis (X) Lateral Axes (Y,Z)
Frequency sub-band [Hz] 5-10 10-15 15-20 2-5 5-10 10-15
Amplitude [g] 0.5 0.6 0.5 0.2-0.5 0.5 0.5-1.0
Duration [sec] 10 30 60 100 100 100

Table 4: Sinusoidal loads for the Dnepr launch vehicle specified by ISIS
All axs (X,Y,Z)
Frequency sub-band [Hz] 5-10 10-15 15-20
Amplitude [g] 0.5 1.0 0.5
Duration [sec] 10 30 60

Random vibrations
The random vibrations induced by Dnepr are characterized by the spectral density of vibro-accelerations. ISIS
specifies the same qualification loads as ISC Kosmotras which are presented in Figure 7. The Root Mean
Square value is 6.5 [g] and the duration of the tests should be 35 seconds per axis.
0.04
power spectral density [g /Hz]
2

0.03

0.02

0.01

0 2 3
10 10
frequency [Hz]

Figure 7: Random vibration qualification spectrum for the Dnepr launch vehicle

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 13 / 51

Figure 8: Axes definition for the Dnepr launch vehicle

Shock loads
In the Dnepr User Manual the source for the shock, or vibro-pulse, loads are identified as the activation of the
separation pyro-devices. The shock spectrum values are presented in Figure 9 and the duration is up to 0.1
sec. However, ISIS does not require any qualification testing for the shock loads [SLR1008].

Figure 9: Shock load spectrum for the Dnepr launch vehicle

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 14 / 51

Quasi-static and dynamic accelerations


Random vibrations occur in three distinct phases of the launch; transportation to the launch pad, 1st stage
burn, 2nd stage burn and 3rd stage burn. The maximum quasi-static and dynamic accelerations experienced at
the Spacecraft (S/C) - launch vehicle (L/V) interface during all phases are listed in table Table 5. The table
also lists the safety factors that should be taken into account during the design and qualification tests.

However, the ISIS ISILaunch3 manual [SLR1008] describes other qualification levels for the quasi-static and
dynamic acceleration. These are presented in

Table 5: The maximum quasi-static and dynamic accelerations experienced during launch with the Dnepr launch vehicle
as stated by ISC Kosmotras
Longitudinal (X) Lateral (Y) Lateral (Z)
Maximum encountered 7.8 ± 0.5 1.0 ± 0.7 0.5 ± 0.5
acceleration [g] (2nd stage burn) (transportation) (transportation)
Safety factor [-] 1.3 2.0 2.0
Acceleration to be 10.14 ± 0.65 2.0 ± 1.4 1±1
qualified for [g]
Gravity is inclusive in the values in Dnepr X-direction (see Figure 8). The dynamic accelerations are indicated by the ± symbol.

Table 6: The maximum quasi-static and dynamic accelerations experienced during launch with the Dnepr launch vehicle
as stated by ISIS for ISILaunch03 campaign
All axes (X,Y,Z)
Maximum encountered 10.5 (in compression)
acceleration [g] 1.3 (in tension)

Stiffness criteria
There are also stiffness criteria defined for the Dnepr launch vehicle. For launch with Dnepr, Delfi-n3Xt should
be designed with a structural stiffness, which ensures that the values of fundamental frequency of the
spacecraft, hard mounted at the separation plane, are not less than 20 Hz along any axis.

5.2 Vibration test campaign at 3ME on RWS prototype


In the early phase of development of the RWS a vibration test campaign was executed at the faculty of
Mechanical, Maritime and Materials Engineering (3ME). This section described the set-up and results.

5.2.1 Test set-up

For testing of the sinusoidal and random vibration loads a Brüel & Kjær type 4809 shaker was used (see
[SLR0768] for datasheet). The shaker was excited using a SigLab router and power amplifier as can also be
seen in Figure 11. The signal was generated using a modified Matlab .m-file, the code of which can be found
in Appendix A.

The bracket that is designed to house the reaction wheels, contains both a reaction wheels in horizontal as
well as one in vertical direction. Two wheels are necessary to test the RWS prototype at the same time in
along axis direction and in the direction perpendicular to the motor axis.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 15 / 51

Test subject configuration

Figure 10: Pictures of the RWS prototype mounted on the 3ME Brüel & Kjær type 4809 shaker.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 16 / 51

Sinusoidal and random vibration test set-up

Figure 11: A schematic representation of the test set-up as used during vibration tests on the RWS prototype

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 17 / 51

Component list
The set-up that is depicted in Figure 10 is the prototype qualification set-up, containing various custom-
designed components, the versions of which are outlined in the image below.

Test subject configuration

Planning of test campaign:


The first priority in the test campaign was to demonstrate that the Faulhaber 1202 004 BH can withstand the
launch loads. After this the reaction wheels could be developed to a more advanced stage. As can be seen in
the image above, the drive electronics are not part of the above set-up and are part of the detailed design
which will have full capabilities.

This prototype design was tested as follows in the week of 25-29 April with a positive result. The test was set
up as following.

1. Detailed motor characterization test

2. Sine-sweep excitation test

3. Random excitation test

4. Detailed motor characterization test

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 18 / 51

5.2.2 Excitations

Sine-sweep
The shaker setup succeeded in producing sine-sweep excitations in close accordance to the specifications of
both the Dnepr and PSLV launcher, which was still under consideration during the vibration test campaign at
3ME. However, the attempts to record the spectrum applied were unsuccessful so no visual proof can be
presented.

Random
By tweaking the Siglab software using the generated .arb-files using the Matlab code presented in Appendix
A, it was possible to excite the reaction wheel assembly according to the specified PSD spectrum to close
approximation. Although the spectrum was programmed to cover both the PSLV and Dnepr power spectral
densities, the spectrum closely resembles that of the Dnepr.

Figure 12: Random vibration loads exerted by the Brüel & Kjær type 4809 shaker shaker, compared with the specfied
random loads for the Dnepr launch vehicle

5.2.3 Results

The torque as a function of rotation speed is plotted in figure Figure 13 for both the horizontally mounted
reaction wheel and the vertically mounted reaction wheel, using a voltage of 3.3 V and maximum torque. This
approach is taken because no more sophisticated control logics are in place in the preliminary design. The
data was generated using a LabView program developed for this purpose [SLR0771].

The results of characterization after the vibration tests in Figure 13 show that the vertically mounted reaction
wheel performed similar after and before the vibration campaign. This indicates that the vibration tests did
not have a degrading effect on the performance of the micromotors. At the time, the program was in a
preliminary phase and the more sophisticated analysis as described in the referenced document was not yet
available. However, for this purpose the analysis demonstrated in Figure 13 suffices.

The horizontally mounted reaction wheel unfortunately sustained damage during the vibration campaign and
it was no longer possible to perform a characterization after the vibration campaign. In analyzing this damage
it was realized that the flex-rigid connectors were allowed to vibrate freely during the vibration test as can be

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 19 / 51

seen in Figure 10. This was due to the fact that the drive electronics were not developed at the time of the
tests and it was not considered to fix these connectors in another way. Inspection of the flex-rigid connector
indeed proved that it was damaged. It was therefore concluded that the flex-rigid connectors are very
vulnerable and should always be fixed during any vibration testing and preferably during handling.

Fortunately, with some effort the connector was temporarily fixed and it had been demonstrated that it could
still reach the same rotation speed. Full characterization was unfortunately not possible.

Figure 13: characterization results for the horizontally- (on the left) and vertically (on the right) mounted reaction wheels.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 20 / 51

5.3 Vibration test campaign at NLR on RWS engineering model


After the promising results of the 3ME vibration test campaign it was decided to go ahead with the RWS
design based on the Faulhaber 1202 004 BH motors. In October 2011 there was the opportunity to piggy-
back on an acceptance test by ISIS at NLR. This was a great opportunity to subject the recently engineering
model that had just been finished to more realistic sinusoidal and random vibration loads. This section
described the set-up and results.

5.3.1 Test set-up

For the test at NLR facilitated by ISIS, the RWS was assembled in GEERTsat. GEERTsat is on its turn put in
position in the ISIPOD in which it is tested. The axes indicated on GEERTsat correspond with the respective
excitation axes. The axes are also indicated in Figure 14, Figure 15 and Figure 16.

Figure 15: The Z-axis with respect to the ISIPOD (and


Figure 14: Orientation of the RWS in GEERTsat
equally for the RWS)

Figure 16: The ISIPOD with the RWS system inside on the NLR shaker table. The image shows the appropriate axes
system and is taken during vibration tests in x-direction.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 21 / 51

5.3.2 Excitations

The excitations on the ISIPOD with the RWS inside are in coherence with the acceptance test values for the
ISIPOD which is designed to meet NASA’s General Environmental Verification Standards (GEVs). Because the
vibration test requirements by NASA GEVs are significantly above those for the Dnepr launch vehicle, the
performed acceptance tests are in accordance with the Dnepr qualification test levels. The performed
vibration campaign is summarized in Table 7. The excitations as measured during the tests can be found in
Appendix B together with comparisons between resonance surveys before and after the vibration campaign in
each direction. A full description of the test can be found in SLR[0989].

Although quasi-static and dynamic accelerations were not tested explicitly, the RMS value of the random
vibration test was approximately 10 [g]

Table 7: Launch loads for the NLR vibration campaign

5.3.3 Results

In the period between the vibration test campaign at 3ME and the more realistic vibration campaign at NLR
not only the RWS itself had developed, but also the software on it as well as the characterization software.
This allowed more elaborate characterization of the reaction wheels before- and after the test campaign. For
benchmarking the test results a theoretical model of the Brushless DC (BLDC) was developed.

Figure 17 and Figure 18 show two characteristic performance graphs for BLDC motors. The theoretical graphs
are also plotted, based on the specification provided by Faulhaber. The graphs for motor B#1 (X-axis in
Figure 16) and B#3 (Z-axis in Figure 16) show great accordance to these theoretical graphs before and after
the vibration tests. The graphs also show that motor B#2 (Y-axis in Figure 16) does not perform as well as
motor B#1 and B#3; it produces significantly lower torques over the full speed range. This reduced
performance is explained by the fact that flex-rigid connector of motor B#2 was damaged during handling
already before the NLR-ISIS test campaign. Again it was concluded that the flex-rigid connector is very
vulnerable. However, the graphs show that the damage has not increased due to the tests. The tests are
therefore a full success.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 22 / 51

Figure 17: The torque vs. rotation speed graphs as part of the characterization of the reaction wheels before and after the
NLR-ISIS vibration campaign

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 23 / 51

Figure 18: The power vs. rotation speed graphs as part of the characterization of the reaction wheels before and after the
NLR-ISIS vibration campaign

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 24 / 51

6 Thermal-vacuum environment
Delfi-n3Xt will operate in space and will therefore be subjected to a vacuum. Additionally, temperatures in
space can be highly variable. The expected temperature range at ADCS-level is described in [SLR0872]. The
design temperature range for the reaction wheels is -15 °C to 45 °C. The purpose of thermal-vacuum testing
is three-fold:

1. Qualify the reaction wheels for operation over the full design temperature range

2. Assess the impact on motor performance characteristics

3. Assess the heat load on the satellite

Due to limitations in the thermal-vacuum chamber of the Space Systems Engineering department, these
objectives are split over two tests. The first test will assess the performance of a single reaction wheel over
the full range of temperatures. The second test will investigate the impact of an elevated temperature in
combination with vacuum, which also provides an opportunity to assess the worst-case heat load on the
satellite.

6.1 Impact of temperature on performance


The design temperature range of the Reaction Wheel System is -15°C to 45°C. A single motor has been
characterized at temperatures -20, -10, 0, 10, 20, 30, 40 and 50 °C to assess the impact of temperature on
the motor performance. Appendix B shows all characteristic curves that had been established from the
measurements together with the correlated theoretical curves and Figure 19. The experimentally determined
values for R, C0 and Cv at each temperature are listed in Table 8.

Figure 19: Test set-up for the thermal environment test

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 25 / 51

Table 8: Experimentally determined values of R, kE, C0 and Cv using by matching the theoretical curves with the
measurements, see
Temperature R [Ω] kE[mV/rpm] C0 [mNm] Cv [mNm/rpm]
-20 16 0.078 0.06 0.52·10-6
-10 16 0.078 0.05 0.52·10-6
0 16.5 0.084 0.035 0.52·10-6
10 16.5 0.094 0.026 0.52·10-6
20 17 0.094 0.026 0.52·10-6
30 17 0.094 0.026 0.52·10-6
40 17 0.094 0.03 0.52·10-6
50 18.2 0.094 0.03 0.52·10-6

With above experimentally derived parameters as inputs, the characteristic performance curves can be
plotted for the respective temperatures, which is done in Figure 20. It is clear that at temperatures below 0°C
the calculated power consumption to deliver the required torque is significantly higher, but drops at
temperatures above 0°C. It should therefore be avoided to operate the reaction wheels at temperatures
below 0°C.

Both at a temperature of <-10°C and >40°C, the maximum useful speed decreases to roughly 25,000 rpm.
For all other speeds the maximum useful speed is above 25,500 rpm. The optimal temperature range is
roughly between 10°C and 30°C where the power consumption is at its minimum. The increase in power
consumption from operation at the nominal temperature of 20°C to operation at 0°C is roughly a 20%
increase in power consumption.

0.35
T = -20
0.3 T = -10
T = 0
0.25 T = 10
T = 20
T = 30
Power [W]

0.2
T = 40
0.15 T = 50

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 20: The characteristic performance curve as introduced in Section 3.2 for temperatures of -20 to 50°C.

6.2 Heat load on the satellite


The worst-case heat load on the satellite is assessed at approximately 45°C by monitoring the temperature
variations around the reaction wheel systems in response to actuation of a reaction wheel at full power. It is
decided to perform the test only for a single reaction wheel, because there is insufficient test hardware to

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 26 / 51

monitor the state of multiple reaction wheels at the same time. Additionally, the three reaction wheels are
sufficiently separated to ignore coupling effects as an estimate of the heat load.

With four thermocouples available, it was decided to distribute them as indicated in Figure 21. The
environment sensor is added to monitor the overall oven temperature, which is used for calibration of the
obtained results. All sensors were calibrated before the test was performed.

Figure 21: Schematic sketch of placement of thermocouple temperature sensors

The test is set up in the following way.

1. Perform characterization test (1) at room temperature

2. Bring environment in the vacuum-oven to a sufficiently stable temperature around 45°C.

3. Pump air from the oven to simulate the vacuum environment of space. This is only done after a
stable temperature is reached, because it is much harder to reach a stable temperature distribution
once a vacuum is in place

4. Start temperature log.

5. Perform characterization test (2)

6. Run reaction wheel at full power for a period of 1 hour and stop temperature log.

7. Perform characterization test (3).

8. Bring environment back to room conditions, i.e. no vacuum and temperature around 22°C.

9. Perform characterization test (4).

With the steps described above the several things can be assessed. These are now introduced one at a time,
where it is described what the conclusions are and how these conclusions are reached.

6.2.1 Heat load of a single reaction wheel.

Using the data acquired between step 4 and 6, we can graph the temperature as a function of time for the
motor level sensor, electronics level sensor and structure level sensor. The goal of the test is to draw
conclusions on the effects of motor operation on the temperature distribution around the reaction wheels.
Therefore we wish to present the measured temperatures with respect to the stable background temperature
of approximately 45°C.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 27 / 51

However it is impossible to maintain a stable background temperature with the vacuum-oven as can be quicly
seen from the graph using the raw data as included in Appendix E. Therefore the results presented in Figure
22 are corrected in two ways:

 Compensation for the approximately linear decrease in temperature over time. From the
reference thermocouple measuring the environment it was concluded that the background
temperature has decreased approximately 6°C throughout the measurement. Similarly, the end
temperatures of the other sensors were lower than the initial temperatures. All measured
temperatures are compensated for this more or less linear temperature decrease.

 Compensation for initial off-set. Although the test set-up was left to settle for over 2 hours, there
was still an off-set in the temperatures measured by the various thermocouples before the test was
started. This can be caused by residual temperature differences but could also be due to less-than-
ideal calibration. Whatever the reason, for easy visual comparison these initial off-sets are
compensated for by a constant. This constant is chosen such that initial temperatures of each sensor
corresponds to the average background temperature during the test; approximately 45°C.

The resulting graphs, shown in Figure 22, indicate that the environment directly around the motor and
electronics increases approximately 2.5°C – 3°C as a result of the actuation of the reaction wheel.

It should be noted that Figure 22 still demonstrate some higher-order effects. These are residual variation in
the temperature unrelated to the reaction wheel. This can be concluded from the fact that the power
consumption of the reaction wheel remained stable throughout the test (approximately 0.09 Watt) and
therefore is not the cause of the steady decrease in the observed temperatures after approximately 40
minutes.

49
Environment
Electronics
48.5
Motor
Structure
48
Temperature [deg Celsius]

47.5

47

46.5

46

45.5

45
0 10 20 30 40 50 60 70 80 90 100
Time [min]

Figure 22: Post-processed results of the heat load from a single reaction wheel. At t = 0 min. three characterization runs
were performed and at t = 20 min. the reaction wheel was turned on at full power until roughly t = 80 min.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 28 / 51

6.3 Impact of vacuum on motor performance.


It is reasoned that vacuum could impact the performance of the reaction wheels in various ways:

1. Influence the bearing lubrication. Although the bearing lubricants are vacuum-rated, their
performance could be influenced by vacuum. This would on its turn impact the viscous damping
coefficient of the reaction wheels.

2. Influence heat dissipation. In vacuum it is more difficult to dissipate heat which could impact the
performance of the motor. The motor is already characterized over the design temperature (up to
50°C) without vacuum in Section 6.1, but vacuum could increase the worst-case steady-state
temperature in the motor beyond this design temperature range. However, as was demonstrated in
Section 6.2.1, the temperature rise is not likely to be more than a few degrees and therefore changes
in heat dissipation are not expected to have a significant impact.

3. Disappearance of air drag. In vacuum there exists no air drag, so if air drag would be responsible for
a big part of the viscous damping coefficient, than vacuum could be beneficial. However, the
observed viscous damping factor is already close to the specified value by the manufacturer and
therefore this is not expected to be a significant factor.

Although no impact is expected from the application of a vacuum, it can be quickly verified by comparison of
the characterization tests (2) and (3). Because any expected impact of vacuum on the performance would
either be independent of temperature or increase the temperature, it makes sense to assess the impact of
vacuum around the maximum design temperature. Figure 23 compares the performance of a motor in
ambient and the same motor in vacuum at a temperature of 43°C.

It can be seen that performance is almost identical, although the motor seems to perform slightly better in
vacuum. In particular at high rotation rates where in vacuum speeds close to 30,000 rpm are reached, which
have not been measured in ambient. Therefore it can be safely concluded that vacuum does not negatively
impact the motor performance.
-4
x 10
1.8 0.8
Measured torque in ambient Measured Pin, max in ambient
1.6 Measured torque in vacuum 0.7 Measured P in vacuum
in, max
Disturbance torque bounds
Measured Pout, max in ambient
1.4
0.6
Measured Pout, max in vacuum
1.2 Disturbance P bounds
0.5 out, max
power [Watt]
Torque [Nm]

1
0.4
0.8
0.3
0.6

0.2
0.4

0.2 0.1

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]

Figure 23: Assessment of impact of vacuum on motor performance at a temperature of 43°C

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 29 / 51

6.4 Degradation testing


The indicated lifetime of the Faulhaber 1202 004 BH motor is 10,000 hours. This equals approximately 400
days of operation. Although no concrete plan of action has been drafted, it is believed that testing for
degradation over time could be interesting. However, this test currently has low priority.

A test plan could simply be to run a motor at constant speed for an extended period of time, while measuring
the power consumption at regular intervals. Chris Verhoeven has proposed to use a microphone to determine
the Fourier Transform of the reaction wheel noise at regular intervals to monitor any alterations in the motor
performance.

It is reasoned that such a test could be interesting to start before launch or at maximum in parallel with
operations in space, because it presents a benchmark for any observed changes in motor performance in
space.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 30 / 51

7 Interference with surrounding hardware


The reaction wheels are mounted on the ADCS PCB, which also incorporates other hardware. Table 9
summarizes the current state of interference testing. The table should be read as the hardware stated in the
header of the row affecting the hardware stated in the header of the column. I.e. the RWS affects the RRS
through vibrations.

Table 9: Summary table of interference between ADCS hardware


RWS RRS MTQ MTM
RWS Issues Conflicting Compatible
RRS
MTQ Compatible EM field
MT

RWS = Reaction Wheel System


RRS = Rotational Rate Sensors
MTQ= Magnetorquers
MT = magnetometers

Pending Compatible Conflicting Issues

7.1 Interference between reaction wheels


Because all three reaction wheels are mounted on the same bracket, the vibrations induced in one wheel will
be very much felt by another wheel. To assess this impact, a characterization test is performed on one wheel
under the influence of a second wheel spinning at maximum speed. The resulting impact on the characteristic
torque and power curves are shown in Figure 24.

-4
x 10 Measured Pin, max (RW2 off)
1.8 0.8
Measured Pin, max (RW2 on)
Measured torque (RW2 off)
1.6 0.7 Measured Pout, max (RW2 off)
Measured torque (RW2 on)
Disturbance torque bounds (RW2 off) Measured Pout, max (RW2 on)
1.4
Disturbance torque bounds (RW2 on) 0.6
Disturbance Pout, max bounds (RW2 off)
1.2 Disturbance Pout, max bounds (RW2 on)
0.5
power [Watt]
Torque [Nm]

1
0.4
0.8
0.3
0.6

0.2
0.4

0.2 0.1

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]

Figure 24: Impact on the characteristic torque and power curves of a reaction wheel due to interference from a second
reaction wheel.

The parameter values that were adjusted to match theoretical graphs with the observed measurements under
the influence of a second reaction wheel turned on are presented in Table 10. No explanation can be given
for this variation except that it has likely to do with the disturbances created by either reaction wheels. The

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 31 / 51

characteristic performance curve for both situations is presented in Figure 25. The power consumption is
estimated to increase from a maximum 160 mW to approximately 180 mW and the maximum useful speed to
decrease from 27,000 to 24,500 due the fact that a second reaction wheel is actuated. Table 10 therefore
also includes the estimated impact on the performance as a percentage of the nominal case.

Table 10: Theoretical interpretation of change in performance of a reaction wheel under the influence of a second
reaction wheel
Parameter RW2 turned off RW2 turned off
Terminal Resistance R [Ohm] 16 17.5
Static friction [Nm] 0.026·10-3 0.045·10-3
Viscous damping factor [Nm/rpm] 0.52·10-9 0
Power consumption 100% 112%
Maximum useful speed 100% 90%

0.25
one wheel actuated
two wheels actuated
0.2

0.15
Power [W]

0.1

0.05

0
0 5,000 10,000 15,000 20,000 25,000 30,000
Rotation speed [rpm]
Figure 25: The characteristic performance curves for the situation where one reaction wheel is actuated and where two
wheels are actuated

7.2 Interference from RWS on RRS


Because it was observed that the reaction wheels induce vibrations in the surrounding structure, it was
hypothesized that this will impact the performance of the Rotation Rate Sensors (RRS). Tests demonstrated
that this indeed is the case and the RRS were omitted from the Delfi-n3Xt ADCS design as a result. See
[SLR0988] for more details about the results on the RRS.

7.3 Interference from RWS on MTM


Tests were performed to measure the influence of the reaction wheels on the magnetometers. [SLR0985]
explains the tests performed and concluded that although the RWS has some impact on the magnetometers,
this impact is acceptable.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 32 / 51

8 Conclusions & recommendations


The vibration test campaign succeeded in subjecting the reaction wheel hardware to targeted sinusoidal and
random vibrations. Shock tests were not executed and should be executed at a later stage. The test can be
considered as very representative of that during a real launch because the full RWS hardware was tested
(including magnetorquers) while the whole system was integrated in an actual 1U cubesat which on its turn
was inserted into the ISIPOD. This is relevant, since the surrounding structure can cause amplification of
vibrations due to resonances. The RWS hardware did not sustain any observable damage and showed the
same performance characteristics after the test campaign as they did before.

Vacuum has no impact on the performance of the reaction wheels, which is to be expected with vacuum-
rated motors. Temperature does have an effect on the performance of the reaction wheels and it was
concluded that the reaction wheels optimal design temperature range is roughly 10°C to 30°C. Outside this
temperature range the power consumption increases and below 0°C this increase is unacceptable. The
allowed temperature range is therefore 0°C - 45°C with an optmimum range of 10°C to 30°C. Outside this
optimum range an increase in power consumption of 20% is expected. The maximum useful speed inside the
optimum temperature range is 26,000 rpm. Outside this range (both above 0°C it is 24,500 rpm).

Interference was also tested, which resulted in the removal of rotational rate sensors from the design and
brought to light that one reaction wheel will interfere with the other reaction wheels. This will increase the
power consumption significantly by roughly 12% and decrease the maximum useful speed by roughly 10%.
Magnetometers also experience some increase in noise due to actuation of the reaction wheels, but this is
within margins.

Because the reaction wheels have such big influence on their own performance, this is something worth
investigating in more detail. Additional the degradation test has not been started, which could give valuable
information about the potential degradation of performance over time and is therefore worth considering.

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 33 / 51

Appendix A. Raw data for consistency tests

2011‐10‐10_B#2_NW_testboard3.3V_torque
Date 10‐10‐2011
Time (at beginning of the  5:17 PM
Measurement time (s) 8
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.015 0 0 0.015 0 0 0.015
0.255 0 0.015 0.254 0 0.015 0.254 0 0.015
0.51 0 0.015 0.508 0 0.015 0.511 0 0.015
0.765 0 0.015 0.762 0 0.015 0.766 0 0.015
1.019 0 0.015 1.016 0 0.015 1.02 0 0.015
1.273 0 0.015 1.268 0 0.015 1.277 0 0.015
1.527 0 0.015 1.522 0 0.015 1.531 0 0.015
1.781 0 0.015 1.778 0 0.015 1.783 0 0.015
2.035 0 0.015 2.032 0 0.015 2.039 0 0.015
2.493 1736.111 0.396 2.489 1794.374 0.395 2.497 1720.184 0.396
3.54 28874.594 0.139 3.528 28871.885 0.134 3.436 28846.154 0.137
4.749 28848.796 0.13 4.818 28871.885 0.126 4.652 28874.594 0.131
5.949 28874.594 0.13 6.018 31055.9 0.128 5.865 28846.157 0.13
7.149 28846.157 0.133 7.216 31250.002 0.131 7.067 31283.383 0.131
8.345 28846.157 0.133 8.417 28820.407 0.135 8.267 28846.157 0.131

2011‐10‐12_B#3_NW_testboard3.3V_torque
Date 12‐10‐2011
Time (at beginning of the  8:57 AM
Measurement time (s) 10
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.015 0 0 0.016 0 0 0.015
0.254 0 0.015 0.256 0 0.016 0.252 0 0.016
0.506 0 0.015 0.51 0 0.015 0.505 0 0.016
0.762 0 0.015 0.764 0 0.015 0.759 0 0.015
1.017 0 0.015 1.018 0 0.015 1.013 0 0.015
1.274 0 0.015 1.27 0 0.015 1.267 0 0.015
1.532 0 0.015 1.523 0 0.016 1.521 0 0.015
1.787 0 0.015 1.777 0 0.016 1.775 0 0.015
2.04 0 0.015 2.032 0 0.016 2.03 0 0.016
2.497 14416.653 0.387 2.491 1689.189 0.385 2.489 1674.007 0.389
3.643 26809.069 0.153 3.435 28873.24 0.154 3.434 28819.12 0.151
4.805 26762.402 0.149 4.601 26785.714 0.149 4.6 26785.714 0.146
5.957 26809.069 0.152 5.749 26809.069 0.146 5.75 26785.714 0.145
7.106 26809.069 0.148 6.9 26810.268 0.144 6.899 26739.13 0.139
8.257 26785.714 0.145 8.051 28846.157 0.147 8.05 28846.154 0.142
9.409 26762.402 0.15 9.2 28846.154 0.142 9.201 28819.12 0.141
10.558 28819.12 0.142 10.35 26809.069 0.147 10.35 28846.154 0.145

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 34 / 51

2011‐10‐12_B#4_NW_testboard3.3V_torque
Date 12‐10‐2011
Time (at beginning of the  9:33 AM
Measurement time (s) 10
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.015 0 0 0.015 0 0 0.015
0.255 0 0.015 0.254 0 0.015 0.254 0 0.015
0.509 0 0.015 0.507 0 0.015 0.511 0 0.015
0.764 0 0.015 0.762 0 0.015 0.765 0 0.015
1.017 0 0.015 1.017 0 0.015 1.02 0 0.015
1.271 0 0.015 1.273 0 0.015 1.273 0 0.015
1.525 0 0.015 1.527 0 0.015 1.527 0 0.015
1.78 0 0.015 1.781 0 0.015 1.782 0 0.015
2.032 0 0.015 2.035 0 0.015 2.039 0 0.015
2.493 15626.223 0.396 2.492 1712.434 0.401 2.496 1720.184 0.395
3.763 28846.157 0.138 3.413 28874.594 0.145 3.435 28846.154 0.145
4.916 28871.882 0.139 4.582 28846.154 0.143 4.613 28873.24 0.131
6.065 28871.882 0.133 5.732 28820.407 0.137 5.764 31277.454 0.142
7.213 28817.831 0.127 6.882 31250.002 0.132 6.914 31250.002 0.13
8.366 28846.154 0.128 8.033 26807.899 0.13 8.067 28873.24 0.129
9.516 30193.392 0.129 9.184 28871.882 0.132 9.215 28846.154 0.128
10.665 28792.134 0.129 10.334 31216.692 0.125 10.368 28846.157 0.125

2011‐10‐12_B#5_NW_testboard3.3V_torque
Date 12‐10‐2011
Time (at beginning of the  10:15 AM
Measurement time (s) 10
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.015 0 0 0.016 0 0 0.015
0.261 0 0.015 0.253 0 0.015 0.252 0 0.015
0.515 0 0.015 0.505 0 0.015 0.508 0 0.015
0.769 0 0.015 0.758 0 0.016 0.762 0 0.016
1.023 0 0.015 1.012 0 0.015 1.016 0 0.015
1.277 0 0.015 1.269 0 0.016 1.27 0 0.015
1.531 0 0.015 1.524 0 0.015 1.522 0 0.015
1.782 0 0.015 1.781 0 0.015 1.776 0 0.015
2.035 0 0.015 2.036 0 0.015 2.029 0 0.015
2.494 18750 0.384 2.497 1728.331 0.383 2.488 5209.308 0.385
3.801 28820.471 0.116 3.364 28846.154 0.13 3.479 26761.238 0.126
4.952 28820.471 0.115 4.546 28874.531 0.119 4.663 31216.692 0.117
6.102 31219.86 0.116 5.698 28900.377 0.116 5.814 28792.134 0.117
7.303 30911.902 0.12 6.848 28846.157 0.12 6.964 28843.513 0.123
8.453 28873.24 0.12 7.998 28819.12 0.123 8.116 28874.594 0.12
9.604 28846.157 0.111 9.149 31216.692 0.12 9.263 28820.407 0.116
10.755 28846.157 0.119 10.298 28871.885 0.121 10.413 28874.594 0.114

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 35 / 51

2011‐10‐12_B#6_NW_testboard3.3V_torque
Date 12‐10‐2011
Time (at beginning of the  10:42 AM
Measurement time (s) 10
Voltage (V) 3.3
Rotational direction RW1 CW
Rotational direction RW2 CW
Rotational direction RW3 CW

Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW) Time (s) Rotation speed (rpm) Power (mW)


0 0 0.016 0 0 0.015 0 0 0.015
0.253 0 0.015 0.254 0 0.015 0.254 0 0.015
0.507 0 0.015 0.508 0 0.015 0.509 0 0.015
0.761 0 0.016 0.765 0 0.015 0.761 0 0.015
1.013 0 0.015 1.02 0 0.015 1.015 0 0.015
1.266 0 0.015 1.274 0 0.015 1.269 0 0.015
1.523 0 0.015 1.528 0 0.015 1.526 0 0.015
1.778 0 0.015 1.782 0 0.015 1.78 0 0.015
2.032 0 0.015 2.037 0 0.015 2.035 0 0.015
2.492 1736.126 0.405 2.496 1752.337 0.401 2.497 17868.038 0.402
3.414 25021.36 0.184 3.431 26763.571 0.165 3.738 26761.238 0.182
4.583 25001.986 0.164 4.616 26785.714 0.152 4.886 26785.717 0.147
5.733 28843.513 0.16 5.765 26832.46 0.147 6.033 28871.885 0.138
6.882 26807.899 0.158 6.917 28846.157 0.139 7.183 28817.77 0.137
8.036 26832.46 0.146 8.069 28874.594 0.138 8.336 28820.471 0.136
9.184 26785.717 0.134 9.218 28848.796 0.139 9.485 28846.157 0.132
10.334 28820.407 0.141 10.366 28846.154 0.131 10.636 28820.471 0.138

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 36 / 51

Appendix B. Characteristic curves over design temperature


range.
This appendix demonstrates the characteristic curves as a function of temperature over the full design
temperature range. The torque bounds that were derived in the Reaction Wheel Design Document [SLR0284]
are also included in the curves.

Power curve x 10
-4
Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]
1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 26: Experimentally matched theoretical characteristic power- and torque curve for temperature -20°C

Power curve x 10
-4
Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]

1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 27: Experimentally matched theoretical characteristic power- and torque curve for temperature -10°C

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 37 / 51

Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]
1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 28: Experimentally matched theoretical characteristic power- and torque curve for temperature 0°C

Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]

1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 29: Experimentally matched theoretical characteristic power- and torque curve for temperature 10°C
Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]

1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 30: Experimentally matched theoretical characteristic power- and torque curve for temperature 20°C

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 38 / 51

Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]
1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 31: Experimentally matched theoretical characteristic power- and torque curve for temperature 30°C

Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]

1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 32: Experimentally matched theoretical characteristic power- and torque curve for temperature 40°C

Power curve x 10
-4 Torque curve
0.8

1.5
0.6
power [Watt]

Torque [Nm]

1
0.4

0.2 0.5

0 0
0 10,000 20,000 30,000 0 10,000 20,000 30,000
rotation speed [rpm] rotation speed [rpm]
Figure 33: Experimentally matched theoretical characteristic power- and torque curve for temperature 50°C

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 39 / 51

Appendix C. Matlab code to generate .arb-files for


execution of sinesweep and random vibration files in
SigLab

function sinesweepGEN(doplot)

% function sinesweepGEN(doplot)
% generates example sinesweepGEN files for excitation on the Brüel & Kjær
% type 4809 shaker. It can only be executed on the laptop at 3ME as SigLab
% is required for its execution. The file will generate two arb-file which
% can be used in SigLab. The files will be placed in the folder C:\Delfi.
%
% The 'chirp_Delfi.arb' function will generate a file for use in a
% sine-sweep test. It might be required to run this .m-file twice before a
% proper file is created. You can play with the variable 'L' to make the
% sine sweep longer/shorter with time, although it was found that it is
% currently tweaked to the longest time (approximately 1 min.). You can
% play with the lengths of the segments and the respective gains to
% construct your own desired sine spectrum.
%
% The 'Delfi_random.arb' creates an .arb file for use in random tests.
% Also in this function, it is possible to play with the variables to
% construct the random vibration spectrum of your liking.
%
% Add your own functions to the list and place the code to make it below.
% creation with a different file name.
%
%
% Teun Hoevenaars, 2011. Delfi Reaction Wheel Engineer
% based on a file by Dick Benson DSP Technology

%include % just execute it (no preprocessor)


vfg_h; % header file with arb_key definition,
vhw_h; % some hardware realities
%end_include

drive = 'c:';
ppath='\Delfi'; % path to store arb files
LF = setstr(10); % LF is linefeed character
nfunc = 2; % number of valid functions currently defined
key = arb_key; % arb file key
Arb_Clk = TB51200c; % must select timebase frequency in Hz
% 50-21 users can select TB50_21c if
% frequencies beyond 20 kHz are needed
dv = INTERP_TBLc; % Available Interpolations (see vhw_h)
% same for both 20-XX and 50-21 systems

% Signal Max Bandwidth = Arb_Clk/(OVS_FACc*dv(Arb_Interp_Index))

for k=1:nfunc

if k==1
file_name = 'chirp_Delfi.arb';
Arb_Interp_Index = 1;

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 40 / 51

L = 2^22; % number of time samples desired


%(rounded to a multiple of 100)
Fstart = 10; Fstop = 100; % start/stop frequencies
L2 = L * Fstop / (Fstop - Fstart);
w = Fstop * pi / (Arb_Clk * L2);
k = L2-L:L2-1;

%Below the segment array lengts are defined.


k_size = size(k);
Sine_size = k_size(1,2) - 4; % manually reduced to a multiple of 100

Sine_seg1 = 2*Sine_size/10 + 4; %to compensate for above reduction


Sine_seg2 = Sine_size/20;
Sine_seg3 = 8*Sine_size/100;
Sine_seg4 = 2*Sine_size/100;
Sine_seg5 = 65*Sine_size/100;

%Here the gain are defined to form the sine sweep profile
gain_mod1 = [1.0 : -0.3./Sine_seg1 : 0.7 + 0.3./Sine_seg1];
gain_mod2 = [0.7 : 0.3./Sine_seg2 : 1.0 - 0.3./Sine_seg2];
gain_mod3 = [1.0 : -0.4./Sine_seg3 : 0.6 + 0.4./Sine_seg3];
gain_mod4 = [0.6 : -0.5./Sine_seg4 : 0.1 + 0.5./Sine_seg4];
gain_mod5 = [0.1 : -0.08./Sine_seg5 : 0.02 + 0.08./Sine_seg5];
Sine_gain = [gain_mod1, gain_mod2, gain_mod3, gain_mod4, gain_mod5];

Arb_Data = Sine_gain.*sin(w * k.^2);


Arb_Text = ['Chirp:' LF 'from 10 to 100 Hz' LF 'with PSLV PSD'];
%figure
%plot(Arb_Data)

elseif k==2
file_name = 'Delfi_random.arb';
Arb_Interp_Index = 1; % max bandwidth
L = 2^20; % number of time samples
desired, must be power of 2 for ifft
Lh = L/2;
Eff = 25/32; % useful portion of Nyquist
rate
Lm = Lh*Eff; % Lm is then equal to the
bandwidth chosen in the VFG (this is designed for BW=2,000Hz).
yfrq = [exp(2j*pi*rand(1,Lm)),zeros(1,Lh-Lm)]; % freq domain data, no DC term
% unit magnitude, random phase
%Below the segment array lengts are defined.
seg1 = 2*Lm/50;
seg2 = 5*Lm/50;
seg3 = 15*Lm/50;
seg4 = 3*Lm/50;
seg5 = 25*Lm/50;
seg6 = Lh-Lm;

%Below the Power Spectral Density profile is defined


yfreq_mod1 = [0.5 : -0.3./seg1 : 0.2 + 0.3./seg1];
yfreq_mod2 = [0.2 : 0.5./seg2 : 0.7 - 0.5./seg2];
yfreq_mod3 = [0.7 : 0.3./seg3 : 1.0 - 0.3./seg3];
yfreq_mod4 = ones(1,seg4);
yfreq_mod5 = [1.0 : -1./seg5 : 0 + 1./seg5];

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 41 / 51

yfreq_mod6 = zeros(1,seg6);

yfrq = yfrq.*[yfreq_mod1, yfreq_mod2, yfreq_mod3, yfreq_mod4, ...


yfreq_mod5, yfreq_mod6]; % shape for PSLV noise
y = ifft([0,yfrq,conj(yfrq(Lh-1:-1:1))]); % append DC term & reflected
yfrq
y = real(y); % should already be real
Arb_Data = y/max(abs(y)); % normalize to 1
Arb_Text = 'Delfi sequence: combined PSLV/DNEPR noise';

end;

if length(Arb_Text)>=ARB_TEXTMAXc
disp('Too many characters in Arb_Text');
end;

eval(['save ',drive,ppath,'\',file_name,...
' key Arb_Clk Arb_Interp_Index Arb_Data Arb_Text']);
if nargin
if doplot==0 | doplot==k
figure('name',file_name,'menu','none',...
'position',[k*30-25,(nfunc-k)*30+5,400,300]);
plot(Arb_Data);
end;
end;
end; % end for loop
% end Arbgen

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 42 / 51

Appendix D. Vibration levels during test campaign at NLR

List of test runs

ISIS-2011-02 FM-02, FM-07, TP-03 and TP-09


Test Run Name Test Test: Test direction: Date/time ready:
31.0 TS-301 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Y-direction 25 okt 2011 16:15
32.0 TS-302 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Sine AT Y-direction 25 okt 2011 16:20
33.0 TS-303 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Y-direction 25 okt 2011 16:25
34.0 TS-304 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Random AT Y-direction 25 okt 2011 16:32
35.0 TS-305 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Y-direction 25 okt 2011 16:43
36.0 TS-306 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Z-direction 26 okt 2011 8:56
37.0 TS-307 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Sine AT Z-direction 26 okt 2011 9:09
38.0 TS-308 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Z-direction 26 okt 2011 9:19
39.0 TS-309 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Random AT Z-direction 26 okt 2011 9:25
40.0 TS-310 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey Z-direction 26 okt 2011 9:32
41.0 TS-311 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey X-direction 26 okt 2011 10:05
42.0 TS-312 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Sine AT X-direction 26 okt 2011 10:10
43.0 TS-313 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey X-direction 26 okt 2011 10:17
44.0 TS-314 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Random AT X-direction 26 okt 2011 10:34
45.0 TS-315 1* 1U + 1 * 2U + 2 * 2U ISIPOD AT Resonance Survey X-direction 26 okt 2011 10:42

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 43 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-304 Run Nr: Random_1
Date / Time : Tue Oct 25 2011 16:31:00 Function class : PSD
Mode : Random Vibration Control Operator : Grijpma R.A.
TestlabVersion 11B
Test level : 0.0 dB Control Strategy : Average Reference 9.99521 g
Degrees of Freedom : 90 No. Ctrl. Chan. : 2 LowAbort 7.07607 g
Window Type : Hanning No. Alarm lines : 24
UpAbort 14.1186 g
Time at level : 0:00:43.96 No. Abort Lines : 0
Spectrum format : PSD Sample Frequency : 6400 Hz Expander_B 10.9692 g

1.00
43

g2/Hz
Log

F PSD Reference
F PSD Low Abort
F PSD UpAbort
F PSD Expander_B:-Y

1.00e-3
20.00 Hz 2000.00

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Random vibration load spectrum in Y-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 44 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-305 Run Nr : Sine_2
Date / Time : Tue Oct 25 2011 16:43:38 Function class : Spectrum
Mode : Sine Vibration Control Operator : Grijpma R.A. Testlab Version 11B
Sweep mode : Log Control Strategy : Average Reference
Sweep Direction : Down No. Ctrl. Chan. : 2
TP-03_Y 1.02542 g
Sweep Rate : 2 Oct/min Number of lines : 375
TP-03_Y 1.26604 g
Sweeps done : 1 Weighting : Linear
Compression factor : 4 Frequency Res. : Variable TP-03_Y 1.12615 g

10.00
44

Log
g

F Spectrum Reference TS-305 (Y) Post RS


F Harmonic Spectrum TP-03_Y:+Y TS-301 (Y) Pre RS
F Harmonic Spectrum TP-03_Y:+Y TS-303 (Y) RS
F Harmonic Spectrum TP-03_Y:+Y TS-305 (Y) Post RS
F Harmonic Spectrum TP-03_Y:+Y TS-305 (Y) Post RS

1.00e-3 711.00
5.00 Hz 2000.00
2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Comparison between resonance survey of before- and after the vibration campaign in Y-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 45 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-307 Run Nr : Sine_1
Date / Time : Wed Oct 26 2011 09:08:54 Function class : Spectrum
Mode : Sine Vibration Control Operator : Grijpma R.A. Testlab Version 11B
Sweep mode : Log Control Strategy : Average Reference
Sweep Direction : Up No. Ctrl. Chan. : 2
LowAbort Property not found
Sweep Rate : 4 Oct/min Number of lines : 830 i t
UpAbort Property not found
Sweeps done : 1 Weighting : Linear i t
Compression factor : 4 Frequency Res. : Variable Sliptable_B 3.06859 g

10.00
45

Log
g

F Spectrum Reference TS-307 (Z) Sine AT


F Spectrum Low Abort TS-307 (Z) Sine AT
F Spectrum UpAbort TS-307 (Z) Sine AT
F Harmonic Spectrum Sliptable_B:+Z TS-307 (Z) Sine AT

0.10
5.00 Hz 100.00
2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Sine sweep in Z-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 46 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-309 Run Nr: Random_1
Date / Time : Wed Oct 26 2011 09:23:34 Function class : PSD
Mode : Random Vibration Control Operator : Grijpma R.A.
TestlabVersion 11B
Test level : 0.0 dB Control Strategy : Average Reference 9.99521 g
Degrees of Freedom : 90 No. Ctrl. Chan. : 2 LowAbort 7.07607 g
Window Type : Hanning No. Alarm lines : 19
UpAbort 14.1186 g
Time at level : 0:00:43.94 No. Abort Lines : 0
Spectrum format : PSD Sample Frequency : 6400 Hz Sliptable_B 9.05176 g

1.00
46

g2/Hz
Log

F PSD Reference
F PSD Low Abort
F PSD UpAbort
F PSD Sliptable_B:+Z

100e-6
20.00 Hz 2000.00

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Random vibration load spectrum in Y-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 47 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-310 Run Nr : Sine_1
Date / Time : Wed Oct 26 2011 09:32:08 Function class : Spectrum
Mode : Sine Vibration Control Operator : Grijpma R.A. Testlab Version 11B
Sweep mode : Log Control Strategy : Average Reference
Sweep Direction : Up No. Ctrl. Chan. : 2
Sliptable_B 0.541227 g
Sweep Rate : 2 Oct/min Number of lines : 2089
Sliptable_B 0.540261 g
Sweeps done : 1 Weighting : Linear
Compression factor : 4 Frequency Res. : Variable Sliptable_B 0.552069 g

1.00
47

Log
g

F Spectrum Reference TS-310 (Z) Post RS


F Harmonic Spectrum Sliptable_B:+Z TS-306 (Z) Pre RS
F Harmonic Spectrum Sliptable_B:+Z TS-308 (Z) RS
F Harmonic Spectrum Sliptable_B:+Z TS-310 (Z) Post RS

0.01
5.00 Hz 2000.00
2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Comparison between resonance survey of before- and after the vibration campaign in Z-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 48 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-312 Run Nr : Sine_1
Date / Time : Wed Oct 26 2011 10:10:24 Function class : Spectrum
Mode : Sine Vibration Control Operator : Grijpma R.A. Testlab Version 11B
Sweep mode : Log Control Strategy : Average Reference
Sweep Direction : Up No. Ctrl. Chan. : 2
LowAbort Property not found
Sweep Rate : 4 Oct/min Number of lines : 830 i t
UpAbort Property not found
Sweeps done : 1 Weighting : Linear i t
Compression factor : 4 Frequency Res. : Variable Sliptable_B 3.07171 g

10.00
48

Log
g

F Spectrum Reference TS-312 (X) Sine AT


F Spectrum Low Abort TS-312 (X) Sine AT
F Spectrum UpAbort TS-312 (X) Sine AT
F Harmonic Spectrum Sliptable_B:+X TS-312 (X) Sine AT

0.10
5.00 Hz 100.00
2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Sine sweep in X-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 49 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-314 Run Nr: Random_1
Date / Time : Wed Oct 26 2011 10:32:47 Function class : PSD
Mode : Random Vibration Control Operator : Grijpma R.A.
TestlabVersion 11B
Test level : 0.0 dB Control Strategy : Average Reference 9.99521 g
Degrees of Freedom : 90 No. Ctrl. Chan. : 2 LowAbort 7.07607 g
Window Type : Hanning No. Alarm lines : 6
UpAbort 14.1186 g
Time at level : 0:00:43.85 No. Abort Lines : 0
Spectrum format : PSD Sample Frequency : 6400 Hz Sliptable_B 9.01661 g

1.00
49

g2/Hz
Log

100e-6
20.00 Hz 2000.00

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Random vibration load spectrum in X-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 50 / 51
.

Plotformat Testlab 11B


.
NLR Vibration and Shock Test laboratory
Project : ISIS-2011-02 Pod Units Test Identifier : TS-315 Run Nr : Sine_1
Date / Time : Wed Oct 26 2011 10:42:15 Function class : Spectrum
Mode : Sine Vibration Control Operator : Grijpma R.A. Testlab Version 11B
Sweep mode : Log Control Strategy : Average Reference
Sweep Direction : Up No. Ctrl. Chan. : 2
Sliptable_B 0.548974 g
Sweep Rate : 2 Oct/min Number of lines : 2089
Sliptable_B 0.537285 g
Sweeps done : 1 Weighting : Linear
Compression factor : 4 Frequency Res. : Variable Sliptable_B 0.554058 g

1.00
50

Log
g

F Spectrum Reference TS-315 (X) Post RS


F Harmonic Spectrum Sliptable_B:+X TS-311 (X) Pre RS
F Harmonic Spectrum Sliptable_B:+X TS-313 (X) RS
F Harmonic Spectrum Sliptable_B:+X TS-315 (X) Post RS

0.01
5.00 Hz 2000.00
2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
Comparison between resonance survey of before- and after the vibration campaign in X-direction
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Document DNX-TUD-TR-0766
Date 21-02-2012
Technical Note Issue 3.3
Page 51 / 51

Appendix E. Graph of the development of temperature over


time before post-processing
The figure below presents the raw temperature measurements as obtained during the heat load test
described in Section 6.2.1. It can be seen that there is both an initial off-set and a steady decrease of
temperature over time.

50
Environment
49 Electronics
Temperature [deg Celsius]

48 Motor
Structure
47

46

45

44

43

42
0 10 20 30 40 50 60 70 80 90 100
Time [min]

2012. All rights reserved. Disclosure to third parties of this document or any part thereof, or the DNX-TUD-TR-0766 [3.3] ADCS -
use of any information contained therein for purposes other than provided for by this document, is Reaction Wheels Test
not permitted, except prior and express written permission of Delft University of Technology. Document.docx
Appendix G
Analysis of Controller Robustness

It is anticipated that the values for the terminal resistance R, the static fric-
tion C0 and the back-EMF constant kE of the motor may change over the
mission lifetime. Especially temperature has a big influence on their value
as was concluded in Appendix F. To assess the robustness of the controller,
the determined values for R, C0 and kE for different temperatures are used
as inputs in the system model and the system response is simulated. Ta-
ble G.1 introduces the determined parameters corresponding to the various
temperatures as a representation for the maximum variation over the mis-
sion lifetime. Figure G.1 demonstrates the system responses for a campaign
including a zero-crossing. It can be seen that the different sets of param-
eters result in different behavior at the zero-crossing but that outside the
zero-crossing region the performance of the controller is very similar. It can
therefore be concluded that the robustness of the controller is adequate.

273
Analysis of Controller Robustness

Table G.1: The variations of motor parameters as a function of temperature.


Temperature [ C] RrΩs C0 [mNm] kE [mV/rpm]

-20 16 0.06 0.078

0 16.5 0.035 0.084

20 17 0.026 0.094

50 18.2 0.03 0.094

System response
2000

Ideal response
Simulated system response at T=−20°C
1500
Simulated system response at T=0°C
Rotational speed [rpm]

Simulated system response at T=20°C


1000 Simulated system response at T=50°C

500

−500
0 5 10 15 20 25 30 35 40 45 50
Time [s]

Figure G.1: The simulated system response for different sets of parameters
as introduced in Table G.1

274 MSC THESIS A.G.L. HOEVENAARS


Appendix H
Graphs on Relative Impact of
Disturbance Sources

This appendix demonstrates graphically the impact of changing the value


of either axle or cog , which were introduced in Figure 5.2 in Section 5.1.

H.1 Impact of Axle Deflection Angle


Figures H.1, H.2 and H.3 demonstrate the impact of changes in the axle
deflection angle axle on respectively the amplitude of the resulting distur-
bance accelerations, -forces and -moments. In all cases the frequency of the
disturbances is equal to the rotational speed and cog is kept constant at 0.4
deg.

275
Graphs on Relative Impact of Disturbance Sources

X− and Y− axis Z− axis


60 2
εaxle=0.85 [deg]
50 εaxle=1.70 [deg]
1.5
εaxle=0.425 [deg]
Amplitude [m/s2]

Amplitude [m/s2]
40
εaxle=0 [deg]
30 1

20
0.5
10

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.1: Impact of changes in the axle deflection angle axle on the
amplitude of disturbance accelerations in the motor reference frame.

X− and Y− axis Z− axis


0.012
εaxle=0.85 [deg]
0.3 0.01 εaxle=1.70 [deg]
0.25 εaxle=0.425 [deg]
0.008
Amplitude [N]

Amplitude [N]

ε =0 [deg]
0.2 axle
0.006
0.15
0.004
0.1

0.05 0.002

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.2: Impact of changes in the axle deflection angle axle on the
amplitude of disturbance forces in the motor reference frame.

276 MSC THESIS A.G.L. HOEVENAARS


H.2. Impact of Manufacturing Errors

x 10
−4 X− and Y− axis x 10
−5 Z− axis
7 2
εaxle=0.85 [deg]
6
εaxle=1.70 [deg]
1.5
5 εaxle=0.425 [deg]
Amplitude [Nm]

Amplitude [Nm]
εaxle=0 [deg]
4
1
3

2
0.5
1

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.3: Impact of changes in the axle deflection angle axle on the
amplitude of disturbance torques in the motor reference frame.

H.2 Impact of Manufacturing Errors


cog was introduced in Section 5.1 to represent the shift of the center of
gravity away from the rotor. Figures H.4, H.5 and H.6 demonstrate the
impact of changes in the angle cog on the amplitude of respectively the
resulting disturbance accelerations, -forces and -moments. In all cases the
frequency of the disturbances is equal to the rotational speed and axle is
kept constant at 0.85 deg.
X− and Y− axis Z− axis
120 2
εcog=0.4 [deg] εcog=0.4 [deg]
100 εcog=0.8 [deg] εcog=0.8 [deg]
1.5
εcog=0.2 [deg] εcog=0.2 [deg]
Amplitude [m/s ]

Amplitude [m/s2]
2

80
ε =0 [deg] ε =0 [deg]
cog cog
60 1

40
0.5
20

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.4: Impact of changes in the axle deflection angle cog on the am-
plitude of disturbance accelerations in the motor reference frame.

DELFI-N3XT REACTION WHEEL SYSTEM 277


Graphs on Relative Impact of Disturbance Sources

X− and Y− axis Z− axis


0.01
εcog=0.4 [deg] εcog=0.4 [deg]
0.6
εcog=0.8 [deg] 0.008 εcog=0.8 [deg]
0.5 εcog=0.2 [deg] εcog=0.2 [deg]
Amplitude [N]

Amplitude [N]
εcog=0 [deg] 0.006 εcog=0 [deg]
0.4

0.3 0.004
0.2
0.002
0.1

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.5: Impact of changes in the axle deflection angle cog on the am-
plitude of disturbance forces in the motor reference frame.

x 10
−3 X− and Y− axis x 10
−5 Z− axis
1.4 2
εcog=0.4 [deg] εcog=0.4 [deg]
1.2
εcog=0.8 [deg] εcog=0.8 [deg]
1.5
1 εcog=0.2 [deg] εcog=0.2 [deg]
Amplitude [Nm]

Amplitude [Nm]

ε =0 [deg] ε =0 [deg]
0.8 cog cog
1
0.6

0.4
0.5
0.2

0 0
0 10,000 20,000 0 10,000 20,000
Rotational speed [rpm] Rotational speed [rpm]

Figure H.6: Impact of changes in the axle deflection angle cog on the am-
plitude of disturbance torques in the motor reference frame.

278 MSC THESIS A.G.L. HOEVENAARS

You might also like