70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.

Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

IAC-19-B4.9.GTS.5

SIMULATION AND SELECTION OF DETUMBLING ALGORITHMS FOR A 3U CUBESAT

Vishnu P Katkooria*, Jivat Neet Kaurb , Tushar Goyalc

,
a
Department of Chemical Engineering, Birla Institute of Technology and Science (BITS), India,
[email protected]
b
Department of Computer Science and Engineering, , Birla Institute of Technology and Science (BITS), India,
[email protected]
c
Department of Electrical and Electronics Engineering, , Birla Institute of Technology and Science (BITS), India,
[email protected]
* Corresponding Author

Abstract
As a satellite is deployed from the launch vehicle, it is subjected to high angular rates which need to be dampened
in order for the satellite to perform its functions as expected. Simple and robust algorithms, such as BDot, are
generally used to provide the required control torque for detumbling the satellite. This paper elucidates the design
process for the detumbling algorithm to be implemented on a nanosatellite currently being developed by Team
Anant, the Student Satellite Team of BITS Pilani. The process commenced with the selection of hardware to be used
on-board the satellite. Magnetometers and Gyroscopes were finalized to be used as sensors. Various commercially
available sensor models were then compared based on power and operating conditions. For actuation, a
magnetorquer system was designed specifically to the requirements of the team. The system comprised of two
magnetorquer rods and a magnetorquer coil aligned in orthogonal directions. The sensors and actuators were then
accurately modelled in MATLAB for further testing. The modelling involved some interesting challenges due to the
magnetic moment retained by the ferromagnetic core. These challenges, and the ways to overcome them have been
also been briefly discussed in the paper. After finalizing the hardware, the team proceeded with implementing
various popular control algorithms for detumbling the satellite. The algorithms were first theoretically analysed, and
then modelled on MATLAB. The simulations took the space environment around the satellite into consideration for
higher accuracy. The algorithms were tested for different initial conditions, using different time-steps and under
different power constraints. The algorithms considered and the conclusions derived from these simulations have also
been discussed elaborately in this paper. The paper concluded by presenting the finalized detumbling algorithm(s) to
be used by Team Anant, and the various conditions devised to ensure efficient use of electrical power. The paper also
presents viable alternatives to the finalized algorithm(s), using other hardware components. These alternatives and
conditions have also been documented in the paper for a better understanding.
Keywords: ADCS, Satellite Dynamics, Detumbling, Magnetic Actuation

Acronyms/Abbreviations
1. Introduction
ADCS - Attitude Determination and Control System The Attitude Determination and Control Subsystem,
(ADCS) is responsible for ensuring the satellite has the
ECI - Earth Centered Inertial correct attitude in space by providing accurate control to
correct any deviation from expected output. Detumbling
ECEF - Earth Centered Earth Fixed is essential to stabilize and dampen the high angular rate
along all three axes of the satellite after deployment
IGRF - International Geomagnetic Reference Frame from the Poly-Picosatellite Orbital Deployer (P-POD).

IMU - Inertial Measurement Unit The paper aims to analyse the different detumbling
algorithms and select the most efficient control
LEO - Low Earth Orbit algorithm for a 3U CubeSat in a sun-synchronous, Low
Earth Orbit (LEO). The first section describes the
P-POD - Poly-Picosatellite Orbital Deployer detumbling algorithms studied and selected for
implementation. The next section discusses the sensors
and actuators employed in the satellite to achieve
efficient detumbling. Subsequently, a discussion on the

IAC-19-B4.9.GTS.5 Page 1 of 6
 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.
Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

simulation methodology and the various modules

implemented to test the algorithms is made. Finally, a Bt  Bt 1
comparison of the different algorithms implemented in Bt dot  (3)
terms of power consumption, detumbling time and tsamp
magnetic moment is made.
Here, t samp represents the sampling time. It is
2. Detumbling Algorithms
Two detumbling algorithms are reviewed within noted that this control law is an approximation to the
this paper. The first control law requires angular control law described by algorithm 1 [1]. In addition,
velocity feedback in addition to the ambient magnetic this control law can only detumble the satellite to the
field vector. The second control law, popularly known order of magnitude of the orbital rate (10-3 rad/s). Since
as the Bdot control law operates exclusively with the feedback term is the rate of change in magnetic field,
angular velocity feedback. the controller will try to stabilize the satellite about the
local magnetic field. This is the cause for the slow
2.1 Algorithm 1 – (ω xB) rotation of the satellite after detumbling is completed.
In this control law, a magnetic moment is This is unlike equation (1), which, in theory, can
generated perpendicular to the angular velocity and the completely stabilize the satellite.
local magnetic field vector. Angular velocity can be
split up into two components - a component which is 2.2 Selection of Gain
along the direction of the local magnetic field and a The gain expression, proposed by Avanzini et. al.
component which is normal to it. The magnetic moment [1] is based on analysing the closed loop dynamics of
for the control law is calculated as follows [2]. the component of angular velocity perpendicular to the
earth’s magnetic field.
k
m ( xB) (1) 4
B k (1  sin  ) J min (4)
Torb
Here, k is a scalar gain, ω is the angular
velocity of the satellite relative to the earth centered Here, Torb represents the orbital time of the
inertial (ECI) frame measured in the body frame, and B
satellite,  represents the inclination of the satellite
is the local magnetic field measured in the body frame.
This particular selection of magnetic moment ensures with respect to the geomagnetic equatorial plane, and
that the torque produced is antiparallel to the component J min is the minimum principle moment of inertia for the
of angular velocity normal to the magnetic field. The satellite [2].
lack of control along the parallel component of angular
velocity is not an issue due to the spatial variation of 3. Hardware
magnetic field throughout the orbit. The spatial
variation of magnetic field is maximized for near polar 3.1 Sensors
orbits. Hence, detumbling via magnetic actuation is a The sensors used during detumbling are
viable option for sun-synchronous, LEO orbits. magnetometers and an IMU. Magnetometers measure
the local magnetic field, whereas the IMU measures
Feedback for this control law comes from the angular velocity, as well as the acceleration of the
Inertial Measurement Unit (IMU), as well as the satellite.
magnetometer.
3.1.1 Magnetometers
2.2 Algorithm 2 – Bdot Magnetometers are widely used in satellites as
The Bdot control law calculates magnetic they are relatively small, inexpensive and lightweight.
moment using the rate of change of the magnetic field. The magnetometers measure a sum of the local
It utilizes feedback exclusively from the magnetometer. magnetic field that is of interest and also the local fields
The control law takes the following form [2]. produced by the satellite. The readings could be
k dot disturbed due to the presence of magnetorquer coils,
m B (2) ferrous materials on board, and other residual magnetic
B
fields. They have to be calculated and compensated for.
As a consequence of this, magnetometer readings are
The rate of change of magnetic field is only reliable when there is no current passing through
calculated by using a finite difference method. the actuators. Therefore, it must be ensured that there is

IAC-19-B4.9.GTS.5 Page 2 of 6
 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.
Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

no current passing through the magnetorquers while consumed for a constant voltage supply. Therefore, the
sampling readings from the magnetometer. In addition, use of thinner wires is preferred for minimal power
the magnetometers must be kept as far away possible consumption.
form ferromagnetic materials, such as the ferromagnetic
core of rod type magnetorquers. For the final configuration, a coil type
magnetorquer and a pair of rod-type magnetorquers was
The HMC6343 is selected as the magnetometer chosen. Perm alloy was chosen as a soft ferromagnetic
for the satellite. It had the primary benefits of being core for the rod-type torquer. Despite having lower
small in size and having a low power requirement of power consumption and a smaller mass, a rod type
19.8 mW. In addition, it has the added advantage space magnetorquer could not be accommodated for the 10cm
heritage, as it has been flown on multiple CubeSat x 10cm face due to volume constraints. The technical
missions. The operating frequencies of the specifications are given in the table below.
magnetometer are set to either, 1Hz or 10Hz.
Table 1. Magnetorquer Specifications
3.1.2 Inertial Measurement Unit (IMU) Rod Type Coil Type
The inertial measurement unit gives the measure of Mass (g) 16.91 77.7
the external forces and the angular rate of the satellite in Max Power (mW) 72.5 438
body frame. The IMU is a combination of two types of Max Moment (Am2) 0.332 0.324
sensors - the 3-axis gyroscopes that measure the angular Winding Material Awg-39 Awg-30
rate of the satellite and 3-axis accelerometers that
measure the acceleration and hence the external forces
on the satellite.
4. Simulation Methodology
The methodology for simulation of the control
The ADIS16334 inertial sensor by Analog Devices
laws was based on partitioning the code into sections
was chosen as the IMU for the satellite. Similar to the
which can be verified independently. The block diagram
selection methodology for the magnetometer, this
given below can help visualize the processes as well as
sensor was chosen for its small size, low power
the input-output structure.
consumption relative to its competitors, and space
heritage. This sensor consumes a significantly larger
amount of power than the magnetometer

3.2 Actuators
Magnetorquers are chosen as actuators during
detumbling. This is due to their low power consumption,
as compared to alternatives such as reaction wheels.
However, this comes at the expense of having an under-
actuated system. An in-house magnetorquer system was
developed by the team. The design methodology was
based upon constraints of mass, power, and size. The
torquers must be lightweight and compact enough to be
accommodated into the internal architecture of the
satellite. They must be powerful enough to detumble the
Fig. 1. Block Diagram for Control Law Simulation
satellite within 3 orbits. In addition, the components of
the magnetorquer system must operate within safety
4.1 Position Update Module
guidelines prescribed by the manufacturer. The torque
Using predefined keplerian orbital elements,
generated by the magnetorquers is represented as
this module updates the position and velocity of the
follows.
satellite in ECI frame. The eccentric anomaly is
computed iteratively using Newton Raphson’s method
  m B (5) [3]. In addition, orbital perturbations are accounted for
by using the SGP model.
3.2.1 Magnetorquer Design
Although aluminium is a more lightweight 4.2 Magnetic Field Update
material, copper is chosen as the winding material The 12th rendition of the International
because it is commercially available at diameters Geomagnetic Reference Field (IGRF) model is used to
smaller than 1.1 mm. The increase in cross sectional compute the local magnetic field vector. The output is in
area decreases resistance, thereby increasing power the Earth Centered Earth Fixed (ECEF) frame, and

IAC-19-B4.9.GTS.5 Page 3 of 6
 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.
Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

hence, must be converted to the ECI frame with the aid Here, I represents the inertia tensor of the satellite,
of quaternion multiplication. The transformation  con is the control torque calculated using (4) , and
between these frames must take into account the
revolution of Earth about its spin axis as well as the  dis represents the disturbance torques.
revolution about the sun. A gaussian error may be
implemented in the magnetometer data in order to Propagation is carried out by using a fourth order Runge
account for fluctuations in sensor measurement. The Kutta (RK4) numerical integration. The iterative
amplitude of this noise will correspond to the scheme is presented as follows.
specifications provided by the sensor manufacturer.
1
4.3 Controller Module t 1  t  (kw1  2kw2  2kw3  kw4 )t (9)
The inputs to the controller module are the 6
angular velocity and the magnetic field vectors
represented in the body frame. For algorithm-2, only 1
magnetic field feedback is required qt 1  qt  (kq1  2kq 2  2kq 3  kq 4 )t (10)
The component wise magnetic moment is 6
calculated by this module. In addition, the current
requirement for the corresponding moment is tabulated dw
in order to keep track of power consumption. If the
ki  , i  o  ak (i 1) t (11)
dt  (i )
calculated moment is larger than the limits set by the
hardware, then the coil is saturated at the maximum
possible value. Errors in the magnetic moment output dq
are may be incorporated as gaussian noise. The error is kqi  , qi  qo  akq (i 1) t (12)
induced because there might be a disparity between the
dq q ( i ), w ( i )
calculated current and the current supplied by the
system.
0 i 1

4.4 State Space Propagation Module a  0.5 i  2,3 (13)
The state variables that are propagated in the 1 i4
simulation are the angular velocity vector, as well as the 
quaternion which represents rotation between the body
and ECI frames. Although the propagation of angular In this scheme the quaternion is normalized at
velocity is intrinsic to detumbling, the rotation after every calculation of kqi in order to maintain unit
quaternion should also be propagated in order to
simulate the magnetic field vector in the body frame. magnitude. The magnetic field vector is also
This component is critical in computing the necessary recalculated in each step so as to simulate the change in
control torque. The non-linear state equations are ambient magnetic field the perspective of the body
represented below. frame. This results in the torque acting on the body
changing as the magnetic field rotates. This is a
reflection of the fact that the torque produced will
d
 I 1 ( x( I  )   con   dis ) (6) attempt to align the magnetic dipole with the local
dt magnetic field. Therefore, a closer representation of
space environment is achieved.
dq 1 Initially simulations were performed using
 ( )q (7) Euler’s method due to ease of implementation. However,
dt 2 the tabulated ordinary differential equations are highly
sensitive and non-linear, and hence, required a much
 0 3 2 1  lower time step as compared to the time step required in
  0 1 2 
the RK4 method in order to achieve convergence. The
( )   3 (8)
lower time step led to a much higher computation time
 2 1 0 3  despite using a simpler numerical method. Therefore, it
  is recommended that a higher order integration scheme
 1 2 3 0 should be employed for propagating satellite dynamics.

4.5 Disturbance Module

IAC-19-B4.9.GTS.5 Page 4 of 6
 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.
Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

Disturbances are introduced into the simulation all three axes was less than or equal to twice the orbital
in order to verify the robustness of the controller under rate.
the influence of perturbations. Within the space
environment the satellite experiences disturbance Table 2. Orbital Parameters
torques due to aerodynamic drag, gravity gradient, solar Parameter Value
radiation pressure, and residual magnetic moments. Two
approaches were undertaken to simulate disturbances. Inclination (deg) 97.8

3.5.1 First Approach Altitude (km) 607

A calculation of the maximum disturbance
torque for each category of external disturbance forces Eccentricity 7.15*(10^-5)
was tabulated. The magnitude of maximum possible
disturbance torque acting on the satellite will be the sum Initial Angular Velocity (.03 , -.03, .03) T
of the individual magnitudes. Note that an added margin
of 30% was included to this value. Now, a random unit
vector is generated, and thereby, a torque of maximum Both algorithms were tested using the first
possible magnitude is generated along this direction. approach for disturbances over 1000 runs for the same
This disturbance torque is added to the satellite right ascension and argument of perigee. For the second
dynamics. approach for disturbances, the right ascension was
However, although this method introduces a varied over 10 different values varying by 36 degrees
high degree of disturbance in the system, it doesn’t have since it will affect the spatial change in the IGRF model.
an intuitive basis. It is probable for a given trial run, that The detumbling time for algorithm 1 was less than the
a majority of the disturbance torques generated are anti- Bdot algorithm in all runs. The disparity in detumbling
parallel to the direction of angular velocity, and hence, time was approximately equal to half of the orbital rate.
aiding the detumbling progress. A remedy to overcome In addition, the magnetorquers weren’t saturated at any
such cases is to run this simulation multiple times under instant for either case. This indicates that the torquer
a different set of disturbances and find an average of the design is powerful enough for the system?
settling time. When comparing the reduction of angular
velocity component wise, the z-component of angular
3.5.2 Second Approach velocity requires the most time because the magnitude
Since the aerodynamic drag is by far the of magnetic field driving the detumbling process is
maximum source of disturbance force for the satellite lower compared to the x and y components. Component
and acts opposite to the direction of velocity, a wise response to detumbling control is given below for
disturbance force vector is generated antiparallel to the a test run using the Bdot algorithm.
direction of velocity. Its magnitude will be the sum of
the maximum magnitude of disturbance force possible
within each category. This disturbance force is then
transformed under rotation and represented in the body
frame. The torque produced by this force will act
perpendicular to the vector between the centre of
pressure and the centre of mass of the satellite. This
method of adding disturbances may not provide
maximum possible perturbations to the satellite, but it
better represents the actual dynamics of the satellite.

5. Results and Discussion

The detumbling algorithms were tested with the
parameters listed in Table 2. Worst case angular
velocity is reported as 10 degrees per axis [4]. A
margin of 50 percent per axis is taken to ensure
robustness. It is assumed that at the start of the
simulation, the ECI and ECEF frames are aligned.
Orbital parameters are selected for a sun-synchronous,
LEO orbit with an altitude of approximately 600 km.
The simulation terminated when the angular velocity on Fig 2. Component-wise Angular Velocity vs Time

IAC-19-B4.9.GTS.5 Page 5 of 6
 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019.
Copyright ©2019 by the International Astronautical Federation (IAF). All rights reserved.

6. Conclusions
The settling time of the controller for algorithm 1 References
was smaller than the Bdot controller for all nominal
cases. However, the difference in detumbling time of [1] Giulio Avanzini and Fabrizio Giulietti, “Magnetic
the two algorithms was not significant as per the Detumbling of a Rigid Spacecraft”, JOURNAL OF
requirements of the satellite. Bdot has the added GUIDANCE, CONTROL, AND DYNAMICS, Vol. 35,
advantage of consuming less power as it does not use No. 4, July–August 2012
the IMU for angular velocity feedback. The IMU, used
in algorithm 1 consumes a significant amount of power. [2] F. Markley and J. Crassidis, Fundamentals of
The energy expenditure of the IMU outweighs the spacecraft attitude determination and control, New York:
advantage of having a smaller detumbling time. Hence, Springer, 2014.
based on our analysis and simulation, the Bdot control
law was selected as a more efficient detumbling [3] Oliver Montenbruck and Eberhard Gill, Satellite
algorithm for a 3U CubeSat. In order to verify if the Orbits, Springer,2005.
satellite is detumbled, the IMU will be used at sparse
intervels. In cases where the angular velocity exceeds a [4] Mr. Rutwik Narendra Jain, et. al., “Modes of
critical amount ( >100 deg/s), the detumbling mode will Operation for
not be initiated and the satellite will detumble naturally. a 3U CubeSat with Hyperspectral Imaging Payload”,
69th International Astronautical Congress, 2018.
7. Future Work
By using the results from the simulation, the Team [5] Goyal, Tushar & Aggarwal, Kushagra. (2019).
is moving towards hardware in the loop testing (HIL). Simulator for Functional Verification and Validation of
This includes the fabrication and testing of the a Nanosatellite. 1-8. 10.1109/AERO.2019.8741886.
magnetorquers. This will be done with the aid of a
Helmholtz cage [6] in order to monitor and regulate the [6] Goyal, Tushar. (2017). Design and Development of
magnetic field of the environment. Here, we can verify a Three-Axis Controlled Helmholtz cage as an In-House
the amount of residual magnetization present in the Magnetic Field Simulator for CubeSats.
ferromagnetic rod. In addition, the team is working on
the accurate determination of Bdot from magnetometer
data which may require the implementation of a low
pass filter to remove unwanted noise.

Acknowledgements

The authors would like to thank the following:

1. Team Anant, the Student Satellite Team of BITS

Pilani for providing the motivation to carryout these
simulations.

2. Dr. Kaushar Vaidya, Rudrakh Panigarhi, and Jeet

Yadav for their constant support and technical expertise
in perusing research for this paper.

3. The administration of BITS Pilani and Indian Space

Research Organization (ISRO) for giving us an
opportunity to work on building a satellite.

IAC-19-B4.9.GTS.5 Page 6 of 6