Design of an ADCS test stand for nanosatellites

Alberto Gallina∗ , Dawid Knapik∗ , Krzysztof Kołek∗ , Maciej Rosół∗ , Andrzej Tutaj∗ and Paweł Zagórski∗
∗ AGH University of Science and Technology,
al. A. Mickiewicza 30,
30-059 Krakow, Poland
Email: [agallina, knapik, kko, mr, tutaj, pzagor]@agh.edu.pl
2022 26th International Conference on Methods and Models in Automation and Robotics (MMAR) | 978-1-6654-6858-9/22/$31.00 ©2022 IEEE | DOI: 10.1109/MMAR55195.2022.9874310

Abstract—ADCS test stands are used for on-the-ground testing and control solution with the rapid controller prototyping
of satellite altitude determination and control systems. Their methodology. Section V describes the realization of the test
main purpose is to simulate time-vaying conditions similar to stand as well as an experimental setup used to verify its
those found in outer space on a geocentric orbit. The paper
presents the main component of the ADCS test stand – a performance together with obtained results. A discussion is
Helmholtz cage used to simulate the orbital magnetic field. The provided in Section VI.
design and integration of mechanical, electrical, electronic and
software components that led to a highly uniform and precisely II. D ESIGN OF THE ADCS TEST STAND
controlled magnetic field is described here. The validation pro- Based on the character of nanosatellites, i.e. small size, often
cess, including 3D scanning of the magnetic field, is presented.
Remarks on the employed technology, limitations of the proposed defined by the CubeSat standard (10 cm × 10 cm × 10 cm) and
solution, some design recommendations, and plans for future their typical working environment, i.e. Low Earth Orbit (LEO),
work are provided. requirements for magnetic field simulator can be specified. The
magnetic field uniformity region was designed to be 20 cm ×
I. I NTRODUCTION 20 cm × 16 cm in the center of the cage to cover the device
To test the dynamics and performance of a satellite and its under test (DUT) and allow the movement of the nanosatellite.
subsystems, such as the Altitude Determination and Control The magnetic flux density inside the cage should mimic the
System (ADCS), ground tests on dedicated test rig are con- LEO environment. The maximum expected value is 60 µT.
ducted before launch [1]–[4]. Simulation of the space environ- To cancel the local magnetic field (at the test stand location)
ment is crucial to achieve success. One of the phenomena that and freely shape the simulated magnetic field, the approximate
has to be considered is the Earth’s magnetic field. To reproduce induction generated by the cage should be equal to at least
in the laboratory the Earth’s magnetic field experienced by a 120 µT. Finally, a 30 µ T margin was added resulting in a
satellite, a Helmholtz cage can be used. The device consists of maximum required magnetic field of 150 µT.
three orthogonal Helmholtz coils, each of them being a pair For better space utilization and to ensure uniformity in
of electromagnets on the same axis. By controlling current the designed area, square coils were selected with respective
flow through the coils, the magnetic field inside the cage can side dimensions d of 760 mm, 800 mm and 840 mm. The
be shaped. The nanosatellite under test is placed inside the spacing of the coils was calculated according to the relation
cage on a spherical air bearing or a rotary table. The first w = 0.5445 d, which maximizes the uniformity of the field
option allows 3 degree of freedom rotational movement and [10]. The number of single coil turns n, the area of the wire
– with accurate balancing – simulation of a frictionless and cross section A, and the length of the coil side d to generate
microgravity space environment [5]–[8]. The second option a sufficiently large and uniform field with limited electrical
enables precise control of movement in one axis, which is power input Pmax was selected with the constraints resulting
beneficial in various sensor calibration scenarios [9]. from the equation.
The primary goal of this paper is to present the develop- 2
Umax A
ment, build, and verification of a Helmholtz cage suitable Pmax = , (1)
for testing nanosatellites. It shows electronics hardware and 4ρnd
software which improves precision of generated magnetic where ρ is the resistivity of the wire material and Umax is the
field and simplifies testing procedures. The presented solution maximum voltage across a single coil. The selected parameters
has been constructed at the AGH-UST in Krakow, Poland. are shown in Table I.
Similar realizations have been described in the past with main
focus on construction aspects or Hardware-in-the-Loop (HIL) III. E LECTRONIC D ESIGN
capabilities. The novelty of this work is the incorporation A block diagram of an electronic circuit developed to con-
of the Rapid Controller Prototyping (RCP) methodology to trol the Helmholtz cage is presented in Figure 1. The system
greatly simplify testing small satellites. incorporates a three-channel power interface responsible for
The article is organized as follows. In Section II the design amplifying input voltage signals (ux , uy , and uz ) controlling
essentials are shown. Section III describes the electronic the coil currents, an analog control and measurement signals
hardware setup, while Section IV presents the data acquisition conditioning blocks, and a power supply block generating

978-1-6654-6858-9/22/$31.00 ©2022 IEEE 459

Authorized licensed use limited to: Akademia Gorniczo-Hutnicza. Downloaded on August 30,2023 at 18:45:39 UTC from IEEE Xplore. Restrictions apply.
 TABLE I
M AIN DESIGN SPECIFICATIONS OF THE H ELMHOLTZ CAGE

dx 760 mm
dy 800 mm
dz 840 mm
wx 413.8 mm
wy 435.6 mm
wz 457.4 mm
Maximum magnetic field 150 µT
Maximum current 1.5 A
Number of single coil turns 36
Copper wire diameter 0.9 mm
Maximum power per channel 52 W

supply voltages required by the all-electronic components. The

input (control) and output (measurement) signals are processed
by an RT-DAC/Zynq control board presented in Section IV.
The supply voltages and currents required by the three coil
pairs Lx , Ly , and Lz are provided by three power analog Fig. 1. Block diagram of the control system.
amplifiers with built-in thermal, overvoltage, and short-circuit
protective circuits. Each driver is powered by the DC voltage
of +/-15 V and is capable of supplying maximum current of matrix and a bias/offset vector using the linear regression (least
2 A (the coil resistance limits this current). The use of a mean square) method. The last row is used for validation of
voltage-analog amplifier significantly reduces the distortions the results obtained. As a result, a simple matrix equation
of the magnetic field, especially compared to PWM H-bridge was obtained, enabling precise measurement of coil currents
controllers. The power interface contains current and voltage and compensation of scale error, zero error, and cross-channel
sensors responsible for measuring currents (Ix , Iy and Iz ) and crosstalk:
voltages (Vx , Vy and Vz ) of individual Helmholtz cage coil i=Ks+b (2)
pairs. Currents are measured using 0.01 Ω shunt resistors with T
tolerance of 0.1 % and a temperature coefficient of 75 ppm/◦ C. where i = [ix , iy , iz ] is a vector or coil currents while s =
T
The analog signal conditioning blocks are responsible for [sx , sy , sz ] is a vector of raw measurement results. Matrix
adjusting voltage values obtained from sensors to measurement K and vector b are results of linear regression.
ranges of ADC and DAC converters incorporated into the RT-
DAC/Zynq board. The measured quantities are converted by IV. DATA ACQUISITION AND C ONTROL S OLUTIONS
this block into voltages within a range of +/-10 V. The coils The RT-DAQ/Zynq control board is responsible for closed-
can be controlled in two modes: voltage and current. In the loop control of coil currents. It is equipped with a Xilinx
case of the Helmhotz cage, it is important to maintain the set Zynq SoC, which includes a two-core ARM processor and
value of the magnetic field intensity, hence the current mode is a reconfigurable FPGA fabric [11].
preferred. It is worth noting that the coils are current-controlled To precisely control the magnetic field, high-accuracy reg-
using a regulator implemented on the RT-DAC/Zynq board. ulation of the coil current is needed. It is achieved by closing
Based on the measured current and the desired current value, the control loop, which includes: the measurement reading,
the controller generates a voltage control signal. Calibration of the control calculation for the power analog amplifiers, and the
conditioning blocks embedded in the power stage (amplifier) analog output signals altering. These operations have to be per-
driving three coil pairs of a 3-D Helmhloltz cage was made formed in real time to guarantee the deterministic behaviour.
programmatically. On the basis of the experimental results, They are assigned to be calculated by the dedicated ARM2
a specific matrix was created. The matrix contains results processor. The performance of this system was characterized
of 22 series of measurements. Each single measurement is in [12] where it showed a very low jitter – parameter which
represented by a single row. Subsequent columns of a matrix describes the deviation of the sampling period from the desired
represent, respectively: value.
• three current setpoints for software current controllers for All non-real-time tasks like the communication, data log-
coil pairs of X, Y and Z axes, ging, and interaction with a user are performed by the ARM1
• coil currents measured with three bench Rigol multime- processor running the Linux operating system. The two pro-
ters (considered reference or ground-truth values), cessors communicate through a shared memory buffer.
• raw output voltages of current transducers of the power The FPGA fabric hosts a dedicated configuration that im-
stage (sampled with a sample period of 0.1 s and averaged plements all functions with time-precision requirements that
over a window of 40 or more seconds). cannot be met by real-time software. This applies to the control
The first 21 rows are used to estimate a corrective gain of the A/D and D/A converter and communication with the

460
Authorized licensed use limited to: Akademia Gorniczo-Hutnicza. Downloaded on August 30,2023 at 18:45:39 UTC from IEEE Xplore. Restrictions apply.
 Fig. 2. Simulink block diagram of the control algorithm prepared for automatic code generation.

measuring units by the UART channels. The FPGA matrix data, a 3D magnetic field scanner was prepared. The scanner
blocks work in parallel and independently for both processors. consists of a small industrial robot (Mitsubishi RV-2F-D) with
The RT-DAQ/Zynq board is fully compatible with the rapid a magnetic sensor fixed to the extreme of a 80cm aluminum
controller prototyping methodology [13]. The control algo- profile attached to the robot wrist.
rithm is developed in an environment that offers a high level of Mitsubishi RV-2F-D [14] is a 6-axis industrial robot with
abstraction, i.e. MATLAB/Simulink, and the implementation a payload capacity of 2 kg. The given pose repeatability
on the hardware platform is achieved by automatic generation (according to JIS 8432) is 0.02 mm.
of code in the target programming language. Such an approach Magnetic field measurement is performed with the MR3
allows for problem solving, while omitting implementation magnetoresistive milligauss meter from Alpha Labs [15]. Its
issues. Figure 2 shows the Simulink block diagram that depicts declared sensitivity error is ±0.5% of the reading and the
the control algorithm for the closed loop control of the coil maximum offset error is ±50 nT. The sensor comprises three
current. The control algorithm consists of a separate discrete- discrete sensing elements (for fields of the X, Y, and Z axes),
time PID regulator for each pair of coils. As the same which are factory calibrated to ensure that the three axes
environment is used for the development and tuning of the are perpendicular. The sensing elements are located a few
control system, the implementation time is greatly reduced. millimeters apart.
Tuning the algorithms proprieties and displaying acquired data The robot and the Helmholtz cage are placed side by side
can be done on-line without the need of code regeneration. to ensure optimal robot range of operation and hence maximal
scanning volume. The 3D printed markers mounted in 3 fixed
V. R EALIZATION AND VALIDATION OF THE H ELMHOLTZ positions inside the cage allow to derive a transformation from
CAGE
the robot reference frame (RRF) to the cage reference frame
The cage structure is based on extruded T-slot aluminum (CRF).
profiles with 20 mm × 20 mm cross section. The profiles are The validation procedure consists of two phases. In the first
connected with rigid fasteners and brackets. The cage coils phase, the operator moves the end-effector of the robot to three
are placed in aluminum U-channels and mounted on the frame markers in known locations inside the Helmholtz cage. The
with 3D printed elements. The base plate made of composite robot position is read to determine the translation between
materials (polymethyl methacrylate (PMMA) and plywood) RRF and CRF frames. In the second phase of the validation
ensures required stiffness of the entire construction. The CAD procedure, the coils are energized with sets of constant current
model and the real-life laboratory setup are presented in Figure to produce a uniform magnetic field. The robot places the MR3
3. sensor at points evenly distributed in volume surrounding the
To ensure correct behaviour and characterize cage perfor- center of the cage. The robot, the cage, and the milligauss
mance, an experimental setup was built. To collect high-quality meter are all controlled by a computer. This allows us to

461
Authorized licensed use limited to: Akademia Gorniczo-Hutnicza. Downloaded on August 30,2023 at 18:45:39 UTC from IEEE Xplore. Restrictions apply.
Fig. 3. Helmholtz cage is powered with computer-controlled current driver. The robot carrying a precise 3-axis magnetometer mounted on the boom can
repeatably measure the magnetic field in predetermined positions inside the cage workspace.

automate the measurement process. It is useful because the

complete 3D scan of the magnetic field workspace inside the
cage is relatively long and takes around 3.5 hours.
To ensure deterministic, repeatable experiments, a MAT-
LAB script has been written that controls the Helmholtz cage
driver and the robot. The test locations are organized in a
5 × 5 × 5 point grid that spans 140 mm in each direction.
At every location, several magnetic field measurements are
performed for different coil currents. After reaching each of
the points, the following steps are executed:
1) Wait for 10 seconds to allow the boom to settle
2) Disable the coils driver and perform 80 measurements
of the background field
3) Enable the required current to the coils and perform 80 Fig. 4. Nonuniformity of the magnetic field in the Helmholtz cage. The
deviation of the field is measured as percent difference between the magnitude
measurements of the field of the field vector at the specific point of the cage and the central point (0,
4) Move the sensor to the next measurement point 0, 0). The angle nonuniformity is measured alike.
The magnetic field nonuniformity experimentally measured
is depicted by cross-sectional heat maps in Figure 4. The
results obtained reveal a maximum deviation of the field measurement and appropriate feedback control algorithm can
magnitude lower than 1.5 % and angular nonuniformity lower further improve the accuracy of the simulated field.
than 0.4°. Those results are consistent with the simulations. R EFERENCES
VI. C ONCLUSION [1] R. C. da Silva, I. S. K. Ishioka, C. Cappelletti, S. Battistini, and
R. A. Borges, “Helmholtz cage design and validation for nanosatellites
The paper described the complete workflow that led to hwil testing,” IEEE Transactions on Aerospace and Electronic Systems,
the construction and commissioning of the Helmholtz cage vol. 55, no. 6, pp. 3050–3061, 2019.
dedicated to nanosatellites ADCS testing. The use of the [2] A. R. Al-Abri, N. I. Al-Khoori, M. S. Al-Marri, A. A. Al-Hamed,
and P. Sreekumar, “Helmholtz cage for attitude testing of cubesat,” in
rapid controller prototyping methodology and the extensible 2022 Advances in Science and Engineering Technology International
data acquisition and control board greatly simplifies usage Conferences (ASET). IEEE, 2022, pp. 1–6.
of the test stand. Moreover, the incorporation of additional [3] T. Goyal, K. Batta, P. Baranwal, A. A. NA, N. More, R. Panigrahi,
and T. Kaushik, “Design and development of a three-axis controlled
sensors and actuators is simplified due to the FPGA fabric helmholtz cage as an in-house magnetic field simulator for cubesats,” in
reconfiguration capabilities. In the next step, a precise LED International Astronautical Congress, 2017.
spotlight with uniform illumination (Mightex PLS-5500-030) [4] Z. P., K. P., and G. A., Batch Algorithm for Balancing the Air Bearing
Platform. Springer, 2020, vol. 1196, pp. 736–746.
will be integrated into the test stand to simulate the position [5] N. M. Hatcher and R. N. Young, “An automatic balancing system for use
of the Sun and thus enable calibration of the sun sensor. The on frictionlessly supported attitude-controlled test platforms,” NASA,
observed external disturbances in the magnetic field suggest Tech. Rep., 1968.
[6] J. Young, “Balancing of a small satellite attitude control simlator on
the need for closed-loop control with magnetic feedback. A an air bearing,” in Proceedings of the Utah Space Grant Consortium
supplementary magnetic field sensor for background noise Symposium, 1998.

462
Authorized licensed use limited to: Akademia Gorniczo-Hutnicza. Downloaded on August 30,2023 at 18:45:39 UTC from IEEE Xplore. Restrictions apply.
 [7] S. Cho and N. H. McClamroch, “Feedback control of triaxial attitude
control testbed actuated by two proof mass devices,” in Proceedings
of the 41st IEEE Conference on Decision and Control, 2002., vol. 1.
IEEE, 2002, pp. 498–503.
[8] J. L. Schwartz and C. D. Hall, “System identification of a spherical air-
bearing spacecraft simulator,” in in Proceedings of the AAS/AIAA Space
Flight Mechanics Conference, no. AAS 04-122, 2004.
[9] P. Mahavarkar, J. John, V. Dhapre, V. Dongre, and S. Labde, “Tri-
axial square helmholtz coil system at the alibag magnetic observatory:
upgraded to a magnetic sensor calibration facility,” Geoscientific Instru-
mentation, Methods and Data Systems, vol. 7, no. 2, pp. 143–149, 2018.
[10] M. E. Rudd and J. R. Craig, “Optimum spacing of square and circular
coil pairs,” Research Papers in Physics and Astronomy, 1968.
[11] INTECO, “Rt-dac/zynq user’s manual,” http://www.inteco.com.pl/Docs/
Rtdac Zynq.pdf, accessed: 2022-04-06.
[12] D. Knapik, K. Kołek, M. Rosół, and A. Turnau, “Autonomous,
reconfigurable mobile vehicle with rapid control prototyping
functionality,” IFAC-PapersOnLine, vol. 52, no. 8, pp. 13–18,
2019, 10th IFAC Symposium on Intelligent Autonomous Vehicles
IAV 2019. [Online]. Available: https://www.sciencedirect.com/science/
article/pii/S2405896319303611
[13] K. Kołek and K. Piatek, “Rapid algorithm prototyping and implemen-
tation for power quality measurement,” EURASIP Journal on Advances
in Signal Processing, vol. 2015, no. 1, pp. 1–12, 2015.
[14] Mitsubishi, “Rv-2f-d series standard specifications manual,” http://www.
geva-roboter.at/files/rv-2f series standard specifications.pdf, accessed:
2022-04-06.
[15] “Magnetometer model mr3,” AlphaLab, Tech. Rep., 2016.
[Online]. Available: https://www.alphalabinc.com/wp-content/uploads/
2018/02/mr3 2016.pdf

463
Authorized licensed use limited to: Akademia Gorniczo-Hutnicza. Downloaded on August 30,2023 at 18:45:39 UTC from IEEE Xplore. Restrictions apply.