Design, Integration and Testing of the TERRIERS Attitude Control and

Determination System



Dr. John Hanson,
Senior Engineer, AeroAstro LLC

Mr. Richard Warner,
Chief Technical Officer, AeroAstro LLC

Mr. Russell Brinkmann,
Senior Engineer, AeroAstro LLC

Abstract

The Tomographic Experiment using Radiative Recombinative Ionospheric EUV and Radio Sources (TERRIERS) is a satellite scheduled for launch in late 1997 or early 1998. The goal of the TERRIERS mission is to demonstrate global ionospheric tomography and to utilize this technique to study ionospheric and thermospheric processes. TERRIERS is a spin-stabilized satellite with magnetic torque control. Attitude information is provided by two magnetometers, two body mounted horizon crossing indicators and a suite of sun sensors. In its science collection mode, it is spinning at 10 RPM in a Thomson Spinner configuration. AeroAstro LLC has been responsible for the design, manufacturing and testing of the spacecraft bus, including the complete attitude control system, under contract to Boston University.
In order to test the integrated attitude control system with the spacecraft bus and payload, a hardware-in-the-loop test is being conducted. In this test, signals to the attitude control sensors (sun sensors, horizon crossing indicators, magnetometers) are simulated with hardware. Spacecraft commands to the magnetic torquers are sensed by an external computer that monitors the spacecraft bus. These commands are used in a simulation of the spacecraft dynamics. This simulation then controls the ACS sensor inputs in real time. Using this experimental apparatus, the performance of the attitude control system is demonstrated and realistic mission simulations are being performed.
This paper will present details of the design of an attitude control system for a small satellite. It will also discuss the testing of the attitude control system, with results from the hardware-in-the-loop test.

Introduction

The Tomographic Experiment using Radiative Recombinative Ionospheric EUV and Radio Sources (TERRIERS, Figure 1) is a satellite scheduled for launch in late 1997 or early 1998. The goal of the TERRIERS mission is to demonstrate global ionospheric tomography and to utilize this technique to study ionospheric and thermospheric processes. Under contract to Boston University, AeroAstro LLC has been responsible for the design, manufacturing and testing of the spacecraft bus. This includes the complete attitude control system (ACS).
The TERRIERS ACS has two general requirements: to move the spacecraft from launch vehicle tip-off to a spin-stabilized, sun-pointing configuration and to maintain an acceptable orientation when the payload is taking data. This requires the ACS to operate in several different modes and to switch autonomously between them (Figure 2). Within one minute of release from the launch vehicle, the flight computer will have booted and the ACS will be running. The ACS Mode Switching Logic will move the spacecraft from an arbitrary initial tumble to a spin-stabilized, sun-pointing configuration. Ground commands will be used to spin the spacecraft up to 10 RPM and to turn the spacecraft to its data collection configuration.
The science mission requires the primary instruments (Tomographic EUV Spectrographs) to scan the earth atmosphere at a frequency of 8 to 12 RPM. Rather than use a scanning platform, when operational, TERRIERS will be spin-stabilized at 10 RPM in a Thomson Spinner configuration, i.e., with the spacecraft spin axis in the same direction as the orbit normal. TERRIERS is required to maintain this orientation to within 1.5 deg with knowledge to 0.25 deg in the direction of spin and 0.5 deg in any other direction. Spin rate information is provided to the payload for preprocessing the tomographic data.


Figure 1: TERRIERS Figure 2: ACS Mode Switching Logic

In order to test the integrated attitude control system with the spacecraft bus and payload, a hardware-in-the-loop (HITL) test is being conducted. In this test, the environment seen by the attitude control sensors (sun sensors, horizon crossing indicators, magnetometers) are simulated with hardware. Spacecraft commands to the magnetic torquers are sensed by an external computer that monitors the spacecraft bus. These commands are used in a simulation of the spacecraft dynamics. This simulation then controls the simulated sensor inputs in real time. Using this experimental apparatus, the functionality of the attitude control system is demonstrated and realistic mission simulations can be conducted. The goal of these tests is two-fold: to demonstrate the stability of all of the control systems and to provide confidence during mission simulations that no interactions between the ACS and the rest of the spacecraft bus and payload exist.

ACS Hardware

TERRIERS has several different sensors to measure attitude. In the Operational Mode, the primary ACS sensors are two body-mounted horizon-crossing indicators (HCI's) that provide roll angle, pitch angle and pitch rate information. One primary and one backup magnetometer are used to measure the local magnetic field. These signals are differentiated to provide a measure of spin-rate when the HCI's cannot be used (e.g., when the spacecraft is not in the Operational Mode). Seven coarse sun sensors and four medium sun sensors measure the direction of the sun-line in spacecraft body coordinates.


Figure 3: TERRIERS ACS Block Diagram

The TERRIERS ACS block diagram (Figure 3) shows the major hardware components of the system and how they are connected.
The sun vector is determined by an array of simple photodiodes called Coarse Sun Sensors (CSS). These sensors utilize the cosine drop off in current from the photodiodes when the sunlight hits at less than normal incidence and provide low accuracy (5 ) but full sky coverage. There are seven CSS's mounted at various points around the spacecraft. Four Medium Sun Sensors (MSS) provide a more accurate measure of the sun vector (0.1 ) over a more limited field of view (~45 from spin axis). These sensors utilize a 4 quadrant photocell illuminated by an offset aperture. By measuring the current output of each quadrant, the aperture centroid can be located in the plane of the photocells. The offset of the centroid from the center of the instrument gives a measure of the sun-line in body axes. The four MSS's are mounted in a package on the front of the solar array and are pointed along the y-axis. Both sun sensors are manufactured by AeroAstro.
A redundant pair of Schonstedt SAM-73C magnetometers (Mag 1, Mag 2) sense the earth magnetic field in three axes. These are fluxgate magnetometers and provide 1% field magnitude accuracy over a 50 mT range and 1 field orientation accuracy. A general concern with magnetometers is error due to spacecraft generated magnetic fields, either due to permanent magnets or current loops. While it is common practice to mount these devices on deployable booms or to place them on the ends of deployable solar arrays to reduce the effects of such interactions, on TERRIERS both magnetometers are mounted on the backside of the solar array. The largest potential interaction is with the magnetic field created by the torque coils. The effect of this interaction is removed in flight by careful calibration on the ground. The ACS software uses knowledge of the state of the three torque coils along with an interaction matrix to remove the effect from the raw magnetometer data. Permanent magnetics are reduced by careful material selection and current loops by careful electrical design.
Two A & M Associates Horizon Crossing Indicators (HCI 1, HCI 2) are used to determine the earth vector. These are infrared sensors operating in the 15 micron wavelength to sense the C02 layer about the earth. As the vehicle spins, these instruments will send a pulse whenever their narrow FOV's cross from the sky-to-earth or from earth-to-sky. These crossings are called "lead" and "trail" crossings, respectively. The two HCI's are mounted in the yz-plane such that they point 30 deg offset from the z-axis. This makes the two instruments scan different paths over the earth as TERRIERS spins.
The actuators used on TERRIERS are an orthogonal set of torque coils that generate moments by interaction with the earth magnetic field. They have a full-scale range of 8A-m2 and are controlled by a very high efficiency pulse width modulator that provides 12 bits plus sign resolution. Each of the three axes has redundant coils and drivers. The driver board has the capability to compensate for coil resistance changes due to temperature and bus voltage changes due to battery level (both of which affect the coil scale factor). The coils and coil drivers are manufactured by AeroAstro.
Analog sensor (sun sensor, magnetometer) signals are wired up to a data acquisition board called the TSAX board. This board contains signal conditioning circuits and a 16 bit A/D converter. The HCI pulse signals are routed through the TSAX board to the main processor, where they generate an interrupt to allow for very accurate timing. This interrupt is handled in real time by the system software. When an interrupt is received, the system software stores the current spacecraft clock reading in a buffer and increments a counter. When the counter reaches four, the HCI's have registered one lead and one trail crossing each and the ACS can use the information on its next software cycle to compute the spacecraft attitude. After processing the HCI data, the ACS resets the counter to zero and the process repeats.
The TSAX board and the coil driver board communicate with the processing system via a redundant, error detecting, serial bus called AUX. Virtually all spacecraft subsystems are controlled by the AUX bus. Some of the advantages of this digital bus approach are ease of expansion and the ability to develop subsystems without needing a spacecraft processor to generate AUX commands (since AUX cycles can be generated by any desktop computer). In addition, during ground testing, the spacecraft AUX bus can be taken over by ground support equipment to allow direct control and interrogation of subsystems. The AUX bus can also be monitored by external hardware without interfering with its operation. This capability is used in the HITL test as described below. The TSAX and processor are manufactured by AeroAstro.

ACS Software

The TERRIERS attitude control system software runs at 4 Hz. Each execution of the software can be divided into several steps:
  1. Read all ACS sensors (magnetometers, sun-sensors, horizon crossing
    indicators).
  2. Process and filter raw sensor data to get usable information.
  3. Based on the processed sensor data, switch into the appropriate control
    mode.
  4. Compute control torques for the control mode.
  5. Issue current commands to magnetic torque coils.
  6. Send appropriate information to the payload.

All of the ACS sensors are sampled through the TSAX board as previously described. The raw data from the sensors are used to generate a usable, filtered reconstruction of the vehicle state.
The magnetometer data is transformed from a voltage measurement into a magnetic field measurement in spacecraft body axes. This is filtered prior to use by the ACS. Additionally, the signal is differentiated to get a measure of the spacecraft spin rate. However, only that portion of the spin vector that is perpendicular to the magnetic field line is observable from magnetic field measurements.
The current measurements from the CSS and MSS package are used to generate an estimate of the sun vector in body axes and to indicate when the spacecraft is passing into or out of eclipse. The former is filtered prior to use by the ACS, while the latter is sent to the payload for use in governing the operation of instruments.
When four HCI interrupts have occurred, the ACS will compute a new estimate of the vehicle attitude. Figure 4 shows a typical scanning path of the two HCI's and the timing diagram for the related interrupts. By averaging the four crossing times, a measure of the time of zenith crossing is derived. By taking the difference in this time between two sequential sets of data, the spin rate is derived. The roll angle of the vehicle is related to the difference in the lengths of the paths swept out by the two HCI's. That is, the roll angle is a function of the difference between the times taken by each HCI to record a lead interrupt and then a trail interrupt. The time of zenith crossing is passed to the payload for preprocessing of the tomographic data.



Figure 4: HCI Operation

Based on the processed sensor data, TERRIERS will autonomously select the appropriate control mode. Transition from one mode to the next is defined by parameters stored in EEPROM. These parameters can be modified in flight to change the characteristics of the mode switching logic.
A typical sequence of this logic follows:

Wait Mode (0) - Within one minute of release from the launch vehicle, the flight computer will have booted and the ACS software will be running. For the first three minutes of the mission, the ACS will be in the Wait Mode where the magnetometers and sun sensors are sampled at 4 Hz and 1 Hz, respectively, but no control torques are applied. This will allow filters on the sensor signals to become initialized and prevent large initial control torques.

Detumble Mode (1) - Once the Wait Mode is completed, the ACS will switch into the Detumble Mode. The rate of change of the magnetometer signal is fed back to the torque coils with the goal of reducing the spin rate of the spacecraft in all axes. While only that portion of the spin rate that is normal to the magnetic field lines is observable, that is also the only portion of the spin rate that is controllable with magnetic torque coils.

Spin-up Mode (2) - When the total angular velocity of the satellite has been reduced, the ACS will shift into Spin-Up Mode. The magnetometer is used to estimate the angular velocity of the vehicle, and the x and z torque coils are actuated alternately to spin the spacecraft up around its y-axis. This controller uses proportional feedback, but the gains are selected to be very large, making it a bang-bang system. The nutation of the spin axis is damped by use of a "minus-K-Bdot" controller (Wertz, p. 630).

Sun-Acquisition Mode (3) - When the spin rate has increased to 1 RPM, the ACS switches into Sun Acquisition Mode and turns the solar panel (+y axis) to point at the sun. Obviously, as the spacecraft passes into eclipse, the ACS will switch out of this mode and back into the Spin-Up Mode, but as soon as the vehicle passes out of eclipse, it will switch back into Sun Acquisition Mode. This controller is a Digital Linear Quadratic Regulator (DLQR) and regulates the spin rate as well as the sun-pointing angles. After TERRIERS has acquired the sun and reduced the angle between the solar panel and the sun-line to zero, it will remain in the Sun Acquisition Mode. This configuration, with the solar panel normal to the sun-line, provides the maximum power to the spacecraft, has a benign thermal environment and will allow all subsystems to be tested prior to moving to the Operational Mode. The ACS Mode Switching Logic will not switch the spacecraft into the Operational Mode. This can only be done by issuing a command from the ground station.

Ground Command Mode (4) - Before it can be placed in a Thomson Spinner configuration, the spin axis must be moved from pointing at the sun to pointing along the orbit normal and the spin rate must be increased from 1 RPM to 10 RPM. Since it is in a 9 am to 9 pm sun-synchronous orbit, the spin axis must be turned by a total of 45 degrees. This is done by issuing time-tagged body torque commands from the ground. While in the Ground Command Mode, the ACS will issue no torque commands. This will keep it from stepping on commands issued by the ground, and is also useful as an emergency mode in case of ACS malfunction. This mode can be entered only by direct command from the ground. The ACS Mode Switching Logic cannot switch into this mode autonomously.

Operational Mode (5) - Once the vehicle spin axis has been turned by about 20 degrees and the spin rate has been increased to 10 RPM (again by time-tagged body torque commands), the ACS will be switched into the Operational Mode (Figure 5) and the spacecraft will reduce its roll and yaw angles autonomously. In this mode the spacecraft attitude is defined relative to local vertical coordinates by a body 3-1-2 rotation. These rotations are called yaw, roll and pitch, respectively. The basis vectors for the local vertical system are nx in the direction of spacecraft velocity, ny in the direction of the orbit normal and nz in the anti-nadir direction. For small yaw and roll angles, roll is a rotation about nx, yaw is a rotation about nz, and pitch is a rotation about ny. The spin-rate, then, is the rate of change of the pitch angle. The horizon crossing indicators measure roll angle, spin rate and the time at which nadir crossing occurs. Since roll and yaw are coupled by the motion of the vehicle in the orbit, controlling roll is sufficient to control both roll and yaw. The roll controller was designed with frequency techniques in the digital domain. This controller has one zero and two poles, which puts the crossover of the closed loop system at approximately 0.0004 Hz and provides positive damping to the system. The torque commands are computed in local vertical coordinates every time an updated attitude is computed by the HCI's. These are transformed into body axes using the estimate of the spacecraft spin rate. The modes associated with the nutation of the spin axis are controlled at 4 Hz with a "minus-K-Bdot" controller. The spin rate control requires only proportional feedback.


Figure 5: Operational Controller

Hardware-in-the-Loop Test

The most exhaustive test of the TERRIERS attitude control system is the hardware-in-the-loop (HITL) test. In this test, the spacecraft is run under conditions as near to actual operating conditions as is feasible. The TERRIERS HITL setup is shown in Figure 6. The central piece of HITL equipment is the simulation computer (Macintosh Quadra 950, with a 33 MHz Motorola 68040 CPU). The simulation computer uses the commanded current to the torque coils to integrate the spacecraft dynamic forward in time. This simulation includes a full non-linear dynamic model of the vehicle, orbit mechanics and potential errors such as misalignments of the actuators and sensors relative to body axes. The simulation is run in a poling loop with the TERRIERS ACS, running on the spacecraft, controlling the rate of the loop. It would have been better to use an independent artificial environment simulator, but budget constraints necessitated the use of the same simulator used to develop control algorithms.
A piece of ground support equipment, called a "snooper," watches the data traveling over the AUX bus, in effect, eavesdropping on spacecraft AUX commands. Commands to each of the three torque coils occur only once per ACS cycle. The snooper presents one CRC (error detection code), one command byte and 16 data bits to the simulation computer for each torque command. The simulation computer sorts this data and integrates the spacecraft equations of motion forward in time for 0.25 seconds. At the end of this integration step, the simulation computer issues commands to stimulators attached to the TERRIERS ACS sensors. These stimulators trick the spacecraft into thinking that it is actually on orbit with appropriate magnetic field, solar and earth stimuli. The simulation computer then waits for the snooper to read the next set of torque commands (indicating that the ACS has completed another 0.25 second cycle) before continuing its integration.
For each of the primary sensors, physical stimulation, controlled by the simulation computer, is provided. For the MSS, a secondary or backup ACS system, the sensors are bypassed and the sensor AD board (TSAX) is stimulated directly. Communication between the MSS and magnetometer stimulators and the simulation computer is done over a common AUX bus (isolated from the spacecraft AUX bus). However, for the HCI and CSS, a parallel bus is used with each bit corresponding to a specific action.
Initially, an attempt was made to stimulate each MSS with a collection of LED's, each controlled by a PWM circuit. Unfortunately, these LED's generate insufficient light to properly stimulate the sensor. Instead, the MSS package is bypassed entirely and each MSS input to the TSAX is stimulated directly by a PWM signal run through a low-pass filter. This system has a rise time of 0.333 sec from 0 to 5 V and a ripple voltage of roughly 8% of full-scale at the PWM frequency of 1.9 kHz. The ripple voltage is filtered by the TSAX anti-aliasing filters, resulting in a millivolt level oscillation in the raw MSS signal, as seen by the ACS.



Figure 6: TERRIERS Hardware-in-the-Loop Setup

Each CSS is stimulated by a single infrared LED. Two bits from the simulation computer control two bias currents through the LED and, consequently, the apparent intensity of the sun seen by each CSS (Figure 7). Since there are only two bias levels to adjust, it is fairly easy to compensate for the nonlinearities of each LED's brightness-current curve.
The active magnetometer is placed in a Helmholtz cage that has three orthogonal electromagnetic coils, each driven by a pulse-width modulated circuit (Figure 8). Note that the Helmholtz driver is just a prototype torquecoil driver, thus reducing design effort. By carefully aligning the magnetometer and cage axes and using dual coils to minimize near field effects, off-axis coupling between the coils and the magnetometers is kept below 10%. Furthermore, the interaction between the stimulation coils and the magnetometer is fully calibrated. Initial tests of this device in the Boston University cleanroom showed large fluctuations of the ambient magnetic field with particularly large oscillations (~1% of steady state) at 60 Hz. In order to reduce this noise to acceptable levels, a mu-metal shield is placed around the cage. The resulting performance of the system is shown in Figure 9. Here, a comparison of expected and measured magnetic field is made when the spacecraft is spinning at 10 RPM and indicates that the Helmholtz cage is quite effective.
As mentioned previously, the HCI's sense the 15 micron emission line of CO2 in the Earth atmosphere. In the HITL test, an HCI is stimulated by opening and closing a shutter between it and a heat source, in this case a halogen lamp (Figure 10). Timing signals are provided by the simulation computer to small stepper motors that move the shutters open and closed in 60 to 70 msec.




Figure 7: CSS Stimulator Figure 8: Magnetometer Stimulator




Figure 9: Magnetometer Stimulator Performance



Figure 10: Horizon Crossing Indicator Stimulator

The shutters are constructed from cardboard because it is light enough for quick movement under small torques, with good insulating properties for improved HCI trigger response, and will not damage the spacecraft in the event of a shutter malfunction. By covering the cardboard shutters with tape, particulate generation can be reduced enough to use them in the TERRIERS class-10,000 clean room. The shutters are connected directly to the HCI's with a friction fitted keyed aluminum mounting plate that prevents them from shifting during HITL runs or changing position between runs.
Shutter timing commands are generated by an autonomous process running continuously and in parallel with the simulation of the vehicle dynamics on the simulation computer. Communication between the simulation and the timing loop is done by shared memory. This process has 1 msec resolution and roughly 24 hour maximum duration. This implementation of the timing loop allows for accurate placement of the shutter commands over short timescales. However, it does introduce a long-term drift in these same timing signals since it is not synchronized with the simulation computer.
Figure 11 shows the performance of the horizon crossing indicator stimulators in the HITL test. Stimulation of the HCI's yield roll angle errors of 1.25 deg to 0.5 deg over a 5 deg range. At zero roll angle, there is approximately zero roll error. The error in spin rate is generally zero mean with a standard deviation of about 0.5 deg/sec. Several outlying data points, however, increase this error to as much at 20 deg/sec in some cases. It is believed that these outlying points are the results of errors in the data collection process and not due to problems with the stimulators, simulation process or the TERRIERS ACS.



Figure 11: Performance of HCI Stimulators

Hardware-in-the-Loop Test Results

Several tests have been conducted using the hardware-in-the-loop test setup. These include a number of short runs simulating various stages of the tip-off to sun-pointing phase of the mission, one six-hour test running from tip-off to sun-pointing and one four-hour operational test. The results of the six-hour test starting at launch vehicle tip-off are presented in Figure 12. Initialized in a tumble with an arbitrary initial attitude, the vehicle moves through its mode switching logic from Wait Mode, briefly into Detumble Mode and then into Spin-Up Mode. The ACS spins up the vehicle to about 4 deg/sec and then switches into sun-pointing mode. In this mode it continues to spin the vehicle up to 6 deg/sec and then regulates the spin rate. While this HITL test was being conducted, other spacecraft tests were performed. These included regular radio contacts and a test of the power shedding logic. At the start of the simulation, the spacecraft was disconnected from its power-supply and run completely off batteries. At approximately 70 minutes into the simulation, the power shedding logic detected a decrease in battery capacity and switched the ACS into Ground Command Mode to conserve power. After validating this code, the power supply was reconnected and the ACS was switched back into its Sun-Acquisition Mode and allowed to continue its operations. However, the spacecraft software was not properly configured and the sun vectors computed by the spacecraft were in error, thus invalidating the sun-acquisition portion of the test.
Fortunately, tests conducted previously indicate that the Sun-Acquisition Mode operates properly (Figure 13). In this case the spacecraft is initialized with a spin-rate of 6 deg/sec and an attitude where the solar panel is pointed away from the sun. The ACS regulates the spin rate and begins to move the vehicle into a sun-pointing configuration, shown by the increase in the component of the sun vector along the body y-axis.
One test of TERRIERS in its Operational Mode was conducted (Figure 14). The simulation computer was initialized with a roll error of 2 deg and an initial spin rate of 60 deg/sec (10 RPM). After four hours, the operational controller had reduced the roll error to less than 1 deg. The yaw angle, because it is coupled to the roll angle, was also reduced to less than 1 deg. The spin rate shows a slow increase from 60 deg/sec to 60.4 deg/sec, but this is within the error in the ability of the HCI stimulators to generate spin rates. Certainly, longer HITL tests will be conducted to verify the stability of the spin rate controller. The spin rates in the x and z axes are kept below 0.01 deg/sec, indicating that the polhode motion of the spin axis was also controlled.

Figure 12: HITL Spin-Up Test Results


Figure 13: HITL Sun-Acquisition Results

Figure 14: HITL Operational Mode Results


Conclusions

Low-cost spacecraft require a balance between what would be desirable for ACS testing and what can be achieved under budgetary constraints. The goal is to achieve a large fraction of the confidence that a more complete HITL test would provide for a small fraction of the effort that such a test would require.
AeroAstro LLC has designed, manufactured and tested the attitude control system for the TERRIERS satellite. In addition, a complete hardware-in-the-loop test setup has been created and validated. Tests using the HITL setup demonstrate the stability and effectiveness of the Operational Mode controllers and the Detumble and Spin-Up Mode controllers. Preliminary results indicate that the Sun-Acquisition controller is also operating properly. Future HITL tests will verify its stability. The HITL setup will be used during mission simulation where the spacecraft and payload will be exercised continuously for weeks at a time, thus building confidence in the total spacecraft system as the TERRIERS program moves to its launch.

Acknowledgements

The authors would like to acknowledge the efforts of several people who have contributed to the design, manufacturing and test of the TERRIERS attitude control system: Frank LaRosa, Scott McDermott, Sarah Wright, Joe Kendall, Angie Cuccio, Eric Aamot and Rob Baltrum.

References

Bate, Roger R., Donald D. Mueller, Jerry E. White. Fundamentals of Astrodynamics .. New York: Dover Publications, Inc., 1971.

Kane, Thomas R., Peter W. Likins, and David A. Levinson. Spacecraft Dynamics . New York: McGraw-Hill, Inc., 1983.

Spacecraft Attitude Determination and Control . James R. Wertz, ed. Netherlands: Kluwer Academic Publishers, 1978.