Over the first two and a half years of the FUSE mission, the satellite attitude control system design using the FES and IDS to provide fine guidance data to the ACS worked as intended. Many parameters were tuned-up during in-orbit checkout: FES to S/C alignment, FES optical distortion calibration, IDS image processing parameters, IDS Star-identification parameters, and FES timing and subarray sizes. Most parameters needed only one adjustment; the image processing parameters needed several iterations over the first year of the mission, however, as various (e.g. crowded, high/extended nebulosity, extremely bright star) star-fields were encountered. Specific star field tables were generated and loaded for particular observations that needed them. Pointing stability was typically a half-arcsecond or better, and slews and target acquisitions were efficient and reliable.
Over the subsequent years of the FUSE mission, however, the ACS subsystem experienced the lion’s share of the mission-critical hardware anomalies. Four of the six gyro channels failed, and each of the four RWAs failed, ultimately resulting in the termination of the mission. The flight software in the FES, IDS, and ACS was rewritten to accommodate the loss of gyros. Subsequent to each wheel failure, the ACS software was further rewritten to compensate for the reduction in attitude control. The target acquisition scripts in the IDS were rewritten on a nearly continuous basis to improve the robustness of operations.
However, in other respects, performance of the ACS remained nominal throughout the mission.
The Tri-axial Magnetometers (TAMs) performance exceeded pre-launch expectations. They were calibrated to better than 0.2 degree, whereas the pre-launch expectation was 2 degrees.
The performance of the Coarse Sun Sensor (CSS) assemblies was similar to other Explorer Platform missions using similar assemblies. Apart from the additional signal from the Earth’s albedo corrupting the CSS readings, there were no anomalies in the CSS subsystem during the mission.
Prior to the advent of one-wheel operations, the solar arrays were typically moved only during slews to new targets. Typically the slews were much smaller than 90 degrees with an average of three slews per day. After one wheel operations began, the Solar Array Drive (SAD) mechanisms were in use for much more time than envisioned pre-launch. Nevertheless, there were no anomalies of the SAD hardware during the mission.
The “gyroless” attitude determination software worked quite well, so the gyro failures had no direct impact on FUSE science data beyond increased overheads needed for target acquisitions. If anything, pointing stability improved because the angular rates determined from FES data had lower noise than those of the gyros.
The loss of reaction wheels, however, did have an impact on the quality of the science data. Pointing of the FUSE spacecraft was originally controlled with all four reaction wheels, which typically maintained a pointing stability of 0.2"-0.3". After the failure of the yaw and pitch reaction wheels, it became necessary to control the spacecraft orientation along the A axis (see Figure 3-1) with MTBs. Spacecraft drift increased along this axis: typical values were 1-2 arcsec, but drifts of many arcsec occurred occasionally. Without correction, such motions of a target within the LWRS aperture (the default aperture in this mode) could degrade the resolution of the spectra, so the CalFUSE pipeline was modified to correct the data for spacecraft motion during each exposure. For time-tag observations of point sources, individual photon events were repositioned based on fine pointing data (FPD) in the telemetry. For histogram observations, it was only possible to correct for the exposure time lost to large excursions of the target outside the aperture.
Following the failure of the roll-axis wheel, the frequency of large pointing drifts increased significantly. However, the transition between stable pointing and unstable pointing was rather abrupt. If the magnetic torque available exceeded the gravity gradient, then the pointing often remained stable to a few arcsec, though excursions of 5-10 arcsec were more common than before. If the gravity gradient exceeded the available magnetic torque, however, then the resulting excursion would move the source far out of the aperture. Thus, the main impact of one-wheel operations was on observing efficiency and limitations on target availability, rather than on spectral resolution.
The chronology of the major events associated with the IRUs over the course of the mission are listed in Table 7.2‑1. Two of the low-intensity flags were set in IRU-A roughly six months after launch, and the third was set three months later. This rapid drop in laser intensity prompted initial thinking about how to operate with reduced or without gyros; with the failure of the first channel in May 2001, planning for gyroless operations began in earnest. The ACS switched autonomously to IRU-B when the first channel failed in IRU-A. The first version of “gyroless” attitude determination and control software was uplinked the week of April 16, 2003. When the first channel (Z) in IRU-B failed in July 2003, the software began using the new rate estimator logic in place of the failed gyro channel. In May 2004, noise in the IRU-B pitch axis began to increase. This was first noticed in August, and reached the point of significantly degrading performance in September. This gyro channel was manually removed from use on September 24, 2004. This failure mode differs from that of the other channels, in that the gyro continued to provide data. Data from this channel might have been usable with suitable filtering, but this possibility was never pursued. When the roll wheel failed in December 2004, IRU-A was powered back on so that there would be gyro data on all axes while trying to get one-wheel control to work. The roll axis in IRU-B failed on April 17, 2005, which ended the use of IRU-B for guiding. It was occasionally powered on thereafter, for thermal or other ancillary reasons. The pitch and yaw channels in IRU-A continued to operate normally until the satellite was decommissioned in October 2007. The failures of the IRU-A Y and IRU-B Y,Z channels were similar in nature: the measured rate dropped abruptly to zero and remained exactly zero. There was no apparent change in performance prior to the failure. This behavior is consistent with cessation of lasing as a result of declining intensity.
|06-24-1999||Launch – IRU A ON|
|01-06-2000||IRU A Z low intensity|
|01-18-2000||IRU A X low intensity|
|04-19-2000||IRU A Y low intensity|
|05-30-2001||IRU A Y fails Switch to IRU B|
|08-31-2001||IRU B X low intensity|
|10-06-2001||IRU B Y low intensity|
|12-10-2002||IRU B Z low intensity|
|07-31-2003||IRU B Z Fails Begin 2-gyro ops|
|09-28-2004||IRU B X too noisy Begin 1-gyro ops|
|12-27-2004||Roll RWA Fails - Power ON IRU A&B|
|04-17-2005||IRU B Y axis fails|
Table 7.2‑1: IRU (gyro) Chronology of Events
FUSE had four reaction wheels to control the attitude and perform slews. Three of the wheels were mounted along the body axes, while the fourth (called “skew”) was mounted at 54.7° to each of the body axes. All of the wheels ultimately experienced failures, with the failure of the fourth wheel leading to the end of the mission. Each wheel suffered two failures: an initial failure after which the wheel could be restarted, and a final failure from which recovery was not possible. Table 7.3‑1 summarizes the major wheel events, listed by wheel.
|1||Yaw (X)||2001:047:13:15||Stopped due to torque checks. Restarted on day 2001:054.|
|2001:329:23:24||Final failure||2||Pitch (Y)||2000:217:10:04||Did not respond to torque commands. Restarted on day 2000:258|
|3||Roll (Z)||2002:351:13:19||Stopped for 2 hours and then autonomously restarted|
|4||Skew (S)||2007:128:16:27||Wheel stopped. Restarted on day 2007:134, stopped again, restarted on day 2007:144.|
: FUSE Reaction Wheel events.
In late 2001, the FUSE project was hit with a near-mission-ending failure of two out of the original four reaction wheel assemblies (RWAs). The yaw wheel failed on Nov. 25, 2001, but normal operations continued with three RWAs. Two weeks later, on Dec. 10, 2001, the pitch RWA also terminated operations. Each wheel had recovered from one previous stiction event. However, the Nov/Dec 2001 events were hard failures and these wheels never worked again. The initial attitude control system required three wheels to maintain three-axis stability for science operations.
After establishing a safe mode for the satellite, the team put its efforts into assessing and implementing a two-wheel control system, using the magnetic torquer bars (MTBs) to replace the missing axis. The gyroless software design was already well underway at this point, but several aspects of the software had to be redesigned to accommodate the interactions of the loss of controlability with the loss of continuous rate information. The safe-modes in particular had to be completely redesigned.
Seven weeks after the loss of the yaw and pitch reaction wheels in November-December 2001, a new ACS control mode was implemented in the flight software to control the satellite in all three axes, using a hybrid of the remaining reaction wheels and the MTBs, acting against the geomagnetic field to control the third axis. A new coordinate system was adopted (SAZ) where the Z-axis is roll, the S-axis is the projection of the skew wheel onto the XY plane, and the A-axis is orthogonal to both the S and Z axis (Figure 3‑1). The S and Z axes were controlled by the remaining wheels (skew and roll) but the A-axis was totally controlled with magnetic torque generated by the MTBs.
On December 27, 2004, telemetry was received indicating that the Roll RWA had stopped. Recovery attempts were unsuccessful. As with the other wheels, this RWA also had one previous stiction event but had been recovered to service for an additional period before the failure. This left only the skew-mounted RWA operational, and once again the ability to perform science operations was threatened. During the first month after this failure, it was difficult to maintain the satellite in a safe configuration while software responses were designed and implemented.
The loss of the roll wheel necessitated another software overhaul, beginning with the need for a new safe pointing mode for the satellite. The control authority with a single wheel was so limited that the satellite would easily fall into regimes where the attitude could not be controlled and a slew to the orbit pole was not possible. The orbit pole itself was often only marginally stable, and the need to use the MTBs to control two axes often resulted in loading the remaining skew wheel with excess momentum even when the gravity gradient itself was small. A non-inertial safe mode was developed where the satellite would be nadir pointed. This mode was largely gravity-gradient-stabilized: once the satellite had settled into this mode, only small additional control torques were needed to maintain control. Settling into this mode was often difficult, and many refinements of the control gains were made before this process was well-behaved. Management of the solar arrays in this mode was likewise refined several times.
The response to this failure followed the general direction of the two-wheel mode, only now the MTBs needed to control two axes while the remaining (skew) reaction wheel controlled the S-axis. In most circumstances the MTBs could not simultaneously provide pointing control while performing momentum unloading; the one-wheel mode operations workaround demanded the use of proper target sequencing as the principal strategy for managing the wheel momentum – momentum build up at one target could be counterbalanced at a succeeding target by torque in the other direction.
In order to accurately predict the wheel momentum in an MPS and avoid wheel saturations or on-the-fly replanning of the timeline (“momentum interventions”), it became necessary to include dynamic simulation capabilities in the Mission Planning MP tools. The final baselined MP operational tool included the modeling of the on-board controller, the torque distribution algorithm, the unloading algorithm, the slew generation algorithm as well as simulation models for attitude dynamics and kinematics.
After uploading revised ACS software and subsequent modifications based upon on-orbit operations, the observatory was finally declared operational again in November 2005, ~10 months after the roll RWA failed. As a result of project expertise gained during the first two months of difficult operations, significant improvements were seen in observing success and efficiency and many fewer trips into various safe modes were made.
The skew RWA had an initial stoppage on May 8, 2007, but did not appear to fail hard. It was restarted temporarily on May 8, within a few days of the stoppage. After making modifications to the safe mode control scheme, the wheel was restarted and over a two week period the excess friction in the wheel decreased to a level similar to that before the anomaly. Operations picked up again on June 12, 2007.
FUSE was operated in a revised single wheel mode for one month before the final wheel failure on July 12, 2007. After one month of wheel restart attempts (including all four wheels), the operational science mission was terminated by NASA on August 17, 2007. After end-of-mission tests, the satellite was decommissioned on Oct. 17, 2007.
In the original conceptual design for FUSE operations, the ACS held the spacecraft inertially-fixed by applying torques to the RWAs so as to keep the angular rates measured by the gyros at zero. The attitude was determined from the TAMs in the absence of FES data.. The TAM data are noisy, and provide instantaneous information only for the two axes perpendicular to the B-field direction. Over a period of many minutes, however, the satellite moves far enough along the path of the orbit that the orientation of the B-field shifts enough for the Kalman filter to accumulate knowledge about all three axes. As the filtered attitude estimate was updated, the ACS applied control torques accordingly to maintain the commanded attitude. When performing science observations, the data from the FES is used to provide absolute attitude information with sub-arcsecond accuracy. The target acquisition sequence employed with this system is described in Section 6.5.2.
Almost every aspect of the original attitude determination and target acquisition operations flow was redesigned to work in the absence of gyroscopes. The critical design considerations for the new system were as follows:
The ACS, IDS, and FES flight software all required major modifications, the highlights of which are summarized briefly here.
The FES had two major new functional requirements:
The major new functions added to the IDS software to support gyroless operations were:
The major updates to the ACS software were:
As experience was gained with on-orbit gyroless operations, and with the idiosyncracies of two and one-wheel control, the interface between the IDS and ACS was augmented further. The ACS status message was modified to contain information on the status of the Kalman filter, the estimated angular rates, and the control law in use. The FPD command codes were expanded to include setting 5 different slew rates, setting the slew algorithm, enabling/disabling momentum unloading, resetting the Kalman filter covariances, setting the momentum bias for the wheel(s), and selecting control law gains.
The combined failure of reaction wheels as well as gyros was particularly challenging, as uncertainties in the propagated rates occasionally resulted in rates that were too high for the FES to acquire stars. Overall, however, the system did work and provided acceptable observing efficiency. Once guide stars were acquired, the low measurement noise of the FES resulted in better pointing stability than with gyros, for axes controlled by reaction wheels. For axes controlled by MTBs, the low bandwidth of the controller limited performance.
The IDS flight scripts that controlled target acquisitions were completely redesigned for the gyroless system. The previous linear sequence of acquisition steps was replaced with a state machine. Autonomous rules operating in the IDS would monitor both the desired acquisition state and the actual state, and would trigger activation of the appropriate script when necessary. This design allowed for steps to be repeated as necessary, and for the system to respond gracefully if tracking was lost at any point. The ability to reacquire guiding autonomously greatly expanded the regions of the sky accessible for observations, as it was no longer necessary to require that a given spacecraft attitude be perfectly stable 100% of the time.
The basic flow of a target acquisition with the gyroless software was as follows:
1) Complete large slew or exit occultation.
2) Wait for angular rates to settle below ~12 arcsec/sec.
3) Do a fast FES autotrack. ACS uses “1Hz” control law. ‘fast autotrack’ means bin the FES image 4×4 and use 0.4 sec image exposure time to obtain centroids as quickly as possible. Request only two guide stars, allowing the subarray size to be 20×20 pixels, increasing the likelihood that stars could be acquired while the attitude drift rates were high.
4) Wait for ACS to report that the attitude estimate has converged to <2°.
5) If pointing error > 2°, command ACS to do a correction slew to remove the error; go back to step 2.
6) Once error is < 2°, repeat FES autotrack to get an image usable for Star_ID: FES image is binned only 2×2 and full FOV exposure time is 1 sec.
7) Process image, identify stars in image, compute pointing error (dq).
8) If pointing error > 12′, do a dead-reckoning unknown-to-known slew, go back to step 2.
9) If pointing error < 12′, do a guided slew to target Q .
10) Change to guide stars selected by Mission Planning.
11) If doing a moving target, do rendezvous slew and begin moving target track.
12) Proceed with the observation.
If pointing control were lost during an exposure, this acquisition sequence would be triggered automatically. The impact on the science data would be a gap in the midst of a time-tag exposure, and a reduction in the effective exposure time for a histogram exposure. CalFUSE corrections for such instances are discussed in the FUSE Data Handbook 2009.
Papers describing gyroless attitude determination and control with only one or two reactions wheels are listed in Sections 12.3 and 12.4.
Next: Satellite to Target Coordinate Transformation Previous: FUSE Instrument Technical Appendix Table of Contents:Table of Contents