ROSAT Status Report #83

Feb. 2 1994

Note: This Status Report was received as ROSAT News from MPE

SATELLITE STATUS: understanding the problems

In the following we give a detailed summary of the present problems with the ROSAT attitude system, including some history of the trials to overcome them (collected by Martin K"urster, MPE). Although repeating some of the info we sent out earlier, it may help to better understand the context.

What happened

After Christmas and in early January there were a number of occasions during which problems in ROSAT's attitude measurement and control system (AMCS) caused the satellite to drift away from its nominal pointing direction. This drift ended up with the safe mode being triggered. Currently, while the problem is being investigated, ROSAT is remaining in safe-mode.

It soon became clear that the problem is related to the new attitude control strategy that became active after the Z-gyro failure that occurred in November 1993. To remind you of the geometry: ROSAT originally had four gyros to measure rotations around three spatial axes; the X-axis is pointing to the sun, the Z-axis is the telescope pointing direction, and the Y-axis is perpendicular to X and Z; the axis of the fourth (redundant) gyro is inclined with respect to all of the other three axes. The Y-gyro was lost earlier in the mission. With the failure of the Z-gyro the attitude solution must now be derived from the following information.

If the star tracker has identified a star pattern, this information is exclusively used for the attitude solution (stable pointing); this is unchanged.
If no star tracker information is available (during slew, earth block, or when the star pattern cannot be identified) the 3-D attitude solution can no longer be derived from the readings of three gyros (as was the case previously), but must be deduced from the combined readings of the X-gyro and S-gyro, the sun sensor(s), and the magnetometers.

All critical situations occurred when at the same time ROSAT was in the Earth's shadow (i.e. no sun-sensor readings could be used) and also no star tracker information was available (e.g. during an Earth block of the sky region the telescope was pointing to). In these cases, the attitude solution must be derived from the two gyros and the magnetometers. Since the gyro measurements have been used routinely during the whole mission, the error must be related to the magnetometer readings.

In the first instances, drifts occurring during Earth blocks on the night side were typically of the order 25 deg. When ROSAT left the Earth's shadow, the fine sun sensor provided the information that the maximum sun angle (20 deg) was exceeded and triggered the safe mode. At this time it was believed that only slow drifts occurred during the night side, and that the sun angle could not be exceeded by much. As a quick measure to prevent frequent safe modes (which can lead up to almost a day of idle time) the attitude control software was modified in such a way that it allowed for 2.5 min (later 5 min) during which the fine sun sensor could try to move the satellite back to nominal sun angle before triggering safe mode.

This measure did not help when a drift of 110 deg occurred, so that after sunrise, the coarse sun sensor (with a field of view of 45 deg radius) also did not `see' the sun and triggered safe mode. After it was clear that such large drifts can occur, it was decided that ROSAT was not to be recovered from safe mode before a detailed understanding of the problem was gained. Otherwise, it could not be excluded that the telescope would accidentally point to the sun with focal plane instruments (FI) unsecured.

What causes the problem

A detailed analysis of the drift events revealed the following: For operational reasons, magnetometer readings can be used only in intervals of two minutes. In the critical shadow-and-Earth-block situation two spatial coordinates are constantly given by the gyros, but the third component is available only in two-minute intervals. If some kind of external torque is present (see below) the satellite begins to drift around some spatial axis, and for two minutes, only two components of this drift (parallel to the X- and S-gyro) are measured and compensated for. The third drift component (perpendicular to the X-S-plane) is only recognized two minutes later. In some unfortunate situations, the satellite is oriented such that the Earth's magnetic field vector is lying near the plane defined by the axes of the X- and S-gyros. Then the third component is not available from the magnetometers, the drift is not recognized and can continue.

While it was clear from the beginning that these critical geometrical situations can occur, it was not realized that torques strong enough to cause significant drifts in a few minutes are being exerted on the satellite. The source of these torques has now been identified to be the magnetic torquers which are used to prevent the reaction wheels from accumulating too-high rotation rates and saturating. The torquers produce a magnetic field that tends to align (by rotation of the satellite) with the terrestrial field thereby making it possible to slow the rotation rate of the reaction wheels (counter-rotation). To make this discharge of the reaction wheels as smooth as possible, it is performed almost permanently in small portions. Thus, during the 2 min. without magnetic field measurements a number of discharge events can take place. As simulations performed by DASA (the company responsible for the AMCS system) show the strength of these torques is of the right order to produce the observed drifts.

Measures taken to solve the problem

The identification of the source of the disturbances leads to a quite straightforward cure of the problem, i.e. the suppression of magnetic reaction wheel discharge during night time. The capacity of the reaction wheels is sufficient to restrict their discharge to the times of sun presence. However, substantial modifications of the onboard software are required, which need to be tested in a simulator on ground. Also, two days of in-orbit tests with secured FI are foreseen.

A temporary reduced pointing phase

To make use of the time before the new software will become available for use it is planned to circumvent the critical situations (nightly earth blocks) by means of an appropriately devised timeline. Three cases can be distinguished:

During full-Sun orbits (a few days once a month) the Sun sensor can be used together with the gyros so that a regular automatic timeline can be applied.
During times where there is a continuous viewing zone in the sky (also a few days per month), targets located in this zone can be observed, because attitude reference is given without interruption by the star tracker.
At all other times those regions in the sky are identified that do not go into Earth block during the shadow phase. Targets located in these regions will then be identified. Since these regions move on the sky, target changes are foreseen once a day. They will be made during sun presence. As a security against the rare situation that after a slew the star tracker does not manage to lock-on to a star pattern during day time, a change of the FI software is currently being made that automatically protects the FI when the night begins.

This temporary observing program will begin on Wednesday, Feb. 2, 1994, i.e. at the beginning of the next full-sun orbit phase.

Planned return to nominal action

Nominal actions will be resumed when the new AMCS software is available and tested. This is estimated to happen at mid-February.

MISSION PLANNING: what was observed in the Dec 93/Jan 94 period ?

In December 1993 and January 1994, ROSAT was only partially operational. Users that would like to know whether a specific observation in the Dec/Jan timeline was performed, can look to the file dec93-feb94.obs in directory timeline in the ROSAT Service Area. For HRI observations, an entry means that the observation looks normal from the aspect protocol, but the useful data time has not yet been evaluated. For PSPC data, the accepted time, as derived from the SASS processing, is noted in the last column.

US note: the file can also be found on the legacy ftp area under rosat/timelines/dec93-feb94.obs

ROSAT DATA ARCHIVE (RDA): now ONLINE accessible

From February 10 on, all publicly available data sets in the RDA will be accessible online from the ROSAT Service Area via anonymous ftp. The data are stored in compressed form on magnetic discs and therefore, after transfer, have to be decompressed at the user's site. Thus, the network load is minimized.

At this time, data retrieval from UNIX systems is straightforward and easy to do. For retrievals from VMS systems, things are heavier and not yet tested. We will try to find also for VMS users a nice solution.

Layout of the data in the archive is as follows:

Data from a specific observation, identified by its sequence number and an additional identifier (p for PSPC, f for PSPC with filter, and h for HRI data), are contained in a proper subdirectory of the directory archive/data. PSPC data with sequence number 123456, for example, are stored in directory archive/data/123456p.

The archive data have been compressed with the gzip utility that is used by most data/software distribution centers. It is publicly available (no cost) and more effective than e.g. the UNIX compress utility. There are versions for most operating systems including VMS.

If the gzip utility is not already installed on your system, ask your system manager to do so. He can get the software either directly from one of the FTP servers or likewise from the ROSAT Service Area (read file gzip.info in directory archive/tools).

You may, of course, also in future send your archive requests via e-mail to us and receive the data on cassette. This is especially recommended if datasets are very long or the network connections not sufficiently stable. A request form is available in directory archive.

This information, plus an example session for UNIX users, can be found in the file getdata.info in directory archive (a file getdata-vms.info will be placed later there to inform on vms specifics). Please send us your comments on experience with the online data retrieval.

WFC EVENT FILES, gaps in - Problems with Exposure Determination

In the course of statistical investigations of a large sample of WFC sources, we discovered discrepancies up to 10 percent -20 percent in the count rates of different observations of bright non-variable sources. This situation was found to be due to data gaps in the event files, not recorded in the TIM_SEL and TIM_INT descriptors. Further analysis showed that the start and end times of each slot (OBI) as recorded in the TIM_SEL descriptor correctly refer to the times when the WFC was on target and recording events. However, events only were copied into the event files delivered to us from the Rutherford-Appleton-Lab for time periods where a valid attitude from the WFC startracker was available, thus resulting in an inconsistency of the data and the descriptors of these files. We are in contact with the UK Science and Data Centres about this problem. The problem results in an over-estimation of the exposure time and, consequently, an underestimation of the source count rates in the affected observations. A detailed investigation of the event files of 421 observations carried out at the GXUVDC showed that 50 percent of the observations are not affected at all, and that for about 90 percent of the observations, the error in the exposure is less than 10 percent (comparable to the uncertainty in the calibration of the effective areas of the WFC). However, about 1 percent of the observations have errors larger than 30 percent.

The standard analysis software run in Tuebingen has been revised (V4.00 2/94), to correctly treat the problem for all data distributed after January 15.

A preliminary correction has been applied to the exposure maps available online in the public ROSAT/WFC archive. Note that values in the TIM_INT and TIM_SEL descriptors were not changed and that a residual error of the exposure values remains (98 % of the corrected exposure maps have errors less than 10 % and about 90 % have errors less than 5 %). A complete reprocessing of the data in the public online archive will occur at a later time.

There are several different ways for owners of WFC data to check the status of their old data and to work around the problem:

The exposure time (descriptor TIM_INT) will be changed to the correct value by running the EXSAS correction package in the attitude mode. Analysis making use of the EXSAS correction package in attitude mode at all times has resulted in the correct count rates.
PIs may order revised data tapes and standard analysis print outs produced with the new software version from the GXUVDC.
For public observations a corrected exposure value may be obtained by down-loading the corrected exposure map from the ROSAT/WFC archive account. Note the above limitations of the corrected exposure maps at the present time.
Observers interested in timing analysis may check for gaps in the data by binning the event file in short (5 sec) intervals and removing the empty bins. We do suggest, however, to order a revised data set from the GXUVDC in such cases.

EXSAS/MIDAS: OSF/1 implementation available

All institutes that returned the EXSAS Questionnaire (you may also get it from the Service Area, directory exsas) to us should have received the new EXSAS software in the meantime.

A version of MIDAS/EXSAS for the OSF/1 operating system on DEC Alpha computers is now in use at the data center and may be requested from us.