Last revised 31 May 2001 Revised 10 February 2000 File EPH_ERROR_11535_11542.TXT Ephemeris Error from mid TJD 11535 through mid TJD 11542. TJD 11535 = 23 December 1999 TJD 11542 = 30 December 1999 On the day after the TF28 Flight Software Clock rollover (which happened on TJD 11534), the onboard ephemeris (spacecraft position, etc.) showed anomalous behavior at approximately SoD 55000. In particular, the altitude showed a downward excursion from its expected values for a short interval, perhaps 20 minutes. It was determined that the ephemeris was off by 10 clock ticks (10 x 1.024 s = 5 packets) from mid TJD 11535 through mid TJD 11542 (also at approx. SoD 55000) when the problem was fixed. (The existence of the whole ephemeris problem was first discovered when observed source occultation times were found to be approximately 10 seconds off from the predicted times.) During the interval when the transition from the wrong ephemeris to the correct ephemeris took place on TJD 11542, again perhaps 20 minutes, the altitude showed an upward excursion. The excursions at the beginning and end of the interval when the ephemeris was wrong are larger than what might be expected for an error of 10.24 s; the excursions probably occurred because of the interpolation method used between data points in a table that are very close to 20 minutes apart. (The interpolation supposedly uses tabulated values of position and velocity.) Triggers occuring during the time when the onboard ephemeris (spacecraft position and geocenter direction) was wrong were 7917 (at TJD/SoD = 11537/41182) - 7927 (at TJD/SoD = 11542/47262). The neighboring triggers before and after the time of bad ephemeris were number 7916 at 11535/39979) and number 7928 at 11542/58553. (The IBDB background interval for number 7928 would include part of the time with the bad ephemeris). The Flight Dynamics Facility (FDF) at GSFC has provided files containing tables with the corrected ephemeris (X,Y,Z values) at each possible packet time on each of the eight days involved. A program was written to compute the geocenter direction angles (GRO azimuth and Z-angle) at each packet time using the new correct position values and then to create new files, HKG_DATA.DAT, with the new correct X,Y,Z values and geocenter angles. The new HKG_DATA.DAT files for TJDs 11535-11542 have an audit trail record. The occultation prerequisite program (higher-level TAE procedure PRE_OCC which runs the program PREOCC), the program to generate the small position history file in the daily dataset (higher-level TAE procedure SMALL_POS_APP which runs the program OHEDITOR-APPEND), and the preliminary occultation anylysis programs (higher-level TAE procedure OC_STD running other higher- level TAE procedures OC_SER which runs the program OCSEARCH and OC_FLUX_CONT which runs the program OCFLUX) were repeated. The effect on burst locations, etc., for the triggers involved is probably not known. ----------------- Message from FDF at GSFC attempting to explain ephemeris problem: CGRO onboard ephemerides are computed by interpolation between grid points approximately 20 minutes apart (1172 counts x 1.024s/count = 1200.128s = 20 min, 0.128s). One table spans 72 such intervals, for a total of 86409.22s, 9.22s longer than one day. Operational procedure is to set the start time of each new table 86409.22s later than the start time of the previous table, from a value stored automatically upon each table generation. This procedure has the effect of separating the two tables by one grid point interval, or approximately 20 minutes. Evidently the OBC smoothes the transition between tables by averaging values in the bridging region between the end of the old table and the beginning of the new table. In performing this averaging, it apparently presumes that the same grid (offset and period) is used for both tables, i.e., that the time of the 1st point in the later table is exactly 1172 counts later than the time of the 72nd point in the earlier table. This presumption imposes the time scale (offset and period) of the earlier table on the later, regardless of its actual values. The period was the same (1172 counts) in the two tables, but the offset was perturbed by the clock rollover. For operational considerations, the reference post-rollover grid point time was selected to be the standard 90 minutes (5274 counts = 5400.576s = 90min, 0.576s) after table initialization (selected as rollover time for this event). This reference time established a new grid that was 10 counts out of phase with the previous operational grid. The first post-rollover table functioned perfectly, as was observed during intensive post-rollover monitoring, because its connection with the previous table was broken. However, each subsequent table generated on the previous operational grid was presumed by the OBC to lie on the first table's grid, resulting in a persistent 10-count ephemeris discrepancy. This discrepancy was noted by the BATSE instrument team as a 10-second difference between predicted and observed occultation times. FDF was able to attribute the difference to ephemeris representation by comparing OBC and ground ephemerides. Effectively, the first post-rollover table established a new grid 10 counts out of phase with the operational grid. It is possible to reconcile the two grids by adjusting the table start time. This adjustment has been made in tables prepared on December 29 for uplink and application on December 30.