IRS AOT Revisions Spitzer Gyro Performance C. Grillmair & H. Teplitz IRS Instrument Support Team, Spitzer Science Center 07/23/04 I. Background Two separate runs of the IRS Large Offset Test as well as independent measurements by the IRS GTO team have demonstrated that, given the current performance of the gyros, it is not possible to execute a significant number of existing IRS AORs without severely compromising the science (ie. partially or completely missing the targets). Despite considerable ongoing effort by the OET and the PCS working group, it has thus far not been possible to identify the root cause of the problem. While in-flight experiments are planned to better characterize gyro performance and to test a hypothesis, the IRS instrument support team has been asked to examine procedural changes or work arounds which would preserve IRS science in the event that hardware performance cannot or will not be improved. The scope of any work around depends both qualitatively and quantitatively on the nature of the problem. An improved bias estimation filter by OET may help to reduce the frequency and magnitude of poor bias estimates, and thus the incidence of high drift rates. However, it is not yet known whether such an improvement is possible, or to what extent it will solve the problem. If the problem stems at least partly from a poor A2X scale factor (as has been suggested from time to time), then the least expensive, most prudent solution is to improve the scale factor on board the spacecraft, rather than modify command sequences on the ground to cope with the problem. Neither of these hypothesis is completely consistent with the available data (the maximum bias error quote by J. Tietz is only 1/3 to ½ of the amount necessary to account for the pointing drifts seen in the IRS large offset test, but a simple error in the A2X scale factor is not consistent with all the data either). II. Case 1: Noisy Gyro Bias Estimates If our pointing problems are primarily the product of inaccurate bias estimates and the solution is not to update the bias estimate for extended periods of time, then the work around for IRS is not costly, and indeed more efficient from an observer’s point of view. All IRS high accuracy modes invoke bias estimation (a 120 second inertial hold followed by a PCS_GYRO_UPDATE) every 15 minutes. Medium and low accuracy AOTs do not invoke internal bias estimates – these are inserted into the schedule between AORs.If the pointing problem can be identified with unreliable bias updates, and a less frequent but more well measured bias estimate is the cure, then the changes to IRS AORs would be limited to simply removing all inertial holds and PCS_GYRO_UPDATEs. A separate (and welcome) action would fall on OPST to delete once-per-three hour invocations of gyro updates from standard scheduling protocol. Recoding of AIRE would take XX days. There is a remote possibility that the fix could be incorporated into the S11 delivery. Integration and test would then be accomplished with little additional cost. However, if we have not identified the source of the pointing problem in the next few weeks, it is unlikely that the fix could be included in the S11 build schedule. The next scheduled delivery would be S12, with full-up implementation by Cycle 3. It is not reasonable to keep a large number of high accuracy or cluster AORs on hold for more than a year so we would need to schedule a point build to accommodate the fix. III. Case 2: Pointing Issues Remain Unresolved or Uncorrected This situation has more serious consequences for observing efficiency and development effort. For a high accuracy staring AOR, a single 60 second exposure time (cycles=1) and INC_POINT_B, we find that after the first dither position on the Short-Lo-1 slit: Sigma = SQRT(0.28”^2 + 0.2”^2)+90s *.004”/s = .704” single axis. Here 0.28” is the single axis uncertainty immediately after a peak-up, 0.2” is the “bump” contributed by star tracker bias in the course of every move, 90 seconds is the amount of time spent slewing and then holding on gyros in INC_POINT_B, and 0.004”/s is the magnitude of the uncorrected gyro drift observed in IRS Campaign 9. (Note that this is about half the magnitude of the drift we saw in IRS Campaign 4). Slew times for most AORs are anywhere from 5 to 60 seconds, and we take 10 seconds as a typical time to slew from one slit to another. INC_POINT handoff times are always 80 seconds. The gyro drift is not added in quadrature since its direction does not appear to be random over the course of a single AOR (indeed, the direction is remarkably constant over many successive bias estimates). We see from the above that the pointing uncertainty has spanned the breadth of the HardPoint1 window in a single move from the peak-up array to the first requested slit. Propagating the uncertainties over the standard sequences of dithers and slits, we find: FOV Sigma (arcseconds single axis) SL1-b 1.1 SL2-a 1.5SL2-b 1.8 SH-a 2.2 SH-b 2.6 We are well beyond even a QuickPoint condition in the course of a single, standard, multislit, staring AOR. One might argue that adding the gyro drift in this way isn’t entirely appropriate. Let us assume that we can’t predict the direction of the drift, but that the magnitude will be around 15”/hr = .004”/sec (e.g. IRS Campaign 9). If the drift is along the slit, it is likely to be inconsequential for science. The situation is more serious if the drift is across the slit. If we adopt 1/sqrt (2)*0.004”/sec = .0028”/sec, then the above table becomes: FOV Sigma (arcseconds single axis) SL1-a 0.6 SL1-b 0.9 SL2-a 1.2 SL2-b 1.4 SH-a 1.7 SH-b 2.0 In both cases we are well beyond the upper bound of the high accuracy mode. If we assume the gyro drift case of 4 mas/sec, then we would need to do a peak-up before every dither position to meet the .71 arcsecond Hardpoint-1 ceiling. Given that peak-ups can take up to 3 minutes per invocation, this option does not warrant further discussion. From a command work around perspective, there are two more viable solutions: (1) Frequent Attitude Resets We can insert PCS_ATT_RESETs before or after every telescope movement (i.e. change of target and/or change of dither position or slit). The costs associated with this solution include: a. A 20 second hold before each attitude reset. b. An approximate doubling of the number of telescope moves for a given number