IRS PCS News and Analysis Page

IRS-900 IER consists of

1.  High accuracy IRS Blue PU on OBS_B.
2.  Switch ATDT_MODE to IRU_ONLY.
3.  Series of CHEAP observations, of the star in the Blue PU aperture.
4.  PCS_ATT_RESET using the Observer, ~30 minutes after the initial PU (after 120 DCEs)
5.  PCS_GYRO_UPDATE and PCS_ATT_RESET, ~15 minutes following the last attitude reset.
6.  Another PCS_GYRO_UPDATE and PCS_ATT_RESET, ~15 minutes after the last gyro update and attitude reset.

 
• Long term drift rates are in the 0.4 - 0.6 milliarcsec/sec range.  See Figure 1 and 2.  These are nominal rates.

Fig.1.  RA (upper) and Dec (lower) with SCLK time in the body frame according to the difference of the attitude according to the STA Observer (P-2201:2204) and the gyro attitude estimate quaternion (P-1000:1003).  Vertical red lines show where the PCS_ATT_RESETs are invoked.  The second two attitude resets are preceded by PCS_GYRO_UPDATE commands.  However, these updates are done while still in IRU_ONLY, and should have been done with intertial holds using the Observer (STA).  IRU_ONLY does not result in valid measurements of the gyro bias rates or misalignment scale factors, but  the radial drift rates over the three legs (indicated between the two panels) are nonetheless nominal, ~0.4,  0.6, and 0.4 milliarcsec/sec.

Fig. 2. Plot by Dan Swanson, showing drifts as the Observer and gyro attitude estimate differences, along the pitch, yaw, and roll axes with SCLK time. This figure and Fig. 1 give the same results for drift rates.  Dan's analysis description of this plot (radial_offset_41500) is here.

The drifts are easily measured from positions of the star in the Blue PU array (centroided by Don Hoard and Harry Teplitz):
 
 

Fig. 3.  Centroided positions of the star on the Blue PU array, in pixels (measured to ~0.1 pix accuracy).  The blue dots are the star positions in the first leg of measurements (120 DCEs), and positions drift from right to left.  Then an attitude reset is done, and the yellow dots indicate the drifting again from right to left at about twice the drift rate (see Fig 1).  A gyro cal update is performed (incorrectly on IRU_ONLY), and the third leg of observations indicated by red dots shows the drift down the array.  Although the gyro cals were not done while attitude was supposed to be controlled by the STA, the drift rates are nominal.	Fig. 4.  The first set of ``science'' DCEs (following the initial two PU exposures), corresponding to the blue dots in Fig. 3, showing the drift with x-pixel, y-pixel location and DCE number.  Outliers are probably centroids skewed by uncorrected cosmic rays or other pixel artifacts.
Figure 5.  Second set of of 60 DCEs, corresponding to yellow dots in Fig. 3.	Fig. 6.  Third and last set of 60 DCEs, corresponding to red dots in Fig. 3.


 
• Short term drift rates are seen, from Figs. 1 and 2, to vary from 3-4 milliarcsec/sec over time intervals of around 100 sec.   Further analaysis of these data show that the drifting is periodic, with a period of about 200 seconds, and a peak to valley amplitude of ~0.1 arcsec.  This manifests itself mainly in the RA (in Fig. 1), corresponding chiefly to pitch (Fig 2.)  This was also seen in gyro attitude data from the IRS_anneal_block_startup sequence IER in IRS Campaign B, where the variability was first seen with the same period and amplitude, manifest then at that pointing chiefly in Dec.

Mark Lacy has followed this up with IRAC data from a slew and settle test in IRAC-R, showing the 200s oscillation:
 

Fig. 7. Power spectrum of y-axis data (~yaw axis) from IRAC-R slew and settle test.  The first peak is an alias, the second is a 207+/-45s period which falls in a gap in the window function and judged to be real.  Plot by M. Lacy.

Mitigation plan: The currently favored explanation for the modulation is related to the IRU controller integral gain settings, which may be higher than needed (for their purpose of compensating the torque imparted to the observatory by the solar wind).  It's suggested by D. Swanson that reducing the gains will reduce or eliminate the noticable modulation, M. Trahn is tasked to determine the best gains, and an onboard configuration parameter update is expected to occur before IRS-P.
 

Incremental offset errors are ~0.4 arcsec, on average, with a spread of ~0.15 arcsec, measured by D. Swanson over IRS-L.  See Fig. 8. Worst-case errors are up to 1 arcsec.  There are, at the moment, no satisfying explanations of this.  STA coefficient bias errors are ruled out, as these are not expected to contribute any more than ~0.16 arcsec (TBC on Oct 31 2003).  A constant error in either the gyro scale factor calibration or the gyro
alignment calibration should produce cumulative errors N times larger than the error on each of N identical moves on gyros, but this is not clearly seen in IRAC data analyzed by J. Stauffer (an IRAC 4 row x 41 column map from the GLIMPSE survey).

It's felt, generally speaking (by Swanson), that IRU calibration has been very accurate since 10/2, but that big errors are sometimes encountered in individual offests over stretches of time.
 

Mitigation Plan:   At the moment, the problem is still in the data analysis stage, hopefully to gain some insight on whether the effects are actual positional errors, or imprecise knowledge handed down to observer and gyro attitude estimate telemetry.  IRS will probably devise an experiement to repetetively move a star between two positions in the PU window on IRU_ONLY.  Normal IRS PU statistics will not be revealing since PUs are done on OBS_B. Keeping the map to two positions will allow repeatibility around the offsets to be derived without dependence on distortion paramters.
 

Settling on gyros after slew is currently still holding at a mean value of 5 sec, and a range of 4-7 sec.  AIRE allows 10 seconds after the SCI or RSCI flag goes to complete (5 sec settle + 5 sec pad), then commences with IRS data collection.

Settling on the observer is ~20 seconds.  This settling is mainly the transient time of OBS_B, used during IRS and PCRS PUs.  The PRI is used to indicate the the S/C is settled to sigma of 0.3 arcsec after the first slew to the PU acquistion spot.  This timing looks nominal.  See Fig. 9.
 
 

Fig. 9. Slew to the IRS PU Blue acquisition and sweet spots, showing slew comlete, PRI ready, and data colletion timing on OBS_B.