CXC Processing of Interval after 2001:265 Bright Star Hold

Summary

I have used the CXC aspect pipeline ground processing software to determine the aspect and bias solution for the period immediately following the transition into bright star hold (BSH) on 2001:265.

The CXC solution using both forward and backward (smoothing) kalman filtering gives a lower estimate of pitch bias than the OBC solution. The value of pitch bias at 2001:265:12:00:00 (about 10 minutes after the transition) matches the value derived by Bill Davis using the OFLS and FSS data. The large transient in bias reported by the OBC immediately after transition into BSH appears to be spurious. This transient is reproduced in the CXC processing when only the forward kalman filter is applied.

The time constant for variations in bias estimate for the CXC filter is empirically determined to be around 1500-2000 seconds.

CXC Processing

The CXC aspect pipeline was used to determine an aspect solution and bias estimate for the time period covering 2001:265:12:00:00 to 2001:265:16:00:00. Entry into bright star hold occurred at around 265:11:48. The CXC pipeline was unable to process data before about 265:12:00 because of the large attitude variations caused by the SIM translations. These variations violate a fixed processing limit in the pipeline.

The plot below shows a comparison of the OBC solution for pitch bias versus the CXC solution derived using both forward kalman filtering and backward kalman smoothing.



At the time near 265:12:00, the OBC solution is much larger than the CXC solution. However, the CXC value of around 0.362 arcsec/sec matches almost precisely the last value of pitch bias derived using the OFLS with FSS data during the gyro hold. This agreement implies that the spike in the OBC value is a spurious transient.

Kalman time constant

In order to empirically determine the Kalman filter time constant in the CXC aspect pipeline, I fudged the input gyro data to introduce a step function of 0.1 arcsec/sec in the pitch bias just after 265:14:00. Running just the forward kalman filter results in the bias estimate shown in the plot below.



From visual inspection, the e-folding time constant is in the neighborhood of 1500 to 2000 seconds. It is interesting to note that the CXC forward filter solution reproduces the transient seen in the OBC solution.



Aspect Information main page



Comments or questions: Tom Aldcroft

Last modified:11/02/01