Summary
This report examines the expected performance of the ACA and ground aspect processing after 15 years of mission life. The key factor under consideration is continued degradation of the ACA CCD due to cosmic radiation damage. A model was developed which characterizes the spectrum of pixel damage and can be used to simulate the future CCD dark current map. A dark current distribution (which includes time-dependent flickering pixels) appropriate for 15 years of mission life was applied to "clean" flight ACA data from 1999. These data were reprocessed through the aspect pipeline and compared to the baseline 1999 outputs.We find the following: If the CCD temperature is held at the
current value of -15C, centroiding errors for faint (9-10th mag)
stars will increase by 20 - 50%, with occasional large outliers. The
overall aspect solution will suffer a modest average degradation
(0.03 arcsec RMS in each axis), with worst case errors of 0.05 arcsec
RMS in each axis. If the CCD temperature is lowered by another 5
degrees to -20C, then there is negligible degradation of centroiding
and aspect solution accuracy.
Dark current
The first step in simulating ACA CCD dark current map at 15 years MET is to model the spectrum of "radiation damage events." We compared dark current maps over the 3.71 year period spanning day 1999:223 (when the CCD was relatively pristine) and day 2003:116. It was assumed that each pixel with a delta dark current of greater than 100 e-/sec was damaged by a single event. Statistically, up to 0.6% of events may be double hits, but this is a small enough fraction that it was ignored. The distribution of damage events was well characterized as the sum of two exponentials, as shown in the figure below. The lower panel has been binned heavily to show the distribution of hot pixels above 1000 e-/sec. The fitting was done with CIAO/Sherpa using Cash statistics.
Simulated ACA data
To characterize the effects on centroiding and aspect solution accuracy due to CCD radiation damage at 15 years, we selected 10 short observations taken early in the mission (near 1999-Nov). At this time the CCD was relatively clean and the fraction of warm pixels was less than 1%. For each observation we processed the data through the aspect pipeline to establish the baseline performance. We then used the 15-year dark current map to artificially "age" the data to 15 years MET by adding appropriate dark current. Three ACA CCD temperature cases were considered: -10C, -15C, and -20C. It is important to note that the added dark current data were based on actual flight data which includes the significant effect of flickering pixels. After "aging" the CCD image data, the standard aspect pipeline was run to generate both star centroids and the final aspect solution.Results
Centroiding
Below we show a comparison of the centroiding accuracy as a function of star magnitude for each of the three cases (-10C: red plus signs, -15C: green circles, and -20C: blue crosses). The centroiding accuracies have been normalized by the corresponding values from the baseline run, so a value of 1.0 corresponds to beginning-of-life performance. For both the -10C and -15C cases, most centroid accuracies are within 20-50% of the baseline, but there are substantial numbers of outliers that are up to a factor of 3.6 times less accurate.
Aspect solution accuracy
For ground processing the bottom line is the accuracy of the final aspect solution. For each observation and temperature case we calculated the time-dependent difference from the baseline aspect solution, and then took the RMS (in pitch and yaw) of this difference. This represents an estimate of the RMS aspect solution error (per axis) incurred because of the damaged CCD. The table below lists the minimum, mean, and maximum values for the 10 observations:| Temperature | Min (arcsec) | Mean (arcsec) | Max (arcsec) |
| -10C | 0.015 | 0.044 | 0.100 |
| -15C | 0.014 | 0.028 | 0.050 |
| -20C | 0.010 | 0.015 | 0.024 |
For the sample of 10 observations, the "Max" column gives a very rough estimate of the 90th percentile of expected errors due to CCD damage. This is more appropriate than the mean for evaluating the impact to Chandra image reconstruction accuracy. Flight data indicate that the image reconstruction accuracy is now typically around 0.06 to 0.10 arcsec (RMS per axis). Therefore an additional error of 0.100 arcsec at -10C would noticably degrade the accuracy. This argues that the CCD temperature cannot be raised back to -10C without degrading Chandra science. At -15C, the degradation would likely be noticable only in a small subset of observations, but is nevertheless undesirable. If the CCD is cooled further to -20C, then there is expected to be no appreciable degradation in the image reconstruction accuracy.
PCAD flight control
One remaining analysis is to evaluate the errors in the on-board attitude determination. Since this relies on first moment centroiding (which is much more sensitive to warm/hot pixels) leaving the CCD at -15C may be problematic if large roll and gyro bias errors occur. This analysis is complicated by the need to understand the ACA flight software status flagging (e.g. when will warm/hot pixels trigger bad status flags) and the OBC Kalman filter data screening (at what level are centroids rejected?). It may prove easier to do the analysis to show that it is safe and effective to cool the CCD another 5 degrees to -20C. In this case there are unlikely to be errors at the level which could compromise on-board attitude determination.
Aspect Information main page
Comments or questions: Aspect Help or Tom Aldcroft
Last modified:12/27/13