VII. 3-Band Cryo Data Release

3. 3-Band Cryo Data Processing

d. Position Reconstruction

In order to determine whether the optical distortion was significantly affected by the temperature changes that defined the beginning of the 3-Band Cryo Period and continued through secondary- tank cryogen depletion, the data were arranged into consecutive 5-day windows, and distortion calibration was performed on each window separately. It was immediately seen that the W3 and W4 distortion was changing rapidly and continued to change over the entire period. W4 was soon declared unusable and dropped from the processing.

The data windows used to compute the distortion models are tagged below by the day number of their center. We count day numbers from the approximate start of the 3-Band-Cryo period, Julian Date 2455411.0, which is actually about half a day before the official start of 3-Band- Cryo, but in order to get five days of data for the first window, which is centered on day 2, we actually include data from as early as Julian Date 2455410.5.

W1 and W2 appeared to be almost unaffected by the temperature changes, with the sole exception of a brief W1 excursion early on that dissipated quickly and does not appear to be important other than to be noted for the record This phenomenon can be seen in the upper left array corner in the second frame of the animated GIF below (see Figure 5). This frame shows the changes in the distortion for the window centered on day 5 relative to the distortion model in the first window, which is centered on day 2. The unusually large fluctuations are gone in the next frame, which is the window centered on day 8. These early models in W1 used 3-day windows instead of 5-day windows in an attempt to track this excursion.

Attempts to achieve high time resolution were thwarted by insufficient sample size to obtain a model with uncertainties small enough to achieve clear statistical significance. Typical radial uncertainties of the corner distortion for 5-day windows are about 95 milli-arcsec and about 123 milli-arcsec for 3-day windows; for a 1-day window, this becomes about 212 milli-arcsec, making the maximum day-5 excursion of 265 milli-arcsec a 1.25-sigma fluctuation, whereas it is about 2.2-sigma relative to the 3-day model. During the attempts to isolate and characterize the event, large but not reliably reproducible fluctuations were found, leaving the conclusion that something real happened but could not be delineated with the usual accuracy.

In order to determine whether distortion drifted even without temperature variations, the cold mission was re-analyzed using five 5-day periods evenly spaced from in-orbit-checkout to a point shortly before the first indications that the secondary cryogen tank was beginning to lose full effectiveness. The results of this analysis were that only fluctuations on the order of the fitting uncertainties occurred. The availability of a 25-day model for each band, however, was used to advantage for the final reprocessing of the cold-mission-data .

The first set of 5-day-window W3 models showed that the distortion was changing by significant amounts over intervals shorter than five days, so experiments were performed in which window centers were spaced closer together than five days with windows of one, three, and five days width. By "significant amounts", we mean on the order of 100 milli-arcsec. The formal uncertainties for the models that were less than five days showed that model errors on the order of 100 milli-arcsec were to be expected, and anything less than five days of data could not produce a model whose own error was less than the variation over the same period.

The best sampling compromises available at the time were made and used for the production processing of the 3-Band-Cryo data. The windows as defined by JD at start, end, and midpoint are shown in the tables below. The attempt to track the early W1 excursion led to the use of 3-day windows during the early portion, since W1 had sufficient data to allow acceptable fits with just three days of data while avoiding smoothing the excursion more than absolutely necessary. A return to 5-day windows would actually have been acceptable sooner than was done, but this became apparent only after the pipeline reprocessing was well underway. The final set of models used 27 windows for W1, 17 for W2, and 34 for W3.

Meanwhile, analysis of time-dependent distortion continued. This further work showed that in W3, windows straddling integration-time changes were introducing artificial smoothing into the models. To correct this, the windows were redefined so that the epochs of integration-time changes were also window boundaries. The result was that the distortion model parameters were seen to change discontinuously at these boundaries. The resultant models produced slightly better astrometry near those boundaries for the test regions, but it was too late to reprocess all the affected 3-Band-Cryo data with these new time-dependent W3 distortion models. Most of the data were not close enough to an integration-time discontinuity to be affected, but for those that were, the smoothing effect can be seen by comparing the time history of the constant terms in the newer models to those used for production. Figure 1 shows the A_0_0 and B_0_0 SIP coefficients in the windows for the models that were used, and Figure 2 shows the same information for the models that avoided having a window straddle an integration-time boundary. In the latter, the one window that was not five days wide was the one for the 2-second integration time, since this period spanned only 3.141 days.

Figure 1 - SIP Constant Coefficients For the Model Windows Used for Processing

Figure 2 - SIP Constant Coefficients For the Model Windows Not Straddling Integration-Time Boundaries

The first and last distortion models for each band are shown below in Figures 3, 4, 6, 7, 9, and 10 as vector-flow diagrams with a scale factor of 10×. Animated GIFs illustrate the changes in the models as functions of time in Figures 5, 8, and 11. These use scale factors of 500× for W1 and W2 and 50× for W3. In order to make the flow more even, not all windows are used, since the window spacing is highly variable, as may be seen in Figures 1 and 2 above. The 14 windows used for each band are listed below by the 3-Band-Cryo day number of their centers, along with the maximum distortion-change vector length in milli-arcsec. The window day centers are shown in the animations with a green progress bar to aid in following the time development.

Figure 3 -Vector Diagram for First W1 Distortion Model Scaled at 10× (Maximum Vector Length = 2.245 pix)

Figure 4 -Vector Diagram for Last W1 Distortion Model Scaled at 10× (Maximum Vector Length = 2.216 pix)

Figure 5 - W1 Distortion Change During 3-Band Cryo Period Scaled at 500× (click to run animation)

Figure 6 -Vector Diagram for First W2 Distortion Model Scaled at 10× (Maximum Vector Length = 2.350 pix)

Figure 7 -Vector Diagram for Last W2 Distortion Model Scaled at 10× (Maximum Vector Length = 2.337 pix)

Figure 8 - W2 Distortion Change During 3-Band Cryo Period Scaled at 500× (click to run animation)

Figure 9 -Vector Diagram for First W3 Distortion Model Scaled at 10× (Maximum Vector Length = 2.312 pix)

Figure 10 -Vector Diagram for Last W3 Distortion Model Scaled at 10× (Maximum Vector Length = 3.029 pix)

Figure 11 - W3 Distortion Change During 3-Band Cryo Period Scaled at 50× (click to run animation)

It can be seen in the W3 distortion-change animated GIF that the largest jump occurred fairly suddenly around the beginning of the second quarter of the 3-Band-Cryo period. Since this happens to coincide with the 4.4-second and 2.2-second integration times, it seems plausible that there is a connection. This may be due to a small dependence on the scan-mirror position, since the W3 integration time ceased to coincide with the usual full survey range of the scan mirror at this time (unlike that of W1 and W2, which continued using the full scan-mirror range). The same is true of the W3 1.1-second integration period, whose distortion is in fact substantially different from that of the cold mission, but since no further changes in integration time were possible, the distortion remained relatively stable during this part of the 3-Band-Cryo period.

Last update: 2012 July 26