Position Uncertainty Adjustments for Problem Scans

One set of scans with underestimated position errors stems from a PosMan problem identified prior to the start of V3 processing. Unfortunately there was insufficient time to make and test a fix prior to the start of processing. Although a fix was later made and tested it was never implemented due to a freeze on code changes. This problem has been referred to as the "low density" problem or sometimes the "float" problem. As implemented the Martin corrections were used only to determine the set of relative frame positions, which was then tied back to the sky with a linear fit against Tycho-2. When only a few Tycho-2 measurements were available a linear error could creep into the solution. There were a half dozen cases where the number of Tycho-2 stars was deemed totally insufficient and the algorithm fell back to use the USNOA catalog. Figure 1 shows an example of a USNOA reconstructed scan. It plots the differences between scan 980310n095 (reconstructed using USNOA) and an overlapping scan. Even when Tycho-2 was used the "low density" problem could sometimes result in large linear errors, as illustrated for scan 981121n069 in Figure 2. By comparing sfpos files to fpos files a list of 179 scans where such errors approach 0.5 asec at some point in the scan has been generated. Uncertainties for these scans are easily adjusted since sfpos:fpos differences are known.

Another set of scans with underestimated position errors was generated by examining the tails of the distribution of x-scan scale factors. Figure 3 presents an example (980626s050) of such a scan showing differences with UCAC as a function of x-scan position. A list of 29 scans with errors approaching 0.5 asec at the edges was generated.

During pipeline processing scans determined to be questionable by the QA were examined in detail and two lists were maintained. The first list (qa_fix) was of scans where reconstruction errors greater than 0.5 asec were sustained for over half a degree of Declination. A total of 10 scans from this list made it into the catalog. Previously mentioned scans 980310n095 and 981121n069 are also members of this list. The second list (qa_sig) was of scans where reconstruction errors appeared to exceed 0.5 asec for less than half a degree of Declination or had a sustained error greater then 0.33 asec. The first was frequently due to rapid error growth near one of the scan ends, such as scan 990113s087 shown in Figure 4 in comparison to the UCAC. A second example from this list is scan 980918s123, shown in Figure 5 with in-scan differences w.r.t. UCAC.

Previous analysis of the Polar1 test field has shown a spike in errors very near the north pole. There are 311 affected scans. The south pole very likely has the same problem which would add another 311 scans.

II. Current Status of Uncertainty Updates: