Scans Containing Some Large Frame Steps
H.L. McCallon 04-22-01
The recent discovery of a pattern-mismatch scan (000116n058) containing a very large frame-to-frame step prompted a search through the fpos files for other large steps. X-scan and in-scan mean step sizes were computed for each scan and deviations (dx and dy) from those means noted. A measure of the total deviation dr [sqrt(dx*dx+dy*dy)] at each step was computed and the scans were ranked in decreasing order of maximum dr within each scan. Making a somewhat arbitrary cut at dr>2.0" yields a list of 41 scans (Big-Step Table) from the v3 candidate list. Each scan in the table is ranked, dx, dy and dr values are tabulated and the two frames associated with the step are identified (fr1-fr2). The RA and Dec of the second frame center are tabulated as well.
I. Real or Reconstruction Problem:
In addition a tentative classification of the source of the "problem" (Clas) is given. Knowing the source is the first step toward determining the impact, as well as what (if anything) can (or should) be done. The classification is based on a comparison of the step histories from the fest and fpos files. The fest file is generated in PFPrep using a different and much simpler algorithm than the one used by PosFrm to generate the fpos file. Since it uses only frame-to-frame data and doesn't attempt to reconcile reference stars, very small errors at each step build over the length of a scan. On the flip side, this also makes it less likely to throw in a spurious step change. The assumption is that if the fest file shows essentially the same step history as the fpos file, that history is very likely "real" and the reconstruction is probably ok. If the step history from the fest looks normal but large steps show up in the fpos, it is likely due to a reconstruction problem and is tagged "recon". When the fest shows one or more large steps but the fpos doesn't show good agreement across the scan, that scan is tagged "both". The idea here being that there is likely real motion of an unusual nature for which the reconstruction is questionable.
The classification determined from the fest:fpos comparison could be verified by over-plotting the U-Scan positions of single-frame extractions from multiple frames around each point in question. The needed positions are included in the pfep file (output by ReFPOS), but I'm not sure when I'll have the time to do it.
Clicking the date/hem/scan (DatScn) number in the Table displays the associated fest and fpos plots. These are 4-panel plots with the fest x-scan step history plotted in the upper-left quadrant and the fest in-scan history in the lower left. The fpos plots are arranged in the same manner on the right. Thus when comparing the two files, compare across the page. Also keep in mind that all the plots are individually autoscaled.
II. What To Do About It:
Note that the vast majority of the scans
in the table are classified as real (34) or have a real component (3). Of the
4 scans which appear to be solely reconstruction problems, the pattern-mismatch
scan (000116n058)
has been rerun with a successful pattern match. It's interesting to note
that in two (
981130n041 and
000113n143) of the three
remaining such scans, the large step comes between the first and second frame.
Not only does this reduce the effect but may provide a clue to the cause.
Assuming the scans in the "real" category are properly classified,
about all that can be done
is to snip out any smeared frames. In all cases, regardless of classification,
it would advisable
to use alternate scans from the v3 candidate list where possible.
As it turns out 18 of the 41 scans have alternates available which are
tabulated in the Alternate Scan Table.
http://spider.ipac.caltech.edu/staff/hlm/2mass/bigstep/bigstep.html
Comments to: Howard McCallon
Last update: 22 April 2001