Proposed Position Reruns
H.L. McCallon 02-06-02
As per Roc Cutri's request, this URL contains my
e-mail sent out 01/30/02 discussing proposed position reruns. Testing since
the e-mail was written has shown that
simply setting the fpos files back to the sfpos files prior
to PFPost does not fix the scans where the two differ.
Adding a routine to PosPts to compute and
apply the needed sfpos:fpos adjustment was tried and works well.
Roc-
Out of the 88% of the v3 processing I've reviewed to date, 17 scans stand out as
most in need of reprocessing. These scans were selected as systematically
exceeding 0.5 asec errors over >1/2 degree of Dec. Of that number, 7 scans have
useable alternates. Alternates should be selected for the following scans:
000223s020 (329591)
000913n030 (029789)
000925n029 (029789)
000925n030 (029790)
980927s042 (302746)
981231s109 (301358)
991016n029 (029761)
Note that scans 000913n030 and 000925n029 both cover tile 029789. Fortunately
there are 9 other scans which also cover that tile.
I strongly recommend reprocessing 9 of the remaining 10 scans which include:
000119s014 (200435)
000308s050 (204394)
000925s138 (200365)
001022s070 (303126)
980202n117 (012811)
980310n095 (015620)
981024s045 (303126)
981121n069 (012195)
990527n018 (010263)
991015n021 (029748)
Note that scans 001022s070 and 981024s045 both cover tile 303126, meaning we'll
only need to rerun one of the two.
Three of the worst scans (001022s070, 980310n095 & 981024s045) fell back to
USNOA as the reference catalog, resulting in systematic errors as high as 0.75
asec and exceeding 0.5 asec over most of the scan. We should be able to repair
these by setting the fpos files back to the sfpos files just before PFPost, thus
forcing the solution back to the Martin corrections. Ron is running a pipeline
test of this technique.
Scan 981121n069 also departed significantly from the Martin corrections with
similarly bad results. The reason for this is unclear, since it did not drop
back to USNOA. The technique mentioned above should benefit this scan as well.
Scans 990527n018 and 990527n018 suffered large deviations mid scan which appear
to be due to bad Martin corrections. These could be repaired by modifying the
Martin corrections or dropping back to a Tycho-2 reconstruction w/o Martin
corrections.
Scans 000925s138, 000119s014, 000308s050 & 991015n021 all have large (~ 1 asec)
deviations at a scan end. The problem with 000925s138 appears to be due to a
bad Tycho measurement. Removing the measurement and modifying Martin correction
should fix the problem. Scans 000119s014, 000308s050 and 991015n021 have
something in common. The spacing of the Martin corrections (1/2 degree
segments) and the frequency of the residual errors are phased such that further
changes to the corrections will be of little help. Dropping back to a Tycho-2
w/o Martin corrections should provide improvement.
All 10 scans have been restored for stand-alone testing. Should be able to meet
your Feb. 6th deadline to have everything in place.
Expect 1 or 2 more really bad scans to pop up during the remaining 12% to be
reviewed. I should also mention that there are several (~ one dozen) other scans
which fell back to USNOA as the reference catalog. All but one of these have
larger than normal systematic errors. They are typically in the 0.25 to 0.5
asec range. We might want to consider fixing these as well so we can claim the
entire catalog is referenced to Tycho.
Regards,
Howard
http://spider.ipac.caltech.edu/staff/hlm/2mass/posrerun/posrerun.html
Comments to: Howard McCallon
Last update: 06 February 2002