From jwf@ipac.caltech.eduSat Mar 29 10:01:19 1997
Date: Fri, 28 Mar 1997 13:07:54 -0800
Subject: WG Mtg #119 Minutes

           IPAC 2MASS Working Group Meeting #119 Minutes

Attendees: R. Beck, R. Cutri, T. Evans, J. Fowler, T. Jarrett,
           B. Light, H. McCallon, S. Wheelock, J. White


1.) 2MAPPS Test Results
2.) Scan Direction Definition
3.) Position Reconstruction Biases


1.) 2MAPPS Test Results

    R. Cutri reported on the results of the latest 2MAPPS test runs. On the
positive side, photometric dispersion of the M67 test fields has been
signficantly reduced, and other previously recognized problems (such as those
related to incomplete Read1 merging) appear to be alleviated. On the negative
side, new problems have appeared: some job steps have returned negative error
codes, and BANDMERGE and POSPTS have both terminated abnormally on some MSX
scans. J. Fowler reported that the cause of the BANDMERGE problem is known and
has been fixed, but whether this will affect the POSPTS problem is unknown.
Work will continue on this and the negative return codes.

[Note added in proof: The POSPTS problem was related to a typographical error
in BANDMERGE which has now been fixed and redelivered; the negative return
codes were generated by the unix operating system when the FORTRAN programs
terminated abnormally and were thus unable to set their own return codes;
wrapper scripts will be modified to convert any negative return codes to
positive 4 return codes, indicating failure of the job step.]

2.) Scan Direction Definition

    The observatory proposal to change the scan direction indication from
ascending vs. descending (A/D) to northward vs. southward (N/S) was accepted
by the working group, subject to additional clarification regarding chip
orientation. The proposal was made in order to eliminate ambiguity about what
an "ascending scan" is in the southern hemisphere. The N/S values for the
parameter named "Scan Direction" achieve this as far as celestial direction is
concerned, but not necessarily as far as the motion of object images through
chronologically ordered frames is concerned. The latter phenomenon is equally
important to 2MAPPS processing. If the southern observatory were to use a chip
orientation that results in south at the top of images as currently created
from data frames, the N/S specification alone would not suffice for successful
analysis of the data.
    The group recommendation was to add the following specification: the (0,0)
pixel (as defined by the current readout and data formatting conventions) will
correspond to the southeast corner of the area imaged in the data frame that
begins a scan. This assumes that scans of tiles with a celestial pole at one
end will always start from the end farthest from the pole; if this turns out
not to be true, then the SE corner definition will be defined as applying at
the time of tile center crossing.

3.) Position Reconstruction Biases

    As reported last week, H. McCallon has been investigating certain position
biases. One of these has been resolved: the Read1/Read2-Read1 bias (as
indicated in the note added to last week's minutes). The other bias is not yet
understood; this was described as a scan-to-overlapping-scan bias, but Howard
has discovered that it involves a bias between positions assigned to sources
in a scan at one point in the pipeline and the corresponding positions of
those sources at other points in the pipeline. Since sources are represented
in different coordinate systems at different pipeline points, a fairly general
utility had to be written to perform various coordinate transformations and
source associations.
    Application of this utility has narrowed the range of where the bias is
generated to somewhere after POSFRM and before BANDMERGE, although some
amplification of the bias in the scan direction may take place in BANDMERGE.
Histograms of the bias indicate that in the few scans studied so far, it runs
from about 0.05 to 0.2 arcsec, with a mean of typically about 0.1 arcsec on
both axes. For comparison of POSFRM output to MAPCOR output, the POSFRM
positions are transformed to band-scan coordinates to match MAPCOR's
convention, and the single-band MAPCOR sources are associated with the multi-
band POSFRM sources, and the latter positions are subtracted from the former.
The result is a positive bias in all bands with a mean of typically 0.1 arcsec
on both axes. After BANDMERGE, U-Scan coordinates are used, and both sets of
sources are multi-band; the X bias remains, and the Y bias appears to be
approximately doubled.
    The bias does not appear to depend on scan direction, but more scans need
to be studied before direction dependence can be ruled out. If the bias is as
systematic as it appears, it will be removed by POSPTS when the latter program
is fully implemented. For now, it is simply desirable to understand the origin
of the bias, which may turn out not to be of concern as long as it is
understood and removed effectively by POSPTS.
    Howard's utility program will be used as a diagnostic to detect any sudden
disappearance of the bias, which could happen when a software redelivery is
made for some other purpose. Correlating such a disappearance to the
redelivery should shed light on the origin of the bias if it is not found
sooner. Some investigation will be done to determine whether the bias varies
along the scan.