From jwf@ipac.caltech.eduTue Mar 25 09:38:22 1997
Date: Sun, 2 Mar 1997 14:48:28 -0800
From: jwf@ipac.caltech.edu
To: 2mass@ipac.caltech.edu
Cc: chas@ipac.caltech.edu, sstrom@ipac.caltech.edu,
    stiening@ipac.caltech.edu
Subject: WG Mtg #116 Minutes

              IPAC 2MASS Working Group Meeting #116 Minutes
                                2/25/97

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

AGENDA

1.) Scan Direction Definition
2.) Telescope/Camera Status
3.) Digitized RA
4.) Sources Missed by BFILL in Partial Coadds
5.) Karloff Dedication to DataBase Usage
6.) Missing Four Corners Files
7.) Coadd Noise


DISCUSSION


1.) Scan Direction Definition

    R. Cutri asked whether the definition of ascending scans is that the scan
direction is northward, independent of hemisphere. No one knew whether the
meaning of the term "ascending scan" was different in the southern hemisphere.
J. Fowler pointed out that the IPAC ADDSCAN, FRESCO, and HiRes programs all
use a convention in which positive scan direction is northward, independent of
other considerations (as in 2MASS, IRAS scan directions were never close to
East-West, so no ambiguity arose). R. Cutri accepted the task of determining
whether the project is consistent throughout in its definition of scan
direction.


2.) Telescope/Camera Status

    R. Cutri reported that optical alignment delays have resulted in slipping
first light to late February. Scattered-light fixups are being performed on
the camera. The survey start date is now May 1. Three-channel data are
expected at IPAC around the end of March.


3.) Digitized RA

    L. Fullmer reported that the digitized differences in right ascension have
ceased to be observed. These were caused by carrying RA as a REAL*4 variable
in BFILL (the band-filling module), which resulted in loss of precision such
that when CONMAN computed position differences, digitization at about 0.1
arcsec was observed. The change to REAL*8 was made, and BFILL was redelivered.
[Note: it was later found that the change in the FORTRAN structure used for
point sources in BFILL had caused the compiler to modify its byte-boundary
alignment fixups, yielding a structure size larger than expected, so that
memory allocation fell short, and segmentation faults caused some scans not to
be finished; the REAL*4 version was redelivered pending tests of memory
allocation to determine the minimum acceptable size, so in the meantime,
digitized RA differences will reappear.]


4.) Sources Missed by BFILL in Partial Coadds

    S. Wheelock reported that point sources missing from the *.bfpts files
(the band-filled point sources) in descending scans were no longer missing.
This problem occurred in the last coadd of descending scans, because the
band-filling module BFILL and the main program GALWORKS were using
inconsistent definitions of the coordinate zero points in these partial
coadds. This was repaired in the same redelivery mentioned in the previous
item.


5.) Karloff Dedication to DataBase Usage

    T. Evans reported that the karloff disks need to be dedicated to data base
usage at this time (or phased in over the next few weeks). The team was polled
for information concerning usage of the protocam data that remain on karloff,
and agreement was reached on phasing out such data, which can be restored from
tape onto lugosi as needed.


6) Missing Four Corners Files

    L. Fullmer reported that the four-corners files were missing for some MSX
scans. H. McCallon will investigate the reason for this and provide the needed 
corrections.


7.) Coadd Noise

    G. Kopan reported that in studies conducted with T. Jarrett, it has been
found that the noise in coadded images is about 20% higher than one would
predict from the noise in the individual frames being coadded. It is possible
that the larger masked-pixel dropout rate of the 4x4 Weinberg kernel (as
compared to the previous 3x3 kernel) may contribute to this. G. Kopan will
investigate further. [Note: G. Kopan subsequently found that the frame noise
is frequently very non-Gaussian, causing the individual frame noise to be
underestimated; this may indicate a problem in the frame-flattening method, or
it may be a simple by-product of the hybrid trimming algorithm in DFLAT; in
any case, G. Kopan has augmented the noise estimator so that it is not thrown
off by the presence of this non-Gaussian noise.]