[evlatests] UVW incorrect sign convention - AM889 deep integration

[Ed Fomalont]

Rick,

    Check to see if the image of the p-band field or any other extended 
source) is reflected about the origin.    That's what we are seeing with 
the deep Lband observations.

Here is our report

=========================================================================

Neil Miller and I have reduced several deep field observations, code
AH889, over the last month, including the July 1 modcomp-free
observation.  The July 1 data were easy to calibrate, but the deep
field image was strange looking.  First, the position of the sources
over the 20'x20' field did not conform to what is in the sky, via the
NVSS, and to sources in overlapping fields.  Second, the image was
much more w-term distorted even with the image faceting that is
generally used.

Both of these problems are caused by:

THE UVW COORDINATES WRITTEN IN THE EVLA ARCHIVE FILE HAVE SWITCHED
SIGNS.

This would not matter if you also flip the sign of the phase, and
define north and east differently (or something like this), but only
the sign of the uvw coordinates were changed. Here is the evidence:

1.  The listing of the u,v,w in the aips data base of a source at the
same sideral time before and after modcomp-free change-over does show
that the u,v,w have the opposite sign but the phases do not.

2.  Running UVFIX on the July 1 data (recalculates the u,v,w using the
antenna file, the time, and the source location), does indeed flip the
sign of u,v,w, without touching the phase---as it is supposed to.

3.  The comparison of the July 1 image before UVFIX and after UVFIX
shows that:

a.  The image before is an approximate reflection through the phase 
center of the 'after UVFIX' data.

b.  The UVFIX data cleans up sources far from the phase center much
better because the sign error in the w-term does not reflect the
image, but sort of doubles the sky distortion.

CONCLUSION:

1.  Change the signs of U,V,W calculated by on-line system and placed in 
the archive data

2.  Running UVFIX after calibration on the SPLIT files corrects for this 
error.

We are still progressing with careful imaging on this field to see if we 
get down to the expected noise level, and if there are any 
correlator-offset residual errors.

Cheers,  Ed