[evlatests] Deep VLA integration test

[George Moellenbrock]

>    I'll next do a more careful comparison of the (u,v,w) coordinates...
>

Rick seems a bit dubious on this issue, now, but here's some additional
info on this question:

I've done a quick spot check on a WIDAR0 dataset from June 18 that I had 
lying around (C0217_000.54990.72014282407), comparing the u,v,w and 
visibility values/signs in the MS and in the post-UVFITS (by CASA) AIPS 
data.  The comparison is fine, for all baselines, at least to within the 
details of converting baseline lengths in meters to wavelengths.  This 
nominally exonerates the UVFITS conversion.  This is consistent with 
failure to achieve expected dynamic range on this and similar data in 
CASA.  (I have also confirmed that the archive is using the most recent 
version of CASA for the UVFITS conversion, which is what I used today for 
this comparison.)

One potential caveat:  I note that the dataset above, as filled from the 
SDM, has all antennas in their natural order, i.e., 
1,2,3,9,18,19,23,24,25.  (I don't know if this is deliberate in the 
SDM/BDF, or by accident, or if the SDM filler attempts to deliver this.) 
This means there is no potential for errors in "flipping" baselines to 
ensure that for each baseline, the first antenna number is lower than the 
second.  We should confirm that this is the case for Rick's data this week 
because I am not sure that the UVFITS output from CASA guarantees a1 < a2 
(and I forget if I ever understood if AIPS insists on or otherwise assumes 
this).  We should be careful to ensure that any a2 < a1 baselines that 
occur on being handled properly.....

If this caveat is ruled out, then the uvw/phase sign convention problem 
occurs upstream of the MS, i.e., in the SDM filler or before.  Presumably 
it is ths same sort of issue as described for my caveat, especially if 
some attempt is being made to sort the antennas into order.  I.e., _some_ 
baselines are getting flipped without the full and proper sign change on 
uvw and phase.  Note also that UVFIXing data in AIPS could yield
data with correct UVW signs and incorrect phase, if an upstream
baseline flip got the phase wrong....

-George


--