[evlatests] Conjugated VLA data!!!

[Rick Perley]

       The almost-annual 'flux densities' run affords us an opportunity 
to carefully scrutinize VLA performance, and to compare this against 
past behavior. 

    The last VLA flux densities run, taken over the weekend of January 
9/11, has demonstrated a puzzling effect whose origin I have only now 
figured out.  To spare the reader a long discourse, I provide below the 
'executive summary':

Executive Summary:

    The data taken before IAT midnight (5PM local time, Sunday January 
10) are complex conjugated at some bands and some IFs.  The data 
spontaneously reverted to proper phase at IAT midnight that day.  The 
earliest data taken in the run was on Friday evening -- these were 
conjugated.   Hence, I cannot tell when this problem began.   Equally 
puzzling is that on the preceding IAT midnight (Saturday afternoon), no 
change was noted. 
    The bands affected are X, K, Ka, and Q.   Both IF pairs are affected 
at X-band.  Only the BD IFs are affected for K, Ka, and Q band. 
    The problem can be easily corrected using the AIPS program 'PHNEG'.  
But we need to know when the problem began.
    I have no idea of why this has happened. 

Some Details:

    The problem was initially noted in the parallel hand calibration.  
For the bands and IFs noted above, a large R-L phase jump occurred 
exactly at IAT midnight, Jan10/11.  Normally, this is completely handled 
through regular calibration.  But in this case, I noted that the 
bands/IFs affected could not be polarization calibrated unless all 
cross-hand data prior to the noted time were deleted.  Clearly, more was 
afoot than a simple phase change, but as the primary goal of the flux 
densities program is to determine flux densities -- and the amplitudes 
are not affected by the problem, I took the easy way out -- flagging the 
cross-hand data.    I hoped this was the final solution. 

    But it wasn't.  When I got to making the images at X-band, a 
different (but, as you will see, related) problem appeared.  The image 
quality for 3C123 (the only target source large enough to be heavily 
resolved at X-band) was very poor.  Self-calibration could not correct 
the problem.  However, it was quickly found that using only the data 
after the IAT midnight time noted above resulted in perfectly correct 
images. 
    This discovery spurred a close look at the X-band images of the 
point-sources 3C48 and 3C138 -- and it was quickly seen that for these, 
all the background sources had been 'symmetrized' -- ghost images appear 
at positions reflected through the origin.  Separating the data into the 
two times periods before and after the IAT midnight noted above showed 
that each gives a correct image, with the preceding time period showing 
a reflected image of the following time period.
    Comparison of these two images to data taken in 2007 (or any other 
period) showed the data *following* IAT midnight on Jan 10/11 are correct. 
    I used the AIPS program PHNEG to phase conjugate the data prior to 
the time, and added it to that following -- the combined/corrected 
dataset appears correct in all ways. 
    Imaging of the flux density data taken at L, S, and C bands shows 
that there are no problems at either IF.  The problem is restricted to 
the frequencies and IFs listed above.