[evlatests] Aborting scripts causes trouble ...

[Rick Perley]

    Barry reported some troubles when aborting a script while in 
2-subarray mode.  This note points out that serious troubles can occur 
in 1-array mode too...

    My standard stress-test has two L-band observations at the beginning 
-- to test if the 'first scan' problem is truly gone -- followed by P, 
C, X, etc. 
    In today's test, antenna 11 was missing from the early data -- it 
turned out that operations forgot to include it -- so I asked them to 
abort the script, add antenna 11, and start it over.  Because of this, I 
got two pairs of L-band observations.  Great oddities were found in all 
four scans ...

    Scan 1:  No VLA antennas, *** except antenna 4 *** gave any fringes 
for the entire 6.3 minutes of this scan.  It's possible the antennas 
were unwrapping, and that for some reason antenna 4 was on the right 
wrap.  The EVLA antennas came to life right on time. 

    Scan 2:  There was 24 seconds of confusion at the beginning 
(essentially, the characteristics of the preceding scan continued, then 
everybody changed), then all 26 working antennas 'sprang to life' 
instantaneously.  The data were fairly stable after this. 

    Scan3:  This was the first scan, following the aborted script.  All 
antennas gave fringes, but the amplitudes were about 3 times higher than 
they should have been, and were very  unstable.  In checking the Tsys, I 
saw that the claimed Tsys values were in the hundreds!  About 10 times 
too high.  Who knows what the system thought it was doing ...

    Scan 4:  This scan followed the one above, and all was well, with 
normal amplitudes and phases for all antennas. 

    In fact, this last scan did have one significant curiosity -- the 
amplitudes on the A/C side *only* were very unstable, with slow (tens of 
seconds) changes by up to a factor of ~50% on most antennas.  Both VLA 
and EVLA were affected.  Examination of the Tsys values showed that 
these were changing in an apparently parallel fashion.  So it seems the 
correlation coefficient was stable, but the synchronous system detected 
gradual changes in system sensitivity, and adjusted the correlator 
values.  None of this happened on the B/D side.  I suppose we can blame 
external influences -- but RFI doesn't behave like this, and I can't 
think of any other mechanism which will do this.