[evlatests] Some Issues Arising from the Gains test

[Rick Perley]

    A few antennas showed significant misbehavior during the 'gains 
test', done Friday night.  Short descriptions follow.

    The run was executed in the following manner:  There was an outer 
loop cycling around the high frequency bands:  X to Ku to K to Ka to Q

    For each band, there were three *different* tunings, the first for 
referenced pointing, and the next two to provide four different 
frequencies.   For example, at X-band:

    Xref -> X1 -> X2

    where 'Xref' was a specific tuning to referenced pointing, and X1 
and X2 were two different tuning setups, to permit better coverage of 
the gain function within the band.  Each tuning provided two spectra 
windows.  This arrangement was used -- rather than utilizing a basic 1 
GHz-wide 8-spectral window setup -- to keep the data volume down. 

    So, on with the issues found:

    1) Zero-mean PDifs.  This means the PDif values were simply noise, 
with near zero mean.  This occured on:
    ea02 at X-band
    ea23 and ea25 at K-band. 

    2) 'Unsteady' PDifs.  This means that each separate observation at a 
given frequency showed a different value of PDif than the others.  The 
culprit is almost certainly the band switches, which are about to be 
replaced.   It has been suggested that the worst offenders be noted, for 
early replacement.  In fact, there are so many antennas/IFs which are 
unsteady that it's not practical to provide a list.  Engineering should 
do the replacements following criteria of their choosing. 

    3) Failures to Tune.  This means that zero-amplitude fringes were 
obtained for some scans for antennas/IFs/tunings that we know are 
working in general.  In all cases, both the RCP and LCP give zero 
amplitude fringes at the same time.  This is an old problem, which (for 
reasons I don't know) I haven't noted in my tests for quite a while.  
But it sure reared its ugly head on this run. 
         a) At X-band, there were numerous 'failures to tune' for the AC 
IF pair, on the first observational scan.  The frequency tuned was 8.150 
GHz (center, with 128 MHz width).  The preceding tuning (X-band 
referenced pointing) was 8.300 GHz.  Six of the 21 separate scans showed 
failures -- each of these had between one and five antennas failing to 
tune correctly. 

    b) At Q-band, there were a very large number of failures, affecting 
both the AC and BD tunings.  The affected antennas were different -- 
meaning that a failure on the AC side was not a predictor of a failure 
on the BD side.  The AC and BD frequencies tuned to were 46.500 and 
40.500 GHz.  The preceding scan was the Q-band referenced pointing scan, 
which used 48.000 and 42.000 GHz, respectively, for the AC and BD 
tunings.  In addition ea27 provided no fringes at all during the run at 
46.5 and 40.5 GHz, but worked fine at the other Q-band tuning (49.5 and 
43.5 GHz). 

    4) ea20 has something seriously wrong with the PDif mechanism, at 
all bands.  There is no connection between variations in the detected 
synchronous power, and the fringe amplitudes as seen in the data.  
Application of the PDifs for this antenna seriously degrade the data.  
The PDif values are constant for any one scan, and have good SNR.  
They're just wrong! 

    5) The Ka-band receiver in ea13 is not working well.  Tsys values 
very high (hundreds of K), and long periods over which zero amplitude 
fringes (and zero PDif as well) were seen.  The problem is not being 
flagged by the system.