[evlatests] Failures to Tune at X-band

[Rick Perley]

     On November 3, I ran a multi-frequency SB, of duration 10 hours.  
The setup was to quickly cycle around four bands, in this order:

     U -> X -> S -> C

     For each frequency change, there were an observation of the target 
(Cygnus A), and a calibrator.  The four-band cycle was repeated 28 
times.  Hence, with 26 antennas, there were 728 band change commands.

The lowest frequency band (S) utilized the 8-bit samplers.  The other 
three bands the 3-bit samplers.  For these three bands, I employed all 
four IF pairs, adjusting the tuning to provide coverage over the subband 
boundaries.

Upon review of the data, an extraordinary number for non-fringing 
antennas were found -- *only at X-band*.  In other words, upon changing 
from Ku to X band, a much higher than normal number of antennas failed 
to fringe at the new frequency setting.  The number of failures was very 
different amongst the four IF pairs -- hence this failure is not due to 
the antenna mispointing or the subreflector not rotating.  The following 
table shows the number of noted failures, per IF pair:

IFPair        Center Frequency          Number
-------------------------------------------------------
A1/C1                9064 MHz                18
A2/C2               11064                         1
B1/D1                 9128                         3
B2/D2               11128                       64 (!!!)
------------------------------------------------------

None of these failures to fringe were flagged by the on-line system.  
The failures always affected both polarizations.  With one exception 
(which was the only failure noted for A2/C2, so this may be unrelated), 
the failure affected the entire scan.  In many cases, the following 
calibrator observation -- with no change in frequency or band in between 
-- showed the previously affected antenna had properly tuned.  There was 
no obvious tendency for the failures to affect certain antennas, with 
the exception of ea01, which had by far the largest number of failures 
for B2/D2.  The failures were uniformly distributed throughout the run.

Any suggestions for what is going wrong here?