[evlatests] An old problem, reappeared

[Rick Perley]

    I ran the usual stress test this afternoon, using the 1 hour of free 
dynamic time which followed software (00:30 -- 1:30 LST).  A curious 
problem in the amplitudes -- not seen for many months -- has reappeared. 
The integration time is 0.417 seconds.  All observations at 50 MHz BW, 
except at P-band (3 MHz).

    Characteristics:

    1)  The amplitudes have a rhythmic pattern with a 1.67 second period:

    2)  Antennas 13, 18, 23, and 25 have alternate states (up, down, up, 
down ...)
         Antennas 14, 16, 19, and 21 have three 'down' states and one up 
state.
        Antennas 17 and 26 have one 'down' state and three 'up' states
        Antennas 1, 11, and 24 don't show the effect at all.
        No VLA antenna shows the effect. 

    3) All bands show the same patterns, as described above. 

    4) The amplitude of the effect is 6% in correlation amplitude.   
There is no evidence that the 'up' state has a lower correlation 
coefficient than normal. 

    5) All four IFs in any one antenna behave identically.    The 
patterns are in synchronism. 

    6) There is no effect visible in the phase. 

    7) This effect was not present yesterday morning, during the stress 
test run around 8AM.  It *was* present this afternoon, during tests by 
Ken and me looking for clues on the origin of the 'short shallow drop' 
problem. 

    This has been seen before -- many months ago -- but I cannot recall 
the solution.  Ken will shortly be trolling the [evlatests] listserver 
email for clues. 

    The loss in sensitivity for user observing is relatively minor -- 
about 3% for the EVLA antennas, so about 1.5% for the array overall.  
Not a critical problem -- but one that needs to be understood and 
corrected.