[evlatests] Vanishing PDif!

[Rick Perley]

    I'm slowly (and painfully) trolling through 1 TB of 'demo science' 
data, taken at L, S, C and X bands, in three 8-hour runs, taken in late 
August and earlier this month.   Two of these runs (Sept 03 and Sept 10) 
were taken at C and X bands. 

    A notable event occurred in the Sep 03 data, at X-band:

    The PDif data abruptly dropped from the nominal value (~200 
millicounts) to nearly zero (about 10 millicounts), about 1/3 of the way 
through the observation.  However -- the PSum data showed no change at 
all.   The noise (scatter) in PDif was the same before and after the 
change.  This happened in both polarizations, and in all subbands.  The 
C-band switched power worked as expected throughout. 
    A subsequent run at the same frequencies was taken Sept 10 -- but 
antenna 5 was out of the array that day ...

    With no change in noise, and no change in total power, it seems that 
the only explanation is that the noise diode dropped its output by a 
factor of about 20. 

    Can anybody suggest why?