[evlatests] Summary of temperature-gain sensitivity

[Rick Perley]

    The 30 hour flux density run, taken during a day of very large 
temperature swing, as given us our best look at our system gain 
sensitivity to temperature.  Here is a summary:

    All data were calibrated using the switched power.  (This was made 
necessary by the constantly changing attenuators -- up to 4 dB in power 
at many bands).  The residual gain variation found represents the 
sensitivity of the noise diode to temperature. 

    a)  L-band.  The gain accuracy is limited by confusion (the data 
were taken in D configuration).  Most antennas show no perceptible 
24-hour periodicity to an accuracy of 2 to 3%.   A coupld (ea05 and 
ea08, notably) show quite striking variations -- 5% pk-pk in voltage (or 
10% in power).  It is not obvious that these are due to temperature 
sensitivities. 

    b) S, C, and X bands.  No obvious 24-h periodicity in gain, to an 
accuracy of 1% (voltage), on most antennas. 

    c) Ku and K bands.  The 24-h periodicity is easily seen on nearly 
all antennas at both bands.  The range is 2 to 3% variation in voltage 
(double that in power). 

    d) Ka band.  No variation visible to better than 2% voltage on any 
antenna. 

    e) Q-band.  No variation visible to better than 5% voltage.  We are 
definitely limited by residual pointing on all antennas here. 

    So I conclude that we should explore fixes to this at Ku and K bands 
first.