[evlatests] VLA Antenna Gain Stability

[Rick Perley]

     The memo I submitted a couple weeks back lays out a method for 
calibrating the Tcals, from which the antenna efficiencies can be 
measured.  The goal is to enable accurate amplitude calibration of the 
visibilities without the need for an external calibrator. Equally 
important -- if the Tcals are accurate, then our estimates of system 
temperature are also accurate, and reliable estimates of data weights 
can be generated for optimal imaging.

     This is a nice and useful goal, but won't be practical if the Tcal 
and antenna efficiencies aren't stable for long periods of time.  We 
certainly expect them to be stable -- but a demonstration would inspire 
some confidence.

     I have analyzed the gains from three observations made on February 
5, Februay 19 and March 21.  Not a long span, but long enough to get an 
idea how inherently stable the antennas are.  The setups were identical 
for all three.   X-band referenced pointing was used.  Switched power 
values were applied to remove effects of changing gains.  Antenna 
voltage gains were derived from an observation of 3C286 at high 
elevation (very high -- 79, 67 and 85 degrees respectively).  Weather 
was excellent for all three observations.  Two spectral windows ('IFs') 
for each band were used, with frequencies chosen to span each band.

     The mean gain and dispersions (agreed -- 3 values don't give robust 
statistics, but are indicative ...) were generated for each antenna, 
spectral window, and polarization.  I then computed the median gain and 
median dispersion over all antennas, both polarization and both spectral 
windows.  (The median, rather than mean, to minimize the effect of 
outliers, of which there are a few).  The results are quite impressive, 
and better than I had expected:


Band   Med.Avg    Med.Disp

---------------------------------------

L             0.981         0.004

S             0.954         0.005

C            0.957          0.005

X            0.951          0.007

Ku           0.969         0.007

K            0.986          0.016

Ka          0.990          0.026

Q            1.013         0.038

------------------------------------------

     For the lower frequency bands, it seems that better than 1% gain 
accuracy is easily within reach.  The rapid rise in the dispersion at 
high frequencies is exactly as expected for residual pointing errors.

     The slightly less than 1.0 average gain means we are slightly 
over-correcting the raw visibilities with the standard values of Tcal 
and/or efficiency.  Q-band higher than 1.0 gain is likely because we are 
using an optimistically high value of efficiency (-- this is again 
likely a pointing effect -- the efficiencies used were based on 
measurements made by Bob Hayward and me, where we carefully did beam 
cuts through the calibrator source -- something that can't be done in 
standard observing).

     Note that separating Tcal from efficiency is not required for 
accurate pre-calibration of the gains -- we could simply apply the 
necessary corrections to the efficiency alone (or the Tcal). (This 
procedure was employed on the old VLA -- the factor was called the 
'peculiar gain', and was periodically measured and applied).  The 
advantage of the extra effort required to independently determine the 
Tcal and efficiencies is the calculation of the true system temperature.

     Not all antennas have been measured yet.   A few more observations 
over the next few months would be useful in determining the long-term 
stability.