[evlatests] Automated Visibility Gain Calibration -- a test

[Rick Perley]

      Providing we know the antenna efficiencies and the value of the 
switched power, we must be able to automatically correct the observed 
visibility amplitudes to put these onto the correct scale. Software (in 
AIPS) to do this has been in place for a number of years, and the 
results of some basic tests were encouraging. However, it was clear that 
for some antennas the corrections are quite far off (more than 10%), and 
that adjustments to either (or both) of the antenna efficiency and Tcal 
would be needed.

     Utilizing some P-band test data taken a few days ago, I have 
generated a simple regimen for measuring, and adjusting the values of 
the efficiency and Tcal for each antenna and both polarizations, such 
that when a special version of the program 'TYAPL' is applied, the 
visibilities for all sources in the test go to their correct values -- 
to within 1%, for all baselines, without any other calibration.

     The test data were taken of Cygnus A, 3C48, and 3C380.  The 
procedure followed was this:

     1) Use the power ratio between Cygnus A and 3C48 to estimate the 
antenna efficiencies.  This was done on two different databases 
(separated by 5 days) -- the results were very consistent between the 
two.  Note that absent a proper adjustment for the different background 
contributions for each of the sources, and knowledge of the flux density 
for 3C405, this measure of the efficiency will have an uncertain 
offset.  I avoided this by simply assuming the very best antenna (ea09) 
had an efficiency of 0.45 (completely arbitrary), and scaled all the 
other antennas.  Resulting values were nearly all between 0.35 and 
0.40.  A few obvious deviant values were clearly due to unsteady, or 
incorrect total power readings.  If ea09's efficiency is not 0.45, the 
error will show up in the corrections needed for the Tcal values (see 
below).

     2) Utilize a special version of TYAPL that Eric generated for this 
purpose, which reads in a file with these adjusted efficiencies.  (The 
default version simply assumed all antennas had the same efficiency).  
Each visibility is modified by the product of a pair of correction 
values -- one of each antenna/IF/polarization given by:

                      sqrt[Tcal/(PDif.Effic)]

     In this operation, the Tcal values were those provided in the 'CD' 
table.

     3) Utilize the known flux density of one source (3C48 in this case) 
to generate antenna-based calibration values.  Since we have adjusted 
the visibilities for varying antenna efficiencies, any deviation of the 
gains from 1.0 should reflect an error in the value of Tcal which was 
applied.

     4) Determine new values of the Tcal, using the old values and the 
determined gains.  The new value is given by:  Tnew = Told*(voltage 
correction)^2.

     5) Re-apply the modified TYAPL program, using both the varying 
efficiency and modified Tcal values.

     Unsurprisingly, the results of this were 'perfect'.  All 
visibilities went straight to their correct values for all three 
sources.  (This only proves the obvious -- the antenna efficiencies and 
switched power values don't change over short time scales, nor are they 
a function of the source strength).   The real test of this procedure is 
in how stable these values are over long periods of time -- months at 
least.  If (as we both  hope and expect) they are stable -- and that the 
antennas with clearly discrepant values of either efficiency or Tcal are 
sorted out -- an improved measurement and application regimen will need 
to be established.