[evlatests] Tsys, Tant, and peculiar Gains

[Rick Perley]

    Another curiosity, of some potential importance...

    From today's data, I have compared the antenna gains from data 
filled two ways:

    a) As correlation coefficients -- data as presented by the correlator
    b) As visibility data seen by the astronomer.  These are correlation 
coefficients which have been adjusted for the system temperature and 
efficiency.  So far as I know, these are the only factors that should 
enter into the computation of the 'peculiar gain' -- the number utilized 
by AIPS to convert the corrrelation coefficient into the visibility. 

    But this is not true.  In making the comparisons, I noted that at 
least two antenna IFs (19A at X-band, and 11B at C-band) had remarkable 
differences in gain between the two data versions.  In both cases, the 
apparent gains changed by about 10%, on timescales of minutes.  For 
antenna 19, the change occured on each side of a referenced pointing 
(and it is clear that the referenced pointing is not the origin, as only 
IF 'A' changed).  For antenna 11B at C-band, the change occurred 
smoothly over 45 seconds of time (and again, no other IF showed this).  
In both cases, the changed levels are in the 'astronomer's version' of 
the data, and no change in correlation coefficient is seen.  So it is 
natural to blame this on an incorrectly computed Tsys. 
       But no change in Tsys is seen in either IF for these antennas!  
The only other factor that I'm aware of that is included in the 
computation of the peculiar gain is antenna efficiency, but this is read 
off a table, and can't change over these timescales. 
     So -- what has happened?  I looked at the 'peculiar gains', and 
sure enough, these do change by the (inverse of the) change in gain.  

    What *actually* goes into the calculation of the 'peculiar gain'?