[evlatests] K-Band system characteristics

[Rick Perley]

    I have completed the K-band data analysis from the 'flux densities' run.

    Some of the larger issues:

    1) System gain stability. 

    a) Tales from Switched Power: 
       -- The diurnal gain variation is seen strongly at this band:  the 
ratio of the max/min varies from 10 to 18%!  But, as at all other bands, 
application of the 'PDif' values removes the visibility changes to less 
than ~2%. 
      -- The change in attenuator setting accompaning the script being 
stopped and restarted was seen in many antennas -- but always at ~1 dB 
-- (unlike L-band).  As always, application of PDif completely removed 
this change. 
      -- Some antennas show very 'ratty' gains, presumably due to the 
switch issue.  Particularly bad are:  1C, 18C, 20C, and 25A. 

    b) Tales from CALIB gains:
          -- There is a strong indication that 'referenced pointing' is 
not working properly.  It was noted that the elevation dependency fits 
(ELINT) were unexpected poor, with large residuals -- far larger than in 
the past.  Plotting of residual gains following these fits show that 
most antennas have *amplitude* residuals about the mean of 3 to 5% -- 
that's a 5 to 10% gain loss, corresponding to 10 to 20 arcseconds 
residual pointing error!  The RCP and LCP residuals always track, and 
the error is the same for all subbands.  It is larger at 22 GHz than at 
18 GHz.  (I also noted that Ku band was about half the gain loss as 
K-band).  All this is consistent with a residual pointing error of about 
this magnitude. 
    If this instability is truly due to pointing, the effect will be 
much worse in Ka-band, which I'm now calibrating.  Stand by ...

   2) Phase Stability. 

    I have not looked closely at parallel-hand gain stability -- it's 
not to absurdly bad, since raw images are sensible.  A closer look will 
be taken later.
    The R-L phase has a curious diurnal variation, looking very much 
like the system gain effect, noted above.  The amplitude is small -- 
typically 5 degrees, so polarimetry is not greatly affected. 
    Some antennas have very poor R-L stability -- notable are:  ea01, 
ea06, ea23. 

    3) Polarization.

    As at the other bands, the PCAL solutions are lovely, and we have 
very accurate measures of the polarization of the standard calibrators. 
    Antenna polarizations, referenced to antenna 10, are generally very 
good, with 'differential D's' typically 5 percent of less.  There are 
some interesting exceptions:
            a)  Antenna 17 at 18.5 GHz:  D > 6% at low edge of band.
            b) Antennas 20 at 18.5 GHz:  D > 8% in 18.0 -- 19.5 GHz band.
            c) Antenna 27, at 18.0 through 19.0 GHz:  D > 13%  
(!!!!!!!)  This one is by far the worst!
     With the 22.3 -- 23.3 GHz tuning (IF 'AC'), all antennas are less 
than 5%, referenced to antenna 10.