[evlatests] Unexpected pointing errors

[Rick Perley]

     I'm reducing C-band data from a 2-hour run taken August 28.   In 
general, the data are of outstanding quality.   The time of the 
observations was from 2 through 4PM.  The weather at the time was 
remarkably benign for a summer afternoon:  Temperatures 25 C, DP 7C 
(dry!), winds light (3 m/s throughout).

     Gain solutions from the local phase calibrator were made every 10 
minutes.

     These solutions showed that some antennas suffered remarkable drops 
in visibility amplitudes.  The characteristics perfectly match those 
given by pointing errors:  Both polarizations identical, and the 
amplitude of the visibilities scale with frequency in the expected manner.

     Referenced pointing *was not* utilized in this observation.

     What is most interesting here is which antennas were affected: 
Without exception, they are all at the ends of the arms.  Here's a 
breakdown, with the size of the apparent pointing errors:

     ea01    N64    1.35 arcminutes.  Three consecutive scans affected.
     ea03    W48    0.6 arcminutes.  One scan only.
     ea05    E72     0.6 arcminutes.   One scan.
     ea07    N72    0.6 arcminutes     Two consecutive scans.
     ea08    E64     2.0, and 1.35 arcminutes.  (Two individual scans, 
40 minutes apart)
     ea16    W64    1.3 arcminutes.  One scan
     ea18    W72    1.7 arcminutes,  Two consecutive scans.
     ea21    E56     0.6 arcminutes.  One scan
     ea25    W56    0.6 arcminutes.  Three consecutive scans.
     ea28    N56     1.1 arcminutes    One scan.

     The apparent mispointings are all at different times!  There is no 
correlation.
     For those mispointings noted as consecutive, the offsets are not 
the same on sequential calibrator observations.

     All other antennas pointed perfectly,  except ea19 (W40).  This one 
was flagged out for nearly an hour, and when it returned its gains were 
notably different.  For this antenna, the gain changes were different 
between polarizations, (although it did scale with frequency as expected 
for a pointing error).