[evlatests] Comments on Referenced Pointing

[Rick Perley]

    A few notes on the apparent efficacy of referenced pointing, as 
determined from the 'flux densities' run from last December. 

    1) It is quite clear that 'referenced pointing' is very beneficial.  
This is most clearly demonstrated by the apparent flux densities of my 
sources at high frequencies during the ~3 hours when the procedure 
failed (due to some software issue).  This occurred during the first 
afternoon of the run, when is when the offsets due to solar heating are 
maximized.   The same sources were observed at the same times the 
following day (this being a 30-hour run), and the difference in 
amplitudes (or gains, if you prefer) are dramatic.   Virtually all 
antennas were positioned more than 30 arseconds off on that first 
afternoon (amplitudes low by half or more), while nearly all were within 
that range (amplitudes within 20% of the correct value) on the second day. 

    2) But it is also very clear from the dispersion in visibility 
amplitudes (or the factored gain solutions) that there is an apparently 
random offset remaining when the referenced pointing is applied.  This 
'dispersion' in observed amplitudes is clearly a pointing effect, and 
not due to an instrumental effect such as an error in the measurement or 
application of 'PDif', since the scatter increases with frequency in the 
manner expected for pointing. 
    The residual is much larger than it should be.  At 48 GHz, 
approximately 10% of the antennas show, by their gain loss, that they 
are mispointed by more than 15 arseconds -- this statistic is generated 
from observations of the calibrators which are 'self-referenced'.  The 
statistics get worse for those extended objects which needed offset 
sources (a nearby calibrator) for referenced pointing.
    With a single exception, it seems that there is no pattern to which 
antennas have an unusually high fraction of mispointings, nor is there 
any correlation with these mispointings to time of day.  'It just 
happens'.  There are many possible origins:  A poor X-band referencing 
solution; a servo/tracking problem, a gust of wind, an erroneous 
collimation value (less likely, I think), etc.  I'm sure others can 
think up other possibilities. 
   
    3) There is a single prominent exception to the lack of any 
correlations in the poor pointing.  This is antenna 28, for which the 
mis-pointings were larger and far more frequent than for any other 
antenna.  I understand that Ken, in reviewing a recent test, has 
concluded that this antenna has a large error in its collimation value. 

    Making a dramatic improvement in referenced pointing is not likely 
to be easy.  A focussed and organized program will be needed, in my 
opinion.  An eager and motivated volunteer would be most welcome!