[evlatests] Results from Referenced Pointing tests

[Rick Perley]

    A test of referenced pointing was made Wednesday afternoon.  Unlike 
previous tests, which inserted known offsets and examined the subsequent 
solutions, this test utilized the offsets introduced by solar heating 
(and pointing errors in general), and judged the result on whether the 
post-corrected amplitudes were increased by the expected factor. 

    The observations were of the strong calibrator 3C345, at an 
elevation of 50 degrees, from 6 to 7 PM, local time.  The sun set during 
this interval, so changes in elevation pointing can be expected 
throughout the test -- these were certainly seen in many antennas. 

    The test consisted of the following sequence:

    1) Observations of the source at C, X, K, Ka, and Q bands.  This was 
to set the 'no correction' gains. 
    2) Referenced pointing at C-band. 
    3) Observations of the source at the above bands, with the C-band 
corrections applied.
    4) Referenced pointing at X-band.
    5) Observations of the source at the above bands, with the X-band 
corrections applied.
    6) Referenced pointing at C, K, Ka, Q and X bands, with the X-band 
corrections applied -- this to determine the collimations offsets which 
apply at this time/place.
    7) Referenced pointing at K-band.
    8) Observations of the source, at the above bands, with the K-band 
corrrections applied. 

    The overall result is:

    Referenced pointing works, and works well for most antennas.  For 
most antennas, when the derived offsets are significant (over 10 
arcseconds), the application of the offset nearly always increases the 
visibility amplitudes by the expected factor.  Note that, in this test, 
although we can state that referenced pointing nearly always helps, we 
cannot *prove* that the corrections are optimal. 

    There are some notable caveats, and secondary results.

    1) Using C-band for determining the pointing offsets resulted in 
many more 'failures' (no improvement, or even a degradation) than using 
X-band or K-band for the determination.  This was particularly notable 
in antennas 6, 15, and 16, for which large c-band offsets were 
determined (20 arcseconds or more), but for which the post-correction 
amplitudes were either not improved, or worsened.   Simple analysis 
shows that, for a given SNR, the accuracy of the determination is 
proportional to frequency (because the beam is narrower).  Some of the 
'failures to improve' for C-band referencing are likely attributed to 
poor solutions.  I recommend using X-band in the future.   The 'success 
rate' is about the same from the K-band referencing as for X-band -- in 
the future, we should carefully consider whether Ku-band is the best 
choice for the referencing. 
    2) Some antennas don't show the expected improvement at all bands, 
using offsets determined at any of the three bands attempted.   These 
antennas are:  13, 15, 16, 18, and 20.  Some of these may be due to 
collimation offsets, but the evidence is not clear. 
    3) One antenna  (ea 22) had a large, constant offset ( ~ 0.5 
arcminutes) in elevation.  Referenced pointing did a superb job in 
removing this! 
    4) In general, Q-band was the toughest to determine whether the 
procedure is working optimally.  This is no surprise -- other factors 
(in particular, whether the 'subreflector rotation trick' is being 
optimally applied) which influence pointing can easily confuse the issue. 

    Ken has 'tweaked' the collimations, and we are about to measure the 
'subreflector rotation trick' offsets today/tonight.  After these 
improvements are implemented, it might be useful to repeat this run.