[evlatests] inter-band collimation

[Ken Sowinski]

A fundamental assumption in the VLA pointing model is that the only
difference between bands is in collimation.  A test of this assumption
is the consistency of the pointing offsets measured for each band
during a collimation pointing run.  A recent collimaton pointing run
suggests that this assumption is not far wrong.

Interband collimations are determined by measuring the pointing offset
at each band in turn after having applied the offset measured at X
band for the same source.  This is done for many sources all over the
sky.  If there are changes in pointing between band other than
collimation it would increase the dispersion of offsets measured over
the collection of sources.

Results from a collimation pointing run of last week were examined.
Antenna ea13 was in the barn and ea05 was out of service for a large
part of the time.  About 15 sets of trials were made at each band.
For each antenna, for each band peek provides the average offset
measured over all trials, and the RMS of that set of offsets.  If the
only difference between bands is collimation this RMS should be
representative of the accuracy in pointing offset determination.  I
have averaged the RMS of offsets for all antennas and bands,
independently for each axis.  There is essentially no dependence on
band.  Elevation offsets have slightly more scatter than azimuth.

The table below presents the average of the RMS of the offsets for
each axis and their RSS in arc-seconds.
     avg    rms
Az  1.6    0.4
El  2.1    0.6
RSS 2.6    0.7

These are much smaller than the the RMS of the pointing model
residuals for an X band pointing run and suggests that there is very
little difference in the pointing model between bands.  It does not
mean that the pointing is this good, but only that difference between
the measured X band pointing and the pointing at another band with X
band pointing offsets applied, when looking at the same source is
small.  This result provides an estimate of the goodness of a single
pointing determination of 2.6/sqrt(2) or 1.8 arc-sec RMS.

A few antennas stand out as being significantly worse that the others
by this measure.  The cause must be something operating on small
time and angular scales such as the well known periodic encoder errors.