[evlatests] Oct 1, 2005 pointing test

[Ken Sowinski]

Summary of X band pointing run on October 1, 2005.  Barry tracked
down the previously reported clunk to improperly negating the azimuth 
collimation term everytime through the pointing algorithm while over 
the top; the main purpose of this test was to verify that the clunk was
no longer there.  Indeed it is not.  Now that it is gone we can look 
at more subtle problems.


1.  Refraction is still wrong for EVLA antennas.  PEEK can now fit 
for this and found the correction to be about 0.13tan(z) the night of 
September 30.  All other terms were fit along with refraction to avoid
biasing the elevation terms at more modest zenith angles.

2.  I claim to have adjusted the azimuth collimation after the
September 9 pointing run, but the same error of about 1 arc-min was
still there.

3.  Over the top observations made sense once I realized that the
half-power offsets applied to the antenna were not reversed, as is
done by the Modcomps, when over the top.  Teaching PEEK to reverse 
the signs of the measured offsets for EVLA antennas when over the top
allowed good solutions over the entire range of elevations.  There may
still be errors in the sign of the centering terms or perpendicualrity
which are too subtle to see.  We have to decide whether to fix this in
the executor, the program which calculates the pointing offsets, or in
PEEK.

4.  The discrepant tilts found when az and el tilts are allowed to
be independent are still there in both antennas.
                    14(CW5) 16(CE7)
-EW_Az - EW_El    -0.27    0.59
 NS_Az - NS_El    -0.31   -0.03
The sums are almost the same and the phase of the effect differs by
about 135 degrees.  The orientation of the antennas differs by 60
degrees.  Software cannot be excluded but the cause is very unlikely
to be there because the coordinates are transfomed by a single three
dimensional matrix which represents the tilts and the Az encoder
offset.  Both antennas had their azimuth bearings changed before
becoming EVLA antennas but, again, it is hard to see a way for this to
be a mechanical problem.  Since there is only a global tilt in the
EVLA pointing model the best we can do is to split the difference and
make both az and el bad.

5.  Fitting only for the standard 'fixed' terms gives a reasonable
fit with post-fit RMSs comparable to VLA antennas even when including 
over the top measurements.  I will update the Az collimation for 16, 
again, today.