[evlatests] Referenced Pointing Problems -- addendum

[Rick Perley]

     The note I sent around concerning referenced pointing issues was 
incomplete.  After a more careful review, I believe the situation is now 
clear (although the cause is not ...):

     The problems I noted only occur when the referenced pointing cycle 
begins before the antennas are on source -- and they only occur for 
'old-ACU' antennas numbered 18 and higher.

     Some details:

     The tests were done on four sources, two of which transit very 
close to the zenith -- 3C286 and OQ208.  All the noted issues occurred 
when observing these object near meridian transit. (Maximum elevations 
were 85 and 84 degrees, respectively).

     We got into trouble because the OPT consistently underestimated the 
time needed to slew between these sources when both were at these high 
elevations.  I had (I thought) conservatively added extra time, above 
the OPT's estimates, but this was not sufficient.

     So, during the execution of the script, the antennas were up to 15 
seconds late in getting on source (almost equal to one 20-second dwell).

     Conclusions from a more careful review:

     1) There were no failures of any kind when the antennas reached the 
source prior to the beginning of the referenced pointing cycle.

     2) There were no failures when the new ACU, and the old-ACU 
antennas numbered up to 17 reached the source during the first offset 
position in the pointing cycle.  These antennas were located in the 
correct *(+Az) position, the off-source flags were correct, so only true 
on-position data were utilized.

     3) The troubles are all with the old-ACU antennas numbered 18 or 
higher, when the antenna slew to source arrived ~ 10 seconds or more 
late.  In this case, the second raster position (-Az) was a combination 
of the +Az and -Az positions -- the 'old' position (+AZ) was observed 
for about the first 10 seconds of the second raster point, and the 'new' 
(correct) position (-Az) was observed in the second half.  The only 
flagged data were in the middle, corresponding to the motion between 
these two positions.

     Because this situation occurred on the second pointing position, it 
would have been counted as a valid offset, but in fact is an average of 
the two azimuth offsets!  (This presumes that the statement made by 
Bryan, that the first pointing position information is never utilized by 
TelCal, is correct).

     Two more important points:

     1) The visibility data clearly show that the *old ACU* antennas do 
not pass through the center of the beam when traveling between the +Az 
and -Az positions, nor do then when traveling between the +El and -El 
positions!!!  In all cases, the visibilities drop to near zero between 
these pointing positions. The situation with the 'new ACU' antennas is 
less clear -- sometimes the visibilities clearly show the passage 
through the beam center, sometimes there is little change in amplitude 
(as if the antenna travelled the circular route ...).  Very strange.

     2) Although the system flags are quite good w.r.t. antenna 
off-source (as when they are in motion), they do not account for the 2 
-- 5 seconds when the antennas are rocking back and forth after having 
reached the intended offset position.  The data show the over/under 
shoot very clearly!  And as these are not flagged, they are being used 
in the offset pointing solution.  The effect is significant, especially 
with the smoothed response of the new ACU antennas (which do not 
oscillate at the natural 2 Hz rate). Hence, I strong recommend:

     * that the first ~5 seconds of any offset pointing position not be 
utilized in the TelCal solution*.