[evlatests] Bad Pointing Behavior during Referenced Pointing

[Rick Perley]

     Alert readers will recall we discovered last summer a strange, new 
feature in referenced pointing, in which a subset of antennas (always 
the high numbered ones) headed far off the beam for the first 10 -- 20 
seconds of a referenced pointing run.

     Some adjustments to timing were made, and a short test in September 
indicated the problem had been fixed.

     But it hasn't, at least not completely.

     The 3C273 science/polarimetry test run, taken last Friday, included 
22 referenced pointing observations.  All provided reasonable solutions, 
and there were none of the failures which led to the discovery of the 
problem noted above.  So the 'fix' put in a couple months ago has 
definitely helped.

     However, closer examination of the data shows we still have a 
residual effect, undoubtedly related to the root cause.

     The best way to see what the antennas are actually doing is to run 
a 'CALIB' on the referenced pointing data.  The resulting gain solutions 
will show the familiar five antenna positions.  The data dump time is 1 
second, and we spend 20 seconds at each position.  I used AIPS, so the 
smallest plotted values refer to the higher (closer to beam center) 
location.

     An example, using four 'new-ACU' antennas is shown in the attached 
figure: 'New-Flags.png'.  These solutions come from the flagged data.  
Except for a modest overshoot in three antennas (near 17:53:22 ), the 
data are excellent.

     It is useful to view the unflagged data from these same four 
antennas -- shown in 'New-NoFlags.png'.  The additional points are all 
lower in amplitude, and hence higher in correlation.  They reflect the 
antenna motion from one offset to the next, when the motion takes the 
antenna beam through the source.

     All 'new-ACU' antennas performed flawlessly.

     Not so the 'old-ACU' antennas.

     Look first at 'Old-Flags.png'.  This shows the gain solutions using 
unflagged data only.  The top two plots (ea12 and ea15) show normal 
behavior for these old systems:  the high gain 'spikes' reflect the 
antennas' overshoot when moving from point to point -- the encoders 
think the antenna is on source, but the antenna's 2 Hz oscillations, 
which last a few seconds, are reflected in the variant gains.

     The bottom two plots show the issue at hand:  For the first offset 
point (centered near 17:52:35), there are *two different values* in the 
same offset position.  Examination of all the plots show that antennas 
numbered less than 18 behave correctly, while all antennas numbered 
higher than 17 behavior badly.

     A better idea of what is going on comes from perusal of the last 
plot:  Old-NoFlags.png.  All antennas show high gain spikes at the 
beginning of the 5-point cycle:  The amplitude of these spikes is large 
-- they tell us the antenna gain is low by a factor of ~ 4 -- or a 
factor of 16 in power!  This is a driven motion -- the antennas 
initially head off away from the beam, but only for about 2 seconds.  
The error is quickly corrected, and the antenna is at the correct 
location about 5 seconds after the start of the cycle.  The preceding 
description is correct for low numbered antennas.

     Antennas 18 and up behave differently:  For these, the antennas 
don't move for 10 seconds, they then move off source for about two 
seconds -- this erroneous motion is quickly corrected, and the antennas 
move to the correct position, about 5 seconds before the end of the 
20-second scan.

     Note that this only happens on the initial 5-point pointing 
raster.  The duration was long enough for a second raster to begin 
(starting at 17:54:03), and for this, all is normal.

     Also note that this errant behavior occurred on about 5 of the 22 
individual referenced pointing runs.

     One more thing to note:  For that first scan with the two different 
values, examination of the amplitudes shows that in every case, the 
amplitude is the same as that of the 5th (and last) raster point!  It 
appears that these antennas are erroneously on the wrong offset position.

-------------------------------------------------------------------------

     What is the effect of this problem?  Happily, it seems to have no 
effect at all.  Bryan looked at the TelCal code, and found that the 
initial pointing position is not used.  (A very wise decision!).

     Because the old antennas oscillate for a few seconds, it would be 
good if the first ~5 seconds of all positions were also not utilized -- 
especially for the 'old-ACU' antennas.  This would prevent the 
'overshoots' (which are not flagged by the system) from sneaking into 
the solution.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: New-NoFlags.png
Type: image/png
Size: 33120 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181107/402677d9/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: New-Flags.png
Type: image/png
Size: 30581 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181107/402677d9/attachment-0005.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Old-Flags.png
Type: image/png
Size: 30387 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181107/402677d9/attachment-0006.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Old-NoFlags.png
Type: image/png
Size: 27345 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181107/402677d9/attachment-0007.png>