[evlatests] Pointing Gone Wild -- May 2020 Edition

[rperley]

We had two 'dish holography' runs, taken at Ka-band, in May 2020 -- on 
the 18th and the 27th.  Weather conditions on both dates were nearly 
idea.  Both runs used the same source (J1743-0350) and the same HA span 
(-3 to +3 hours).  Both observations were in the dead of night.

Generally, the data quality was extremely good.  But two antennas -- 
ea25 and ea26 -- behaved badly for short periods of time on both nights. 
  Attached are the gain plots for both nights, showing the amplitude 
correction factors needed to return the on-axis amplitudes to the 
nominal value.  So, high values in the plot indicate low amplitudes.  
The x-axis is hour angle.

Referenced pointing was determined about once an hour.  The solutions 
obtained were 'differential', meaning that they are the corrections 
needed referenced to the previous corrections that were applied.

Both plots show the same behavior for both antennas:

* EA26 On both days, at HA ~ -01:45, an elevation pointing error of -24 
arcseconds was detected and corrected.  The effect in the plots is a 
steady rise in the gain values, indicating that the antenna was steadily 
moving off source throughout the preceding hour.  (On May 27, there was 
considerable variability in the gains for the first hour of the 
observation.  This is also likely a pointing problem).  The steady rise 
in the gains (from the increasing offset in the pointing) is easy to see 
in the May 18 plot).  the pointing error at the time it was detected in 
the referenced pointing was -24 arcseconds on May 18, and -38 arcseconds 
on May 27.  Note that this is a single event -- the pointing was stable 
for the rest of the run.

* EA25.  On both days, this antenna developed an enormous 60 arcsecond 
pointing error, within the one-hour window between referenced pointing.  
In this case, the pointing error began at about 00:30 HA, and was 
corrected at HA~ 01:20.  The amplitude of the error was 60 arcsecond on 
May 18, and 66 arcsecond on May 27.  In both cases, the gain solutions 
show the pointing offset grew about linearly over time between the two 
referenced pointing solutions.  One arcminute at Ka-band is close to the 
entire half-width of the antenna -- it was close to the beam null (as 
shown by the amplitudes -- a factor of 7 in amplitude, hence 50 in 
power).

This phenomenon -- of a steadily increasing pointing error between 
referenced pointing solutions is (sadly) not rare.  I have shown the two 
most prominent cases here.  There are a half dozen more from the May 
2020 data -- all of lesser amplitude, but similar form.

In the 2019 data, there are many more examples of this -- all on 
different antennas!  The most egregious then was ea18 -- three 
consecutive occurrences, of -42, 36, and +16 arcseconds -- all in 
elevation.

For all the ~dozen or so occurrences that I matched to the pointing 
results, *all* offsets were in elevation alone.  There was no dependency 
on elevation or azimuth.  This effect can happen on any any antenna, at 
any time, anywhere.  But there does appear there is a tendency to 
repeat.

The amplitude of this effect (notably ea25 in the 2020 data, and on ea18 
in the 2019 data) suggests that this is a software-driven offset -- the 
antennas, for whatever reason, were told to move off source.  But 
perhaps somebody can provide an alternate explanation.

Rick
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ea25+26-BadPnt-27May2020.pdf
Type: application/pdf
Size: 52017 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20200701/6c2beddc/attachment-0002.pdf>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ea25+26-BadPnt-18May2020.pdf
Type: application/pdf
Size: 45664 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20200701/6c2beddc/attachment-0003.pdf>