[evlatests] Summary of T303 (U-X Converter) Switch Problems

[Rick Perley]

     Ephraim has asked for a short summary of our remaining switch 
issues, to help him argue for the funds needed to replace the switches.  
I thought it would be useful to send it to everybody.

     A demanding band switch test was run early in November:  This 
comprised of 28 10-minute observations of a strong calibrator.  In each 
10-minute span, we switched back and forth between two bands, with one 
minute on each band.  There are 28 pairs to test from the 8 Cassegrain 
bands.   The switched power data were examined directly for evidence of 
instability.  In addition, the antenna gains were determined both with, 
and without, the switched power applied, to test the efficacy of the 
switched power in removing gain variations.

     Summary:

     1) All low-frequency bands (L, S, C, and X) showed exemplary 
behavior.  No issues relating to frequency changes were found.

     2) All high-frequency bands showed copious evidence of 
instabilities on many antennas.  The instabilities are much worse on the 
'AC' side.  The instabilities are very large -- *up to 40% in power in 
the worst cases*.  The spread in timescales are broad -- from <1 second 
through minutes.

     3) ea10, which was one of the worst behaving antennas in earlier 
tests, and which has new switches installed in the T303, shows exemplary 
behavior.  Not a single instance of instability was seen.

     Comments:

     Attached are two plots to show the range of phenomena, plus a plot 
for ea10, showing what we should be seeing.  These plots show the 
switched power values, logged at 1 Hz.  The plots of total power are 
identical (except for noise), showing that these variations are indeed 
due to gain instabilities.  Applying the switched power *does indeed* 
'repair' the gain variations -- however, the cost is a significantly 
increased noise in the corrected visibilities, since no smoothing can be 
done with such rapid variability.

     The plots show 15 one-minute 'chunks' of PDif values, all from 
Ku-band data.  In the gaps between the 'chunks', the receivers were at 
the alternate band:  For the first 5, the alternate band was K, for the 
second 5, it was Q, and for the last 5, it was Ku.

    Not all antennas behave as badly as those shown in the accompanying 
plots -- indeed, quite a few, perhaps 1/3 to 1/2, are quite stable.  
(And, I admit that the ones attached are the worst I could find).   
Thus, if the preferred solution of replacing all the switches all at 
once is not tenable, I can provide a ranked list of the worst 
offenders.  (ea07 will certainly be at the top of the list).

     The test is very easy to run and analyze -- and can be considerably 
shortened to do only the high frequency bands -- only six pairs, so one 
hour.

     My personal opinion is that this retrofit is extremely important, 
and should be done as soon as possible.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ku-RCP-BadGain.png
Type: image/png
Size: 39306 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181218/b7936e68/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ku-LCP-BadGain.png
Type: image/png
Size: 43054 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181218/b7936e68/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ku-RCP-Fixed.png
Type: image/png
Size: 48769 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20181218/b7936e68/attachment-0005.png>