[evlatests] Recent L-band Performance (and a check on PDif Compression ...)

[Rick Perley]

     The 'Cygnus A Project' has been extended downwards in frequency to 
include L-band.  We have taken the C and B configuration data for this 
(15 Dec 2018 and 18 Apr 2019).  I just recently got the time to look at 
these datasets.

     The basic summaries are given below:

     1) Gain (Power) Stability.  Simply Excellent!  With only one or two 
exceptions, all antennas remained stable (to better than 1%) in power 
gain throughout the two runs.  There was no need to apply the switched 
power.  (But see below for comments on this).

     2) Phase Stability.  Also excellent -- fluctuations of 5 to 10 
degrees at most, and easily attributable to the atmosphere/ionosphere.

     3) R-L Phase stability.  This one is important for good 
polarization calibration.  In general -- very good, with maximum 
differences of 5 degrees.  However, the deviations, when seen, are not 
random -- they exhibit clear symmetry -- or anti-symmetry -- about the 
time of meridian transit.  I presume these are the same phenomena 
reported by Frank.  The effect is stronger in the B configuration data 
than the C configuration data.  It will be interesting to see what our A 
configuration data show (data date TBD).

     4) Bandpasses.  All good, except for 24R in the C configuration 
data, and 28R and the B configuration data.  For these, the solutions 
showed a very rapid amplitude and phase oscillation (in frequency) with 
period of 2 MHz. (!).  If a reflection, the path length between 
reflections is an unlikely 75 meters.  The oscillatory effect was 
amplified in the cross-polarization (for reasons not understood by me), 
so that all data associated with these IFs had to be flagged.  I'm 
blaming the correlator (for lack of a better idea).

     5) Switched Power.  For the SPWs not littered with RFI (about half 
of them), the data reflect the gain stability noted above -- the PDifs 
are perfectly flat (with only a couple of exceptions, both due to real 
gain changes, as seen in the gain solutions). These data provide me the 
chance to review the 'PDif Compression' problem.

     To review:  The demodulated switched power (PDifs) have always 
excepted a false drop ('compression') when observing a very strong 
source (or observing at a low elevation at Q-band).  The effect is not 
large, but it is present at all bands.  It is only significant for 
observations which more than double the system temperature.

     Cygnus A more than quadruples the system temperature at L-band, so 
provides us a good check on this problem:  About half of the antennas 
show 'compression' by up to 5% (i.e., PDifs are about 0.95 the values 
seen on the calibrators).  Two or three antennas actually show 
'expansion' (PDifs are higher than on cold sky), but here the ratios are 
much smaller -- 1 -- 2%).  The compressions are correlated quite well 
between polarizations, and almost exactly repeat on the two different 
observing days.

     Two or three antennas have much larger 'compression' than the 
others:  ea17 and ea20 -- both near 10% or more.  There is nothing to 
distinguish these two antennas from the others -- their power levels 
were set correctly, and the jump in Tsys when on Cygnus A is the same as 
the other antennas.

     It should be noted here that to prevent any sort of digital 
overflow, the system was told to 'set and remember' while on Cygnus A, 
and not (as is usual) on cold sky.  This resulted in 'PSum' values of 
about 2 to 4 (depending on spectral window), rather than the usual 12 to 
14.

     I can guarantee that the compression is false, since if the 
amplifier gains really had dropped by the amounts suggested by the 
PDifs, the resulting self-calibration gains would have shown them.  They 
did not.  The amplitude gains for all antennas (including ea17 and ea20) 
were within 2% of 1.000 (and most, including ea20 within 1%).  (As 
noted, above, I did not apply the switched power to the data).

6) Antenna cross-polarization.  Some problems here.   Using ea03 as a 
reference (it seems well behaved), most antennas show a cross 
polarization of 2 -- 4%.  Some however, are much, much higher: There 
was, as expected, very high correlation between solutions between the 
two days.  Bad behavior is seen in:  ea06 (up to 12%), ea14 (4 - 10%), 
and ea19.  For this antenna, there was poor reproduction of the 
solutions between the two dates (the only one).  For the earlier run, 
the cross-polarization was up to 20% at the low frequency end, declining 
to about 8% for most of the band.  The later run did not show the low 
frequency rise.

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

     A final note, dealing with calibration.  The calibrator source 
chosen, J2015+3710, has low polarization (about 1%), but an 
extraordinary RM -- +1040 rad/m^2.

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