[evlatests] Double-D term at C-band?

[rperley at nrao.edu]

I reported yesterday on a strong temporal variation in electronic gains at
C-band, most likely due to temperature sensitivity, which is mostly -- but
not completely -- removed by application of switched power.  The basic
effect gives a ~ 7% variation in gain.  After correction, about 1%
variation remains.  Nearly all antennas show this effect, to the levels
quoted.

It proves to be difficult to calibrate more accurately than about 1% in
the face of this variation.  In particular, separating temporal from
elevation effects will require software more sophisticated than what we
have now.  I was able to approximately effect this for the C-band data by
utilizing the night-time hours (during which the gains were fairly stable)
to solve for the calibrator fluxes and elevation effects, after which the
residual temporal changes are easier to identify.

Doing this to the C-band data revealed a most curious effect in the gain
residuals:  These are not randomly distributed, but often show an opposite
pattern between two sources -- e.g. source 'A' being 1% above the mean,
source 'B' being 1% below, for a period of a few hours.  If the sources
are observed altenately, the result is a zig-zag pattern.  The really
interesting thing is the opposite polarizations have opposite residuals: 
so in the example noted above, the RCP residuals are of the sense noted,
while the LCP residuals are of equal and opposite sign!  This is seen in
all antennas, throughout the entire 30 hour run!!

The magnitude of the effect is about right for the 'double-D'
cross-polarization -- which leaks RCP into LCP and vice versa.  With the
cross-polarization currently at 10%, the 'double-D' is expected to show up
at the 1% level.

Note that this effect is not seen in the X-band data -- for which the 'D'
terms are at the 1 -- 3% level.

The opposing sense of this for RCP and LCP (which I would not have
expected offhand) means that plots of Stokes 'I' vs (say) UV distance look
very nice, but individuals plots in RCP or LCP are quite a bit worse.  The
difference of course shows up well in plots of Stokes 'V'.

Further efforts to reduce the 'calibrator run' will have to await my
return from holiday and the AAS, in mid January.