[evlatests] Images from WIDAR, C band

[George Moellenbrock]

Vivek-

> Lastly, I did no polarization calibration, and 3C286 is ~10% polarized,
> so I am a bit surprised that split-and-calibrate is able to get within
> a factor 2 of the floor. I expected closure errors from polarization
> but perhaps they do factor by antenna for a short ~1hr observation.

When you say "factor by antenna", presumably you are referring to how well 
the (additive) instrumental polarization contribution to the parallel 
hands factors out _multiplicatively_ as simple gain & bandpass 
calibration....  I would contend that the factorization details are 
probably not the relevant property, though.

Firstly, I wouldn't be so surprised that ignoring the polarization issue 
"worked"---the instr pol effects shouldn't kick in until after another 
factor of at least a few in dynamic range, i.e., considerably deeper than 
the 5000:1 you've observed---UNLESS the D-terms are larger and (more to 
the point) more widely distributed (degree of non-parallelness among 
antennas) than we are used to.  In other words, if the D-terms are large 
and distributed, they could very well be kicking at the 5000-10000:1 level 
(especially considering the highish polarization of this source) via the 
combined effect of tweaking the gain calibration and leaving behind 
non-closing residuals.  Otherwise, the details of how they factor 
shouldn't be important yet.

As regards the parallel hand D-term contribution itself and timescales, 
the relevant timescale for how "well" they factor as gains this way is not 
the length of the observation, but rather the solution interval on the 
gain calibration, which must be sufficiently short to "track" (the 
factorable part of) the variations (which nominally come from the terms 
proportional to the source polarization and which are faster near transit, 
but which may also originate in time-dep cross-talk in the 
electronics...).  And about all you can really count on by depending upon 
gains to soak up D-terms is the recovery of the scale of you calibration 
model; otherwise it is a very speculative "advantage" for the overall 
calibration (unless the feeds are considerably more parallel than pure). 
Note also that a longer observation (with sufficiently _uniform_ 
parallactic angle coverage) will benefit from a tendency of the variable 
additive terms to spin down somewhat.  I.e., a longer observation would 
tend to help, not hurt, the resulting dynamic range and in a way (mostly) 
unrelated to the degree of gain-like factorability of the instrumental 
contribution.

-George

--