[evlatests] Closure and polarization

[Barry Clark]

I think that the effect Rick is seeing, of a residual polarization that
sort of wanders around, but mostly just has a fairly constant amplitude
and a phase that goes with parallactic angle, is just a baseline dependent
instrumental polarization on each baseline arising from closure errors.
This is expected behavior if you get the leakage terms a little wrong, which
you will if there are closure errors.

As usual, closure errors probably arise from differences across the band,
so things should improve if we observe in polarization spectral line mode,
and calibrate the leakage as a function of frequency.

If somebody can figure out how to compute and print it, there is an 
instantaneous closure relation for leakage, analogous to that for parallel
hands, that might be interesting to look at.  It is

L1R2 - R1L2 - L1R3 + R1L3 + L2R3 - R2L3 = 0

where each term is the complex cross hand correlator output from the two 
antennas given by the numbers.

(Strong source polarization may mess this up, as there is no guarantee
that the spectral shape of the leakage terms is at all similar to that of 
the source polarization.)