[evlatests] 3bit PDif Compression -- effects on the visibilities

[Michael Rupen]

I think the real question is whether the _raw visibilities_, with no Pdif
correction, behave identically for the 3- and the 8-bit paths.  That is,
forget about Pdiff; do the 3-bit uv-data track the 8-bit uv-data,
apart from a single antenna-and-IF-dependent scale factor?  The key question
is whether *any* compression/expansion is seen in the raw visibilities.

One approach: run CALIB on 3-bit and on 8-bit for the same subband, and
divide one set of solutions by the other -- is the result constant  for
all sources, elevations, and times?  If that's hard to do, just plot
the CALIB results atop one another.  It's critical to know whether
the visibilities themselves are "clean".

I think this was Ken's suggestion, but it seems a vital test, in deciding
whether the problem is purely one of switched-power measurement, or inherent
in the signal we're measuring.  While I believe you've reported that whatever
effect there is in the visibilities, is much less than in the switched
power data, I am not sure whether that statement is true at the <1% level.

Cheers,

             Michael