[evlatests] 3-bit polarization/stability test

[George Moellenbrock]

Rick-

On Wed, Nov 9, 2011 at 9:27 AM, Rick Perley <rperley at aoc.nrao.edu> wrote:
>   I think we don't have to worry excessively about this.  The tests will be
> done at Ku-band, and we have abundant evidence that the gain variability (in
> outstanding weather, which we know we'll get this evening) at this band is
> very, very low, following correction by the switched power.  (And of course,
> the ionosphere is of no concern at 14 GHz and D configuration).    In any
> event, we have to try.   The polarizers are extremely good at this band and
> frequency (should be better than 2%, and superbly stable), and the source
> polarization is less than .5%.  The devil may well be in the details, but we
> have to look nonetheless ...

You miss my point.  Indeed, the ionosphere is of no concern at Ku-band.
And the tropo weather is unpolarized (so commutes with everything).  The problem
is that applying the net gain calibration (switched power, plus
whatever a standard
'CALIB' yields) as if it were all downstream of the samplers, when, in
fact, some of
the pol-dep gain(t) originates upstream of them,
will substantially complicate your analysis.  What one might conclude
is variability in the
cross-talk in the samplers might just be the upstream gain(pol,t) not
commuting with it.
(I noted ionosphere merely to point out another context where this
sort of thing
causes problems.)   Basically, pol-dep gain(t) (incl phase) upstream
of the polarizing
element _heuristically_ prohibits the assumption of stable leakage
that we like to
make, or equivalently, frustrate attempts to test this assumption.
This is so even
if the leakage is _physically_ stable.  It is not insoluble, and
knowing something
about the upstram gain (sw power) as an isolated factor helps.   That
there is gain
variability occurring _between_  two (or more?) imperfect polarizing
elements--even if they
are physically stable--only further complicates the description of
'net D-terms'.
It is just not a scalar problem, and so not simply factorable in the
traditional manner.

I don't intend to discourage; by all means, take the data!  (Just
don't jump to conclusions.)

-George