[evlatests] More on noise for mode 'PA'

[Ken Sowinski]

On Mon, 14 Apr 2008, rperley at nrao.edu wrote:

> I then remembered that I had taken data in mode 'PA' for observations of
> 3C273, back in 1997, with the same BW (12.5 MHz) as the recent tests.  I
> have loaded these data, and did a quick calibration.  The result of this
> is that there is *no* difference in the noise, as shown by the histrogram
> distributions in real and imaginary parts, of the four correlations.

I want to confirm our understanding of the symptoms to make sure
we are all chasing the right problem.  I welcome corrections to
the following summary.

It has been seen that in 12.5 MHz line polarization mode PA that
the noise in the parralel habd products is about twice as large
as that in the crossed hand products.  The correlated signals,
when observing a polarized source, are related as expected; that
is there is no factor of two difference.  This can be seen even
if the uncalibrated "raw" data is examined.  It is the parallel
ahnd noise which is too large, rather than the noise in the crossed
hand too small.

We can rule out a scaling error in the system controller because
the calibrated parallel and crossed hand fluxes are what is expected.
Something must be adding noise to the parallel hand correlation
products and it must be happening after the recirculator because
the same recirculator output is feed to the multiplier which produces
A*A and to the multiplier which produces A*C.   That leaves us with
configuring the multiplier quadrants wrongly or something wrong in
conveying the multipler products to the integrator, or something wrong
in the array processer software.

Gustaaf has planned a test observation to verify Ricks's report and
to see how far it extends to other line correlator modes and
bandwidrths.