[evlatests] Still more on spectral line noise

[George Moellenbrock]

Rick-

   The constant correlator coefficients indicate that the actual SNR
> remains the same.


By SNR, you mean the SNR of the signals _entering_ the correlator.  Again,
the fractional pol claims support that the _input_ SNR is ok.  The
apparent SNR coming out of the correlator must be less, if, as you say,
the calibration factors are independent of BW but the calibrated noise
is greater for 12.5 MHz.

Note that this cannot be a coherence loss either (like bad delay or delay
tracking), because that would decrease the signal, and you would
only realize the excess noise by scaling with larger calibration factors.
It is also hard for this to be a bandpass effect like effective channel
width
less than optimal, because this would also decrease the corr coeff, and
you would only realize the loss in SNR when you find a larger cal scaling
factor.

Yes, testing a range of integration times could be interesting.  The current
evidence is entirely consistent with only fraction of the bits within the
integration
time actually getting accumulated---except for the PA mode cross-hands.

-George
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20080417/807a46d4/attachment.html>