[evlatests] Serious Problem with default gain corrections

[George Moellenbrock]

Hi Rick,

Because there has been some uncertainty and confusion about the
gaincurves and their normalization, CASA is not yet automatically
applying them under the hood.   In fact, the latest values are not
yet even available to CASA users through the CASA data repository;
I was waiting for a clearer assertion of their veracity before
making them available.  I think I'll wait just a bit longer...

You say the normalization at the zenith is "ok", but this is exactly
the sort of thing that will cause excessive _antenna-dependence_
in the residual gains, especially at the higher-freq bands.  Basically,
the antennas are probably more alike at middling zenith angles
than at the zenith.   Note that the new gaincurves (EVLA.GAINS
that Bryan supplied last week) are now (unlike old versions)
all normalized at the zenith, including for the subreflector-rotation
trick OFF period.   Normalization at the zenith is more easily
argued as appropriate for the OFF period, which is why that
data worked so well for you, probably.

Assuming that the EVLA.GAINS file is otherwise correct,
I'd bet that the even larger _elevation-dependence_
discrepancy that you note has something to do with the time-
and/or freq- interpolation of the gain coeffs---an
issue Bryan expressed some remaining uncertainty about in his
description of the EVLA.GAINS files last week...

(Incidentally, it strikes me that you wouldn't want to pre-apply
the old gain curves to calculate the new, especially since
the interpolation scheme might hide from you what you
should subtract to yield the new gaincurve coeffs.....)

-George

On Mon, Sep 24, 2012 at 10:55 AM, Rick Perley <rperley at nrao.edu> wrote:

>     We ran on Friday night the script for establishing the elevation
> dependency of the antenna gains.   The weather was perfect -- clear, dry
> (DP ~ 30), calm.  Temperatures were nearly steady near 40F during the
> duration.
>
>     Immediately apparent is that the default gain corrections being
> routinely applied to the data are spectacularly in error.  (For my
> processing, I'm using the OBIT path, but I presume the same gain
> corrections are being applied through CASA also ...)
>
>     The amplitude gain corrections being applied show a huge elevation
> dependency.  They are different for each antenna, but all follow the
> same trend.   All appear to be normalized at the zenith (which is
> o.k.).  Below I give a short table showing the *amplitude* correction
> range (amongst antennas) at an elevation of 12 degrees.  The power
> correction is the square of this value.
>
>     X         1.06 -- 1.12
>     Ku       1.25 -- 1.35
>     K         1.45 -- 1.80
>     Ka       1.40 -- 2.00
>     Q         2.20 -- 5.00  (corresponding to a power correction by a
> factor as large as 25!!!!!)
>
>     These corrections are completely wrong!    Where do they come from?
> It is not from the calculation of the opacity (provided the OBIT
> calculation is the same as the AIPS calculation) -- I regenerated the
> corrections using 'INDXR', turning off the antenna gain correction, and
> leaving in the opacity.  For the Q-band data, the resulting correction
> (at 46 GHz) is 1.5 at elevation = 8 -- a reasonable value.
>     So the problem must lie with the default antenna gain curves.   One
> might suspect the correction being applied is that required for the
> period of time where the 'subreflector trick' was turned off.  But this
> is not the case -- I checked the gain corrections from the Jan 2012 flux
> density run (when the subreflector rotation was disabled) -- at X-band,
> there are no antennas requiring a correction larger than 1%!
>
>     I should be able to generate new corrections by the end of the day.
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20120924/e79247a9/attachment.html>