[evlatests] Some off-center imaging tests/problems

[Michael Rupen]

Hi Ed --

>      THE CONSECUTIVE FREQUENCY CHANNELS GO DOWN in FREQUENCY!
>      WAS THIS THE INTENDED SETUP?


Yes, this is expected -- X band is lower sideband, hence shows up inverted.
We could change this I suppose but I rather like knowing which sideband
I'm in.


> --- Standard editing, calibration, in AIPS.

Exactly how did you do this?  Rick's results on Friday (see his
evlatests e-mail) suggest that averaging over frequencies leads to
significant errors, so we need to keep track of whether images are
based on single channels, summing over subbands, or whatever.


> --- Image of 0217+738 looks good.  1000:1 dynamic range.  Should be
>     much better, but this is a detail for now.  Normalized to 1 Jy.
>     Since this source was self-caled, this is not surprising.



0217 is actually about 4 Jy.  Just so's you know...


> --- Used EA02 as antenna reference.  There is good agreement between
>     model phases and measured phases for all antennas
>     EXCEPT FOR EA24 and EA28???
>

I think you mean ea25 and ea28, right?

I notice that this offset seems to be ~180 degrees, and that for
ea28 the observed phase seems to be constant...not sure that means anything
but those seem suspiciously "magical" values.


> --- Application of the blue phase to the offset data base produced a
>     good quality image at the phase center.  Hence, there is no
>     indication of closure errors.

What does "good quality" mean?  Are you doing better than the rather
poor 1000:1 noted above?

Note that the wonderful Eric has upgraded CLPLT and added CAPLT for
plotting closure amplitudes :)



>     BUT, EA24 and EA28 have slowly changing large phases offsets for
>     the offset position, but were calibrated to zero for the onaxis
>     source.  So, this is a differential problem.


Again from your plot this is ea25 and ea28, right?


>     a. The model source of perfect data is not as good as the real
>        source model.  Somewhere, the aips task UVMOD is getting
>        somewhat incorrect information on how to move a source 60".
>        Will check antenna file and other methods like uvfix.  However,
>        this might be consistent with a large antenna-offset error,
>        causing the model phases to be in error.  To produce an offset
>        of 180 deg phase for a position change of 60" requires an
>        antenna error of ~1700 wavelengths, or only 60 meters.  Check
>        that the correlator has not changed these two antenna positions.


The correlator does not do anything with the antenna positions; those
flow directly from the Executor to MCAF, and thence to the SDM.  The
corresponding uvw are calculated by the asdm2MS filler.  That filler
agrees with AIPS' UVFIX, and Rick has shown that for the same LST and 
source you get the same uvw using the old path (VLA correlator + IDCAF) 
and from the new (WIDAR correlator + MCAF).  So this is quite an interesting
result.  Could something weird in the header muck up the calculation???


Can you generate a similar model in CASA, and see whether that gives
better results?


Thanks very much for looking at this...lots of work to do!!


                      Michael