[daip] [#SJM-103823]: extra Stokes I columns in uv data

[Eric Greisen]

Lynn D. Matthews wrote:
>> I suspect that "hidden" I pols might be multiple spectral channels, 
>> IFs or ??
> 
> Indeed. PRTUV and UVPRT handle "CHANNEL" differently, so PRTUV was 
> simply printing CHANNEL, plus the next three in this case where there 
> were not multiple Stokes parameters.
> 
> That "clue" eliminated, this leads me back to worrying about a more 
> fundamental problem with the model data, so I don't want to waste your 
> time if that's the case. However, I did generate the PRTTP output you 
> requested (attached) and compared this to the real data, and had one
> question related to this.
> 
> In the case where the correlator phase center and the pointing center 
> are different, should this difference be reflected in the header? I 
> notice that OBSRA and OBSDEC are the same as RA and DEC (but both in the 
> real data and the model).
> 
> A second complication is that this is a heterogenous array (25m and 32m 
> dishes), but I assume that this should have no relevance for the
> creation of the FITS file.
>
Unless you invoke frequency dependent options in IMAGR, the pointing 
center does not matter and I am not sure that there is an agreed upon 
way to transmit that anyway.  AIPS just does it in some way that its 
readers know about.  Also the heterogeneous nature of the array does not 
matter so long as IMAGRPARM(1) is 0.  I would worry that there is 
something wrong with the model data.  I can't think of anything obvious 
that would make a PSF-like but unCleanable feature.  The single Stokes 
is Stokes I which is what IMAGR does by default with an RR/LL data set 
so that should be good.  If you have a good Clean model from the data, 
you could try subtracting it from the model visiblities with UVSUB.
That might show something.

It would probably be wise to pick a single spectral channel and work 
with it rather than 512 channels/models.

Good luck,

Eric Greisen