[daip] Re: CXPOLN

[Denise Gabuzda]

Dear Uwe,

Good work for finding this problem!

This bug was actually already fixed once back in 1998, when
the first VSOP polarization maps were being made -- then the
input names for the beams were updated from QBEM, UBEM [MX
output] to QBEAM, UBEAM [old IMAGR output]. Is there not indeed
some way to have the procedure automatically detect the correct
class? Failing that, perhaps it would be better to have the
class as an input parameter for the procedure?

Regards,

	Denise


On Wed, 26 Jan 2005, Uwe Bach wrote:

> Dear all,
> 
> following the suggestion by Bill Cotton I looked more deeply what the 
> procedure CXPOLN is doing or better what it is not doing. It seems that 
> at some time in the past the OUTCLASS style for images was changed.
> The procedure CXPOLN is looking for the MX output files QBEAM, UBEAM, 
> QMAP and UMAP, but now the OUTCLASS of dirty images is something like 
> [Stokes]IM001 e.g., QIM001 for Stokes Q and UIM001 for Stokes U and the 
> beam files have QBM001/UBM001.
> 
> I should have noticed that before, because when you run CXPOLN it 
> comments that a file is missing but since all the output files were 
> there I did not care. I just noticed that I have to rename the OUTCLASS 
> of the files, because CXCLN still needs the old style. But due to the 
> wrong file names an important step in CXPOLN was not performed. After 
> the creation of the complex dirty images and beams in Q and U one has to 
> rescale them, so that the reference pixel of the complex beam is 1.0.
> 
> In our experiment these factors were always around 1.4 so that we got 
> always about 40% less linear polarization and also a rotation of the 
> E-vector. I attached a corrected version of the procedure, but I'm not 
> sure if this is now valid in any possible case. I changed the OUTCLASS 
> from the old to the new style, where it was necessary, but I don't know 
> if there are any cases were the input file could be QBM002 instead of 
> QBM001 or so. In that case one has to think about a more complex procedure.
> 
> I inserted also a section at the end where all files are renamed to the 
> old style, since CXCLN needs the files with OUTCLASS: QBEAM, QMAP, UBEAM 
> and UMAP. I checked the procedure with several other experiments, where 
> I have all antennas with full polarization and the CXCLN images are all 
> comparable to the images produced with IMAGR, therefore I'm nearly 
> confident that this procedure works properly now.
> 
> Thanks for all your help and suggestions, kind regards
> Uwe
> 
> P.S. Eric Greisen, please tell me if you need a more detailed or more 
> structured description of what I changed in CXPOLN for the AIPS changes 
> documentation.
> 
>