[daip] PCAL and models

[Eric Greisen]

Martin Hardcastle writes:
 > Hi folks
 > 
 > We have been trying here to do polarization calibration for VLA data
 > where there is poor parallactic angle coverage by calibrators, and
 > have been experimenting with using a model (PMODEL) of a strong
 > polarized source within PCAL.
 > 
 > The documentation for PCAL certainly suggests that this can be done
 > for all solution types:
 > 
 > PCAL
 > Task:  This task reads a UV file, calibrates, subtracts a model and
 >        determines the effective feed parameters for each antenna and
 >        IF.  These parameters are then placed in the antenna (AN)
 >        table.  Polarization corrections can then be applied by setting
 >        DOPOL=1 in LISTR or SPLIT.
 >             If a polarized model is given then only the feed
 >             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 >        parameters are determined.  If no model is given then a point
 >        ^^^^^^^^^^^^^^^^^^^^^^^^^^
 >        source is assumed for SOLTYPE='APPR'.  For SOLTYPEs 'ORI-' and
 >        'RAPR' the source may be resolved but the polarized flux is
 >        assumed to have the same distribution as the total intensity
 >        and the polarization angle is assumed to be constant.
 > 
 > [...]
 > 
 >   PMODEL.....A single component model to be used instead of a
 >              CLEAN components model; if PMODEL(1) > 0 then
 >                                      ^^^^^^^^^^^^^^^^^^^^^
 >              use of this model is requested.
 >              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 > 
 > 
 > ... you'd think from this that the model can be given for any one of
 > the solution types. But in fact the code makes it clear that only
 > 'ORI-' uses any of the inputs to PMODEL (other than Stokes I). If
 > PMODEL is used with SOLTYPE 'APPR' the Stokes I used is supplied but
 > Q, U and V are fixed to zero in the fits, which of course gives
 > seriously incorrect solutions for the feed parameters. Is this a bug
 > in the documentation, or the code?
 > 

Actually subroutine PCLFLX is called for all modes and uses PMODEL
(XSMOD inside the task).  PCLFLX exits abruptly on ORI- and does a lot
more on the other modes.  PCLFLX is called by PCLSEL from PLCUV ahead
of the long if mode ORI- region.  For mode ORI- XSMOD is used within
that region instead.

Having said all this, do not think that I have any idea what the task
is actually doing.  I am including Bill Cotton in the reply because he
may have some memory of this.

I am completely ignorant about polarization - if you find things that
need fixing let me know and I will see hwat I can do.

Eric Greisen