[daip] Re: CXPOLN and CXCLN: missing linear polarization?

[Bill Cotton]

Uwe,
   I haven't used this combination for several years so I'm unsure as
to what state it's in.  What you report sounds like a bug
(suspiciously close to sqrt(2)) however there are several
possibilities.  CXPOLN generates the real and imaginary parts which in
your test case of symmetric data should be Q and U.  You can compare
them with what IMAGR gives without CLEANing.  Also CXCLN generates Q
and U component lists which can be compared with the results of IMAGR.
My guess is that the problem is in CXPOLN as it's more complicated.
The polarization angle rotation may (or may not) be unrelated.

-Bill

Uwe Bach writes:
 > Dear AIPS group, dear Bill Cotton,
 > 
 > recently I reduced some VSOP observations of 0716+714 at 5GHz in full 
 > polarization together with Thomas Krichbaum at the MPI in Bonn. We did 
 > the polarization calibration (D-terms) following the AIPS Memo by Bill 
 > Cotton with total intensity models and LPCAL.  The d-terms were 
 > comparable to previous observation and were also reproducible on 
 > different calibrators.
 > 
 > The polarization maps of the ground array were produced with IMAGR, but 
 > the Q and U images of the observations including HALCA had to be imaged 
 > using CXPOLN and CXCLN, since HALCA recorded only left circular 
 > polarization. From the comparison of the ground  array only images with 
 > the images including HALCA we found that the core of 0716+714 (which is 
 > very compact) had about 40% less polarized flux in the VSOP images then 
 > in the ground array images and that the E-vectors were rotated by about 
 > 24 deg. This happened in all 3 epochs.
 > 
 > To test if this is due to the higher resolution of the VSOP images we 
 > also imaged the ground array data using CXPOLN and CXCLN. The resulting 
 > maps also contain 40% less linear polarization and show E-vectors 
 > rotated by about 24 deg with respect to the IMAGR maps, although the 
 > number of antennas, resolution and weighting are completely the same. Do 
 > you have any idea where this comes from? Is it a bug in CXPOLN or CXCLN 
 > or is it just due to the fact that a complex clean uses only one of the 
 > cross polarizations (LR or RL)? But the last point might only explain 
 > the rotated E-vector and the higher noise, but not why there is always 
 > less linear polarization. We observed this also in another VSOP 
 > experiment of 0954+62.
 > 
 > Thanks for your help and kind regards
 > Uwe
 > 
 > -- 
 > _________________________________________
 > Uwe Bach
 > INAF - Osservatorio Astronomico di Torino
 > Via Osservatorio 20
 > 10025 Pino Torinese (TO), ITALY
 > http://www.to.astro.it/blazars/bach
 > phone: +39 011 8101912
 > fax: +39 011 8101930