[evlatests] A shorter summary of recent imaging experiments

[Rick Perley]

Herein, a short summary of the results of the polarimetric imaging 
experiments.  These are from a 5-hour observation of 3C286, OQ208, 
3C287, and 3C273 done in January, 2019, in C configuration.

*1) Total Intensity. *

     All four sources show the effects of 'non-closing' errors -- errors 
which cannot be factored out by antenna.  The achieved DRs are 15 to 25 
thousand:1 -- not very good.

     The baseline-based solution program 'BLCHN' significantly improves 
the total intensity images.  The improvement factor is related to the 
polarized flux, and flux density, of the source -- near perfect results 
for OQ208 (unpolarized) and 3C287 (3% polarized).  For 3C286 and 3C273 
(both 10% polarized, or more), clear residual remain, especially for 
3C273 (which is by far the strongest source).

     Note that I ran BLCHN to solve for non-changing errors (single 
solution over the entire duration).  It is very risky running the 
program with shorter time solutions, as it will cheerfully give you back 
the solutions required to make your data perfectly match the model you 
gave it.

*2) Polarized intensity*

     BLCHN does nothing for Stokes 'Q' and 'U' images.  These clearly 
show similar imaging errors to those seen in 'I'.  For 3C286, it is also 
clear that a dominant error is that due to a changing R-L phase.  Lesser 
evidence for this is seen in 3C287.

     The program RLCAL finds and removes 'R-L' phase differences. 
Application of this program made significant improvements in the 3C286 
polarimetric images (dec = 30.5), small improvements to 3C287 (dec = 
25.2), and no improvements for 3C273 (dec = 2).

     The R-L corrections found by RLCAL appear to be independent of the 
reference antenna used in the calibration.  This is a most puzzling 
result, as it is inconsistent with the R-L phases found by differencing 
the parallel-hand solutions.

*3) Bottom Line*

     My best images, following application of the best calibration 
practices (including the specialized programs noted above), are far from 
thermal noise.  For 3C273, the final images are a factor of 15 noisier 
than the expected thermal noise (in all Stokes' parameters). (!!!).  
These images are far worse than those I obtained in the 1980s using the 
'old' VLA.  For 3C286, the 'I' image is about 50% noisier than thermal, 
the Q and U are a factor of 4 noisier than thermal.  The higher noise is 
seen everywhere in the images -- not just  in restricted areas (like the 
N-S 'stripe' seen in the 3C273 images) -- although it is worst in these 
areas.

     The `R-L' issue has been an interesting diversion -- but it is not 
the central reason for the poor results for 3C273.  It's quite clear 
that the R-L problem (whatever its origin) can be avoided by not 
observing within ~10 degrees of the zenith.

     I'm personally convinced that the basic problem causing the poor 
imaging results for 3C273 is due to coupling of the Q and U into I (and 
vice versa), and thus results from insufficient/incomplete polarization 
calibration and subsequent imaging.  How best to correct this remains an 
open question.

Rick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20220412/f113988f/attachment.html>