[evlatests] Some Good WIDAR news

[George Moellenbrock]

Re: image symmetrization.  3C286, L-band, RR, Jul 30.

In CASA, I've imaged a narrow channel range of subband 4 nearer the edge 
(channels 210-220 out of 256, rather than channels 50-220 which I imaged 
previously).  I see no symmetrization effects.  This data has already been 
calibrated with a time-dependent (10s) bandpass using a good resolved 
model, so this result is consistent with Rick's (updated) claim that
self-cal can remove the effect.

For bw=5.5 MHz (11 observed channels), I get an "optimistic dynamic range" 
(i.e., noise measured well off-source) of ~50k:1.  This is roughly the 
factor of 4 worse that I should expect from ~1/16 the bandwidth cf my 
earlier, 85.5 MHz image (220k:1).  If anything, I am doing better with the 
wider bandwidth than with the narrower, i.e., I'm marginally more 
seriously dynamic range limited at the narrower bandwidth.  I don't see 
symmetrization, but perhaps something else untoward is lurking..... 
(e.g., maybe the polarized contribution to the closure errors just is a 
bit worse than average for these channels, or the radially improved 
uv-coverage is paying off a bit, etc.)

I note also that Rick's most recent report of the symmetrization effect 
was for subband 1 of this dataset, which, at a lower frequency, has a 
significantly wider (~25% more) _fractional_ bandwidth.  I wonder if, in 
Rick's reduction, BPASS is calculating the source model at each channel 
separately (freq-dep), or just at the reference (or central) frequency? 
At wider fractional bandwidths, the visibility change with frequency may 
be important at the level of the symmetrization effect?  It is not 
immediately clear to me that this should cause symmetrization of the sort 
Rick has observed....  I note also that I have not done any pre-imaging 
(or pre-calibration) channel averaging to reduce the dataset size, as Rick 
has done.  Depending on how and when it is done, such averaging could 
"lock in" problems....

CASA calculates the model on a per-channel basis, and the model of this 
source varies ~sinusoidally by typically 1 cycle over the 85.5 MHz I 
originally imaged (subband 4), with an amplitude of up to a few % (all 
depending on the baseline---I looked at baseline 1-2 specifically).  (A 
few % is what one would expect from a background source of a few hundred 
mJy near 3C286 at this frequency, and we do indeed have the bandwidth to 
detect the visibility variation....)

-George

--