[evlatests] Cancel Red Alert ...

[George Moellenbrock]

Rick-

>    2) The appallingly poor L-band image that I showed was a result of
> BLCAL not having been applied.  When BLCAL was computed and applied,
> (and the data self-calibrated with my best VLA image), the output image
> immediately shows a spiffy 180,000:1 dynamic range!  Better than the VLA
> image.   But still a long ways from thermal noise.

Do you stand by your earlier statement that the BLCAL solutions were 
"insignificant"?  Presumably not.  So then, are their magnitudes and 
distribution consistent with the few 10000s:1 you get without applying the 
BLCAL solution?  Must be.  And (you knew I would ask!), are they 
consistent with what we might expect for the instrumental polarization 
(~constant <D*D>) contribution at L-band? (I think all current EVLA/WIDAR 
antennas still have the "Hybrid L-band" feed/polarizer or otherwise 
not-yet-EVLA-compliant L-band RX.)

Also, the above might imply that you did the BLCAL before the ordinary 
selfcal (both, presumably, with the VLA model).  Is this so?  This tactic 
is subject to how the time-dep antenna-based calibration residuals average 
up for the (constant) BLCAL solution.  It is probably wiser (but an 
absolute ideal is not possible) to do the ordinary selfcal first, and let 
it also soak up longer-timescale (constant) _antenna-based_ errors, before 
letting BLCAL obliterate such information.  This point is probably far 
more important when you are actually deriving the model from the data 
itself in between these solves (rather than uniformly using a prior VLA 
model).  In the present case, it merely confuses what is antenna-based and 
what is not for the long-timescale (i.e., constant) stuff.  (To be sure, 
closure errors can also leak into the antenna-based solution, but they do 
so less efficiently by a factor Nant, and we should expect to be dominated 
by antenna-based stuff in any case.)

-George




--