[evlatests] Deep Imaging!

[George Moellenbrock]

>
> D-terms.  For finite D-terms that differ among the antennas, dynamic range 
> will be limited to something like 1/(<D*D>*sqrt(N)) where <D*D> is the mean 
> square of the typical residual D-term (that part not partially absorbed by 
> the gain calibration), and N is the number of baselines.

Oops.  Actually, the (BLCAL-less) dynamic range is limited to:

DRmax ~ sqrt(N)/<D*D>

I.e., more baselines increases the limiting dynamic range.  This assumes 
that the D*D are ~randomly distributed over baselines such that the 
averaging over baselines works for you in the conventional way (e.g., 
probably not the case when there were ~same number of VLA and EVLA 
antennas).  Since the D*D terms are constant in time, there is nothing to 
be gained (except better detections of the BLCAL solution) by more net 
integration time.  More integration time tends to help ameliorate the 
rotating terms (D*P) for polarized sources, but it will be a rare case 
where the dampening is especially efficient, and the D*D will of course 
still be present.

(Correcting for the D contribution formally turns out to be quite
difficult for a small array [constant parallactic angle] like the
VLA.  This is a discussion for another day...)

-George