[daip] problems with CCMOD, UVSUB, IMAGR

[Eric Greisen]

I have now done the test of imagr to make a residual image, imagr to
make a clean image (BTW the two sets of CC files are identical), UVSUB
the second set of CC files, and then imagr with no clean.

The values found:

Clean image peak 100000.   microJy
Resid peak         2000.
Diff peak             0.7
Diff normal value     0.1

I would say that this is an awfully good result.  The difference image
showed major differences at the corners of the facets which is where
minor differences in subtraction would be amplified.  IMAGR will have
used DFT subtraction in the early rounds and gridded later, while
UVSUB gets merged CC files and does gridded with the whole lot all at
once.  An average error of 1 part in a million is pretty good with
singole precision floating point numbers.

Eric Greisen