[daip] [Mnj] 31DEC07 Patch

[Eric Greisen]

R. Craig Walker wrote:
> Eric,
> 
> I'm trying to understand if this affected me since I use multi-resolution
> clean.  I think I generally use the FFT division in CALIB so I doubt I got
> nailed there, but the history doesn't seem to say.  For IMAGR, you mention
> a "restart".  Does that mean any major cycle, or only when you restart
> IMAGR using the results from a previous imaging process?  I see from the
> history that IMAGR uses DFT subtraction in the major cycles, but the
> components in one major cycle will all be the same type so was it OK?  I
> do think I noticed an effect where, if there was too much flux in the
> large components, things didn't work too well, but I never thought it was
> due to the components not being subtracted.  Also I would have thought the
> effect of non-subtraction would have been more dramatic.

The problem affected things only when all facets were being 
subtracted/divided at once.  This would occur in CALIB or UVSUB or in a 
full restart in IMAGR using the results of the previous IMAGR (i.e. 
BCOMP > 0, OUTSEQ, OUTVER all set to point to previous),  Of course, the 
problem also occurred only when DFT rather than GRID was used.  As you 
say, the affect should have been fairly dramatic.

Eric Greisen