[daip] IMAGR problem

[Cormac Reynolds]

On Tue, 13 Feb 2007, Eric Greisen wrote:

> The problem is as I surmised related to a pathological beam.  The beam
> pattern appears to be (nearly) one with peak bumps at regular
> intervals in all directions.  It amazed me that a wide open clean
> would find an object as well as it does.  When outer points in a beam
> exceed 1, then the standard "histogram" giving the peak sidelobe
> outside radius r becomes a constant > 1 for all r out to that highest
> outer point.  The algorithm to find how big an area to clean fails in
> this case.  31DEC07 already has a change that limited the histogram to
> the inscribed circle which stops the failure in your case.  I added
> today more code to look for a failure and to take corrective
> (undesirable otherwise) action.
> 
> I guess you need 31DEC07....

O.k. great. Actually, it's good just to know why it failed. There's a
lot more going on in these algorithms than most of us users realise...

thanks,
Cormac.