[Almasci] Re: [mmaimcal]Re: Comment in ALMA memo 489

[Mark Holdaway]

> A couple of side points which may help improving the algorithm(s) or our understanding of these algorithms
> 
> 1) CLEAN with or without the zero spacing is not the same
>         - because we add an offset into the map, the positive and
> negative components are different. So CLEAN converges toward a quite
> different solution as soon as (non null) zero-spacing flux is added. The
> zero spacing flux, despite the fact it is not "localized" helps solving
> for other scales in the image !
>         - the above point is perhaps obvious to most of us, but in the discussion, I had the impression the non-linear
> nature of CLEAN was lost.
> 
> 2) So far, there is no way to tell CLEAN where to locate the extended
> emission
>      - yet, we A PRIORI KNOW this extended emission is not entirely flat: there is the primary beam which attenuates the
> sky brightness...
> 
>     Has anybody an idea on how to incoporate that "a priori" information
> into CLEAN ?
>     Perhaps let the final "missing" flux be localized just as the
> (primary) beam ? At least, that would look a better guess...

Long ago, when I was but a child, people used something called "Clean
Boxes" to restrict the model brightness distribution.  But now that I am
grown, the algorithms have grown up too, and (at least in some
packages), you can provide an arbitrary mask image to function as
the "clean box".  That mask image can restrict clean components to
be within the generalized primary beam pattern for the mosaic.

   -M


> 
>         Stephane
> 
> 
>