[alma-config] Difference images for the simulations using the SIL

[Mark Holdaway]

> Hello. Of late I have been wondering what the requirements, and indeed
> definitions, are for fractional difference maps using the CLEAN images and
> convolved models. More immediately my concern has been how to even create
> simple subtractive difference maps for array resolutions other than A, where
> the model and final CLEANed image are different sizes and have different
> pixel scales. Registration of these maps has proved problematic due to
> downsampling, windowing etc., and to show where I am at, I have put together
> a page on my website which demonstrates the current state of the art
> graphically. This may be found by following a link in the 8th November news
> item at the top of my usual index page at
> http://www.stevenheddle.co.uk/ALMA/ALMA_IND.HTM
> 
> I would be grateful if you could have a look at it.
> 

You have to regrid using HGEOM or similar.  HOWEVER, HGEOM's errors will
probably be larger than the errors you make in the deconvolutions.

This is one of the reasons why I regrid the model image to the desired
simulated cell size BEFORE I run the simulation.  (That and also the fact
that simulations with a smaller number of pixels will run faster.)  There
are still differences in the regridded model and the original model due to
HGEOM, but we don't care about them.  Typically, when using HGEOM on a
hand-groomed model image which actually goes to zero flux level outside
some finite region (as opposed to some generic image which will have
noise throuout), you find that the HGEOM'ed image will have some
negative pixels, which give you a handle on how large the errors are.  I
then tidy up that model image (ie, clip the negatives) and use that tidied
up, regridded model as the model image for the simulations at the new
resolution.

Anyway, before you redo all your work, you should try HGEOM and estimate
the errors to verify that my words are true for your particular situation.

	-Mark