[alma-config] simulated imaging test

[John Conway]

Hi Min et al ,

OK well lets go ahead with the approach you have adopted and see what
happens.

I'm afraid my brain must have been still partly on holiday yesterday.
I agree we don't want to try any simulations of sources bigger than 
the beam.  This means in the case of the input Cygnus A image that 
there is no room to have bigger versions for smaller arrays, the original 
image field is already 10" square (and beam FWHM at 345GHz is  about 15" I
guess).  For the smaller test images (eg protoplanatary disk) in which
the A-Array version of the test model is significantly smaller than the
primary beam then there perhaps would be an argument for using somewhat
larger versions for the C-array tests, to have almost as complex test
sources... but .. OK for now its simpler to try fixed size model images
and see what happens.

My concern I guess is that the approach adopted gives too much weight 
to A-array, because in the test suite these will have the most complex
images in terms of filled beam areas which is the opposite of what 
will happen in reality. We have to watch this doesn't happen. It may 
be that the C-array tests are not a good proxy for B and D array, 
maybe we should therefore conduct B abd D array tests as well. B,C and
D are the main arrays in which the two designs differ most in their
properties and this is why tests in these arrays are most important.
Once an automatic script is doing the simulations then
perhaps it would be quite easy to do simulations for all arrays.

   John.

P.S I see that the +25 dec simulation on Simons page used the 
pure C-array, I think we should consider doing some hybrid array
BnC  simulations.  
  
P.P.S On  Simons page it might be useful to have the uv coverage 
plots have the  same scale for both the 'K' and 'C' designs.


> Instead, what I am advocating
> is to use the test images that are the most complex we can have for
> the A array and see how the image degrades as the resolution is reduced.
> One advantage is the ease of bookkeeping as I mentioned bofore -- we need
> only a set of five well defined test images to do the entire variety of 
> imaging tests, rather than having variations of each test image for each
> configurations (and agreeing on what that should be again and again).
> There is a price to pay, which is that the particular test image may not be
> the most complex image one can imaging for some other array.
> However, that is why we have 5 different types of images, each of
> which presumably have structures and challenges at different scales.
> 
>