[alma-config] Re: UV sampling metric

[Min Yun]

It seems to me that there are several aspects to this topic, and these
exchanges are making that point abundantly clear.  A constructive next
step might be looking into the new frontier -- what do other people 
outside radio astronomy do for image reconstruction?  In particular,
what do medical imaging people do?  Does anyone have some familiarity
with that?  Eric Keto started looking into this many years ago, but 
it did not go very far, as far as I know.  My personal recollection 
is that medical imaging people have developed image reconstruction 
algorithms as much or more sophisticated than we have 
although the problems at hands are somewhat different.  
I just don't know enough to say where we could start.
Bryan has some insight into this from his radar work, and there must be
other fields where we can learn a new thing or two....

The point Dave Woody brought up, I thought, was not so much worrying
about the data distribution at the extremities but within the sampling
plane -- the most recent discussions involving Mark, John, and Bryan
are also important but distinct.  With so many antennas for ALMA,
we should be filling all uv cells in the compact configurations very
quickly, but it won't be the case for the largest configurations.
So his point, I thought, was to include complete 
uv-coverage as part of configuration design.  As Ed reminded us,
elegant but otherwise stupid designs can in fact bunch up and crowd
the uv cells and make the problem harder than it should be.  I like
Leonia's minimum sidelobe algorithm for this reason that it prevents
such crowding, at high spatial frequencies where we have hard time
doing mental 2-D FFT.  The merit of "complete uv-coverage" should
apply here.

These two sets of problems are certainly related since we have the
finite numbers of visibilities to fill the finite area of uv plane.
Can clever algorithms allow us to maximize the resolution allowed by the
real space (i.e. the science reserve) or do we build smaller arrays
(and leave no gaps in the uv space) because no extrapolation is safe.



					-- Min