[alma-config] constraints

[mel wright 456]

Hello Min,

        some comments after the discussion this morning:

1. For sparse arrays with a master plan (or severe constraints),
the brain is reasonably effective in finding good antenna configurations. 
With no plan (or modest constraints), Keto's neural network code was better.

2. For large arrays, such as the 1HT with 500-1000 antennas,
the science requirements determine a density distribution of
antennas; the detailed antenna positions do not matter so much.

3. ALMA, with 64+ antennas falls in between. The most
compact configurations are determined by the maximum antenna density.
The largest array is a ring to give the maximum instantaneous
resolution, and happily can give a reasonably uniform uv-distribution
down to a minimum set by the number of antennas.

4. The intermediate configurations should be subjected to
a cost-benifit analysis. 

5. The configurations should be robust; 
there may be 5% of the antennas out of use at any time.
How much is the image fidelity degraded
as the array is moved away from any 'best' configuration.

	cheers, Melvyn