[mmaimcal] Text of AI on antenna quardupod design from Beasley

[Steven T. Myers]

> I still worry about polarimetry, though, as being a possible "showstopper" 
> for the two antenna problem.  As Steve is our expert, I hope he chimes-in on 
> this issue.

I don't see that its a show-stopper.  As I estimated before, I think it 
does mean you have to deal with the differing primary beams for more cases
rather than just the highest dynamic ranges (where you will probably have
to model/measure the primary very carefully).  It increases the amount of
bookkeeping in the software (carrying around V1, V2 voltage patterns or
B1^2, B2^2, and B1B2 beams) and there are storage and computation issues.
Therefore the upshot is probably about 1-2 total FTE-years extra 
computing effort that you might have to spend earlier rather than later,
and maybe a 25% or so hit on effective data rates (which will occur
during gridding and degridding stages) which you either have to spend
extra money for computing or eat in throughput on the science cases
that need it (relatively high dynamic ranges for filled beam or mosaics).

In the end, you will have to deal with effects which need an accurate
primary beam, and as far as the processing goes you build in the 
capability for handing differing antenna beams anyway (after all you
want to handle ACA-7m and ACA-12m also).  Note that this isnt specific
to polarimetry, but again the science cases for accurate polarimetry may
drive you to spend this coding and computing effort earlier than you
would otherwise.

My 2p worth...

   -s