[mmaimcal]Sci IPT minutes

[Mark Holdaway]

The minutes on the Y+ configuration are not
accurate, I must have spoken too quickly, or the
meeting must have gone on too long.  Of course, 
the statement here is TOO DETAILED for the minutes.


Concerning Y+:
The optimization is done.

I was just checking to see if the incrementally optimized configurations
are better than when we optimize for all 44 antennas (16 in common with
Conway) and then make the best intermediate configurations given what the
full resolution Y+ then dictates.  I consider 11 different Y+
configurations (each made by moving 4 antennas, or one day's worth of
moving, the final being full resolution), and given the station locations
defined by the full optimization, I've found the intermediate
configurations which provide the best inner sidelobes -- which are up to
13% in the middle, back down to <8% for the full Y+.  (I should note that
Conway's configuration actually has a 5% inner sidelobe level, as defined
by the snapshot dirty beam minus the best-fit Gaussian.  That 5% sidelobe
level is hidden in the wings of the Gaussian-like dirty beam, and the
largest sidelobes which are distinct from the main lobe are much lower for
Conway's arrays.)  Minimizing far out sidelobes and obtaining a good main
lobe beam shape also entered into the optimization as secondary factors. 
Far out sidelobes don't matter so much because earth rotation synthesis
very quickly smears them out, while the near-in sidelobes are typically
reduced by only a factor of 2 over a 4 hour integration.

When we optimize the station locations 4-at-a-time (ie, incremental
optimization), we actually do much worse, and the inner sidelobes exceed
30% for the full resolution Y+ array and are never better than the
full-resolution-optimized intermediate Y+ configurations.  Why is this so
much worse? When we optimize the locations of all 44 pads, we are doing
some pretty complex fringe cancellation to get the low (<8%) sidelobes,
but when we only optimize 4 antennas at a time, we apparently don't have
enough degrees of freedom to do what we need to get low sidelobes.  The
4-at-a-time incrementally optimized arrays have main lobes which are
closer to the desired circularity (10% N-S elongation, or circular for a
snapshot at dec=-48 deg) than do the intermediate arrays for the single
full resolution Y+ optimization.  The 4-at-a-time optimization typically
gives us about 6% away from the optimal beam elongation, while some of the
intermediate Y+ configurations of the full-resolution optimization are off
by 15% in beam elongation. 

As the near-in sidelobes are the biggest problem for the Y+ configuration
set, optimization of the full resolution Y+ array followed by optimizing
the order of antennas to move is the best way to go.  The exercise
permits me to say "I think this is a good configuration set", mainly
because I have worked hard to find a better set, and that one turned out
to be significantly worse.

            -Mark