[alma-config] Zoom design

[John Conway]

Hi,

 I have created a zoom preliminary design which can be found 
at http://www.oso.chalmers.se/~jconway/ALMA/SIMULATIONS/SIM4/
Some points.

1) I have incorporated what I think are the best bits of the 
NRAO strawperson design (the outer ring and E-array design) 
to create something of a hybrid. Note I have used the old 5deg 
mask which in the region of the <3km baselines seems identical 
to the new 5deg mask


2) To facilitate intercomparision of the two array styles
I have kept the same array center and have virtually the 
same number of pads and antenna moves (194 pads are listed,
but the configurations uses only 191; I only noticed this at a late
stage when it was hard to change). The number of antenna moves in
going from E to A array is 123 (5 antennas don't move).
The resource requirements are  virtually the same as for the
double ring design (I believe there  are 192 pads and 128 moves 
for the NRAO strawperson)

I have also tried to keep the same degree of sampling of the short
spacings for the zoom as for the double-ring arrays,


3) What I think is very interesting is that if one compares 
the radial distribution of uv points (see above web site) for the two
styles  of array, they are now quite similar, i.e. the double ring has
almost the same taper as the zoom spiral. It seems that given the 
number of pads we are allowing ourselves and  50%  pad sharing 
between configurations a factor of 2 apart in resolution then 
the tapering is only a weak function of the exact array geometry. 
(this is in contrst with the interpomprsion of the double ring 
designs of a year ago when the radial distributions were much 
different). Note that The fraction of zoom spiral uv points beyond 
the largest baseline  of the equivalent resolution double ring array is
now only  between 1% and  5%. 


4) I have for each of the A-D arrays minimised the snapshot sidelobes
within 22 beams, using a version of the Kogan algorithm in my own
software, it attempts to simultaneously reduce  several (5 -20 
usually) beam peaks while giving the least square change. Within
1 - 22 beams radius for snapshots I can get minima around 0.05 (i.e.
similar to the case for the twin rings). Making dirty beams in AIPS I
appear by  using IMEAN to give similar peaks, I have tried to confirm this 
with CONFI but have trouble getting it to work. There may be 
slight differences based on the difference in how AIPS fits FWHM 
and the inner radius that is  used in evluating peak sidelobe.

Roughly though it seems the snapshot peak sidelobes are similar
for the two designs. However  for the long tracks I find peak 
sidelobes  which are somewhat smaller (i.e. 2.5% compared to 
4.5%).



5) I am mot 100% happy yet with my A-Array design, although it gives
good long track beams and sidelobes. I found that  keeping some 
antennas on the 150m diameter circle gave much needed short spacings 
and also gave an extra density of point at a radius in the uv plane
which was 50% of the maximum radius. However  these extra points  seem
 to give a sidelobe contribution which parthy cancels out 
the first near-in sidelobes from the otherwise smooth distribution
distribution plus sharp outer edge.



6) I think in terms of scientific capabilities that it would
be good for both designs to have an even higher resolution (<3km) 
array than A-array. It seems to me given a fixed amount of real extate on
which  antennas can be placed one should have an array  which gives the 
maximum possible resolution from that area, this can be accomplished by
putting antennas on the perimeter of the allowed area (its also  
not good to have too much a gap in resolution between the 3km and 10km
array)

An extra 32 pads
on the 3km ring would allow such an A+ configuration with resolution
significantly larger than A-array, Perhaps in terms of saveing pads and
moves an extra 16 pads on the outer ring would suffice, giving a
option to put 3/4 of the antennas on the outer ring. 
 

8) I have included one example of a N/S hybrid array for a zoom
spiral on my web page, which has good uv coverage and quite round 
beams up to dec +30.  There
is a slight pedestal in the N-S direction to the main lobe but 
I don't think this is too serious in terms of imaging. 
I suspect the hybrid double ring will give slightly more N-S elongation 
but larger sidelobes.


9) From tomorrow I will be out of contact for a week or so.
Hopefully though I have provided enough to allow the imaging simulations 
to go ahead quickly (I have not had time to do any myself). 
It will be interesting to see what happens,
I expect given the degree of convergence that has occured there
will not be really dramatic differences in image quality between the 
two designs, but it will be good to know for reference what the 
level of imaging error from uv covergae are to compare with other
error sources (pointing, phase-cal etc).


I get a feeling that a significant convergence between the 
two designs is occuring - I hope others agree.

 Cheers
   John.