[mmaimcal] ACA sims

[Mark Holdaway]

Yet another installment of simulation results can be found
at:

http://www.tuc.nrao.edu/~mholdawa/

I'm also including my "results" writeup:



The results are qualitatively similar to the results obtained by the
French group, in that the ALMA + ACA (option 1) is superior.  In the
case of the CLUSETR model, the improvement with the ACA is marginal
(typically 10-30\%), while in the case of the DEBRIS model, the
improvement is substantial (more like a factor of 2 for a clip of
0.33\%, the only clip value which samples most of the disk).  Note
that a simple position-dependent self-cal algorithm could probably
improve the DEBRIS reconstructions greatly, as they are largely
dominated by errors scattered from the bright central source.

The HCO(4-3) model image shows fidelity curves which are pretty flat with
frequency, indicating that HCO(4-3) image is being limited by something
besides pointing errors (ie, perhaps MEM is just having trouble
deconvolving
this image).  Two notable features of the HCO simulations are that the
short spacings are more reliably reconstructed than the long baselines,
and using a short baseline array such as the ACA or SMA increases the
short baseline fidelity by about 50\% over the homogeneous array.

Using the SMA for short baselines doesn't work quite as well as using
the ACA, but it would be an improvement over the homogeneous array
ALMA alone.  However, it is very difficult to justify either the ACA
above the homogeneous array alone, or the ACA above the SMA.

The CLUSTER model had homogeneous array simulations with both 9 and 25
pointings.  The 25 pointing simulations produced results which indeed
were superior to the 9 pointing results, and which in many aspects
were as good or better than the ALMA + ACA or ALMA + SMA simulations.
This indicates that, to some extent, one can achieve the capabilities
of the ACA simply by observing a wider field of view.  In other words,
the investment of extra observing time for the guard band could be
traded off against the extra expense of the ACA.  However, it seems
pretty clear that the ALMA + ACA case would also benefit from a guard
band of interferometric pointings.  As evidence of this, look at the
CLUSTER model's UV fidelity in the 15-37 m, 37-75 m, and 75-150 m
cases.  All curves lie more or less on top of each other (as it should
be, because the measurement of the short baselines should not really
impact the reconstruction of the long baselines.  However, the 25
pointing ALMA HA curve is significantly better than the other curves.
This indicates that a 25 pointing mosaic made with the ALMA + ACA
would also improve in image quality, likely at all spatial
frequencies.