[alma-config]misc config comments

[Stephane Guilloteau]

Back to the three points mentionned in the last 3 messages

1) Short spacings in all possible configurations: YES
    Mark is right here. It does not cost much, and saves a lot of effort.
This is very cost efficient. The only valid exceptions are the largest
configuration, where it is not only the "shortest" spacing, but quite a few
"shorter" spacings which would be needed. And in general, I doubt people
will be granted time on the 14 km array to make IMAGES if they haven't
already done a pretty good 3 km image... (For size measurement and compact
sources, I would not say so...)

2) Positive inner sidelobes: NO, must have a ZERO MEAN INNER SIDELOBES
    I disagree with Dave's statement. In our imaging simulations, we
actually found out that the best deconvolution results were obtained when
zero spacing data was added with such a weight as to produce practically a
"null" region for the inner sidelobes, and that deconvolution failed as soon
as any significant positive "plateau" was present.
    I concur with Mel in his statement that this ressembles pretty much the
"error beam" problem in single-dish.

    Using this sort of constraint to drive the optimisation towards getting
some short spacing is good, but obtaining a real positive plateau of inner
sidelobes is bad for deconvolution. The plateau should have zero mean.
Having a zero mean plateau does not imply you miss the extended structures:
the main beam remains sensitive to these structures.

3) Sidelobes level: SMALL
    Agree with Mark and Mel here. Algorithms play a significant role. From a
mathematical view point, high sidelobes result in low eigenvalue for the
inverse problem, so are always a nuisance when noise has to be considered.
The noise free case is a bit academic (unfortunately).

    My own 3 cents (of Euros)

        Stephane






-----Original Message-----
From: Mark Holdaway <mholdawa at nrao.edu>
To: mel wright 456 <wright at astron.berkeley.edu>
Cc: alma-config at donar.cv.nrao.edu <alma-config at donar.cv.nrao.edu>
Date: Wednesday, October 03, 2001 12:26 AM
Subject: Re: [alma-config]misc config comments


>
>
>> I don't think large array configs need to have short spacings.
>> Some overlap in spatial frequencies for multiple configuration
>> cross calibrations is good, but no need to overdo it. One
>> could make joint observations with ACA if there is a case for
>> short and long at the same epoch.
>>
>
>Sorry, this message is TOO LONG, but here it is anyway.
>
>Short spacings are required for imaging rather than cross-calibration.
>
>The idea behind extra short spacings in all configurations dates back to
>about 1993, and has become a generally accepted constraint in the
>configuration work.  In 1993, Robert Braun suggested that the VLA change
>it's configurations by adding more short baseline information, reusing
>pads in the C, B, and A arrays.  Also at that time, I realized that the
>Keto triangle configurations, even for 40 antennas, had a much worse
>central hole than the VLA's 27 antennas -- ie, in at least this one
>respect, we were going BACKWARDS.
>
>The idea is that the VLA's implementation of multiple configurations
>scales the inner hole as well as the longest spacing.  This basically
>requires that many "B" array observations also need to have some "C"
>array, or even "D" array time.  From SNR equalization arguments, the time
>required in "C" array will be about 1/100 the "B" array time (never done
>in practice; usually a minimum of a couple of hours is granted in the
>smaller configuration), and much less for "D" array, but the astronomer
>must wait and organize multiple data sets.
>
>Now: if so little time is required to "fill the hole" in the smaller
>configurations, and that hole is really such a small part of the Fourier
>plane (just happens to be the MOST IMPORTANT PART!), it seems that it
>would take very few antennas being moved closer toegther to fill the hole
>using the full time available to the more extended array.  The VLA has now
>gone to the shortened C array (a modification of the C array made by
>moving two antennas to otherwise vacant D-array pads) to implement this
>idea.
>
>Simulations by EVERYONE WHO DOES THEM (except for mosaicing simulations
>with total power) basically indicate that if you don't simulate either
>multi-configuration data or use an array with extra short spacings, you
>make bad images of complex objects.  A minor amount of care in assuring
>the presence of short spacings gives you excellent imaging. In fact, the
>short spacings often dominate simulation results.
>
>If one wants to act like a black belt, one may have theories about how to
>shape the exact Fourier distribution by compining different
>configurations' data with different observing times and different
>weightings.  Or, if one feels they are really pushing the abilities of the
>array on a very complex object the imaging of which would otherwise be
>dominated by errors in the extended emission, one can require multiple
>configuration data. However, for MANY projects which would otherwise
>require some multi-configuration time, a single configuration with some
>short spacings will do the job as required.  This is quick for the
>astronomer and simple for the support staff, streamlining operations in
>a telescope which is otherwise becoming bogged down by it's own
>complexity.
>
>Joint ACA-ALMA observations also go in the direction of complication,
>while the single-configuration-imaging idea is trying to simplify life for
>the astronomer (or scheduling/pipeline coders). Furthermore, if no care is
>taken for short baselines in the intermediate configurations, there will
>be a gap between the ACA's longest and the ALMA's shortest baselines, so
>it won't work.
>
> -Mark
>
>
>_______________________________________________
>Alma-config mailing list
>Alma-config at listmgr.cv.nrao.edu
>http://listmgr.cv.nrao.edu/mailman/listinfo/alma-config
>