[alma-config]misc config comments

[David Woody]

Hi Mark

This is a tricky and important issue and I don't think there is a single
or simple answer.  The problem I was trying to solve
is that for the larger configurations a Gaussian UV distribution
or any bell shaped distribution actually has very few short
baselines and so extended objects disappear, i.e. get resolved
out, and the single dish or smaller configuration data is required
to see them at all.  There is a desire for making all but the
extended configuration self sufficient, i.e. good images without
requiring other data.  Also it is generally accepted that the
single dish data will be inferior to interferometer data.  (This
is one reason for the ACA.)  Thus we should try to obtain
over as wide of a scale size as possible.

Everyone says we need more short spacing data, but none of the
optimization routines automatically drive the configurations in this
direction.
But I found that my sequential optimization did drive in this
direction if I tried to produce finite near sidelobes.  In my treatment
the single dish data is always included so that sidelobes=1/(N-1)
are actually zero in the absence of the single dish data.
So a better statement of the goal would be zero sidelobes
out to some radius for interferometer data only.

This comes down to how much are we going to depend upon single
dish data and if the PSF is used as a metric should it include the
single dish data or just interferometric data.  This needs to also
carry over into image simulations that attempt to include noise.
What noise level should it be assigned to the U=V=0 data.

Cheers
David

****************************************
| Owens Valley Radio Observatory
| P.O. Box 968, 100 Leighton Lane
| Big Pine, CA 93513, USA
| Phone 760-938-2075ext111, FAX 760-938-2075
|dwoody at caltech.edu
****************************************
----- Original Message -----
From: "Mark Holdaway" <mholdawa at nrao.edu>
To: <alma-config at donar.cv.nrao.edu>
Sent: Tuesday, October 02, 2001 1:02 PM
Subject: [alma-config]misc config comments


>
> Sorry I missed the telecon.
>
> I am not at all sure that one wants to have (as Dave Woody suggested), on
> average, positive sidelobes over some region of the PSF (like the inner
> part of the PSF).
>
> If we observe a source which is just a bunch of low resolution blobs
> (or similarly, if we taper the (u,v) coverage or smooth the dirty
> image to the low resolution), any region which has, on average,
> positive or negative sidelobes will show up as a large sidelobe at
> this low resolution.  I think that If we mandate that the sidelobes
> all be positive somewhere in the PSF, we will end up with a huge low
> resolution sidelobe there and large errors of extended structure will
> result.
>
> So, while the approach of iteratively moving out to more distant sidelobes
> sounds very good to me, that approach seems to be missing something:
> looking at various spacings in the (u,v) plane as well.
>
> Frederic said:
>
> > Thus, in interferometry, a given distribution of
> > Fourier samples yielding high sidelobes may allow to
> > reconstruct the brightness distribution with high
> > precision provided that the uv-plane is sampled with
> > the required accuracy.
>
> I think this statement is not necesarily true, as images are
> constructed via concrete heuristic algorithms.  These algorithms do
> not take advantage of any knowledge of the completeness of the (u,v)
> coverage, though they make minor use of the PSF sidelobe level.  CLEAN
> is more susceptable to sidelobe level than MEM-type algorithms which
> change many pixel fluxes simultaneously.  This speaks to the problem:
> we may imagine a future imaging algorithm which makes explicit use of
> the completeness of coverage of the Fourier plane, and this algorithm
> will speak to our desire to have a completely sampled Fourier plane,
> but that algorithm doesn't yet exist in the radio-astronomical lexicon.
>
>
> -Mark
>
>
>
> _______________________________________________
> Alma-config mailing list
> Alma-config at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/alma-config
>