[alma-config] From David Woody

[Al Wootten]

Reply-To: "David Woody" <dwoody at ovro.caltech.edu>
From: "David Woody" <dwoody at ovro.caltech.edu>
To: "Bryan Butler" <bbutler at aoc.nrao.edu>, <alma-config at nrao.edu>
References: <200101292147.OAA19520 at planetas.aoc.nrao.edu>
Subject: Re: [alma-config] thoughts on simulation metrics
Date: Tue, 30 Jan 2001 13:26:34 -0800
Organization: Caltech
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4133.2400
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400

I agree with Bryan that we need to go back to 
looking at selecting useful metrics.

Adding to Bryan's comments

1.  UV metrics, or equivalently dirty beam metrics, are
the only way we can hope to evaluate the myriad of configuration
options and operational issues in a finite amount of time.
It is also the only way to generate an "optimum" configuration.

2.  It is important to find UV metrics that are 
good proxies for the kinds of image quality issues that 
astronomers are interested in.

3.  We need to keep in mind that the imaging algorithms
necessarily use assumptions about how to fill in the missing
UV data and that different assumptions will lead to
different algorithms and imaging artifacts.  The algorithms
used by the next generation of astronomers will presumably
be better, or at least different, from the current algorithms.

4.  Maximizing the UV coverage (minimum missing UV data)
must minimize the artifacts in the absence of noise.
If every measured UV data point has infinite S/N, then
you can give it arbitrary weight to produce whatever
sky beam you want consistent with the extent of the 
UV-plane coverage.

5.  The importance of the UV distribution is that it determines
the integration time at each point in the UV-plane and
hence the noise at each UV point.  For a given desired 
sky beam shape, the S/N on source in the image
is maximized when the UV distribution matches the
FT of the desired sky beam.

6.  It is difficult to produce test images that do
not prejudice the configuration selection.
Starting with a high resolution test image and 
convolving it with a beam shape function to produce
another test image at lower resolution is  not a fair test.  The best
configuration will be the one that best mimics the
beam shape used to produce the lower resolution
test image.  The same can be said for comparing the
reconstructed image with the test image convolved
with the "clean" beam, i.e. an image convolved with
a gaussian beam will necessarily be best reconstructed
from a gaussian UV-distribution.
 
****************************************
| David Woody
| P.O. Box 968, 100 Leighton Lane                         
| Big Pine, CA 93513, USA                                  
| Phone 760-938-2075ext111, FAX 760-938-2075
|dwoody at caltech.edu 
****************************************