[alma-config] Re: Keto's message

[Stephane Guilloteau]

I read Eric Keto's message and would like to comment on the points he
mentionned

1) The spiral array must 2 times larger than the ring to achieve same
angular
    resolution
        True: but so what ? What we should compare is the brightness
sensitivity
        at a given angular resolution, not the apparent size of the array
(i.e. the maximum
        baseline)

   Atmospheric decorrelation will be larger, and noise too.
        False: If we adopt either the fast switching, or the WVR correction,
the phase
        noise no longer increases on the longest baseline

2) IMAGR limitations (or other deconvolution techniques)

    As pointed out by Eric, most deconvolutions techniques use Gaussian
beams, which
    is not the best one can do for angular resolution.

    The best compromise would actually be spheroidal functions (which
optimize the
    support area in both U,V and image plane). As far as I know, only the
WIPE algorithm
    developped by Andre Lannes and collaborators does something similar.
Nobody has
    used it for real work yet, however.

    It is possible indeed to develop other deconvolution methods, or to use
something else
    than a Gaussian beam. This would be desirable, since Gaussian beams have
fairly
    extended wings which can confuse low level emission...

3) The large sidelobe level is indeed due to the outer cutoff. Moreover it
is also true that
    extrapolation is always more difficult than interpolation. Hence, for
good quality
    imaging, it is actually advantageous to have a few outer baselines /
antennas
    which indeed do not count very much in sensitivity terms, but prevent
prevent large
    interpolation errors.

    In the deconvolution theory, this is a better conditionned problem which
results in less
    amplification of the errors during the deconvolution. While the
non-deconvolved noise
    may be slightly higher, the lower amplification factor result in a lower
noise
    after deconvolution.

    The condition number (amplification factor) can be tested using WIPE.
The main
    problem is computation speed, but this should be affordable for a few
test cases.

4) SMA and ALMA are different problems
    - the number of antennas is completely different. SMA has to rely on
earth rotation
    for aperture synthesis. ALMA can do without for a fraction of the
projects.
    - SMA configuration was studied with one optimisation criterium
        "Maximum signal to noise for the highest angular resolution"
      Should ALMA use only that criterium ?

5) Many astronomical images actually have  structures on all scales. I
suggest the best
    criterium for INTERMEDIATE configurations is to match sensitivity to the
brightness
    expected on the various scales.

    This can be done in several ways, including multi-configurations
observations if
    necessary.

6) The criterium the the LARGE configurations can of course be different,
and the one
    used for SMA looks reasonable for these.

    However, what is considered a LARGE configuration ?


So, one  of the big questions for our meeting next week is to suggest
several possible
criteria, and then (unless we can agree totally among ourselves), to get a
recommendation on which one(s) are scientifically the most interesting.
Operation
constraints should also be taken into account.

        See you soon in Tucson

            Stephane