[alma-config] Guess it...
Bryan Butler
bbutler at aoc.nrao.edu
Thu Mar 16 11:29:15 EST 2000
stephane writes:
>
> To Bryan's point
> 1) given a fixed physical size limit, what flavor of array would you
> design then?
>
> That is indeed the right question, beyond any doubt
>
> 2) now, if you can come up with a ring/donut which has nearly as good an
> imaging capability as the more condensed spiral, and that has twice the
> resolution, which would you choose?
>
> My reply to that is that having "as good imaging capability" and "twice
> the resolution" is a conflicting request. To get resolution, you have to
> sacrifice something (brightness sensitivity, sidelobes, deconvolution
> stability, ...). The best of both worlds does not exist, so that we have
> to define an acceptable compromise.
>
> That's our job. Moreover, we don't have yet the definition of "good imaging
> quality" (as pointed out by Eric Keto, current estimates of image quality may
> depend on the imaging tool)
>
to which i reply:
i agree entirely that you don't get something for nothing. the general tradeoff
in these types of problems (which involve fourier transforms of bounded samples)
is width of central lobe ("resolution") vs. peak sidelobe level (sidelobe
rolloff is also in there, but is of somewhat less importance). the key word
that you left out in my original statement was "nearly". i think john and
leonia's simulations are showing that the imaging performance (by what metrics
we are currently using, at least) is at least grossly similar for spirals and
rings/donuts (within 10's of percent, i would guess, and possibly even closer).
so, let me rephrase my original statement slightly differently: comparing a
spiral and ring/donut array, and assuming that the ring/donut has twice the
resolution and only 10% "worse" (by whatever measure you devise) imaging, which
do you choose? if you like, replace "10%" above by whatever number you choose -
20%, 50%, 100%, whatever.
now, as you point out, we still have no good idea of what "good imaging quality"
is and how it is quantified. i agree entirely. we will discuss this in tucson
a bit, i guess. this is how we arrive at the XX% number in the above statement.
but we can imagine that there might be a cutoff in the decision, and discuss in
broad general terms where that cutoff might be. for instance, you might say
that if the number is 10%, then you would choose the ring/donut. if the number
is 100%, then you would choose the spiral. but, if it is, say, 30%, which do
you choose? i suspect that we will find ourselves in this muddy zone in the
middle where there is no clear choice (we always seem to in these configuration
discussions), but at least we can discuss it...
-bryan
More information about the Alma-config
mailing list