[alma-config] Re: UV sampling metric

[Bryan Butler]

>Mark asserts:
>True: but to accomplish this, we need an intrinsically tapered
>(u,v) sampling envelope, or similarly, minimum sidelobe PSF's.
>But not many are arguing against this.  It does argue against
>"uniform" coverage.

i guess i still don't see why the tapered u-v sampling envelope
is a requirement.  i must have some misunderstanding about how
this works, given that you understand it much better than i do...

>Mark asserts:
>The Lamba function is the PSF, no?

yes.

>The characteristic width of the Lambda function will be the
>oversampling of the beam in image space, typically 3 pixels, so

i don't follow this.  could you expound on what you mean more here?
my thinking is that the characteristic width of the Lambda fn
is quite narrow compared to the "pixel" spacing.  maybe i'm just
way off here, but it seems like the Lambda fn width scales like
1/Umax and the image "pixel" spacing scales like 1/Usep (Usep is 
the spacing between u-v cell centers).  since Umax >> Usep, it
seems like the width of the Lambda fn. is << 1 "pixel".  where am
i going wrong here?

>Mark retorts:
>I think that means tapering the resolution to the extent that the
>positive and negative sidelobes blur together, meaning you've thrown
>away something like half your resolution, 3/4 of your baselines, and
>half your sensitivity.

no.  it means tapering it by less than 10% or so in linear resolution,
at least in the way i'm thinking about it (which may be wrong...).

>Mark philosophizes:
>We are in a bind.  True, the ALMA will spawn a variety of imaging
>algorithms that will improve the imaging and calibration; but we
>cannot design an array based on the characteristics of some unknown
>algorithm which meets some abtsract criteria.  On the other hand, it is
>not too good to design the ALMA to work well for our current
>inferior algorithms.  How far are you willing to guess we will progress?

well, what i'm saying is that the algorithm might not be so "unknown"
and that the criteria might not be so "abstract".  maybe we could
figure out what the proper algorithm should be in the case of complete
u-v coverage...

	-bryan