[alma-config] CONFI

[John Conway]

> 
> Some comments to your and Min's message:
> 
> For other configuration it is not clear which side lobes affect more 
> the image performance. I discussed this question with some people here
> and there is an opinion that the far sidelobes are the less they
> affect.
> 

I have the same feeling, that even for the snapshot
beam descresing the near-in sidelobes is much more useful 
for descreasing  imaging errors.

In addition as one goes from snapshot
to long track the far sidelobes seem to descrease faster than the 
near-in ones.  If one thinks  of the long track dirty beam as
the sum of rotated/strectched versions of the 
zenith snapshop beams and if one approximates sidelobes peaks as
isolated peaks of constant width (2 to 3 pixels) then one would expect
sidelobes to decrease by a factor  1/r when going from snapshot to long
track (I admit though this argument is very approximate, but up to 
a point its roughly  confirmed in simulations).

>From both of the above points of view it might be best when -finding - the 
peak sidelobe in CONFI which one attempts to minimise to first multiply
the dirty beam by (r_{o}/r) outside some radius r_{o}. Then minimise this 
sidelobe. I am thinking of adding this modification to CONFI and see what
happens. Doing this would  also  mean than one does not get a sharp
transition between  low sidelobes and large ones at some radius as now.
For the compact configurations as you say you can arrange the radius at
which the large sidelobes lie to be outside of the primary beam, this would be 
OK for single pointing observations. Then again when one mosaiced one
would tend to get these sidelobes back. The reduction in the sidelobes is coming
from the fact that an interferometer samples not a uv point but an area in the uv
plane (or the uv plane convolved by the single dish size and then
resampled), mosaicing on the long baselines allows you to reconstruct the
unsmoothed uv data (and for the short baselines of course get uv spacings
less than a dish diameter). Still minimisiing within the primary beam is at least
not an arbitary  choice, and the cross-sidelobes in mosaiced fields might
be small enough anyway. 


> Optimising side lobes we automatically produce tapering in the UV coverage 
> and usually achieve a smooth UV coverage.
> 

Yes I agree with you and Min that - in practice-  beam minimimization
seems to produce a smoother and somewhat tapered uv coverage. How it makes
the uv smoothing is not obvious  athough it is clear it is an interative
method forcing toward a beam with nice main lobe and zero sidelobes whose
FT is smooth. It effectivly stomps on anything else, and forces the beam
to this ideal. It  has the  same sort of relationship to uv metrics that CLEAN has
to MEM,and we know that even though  CLEAN is not as theoretically as nice
as MEM it often gives better results. CONFI also gives some degree of uv
tapering, it has to do this to reduce the near-in sidelobes from 15% to
around 5%  but after this it starts concentrating on the far sidelobes
since these  are larger than 5%. I would argue from the point of view of 
imaging simulations and the size of the long track near-in sidelobes that
somewhat more tapering than  what  the present CONFI gives is desirable.  By 
adapting  CONFI to give more weight to minimising near-in sidelobes and
starting with  input uv distributions which were already tapered I'm sure 
that more tapered outputs would result.


      John.