[alma-config] Comments on Marks proto-memo

[John Conway]

>
> We also need to distinguish between ON-SOURCE errors and OFF-SOURCE
> errors.
> 
> ON-SOURCE errors are much larger than OFF-SOURCE errors.  In mosaicing, at
> least, the off-source errors are roughly of the magnitude of the on-source
> errors times the rms PSF sidelobe level.  From mosaicing pointing error
> simulations, we found the off-source errors were like 1 part in 500 or
> 1 part in 1000.  The ON-SOURCE errors were like one part in 20.
> 
> 	-Mark
> 

1) Yes this is a very good point, which I did indeed gloss
over. The two types of error need to
be distinguished, and I have been worrying about this in terms of 
what imaging metrics one should use.

In the M51CO simulations I described, testing pure deconvolution 
errors  (on quite a simple source and a long track) the peak ON-SOURCE
errors were about 3%
(1 part in 30), I don't have a number for the OFF SOURCE errors, because
they are probably masked by the thermal noise I added and the model may
not be high DR enough. In my CLEAN Cygnus A similations however the 
off source errrors due to deconvolution were a factor or 10-20 less than
the on source errors. My point yesterday is that the pure reconstruction
errors may often larger than what people expect, and it would be nice to
have a handle on what they are.

The  balance of importance between these pure reconstruction errors
and pointing or other errors will  be strongly frequency dependent as well as
source structure dependent, because of course the fractional pointing error is
strongly frequency dependent.  I believe there will be a regime 
of imaging at  90GHz to  23GGHz of sources which do not fill the 
beam where the effects of pointing errors are small 
(by 'pointing errors' I mean errors which are both image position 
and antenna dependent, not just antenna dependant, the later can be
fixed up by amplitude self-cal). This regime is worth simulating.
The opposite regime of full mosaicing is also worth simulationg. If the 
same array style come out on top for both regimes thats great, if
not we have to weight the two tests (50/50?). Lets hope life is 
kind to us.


2) On thermal noise. I have also done some similmar tests as descirbed
by Mark, where I cleaned a pure noise image for gausian and ring uv
covergages, I was hoping that the gaussian might give lower noise 
in the clean mao (the argument  being thet the CLEAN map is convolved
by a restoring beam, which has a bigger attenuation effect on the uv data
for more uniform coverage). I found that the noise in mJy/beam were
virtually  the same.. I don't know if I understand it either.,, but
it seems to be the case.  


I guess we can talk about these points at the telecon. 

  Cheers
     John.