[daip] UVLSF in wide-field imaging

[Eric Greisen]

Andy Biggs wrote:
> Hi Eric. I have a question about subtracting a continuum from a 
> wide-field dataset. I have cleaned my GMRT data and subtracted the clean 
> components - this gets rid of all the obvious continuum. If I then stack 
> the data I get a weak continuum source (good!) but ideally I want to 
> remove this from the spectral data.
> 
> The way I've been doing the stacking of late is to shift each source to 
> the phase centre and then make a map. This seems to offer an advantage 
> over using a FLATN'd image as the sources are all perfectly centred in 
> their own image and I can also run UVLSF on each shifted data set. As I 
> understand it, UVLSF only works correctly for sources at the phase 
> centre, but can you see anything wrong with this approach?
> 
> I ask because the noise from a stacked continuum image of the UVLSF data 
> has extremely low noise, about 10 times less than I would expect. I've 
> only been using a zeroth or first order fit to be on the safe side.
> 
> And what about running UVLSF on a non uv-shifted dataset? i.e. just 
> running it once in the hope of removing all the remaining continuum in 
> from all the sources in the field in one fell swoop. Will this do 
> anything sensible? This has been done by some other researchers who are 
> working on a similar project. Given that UVLSF only works on a source at 
> the phase centre, this just seems plain wrong.
> 
> Thanks as always for your help.
> 
> Andy
> 

The issue is whether the continuum vis is sensibly a straight line in 
sin and cos vs frequency at each visibility.  A strong source well away 
from the center will make a sine wave which is not always a straight 
line.  Move that source to the center and it becomes perfectly straight 
- move close and it is probably good enough.  the best continuum 
subtraction is done with UVSUB for wide field images.  UVLSF is really 
for modestly simple cases.

Eric