[daip] clean components and image restoration

[Eric Greisen]

sdt4 at njit.edu wrote:
> Hello Dr. Greisen,
> 
> My name is Samuel Tun, and currently I am working on some imaging programs for
> the Owens Valley Radio Observatory. Along the way, I get all the CLEAN maps to
> come out well, but I am having problems being able to go back and forth between
> the image and uv planes using the CLEAN components (I'll need this for my
> implementation of self calibration). I read a thread on the public mailing lists
> on listmgr.cv.nrao.edu. that refers a user to task RSTOR in AIPS, and made me
> think you might know the answer to my question. 
>  
> Simply put, is there a standard procedure/method for recreating the CLEAN map
> from visibilities derived from the collection of CLEAN components? Simply using
> the CLEAN components gives me a map that matches my CLEAN map in appearance, but
> not in flux values. I have carefully accounted for all the factors used
> throughout the CLEAN routine, but am still off the mark by about a factor of 10.
> Any help and direction is most welcome and appreciated.

The Clean image is not restored from UV data.  The Clean image is 
comprised of the residual image left at the end of the Clean with the 
components restored in the form of Gaussians.  They are put in - 
assuming no deliberate change of units - with a peak equal to the 
component flux and a size equal to the "beam" and the units are then 
Jy/beam area.  One does have to be careful that the beam in restoration 
is about the actual synthesized beam.  If it is not, the units of the 
residual image must be re-scaled to go from Jy/ observed beam to Jy/ new 
beam.

For self-cal and modeling, the image itself is totally ignored and only 
the Clean components as points (unless multi-resolution was used) are 
subtracted from or divided into the visibility data.  AIPS tasks like 
CALIB, sum up the model visibilities at each observed point and then 
divide the sum into the data to get the apparent gains on each baseline.

RSTOR and CCRES in AIPS are tasks that rearrange the component Gaussian 
and residual portions of the image.  They ignore the uv data entirely. 
One can take out a particular Gaussian and put in another etc.  I have 
no idea where your error of a factor of 10 might be coming from, but a 
beam area is often 10-20 pixels so it may be related to units of 
Jy/pixel vs Jy/beam.

Eric Greisen