[daip] AIPS: SAD for mosaicked data

[Eric Greisen]

Matthew Sharp writes:
 > Dear Designated AIPs,
 > 
 >   I've been working on sets of mosaic maps made with 8GHz VLA data
 > (thanks again for all of your help getting this far!).  I seem to be
 > able to generate calibrated SNR maps of my mosaicked fields which take
 > into account the mosaicked beams accurately.
 > 
 >    I want to find a catalog of sources in this map.  My mosaic has
 > non-uniform sensitivity, and so I need to look for peaks in this SNR
 > map, and not in the actual, Jy/beam
 >  signal map.
 > 
 >    using SAD, I can get a list of all sources that show up at 5sigma
 > in my mosaic, but the flux that is reported is in units of the ratio
 > of signal over noise that I used to compose the map.  I'm wondering
 > about how best to find the real, beam corrected fluxes of these
 > sources in Jy/beam.
 > 
 >    My thought is to use SAD to generate a residual SNR map, and then
 > to subtract the residuals from the map itself, leaving zeros except
 > for my 5 sigma sources.  I can then multiply this difference map by
 > the noise map, yielding a map with only my bright sources, but in Jy.
 > 
 >    As for the primary beam correction, I understand that FLATN can do
 > this for me when i generate the maps, but I'm not certain if it's
 > necessary to set PBPARM(1)>0, or if this correction happens no matter
 > what.
 > 
 >   Now I should have a map, in Jy, of my mosaic field, with zeros
 > except for 5 sigma sources, which should remain as gaussians.  I can
 > then use SAD a second time to extract fluxes and other parameters from
 > this map.
 > 
 >    I'm wondering if this method is likely to produce what I want, or
 > if there is perhaps a more straightforward way of extracting source
 > fluxes from this kind of signal to noise map.
 > 
I do not know how SAD would respond in this case - it would be easy
enough to try.  My bet is that it will not get the same answers.  To
pursue what you are doing in a better way - you could write a
procedure that reads to MF file for center coordinate in pixels, reads
the noise map at that pixel, then reads the MF file peak, peak error,
etc columns to multiply them by the noise and put them back.  (Of
course copy your MF file to a new version before doing this for
safety's sake.)

A thought has occurred to me.  If a source occurs in a location where
the noise image is changing with position, then a Gaussian in the
intensity image will not be a Gaussian in the S/N image.  This may not
be important depnding on the rapidity with which the noise image
changes compared to the source sizes, but it is a worry.

Eric Greisen