[daip] Request

[Eric Greisen]

Glenn Morrison writes:
 > I see your point. Let me ask the following question of measuring SNR  
 > of a source.
 > 
 > For single pointing of the VLA I have been using the SNR=(peak flux/ 
 > error in the peak flux -assuming that the errors are 1-sigma) where  
 > the measured values are from JMFIT or SAD. Now I was using the non- 
 > PBCORed map value for the peak flux and error in peak flux to derive  
 > the SNR. However, in comparing the measured values from the PBCORed  
 > map, I'm find that the calculated SNR are almost identical.
 > 
 > My question is --is there any reason you can think of for not using  
 > the PBCORed map for measuring the SNR values?
 > 

On a PBCORed image, the noise level is a function of position.  JMFIT
and SAD use the keyword in the header to give it the noise - and that
keyword is usually set prior to PBCOR.  If the keyword (from IMEAN) is
not present, they will by default use the entire image to estimate an
RMS.  Since that RMS is not constant over the image this is not a good
thing to do.  The uncertainty in the peak value of a Gaussian fit is
just this rms.  JMFIT will allow you to fit an rms over a small
region around the source which may help.   In general estimating the
S/N is better done on the non-PBCORed images.  But in overlap regions
of multiple pointings this only gives the S/N of the individual
pointings, not the final image.

Eric Greisen