[daip] Re: AIPS deconvolution

[Eric Greisen]

Glenn Laurent writes:

 > I'm a 4th year graduate student at the University of Colorado, and I'm 
 > attempting to deconvolve a 1.1mm map of the galactic center with a map of the 
 > beam.  Our sysadmin, Janet Shaw, gave me your email address.
 > 
 > I am attempting to use the task "APCLN" in AIPS to perform this operation, but I 
 > am getting some errors.  Perhaps you could shed some light on this problem.
 > 
 > I am using APCLN becasue this task allows me to input both a dirty image and an 
 > image of my beam.
 > 
 > Understand that the .FITS image-plane images that I have were NOT created from 
 > uv-plane data.  These images are from BOLOCAM, a millimeter-wave bolometric 
 > instrument used at the CSO on Mauna-Kea.
 > 
 > I have padded the dirty image and dirty beam with zeros to be 1024x1024 becasue 
 > APCLN requires the beam image to be powers of 2 (due to the FFT operation???) 
 > and the beam image and dirty image to be the same size.

    The usual problem is that the images are too small - and yes FFTs
are used a lot.

 > 
 > Regardless of the input parameters, I get the following errors:
 > 
 > >go
 > APCLN2: Task APCLN  (release of 31DEC02) begins
 > APCLN2: ROTFND: CAN'T FIND PROPER AXIS TYPE FOR MAP ROTATION

     It is looking in the header for a proper RA-DEC pair of axes to
check for rotation of the axes.  Probably not important.

 > APCLN2: Create BOB         .ICL001.   1 (MA)  on disk  1  cno    6
 > AIPS 2: Resumes
 > APCLN2: BMSHP ERROR: CENTER OF BEAM =  0.26312E+00

     This however is serious.  How can you have a beam that is not 1.0
at its peak?  This should be the image of a 1.0 (whatever units) point
source at the coordinate center and properly calibrated.  Therefore it
has value 1.0 in brightness.  That is what APCLN is checking.  It is
checking pixel 512,513 (in your 1024x1024) case since that is where
the peak will be in an FFTed image in AIPS.

 > Now I have no idea what rotation axis the error is referring to.  Both images 
 > have the same rotation angle.

    I would need to see the output of IMHEADER on the images to tell -
but I suspect that it does not matter.

 > 
 > APCLN seems to be finding the maximum of the beam correctly (0.26312E+00), but 
 > I'm not sure why it is causing an error.
 > 
 > I'm suspecting that since I am using images that have not been created from 
 > uv-plane data, the FITS headers may be missing some information.  
 > 
 > Now I would not ordinarily use this software, but I have been told that this 
 > method has worked for cleaning old HST images (before the optics were 
 > corrected).

In principle, APCLN requires a symmetric beam - i.e .one whose FFT is
real.  It seems to sometimes help quite a bit with other cases but
cannot deconvolve really deeply in any case.

You will probably need to RESCALE the beam image.  If there are more
problems send me the IMHEADERs and any other info you can.

Eric Greisen