[daip] New Ticket - [!BOX-227648]: CONVL

[Michael Bietenholz]

New Ticket: CONVL

Something strange is happening in CONVL.

I tried making running conv on an image.  This image had no clean beam (it was basically
IMMOD output, and I'd removed any BMAJ,BMIN values from the header), so its
effecdtively in Jy/pixel (rather than Jy/beam).  I'm trying to convolve it with a clean
beam.  CONVL seems to change the pixel brightnesses much as you'd expect,
so it chances from Jy/pixel to Jy/beam, but doesn't convolve anything.  Apart from scale
change (of 12.69 in the example below) the output image seems identical to the input image.


The messages give some clue that something is going wrong:
CONVL1: Task CONVL  (release of 31DEC10) begins
CONVL1: CLEAN beam   : MAJ:   0.0000 MIN:   0.0000 PA:   22.3
CONVL1: Convolve with: MAJ:   0.0010 MIN:   0.0010 PA:   22.3
CONVL1: Using factor = 1.26949E+01 to scale image
CONVL1: Create S93J-3SNSH15.CONVL .   1 (MA)  on disk  1  cno    4
CONVL1: QINIT: did a GET  of      5120 Kwords, OFF       -341920173
 -> this is the desired clean beam size (its really 1.02 x 0.98 mas, 
     not quite round.
CONVL1: Using factor = 1.26949E+01 to scale image
CONVL1: Create S93J-3SNSH15.CONVL .   1 (MA)  on disk  1  cno    4
CONVL1: QINIT: did a GET  of      5120 Kwords, OFF       -341920173
CONVL1: CONGAU: plane     1 convolve with   0.00002 x   0.00000 at  22.3
 -> this doesn't seem right - there is NO clean beam to deconvolve for this image, so we
     should just be convolving with a 1.02 x 0.98 mas @ 22.3 degs Gaussian 
    (or, in pixels for this image: 3.40 x 3.27 pix @ 22.3 degs).
    This convulution kernel reported in the CONGAU line is far too small.  If that's what
    CONVL is really using, then it would explain why the output image is essentially identical
    to the input.
CONVL1: BMAJ=  0.0010 BMIN=  0.0010 BPA=  22.3/Output beam

This behaviour seems to depend on the exact nature of the image, if I add a point source
to the same image as above, CONV behaves normally (ie. does convolve).  I can supply you with FITS files if you want to try and reproduce the behaviour.

This image *used to have* a clean beam very close to the desired one.  I've reset
all of the header keywords BMAJ,BMIN and BPA to have values of 0.
Is it possible that CONVL is using some other means of determining the existing clean beam?  Then it would find that the clean beam is already equal to the desired output beam and that no convolution is required.

In one sense, its IMMOD that sort of gets it wrong - the output of IMMOD has the same clean beam in the header as the input image did, however, the pixel-values for IMMOD have not been convolved with any clean beam - working around this inconsistency is what got me to
this point.

               

Ticket Details
===================
Ticket ID: BOX-227648
Department: AIPS Data Processing
Priority: Default
Status: Open
Link:  https://help.nrao.edu/staff/index.php?_m=tickets&_a=viewticket&ticketid=416