[daip] Mosaicing-FLATN

[Eric Greisen]

Mpati Ramatsoku wrote:

> I have 2 pointings at 21cm from WSRT and want to combine them.

> AIPS 1: NFIELD        1                    Max number of fields per
> AIPS 1:                                       pointing
> AIPS 1: NMAPS         2                    Number of pointings
> AIPS 1:                                    Output image
>
> AIPS 1: IMSIZE        0           0        Output image size in pixels
> AIPS 1: COORDINA   *all 0                  Central pixel coordinate
> AIPS 1:                                    all 0 => use observed
> AIPS 1: COOTYPE    ' '                     Desired projection type
> AIPS 1: COOINC        0           0        Desired coordinate increment
> AIPS 1: COOREF        0           0        Desired reference pixel

> I have both my pointings included, but only corners are included  and 
> all the pixels within the round shape are blanked. It should be the 
> other way around.
> 
> Am I missing something important here?

I looked at your headers and saw nothing unusual with them.  The 
observed pointing in each is either 0 or exactly matches to reference RA 
and Dec.  So I tried a few tests taking copies of one of my cubes and 
changing the headers around some.  I changed the pointing and reference 
Dec in one to most of an image size away and then combined them with 
FLATN.  I tried on a Linux GNU compiler, 32-bit and 64-bit Linux Intel 
compilers, and also on my Mac (the one that computes the binaries for 
shipping).  In all cases beam corrections were made and the images 
looked as they should - I even tried changing telescope to blank and 
instrument to WSRT as in your headers, but this made no difference.

The only issue I encountered was that FLATN with your adverbs barely 
included the second image.  I needed to set IMSIZE and COOREF suitably 
in order to get both images to be part of the output.  I don't know what 
would be right for your images, mine needed IMSIZE = 512,1024 and
cooref = 256,257  or 256,769 (depending on whether image 2 was north or 
south of image 1).

I wish I could be of more help.  To speed any tests, try stripping just 
a few spectral channel off form each cube.  Your vary large cubes will 
be slow and the result should not care about the number of spectral 
channels.

Eric Greisen