[daip] Rescaling at end of IMAGR

[Eric Greisen]

Andrew Biggs wrote:
> Dear daip. I have a question regarding the rescaling that goes on at the
> end of IMAGR when multiple channels have been imaged separately. After
> each channel is imaged (I do not clean as the continuum has been
> subtracted) I'm informed that the field is being rescaled to fix the units
> and I think that this is something to do with the different size of the
> dirty beams compared to what is written in the header. This has the
> effect, I think, of making the measurement of a flux in the map correct. I
> can check this by running imagr on all channels and then on just one. If I
> run an IMSTAT on the same channel in each, the reported flux is the same
> in each case.
> 
> However, and this is where I worry, the rms and max/min are not. Obviously
> this must be the case given that the multi-channel images have been scaled
> by some factor. However, I am going to stack these images (weighted by the
> noise) and worry that I'm artificially inflating the noise unnecessarily.
> For example, I can run IMAGR specifying a restoring beam (using BMAJ and
> BMIN) and making a map of the same channel in the two different ways
> produces identical images as the pixels weren't scaled during the
> multi-channel imaging, but no actual restoring was done in either case (as
> nothing is being cleaned). Should I in fact be preventing any rescaling
> going on? Or would convolving each source to the same resolution,
> post-IMAGR, be tantamount to the same thing?
> 
> The stacking, by the way, is of the 3-d variety, the map of a channel from
> one source being combined with the map of a different channel from another
> source. This is all taken from a wide-field dataset. I'm making small maps
> of the individual sources after rotating them to the phase centre and then
> stacking them in IDL. This, among other things, presumably allows UVLSF to
> work better with a low-order fit as all the sources will be at the phase
> centre.

The rms and max/min are in units of Jy per the beam listed in the header 
of the image and so they differ depending on whether you make the full 
cube or just a few planes.  To stack as you propose, I would make the 
image without specifying BMAJ, BMIN, BPA and then run CONVL on it 
specifying the fattest BMAJ, BMIN, BPA of all of the channels (usually 
that will be the one in the header).  In that way each channel image 
gets the same resolution and all the noise is handled appropriately.

Eric Greisen