[daip] Regarding an issue in using 'spectral index' image with AIPS task IMAGR {External} {External} {External}

[egreisen]

On 2023-05-29 13:23, Subhashis Roy wrote:
> Hi Eric,
> 
> I have been playing around to check how wideband Clean of Aips via
> IN3NAME works for multi-facet images [of-course setting Imagrprm(8)=0
> for the time-being].
> 
> As I understand, with typical observations, one need to make a image
> cube from sub-bands and Clean it with BMAJ, BMIN, BPA set to be the
> same during Clean for the cube (without the 3 parameters set or
> convolution, SPIXR has trouble to compute spectral index correctly due
> to differeing beam sizes as a function of frequency) with NFIELD set
> to N (as needed).
> 
>> From the Clean image cube having N facets, one uses FLATN to get an 
>> image
> cube for the whole Field of view. Then run Trans, and Spixr on Trans
> output to get the spectral index map. Finally, use the Spectral index
> map generated in the last step as IN3NAME, and set IMAGRPRM(17) (I am
> setting it as 2).
> In the next run of Imagr for wideband Clean, set Echan=M, NCHAV=M
> (where M is the total no. of frequency channels). NFIELD=N set as
> before.
> 
> The above steps seemed to work well with N=3, M=5. Differences of the
> new Model from UVMOD w.r.t. the previous one are: the 2 point sources
> were put +/-600" away along North and South from the centre, and
> spectral index of the central Gaussian was set to -0.2 (earlier was
> 0.0). I also made Flux (UVMOD) to go down by 4 as compared to the
> earlier.
> 
> The result is attached (gzipped Fits image). You can see that the
> Clean problem around the compact sources are gone due imaging them at
> the centre of separate facets.

I have done more tests.  The filtering option does use the spectral 
index.
That has been set up and should run correctly. I think the filtering 
itself
is not a good thing.  If one  cleans the whole image without Clean boxes
then it can be good to get rid of components that would never have
appeared in boxes set properly.

The defects around the northern and southern sources have looked to me 
like
3D issues, but SETFC insists that your original model only needs 1 
facet.
Going to 600 arc sec will use more facets according to SETFC.  I suspect
that the filtering option should really use a radius of 10 or 20 cells 
not
just 3.  I notice when filtering that a bright rim appeared around the
central Gaussian indicating that the filter simply nibbled away at the 
edges
rather than getting "isolated" components.

I guess I should try your more extreme model just for fun.

Note, you are better off running IMAGR with BMAJ 0.  Then run CONVL on 
the
output cube with operation GAUS and BMAJ 0.  That will smooth each plane
including the residuals to the basic (poorest) resolution.  Setting BMAJ 
in
IMAGR smooths the Clean components but not the residuals to the desired
resolution.

Eric Greisen