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

[Subhashis Roy]

Hi Eric,

Thanks for your effort and time to check it out.
I now agree that using Spectral index option through In3name and
Imagrprm(17) does significantly improves the outcome for broadband data.
The model that I mailed could give me a factor of 3 improvement while 
using Spectral index image made from multichannel UV data.

However, now I believe to have been able to isolate the issue that
causes no improvement despite Imagrprm(17) and In3name I reported
earlier.  I believe you had switched off Imagrprm(8) [filter
components] during the imaging that you tried.

After forcing the above option near the end, I could see the appearance
of significant diffraction rings around the sources with high magnitude
of spectral index, which did not go away with Clean later. When I
switched it off, the problem mostly goes away.
Are filtered CCs are not getting scaled by the spectral index for
different frequency channels in this case ?

I have been setting IMAGRPRM(8) to ~3-5*rms, and use it manually couple of times
when deep clean has been done on extended emission. It does make the final
emission significantly smoother.

Regards,
Subhashis

On Thu, 25 May 2023, egreisen wrote:

> On 2023-05-22 05:42, Subhashis Roy wrote:
>> Hi Eric,
>> 
>> IMAGRPRM(17) in IMAGR is a very useful option for imaging wideband
>> data, for which spectral index image has already been made by imaging
>> from each sub-bands.
>
> I have reached the conclusion that IMAGR is doing the best it can.
> When I made a model without the second central source and corrected
> UVMOD (it did bad things if FQCENTER > 0) and lowered FLUX, the
> IMAGR with SPIX gave an rms 4-5 times better than without SPIX.
> There are still low level rings around the north and south points
> but they are significantly reduced.  I noticed that IMAGR had some
> trouble with these sources even when doing a single channel so I
> am not surprised that some trouble remains.  The FLUX level you chose
> made the random noise a significant contributor to the image rms
> with both SPIX choices.
>
> Some of the issues I reported regarding SPIXR were due to my not remembering
> to run CONVL on the cube in advance.  Of course, my IMMOD spectral index 
> image
> was guaranteed to be better than from the IMAGR cube.
>
> Eric Greisen
>