[daip] frequency dependent correction in IMAGR

[Eric Greisen]

On 06/30/2014 08:13 AM, Nicholas Wrigley wrote:
> Hi Eric,
> Sorry for the delay getting back to you, the machine I have been using is having problems, so I still cannot get at the header at the moment. But I’ll tell you what I’ve been doing..
> The array is the e-MERLIN array and the data was split up into multiple facets, 64 in total, by phase rotating and averaging in time and frequency such that each local facet is a UV dataset with 128 channels split up into 8 IFs from 1.25 GHZ to 1.75 GHz. ie. each channel width is 4MHz with 4s averaging time. So these 64 uvfiles have phase centres spread over a chessboard-like grid covering the primary beam of the array, to be imaged individually (for speed). I can image them all fine without the imagrprm(1), but as I want to correct for the PB induced spectral index, I was hoping to use this feature. There don’t appear to be any error messages when it fails - it just stops and hangs after what seems to be a major cleaning cycle. For example if I choose niter=10000 then it would stop somewhere in the 9000s. I then have to kill the process. If I increase the NITER to 20000 then it fails somewhere in the 17000s or 18000s (I think). It seems to vary.
> I’ll get the header etc ASAP.
> Any ideas in the meantime? To be honest, I don’t quite understand how this feature works, and I didn’t specify any FQTOL - maybe I should have?
> Thanks

The feature works by figuring out the average contribution of each 
spectral channel to a particular component and then scaling the 
subtraction of each channel to correct for the beam.

The reason why one usually does all facets at once is that sources in 
one facet make sidelobes in another.  When imaging with one dataset all 
facets then when one source is subtracted from facet N its affects 
disappear from all facets.  The scheme you propose may be faster but it 
is not better.

I will try this option myself on some data and see what happens.  What 
architecture (Linux, Linux64, Mac) are you using?

Eric Greisen