UVGLUE query

[Laura]

I am reducing some 16cm 2GHz bandwidth mosaic data. It was suggested that I split it into frequency chunks for the calibration - which I did, and the resulting image is much better than when I did it all together!

I did the splitting in frequency with
uvsplit maxwidth=0.128

Then I calibrate each frequency chunk separately, and eventually invert it with
invert vis='source.????.i'
where each calibrated frequency chunk is in source.1357.i, source.1489.i, etc.

That actually works very well! But I eventually want to be able to uvmodel subtract the point sources from the calibrated data, to look for extended structure. So in order to do that I uvglue'd the calibrated source files together with
uvglue vis=vis
where I made soft links to the various source files as vis_1, vis_2 etc. so that they could follow the naming convention required by uvglue.

But when I image the file made by uvglue, with all the same inputs as the previous invert, except with vis=the uvglue output file, the map is considerably worse, with higher noise.

Is there something you have to re-set if you combine files with uvglue? Or is there a better way to combine the calibrated source files?

I had a look at the uvglued file with uvspec, and I can't see anything obviously wrong with it. Maybe I am just setting some input wrong somewhere!!

Thanks!

[Mark.Wieringa]

Hi Laura,

there was a discussion on this topic earlier - http://atcaforum.atnf.csiro.au/viewtopi ... lit=uvglue

There are quite a few pitfalls:
- make sure you apply the calibration to the data before recombining (use uvaver on each file)
- make sure all the split files have exactly the same number of channels (the last file tends to get an extra one when uvsplit with maxwidth is used)
- make sure the files are in the right order - check if the frequency goes up or down with channel number and order the input files for uvglue accordingly

Note that you can now also use nfbin in gpcal to do frequency binned calibration on the whole spectrum, this avoids the split/glue process.

Cheers,

Mark

[Laura]

Hi Mark
Thanks for the reply!

I tried using nfbin in gpcal rather (as you suggested!) It gives a great image for the central panel of the mosaic I am working with. But the invert map of the surrounding panels are horrible - the maximum flux is about half of what it should be (as compared with the previous images of those panels, that I made using the frequency splitted data and inverting with invert vis='source.????.i' ).

I have been trying to find out what is causing this, but not having any luck so far. Do you have any idea what the problem could be, or suggestions for where to look?

Thanks!

[Mark.Wieringa]

Hi Laura,

No, I don't know how using nfbin in gpcal can affect mosaic fields in different ways. It seems more likely something has gone wrong in the copying or application of calibration tables.
Are all your mosaic data in one visibility file or are there many?

If you are unable to resolve this, please point me at some data on an ATNF server and list the task inputs used, so I can try to reproduce the problem.

In the meantime Jamie says he is going to try nfbin in gpcal on his mosaic observation. We'll let you know what we find.

Cheers,

Mark