problem with UVSPLIT(?)

[marcella]

Hi all,
my pipeline of data reduction have stopped working about 20 days ago. However I've verified that the problem is not data related and does not depend on the useage of the pipeline. I realized that the problem is still there even if I try on data that I had already analyzed in the past successfully. The problem does not depend on my pipeline too because it is still there passing the command through the MIRIAD command line (i.e. not running the pipeline). I think it appeared after we run a mirsync to upload MIRIAD. In any case after rerunning mirsync today it hasn't disappeared.

The issue is the following:
After splitting the visibilities and flagging all the bad channels I run MFCAL on a source
mfcal vis=1934-638.18000 interval=0.1 stokes=i
whatever I do on the resulting data is ok up to this stage (e.g. if I uvplt amp vs freq I don't have any error message and everything works, and I don't see any problem neither with UVFLUX or CALRED for example).

Then I have to split my band in 4 channels and usually use UVSPLIT
uvsplit vis=1934-638.18000 maxwidth=0.512
it works creating the splitted files 1934-638.17232 1934-638.17744 1934-638.18256 1934-638.18768 as usual. In the folders there are the
bandpass and gains as usual. However almost whatever I try to do (e.g. try uvplt amp vs freq) on these
data I got the error message
#Fatal Error: Bandpass table size is incorrect. This happens when the inputs for mfcal and gpcal are inconsistent
I get this message at least with uvplt, uvflux and calred applied to splitted vis.
However, I verified that I can do gpboot with a similarly splitted subbands for calibrator and source without any message.
On the other hand, a similar message appears if I split source and URANUS and try to do mfboot at 7mm.

In any case everything works fine for unsplitted files.. for this reason I think it has something to do with uvsplit...

Do you know if anything has changed with uvsplit?

I thank you a lot for any help you can provide, since I really don't have any clue on what else I can try and I need to split the band to get my
results properly calibrated. In any case if it doesn't work any more I have to rethink all the pipeline from the beginning...and I'd prefer to
avoid it... Do you have any suggestion?

Marcella

[Mark.Wieringa]

Hi Marcella,

The message you see is related to the fact that the bandpass solution can now be time dependent. I'm not quite sure why it pops up in your case, but one thing I noticed it that you seem to be splitting the file after calibration, but without applying the calibration to the data first.
I suspect it is related to this. I'm not certain uvsplit can correctly divide up the bandpass table in frequency chunks.
Is it possible to move uvsplit earlier in the pipeline, before bandpass calibration?

Let me know how it goes..

Cheers,

Mark

[marcella]

Hi Mark,
thanks for your reply
is that something recently done?
I usually do the bandpass calibration copy it to the primary and to the targets then split both calibrators and targets and apply the flux calibration (mfboot or gpboot) to the splitted data.
It has always worked, so far...
So you suggest to split the raw data both for calibrators and targets and do all the calibrations (from mfcal on) on splitted chuncks?
I can try...but that means to change all my pipeline and rerun all the calibrations done so far... Can you see any alternative solution?
Cheers
Marcella

[Mark.Wieringa]

HI Marcella,

the quickest solution would probably be to add a uvaver step after the gpcopy of the bandpass, i.e., apply the bandpass to the data and then continue the script as before.
I can have a look at what's going on in more detail later and see if the fatal error can be avoided.

Cheers,

Mark

[MaximVoronkov]

I actually prefer to use uvaver instead of uvsplit in most my scripts (of course, without averaging in time or frequency). It gives a bit better control on what you want to split out.

[rchhetri]

Marcella,
How did you go with using UVAVER?

I am having the same problem as you are. Even using UVAVER did not work for me. The only way it works is that I have to split the data per source first, then GPCOPY my bandpass solutions to the source. It works after that. I tried copying bandpass solutions to visibility file, applying UVAVER and then splitting. I've also tried copying bandpass solution, splitting and using UVAVER. Both the approaches failed with the same sort of output as Marcella's. Mark did you have any luck?

Cheers,
Rajan