Bandpass solutions for autocorrelations

Got a calibration problem? Discuss it here.

Moderator: Mark.Wieringa

Post Reply
jor061
Posts: 3
Joined: Mon Sep 08, 2014 5:52 pm

Bandpass solutions for autocorrelations

Post by jor061 »

Hi all,

I'm looking for some help to get a spectral line out of my data. The short story is: I've run atlod with nocross, and wish to use mfcal to solve for a bandpass solution, but mfcal complains that it cannot find any correlations.

The longer story: I've got some data of maser regions that have zooms positioned on more spectral lines than just the maser. As the array configuration is rather large, and other spectral lines I'm interested in are extended, I expect that emission is "resolved out." I want to do a quick data reduction for just single dish data, which would be neat to show detection via uvspec or somesuch. The subsequent spectra might be useful in a publication.

Any ideas?

Thanks,
Chris
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: Bandpass solutions for autocorrelations

Post by Mark.Wieringa »

Hi Chris,

mfcal tries to determine bandpasses by factoring the antenna gains & bandpasses out of the visibilities. Without cross correlations it can't do this. If you have only autocorrelations, you need an on-off type scheme to remove the bandpass. I don't think Miriad provides this - single dish reduction packages do. Alternatively, you can load all the data, solve for the bandpass normally and then plot the auto correlations - they will be bandpass corrected, so you should be able to see features due to large scale emission in the spectra.

Cheers,

Mark
jor061
Posts: 3
Joined: Mon Sep 08, 2014 5:52 pm

Re: Bandpass solutions for autocorrelations

Post by jor061 »

Hi Mark,

Thanks for your help. As per your advice, I have loaded in all the data (i.e. "nocross" and "noauto" are not specified to atlod), then proceeded with a normal calibration. However, it appears that bandpass solutions were not applied to the auto-correlations; see the attached figure. Note that a channel spike around 2049 is due to stitched CABB zooms.

Any idea if this is still possible, or I'm doing something wrong? The history for one such source is as below.

Code: Select all

ATLOD: Miriad atlod: Revision 1.47, 2014/01/15 04:28:11 UTC
ATLOD: Executed on: 2014-09-09T11:13:32.0
ATLOD: Command line inputs follow:
ATLOD:   in=../../../RAW/METHANOL/DAY2/*
ATLOD:   out=atlod.if08
ATLOD:   ifsel=8
ATLOD:   restfreq=43.42376
ATLOD:   options=rfiflag,birdie,xycorr,opcorr,unflag
...
QVACK: Executed on: 2014-09-09T11:58:10.0
QVACK: Command line inputs follow:
QVACK:   vis=atlod.if08
QVACK:   interval=0.1
QVACK:   force=0
QVACK:   mode=source
ATFIX: Miriad atfix: Revision 1.6, 2013/09/09 03:03:04 UTC
ATFIX: Executed on: 2014-09-09T11:58:10.0
ATFIX: Command line inputs follow:
ATFIX:   vis=atlod.if08
ATFIX:   out=atlod.if08.atfix
UVSPLIT: Miriad uvsplit: Revision 1.15, 2013/08/30 01:49:21 UTC
UVSPLIT: Executed on: 2014-09-09T12:00:14.0
UVSPLIT: Command line inputs follow:
UVSPLIT:   vis=atlod.if08.atfix
UVSPLIT:   select=-ant(6)
BLFLAG: Miriad blflag: Revision 1.20, 2013/05/23 03:43:33 UTC
BLFLAG: Executed on: 2014-09-09T13:13:15.0
BLFLAG: Command line inputs follow:
BLFLAG:   vis=ch3oh_12.43440
BLFLAG:   stokes=xx
BLFLAG:   options=nobase
BLFLAG:   device=/xs
BLFLAG: Number of correlations flagged: 22968
BLFLAG: Miriad blflag: Revision 1.20, 2013/05/23 03:43:33 UTC
BLFLAG: Executed on: 2014-09-09T13:13:18.0
BLFLAG: Command line inputs follow:
BLFLAG:   vis=ch3oh_12.43440
BLFLAG:   stokes=yx
BLFLAG:   options=nobase
BLFLAG:   device=/xs
BLFLAG: Number of correlations flagged: 11484
GPCOPY: Miriad gpcopy: Revision 1.11, 2013/11/21 01:28:24 UTC
GPCOPY: Executed on: 2014-09-09T13:29:52.0
GPCOPY: Command line inputs follow:
GPCOPY:   vis=../CALIBRATORS/1646-50_c.43440
GPCOPY:   out=ch3oh_12.43440
EDIT: Sorry, this should've been the history for the bandpass calibrator, which is below.

Code: Select all

MFCAL: MfCal: $Revision: 1.16 $, $Date: 2014/04/29 00:50:17 $
MFCAL: Executed on: 2014-09-09T13:29:40.0
MFCAL: Command line inputs follow:
MFCAL:   vis=1253-055.43440
MFCAL:   interval=0.1
MFCAL:   refant=2
Apologies if I'm being naive and this is an obvious solution, I've been unable to find help elsewhere. Unfortunately, these observations did not record a reference scan (for an "on-off" technique), but I'm optimistic that a bandpass could be removed from these auto-correlations.

Cheers,
Chris
Attachments
2014-09-09-134257_850x988_scrot.png
2014-09-09-134257_850x988_scrot.png (39.52 KiB) Viewed 16217 times
Mark.Wieringa
ATCA Expert
Posts: 297
Joined: Mon Feb 08, 2010 1:37 pm

Re: Bandpass solutions for autocorrelations

Post by Mark.Wieringa »

Hi Chris,

I just gave this a try - applying bandpass calibration to data with auto correlations - and it looks to me that it does apply the calibration, but it may not give the results you want. Any signal in the autos that is absent in the cross correlation bandpass will remain, this includes the most of the Tsys structure like the slope on 1-1 and 2-2. The ripple on 3-3 is a bit of a mystery, although we do see similar effects at lower frequencies.
I think we're seeing the reason there are separate single dish packages and observing modes.

One other thing to try is the task uvgains - it claims to do a lot of things with auto correlations and bandpasses, including polynomial fitting of baselines etc., but I've never tried it.

Cheers,

Mark
jor061
Posts: 3
Joined: Mon Sep 08, 2014 5:52 pm

Re: Bandpass solutions for autocorrelations

Post by jor061 »

Hi Mark,

After being distracted by other things for a few months, I've finally had time to revisit this. Your suggestion of uvgains was a good one!

For reference, my procedure was:

Code: Select all

uvgains vis=<dataset with only autocorrelations> stokes=i out=<something> options=window,polyfit device=/xs npoly=2,2 endchan=100
I'm not sure if the bandpass calibration being copied to the source file before uvgains will affect things or not.

Following this, a bright spectral line is visible in uvspec with options=nobase,avall. This is a little unfair on the data, so I will script uvspec for each antenna before subtracting continuum and stacking.

Cheers for your help!
Chris
Post Reply