Friday, January 4, 2008

Double Sources-- and not the AGN kind

I'm having this problem with some archival data from 1990. I have two nights of data on a galaxy, spaced a few nights apart from one another. I calibrate the nights separately, and image them separately. It rapidly becomes obvious that the images from the two nights are shifted relative to one another, but not in any simple way. The galaxy center, at the image center, is at the same position in both, but as you go out radially, the first night's sources are further out, radially. It's as if the image is stretched one night, relative to the other.

AIPS thinks that both images have the same pixel scale, so that means a source will have different R.A. and Dec (by about an arcminute, pretty bad!) between the two nights' images. If I DBCON the two nights, I get lots of double sources.

The observational setup claims to be exactly the same between the two nights-- same pointing center, same correlator set-up.

Anyone have any idea what might be causing this? I think I've had a similar problem with GMRT data before, and gave up temporarily and went on to a different project. I'm guessing it's a calibration issue?


amanda said...

Totally weird. The first thought I had was that it was a frequency thing, but you said the correlator set up was the same. I'd double check the header to make sure. If anything else comes to me, I'll let you know.

Laura said...

So, it appears to be a mistake with the archive, or a mistake with how the correlator set-up was recorded at the telescope. In 'observations summary table', the archive claims there are 2 IFs, one at 1465 MHz and one at 1665 MHz. This is clearly untrue, cause in the list of scans, it says all the data is 1465 MHz, and the data was taken in full stokes modes so it must only have one IF.

But the kicker is, the offset is consistent with one day being observed at 1465 MHz, and the other day being observed at 1665 MHZ. Unfortunately, for some reason AIPS thinks it was all observed at 1465 MHz, and it is apparently non-trivial to change what AIPS thinks. I talked to Miller Goss and Miller is going to talk to Eric Greisen, and then hopefully we should have a nice fix for this problem.

amanda said...

Glad to know that a fix is on the way.

However, in my full stokes data, I get two IFs at two different frequencies by default.

Laura said...

Ok, so the solution was a bug in FILLM! For some (probably mistaken) reason, at some point the observers switched from observing in IF AC mode to IF BD mode. I don't really know what this means, but it resulted in them switching from 1465 MHz to 1665 MHz. FILLM, however, didn't realize this. It assigned all data to 1465 MHz.

The solution was just to FILLM the 1665 day separately from the 1465 day. When i did this, the correlator information was properly represented in LISTR.

Miller claims that Eric Greisen has been alerted to this bug, and will try to fix it.

Thanks, Amanda!