gijzelaerr
commented
10 years ago Original comment thread migrated from bugzilla
at 2011-08-10 17:04:11 Oleg Smirnov replied:
Originally reported by Peter Williams:
<<<Hi Oleg,
I've been attempting to use MeqTrees on some of my data and have run into what I think is a bug somewhere in Siamese.OMS.fitsimage_sky. For some cross-correlations, the model prediction derived from the FITS image is wrong, and huge -- 315 Jy. This number is almost exactly the total flux in the model I'm using.
Attached is a small test case. It contains a tiny MS with two consecutive 10-second dumps, and a script (process.py) that generates corrected residuals from the model image using calico-wsrt.py. (You need to give it the path to that script as an argument.) All of the data for the first dump come out fine. For the second dump, many of the higher-numbered baselines come out with residuals of ~315 Jy due to the very large model prediction. The last valid baseline, 38-41, is one example. It's a short baseline, but not that short.
I don't really know enough about MeqTrees internals to debug this. Can you think of what might be going wrong here? I'm using SVN builds for pretty much the whole stack.
Cheers,
Peter
Root cause is null UVWs in the second timeslot (MS values are OK). bugdemo.tar.gz is on oms@cedar:tmp/pw.
at 2011-08-11 17:51:31 Oleg Smirnov replied:
OK, this is a case of baselines to the reference antenna missing in the second timeslot! So not really a bug at all.
On the other hand, we need a more graceful way to detect this failure condition. Can't fix it for R1.2 though, but will rename and leave the bug open for the future.
at 2011-08-10 17:04:11 Oleg Smirnov reported:
Need proper diagnostics for when reference antenna is not fully present