[daip] Aips calibration seems to have changed

[Eric Greisen]

Roderick Johnstone wrote:
> Hi,
> 
> We have been reducing some GMRT data for some weeks now, progressively
> cleaning and calibrating the data. After changing the reference antenna
> (to look for a better resoution) and then moving back to the original
> reference antenna the data have become much worse than they were.
> 
> For example, when using the wiper task in IMAG vs REAL, we were getting
> down to a nice circular region of uv data points with a dispersion of
> about 15. Now the data are no longer circularly symmetric and have a
> dispersion of about 40.
> 
> (To do the calibrations we are using a script which takes the original
> data, does all the bandpass and calibrations using the specified
> reference antenna, applies the latest flag table which is saved as a
> .txt file and spits out a uv data set with the new corresponding
> calibration tables....so it does not modify the original data set in any
> way, it just deletes and re-computes the tables attached to it).
> 
> We have done a lot of tests and are unable to get back to the original
> good data, so need some expert advice on where to look for the problem.
> 
> It seems like AIPS is no longer giving the previous calibrated data we
> had. We've tried reloading the uv data set into aips, using previous
> flag tables, calibrating manually without the script, but nothing works.
> 
> Right before re-running the script with the new reference antenna, we
> had to shut down aips (which had been open for some weeks) and
> reboot the computer, but we cant see how that could be the problem.
> 
> The reduction has all been run with the DEC11 version of aips on Linux
> x86_64 (Fedora 14). The midnight job has not been run since starting the
> reduction so we are pretty sure that aips itself has not changed.
> 
> We are keen to find out what went wrong, rather than just start over
> because it takes a long time to flag these data and don't want it to
> happen again, so would appreciate some pointers.

I am going to bcc this to a couple of expert aips users to see if they 
have any suggestion.  Your subject line had suggested something else in 
that the flux models in SETJY for 3C286 and friends has changed.  But 
you have WISELY not run a MNJ and I suggest that you do not until we 
figure something out.

Clearly something has changed and you will need to investigate what

1. Did the flag information get something more that could have deleted 
more data than intended?  Or lose some info and re-insert bad data?

2. Can you compare the old SN tables with new ones?  Old BP tables with 
new ones?  Old CL tables with new ones?  My suspicion is that the 
solutions have gotten raunchy for some reason.  Even if you can't 
compare the old "good" ones to the new "not good" ones, use POSSM (for 
BP) and SNPLT (for CL and SN) to look at the calibrations to see if 
something is odd.

Eric Greisen