[daip] SU table after UVAVG ?

[Eric Greisen]

Nissim Kanekar wrote:

> 
> Here's the PRTAB output of the NX table of the .UVAVG file. I didn't do 
> any time-dependent flagging before running UVAVG; just flagged out dead 
> antennas and then ran FRING to correct delays, before running UVAVG. Of 
> course, it's possible that the correlator might have produced zeros at 
> the start time for some baselines, which would have gotten flagged 
> automatically in FITLD.
> 
> I guess a gap might be defined as a timerange significantly larger than 
> the record time (e.g. 3*record time or more), over all antennas. The 
> record time might be a required entry in a new CPARM element in INDXR 
> (or can it be read directly from the header ?).

The problem is that the routines did see times that differed by more 
than 0.2 sec and so took that to mean that the normal interval was ~0.2 
sec.  A 5 sec interval is then 10 times that and so a scan break.  I 
changed the routine to require 2 gaps before being willing to decide 
about normal gaps and rearranged things to allow the calling task to 
provide information.  UVAVG does know something and so can provide a 
better suggestion than the 0.2 sec.  The MNJ tomorrow should bring this 
change.

Your NX table showed that some "scans" had a non-zero length esp the 1st 
two.  That is what confused the task and was probably due to a ragged start.

Hopefully this will be better but there is no perfect solution other 
than to carry around an integration time on everything which we decided 
long ago not to do (although CASA is trying to override that decision).

Thanks,

Eric