[daip] Re: BPASS and flagging

[Eric Greisen]

Robert Braun writes:

 > I seem to be having a problem with BPASS, in those cases where I have FG
 > tables with a large number of entries (more than about 1000). It looks
 > like the resulting BP solutions do not take account of the flags correctly
 > and so do not have the correct normalization across the band. If I run
 > SPLIT first and then redo the BPASS, the problem seems to go away, but I
 > would sooner not have to do all the SPLITing, MULTIing and DBCONing.
 > 
 > I know that SPLIT seems to have a limit of some 5000 flags per time stamp
 > (which I also sometimes encounter as a problem), but it seems that BPASS
 > gets into trouble even sooner.

         UVCOP has the limit increased to 50000 to deal with this
problem.

BPASS uses the same software identically as SPLIT et al to read the
data.  I do not see how there can be any difference.  I will have to
see a great deal more information to credit this claim - the fact that
I take you seriously is not enough in this case.

 > 
 > Are there system parameters I could reset and then recompile in order
 > to circumvent these limitations?
 > 
 > On a slightly related note, it would be really wonderful if AVSPC were to
 > support application of a BP correction while averaging. It is really
 > critical to do the BP first to make a decent "continuum" channel, so I've
 > been forced to jumop through some hoops to make this happen in practise
 > (SPLIT, MULTI, DBCON, AVSPC).

You also seem unaware of SPLAT which would help deal with such issues.
I will add to the wish list to add the calibration package to AVSPC.

Eric Greisen