Re: [BLAST_ANAWARE] wch efficiency

From: Chi Zhang (zhangchi@MIT.EDU)
Date: Wed Aug 04 2004 - 21:53:45 EDT


Hey pal,

Thanks for trying to save me the embarrassment. do believe I myself
put the code that way. but if you are going to take the bullets for me,
you are welcome. :) Anyway, it s fixed and we shall see what
impact it has on rates and stuff. And I had only to replace two lines in
lrn_event.cc.

BTW, this is the only change made to blast/cvs_v3.0/BlastLib2 since July
23.

Chi

On Wed, 4 Aug 2004, vitaliy ziskin wrote:

> Chi,
> Don't be so hard on yourself. I believe that I might be the one who did
> that. I was under impresion that all neutral tracks are done first.
> Then they are followed by charged once with the qwl or qwr overwritten
> if a charge track is found. Perhaps this needs to done that way, even
> if my first stab at it was an awkward one.
>
>
> Cheers, Vitaliy
>
> Chi Zhang wrote:
>
> >Hi,
> >
> >After Genya showed me the data, I found that there IS a bug in the way
> >ntuple is filled. When NC/LADS are fired, qwl/qwr is set to 0, charge
> >info from WC tracks are overwritten. This is certainly incorrect behavior.
> >Sorry I screw up this one.
> >
> >I fixed the code in head version and crunched run 9719 in
> >~blast/cvs_v3.0/BlastLib2_dev. The problem is gone.
> >
> >To estimate the damage: for run 9719, please see the table below:
> >charge of track left(p>0.1) righ(p>0.1)
> > - 4721 3437
> > + 25933 15817
> > 0 1619 5144
> > 5% 21%
> >the last line is the number of wrong charge over the sum of three rows
> >above.
> >
> >on left sector, 5% of charged tracks had qwl overwritten to 0, on the
> >right, 21% of them were affected. Also, compare e-right to e-left case,
> >one also finds 27% difficiency.
> >
> >I hope this explains the huge difference in tracking efficiency Tancredi
> >sees(e-right 50%, e-left 70% compare to pure TOF cuts).
> >
> >I hope this also explains the missing ed events with electron going
> >right but I need to see also.
> >
> >The bug is fixed in ~blast/cvs_v3.0/BlastLib2, lrn, lrd and
> >libBlast.so are recompiled and copied to bin(lib)/lrn_v3.0c (lrd_v3.0c and
> >libBlast.so_v3.0c respectively) and then linked to lrn (lrd, libBlast.so).
> >
> >the run currently going on is 9801, it will be crunched with new code.
> >
> >Thanks to Genya who spotted this problem.
> >
> >Chi
> >
> >
>
>



This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:31 EST