From: Yin Zhongbao (yin.zhongbao@fi.uib.no)
Date: Wed Nov 20 2002 - 04:46:59 EST
Dear brahms-dev, I think it could be done in preloop to evaluate 1/beta-1/beta(calculated) and R-R(calculated). And it would be nice since it gives momentum dependent values. But for high momentum case or some other setting, I am not sure if we have enough statistics, may one fit it with a function when there are enough statistics and give a general function good to extrapolate to low statistic case. My concern goes to the method CalculateMomentum. There may have an assumption that the track has passed through all of detectors in FS and momenta pd2, pd3 and pd4 are reconstructed. Otherwise, for example, if pd4 = 0 it may give a wrong value. Is it not possible to happen? Thanks for discussion. Best regards, Zhongbao On Wed, 20 Nov 2002, Djamel Ouerdane wrote: > > The preloop should make sure it only attempts to determine constants > > (offsets) in region of momenat where these makes sense. One should not try > > to get a MassSq for protons where p>=8 GeV in FS and at lower value in FFS. > > And it seems to me much of this is in fact done better using > > 1/beta-1/beta(calc) for specific species. > > I did that a long time ago and thought it was much better but then, i > thought about this beta(calc), which in fact uses the experimental > momentum, which made me doubt about this stuff. > > Then I did something a bit more complicated : I drew the real expectation > curve 1/beta vs p for each particle specie. Then I evaluated the closest > distance in this particular space between an experimental point and the > curves. I checked this distance as a function of p (note: not the > experimental p), I could then have a "pid" resolution (mixture of beta and > experimental p). > > This worked pretty well but Peter was not confident about it because he > claimed that the difference in the units between beta and p fooled this > more geometrical procedure (I disagree with him since I use basic > equations, minimize the distance with these equations, etc unless > someone clearly demonstrates that it is wrong to do so). > > But then Claus came up with his procedure which is used now in the > bdst stuff, and I never really agreed with it. It worked ok for ratio > analyses but for what we're doing now, it's certainly not optimal, not to > say wrong... > > > Talking about calibrations, there is clearly some issue about its quality. > First, we were very few to do some calibrations and we certainly missed > some, underchecked some, etc. > Second, there is no bookkeeping ala shift reports and this is really the > main issue as far as I see it. It's almost impossible to quickly figure > when, how, by whom, with what data a calibration was done and committed > unless you're an sql hacker (and even so, you cannot retrieve all info you > wish unless the revision string comment details this info). > > As an example of this, I noticed that dY between T1 and T2 for runs 5581 > to 5656 was weird (double peak) while I was checking some matching offsets > 2 weeks ago. I thought then that the vdrift in the DB was somewhat wrong. > I wanted to check who did that and how. I got very little hint from the DB > but I was lucky enough to guess how things were done since this > calibration could have only be done at NBI according to these hints. I > could then ask the concerned people (by the way, this calibration has not > yet been rechecked, I will launch the reduction script on a couple of > these runs). > > Djam > > > > -- ----------------------------------------------------- Zhongbao YIN Phone: +47-55-582792 (O) Address: +47-55-276803 (H) Fantoftveien 14G 466 E-mail: 5075 Bergen Yin.Zhongbao@fi.uib.no ------------------------------------------------------
This archive was generated by hypermail 2.1.5 : Wed Nov 20 2002 - 04:47:50 EST