The rapidity distributions of pi- normalized to 1 LVL1 event, efficiency corrected, differ by 30% between two analyses:
one made by Alexander Schmah based on velocity-momentum and energy loss-momentum, standard cut analysis (marked "schmah" in the picture)
and one made by Jehad and Pavel based on "Bayesian" PID using velocity-momentum (marked "tlusty")

• ArKCl pi- rapidity:
  

I suggest to compare the numbers of pi- before and after efficiency corrections for "safe" region mom>125MeV/c and theta>40 deg. In this region there are 50% of pi-, and they have hit in TOF (so we avoid problems with TOFINO multihits). As seen, this region covers the backward rapidity part, see next figure. Our rapidity distributions are symmetric, so obviously we do not have problems with TOFINO anyway.

In following I show figures and numbers for "Bayesian" analysis based on velocity-momentum PID, in previous picture marked "tlusty":

• pt vs rapidity for pi- in ArKCl (efficiency corrected), Contours in polar angle are shown in red (10 deg step) and black (2 deg step).:
  

The table shows track and pi- multiplicities per one LVL1 event (M16 trigger). For comparison I show also results of analysis of UrQMD events - dsts with M16 trigger emulation:

Multiplicities per one LVL1 M16 event:
                                          data            !UrQMD
charged track                            10.1             10.8
negative track                            1.40             1.60
negative track & theta>40                 0.667            0.793
negative track & theta>40 & mom>125       0.647            0.778
identified pi- & theta>40 & mom>125       0.622            0.757
identified&true pi- & theta>40 & mom>125                   0.703 PID purity = identified&true/identified = 0.928% due to pi- in-flight decay
true pi- & theta>40 & mom>125                              0.704 PID efficiency = identified&true/true = 1.0

identified pi- & theta>40 & mom>125 &&
efficiency corrected                      1.01             1.22

Track selection: chi2_RK<1000 && chi2_mdc0>=0 && meta_quality<2.5. Same condition was used for track efficiency calculation.

Track efficiency: from simulation using UrQMD events -> Geant -> DST with M16 trigger
For pi- with mom>150 and theta>40 the efficency is 0.62 including missing acceptance due to dead space behind magnet coils (~15% dead space), so the real tracking efficiency is 0.62/0.85=0.73.
Normalization to one LVL1 M16 event: in one hld file there is typically 190k events, from which 30% is LVL1&M16. We did an additional subtraction of 3.6% "empty target" events, so we select about 54k events for pion analysis.

The numbers in the Table are consistent with points shown in the rapidity distribution in the first figure. For the rapidity distribution the missing low pt region was included by extrapolation of the measured distribution, the correction is mostly less than 10%.
The observed difference can be due various reasons, the first candidates are the normalization to one event, or efficiency corrections.

Following pictures show that pi- identification in selected region is easy, 96% of all negative tracks are pi-:
Velocity distribution of negative tracks and identified pions is shown for 200 < mom < 300 and for mom > 125, for both data and UrQMD.

• Data: pi- velocity for theta>40 and 200<mom<300, black - all negative tracks, green - identified pi-:
  

• Data: pi- velocity for theta>40 and mom>125, black - all negative tracks, green - identified pi-:
  

• UrQMD: pi- velocity for theta>40 and 200<mom<300, black - all negative tracks, red - identified pi-, green - true identified pi-:
  

• UrQMD: pi- velocity for theta>40 and mom>125, black - all negative tracks, red - identified pi-, green - true identified pi-:
  

-- PavelTlusty - 23 Jun 2009