See:  http://drupal.star.bnl.gov/STAR/blog/rjreed/2009/sep/22/ppv-central-membrane-crossing-issue for the summary of the issue.

mu->PrimaryVertices()->nCrossCentralMembrane() always returned zero.  In fact, no tracks were considered to have crossed the central membrane.

In version 1.30 of PPV, the central membrane crossing check was done in:

StPPVertexFinder::matchTrack2Membrane(const StiKalmanTrack* track,TrackData &t)

with the last line in the function as:

if(mTestMode!=4) t.scanNodes(hitPatt,jz0); // if central membrane is crossed

Where t is of the class TrackData used in PPV.  The associated function scanNodes takes the vector<int> hitPatt and the number of hits in that the cross appeared to happen, jz0 and checks to see if the track is in time and crossed the central membrane.  If this is the case, t.mTpc is set to 1 for that track.

In version 1.31 this line was removed during code clean up.  Fortunately, this does not effect the 2008 production as it was done with 1.30.  The 2009 production has not been run yet.  The line will be added and back propagated so that any further production done with PPV will be correct.

Checking this against a W triggered daq:

(library is SL09b)

'bfc.C(1,100,"VtxSeedCalG pp2009a ITTF BEmcChkStat QAalltrigs btofDat Corr3 OSpaceZ2 OGridLeak3D beamline -VFMinuit  VFPPVnoCTB -dstout evout","/star/data05/scratch/rjreed/PPV2009/tpccm2/st_W_10095048_raw_1180001.daq")'

I ran over the same file with and without the line to check for TPC CM tracks.

Figure 1: # of TPC CM crossing tracks in index 0 vertex with rank > 0.  This matches what had been seen in 2008.  (See figure 5 of http://drupal.star.bnl.gov/STAR/blog-entry/rjreed/2008/sep/15/ppv-revision-1-29-low-pt-high-luminosity-fmsslow-trigger-evaluation-part-2 )

Figure 2: Vz of the index 0 vertex with rank > 0.  The addition of the TPC CM tracks caused an index 1 and index 0 vertex to switch in 1 out of the 100 events.

Figure 3: Rank of index 0 vertices with rank > 0.  The addition of the TPC CM tracks increased the rank of some vertices, as to be expected.

This fix should affect the minbias triggers more.  However, since no production has been run I do not think an indepth analysis is needed unless requested.