Trackers evaluation : Reconstruction versus Simulation
Data samples:
- y2008e_dAu200
- Comments:
- Reconstruction
- Sti has global track χ2/NDF dependence on of fit.points, for Stv this dependence much weaker.
- Both Sti and Stv have χ2/NDF increasing with no. of bad points. For global tracks Stv has
χ2/NDF = 1.5 and Sti χ2/NDF = 1 for 0 bad points.
- Increment of χ2Vx
in Sti for including Vertex ~3 corresponds 3 measurements coming with vertex (o.k.).
- Sti and Stv have opposite dependence of χ2 on η.
This possibly reflects different hit error parametrization (?). The same is true for primary tracks.
- χ2 dependence on pT for both global and
primary tracks
supposes that accounting for multiple scattering has problem in Stv.
- The difference 1/pT between RC and MC for Sti shows
degradation of 1/pT resolution when no. of bad fit points > 4.
- The difference tan(λ) between RC and MC for Sti shows
resolution degradation for no. of bad fit points > 2. Stv is more stable.
- ΔDcaXY shows track charge dependence versus η(?).
- ΔΨ shows track charge dependence versus η which has been shown for DcaXY(?).
- The above charge dependencies are mainly due to low pT (< 200 MeV/c) tracks.
- The Sti pull for Ψ has very strange bump at No. of fit points ~ 25.
Stv does not have it. The pulls for Sti are ~1, for Stv ~3 i.e. Stv significantly underestimates errors in Ψ
- The Sti pull for η has the same strange bump at No. of fit points ~ 25.
Stv does not have it.
- The Sti pull for primary track η has very strange η dependence.
Stv reproduces this dependence with magnification factor ~6.
- There is no significant difference for track parameters by comparing "All" MC particles and &pion; only.
- Efficiencies
- Geometrical acceptance:
- Tpc geometrical acceptance within |η|<1 and pT > 0.11 GeV/c region for
global tracks is ~70% for positive and ~80% for negative ones,
respectively.
- Tpc geometrical acceptance within |η|<1 and pT > 0.11 GeV/c region for
primary tracks is ~90% for both positive and negative ones.
- 1 MC track matched with 1 RC track
- Global track reconstruction efficiency (which includes geometrical acceptance) shows that
- Sti has higher efficiency than Stv by 15%.
- CA for both cases increases efficiency by a few %.
- Stv shows that the difference in efficiency is mainly coming
from degradation of Stv for positive tracks with higher momentum => wrong Multiple scattering.
- Primary track reconstruction efficiencies for Sti and Stv are very close.
The difference is coming from low pT tracks mainly. CA increases efficiency by a few %.
- Global and primary track
reconstruction efficiency (with respect to track within geometrical acceptance)
shows the same behavior as above (with scale 1/0.9).
- 1 MC track matched with > 1 RC track (Clone)
- Sti has ~5% of clones. Stv practically does not have them.
- The Sti clones are mainly coming from loopers (pT < 200 MeV/c);
- These loopers look more prominent for primary tracks.
- 1 MC track has no a match with RC track (Lost).
Over all losses for global tracks for Sti is ~10% and for Stv is ~35%.
The over all losses for primary tracks are ~5% for Sti and ~10% for Stv, respectively.
The losses with respect to geometrical acceptance
for primary tracks are ~8% for Sti and ~14% for Stv, respectively.
- No MC track has match with RC track (Ghost).
Both Sti and Stv have ghost tracks for η > 0.5 with reconstructed pT > 2 GeV/c.
In Stv rate of ghost track is a factor of 2 less than in Sti.
- y2009c_AuAu11
- Summary: There are no significant differences with respect to dAu200 sample.
- y2009c_pp500_pileup
- Comments: There is no big sense to compare global tracks with MC with pile-up.
- Questions:
- There is a noticeable difference in χ2Vx distribution versus No. of fit points for dAu200 and
pp500. For the primary track fit in pp500 it is used only track with more than 20 fit points.
χ2 is a factor 6 in pp500 greater than in dAu. Wrong hit errors ? Wrong vertex errors ?
- Sti/StiCA shows a better momentum resolution versus no. of fit points and
momentum
for primary tracks than Stv/StvCA.
- Efficiencies
- MC track matched with 1 RC track (Unique Match):
StiCA shows significant increase efficiency with respect to Sti.
The Stv and StvCA efficiencies are lower that Sti.
- MC track matched with 0 RC track (Lost): 30% loss for Sti and 40% loss for Stv.
My guess is that higher rate of lost tracks in pp with respect to dAu200 (~10%) is due to
wrong assignment tracks to the primary vertex.
- No MC track has a match with RC global track (Ghost):
This distribution has a very strange pattern with a step at η = 0.5 (?).
- y2009c_pp500W_pileup
- y2010c_AuAu200
- y2011_AuAu200
- Questions: The list of questions is the same as for y2010c_AuAu200.
- Pulls.
- DCA.
Conclusions
- Sti shows some strange dependencies in track parameters and their pulls. But these dependencies are within a reasonable margin.
CA improves efficiencies.
- Stv. We have to keep in mind that this is the first massive test of Stv. Stv is working and producing results. This is a big step forward.
- But Stv does show weird dependencies in track parameters. This demonstrates that there are problems with material accounting.
The main goal of Stv is the proper material accounting. The reconstruction efficiency also show there are problems with fit
(The pattern recognition done by CA is the same for both Sti and Stv).
Yuri Fisyak
Last modified: Mon Nov 21 12:18:35 EST 2011