Performance Comparison

The CATS package was extensively tested on simulated data using as references two other track reconstruction packages, RANGER and TEMA. The evaluation procedure, criteria of the evaluation and results are summarized below. All described tests were performed for the HERA-B setup 2000, with ITR and OTR chambers in the MC area.

It should be mentioned that all three packages work within the standard reconstruction framework of HERA-B, which includes such modules as:

• hit preparation for the Inner and Outer tracker,
• pattern recognition in the PC area (can be used: CATS, RANGER, TEMA),
• track prolongation into the MC and TC area,
• usage of SLT tracks as seeds for pattern recognition in PC area,
• matching of track segments reconstructed in different areas,
• track creation (RTRA) and refitting with full knowledge about traversed material.

The track reconstruction strongly depends on the pattern recognition, because results of the track propagation, matching and refitting are strongly depended on the quality of the reconstructed track (fraction of wrong associated hits, number of wrong reconstructed segments).

Tests for the evaluation of the pattern recognition packages are selected according to the physical program of the HERA-B experiment. The physical program is described in Chapter 1. The topics of the physical program made clear that the following set of tracks are important:

• lepton tracks from $J/\psi$ decays ($J/\psi$ prompt or B-meson decays), a high efficiency and good understanding of the ghost rate is needed,
• all tracks which originate from primary decays are important, because the resolution of the primary vertices depends on the number of associated tracks,
• decay tracks from B-decays are important, the efficiencies and ghost rates are important to calculate an efficiency for $J/\psi X$.

Decay tracks from long lived particles were not included in the official test program but they were checked as well.

The number of interactions per event is chosen to be one $J/\Psi \rightarrow \mu ^+\mu ^-$ event and two superimposed Poisson distributed inelastic events with the aim to represent the track multiplicity in the triggered data of 2000. But the performance is checked also for up to four superimposed interactions per event. The sample of simulated data consists of 60.000 minimum bias events and 100.000 prompt $J/\Psi \rightarrow \mu ^+\mu ^-$ events. All events were generated using the same carbon target wire ``Inner 2''.

For the detector simulation two scenarios are defined: generic and realistic, according to the measured performance during the run period in 2000. Table 4.1 summarizes the parameters. For the OTR hit preparation dead and noisy channels of the OTR were masked according to the run 14577, which belongs to the ``Golden Minimum Bias'' data taking period in 2000.

For the tests on data, the ``Golden'' minimum bias run 14577 is used. For $J/\psi$ tracks a sample of preselected data is used. This sample contains about 2000 events with at least one $J/\psi$ candidate per event.

Table 4.1: Parameters used for detector simulation.

	generic	performance of 2000
Hit resolution for OTR, $\mu$m	350	500
Hit resolution for ITR, $\mu$m	150	200
Hit efficiency for OTR, %	95	90
Hit efficiency for ITR, %	90	86
Dead strip simulation	-	according to
		the run 14577