L3 Tracking FAQ

This document is intended to provide the end user of the tracking algorithm, such as a physicist running D0TrigSim, with a guide to jumping the temporary technical hurdles required for running correctly in various environments. These hurdles arise from issues in handling the geometry and calibration for clustering the data from the Central Fiber Tracker and Silicon Microstrip Tracker.

The need for this document underscores the importance of a long-term solution which would require no knowledge in the consumer.

Please send me corrections/suggestions/clarifications...

danielw@fnal.gov
August 01, 2002

This page contains the following useful items:

Brief description of L3 tracking algorithm
Common problems encountered
Cookbook for running on data or MC
Track quality macro and reference plots for data and MC

Overview of tracking algorithm

The L3 Global Tracking algorithm is described in detail in D0Note 3808. Please see that document for a thorough understanding of the tracking and its performance in MC events. A note describing its performance on real data is in preparation. There are two central features of the tracking algorithm:

Each request for track finding is accompanied by a specified minimum P_T threshold, above which maximum efficiency is desired and below which efficiency may be sacrificed for speed.
The tracking is performed in two stages: axial tracking in both CFT and SMT detectors, followed by a search for stereo tracks for those axial tracks which have a PT above the threshold .

The tracking begins with track candidates formed with pairs of clusters in the outer two layers of the CFT. These candidates are propagated toward the center of the detector, through each axial layer of the CFT and the SMT barrels. At each layer, the track is allowed to pick up hits within a cone (the size of which is determined by the PT threshold) or to miss hits in that layer.

Axial tracks are kept if they have either

8 CFT hits
7 CFT hits plus at least 2 SMT hits

For those axial tracks with a PT larger than the threshold, a histogramming algorithm is used to search for stereo information in both the CFT and SMT. If no stereo information is found, the track is kept as an axial-only track.

Common problems

If you have problems with the tracks at level 3, please run the reference macro and compare the plots to those below. You may be suffering from one of these easy-to-fix problems:

Problem	Most of my tracks have only 8 hits
Cause:	The geometry or calibration files are incorrect or missing. Make sure you've carefully followed the cookbook for data or MC

Problem:	My track phi distribution has big holes!
Cause:	You may be using a MC setup on real data. Specifically, make sure that the REALDATA parameter is set to TRUE for data. Otherwise, it may represent detector problems.Check your CFT configuration for data, or try a different run.

Problem:	I see lots of axial only tracks: spikes in Z or tan(lambda)
Cause:	If you've ensure that the system is configured correctly (check the distribution of the number of hits on the track, and see above if it peaks at 8 hits. Otherwise, this may be the desired performance -- many axial tracks are found below the threshold and no stereo tracking is performed. Check the correlation between the pt and the stereo information

Problem:	My tracks don't match my electrons/muons/etc
Cause:	If you're using the MC geometry on a data file or vice versa, then you'll see this problem. To confirm it, you will see very few SMT hits because the CFT is rotated with respect to the SMT and the rest of the geometry.

Problem:	My efficiency is very low!
Cause:	If you've ensured that the system is configured correctly (check the distribution of the number of hits on the track, and see above if it peaks at 8 hits), then make sure that you're calculating your efficiency in the CFT fiducial region. Remember that unlike offline tracking, online tracking is only performed in the central region, to an eta of roughly 1.1.

Problem:	My DCA resolution is very large
Cause:	The beam spot is shifted with respect to the center of the detector, so that the tracks will typically not come from (0,0). Unlike the offline tracking, the track parameters are always calculated with respect to the detector (0,0), so a plot of track DCA versus track angle will show a sine wave (see the reference plots). If you fit this sine wave and calculate the DCA relative to the beam spot, you should be able to get a DCA resolution of ~65 microns for tracks above 1 GeV.

Problem:	My Pt resolution is poor
Cause:	The Pt resolution is greatly enhanced by the addition of SMT information. Make sure that you have correctly configured the SMT so that the most likely number of hits on a track is 10 or 11 rather than 8 (see above).

Cookbook for running over MC

Specify the correct geometry. To do this, you need to know the version which was used to create the MC file. This must be specified in the L3 Geometry Management tool. Your level3.sim file (which is generated from the trigger list) must be configured as such (this example uses the tag p10.06.01):
```
 GEO L3GeometryManagement(MCTag="p10.06.01")|
 CFTUnp   L3TCFTUnpack(realdata=FALSE)|
 SmtUnp   L3TSmtUnpack|
```
Use the correct calibration. For MC files, only the SMT needs to be calibrated, and the correct file is automatically selected from the package l3fsmtunptool.

Cookbook for running over real data

To run over real data, you must do the following:

Specify the correct geometry. In the level3.sim file, the geometry and clustering tools should appear as below. This is the configuration for all triggers run online:
```
GEO L3GeometryManagement(
 		tooltype="utility", 
 		RunNo=-1, 
 		MCTag="xxx.xx.xx")|
CFTUnp L3TCFTUnpack(
 		tooltype="unpack",
 		adcthreshold=20,
 		maxfibers=8,
 		realdata=TRUE)|
 	SmtUnp L3TSmtUnpack(
 		chanthreshold=.03,
 		maxstrips=16,
 		adcthreshold=10,
 		clusterthreshold=.03,
 		tooltype="unpack")|
```
Note that "RunNo=-1" will give you the latest geometry available. If you instead insert a specific run number (like 123456), then it will use the geometry from the DB which was available for that run.
Specify the correct map and calibration information. For running on real data, one needs to have calibration/mapping information included in the L3 calibration file, called "l3calib.ev."
The CFT and SMT information must be merged by hand into one file. This file can be obtained from cvs: gtr_l3/bin/l3calib.merged.cftstereo2.ev and must be copied into your "trigsim" directory if you are using runD0TrigSim, or into your local directory if you are using another executable:
```
 $ cp /d0dist/dist/packages/gtr_l3/devel/bin/l3calib.merged.cftstereo2.ev ./l3calib.ev
```

Macro and reference plots

A simple ROOT macro to produce simple diagnotic L3 tracking plots.

Reference plots for data: run 160588

Note some important items:

The number of hits peaks at 10, indicating good SMT and CFT geometrical configuration and calibration
There is a small dip in phi near 3.5, this is due to problems with the detector/readout.
The dca vs phi plot shows a clear beam displacement.

Reference plots for MC: j/psi

From Abid Patwa and Avdhesh Chandra
Note some important items:

The number of hits peaks at 12, indicating good SMT and CFT geometrical configuration and calibration
The phi distribution is flat.
The dca vs phi plot shows no beam displacement.