Physics Reconstruction Documentation

Overview

The HPS physics reconstruction is implemented within the HPS Java project as a chain of org.lcsim Drivers (event processors) which process LCIO Data Format events that are accessed through the EventHeader API.  The Drivers are described by an lcsim xml steering file, which is read by the lcsim job manager.  Detector Conditions including per channel calibrations are read and applied in the initialization phase.  The data from each event is processed to produce additional output collections containing reconstructed objects such as tracks and calorimeter clusters. 

First, the raw data must be converted from EVIO Data Format using an LCSimEventBuilder such as LCSimEngRunEventBuilder.  The EvioToLcio command line tool is used to read EvioEvents with an EvioReader and then uses the event builder to create LCIO raw data events.  The full Driver chain is run on the raw data collections to perform the physics reconstruction.  These combined events are written out to an LCIO file, which can subsequently be converted to ROOT DST Data Format.  The LCIO events may also be loaded back into HPS Java for analysis. 

Basic Recipe

This is the basic recipe for reconstructing HPS data:

1.  EVIO data files are read in and converted to LCIO events containing the raw data collections.
2. The HPS Java reconstruction, defined by a list of Drivers in an XML steering file, adds additional reconstructed object collections to the events.
3. The events are written to an LCIO file containing the results of the recon.
4. The output LCIO file is converted to ROOT DST Data Format for analysis.

Steps 1-3 are typically performed in the same job (process) using the EvioToLcio command line utility.

Reconstruction Drivers 

The reconstruction Driver chain is defined in production steering files such as EngineeringRun2015FullRecon.lcsim, which are kept in this SVN folder and typically accessed as a class resource from a jar file.

Data Collections

Algorithm Details

Data Conversion

The LCSimEventBuilder defines an interface for converting from EVIO to LCIO events, with the LCSimEngRunEventBuilder providing the current implementation of this conversion process.  EVIO collections are processed by a reader which gets raw bank data and converts it into a typed LCIO collection.

SVT data banks are handled by an SvtEvioReader and converted into RawTrackerHit and GenericObject collections.

Various modes of EVIO ECal data are converted using the EcalEvioReader.

The default builder will also convert and write DAQ config information, EPICS data, and scaler bank data into the output LCSim events, if these banks are present in the EVIO data.

Track Reconstruction

These are the steps in the HPS Java track reconstruction:

1. RawTrackerHitFitterDriver converts SVT raw data into fitted hits collection by fitting ADC vs time signals.
2. DataTrackerHitDriver creates stereo hit pairs from the fitted strip hits.
3. HelicalTrackHitDriver creates 3D hits (clusters) from input stereo hits.
4. TrackerReconDriver runs track finding on the 3D hit collection.
   1. Track finding is run multiple times with different tracking strategy files, creating a track collection for each strategy.
5. The MergeTrackCollections Driver is used to merge the multiple track collections into a single output collection.
6. The GBLRefitterDriver refits the tracks using GBL.
7. TrackDataDriver adds a Generic Object collection containing additional information about the track for persistency.

The tracking packages in lcsim form the basis for HPS's tracking algorithms through usage and extension.  Seed Tracker is used for track finding using a set of input tracking strategies.  The track fit from lcsim is further refined using a Java implementation (port) of the GBL C++ algorithm.

Additional References

This paper describes the LCIO track parameters.

These slides provide some details about how tracking strategies are used (see pages 4 & 8).

Cluster Reconstruction

These are the basic steps of the ECal reconstruction:

1. EcalRawConverterDriver converts RawTrackerHit input collection into CalorimeterHit collection using the EcalRawConverter.
2. ReconClusterDriver uses the ReconClusterer to create calorimeter Cluster collection from input hits collection.
3. CopyClusterCollectionDriver copies the clusters (with raw energies) to a different collection.

The copied collection will be updated with corrected energies in the step which creates recon particles.

Additional References

This CLAS Note describes the basic clustering algorithm.

This HPS Note covers position corrections and other analysis.

Reconstructed Particles

The ReconParticleDriver creates ReconstructedParticle objects representing the final state particles from the event reconstruction.  These are tracks with matching clusters (when applicable).  It also performs vertex reconstruction and creates a number of candidate particle collections.

The ReconParticleDriver is sub-classed by the actual HpsReconParticleDriver from the steering which adds Møller candidate collections.

Java Packages

HPS Java Reconstruction Packages

The  HPS Java Documentation can be used to browse the packages and classes used for physics reconstruction.

LCSim Packages

These lcsim packages are used extensively within the HPS Java reconstruction code.