Skip to content

reconstruction-pipeline.md

Latest commit

 

History

History
57 lines (40 loc) · 2.6 KB

reconstruction-pipeline.md

File metadata and controls

57 lines (40 loc) · 2.6 KB

Reconstruction Pipeline

The reconstruction pipeline converts raw pad traces into hit clouds, candidate tracks, and fitted track results. Each stage is usually a FairTask that reads and writes branches through FairRootManager.

In the current tree, those branches are usually TClonesArray containers. Tasks typically fetch the branch container from FairRootManager, then read or construct the event object at slot 0.

Shared Early Stages

AtUnpackTask
  │  reads raw GET/GRAW, HDF5, or ROOT data
  └─ output branch: AtRawEvent -> TClonesArray[AtRawEvent]
        ▼
AtFilterTask   [optional]
  │  filters raw traces
  └─ output branch: AtRawEventFiltered -> TClonesArray[AtRawEvent]
        ▼
AtPSAtask
  │  converts traces into hits
  └─ output branch: AtEventH -> TClonesArray[AtEvent]
        ▼
AtDataCleaningTask   [optional]
  │  removes or adjusts hits before pattern recognition
  └─ output branch: AtEventCleaned -> TClonesArray[AtEvent]

The exact branch names and defaults are documented in branch-io-contracts.md.

Pattern Recognition

Two supported paths produce AtPatternEvent → TClonesArray[AtPatternEvent]. Both populate fHitArray on each track and fNoise on the event with unassigned hits. They differ in which AtTrack fields they fill:

AtPRAtask AtSampleConsensusTask
Algorithm triplet/hierarchical clustering (TriplClust default; SetPRAlgorithm(n)) sample consensus (SetPatternType, SetEstimator)
fPattern always AtPatternCircle2D configurable: kLine, kCircle2D, kRay, kY
fGeoRadius, fGeoCenter, fGeoTheta, fGeoPhi filled — via internal RANSAC after clustering not filled (remain zero)
fHitClusterArray populated not populated

The fGeo* fields are the momentum-seed inputs for GenFit-based fitters. AtSampleConsensusTask leaves them empty, so it requires additional processing (e.g. AtPatternModificationTask) before a GenFit-based fitter can run.

Fitting

After pattern recognition, fitting produces the AtTrackingEvent branch container.

Current fit-related tasks include:

  • AtFitterTask
  • AtMCFitterTask

Fitting is unstable, so this page intentionally stays shallow. See fitting-status.md.

Running Reconstruction

A typical macro creates a FairRunAna, adds tasks in order, and calls Run(...).

For runtime objects, see data-model.md. For example macros, see macro-cookbook.md.