50 likes | 96 Vues
Types of Data. SimTracks: generated particles SimHits: energy depositions in a detector volume Digis: Single-channel pieces of the detector’s raw binary output RecHits: Reconstructed energy and position for a single detector elements
E N D
Types of Data • SimTracks: generated particles • SimHits: energy depositions in a detector volume • Digis: Single-channel pieces of the detector’s raw binary output • RecHits: Reconstructed energy and position for a single detector elements • DST Objects: Higher level things, like jets & tracks
Software Components • Generator (such as PYTHIA) • Creates SimTracks • Stores in PAW ntuples in a standard format (HEPEVT) • Simulation (OSCAR) • Swims the tracks through the detector, leaving energy deposits (SimHits) • Fast simulation (FAMOS) • Uses parametrized resolutions to smear the SimTracks, and make higher level (DST) objects • Reconstruction (ORCA) • Simulates detector signals (Digis) from the SimHits • Reconstructs (RecHits) & higher level (DST) Objects • Can run on these objects to make ROOT trees • Visualization (IGUANA)
Starting out with CMS software • First place to start is the ROOT file • Get account on Fermilab analysis machines • Pick your favorite dataset • Single-particle sets are fine for now. • Start with a “Digi” or “DST” dataset, not “SimHits” • Follow the instructions in Eric’s talk to run ExRootAnalysis • Make ROOT plots of everything that might interest you • Pick out the calorimetry clusters in some eta/phi radius around certain type of generated particle • See what kind of energy resolution you’re getting • Can you see brem effects in electrons? • How is energy shared between ECAL & HCAL? • How does it depend on eta? • How many generated tracks are there around this particle? • Do they come from pileup? Underlying event? • How many reconstructed tracks?
Learning about reconstruction algorithms • Try to figure out which algorithms make your favorite reconstructed object • “grep”ping through the codebase is your easiest option • DON’T try to follow the logic of how ORCA actually calls that algorithm • Look through CMS notes for reconstruction algorithms • (don’t expect much documentation for the code) • See what parameters can be varied in the .orcarc • Put some “cout” statements in, scram build, and run ExRootAnalysis again • Run from a “Digi” dataset this time, to make sure everything gets rebuilt. The “DST” dataset will already have the objects, so they won’t get rebuilt.
New Framework • Timescale is spring/summer 2006, for cosmics tests • More rigidly-structured than ORCA • Data is “pushed”, not “pulled” • Stored in ROOT format, so you can plot the data directly from ROOT