1 / 25

Calorimeter Calibration Workshop

G.Unal 19/10/2006. Calorimeter Calibration Workshop. 3 days early September, ~65 talks http://indico.cern.ch/conferenceDisplay.py?confId=a062339 Review everything from ADC’s to final calibrated energies (electrons, photons, jets, Etmiss) Including: Trigger Use of in-situ tools

latashah
Download Presentation

Calorimeter Calibration Workshop

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. G.Unal 19/10/2006 Calorimeter Calibration Workshop • 3 days early September, ~65 talks • http://indico.cern.ch/conferenceDisplay.py?confId=a062339 • Review everything from ADC’s to final calibrated energies(electrons, photons, jets, Etmiss) • Including: • Trigger • Use of in-situ tools • Plans for CDC (Calibration Data Challenge) • Test-Beam data/MC comparisons • Only some aspects discussed in this “summary”

  2. (simplified) overall picture: L1 trigger Detector FrontEnd: ADC Pulse reconstruction Electronic calibration EM scale Online ROD: RawChannel (GeV) L2 trigger Cell: Energy (GeV) TopoClusters Towers EM clusters (sliding window) Weights DM corrections Weights (per cell or per layer) Calibrated TopoClusters Specific corrections Electrons/Photons Jets/Etmiss ( & tau’s)

  3. Zoom on jet part… (from P.Loch @ Barcelona)

  4. Electronic calibration • E = F.ai(ADCi-Ped) • Ped = Pedestal • F = Conversion from ADC to uA * A to MeV (A = initial current of physics signal) • ai = “optimal filter” coefficients: estimate peak amplitude from 5 samples, minimizing the noise contribution as well as time jitter effect. • Need knowledge of pulse shape for physics events (the tricky part) • Need also noise time autocorrelation function (easier, should include pileup at high Lumi) • Beware of cross-talk effects in the strips of the EM calorimeter

  5. Mphys Mcsli Difficulty:Physics and calibration signal see differently the detector cell (L, C) => different pulse shape => bias on amplitude (not the same for all cells!) This has to be taken into account to reach <1% constant term Need to be able to predict physics pulse shape (M.Delmastro talk)

  6. EM energy scale for cells • Extracted from Test-Beam (electrons) • MeV/nA accounts for energy to current conversion and sampling fraction • When using sampling fraction from G4.8, the energy to current conversion from the data is close to what is expected from the simulation (including detailed modeling of electric field in the accordion folds) • Main uncertainty when extrapolating to Atlas: • Knowledge of absolute LAr temperature: • Response variation is ~2% / Kelvin

  7. Example of Data/MC comparison for EMEC Gap varies with radius 9 HV region to ~ equalize average response with eta Outer Wheel (ECC0) Inner Wheel (ECC5) (normalized) HV default ??? ( ~50 V / 1200 V ) V.Niess

  8. Electron/Photon calibration • Use fixed size clusters of electrons and photons • For instance 3x7 cells for electrons in barrel (in middle cells size) • Several corrections have to be applied to “raw” cluster energy to estimate true energy of incident particle • Energy lost upstream the calorimeter • Lateral leakage outside the cluster • Downstream leakage (at high energy / low eta) • Modulation of energy with position within the cell from accordion geometry effect + variation of containment

  9. Leakage correction • Derived from MC. Typical formula: • Erec = (a + b.Eps + E1+E2+w3*E3) • Either “simple” minimization of (Erec-Etrue) (with a,b,,w3 only function of eta) • Or more detailed analysis of energy lost in various dead material layers, correcting effect by effect (parameters are also function of energy or shower depth) • Different corrections for electrons and photons (new in 12.0.3) • First method currently implemented in athena for Atlas simulation (also applied to CTB) • Short term: need new constants for 12.0.4 • Second method developed on test-beam and also studied for Atlas simulation. Probably more accurate at the end • Compare both methods with CSC samples • Required MC accuracy for 0.1% non-linearity above 10-20 GeV • O(10%) X0 for upstream matter • <5% X0 for matter between presampler and strips => CDC simulation with added extra-material

  10. (preliminary) Performances for “unconverted” photons 11.0.4x simulation, 12.0.3 reconstruction (to be re-done with 12.0.3 simulation) (<E>-Etrue)/Etrue vs  (E)/E vs  25 GeV 50 GeV 100 GeV 200 GeV Medium term work:understand how to treat “converted” photons L.Carminati talk

  11. Zee intercalibration studies 384 regions in -. Inject ~2% region to region variations Use Z mass constraint to recover intercalibration between regions Effect of “imperfect calibration” (11.0.4 simulation and reconstruction => old “Rome” corrections for dead matter) From ~70K Z (O(150pb-1)), ~0.4% stat. accurary on intercalibration/region. But bias from absorbing wrong DM correction: ~1% effect With more stat, Z could also be used for non-linearity studies, … K.Tschann-Grim talk 384 zones

  12. Hadronic calibration • Jets are complicated objects • Should avoid large tails, get correct mean, optimize resolution • Current athena “default”: • Jets are made from tower • (Or towers in TopoCluster to reduce noise, studies of efficiency shown at Barcelona) • Different jet algorithms (cone, Kt, etc…) not discussed here • Weights applied to “restore” compensation and absorb dead matter effects (minimizing resolution) • Compare reconstructed jet and “truth” particle jet • Several weighting schemes • Per layer (simpler for trigger ?) or per cell • Weights = f(Cell Energy density) (“H1 weights”) or f(Cell Energy density,Ejet) • Method used for most physic studies so far

  13. Jet Resolution standard towers • Multi-jet event - Cone 0.7 - Use towers that include only those cells that belong to topoclusters - H1 weights. standard towers Un calibrated (barrel region) towers from topos Calibrated (standard towers) towers from topos Calibrated (towers from topos) A.Gupta talk

  14. Local hadronic calibration • Mainstream method from the hadronic calibration group • Try to disentangle various effects and optimize noise suppression • Topocluster with S/N cuts for seed & expansion. “Ideal” world: 1 EM particule in 1 topo cluster • Cluster splitting using local maxima • Classify EM/Had nature of cluster using shower shape related variables • Cell Weights on Had. Cluster to go back to true energy in cluster (correcting for “invisible” and “escaped” energies from Geant4) • Explicit corrections for Dead Matter energy lost • Make jets from those calibrated Topocluster • Requires to understand many complicated ingredients (not a “brute force” minimization) • Is it more (or less) MC dependent than other approaches ? • Factorization of weighting and DM is an approximation

  15. Ncluster in “empty” events follow cell granularity Effect of PileUp on TopoCluster still needs investigations J4 dijet sample (140-280 GeV pt): In average 1.5 truth particles per TopoCluster. How does it behave at larger energy ? S.Menke talk

  16. Cluster classification: Use 2 variables: Shower depth => later shower are more hadronics Energy weighted average cell density => EM shower “denser” Classify as EM True EM fraction from G4 deposits 1 bin in eta and Eclus S.Menke talk

  17. Weights for non-EM clusters 2<eta<2.2 emec2 Applied at cell Level as a function of: Eclus and cell energy density for different bins in eta and different layers (only for cells with >2 sigmas of noise) Computed from true G4 energy deposits: (Einvisible+Eescaped+Evisible)/Evisible hec1 S.Menke talk

  18. What does it do on 50 GeV single pions ? EM scale After weighting Ereconstruced/(Etrue in calo) Weighting works better at high energy Dead Matter energy loss is also important S.Menke talk

  19. G. Pospelov talk

  20. DM corrections on 100 GeV pions No correction Weighted Weighted+DM corr Weighted+(all)true DM G. Pospelov talk

  21. Does it work on jets ? (kt on Dijet events ~150 GeV pt) EM scale Weighting Weighting+DM Energy resolution improves from 7.3 to 6.5% Average get closer to 1 Need to complete full set of corrections for 12.0.4 (various bugs in productions done before Barcelona) S.Menke talk

  22. Alternative: Weighting as a function of shower development (decompose events on eigenvectors of covariance matrix of correlation between samples and apply weight/layer) On CBT simulation (pions) Does it also work on jets ? Along similar lines: idea of radial weighting in Fcal T.Carli talk

  23. ATHENA MissingET : EtMiss Reconstruction and Calibration All Calorimeter cells in ||<5 |Ecell | > 2(noise) H1-like weights from jets OR OR OR MET_Final = MET_Calib + MET_Cryo + MET_ Muon Estimated energy loss in cryostat between LAr and Tile from reco jets (cone07) w*sqrt(EM3*TILE1) from MOORE mu ||<2.7 pT from external Spectrometer BestMatch /good quality Calorimeter cells in Topoclusters 4/2/0 H1-like weights from jets from MuonBoy muons in || < 2.7 D. Cavalli, S.Resconi from reco jets (kT) Calorimeter cells in Topoclusters 4/2/0 LocalHadronCalibration (under test) Topoclusters 4/2/0 LocalHadronCalibration (under test) Refinement of Calibration for cells in different objects (e/g, t, m ,jets...) a la TDR (implemented in 12.0.2 – under test) : Present default in MissingET D.Cavalli talk

  24. ATHENA MissingETPerformance Good EtMiss measurement crucial for mass reco in Z   and H   EtMiss resolution vs SumET EtMiss resolution vs SumET CSC minimum bias N. Kanaya CSC data EtMiss resolution vs SumET EtMiss resol vs SumET CSC data CSC data Pythiaand Herwig QCD jets Resolution different for diff event samples and slightly worse respect to TDR parametrisation (s(EtMiss)=0.46*sqrt(SumET) from Z/A/H tt) No Montecarlo dependence

  25. Conclusions • Many topics not summarized here • Calibration at trigger level • Data streams for calibration • In situ samples for hadronic calibration • Test-Beam still very useful for many studies and MC validation • Many studies planned with CSC/CDC • Should quantify MC dependence of various methods • Maybe difficult to converge on one single method for everything before having data.

More Related