1 / 26

On Calorimeter Thresholds for Jet Reconstruction

On Calorimeter Thresholds for Jet Reconstruction. Marek Zieli ński ( Rochester) Jet MET, 12 February 2008. Introduction. Hierarchy of thresholds: Online Zero Suppression Driven by channel occupancy requirements “Typical” ZS: 1-3 ADC; in HB 3ADC ~500 MeV avg, 300-900 range

jodie
Download Presentation

On Calorimeter Thresholds for Jet Reconstruction

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. On Calorimeter Thresholdsfor Jet Reconstruction Marek Zieliński (Rochester) Jet MET, 12 February 2008

  2. Introduction • Hierarchy of thresholds: • Online Zero Suppression • Driven by channel occupancy requirements • “Typical” ZS: 1-3 ADC; in HB 3ADC ~500 MeV avg, 300-900 range • Offline thresholds for cells (RecHits) and towers • Thresholds for jet and MET reco could be different; should they? • Current scheme for offline thresholds • Energy cell thresholds to cut out majority of noise • ET tower thresholds to cut out majority of pileup & UE • Default in CMSSW: Scheme B for cells + tower ET>0.5 GeV Marek Zieliński

  3. Goals and Tools • Revise the value of cell thresholds • Cell thresholds can have a significant impact on the observed response to low energy particles • Optimization of tower ET threshold best done as part of pileup studies • Scheme B was derived based on ORCA 871 MC • CMS Note-2006/020 • Many changes in HCAL simulation since then • Information to consider (per subdetector) • Amount of noise within jet cone vs threshold • Amount of real jet energy removed vs threshold • Jet resolution vs threshold • Jet rapidity distribution at low pT • … no clear single measure for optimization… Marek Zieliński

  4. “A Brief History of Noise…” • Measurements and simulation of noise are performed by members of HCAL DPG using a variety of data • Description of noise changed significantly in CMSSW 152 • From ORCA times till 14x, noise patterns had a discrete pattern per HCAL subdetector • From 152 onward, realistic HCAL calibrations (based on DB values) • ADC/GeV, snoise and other constants improved • Corrections for ZS/threshold effects will be done by JEC group HB CMSSW 152 RecHits Salavat CMSSW 140 RecHits Salavat ORCA 871 Marek Zieliński

  5. The Real Data: GREN • Analysis of GREN RecHit data from Daniel Miner • GR run 30333 • New results and MC much closer (need to separate HB & HE) • While the dust settles, I’ll continue with 1.5.2 MC… “Old” gains and pedestals “New” gains and pedestals Mean: 0.36 Sigma: 0.78 Mean: 0.003 Sigma: 0.297 Marek Zieliński

  6. “Cell Thresholds” and Noise (152) • A “first look” using the HCAL constants in 152 • effects of cell thresholds on noise in jets for five different regions of the calorimeter: HB, HO, HE, EB, and EE (HF not considered here) • No simulation of online ZS • A pure noise file generated in CMSSW 152 was used • All tower-level thresholds were removed including those on crystals • Checked against a CMSSW 152 single muon plus noise RelVal sample with no thresholds and compared to Salavat’s plots • Single muon plus noise files are dominated by noise in HCAL and ECAL • To estimate noise contributing to a jet, energy was summed from cells within a cone R=0.5 • Two methods used • Thresholds applied to emEnergy, hadEnergy, outerEnergy Marek Zieliński

  7. HAD Noise per HB Tower in 152 • Noise level in the “pure noise” file agrees with noise seen in RelVal single-muon file and shown in HCAL DPG meetings Noise energy in HB cells Salavat The 152 single-m file (RelVal sample) The 152 noise file (private production) The 152 RecHits (HCAL DPG study) Marek Zieliński

  8. Average of all HB towers, normalized to cone area (better accuracy for mean value) Used for Noise vs Threshold plots Noise in Cone in HB (Threshold = 0) • Two methods • to sum the noise contributions within cone R = 0.5 One cone per event moved randomly in HB (illustrates expected spread of jet noise values) Marek Zieliński

  9. HO HE The 152 noise file Noise in Cone in HE & HO (Threshold = 0) • Sum of hadEnergy contributions from all HE/HO towers, normalized to cone area • In HE, different from applying thresholds per cell Marek Zieliński

  10. Scheme B “HAD Thresholds” and Noise in Cone • Average noise in R=0.5 cone vs threshold • Compare to Scheme B Thresholds: HB 0.9, HE 1.4, HO 1.1 GeV • Scheme B thresholds seem larger than necessary for all regions Pat Buchinski Marek Zieliński

  11. Question: How much real jet energy is lost by imposing calorimeter cell energy thresholds? Analysis: QCD ptHat > 90 GeV in CMSSW 152 (1000 events), jet pT > 40 GeV No noise No thresholds except optional Tower ET>0.5 GeV Apply thresholds on emEnergy, hadEnergy, outerEnergy “by hand” Recalculate jet energy components Results change when tower ET>0.5 GeV is applied (default in jet reco) “Cell Thresholds” and Energy Loss in Jets Zhen Qi Marek Zieliński

  12. Thresholds and Jet Energy Loss in Barrel • Energy loss = Subtract energy with thresholds from energy with threshold=0 • Jet energy loss could be reduced by lowering thresholds, esp in HB • Tower ET>0.5 GeV threshold has big impact • Energy loss due to crystal cuts not studied (in EB or EE) Zhen Qi Marek Zieliński

  13. Jet Resolution vs Thresholds • Methodology: • Used 50 < pThat< 80 GeV dijet sample • Defined “MC truth” response and resolution: • mean response m = mean(pTcorr/pTgen) • scaled resolution: s(pTcorr/pTgen)/m • Derived m and s/m for 5 threshold schemes • T0: HB:0. HO:0. HES:0. HED:0. • T1: HB:0.3 HO:0.3 HES:0.3 HED:0.3 • T2: HB:0.5 HO:0.5 HES:0.6 HED:0.6 • T3: HB:0.7 HO:0.8 HES:1.0 HED:1.0 • T4: HB:0.9 HO:1.1 HES:1.4 HED:1.4 = Scheme B • Kept EM cuts and tower ET cut at default values • 100k events each • MCJet corrections applied (same • Two pTgen bins: 20-40 and 50-80 GeV • Separately for jets in Barrel and Endcap Marek Zieliński

  14. HB: Jet Response & Resolution vs T Zhen Qi 20<pTgen<40 GeV m s/m 50<pTgen<80 GeV Reprocessing without the Tower ET cut Marek Zieliński

  15. HE: Jet Response & Resolution vs T Zhen Qi 20<pTgen<40 GeV m s/m 50<pTgen<80 GeV Reprocessing without the Tower ET cut Marek Zieliński

  16. Jet h Distributions • The impact of low cell thresholds of 0.5 GeV was HUGE in ORCA • default cell energy threshold E>0.5 GeV (black) • Higher threshold 0.8 GeV (red) was similar to Scheme B in Barrel • nominal tower threshold ET>0.5 GeV for both cases • Not observed in 1.5.2 • consistent with significantly lower noise • tower ET>0.5 GeV cut seems to suppress the noise by itself • tuning noise thresholds coupled with the choice of tower cut for PU ORCA CMSSW 1.5.2 Zhen Qi Marek Zieliński

  17. Jet Matching Efficiency & “Fakes” • ereco is the standard jet matching efficiency • No significant effect of lowering the thresholds from Scheme B (black) to flat 0.5 GeV cut (red) • efake is the standard jet mismatch rate (aka “fake rate”) • a small effect near h = 0 for lower thresholds – a few real fake jets?! Cosmin Dragoiu Marek Zieliński

  18. Conclusions & Outlook • A first look at Noise-in-Cone and Jet-Energy-Loss vs calorimeter “cell” thresholds a la ORCA study • Scheme B thresholds seem to be higher than necessary • Value of JEL strongly affected by tower ET cut • Jet resolutions improve for lower thresholds • Things to investigate: • Effects of online ZS • Correlation of tower ET and cell thresholds • Re-optimization of ECAL cuts • Impact of thresholds on jet “efficiencies” and “fake rates” • Crucial to coordinate threshold studies with • HCAL improvements of pedestals/noise (HCAL DPG) • GR noise studies (Daniel Miner, Efe,…) • JEC offset studies (Daniel, Ia Iashvili…) • Pileup suppression (Ia, Olga,…) • Looking for help/volunteers Marek Zieliński

  19. Some Questions… • What HCAL ZS settings are expected for real data? • when will be known? • Are ECAL thresholds & noise finalized in simulation? • What CMSSW version to use? • Salavat: may want to wait for 2xx • What MC samples are required? • What additional info/studies are needed to achieve consensus about thresholds? • What is current thinking about “safety factors” when facing the unknowns of real data taking? Marek Zieliński

  20. Backups

  21. Considering Zero-Suppression From Davide, 27 Aug 2007 Marek Zieliński

  22. HAD Noise per Tower in HE & HO • Difference in HE due to multiple cells per tower in HE The 152 noise file HE HO HO HE The 152 RecHits Salavat Marek Zieliński

  23. “HAD Thresholds” and Noise in Cone • Average noise in R=0.5 cone vs threshold • Compare to Scheme B Thresholds: HB 0.9, HE 1.4, HO 1.1 GeV • Scheme B thresholds seem larger than necessary for all regions Scheme B Marek Zieliński

  24. “EM Thresholds” and Noise in Cone • Average noise in R=0.5 cone vs threshold on SumEB/SumEE • Compare to Scheme B Thresholds: Sum EB 0.2, SumEE 0.45 GeV • “SumEM” thresholds depend on crystal thresholds; are these final? EE EB No crystal thresholds applied when making CaloTowers. “Standard” crystal thresholds applied in CaloTowers: EB 0.9 / EE 0.45 GeV EB EE Results slightly different than in ORCA: 1.9 (EB) and 0.2 (EE) GeV for SumEM threshold = 0. Is this consistent with changes in EM noise and/or thresholds since ORCA?? Marek Zieliński

  25. Avg Jet Energies in Endcap vs Thresholds • Jet energy loss could be reduced by lowering thresholds Zhen Qi Marek Zieliński

  26. “Cell Thresholds” and Energy Loss in Jets • Jet energy loss could be reduced by lowering thresholds • But not much gain in Endcap after tower ET>0.5 GeV threshold Zhen Qi Marek Zieliński

More Related