sipm noise in fastsim with radiation damage
play

SiPM Noise in FastSim with Radiation Damage Kevin Pedro UMD CMS - PowerPoint PPT Presentation

SiPM Noise in FastSim with Radiation Damage Kevin Pedro UMD CMS Group December 6, 2012 SiPM Noise Model SiPM noise, in photoelectrons (pe), scales as the square root of radiation dose: noise SiPM = 15( lumi/3000) This is based on a


  1. SiPM Noise in FastSim with Radiation Damage Kevin Pedro UMD CMS Group December 6, 2012

  2. SiPM Noise Model • SiPM noise, in photoelectrons (pe), scales as the square root of radiation dose: noise SiPM = 15∙√( lumi/3000) This is based on a reference measurement of ~15 pe at 3000 fb -1 . • In the current FastSim, longitudinal segmentation is not yet implemented. Each tower will have 5 SiPMs, so the noise term should take that into account: noise SiPM → √(5)∙ noise SiPM • There is also a constant term of ~2 fC from the QIE-10. The QIE- 10 will experience an effective gain of 60,000, implying a conversion of 3 fC/pe: noise QIE-10 = 0.667 2

  3. SiPM Noise Model • These two noise terms should be added in quadrature: noise = √(noise SiPM ² + noise QIE-10 ²) • The total noise term must be converted from pe to GeV so it can be used to add noise to the SimHits when creating RecHits. • This conversion factor differs between HB and HE due to the difference in sampling factor: HB: 30 pe/GeV (sampling factor ~120) HE: 20 pe/GeV (sampling factor ~180) • The SiPM noise will be ~500 MeV in HB and ~750 MeV in HE at 3000 fb -1 . For comparison, the HPD noise is ~270 MeV. (SiPMs in HO have not yet been implemented.) 3

  4. SiPM Signal Reduction • SiPMs experience signal reduction with increasing radiation dose: signal = 1 – 0.1∙( lumi/3000) This is based on a reference measurement of ~10% reduction at 3000 fb -1 , with linear scaling. • Signal reduction is applied to SimHits (only real energy, no noise) before noise is added: energy → energy∙signal • After noise is added to create RecHits, turning up the gain to correct for the signal reduction is simulated: ( energy + noise) → (energy + noise)/signal 4

  5. Different Dose for HB vs. HE? • HB and HE front-end electronics (FEE) are located fairly close together. Is there an appreciable difference in dose between the two locations? 5

  6. Physics Test: Jets • Use Pythia6 to generate jets from d quarks at various p T , all uniformly distributed in j , with several samples shot at different h values (2.0, 2.5, 3.0). • Match ak5 CaloJets to the GenJet within a cone of dR < 0.5, to examine calorimeter damage effects on jet p T response, p T resolution, and position resolution. (CaloJets do do no not have jet energy scale corrections applied.) • Jet pT response and resolution are found from a Gaussian fit to ±1 RMS around the mean of the pT distribution. Position resolutions are found from just the RMS of the η and φ distributions. 6

  7. Physics Test Settings • Simulate the detector response in FastSim for several integrated lumi values and jet p T values. • Noise settings tested: zero noise, HPD noise, SiPM noise • Instantaneous luminosity is set to 5.0 × 10 34 cm -2 s -1 for all runs. • Radiation damage enabled in both ECAL and HCAL for lumi > 0. (HF is considered to be ideally rad-hard for these tests, since a radiation damage model is not yet available for it.) 7

  8. Sample SiPM Results ( η = 2.5) 8

  9. Sample HPD Results ( η = 2.5) 9

  10. Sample “no noise” Results ( η = 2.5) 10

  11. Physics Test Conclusions • Effects on p T and position resolution are largest for low-p T jets • Noise thresholds for CaloTower reconstruction need to be tuned carefully for response values to be meaningful • The scaling of SiPM noise with integrated luminosity is an important effect: SiPM noise at Phase 2 luminosities will be larger than HPD noise, and this contributes significantly to jet response and resolution • Lower operation temperature for SiPMs (~4°C vs. default ~25°C) could reduce the noise levels by a factor of ~3. (Noise drops by a factor of ~√(2) for every 7 °C drop in temperature.) • Many thanks to Jake Anderson, Yuri Musienko, Chris Tully, and Adriaan Heering for answering many questions about the SiPMs 11

  12. Backup

  13. HE Radiation Damage Model There is an existing model in CMSSW full sim which applies • radiation damage darkening to the HE sensitive layers (developed by Salavat Abdoulline, Petr Moisenz, and Anatoli Zarubin) This model is defined in the class HEDarkening (in • SimG4CMS/Calo ) and implemented in the function HCalSD::GetEnergyDeposit() Radiation dose is parameterized by layer and transverse radius, • and then multiplied by integrated luminosity (in fb -1 ) to give a value in MRad for use in an exponentially decaying weight factor (applied to SimHit energy): w = exp(-MRad/6.4) 13

  14. HE Radiation Damage Model This model has been adapted to work in FastSim by applying the • same darkening to energy spots in HE The code has been combined with the ECAL radiation damage • model implemented by Alexander Ledovskoy and Brian Francis, so radiation damage can be applied to both ECAL and HE simultaneously (HF model is in the works) Currently, the software is available as a private FastSim release, • with details and instructions at this twiki page: https://twiki.cern.ch/twiki/bin/view/CMS/FCALSimSLHCFastSim Aging The results for single pion energy response have been validated • against CMSSW full sim, using samples of 10,000 pions 14

  15. Validation of Darkening vs. Full Sim (Response) In this plot, FastSim points are shifted slightly on the x-axis for easier visual comparison. 15

  16. Validation of Darkening vs. Full Sim (Resolution) In this plot, FastSim points are shifted slightly on the x-axis for easier visual comparison. 16

  17. Recalibration Factors In practice, the photodetector gain in HE will be turned up to compensate for darkening and keep the energy response constant. In FastSim, the recalibration factors are calculated as follows: • Given the mean energy in each endcap i η tower for a sample of 100,000 pions at 50 GeV uniformly distributed in 1.6 < η < 3.0 (with magnetic field turned off) at a given darkening luminosity L: <E(L, i η )> • The factor is then: f(L, i η ) = <E(0, i η )>/<E(L, i η )> • This scales the mean energy in a darkened sample up to the mean energy in the undarkened sample. • Factors are calculated for various lumi values and interpolated for intervening values. • Factors are also applied to HCAL noise in the reconstruction process for physical consistency. 17

  18. Recalibration Factors The first few layers of the last HE tower (2.9 < η < 3.0) become completely dark at high lumi, leading to very large factors. 18

Recommend


More recommend