ecal simulation status report brian heltsley full simulation mtg cms week 14 june 2010
DESCRIPTION
Ecal Simulation Status Report Brian Heltsley Full Simulation Mtg CMS Week 14 June 2010 Focus for last 3 months Improved noise models APD simulation studies Housekeeping. Measured Correlations. Numbers from Sasha Ledovskoy EB & EE have different correlations - PowerPoint PPT PresentationTRANSCRIPT
B. Heltsley Simulation Mtg - 15 Jun 20101
Ecal Simulation
Status ReportBrian Heltsley
Full Simulation Mtg
CMS Week
14 June 2010
Focus for last 3 months• Improved noise models• APD simulation studies
• Housekeeping
Measured Correlations
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20102
Numbers from Sasha LedovskoyEB & EE have different correlations
Noise frequency is higher in EE than EBBoth have high correlations in neighboring samples
RecHit Energy Spectrum MC/data agreement improves using measured
(fixed) ’s, online ZS/SR:(EB)=1.089, ZS/SR(EB)=78.75 MeV;(EE)=2.018, ZS/SR(EE)=225 MeV
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20103
Recent Noise ProgressStructural code changes allowing separate EB, EE correlation matrices completed in CMSSW_3_6_0_pre3, but values still set to identity
Tag set prepared using – measured correlations– Fixed, measured ’s (i.e. same for all channels) – online ZS/SR thresholds
Above tag set fully validated with REDIGI of MIN_BIAS MC (next slides) (lots of work by many people!!)
Was taken for CMSSW_3_7_0_pre1 & back-ported to 3_6_1
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20104
Commissioning-PromptReco-v7 vs REDIGI_MinBias7TeV
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20105
EB EE+ EE-
OccupancyModelingImproves
Commissioning-PromptReco-v7 vs REDIGI_MinBias7TeV
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20106
EB EE+ EE-
354p1 354p1
Timing ResolutionModelingImproves
Data
ES DevelopmentsChanges & validations from CM Kuo
Up until now, several ES MC values have been set in the cfg file only– Pedestals
– Gain (hi or lo)
– MIPtoADC
– MIPtokeV
– ZS threshold
ZS algorithm not same as online
Tag set & DB payloads created which fix the above
Installed for CMSSW_3_7_0_pre1 & back-ported to 3_6_1
B. Heltsley Ecal Status@Sim Mtg - 12 Apr 20107
Data
New MC
Old MC
APD SimulationJanuary– Could EB spikes be from energy deposition in APDs (2 behind
each crystal)? Yes
February & March– Are geometry & material details important? Yes
April– Do existing GEANT cutoffs hurt APD simulation? Yes
May– Can we tune APD digitization parameters using data with
a large MIN_BIAS sample? Yes
June– Can APD simulation be an option in CMSSW? Yes
B. Heltsley Simulation Mtg - 15 Jun 20108
Many Contributors!ASCTF (T. Virdee) & EcalDPG (T. Tabarelli, A. Bornheim)– Guidance, priorities
S. Banerjee– GEANT implementation & performance
I. Musienko, Q. Ingram– Specifications & materials
B. Heltsley– Geometry implementation & Digitization
A. Holzner– Analysis & MIN_BIAS production
J. Branson, M. Malberti, L. Di Matteo, D. Petyt– Analysis, comparisons with data
Probably others!! (apologies for omissions)B. Heltsley Simulation Mtg - 15 Jun 2010
9
B. Heltsley Ecal Status@Sim Mtg - 16 Mar 201010
5x5 mm2 APDs initially present in simulation as shown
APD thickness magnified greatly for visualization purposes – actual (potentially active) thickness in original simulation is 5m. Material is Si.
Inactive by default
Origin of APD deposition?
B. Heltsley Simulation Mtg - 15 Jun 201011
-
+
KLn
n
-
+n
nKL
From J. Branson
B. Heltsley Simulation Mtg - 15 Jun 201012
APD Layout 3G (vertical not to scale)
B. Heltsley Simulation Mtg - 15 Jun 201013
CrystalSilicone GlueEpoxy
Non-depeleted SiHi-gain active Si
Lo-gain active Si
Bulk SiCeramic
Capsule
1.65 mm
30 m
575m0.1 m
5 m45 m 400 m
Crystal
5mm
3_7_0_pre3
The Real Thing
Why local materials are important to simulation
B. Heltsley Simulation Mtg - 15 Jun 201014
Z
APD
Epoxy
100 m
From J. Branson
Digitization parametersSimCalorimetry/EcalSimProducers/python/
apdSimParameters_cff.pyapdAddToBarrel = cms.bool(False), apdSeparateDigi = cms.bool(False), apdSimToPELow = cms.double(4.41e6), apdSimToPEHigh = cms.double(157.5e6), apdTimeOffset = cms.double(-10.0), apdDoPEStats = cms.bool(True), apdDigiTag = cms.string("APD"), apdShapeTstart = cms.double( 74.5 ), apdShapeTau = cms.double( 40.5 )
Shape: S(t) = A(t) ( 1 – exp[A(t)] ) where A(t) = (t – tS)/
B. Heltsley Simulation Mtg - 15 Jun 201015
Study of APD-MC Re-Digi
B. Heltsley Simulation Mtg - 15 Jun 201016
From
M. M
alber
ti
Number
B. Heltsley Simulation Mtg - 15 Jun 201017Fro
m M
. Mal
berti
Energy
B. Heltsley Simulation Mtg - 15 Jun 201018
APD energies probably a bit too high
From
M. M
alber
ti
E4/E1
B. Heltsley Simulation Mtg - 15 Jun 201019
Too many spikes!
From
M. M
alber
ti
GEANT Performance
B. Heltsley Simulation Mtg - 15 Jun 201020
From
S. B
aner
jee
Timing
B. Heltsley Simulation Mtg - 15 Jun 201021
GEANT Performance
B. Heltsley Simulation Mtg - 15 Jun 201022
From
S. B
aner
jee
+17%
+5%
+9%
APD Simulation Conclusions
Many qualitative features of the data which were previously completely missing from simulation can now be modeled (surprisingly well)
As of CMSSW_3_8_0_pre3– APD’s will be ACTIVE at GEANT stage of simulation
• Permits later reDigi+reReco when those are understood better
• Modest CPU penalty
– Digitization of APD SimHits will be performed, but saved only in a dedicated, tagged collection (which is small)
• Experts can study APD even in default generation
• We have moved APD pulse backward by 5 ns additional
• We have reduced both layer gains by 20%
B. Heltsley Simulation Mtg - 15 Jun 201023
Other Sim changes in progress(unchanged since last time)
Test containment factor– Sets overall MC energy normalization for EB & EE separately by
looking at simulated single 50 GeV photons– Not sure if GEANT has changed since this was updated
Change from fixed, uniform single-sample ADC noise to measured channel-by-channel values? – Studies in progress by T. Tabarelli & F. de Guio
Correlation matrix for ES (which has 3 time samples)– Ming & I are examining this– Would be much more efficient than current CPU & memory-
intensive so-called “fast” method which tries to account for some of the correlations (ironically, with slower code)
B. Heltsley Simulation Mtg - 15 Jun 201024
Simulation Housekeeping(Unchanged since last time)
In progress– Reform digitized hit response handline
• Has same problem as analog version, too many new ‘s & delete’s
– Add ES digi creation to EB/EE templatized version, if possible– Inefficient ES Digi object (collection of std::vector<int> )
• Should look more like EB & EE Digi objects (long optimized)
• F. Ferri & P. Gras agree
• Migration would have to be carefully orchestrated
B. Heltsley Simulation Mtg - 15 Jun 201025