SplitClusterStudy < Main

Main Web>TWikiUsers>AdamHunyadi>SplitClusterStudy (2017-12-19, AdamHunyadi) (raw view)

---+ Split Cluster Study
---++ Links
   * [[https://github.com/hunyadix/PhaseISplitClusterAnalyzer][Git repository for the analyzer]]
   * [[https://github.com/hunyadix/cmssw/tree/ModForPixelMerger][Git repository for the CMSSW modifications]]
   * [[https://github.com/hunyadix/Histogram-Merger][Git repository for the histogram merger tool]]

---++ Issues
   * The pileup mixing seems to throw away all the digis that are coming from the mixed pileup. This makes it difficult to add the digi markers.
   * The digi collection does not _exactly_ contain the same pixels as the cluster collection. Mostly the digis is a bigger set (with all the cluster pixels available in it), but for now on I create a lookup table and add zero markers for any extra cluster digis that might turn up. The rate of "mocked digis" is a good number to pay attention to when running the simulations even on the grid.

---++ TODOs (problem fixes, find plans at the end of this page)
   * Add modifications to the pileup mixing modules so that it propagates the digi flag collection from the mixed data and it can be combined to a correct flag collection
   * Modify the histogram merger code not to merge the "ModuleClusterPlots" even if they are from the same run/lumiblock/event

---++ Recipe to run the framework

---+++ Deploy script

<verbatim>
export SCRAM_ARCH=slc6_amd64_gcc530
cmsrel CMSSW_9_3_0_pre1
cd CMSSW_9_3_0_pre1/src
cmsenv
git cms init
git cms-addpkg SimTracker/SiPixelDigitizer
git cms-addpkg RecoLocalTracker/SiPixelClusterizer
git remote add hunyadix git@github.com:hunyadix/cmssw.git
git fetch hunyadix
git checkout -t hunyadix/ModForPixelMerger
git clone git@github.com:hunyadix/PhaseISplitClusterAnalyzer.git DPGAnalysis/PhaseISplitClusterAnalyzer
scram b -j 20
</verbatim>

---+++ Generating a CMSSW config file

<verbatim>
cmsDriver.py \
-s GEN,SIM,DIGI,L1,DIGI2RAW,RAW2DIGI,RECO \
--evt_type SingleNuE10_cfi \
--process PrivateRECO \
--conditions auto:phase1_2017_realistic \
--era Run2_2017 \
--geometry DB:Extended \
--pileup Flat_0_50_25ns \
--relval 9000,50 \
--beamspot Realistic25ns13TeVEarly2017Collision \
--filein /store/relval/CMSSW_9_0_0_pre4/RelValMinBias_13/GEN-SIM/90X_upgrade2017_realistic_v6-v1/10000/0047A0C1-BDEC-E611-8FD9-0CC47A4D7630.root \
--pileup_input das:/RelValMinBias_13/CMSSW_9_2_3-92X_upgrade2017_realistic_v1_earlyBS2017-v1/GEN-SIM \
--eventcontent RECOSIM \
--datatier GEN-SIM-RECO \
--fileout file:GenNu_FlatPileup0to50_GENSIMRECO.root \
--python_filename PhaseI_GenNu_FlatPileup0to50_cfg.py \
--runUnscheduled \
-n 10 \
--no_exec
</verbatim>

For no PU, I usually generate TTbar events.

---++++ How to add dynamic inefficiency

The tag is set when the db file is generated.

<verbatim>
# Add dynamic inefficiency to simulate dcol losses
dynIneff_db     = 'sqlite_file:phase1_efficiencies_95_pure.db'
dynIneff_tag    = 'SiPixelDynamicInefficiency_v1'

process.DynIneffReader = cms.ESSource("PoolDBESSource",
    DBParameters = cms.PSet(
        messageLevel = cms.untracked.int32(0),
        authenticationPath = cms.untracked.string('')),
    toGet = cms.VPSet(cms.PSet(
        record = cms.string("SiPixelDynamicInefficiencyRcd"),
        tag = cms.string(dynIneff_tag))),
    connect = cms.string(dynIneff_db))
process.dynineffprefer = cms.ESPrefer("PoolDBESSource", "DynIneffReader")
</verbatim>

---+++Calculating the pileup for data files

Example:
<verbatim>
brilcalc lumi --byls -u /nb --minBiasXsec 78100 --begin 5659 --end 9999 | head -n-8 | tail -n+5 | sed "s;|;;g;s;:; ;g"):
</verbatim>

This bash-script can be called from the SplitClusterStudy analyzer (currently commented out), but opening a socket to store the string as a result still regularly fails (and also takes up too much time when testing).

---+++ Running the HistogramMerger

Running the HistogramMerger is as simple as checking out the git repository and specifying the target files and the output name in src/main.cxx. 

---+++ Tests

   * The pixel markers are seemingly present for all the cluster pixels when the pileup is 0 (1). The function checkDetIdToMarkerPtrMap() is currently run for each of the events and it prints out how many of the clusters have a corresponding marker.
   * In the analyzer testPairFinding() tests the private areClustersPair() for pair finding efficiency. Seems like the pair finding works just fine if the cluster objects have properly set x, y and size fields. 

---++ Results

TTbar events, no pileup, 95% dcol efficiency statistics files: 

<verbatim>
/store/user/ahunyadi/SCS/Run2_CMSSW93X/SCS_93X_mcRun2_DynIneff_95_GEN_SIM_RAWTODIGI_evt100000
</verbatim>

For users on the ui3 of the Wigner institute:
<verbatim>
/data/gridout/ahunyadi/SCS/Run2_CMSSW93X/SCS_93X_mcRun2_DynIneff_95_GEN_SIM_RAWTODIGI_evt100000/171212_101805/0000/
</verbatim>

---+++ Validation
Examples for module cluster plots of single events are shown below. Nothing unordinary to be seen.

%TABLE{headeralign="center" dataalign="center" tablewidth="100%"}%
| *Low eta module, 50 events merged* | *High eta module, 50 events merged* |
| <img src="%ATTACHURL%/Low_eta_example.png" width="650px"> | <img src="%ATTACHURL%/High_eta_example.png" width="650px"> |

---+++ Fluctuations in the leftmost and the rightmost pixels 

Positions for the boundary clusters in pixels:

%TABLE{headeralign="center" dataalign="center" tablewidth="100%"}%
| *All boundary pixels* | *Leftmost pixels* | *Rightmost pixels* |
| <img src="%ATTACHURL%/CLUSTER_BOUNDARY_PIX_COLS_BPIX.png" width="475px"> | <img src="%ATTACHURL%/CLUSTER_LEFTMOST_PIX_COLS_BPIX.png" width="475px"> | <img src="%ATTACHURL%/CLUSTER_RIGHTMOST_PIX_COLS_BPIX.png" width="475px"> |
| Since the dcol loss has the opposite effect on the rightmost and leftmost pixels, the total boundary pixel distribution does not show any fluctuation effect. | For 95% dcol efficiency the leftmost cell shows ~8.5-10.5% less pixels in the odd columns  | For 95% dcol efficiency the rightmost cell shows ~8.5-10.5% less pixels in the even columns |
---++ Plans for the analysis

   * Write a tool to properly calculate the percentage differences in the even and odd cols (excluding the fidicual areas) 
   * Measure the correspondence of the fluctuation in the leftmost and rightmost pixel position to the dcol efficiency correspondence by creating an efficiency scan with the SplitClusterStudy plugin
   * Measure distributions for the left- and rightmost pixels in the data and give estimation for dcol efficiency loss. 

-- Main.AdamHunyadi - 2017-12-04

Attachments

Topic attachments
I	Attachment	History	Action	Size	Date	Who	Comment
png	CLUSTER_BOUNDARY_PIX_COLS_BPIX.png	r1	manage	12.5 K	2017-12-15 - 13:43	AdamHunyadi	Boundary cols. BPIX example
png	CLUSTER_LEFTMOST_PIX_COLS_BPIX.png	r1	manage	12.7 K	2017-12-15 - 13:49	AdamHunyadi	Leftmost pixels, BPIX example
png	CLUSTER_RIGHTMOST_PIX_COLS_BPIX.png	r1	manage	12.3 K	2017-12-15 - 13:50	AdamHunyadi	Rightmost pixels, BPIX example
png	High_eta_example.png	r1	manage	30.7 K	2017-12-15 - 14:42	AdamHunyadi	High eta module cluster plot example
png	Low_eta_example.png	r1	manage	31.6 K	2017-12-15 - 14:42	AdamHunyadi	Low eta module cluster plot example

Topic revision: r11 - 2017-12-19 - AdamHunyadi

Main

Copyright &© by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback