TagAndProbe

Set of tools to evaluate eg/tau trigger performance on T&P, and produce Layer-2 calibrations This is based on the following two prior tools:

1. TauTagAndProbe
2. TauObjectsOptimization

Install instructions

cmsrel CMSSW_13_3_0
cd CMSSW_13_3_0/src
cmsenv
git cms-init
git cms-addpkg L1Trigger/L1TCalorimeter
git cms-addpkg L1Trigger/L1TNtuples
git cms-addpkg L1Trigger/Configuration
git cms-addpkg L1Trigger/L1TGlobal
git cms-addpkg L1Trigger/L1TCommon
git cms-addpkg L1Trigger/L1TZDC
mkdir L1Trigger/L1TZDC/data
cd L1Trigger/L1TZDC/data
wget https://raw.githubusercontent.com/cms-data/L1Trigger-L1TCalorimeter/master/zdcLUT_HI_v0_1.txt
cd -
git clone https://github.com/cms-l1t-offline/L1Trigger-L1TCalorimeter.git L1Trigger/L1TCalorimeter/data

git clone git@github.com:mchiusi/TagAndProbeIntegrated.git -b CMSSW_13_3_0

git cms-checkdeps -A -a
scram b -j 8

L1T emulation relevant GlobalTags in CMSSW_13_3_0 are stored here.

• for run3 data, use latest prompt GT here

Tool utilization

To do the optimization two things are needed:

• L1 objects (sometimes re-emulated) that are extracted from the RAW tier (in principle, non-re-emulated L1 objects can also be extracted from MiniAOD, but for consistency we never do that)
• Offline objects that are extracted from the AOD or MiniAOD tier

Production of the input objects

To produce the input NTuples to the optimization the TauTagAndProbe package is used. The useful scripts for this are mainly in the test subfolder.

Jobs on RAW are submitted using submitOnTier3_reEmulL1_zeroBias.py which in turn launches reEmulL1_X.py Before launching this you need to fix

• the Global Tag
• the configuration of the L1Calorimeter (process.load("L1Trigger.L1TCalorimeter.caloStage2Params_20XX_vX_X_XXX_cfi"))

Jobs on MiniAOD are submitted using submitOnTier3.py which in turn launches test_noTagAndProbe.py Before launching this you need to fix

• the Global Tag

Jobs on Data are submitted using submitOnTier3_reEmulL1_zeroBias.py which in turn launches reEmulL1_X.py. Before launching this you need to fix

• the isMC flag
• the input folder and file list

For Monte Carlo (MC), we implemented a truth matching rather than a Tag & Probe technique which would dramatically and artificially decrease the available statistics.

After having produced the input object hadd all the files.

Optimization

The optimization is run in several sequential steps:

• Merge of the two inputs, match of the L1 objects to the offline ones, compression of the variables
• Calculation of the calibration, pruduction of its LUTs, and its application
• Calculation of the isolation, pruduction of its LUTs, and its application
• Prodution of turnon curves
• Evaluation of the L1 rate

Due to the package's lack of forward-compatibility with CMSSW, the optimzation is run into CMSSW_11_0_2. Check this link in order to keep track of the lastest updates in CMSSW.