Effect of HF TP on L1T

[rekovic]

There are two pending PRs concerning HF LUTs.
(making TPs LSB for 1x1 same as for 3x2 (0.5 GeV))

cms-sw#13252
cms-sw#13704

Their effect on L1T is investigated.

[rekovic]

The two PRs applied on top of l1t-tsg-v4 have a small effect on L1T.

cms-sw#13704 makes jets only slightly harder.
This causes a small increase (~%) in L1T jet rates compared to l1t-tsg-v4, where layer-1 is turned off.
Here are the diff of L1T counts after re-emulating on 1000 data events of Run2015D/MuonEG/

<       17                             L1_SingleJet68       724    724    724         1          1        0
<       18                           L1_SingleS1Jet92       450    450    450         1          1        0
<       19                            L1_SingleJet128       219    219    219         1          1        0
<       20                          L1_SingleS1Jet176       100    100    100         1          1        0
<       21                          L1_SingleS1Jet200        77     77     77         1          1        0
---
>       17                             L1_SingleJet68       723    723    723         1          1        0
>       18                           L1_SingleS1Jet92       448    448    448         1          1        0
>       19                            L1_SingleJet128       216    216    216         1          1        0
>       20                          L1_SingleS1Jet176        99     99     99         1          1        0
>       21                          L1_SingleS1Jet200        75     75     75         1          1        0
106,107c106,107
<      100                  L1_TripleJet_92_76_64_VBF       124    124    124         1          1        0
<      101                  L1_TripleJet_84_68_48_VBF       202    202    202         1          1        0
---
>      100                  L1_TripleJet_92_76_64_VBF       123    123    123         1          1        0
>      101                  L1_TripleJet_84_68_48_VBF       201    201    201         1          1        0

JetEt plot (dots are from l1t-tsg-v4) available here
/afs/cern.ch/user/r/rekovic/public/forL1TStage2/l1t-integration/compare_l1t-tsg-v4+PR_13252+13704_AND_l1t-tsg-v4/discrep_mc_jetEt_0.pdf

More details available here /afs/cern.ch/user/r/rekovic/public/forL1TStage2/l1t-integration/compare_l1t-tsg-v4+PR_13252+13704_AND_l1t-tsg-v4/

[bortigno]

from the log it seems there is no big discrepancy between the two. is this expected? I am not sure I understand your plots though - what are you plotting in the y-axis?

[rekovic]

The point is that the differences are small. I do not know if this (small) change is expected.
HCAL experts should comment @nsmith-

The plots are filled with l1t object found when running re-emulation of L1T trigger on

• 1000 events of Run2015D/MuonEG DATA
• 500 events of TTbar MC

l1-jet Et in DATA [1] and MC [2]
l1-jet Eta in MC [3]

line = l1t-tsg-v4
points = l1t-tsg-v4 + PR cms-sw#13252 + PR cms-sw#13704

The hardening jet Et distribution is very minor that you cannot see it easily on the plots.
(meant to spot big differences)
In future will supplement the tests with the plots to include relative differences, but for now,
here are differences by bins in DATA re-emulation, showing small differences

All Jets:
discrepancy in jetEt bin: 7 266 vs 268
discrepancy in jetEt bin: 8 260 vs 262
discrepancy in jetEt bin: 9 201 vs 203
discrepancy in jetEt bin: 10 176 vs 174
discrepancy in jetEt bin: 12 77 vs 76
discrepancy in jetEt bin: 13 72 vs 73
discrepancy in jetEt bin: 14 49 vs 51
discrepancy in jetEt bin: 15 43 vs 42
discrepancy in jetEt bin: 16 34 vs 35
discrepancy in jetEt bin: 17 25 vs 24
discrepancy in jetEt bin: 21 111 vs 106

[1]
/afs/cern.ch/user/r/rekovic/public/forL1TStage2/l1t-integration/compare_l1t-tsg-v4+PR_13252+13704_AND_l1t-tsg-v4/discrep_reemul_jetEt_0.pdf

[2]
/afs/cern.ch/user/r/rekovic/public/forL1TStage2/l1t-integration/compare_l1t-tsg-v4+PR_13252+13704_AND_l1t-tsg-v4/discrep_mc_jetEt_0.pdf

[3]
/afs/cern.ch/user/r/rekovic/public/forL1TStage2/l1t-integration/compare_l1t-tsg-v4+PR_13252+13704_AND_l1t-tsg-v4/discrep_mc_jetEta_1.pdf

[bortigno]

thank you, now it is clear. let's wait to hear from layer1 layer2 experts ( @nsmith- @jimbrooke ) if that is what they expect from it. btw given the small effect of jets I think that we can safely trust the results from Zhenbin on the menu.

[nsmith-]

One can clearly see the effect of the 1/8 GeV LSB on the HF tower et saturation threshold of 32GeV (1/8 * 256):

Here, I ran on a JetHT data sample,

cmsDriver.py l1Ntuple -s RAW2DIGI --era=Run2_2016 --geometry=Extended2016,Extended2016Reco --customise=L1Trigger/Configuration/customiseReEmul.L1TEventSetupForHF1x1TPs --customise=L1Trigger/Configuration/customiseReEmul.L1TReEmulFromRAW --customise=L1Trigger/L1TNtuples/customiseL1Ntuple.L1NtupleEMU --customise=L1Trigger/Configuration/customiseUtils.L1TTurnOffUnpackStage2GtGmtAndCalo --conditions=auto:run2_data -n 1000 --data --no_exec --no_output --filein=/store/data/Run2015D/JetHT/RAW/v1/000/256/843/00000/42521405-A35E-E511-A7C8-02163E016FEC.root

Vladimir, using MuonEG sample to compare rates will understate the changes in jet triggers, I expect the rate diff in a JetHT sample to be larger. Really, though, rates only mean anything on ZeroBias, so perhaps you can run on that? (the downside is that with ZeroBias, we will need much more than 1000 events to get good predictions...)

[nsmith-]

FYI @rekovic @mulhearn, these two HF TP fix PRs require an L1 signature as well