add sprucings for detached Jpsi, psi2S, tight prompt Jpsi, psi2S
Motivation
Create sprucing lines with full persistency for detached Jpsi->mumu, psi2S->mumu, and prompt (tight pT cut) Jpsi->mumu, psi2S->mumu decays.
The selection is tighter than the corresponding hlt2 lines, to reach the requirement of the disk storage. Selections are mainly taken from
Note that the mass window of the Jpsi is much tighter than the trigger line, considering that these sprucing lines are not used for the study of Jpsi itself, but for Jpsi + X combinations.
Rate test of the new lines:
The test is based on samples in https://gitlab.cern.ch/lhcb-b2oc/b2ocupgrade/-/blob/master/Moore_input_files/minbias_hlt1filt.ldst, corresponding to 1MHz rate. In total 50,000 Minbias events are used for the test. First we run hlt2_persistreco_realtime to get the output of following hlt2 lines whose persistReco=True:
lines += [twobody_line(persistreco=True)]
lines += [threebody_line(persistreco=True)]
lines += [DiMuonInc_line()]
lines += [DiMuonSameSignInc_line()]
lines += [JpsiToMuMuDetached_line()]
lines += [Psi2SToMuMuDetached_line()]
lines += [JpsiToMuMuTight_line()]
lines += [Psi2SToMuMuTight_line()]
lines += [z_to_e_e_line()]
lines += [same_sign_dielec_line()]
lines += [DiMuonHighMass_line()]
lines += [DiMuonSameSignHighMass_line()]
lines += [DiMuonIncHighPT_line()]
lines += [DiMuonSameSignIncHighPT_line()]
lines += [DiMuonSoft_line()]
lines += [UpsilonToMuMu_line()]
lines += [BcToJpsiPip_JpsiToMuMu_line()]
lines += [bc2jpsimu_line()]Then we run the bandq sprucing, and the selected number of event for each lines are:
SpruceBandQ_JpsiToMuMuDetached_Line:13; SpruceBandQ_Psi2SToMuMuDetached_Line:7; SpruceBandQ_JpsiToMuMuTightPrompt_Line:14; SpruceBandQ_Psi2SToMuMuTightPrompt_Line: 2.
The rates for the Jpsi lines are at 300Hz level. For the Psi2S lines, the rate is at 100Hz level.
Efficiency test of the new lines:
For the efficiency estimation, three samples are used: inclusive Jpsi2MuMu for tight prompt Jpsi line, inclusive Psi2S2MuMu for tight prompt psi2S line, B2JpsiK for detached Jpsi line. 500 events are used for each pattern.
Efficiency of SpruceBandQ_JpsiToMuMuTightPrompt_Line on inclusive Jpsi sample: 10%
Efficiency of SpruceBandQ_Psi2SToMuMuTightPrompt_Line on inclusive Jpsi sample: 26%
Efficiency of SpruceBandQ_JpsiToMuMuDetached_Line on inclusive Jpsi sample: 40%
Contribution to size/event for the new lines
To minimize the influence of stat. fluctuation, the size/event cost is estimated using the efficiency-estimation samples. For the three samples mentioned above, the average DstData bank size (bytes) values are slightly different (+-10%), and in general the value is roughly 100kB.
Rate and bandwidth estimation of relevant hlt2 lines
These four new lines mainly rely on four corresponding hlt2 lines. And the rates of them are:
Hlt2_Psi2SToMuMuDetachedLine: 8/50000 * 1MHz = 160Hz;
Hlt2_JpsiToMuMuDetachedLine: 19/50000 * 1MHz = 380Hz;
Hlt2_DiMuonPsi2STightLine: 2/50000 * 1MHz = 40Hz;
Hlt2_DiMuonJPsiTightLine: 14/50000 * 1MHz = 280Hz;
We take the size/event estimation from that of the fully-persisted sprucing lines： 100kB/event
Then the bandwidth cost should be roughly 90MB/s, which is negligible compared to the 5.9GB/s given in Table.4.3 of http://cds.cern.ch/record/2319756/files/LHCB-TDR-018.pdf?version=4 .
Summary
Several dimuon sprucings are added to further select the triggered events and save the disk storage. Selections are mainly taken from corresponding Run2 strippings and should be fine for most of the analyses. Efficiencies are tested, and we did not see crazy numbers indicating serious bugs in the code. For the storage cost: rate for psi2S lines are relatively small; rate for all the four lines are in total roughly 700Hz, and the badwidth cost is roughly 70MB/second.
Comments and suggestions ?
c.c. to BandQ and B2CC run3 migration proponents: @peilian @gcavalle @lan @ngrieser
Comments or suggestions from RTA and DPA experts, especially the ones about the bandwidth cost, would be appreciated: @nskidmor @shunan @abertoli @poluekt @mvesteri
Related to DPA grand task https://gitlab.cern.ch/lhcb-dpa/project/-/issues/78 on ensemble of Sprucing selections.