Transcript Diapositiva 1 - unisalento.it

Event Filter and Trigger Menu
L=1031cm-2s-1
Sergio Grancagnolo
on behalf of the Muon Trigger
software group
Outline
• News on EF algorithm implementation (for
rel. 13.x.x)
• HLT Data Quality assessment
• EF performance studies
– resolution/efficiency
– EF rates
– status of the studies of muons from p decay
• Muon trigger rates and trigger menu for LHC
startup
TrigMoore – a reminder
Two different running modes:
Seeded
Reconstruction performed only in the geometrical
regions provided by the RoIs of previous levels.
Full scan
Full reconstruction, ~equivalent to the offline
working mode
LVL1
LVL2 (muFast)
Seeding Algorithms assume the seed is
from LVL2 or a LVL1 ROI
LVL2 (muComb)
• Full functionality in barrel and endcaps
Seeding Algs
• 3 istances of TrigMoore called by
the steering, for reconstruction in
the MS, extrapolation to the IP and
Moore Algs
combination with ID tracks
• TrigMoore attaches to the TE a
"TrigMooreFeature" for each ROI, TrigMoore
MuIdStandAlone
accessed by TrigMooreHypo for pT
Algs
test
• TrigMoore records in SG the
MuIdCombined
TrigMooreFeature per each ROI and
Algs
all reconstructed tracks in the event
in a single container for conversion
Offline ID
in Trk:Track format and subsequent
output in ESD and AOD
LVL2 ID
Hypo Alg
Hypo Alg
Hypo Alg
MuonEF in Rel 13.0.10
• Migration to new HLT Steering (DONE)
– Both FEX and Hypothesis algorithms
• Migration to Configurables (DONE)
– TrigMoore configurables py classes available in 3 flavours
(MuonSpectrometer, Extrapolated and Combined Tracks)
– New style EF muon sequences (including ID) added in
TrigHypothesis/TrigMuonHypo
• Migration to new EDM (DONE)
– TrigMoore uses the standard reconstruction input object (PrepRawData).
– The Moore offline algorithm organization has been adopted also by
TrigMoore
• Use of EF version of ID for combined MS-ID tracks (DONE)
– MuidCombined adapted to get as input EF-ID tracks.
– Replacement of the previous implementation in wich EF ID and offline ID
(New Tracking or iPatRec) was used.
MuonEF in Rel 13.0.10
• Implementation of REGTESTs for Monitoring and Validation
(DONE)
• Monitoring Histo in Hypo Algo (DONE)
• Definition of jobOptions to include Muon EF in ATN test
(DONE)
• Bug Fix- Fix for duplicated python modules in genConf (DONEto be collected right after 13.0.10)
• Bug Fix- Problem in Persistification (appeared in 13.0.X nightly)
Fixed in 13.0.10
Muon HLT Data Quality
Muon Slice Data Quality (I)
Preliminary considerations on DQ monitoring and assessment for the
muon slice presented at TDAQ Data Quality Wshop in Zeuthen (February
28-March 1 2007), focusing on LVL2 and EF, in online/offline
e.g. for EF online monitoring:
• hit #/track per tecnology;
• geometrical track
parameters;
• pt spectrum;
• track quality (2);
•
•
•
•
•
residual distributions;
matching with ID;
combined tracks/moore tracks;
ratio of positive/negative tracks;
matched hits/total hits in the seed
region;
Monitoring activities are going to be organized for LVL1/2/3 in the
appropriate environments, to use general tools and DQ Monitoring
framework, to move the first steps towards muon trigger slice DQ
Nothing exists for the moment for Muon Slice DQA but what implemented for
monitoring during 2004 test beam (for LVL2) and the test of the trigger slices on the
pre-series machines at Point 1 in december 2006 (for EF) can be a starting point for
Data Quality Monitoring
Muon Slice Data Quality (II)
e.g. TrigMoore histos from the last technical run
Trigger/TrigAlgorithms/TrigMoore/src/TrigMooreHisto.cxx
(here obtained running the jobOptions prepared for the on-line with a bytestream file containing
50 top events as input: muons are selected by the LVL2 and the EF muon algorithms)
completed in release 13 for TrigMoore Hypothesis algos
EF Performance studies
Sample and releases
• Single mu sample centrally simulated on lcg-dq grid (CSC
production)
– mixed charge;
– pT from 2 to 200 GeV/c
– total statistics ……
efficiency, resolution,
single muon trigger rates
• Reconstruction with a dedicated sw chain in 12.0.6
– Muon Trigger slice only
• LVL1 emulation, (100% of the statistics processed with standard LVL1
configuration, part of the low pt points re-processed with low thresholds)
• LVL2: mufast, IDscan, muComb
• EF(seeded by LVL2): trigmoore + EF ID algo
• Hypothesis algorithms disabled
– Private AA NTuple production using grid tools (GANGA) on Napoli
Tier2 storage elements (mostly by Napoli group)
– Used for both LVL1 (Napoli, Tokyo) and EF (Lecce, Napoli) studies
– Main goal: CSC note
Efficiency and resolution curves
(standard LVL1 configuration)
barrel EF MuId Combined efficiency
curves w.r.t. LVL2 (muComb)
6 GeV
8 GeV
20 GeV 40 GeV
pT (GeV)
barrel EF resolutions
pT (GeV)
MuidStandAlone
TrigMoore
MuidCombined
EF MuId Combined Resolutions
Endcap
Overlap
Very Fwd
All
Other EF Muon Trigger rates
6 GeV
1033cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
8 GeV
1033cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
beauty
641
845
beauty
172
291
charm
327
426
charm
77
134
top
0.05
0.07
top
0.06
0.05
W
2.9
4.0
W
2.8
3.9
p/K
1918
1462
p/K
281
313
TOTAL
2889
2737
TOTAL
533
742
20 GeV
1034cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
40 GeV
1034cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
beauty
73
118
beauty
2.5
4.5
charm
28
46
charm
0.87
1.6
top
0.27
0.32
top
0.07
0.07
W
22.3
32.6
W
3.9
7.1
p/K
50
48
p/K
0.2
0.3
TOTAL
173
244
TOTAL
7.5
13.6
EF Efficiency curves (for low pt regime)
EF (MuId CB) efficiency w.r.t. LVL2 (muComb)
barrel
barrel
“4 GeV”
3
5
pT (GeV)
5 GeV
3
5
pT threshold at 4 GeV corresponds to a totally open LVL1
window – EF effective threshold is 3 GeV/c
pT threshold at 5 GeV has a cut lowered by 2 times the
resolution corresponding at 5 GeV (as for higher thresholds)
pT (GeV)
Low pT thresholds single muon rates
Luminosity: L = 1031cm-2s-1
“4 GeV”
1031cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
5 GeV
1031cm-2s-1
Barrel
(Hz)
Endcap
(Hz)
p/K
133
80
p/K
44
30
beauty
19
21
beauty
11
13
charm
11
12
charm
6
7
top
6∙10-4
8∙10-4
top
5∙10-4
7∙10-4
W
0.03
0.04
W
0.03
0.04
TOTAL
163
113
TOTAL
61
49
Trigger Menus Proposal for L=1031 cm-2s-1
L1 trig.
item
Rate
LVL2
Rate
μComb
MU4
MU6
227 Hz
112 Hz
DiMuon
1 Hz
PT
1 Hz
μComb
80.8 Hz TrigMoore 56.3 Hz
InDet only
PS/PT
μFast only 99.7 Hz
PS/PT
PS/PT
MU10
Rate
TrigMoore
~1 kHz TrigDiMuon
PS/PT
EF
1 Hz
μComb
PS/PT
PT
1 Hz
TrigMoore ~10 Hz
1 Hz
PT
1 Hz
Trigger Menus Proposal for L=1031 cm-2s-1
L1 trig.
item
Express streams
MU15
MU20
MU40
2MU4
2MU6
2MU10
2MU20
Rate
LVL2
Rate
EF
Rate
μComb
TrigMoore ~2 Hz
InDet only
PT
19 Hz
μFast only
PT
PS/PT
1 Hz
PT
1 Hz
14 Hz
PT
14 Hz
PT
14 Hz
8 Hz
PT
8 Hz
PT
8 Hz
~9 Hz
???
PT
4 Hz
PT
4 Hz
PT
4 Hz
~1 Hz
PT
~1 Hz
PT
~1 Hz
< 1 Hz
PT
< 1 Hz
PT
< 1 Hz
Trigger efficiency from Z → μ+μDouble Object (DO) method
Backgrounds from BBμμX, Wμv, Zττ
Different reconstruction modes
1-2 % statistical uncertainty with few pb-1 data
Differential (η,φ) trigger efficiency determination
CSC AOD analysis is going to be finalized
Double Object with orthogonal
Signature (DOS) method
Muons from p/K
Strategy
• extract from minimum bias and dijets of various
energy a global pT vs h distribution for pions and
kaons, re-weighting to a common value of the
integrated luminosity
• Use the particle-generator to generate single
pions according to the previous distribution
• simulate the decay of pions with G4 saving only
events were the pion decay before the muon
spectrometer
• study a procedure to identify and reject this kind
of muons
Event Sample
Event
Sample
File Name
Events
Cross Section
(mb)
minimum
bias
csc11.005001.pythia_minbias.evgen.EVNT.v11004202._xxxxx
124200
92E00
J1
csc11.005010.J1_pythia_jetjet.evgen.EVNT.v11004201._xxxxx
242287
1.376E00
J2
csc11.005011.J2_pythia_jetjet.evgen.EVNT.v11004201._xxxxx
257508
9.327E-02
J3
csc11.005012.J3_pythia_jetjet.evgen.EVNT.v11004201._xxxxx
30299
5.884E-03
J4
csc11.005013.J4_pythia_jetjet.evgen.EVNT.v11000401._xxxxx
40400
3.084E-04
J5
csc11.005014.J5_pythia_jetjet.evgen.EVNT.v11000401._xxxxx
5050
1.247E-05
J6
csc11.005015.J6_pythia_jetjet.evgen.EVNT.v11000401._xxxxx
50000
3.604E-07
J7
csc11.005016.J7_pythia_jetjet.evgen.EVNT.v11004201._xxxxx
5050
5.707E-09
J8
csc11.005017.J8_pythia_jetjet.evgen.EVNT.v11004201._xxxxx
10100
2.444E-11
Particle selection requiring:
• p PDG code
• |h|<2.7
• pT>500MeV
• generation at the interaction point (0.,0.,0.)
and filling a two-dimensional histogram of pT vs h, one entry for each pion.
pT vs h distributions
Minimum bias
Dijets J8
Some existing correlation between pT and h requires
a generation according to the 2-D distribution
The solution that I adopted is to slice the 2-D
histograms in pT bins and simulate using the 1-D
projections in pT and h
Supposing an efficiency e=10-4 between 2.5 GeV
and 7.5 GeV, and a granularity of 500 MeV, to
extimate an error of s(e)10%
 106 events needed in each bin
The red curve, mu6 eff. at EF, is
estimated with 10 pT points (2GeV/c –
8 GeV/c) with 106 – 0.5·106 single
muon each but events passing LVL2
are 30 at 3 GeV/c, 500 at 4 GeV/c,
3600 at 5 GeV/c, 350k at 6 GeV/c
If only ~1% of the pions decays in the
volume before the calorimeter
at least 109 events should be generated
Single muon
efficiency
A tool was developed to force all
the generated pions to decay:
the PionDecayer
Produced sample
• no available tools to randomly extract from a 2D
distribution
– solution: slice the 2D distribution in pt bins (of ~same content)
and generate according to the 1D distributions of pT and Eta
• From PANDA, centrally simulated files:
Task name
Task IDReq JobsDone JobsTotal events Prio GridStatus Timestamp
misal1_mc12.007180.singlePion_pTSlice_20_of_30.digit.v12000502 7822 500 500 25000 100 osg done Apr 18 10:04
misal1_mc12.007162.singlePion_pTSlice_2_of_30.digit.v12000502 6296 1600 1600 80000 100 osg done Mar 1 12:34
misal1_mc12.007163.singlePion_pTSlice_3_of_30.digit.v12000502 6295 1400 1400 70000 100 osg done Mar 1 12:33
………
misal1_mc12.007189.singlePion_pTSlice_29_of_30.digit.v12000502 6270 100 100 5000 100 osg done Mar 1 11:46
misal1_mc12.007190.singlePion_pTSlice_30_of_30.digit.v12000502 6269 100 100 5000 100 osg done Mar 1 11:45
TOTAL: 25.800 files 1.065.000 events ~1.7Tb
Since, for a given pion direction, the
decay is forced into a definite path
length L, the single event probability is
LM
P ( p, L ) 
p
Following steps…
• Since full statistics has become available all efforts have been directed
to apply the muon slice and the standard ntuple analysis to the single p
sample
– with some addition specific to the pm case
• keep track of the muon mother and its decay vertex
• keep track of the links between a combined reconstructed muon and its ID and
MS seeds, etc...)
• Fixes made by Alan and migration of methods from MuidMatch to MuidTrack
• first attempt:
– copy from osg grid to castor, submit jobs to lxplus queues for running the
specific reconstruction + analysis code
 turned out to be unaffordable mostly due to castor access too slow and unstable
• use of grid infrastructures absolutely necessary !
– the solution: use lcg-dq after replicating the files to Napoli tier-2
• affordable but still very time consuming;
• copy from castor to NA tier-2 took 4/5 days only for higher pT slices (20-30)
– reconstruction jobs submitted on a small part of these replicas
• A long process for the study of a critical issue: now all necessary input
ready to perform the analysis and get some results
Problems
• Many problems encountered to complete
the analysis chain:
– Last simulation job finished ~ half april
– Copy output to castor from US grid ~1 week
– Many attempts to run the muon trigger slice
•
•
•
•
•
Run athena jobs directly on castor  failed
Copy from castor to local nfslocal disk failed
Replicate from castor to european gridongoing
Running muon trigger sliceongoing
Run analysis macroongoing
Conclusions
• Many important steps since last T&P week
• Performances under control
• Data quality and trigger menus under
study
• Still waiting to finish the study on muons
from single pion sample, but many
technical problems
– grid tools are already very important
• Stay tuned for new results on this topic