Upload
sheryl-lamb
View
219
Download
0
Tags:
Embed Size (px)
Citation preview
Muon Event FilterStatus Report
Sergio Grancagnolofor the muon EF working
groupINFN Lecce & Salento University
Outline
• Performance in releases 12, 13, 14• Robustness to misalignments• Muon EF wrappers for offline
supertools• Converters• Menus• Muon trigger CSC note status
Muon Slice performanceFrom CSC Note – Muon Rates - Convolution method 12.0.6 trigger swLVL1 emulationmuFast+hypomuComb+hypoTrigMooreMSTrigMooreSATrigMooreCB+hypo
Pythia 6.4 for b,c DPMJET for /K(conservative choice)large theoretical uncertaintiesat pT below 6 GeV
must focus on /K rejection at low pT
mu6 needs prescaling mu20/40 need isolation
or prescaling at 1034
Efficiencies
• a drop of the CB trigger eff. at low pT has been observed in 13.0.30 – recovered in 14.x.0 nightlies – appears related to EFID rather than Muon Algos
• Stable eff. at high pT
12.0.613.0.30.3
TrigMooreCB w.r.t. muCombsingle muons
12.0.613.0.40.214.0.0 nightly
13.0.40.214.0.0 nightly
EFId w.r.t. muCombsingle muons
ZZ
Single muons mc12.*.RDO.* Single muons mc12.*.RDO.* • ATLAS-CSC-01-02-00ATLAS-CSC-01-02-00
– 12.0.612.0.6– 13.0.40.213.0.40.2– 14.X.0 devval, rel_3 (12 Mar 2008)14.X.0 devval, rel_3 (12 Mar 2008)
Resolution 1/pT
12.0.6
13.0.40.2
14.X.0
MuIdSA MuIdCB
MuIdSA performance is more or less constant for all releases
MuIdCB shows slight deterioration in 14.X.0 wrt previously
Resolution and 12.0.6
13.0.40.2
14.X.0
MuIdCBMuIdSA
and resolution stable with releases
MuIdSA MuIdCB
/K rejection
single single
min.bias K
min.bias
K
From the CSC note
with current set of cuts (not extensively optimized)– 75% eff. @ 4GeV for (1-effBG)=35%
– 90% eff. @ 20GeV for (1-effBG)=25%
Inner Detector d0 ≤ 0.15ID n. of pixel hits ≥ 3ID n. of B layer hits ≥ 1ID n. of SCT hits ≥ 6pT (ID) / pT (SA) ≤ 1.25Combined match 2 ≤ 26
present cuts
to be implemented in a new hypo algorithm for 14.x.x
Robustness against misalignments
• RDO reconstruction and AANT/AOD production performed with Athena 13.0.40.2
• Four reconstruction scenarios considered with different knowledge of ID and/or MS alignment:– Aligned ID and aligned MS Aligned ID and aligned MS (perfect correction of misalignments)– Aligned ID and misaligned MSAligned ID and misaligned MS (crude MS misalignments ~1mm)– Misaligned ID and misaligned MS Misaligned ID and misaligned MS (ID as in FDR1 + MS as above)– Misaligned ID and aligned MSMisaligned ID and aligned MS (ID as in FDR1 + perfectly known
MS)
EF_InDet EF_InDet wrt LVL2wrt LVL2 EF_MuidCB EF_MuidCB wrt LVL2wrt LVL2
Dramatic loss of EF algos’ efficiency with misal ID.
Misal MS even worse at high pT
Event-based efficiencies applying lowest muon threshold
Efficiencies
Resolution 1/pT
Standard deviation of pT(Reco) / pT(Gen)
EF EF InDetInDet
EF EF MuidSAMuidSA EF EF MuidCBMuidCB
3%
5%
5%
3%
5%
3%
2%
EF InDet pT reso 2-3 times larger with misal ID
IDal-MSal • IDal-MSmisal • IDmisal-MSal • IDmisal-MSmisal
MuidCB pT reso below 10% with misal ID, out of control at >100 GeV with Misal MS
Resolution
Standard deviation of (Reco) – (Gen)
EF EF InDetInDet
EF EF MuidSAMuidSA EF EF MuidCBMuidCB
3x10-3
5x10-4
2%
3%
5%
5x10-3
3x10-3
2x10-3
2x10-4
3x10-4
5x10-4
MuidCB resolution 2-3 times larger with
misal ID/MS
MuidSA performance unchanged
EF_ID resolution constant with misal
ID
IDal-MSal • IDal-MSmisal • IDmisal-MSal • IDmisal-MSmisal
Resolution:
Standard deviation of (Reco) – (Gen)
EF EF InDetInDet
EF EF MuidSAMuidSA EF EF MuidCBMuidCB
2%
3%
5x10-3
3x10-3
2x10-4
3x10-4
5x10-4
2x10-3
2x10-3
2x10-4
3x10-4
5x10-4
MuidSA performance unchanged
EF_ID resolution constant with misal
ID
MuidCB worse at high pT with misal MS, resolution ~
mrad with misal ID
IDal-MSal • IDal-MSmisal • IDmisal-MSal • IDmisal-MSmisal
Robustness considerations Tests on muon trigger robustness
performed on (mis)aligned data, running the full Muon Slice shows that With crude non-realistic (~mm) MS
misalignments + realistic ID misalignments (FDR1): EF efficiencies deteriorate significantly EF spatial resolution under control, but pT gets
worse above 50-100 GeV (below 10% with ID misal. only)
Converters – state as today• MuonCnv/MuonByteStream
– bytestream Converter to RDO for Mdt, Rpc, Tgc, Csc– bytestream Converter to Mdt PrepData– bytestream Converter to Rpc PrepData but no
ambiguity/redundancies solving• For trigger technologies online data format very different wrt offline
– no bytestream to PrepData Converter for Tgc, Csc today• MuonCnv/MuonRdoToPrepData
– Conversion Algorithms from RDO to PrepData for all technologies • uniform in general structure among technologies • implemented for the MC: input is the RDO !• can be run on bytestream: when the algorithm retrieve the
RDO, the bytestream to RDO Converters are activated and the RDO become available in memory for further conversion (as in M5)
• converting full event even when working in a region of interest !!
• for RPC the reason is having clean RIO collections – no redundancies/ambiguities
• NEED to preserve the RoI driven data access
New data access schema for New data access schema for RPC (I)RPC (I)
• @ the EF, the algorithm (FeX) requests RPC PRD in a x region (RoI)– get from the region selector a list of data collection hash id (offline) in
the RoI– get from cabling service the corresponding list of pad identifiers– Retrieve from StoreGate the pad container and find (i.e. decode into
RDO) each pad in the selected list output is the RDO for the RoI• i.e. just use the byte-stream to RDO converter
– convert the RDO into PRD by applying the redundancy and ambiguity removal currently implemented in the offline algorithm for RDO PRD
• First implementation of the schema proposed for RPC offline / Event Filter is ready / under test
• in the offline (re-use not duplicate code) MC and DATA– RDO to PRD delegate all the work to the new AlgTool used over the
whole event allow to use the StatusCode + an attached string describing the kind and severity of the problem in data preparation robustness
• The schema can be adopted for all muon technologiesThe schema can be adopted for all muon technologies– uniform way of accessing the data
embed all that into an AlgToolembed all that into an AlgTool
TrigMoore implementation
LVL2 (muFast)LVL2 (muFast)
Moore AlgsMoore Algs
LVL1LVL1
MuIdStandAloneMuIdStandAloneAlgs Algs
TrigMooreTrigMoore
Seeding AlgsSeeding Algs
MuIdCombinedMuIdCombinedAlgs Algs
Hypo AlgHypo Alg
Hypo AlgHypo Alg
Hypo AlgHypo Alg
LVL2 (muComb)LVL2 (muComb)
Inner DetectorInner DetectorAlgs Algs
• Seeding Algorithms assume the seed from LVL1 or LVL2, barrel or endcap
– Reconstruction only in the geometrical regions provided by the RoIs of previous levels
• 3 instances of TrigMoore are called by the steering for:
– Reconstruction in MS– Extrapolation to the IP– Combination with ID tracks
• Each TrigMooreFeature is accessed by TrigMooreHypo to test hypotheses
Follow up with the new Moore style for
14.0.0 • A new package: Trigger/TrigAlgorithms/TrigMuonEF hosts 4
HLTalgorithms: – TrigMuonEFSegmentFinder wrapper for MooSegmentCombinationFinder (offline
tool invoqued with vector of pointers to prd collections) DONE – TrigMuonEFTrackBuilder wrapper for the offline tool MuonCombiTrackMaker
(invoqued with vector of pointers to segment combination) DONE – TrigMuonEFExtrapolator wrapper for MuidBackTracker DONE– TrigMuonEFCombiner wrapper TO BE DONE for the main tool of MuidCB
(COMING SOON ?)
• Running with release 14.0.0 - producing standalone EF muons propagated to the interaction point
• Most work is just in configuration– requires dedicated configuration of the tools for running in trigger
• The new chain (up to the available implementation) will be in 14.1.0 – defined temporary extra muon trigger sequences to be run concurrently with
the standard EF (TrigMoore)– needs to revise the trigger AOD for the EF
New EF wrappers
LVL2 (muFast)LVL2 (muFast)
TrigMuonEFTrackBuilderTrigMuonEFTrackBuilder
LVL1LVL1
TrigMuonEFExtrapolatorTrigMuonEFExtrapolatorTrigMuonEFTrigMuonEF
TrigMuonEFSegmentFinderTrigMuonEFSegmentFinder
TrigMuonEFCombinerTrigMuonEFCombiner
Hypo AlgHypo Alg
Hypo AlgHypo Alg
Hypo AlgHypo Alg
LVL2 (muComb)LVL2 (muComb)
Inner DetectorInner DetectorAlgs Algs
MooSegmentCombinationFinder
MuonCombiTrackMaker
New implementation schema of muon EF code (to be implemented for 14.1.0)With wrappers for offline tools
MuidBackTracker
Hypo AlgHypo Alg
Combination tool
In red: to be implemented
The slow particle trigger TrigMuGirl will be also included in 14.1.0 or 14.2.0
Revision of muon EF trigger AOD
• The separate step of segment finding allow for another early hypothesis algo– As an example if the number of segments in
inner/middle/outer MS layers is stored– Fit 2 could be saved at different track building
steps• To implement hypo algos for /K rejection
– At least combined match 2 fit must be stored• Link to the ID track used in combination
– Use of ElementLink if ID info is already stored in AOD
– Store ID info if missing (21 bytes)• Proposal made @ last Muon Slice Trigger
Software (3 Apr) for a new schema– Possibility under study
Muon Trigger Menus
• A lot of effort has been put in cross checking the trigger rate estimated with the convolution method and with the counting method (on min. bias Pythia 6.4 samples)– a basic agreement is observed taking into account the
different theory input for /K rates– full details can be found in reports at Nov. Trigger and
Physics week and at Dec.2007 TDAQ week• Based on estimated trigger rates the muon trigger menus
have been defined for 1031 and 1032 assuming standard muon slice algorithms (no isolation yet, no /K rejection strategies)
• first though for 1033 based on extrapolation of 1032
– Need to account for minimum bias pile-up – Reliable samples needed – implications of 1033 luminosity at 75 ns bunch
crossing spacing
PHYSICS MENU -13.0.40.4 (PHYSICS MENU -13.0.40.4 (TriggerMenuPython-00-00-47-34TriggerMenuPython-00-00-47-34) + rates) + rates
SIGNATURESSIGNATURES E32E32
LVL1LVL1 LVL2LVL2 EFEF RATES (Hz)RATES (Hz)
PSPS PTPT PSPS PTPT PS PS ratesrates
PT PT ratesrates
Total Total ratesrates
mu4 1 5*10 400*10 300 0 0.2 0 0.2
mu6 1 1*10 200*10 30 0 3.5 0.1 3.6
mu10 1 1 100*10 1*10 0 18 1 19
mu15 1 1 20*10 1*10 0 5.7 0.2 5.9
mu20 1 1 0 1 0 19 0 19
mu40 1 1 1*10 1 0 1 14 15
mu20_passHLT 1 1000 20*10 1 0 0.02 ~0.2 0.22
2mu4 1 1*10 5*10 1 0 2.4 1.3 3.7
2mu6 1 1 1*10 1 0 7 10 17
mu4_mu6 1 1 1*10 1 0 14 33 47
2mu10 1 1 0 1 0 1 0 1
2mu20 1 1 1 1 0 0 2 2
3mu6 1 1 1 1 0 0 0 0
Menus (13.0.40.4) & Rates for of 10Menus (13.0.40.4) & Rates for of 1032 32 (13 (13th th February)February)
PHYSICS MENU -13.0.40.4 + ratesPHYSICS MENU -13.0.40.4 + rates
SIGNATURESSIGNATURES E32E32
LVL1LVL1 LVL2LVL2 EFEF RATES (Hz)RATES (Hz)
PSPS PTPT PSPS PTPT PS PS ratesrates
PT PT ratesrates
Total Total ratesrates
mu4 1 5*10 400*10 300 0 0.2 0 0.2
mu6 1 1*10 200*10 30 0 3.5 0.1 3.6
mu10 1 10 100*10 1*10 0 1.8 <1 <2.8
mu15 1 1 20*10 1*10 0 5.7 0.2 5..9
mu20 1 1 0 1 0 19 0 19
mu40 1 1 100 1 0 1 1.4 2.4
mu20_passHLT 1 1000 20*10 1 0 0.02 ~0.2 0.22
2mu4 1 1*10 5*10 1 0 2.4 1.3 3.7
2mu6 1 1 100 1 0 7 1 8
mu4_mu6 1 10 200 1 0 1.4 <1.5 <2.9
2mu10 1 1 0 1 0 1 0 1
2mu20 1 1 1 1 0 0 2 2
3mu6 1 1 1 1 0 0 0 0
mu20i mu20i 1 1 0 1 0 5 0 5new
PROPOSAL FOR OPTIMIZATION OF 10PROPOSAL FOR OPTIMIZATION OF 103232
Very fast estimate of the cumulative rate gives : Very fast estimate of the cumulative rate gives : ~50 Hz (against 112 as before)~50 Hz (against 112 as before)
FDR-2 Release13.0.40.4 ( FDR-2 Release13.0.40.4 ( tag 00-00-47-33tag 00-00-47-33))
SIGNATURESSIGNATURES LVL1LVL1 E32E32 E33E33
E32E32 E33E33 LVL2LVL2 EFEF LVL2LVL2 EFEF
PSPS PTPT PSPS PTPT PSPS PTPT PSPS PTPT
mu10 100 100 * 10 1 0 1 0 1 0 1 0
mu20i/ mu20i_tight ( for1033)
1 1 1 0 1 0 1 0 1 0
mu40 / mu40_tight (for 1033)
1 1 1 0 1 0 1 0 1 0
2mu4 (E32) 10 No 1 0 1 0 No No No No
2mu6 1 1 * 10 1 0 1 0 1 0 1 0
2mu10 (E32) 1 No 1 0 1 0 No No No No
2mu15 (E33) No 1 No No No No 1 0 1 0
Not yet inside the menu definition
Status of Muon Trigger CSC Note
• Comments received from referees before Easter– Almost all comments have been implemented
and the remaining ones are going to be finalized
• Mainly editorial work– Figures to be uniformed in style
• New draft sent to referees for further comments
Conclusions • Muon EF performs satisfactory in releases 13 and
14• Several studies of robustness against misalignment
– Need to better understand how much realistic are the considered scenarios
– effort on-going for having realistic (day 0) MS misalignments for FDR2 (~100 microns)
• Muon EF wrappers for offline supertools planned to enter in release 14.1.0– Muon EF AOD under study
• Reenginering of Converters: common effort for different technologies
• Muon trigger menus @ 1032 under optimization– first thoughs for 1033
• Muon trigger CSC note in final shape
Backup
pT(Reco) / pT(Gen) distributions 6 GeV Single Muons Resolution : pT
MuIdCB
12.0.6 13.0.40.2
14.X.0
pT(Reco) / pT(Gen) distributions
19 GeV Single Muons Resolution : pT
MuIdCB
12.0.6 13.0.40.2
14.X.0
Systematic underestimate for release 12.0.6 . This problem seems overcome in the subsequent releases
pT(Reco) / pT(Gen) distributions
100 GeV Single Muons Resolution : pT
MuIdCB
12.0.6 13.0.40.2
14.X.0
pT resolution looks stable wrt releases at high transverse momentum
Mean of pT(Reco) / pT(Gen) gaussian fit
Mean 1/pT : EF MuidSA & MuidCB
12.0.6
13.0.40.2
14.X.0
MuIdSA MuIdCB
No offset appear in reconstructed EF pT
Beam halo
• Samples by A. Stradling:– /castor/cern.ch/user/s/stradlin/PileupNM/
misal1_mc12.007499.singlepart_empty/halo
The Muon Trigger seems unaffected by muons in beam halo (statistical significance to be evaluated according to the luminosity of the simulated sample)
54294 generated events 0 at LVL1 & HLT
ge
n
pT gen (MeV)
pT gen (MeV) gen (rad)
RPC Raw Data and Prep Raw Data
On-line: RDO
Offline: PrepRawData
on-line hits different from off-line hits h.w. duplicated hits due to cabling overlap s.w duplicated hits due to wired-or and logical-or 3 types of trigger hits in the readout
We have to deal with a very different structure of the online and offline data model
Current path to RPC PrepData @ EF
• RPC bytestreamRDOprep data i.e. running the offline algorithm for RDOPRD conversion which, as first step, activates the converter bytestreamRDO– run before the steering (not trigger driven!!!)– converting full event even when working in a region of interest
• inefficient – the reason is having clean PRD collections
• no redundancies /ambiguities– current byte-stream to PRD converter does not solve
overlaps/ambiguities because it scans sequentially the byte-stream and output a PRD for each relevant CM hit
– data clean-up requires to apply some logic on top of raw hits (which must necessarily be organized and stored in some format)
– the most natural and well tested format for that is the RDO