Download - Artemis School On Calibration and Performance of ATLAS Detectors Jörg Stelzer / David Berge
Artemis SchoolOn Calibration and Performance of ATLAS DetectorsJörg Stelzer / David Berge
3 Level Trigger System Level 1:
Hardware based Coarse granularity detector data Calorimeter and muons only Latency 2.2 s Output rate up to ~75 kHz
Level 2: ~500 farm nodes(*) Seeded by level 1 Only detector ”Regions of Interest”
(RoI) processed Full detector granularity in RoIs Fast reconstruction Average execution time ~40 ms Output rate up to ~2 kHz
Event Builder: ~100 farm nodes(*)
Event Filter (EF):~1600 farm nodes(*) Seeded by level 2 Full detector granularity Potential full event access Offline algorithms Average execution time ~4s Output rate up to ~200 Hz
(*) four dual-CPU cores at ~2GHz per farm node19. Sep 2008Trigger Tutorial Introduction 2
The Level 1 Trigger
Pre-processor
ClusterProcessor
Jet/EnergyProcessor
End-cap MuonTrigger (TGC)
Barrel MuonTrigger (RPC)
Muon-CTP-Interface (MuCTPI)
Central Trigger Processor (CTP)Central Trigger Processor (CTP)
LTP
BusyTTC
Detector Front-Ends/Read-out
LTP
BusyTTC
Muon DetectorsCalorimeter Detectors
Common Merger Modules
Trigger objects: Muons, EM and hadronic clusters, jets, total and missing ET
CTP
Available in AOD:
19. Sep 2008 3Trigger Tutorial Introduction
The High Level Trigger
HLT configuration described by the trigger menu Trigger chains: signatures (1 electron and
some missing ET) Description how electrons and missing ET
are calculated and what selection criteria are applied: sequence of algorithms
Algorithm properties are also part of trigger configuration Information not available for analysis
19. Sep 2008 4Trigger Tutorial Introduction
HLT Configuration Concept of Chains (trigger
lines) Chain: list of trigger conditions
(multiplicities of HLT TriggerElements) to be evaluated in order
Sequence: Description how algorithms produce TriggerElements example: EM3 ClusterFinder &
Hypo L2_e5cl
Collection of Chains (each with prescale and path-through rates) is called HLT Menu
Available in AOD: chains, trigger conditions,
TriggerElements, prescales, pass-through
Not available: algorithm names
signature (2e j)
sequence (e) [EM “e-FEX, e-Hypo” e]
sequence (j) [JET “j-FEX, j-Hypo” j]
signature (e’ j’)
Chain (2EJ-L2) input = “2EMJET”
Chain (2EJ-EF) input = “2EJ-L2”
Lvl1 Trigger Item 2EMJET
y/n
y/n
y/n
y/n
L2
EF
HLT Chain
19. Sep 2008 5Trigger Tutorial Introduction
TriggerDB
Easy trigger configuration via 3 integer keys
Easy reproducibility Configuration history
19. Sep 2008Trigger Tutorial Introduction 6
Level 1 Menu + PrescalesHLT Menu + PrescalesAlgorithm parametersRelease version
ConditionsDatabase
(COOL)
TriggerDB Schemareflects trigger design
Configuration data becomes conditions data when used for data taking
Relational database stores trigger configuration – TriggerDB
Trigger Features and TriggerElements Sequence: description how algorithms produce
TriggerElements, example (in xml): <SEQENCE inputTEs=“EM3”
algorithms=“T2CaloCluster T2CaloHypo_5” outputTE=“L2_Photon_5”/>
At least two algorithms in sequence First is a FEX_algo which calculates a “trigger feature” e.g. a
TopoCluster This feature is attached to the TriggerElement “L2_Photon_5” More features can be calculated and attached to this TE Last is a Hypo_alg which checks the feature (e.g. “energy of
TopoCluster>5?”) and enables/disables TriggerElement
Trigger features are ‘physics objects’. TriggerElements are logical objects (feature based selection decision)
19. Sep 2008 7Trigger Tutorial Introduction
Two More Terms
HLT Steering (during trigger running): Framework that executes chains of a
menu (after application of prescales/pass-throughs)
Stops chain execution if condition is failed Caches results of algorithms that are called
multiple times
HLT Navigation (during AOD analysis): Framework that enables access to
trigger features through chains or trigger elements
19. Sep 2008 8Trigger Tutorial Introduction
The Flow of Trigger Data1.TriggerDB to
configure trigger for data taking
2.Configuration data to COOL
3.Trigger result in each event
4.Shipped to reconstruct-ion sites
5.ESD, AOD, TAG for trigger aware analysis
19. Sep 2008Trigger Tutorial Introduction 9
ES
D 1
00
MB
/s
AO
D 2
0M
B/s
ESDAOD
TAG files/DB
L1Result
to
Tier0
express calib
Tie
r 1
tr
an
sfer
Tier 0
Prompt Reconstruction
Express Reconstruction,calibration
Tier 1 ReprocessingTier 2MC production
ConditionsDatabase
Tri
gger
Menus
into
Condit
ions
DB
ConditionsDatabase
TriggerDB
DbProxy
LVL2 Result
EFResult
RODs
Front-end
LVL2
Subfarm Input
EF EF EF
Event Builder
LVL1/CTP
Subfarm Output
EF
Trigger Result
Trigger ObjectsTriggerDBReplication
DPD
Trigger Data for Analysis
19. Sep 2008Trigger Tutorial Introduction 10
TriggerDBAll configuration
data
OnlineConditionsDatabase
COOL
ESD
AOD
TAG
Configure
s Store
s
decod
e
d
Trig
ger
Men
u
DPD
With
decre
asin
gam
ou
nt o
f deta
il
The TrigDecisionTool InterfaceAccess to Trigger result (L1 items passed, prescaled,
vetoed; HLT chains passed, prescaled, pass-through) HLT::CTPItem, HLT::Chain
Trigger features through TrigNavigation
Trigger configuration TrigConf::TriggerItem, TrigConf::HLTChain Not mentioned yet: access to Streams,
ChainGroups19. Sep 2008 11Trigger Tutorial Introduction
Goal of the Tutorial
1. Check out and run standard UserAnalysis
2. Simple printout of trigger menu
3. Print full chain L1 – L2 – EF
4. Print simple trigger statistics of a few triggers
5. Determine trigger efficiencies with respect to an offline selection
6. Plot trigger efficiencies as function of pT and
η
Homework: Look at overlap of different triggers 19. Sep 2008 12Trigger Tutorial Introduction