Transcript Document

Muon Event Filter
Stefania Spagnolo
Dip. Fisica, Univ. Lecce e Sezione INFN, Lecce
on behalf of
the Muon HLT working group
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Outline
Brief reminder
the requirements; the implementation; the studies so far
Status of the TrigMoore package
Performance on recent releases (focus on the End-Cap region)
resolution and efficiency
Trigger rates at the EF vs threshold
Plans
sw developments and physics studies
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Brief reminder
Physics and system Requirements
EF must reduce the total data throughput to 200 Hz
EF total latency ~ 1-2 s
PT reach of the lvl1 muon trigger  5 GeV/c (lower limit)
Muon rate is dominated, at low pt, by decays in flight of light mesons,
b and c quarks
LVL1 muon rate for pT > 6GeV  25 kHz @ 1033 cm-2s-1
Cavern background and pileup at high luminosity are a source of fake
muons
Physics needs high efficiency and selectivity, flexibility in order to
cope with different background scenarios
The implementation
TrigMoore: an interface to the HLT Steering for the offline muon
reconstruction algorithms: MOORE and MuId
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Brief reminder
TrigMoore:
LVL1
a schematic view
Two different running
modes:
Seeded
Reconstruction
performed only in the
geometrical regions
provided by the RoIs of
previous levels.
Full scan
TrigMoore
Full reconstruction,
~equivalent to the
offline working mode
LVL2 (muFast)
Seeding Algs
Thanks to the
ID tracking group;
particularly to
Alan Poppleton
(iPatRec default
ID reco)
Moore Algs
Hypo Alg
MuIdStandAlone
Algs
Hypo Alg
MuIdCombined
Algs
Full functionality in barrel and end-caps
Seed can come from LVL1 or LVL2
Hypo Alg
ESD and AOD from the muon EF available
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Brief reminder
Status at the last T&P week
from the side of the algorithm:
TrigMooreHypo – first implementation
developments for muon EF output in ESD and AOD; validated by
comparing TrigMoore AOD with Moore AOD in the same job (11.0.5)
(slightly out of the scope of HLT) MuId migrated to use as input the
generic track object: Trk:Track; Allows to choose among different
reconstruction algorithms in the ID
From the side of the performance:
timing measurements on single muons with pile-up and nominal
cavern background (400 ms average, 500 ms RMS) – room for
improvement !
trigger rates estimated in the BARREL shown at the March T&P week
(on Rome-workshop data) confirmed by with more recent simulation
samples and software suite
studies of fakes at high luminosity
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Status of the TrigMoore package
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Status of the TrigMoore package
in 12.0.3
Seeding Algorithms assume the seed is a
MuonFeature (from muFast @ LVL2) or a LVL1
roi (seed from LVL1)
3 istances of TrigMoore called by the steering,
for reconstruction in the MS, extrapolation to the Seeding Algs
IP and combination with ID tracks, allow hypo
algos to be run at each step (if requested)
Moore Algs
TrigMoore records in the TE history a
MooTrackContainer or a MuidTrackContainer for
each seeding ROI and they are accessed by
MuIdStandAlone
Algs
TrigMooreHypo for pT test
TrigMoore records all reconstructed tracks in the
MuIdCombined
event in a single container in StoreGate for
Algs
conversion in Trk:Track format and subsequent
output in ESD and AOD
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
Hypo
Hypo
Hypo
CERN, Nov. 2006, ATLAS Trigger and Physics week
Status of the TrigMoore package
Seeding Algorithms assume the seed is a
a problem for the
MuonFeature (from muFast @ LVL2) or a LVL1
integration in the online
roi (seed from LVL1)
3 istances of TrigMoore called by the steering,
not very flexible:
for reconstruction in the MS, extrapolation to the
for 13.0.0 will separate
IP and combination with ID tracks, allow hypo
into 3 packages
algos to be run at each step (if requested)
TrigMoore records in the TE history a
MooTrackContainer or a MuidTrackContainer for
Trigger Aware analyses can
each seeding ROI and they are accessed by
use the muon EF result
TrigMooreHypo for pT test
by looking at the AOD/ESD
TrigMoore records all reconstructed tracks in the
content but cannot run the
event in a single container in StoreGate for
hypothesis algos on AOD
conversion in Trk:Track format and subsequent
output in ESD and AOD
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Status of the TrigMoore package
First use of the TrigRoIDescriptor
for the online integration and for running more general sequences (seed
can come from whatever LVL2 HLT algorithm)
new tag (TrigMoore-00-00-68) collected in 12.0.X and 12.0.3-LST
Allow running the hypothesis algorithm on AOD will be in 12.0.4
Convert Moo(Muid)Tracks into Trk:Track for each ROI and assign
KEY2KEY to each of them in order to allow navigability
Change input toTrigMooreHypo (to Containers of Trk:Track)
Ongoing work to increase modularity:
we’ll have 3 HLTalgos instead of 3 instances of TrigMoore by 13.0.0
Online integration
problem in the retrieving of L2Result information from ByteStream solved with
the use of the TrigRoIDescriptor
jobOptions for the online
works fine in the offline; tested in the online (athenaPT) OK
Event Filter seeded by LVL1 under test
Also preliminary check for pre-series tests OK
next week the Large Scale Test will start – TrigMoore there !
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
The Sample
Single muons with fixed pT between 2 and 75 GeV/c
CSC production with Athena 11.0.42 (Napoli Tier2 + CERN); Muon Layout Q02
Reconstruction in Athena 12.0.3, TrigMoore seeded by LVL1
skip LVL2 in order to test EF coherently in barrel and endcaps
Check pT rec / pT gen vs h
No relevant bias within
statistical errors
MEAN (of a gaussian fit) shows
reconstruction biases
SIGMA gives relative pT
resolution
Blue = MOORE, for tracks with c2 <3
Red = MuidSA, for tracks successfully fitted
Pink = MuidCB, for tracks successfully fitted
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
Pretty stable with pT
h
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
Relative pT resolution as a function of h
from s(pTrec/pTgen)
Expected trend observed: degraded resolution in the transition
region between the barrel and endcap toroidal fields
s(pT)/pT vs h
pT = 8 GeV/c
pT = 40 GeV/c
h
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
s(pT)/pT vs h
h
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
Relative pT resolution as a function of pT
pT = 8 GeV/c
MOORE
from s(pTrec/pTgen)
stable compared to results in previous releases
pT = 8 GeV/c
MuidSA
pT = 8 GeV/c
MuidCB
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
Moore Efficiency
Efficiencies are defined wrt LVL1
accepted muons
ETA
edge at h = 2.4
PHI
inefficiencies in the feet region
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Performance on recent releases
Efficiency vs pT
Moore, MuidSA OK
Unexpected efficiency loss (5%) of MuidCB compared to MuidSA
similar behaviour in the offline package (efficiency wrt MC-truth)
http://muondoc.web.cern.ch/muondoc/Software/Validation/singlemu/12.0.1/singlemu.html
Under investigation:
- tighter cuts in the machting
account for ~2 % eff. loss
- no obvious trend with h and f
- what else ?
plan to fix by 12.0.4
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
Blue = MOORE, for tracks with c2 <3
Red = MuidSA, for tracks successfully fitted
Pink = MuidCB, for tracks successfully fitted
CERN, Nov. 2006, ATLAS Trigger and Physics week
Trigger Rates
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Trigger Rates
Rates at the EF can be estimated
by folding cross sections with
Efficiency of LVL1 for a given
pT threshold
Efficiency of LVL2 w.r.t. LVL1
seed (skipped, i.e. e  1)
Ri  L
pT _cutoff

pT _ inf
dσ i
 ε (pT ) dpT
dpT
Efficiency of EF w.r.t. seeding
algo (LVL1 in this study)
The effect of skipping the LVL2 selection step:
deserve detailed study
test in the barrel
rates LVL1+EF increase by 10% for mu(6) with respect to LVL1+LVL2+EF
30% for mu(20)
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Trigger rates
First exercise to look at rates at pT thresholds different than
the typical scenarios: 6 and 20 GeV/c
Luminosity set to 1033 cm-2s-1
efficiencies for LVL1
from F. Conventi for
6, 8, 10 GeV/c
(in good agreement with
our estimates)
our estimates for
11, 20, 40 GeV/c
Typical scenarios:
mu(6)
3.0 kHz
mu(20) 25 Hz
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Trigger rates
First exercise to look at rates at the EF in the END CAPS
and vs pT thresholds
Luminosity set to 1033 cm-2s-1
Efficiencies for LVL1
from our estimate
maybe slightly
inaccurate for the
known bugs in
TGCCabling
expected effects at EF <10%
Typical scenarios:
mu(6)
3.1 kHz
mu(20) 27 Hz
Barrel + EndCap EF Rate @ 1033 cm-2s-1
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
mu(6) 6.1 kHz
mu(20) 52 Hz
CERN, Nov. 2006, ATLAS Trigger and Physics week
Trigger Rates
Rate for mu(5) at the EF in the barrel
use LVL1 efficiency for the barrel with the trigger
re-configured to have 5GeV/c as lowest threshold
Luminosity set to 1033 cm-2s-1
Muon
sources
5 GeV/c
6 GeV/c
threshold threshold
kHz
kHz
/K
5.0
1.9
b
1.2
0.67
c
0.65
0.34
W
0.003
0.003
t
negligible
negligible
Total
6.9 kHz
3.0 kHz
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week
Conclusions
TrigMoore provides the requested functionality of the muon EF
from the sw architecture point of view (increasing modularity, compliant
with trigger aware analysis needs)
plan to follow up the successful Moore migration to the new EDM
from the physics performance point of view (but room for refinements
exists !)
First exercise to provide rates vs threshold: input for the
definition of trigger menus
integrate the LVL2 (standalone and combined) in the chain
Plans
revise studies with cavern background and pile-up
revise studies of rejection of muons from /K decays
dedicated CSC simulations requested
pursue timing optimization
Stefania.Spagnolo, Dip. Fisica, Univ. Lecce
CERN, Nov. 2006, ATLAS Trigger and Physics week