trigger peter jacobs, lbnl
DESCRIPTION
10 Hz. Trigger Peter Jacobs, LBNL. Pb+Pb interaction rate ~ 4-8 kHz Hard process trigger ~ 10 Hz need rejection factor 400-800. p 0 : 10 Hz p T ~20 GeV/c Inclusive jets: 10 Hz E T ~50 GeV/c. EMCal trigger: enhancement factor relative to minbias trigger + TPC. s=0: p 0 vs p ±. - PowerPoint PPT PresentationTRANSCRIPT
An EMC for ALICE 1ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
TriggerPeter Jacobs, LBNL
0: 10 Hz pT~20 GeV/c
Inclusive jets:10 Hz ET~50 GeV/c
10 Hz
Pb+Pb interaction rate ~ 4-8 kHzHard process trigger ~ 10 Hzneed rejection factor 400-800
An EMC for ALICE 2ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
EMCal trigger: enhancement factor relative tominbias trigger + TPC
Live time: • DAQ dynamic scaledown of common triggers (“central”, “minbias”)• Pierre v.d.Vyvre: reasonable to expect 90% livetime
sacc
saccefffLE
TPC
EMCtrigmbmbAA
Trig
tapetorateevtmax
s=0: 0 vs p±
An EMC for ALICE 3ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Why bother with jets in p+p?
Is ALICE irrelevant for jet studies in p+p? Maybe not.
• p+p at top luminosity: 1034/cm2/s x 100 mb ~ 109 Hz
20 ns bunch spacing 20 minbias interactions per bunch crossing
• ALICE runs at L~1031 much cleaner environment in low to intermediate pT region
ALICE may have a unique niche in p+p: detailed studies of jet fragmentation down to low z important reference for A+A program
An EMC for ALICE 4ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
PHOS L1 Trigger
L1 (6 s): 2x2 tower analog sum TRU 4x4 tower peak finderEMCal: 12 FEE/GTL bus 12*32=384 towers/TRU
(x ~0.24 x 0.36)
An EMC for ALICE 5ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Charged jets (HLT study)
ETcharged>m
Charged jets: poor energy resolution, slow turn-on above trigger threshold, highly biased
C. Loizides, FfM
An EMC for ALICE 6ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Conjecture: better solution is Level 1 EMCal + HLT TPC+EMCal
Some rough numbers:
Minbias data rate ~ 20 MB/evt*4 KHz ~ 80 GB/sHLT input bandwidth ~ 15 GB/s
Least-biased efficient trigger algorithm:
• EMCAL@L1: mildly biased jet patch trigger to cut minbias rate by factor 10 ~ 8 GB/s
• do the rest in HLT incorporating charged tracks, neutral energy from emcal, dijet topologies (?), etc
An EMC for ALICE 7ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Jet patch trigger in p+p
0.21x0.21
Strong biases for ETmax>10 GeV
PYTHIA
An EMC for ALICE 8ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Jet Patch Trigger SimulationsAndre Mischke (Utrecht)
• Pythia jet (ET~50 GeV) + HIJING background• square candidate jet patches x s x s • sweep patch quasi-continuously over detector, find maximum ET
max
An EMC for ALICE 9ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Why a large-ish trigger patch in A+A?Leading particle (0) trigger strongly biased
• Fragmentation bias (prefer low ET jets fragmenting hard)• Geometric bias (prefer lower than average energy loss surface emission)
PYTHIA fragmentation requires relatively small patches (0.1 x 0.1) for efficient triggering (Bill, Chris)
fragmentation fluctuations primarily in distribution of few hardest (colinear) hadrons
But the physics we are after is the modification of the fragmentation, including potentially large jet broadening effects (and correspondingly smaller background fluctuations?)
no really solid theory guidance prudent experimental design requires flexible patch trigger also for lighter systems than Pb+Pb
An EMC for ALICE 10ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Jet patch trigger in Pb+Pb
bkgd
jet (ET~50 GeV)
• Centrality dependent pedestal (no surprise)• Background and fragmentation fluctuations of similar magnitude
Weighted by Nch data volume
PYTHIA+HIJING
0.21x0.21
ET cut for 80% trigger efficiency @ 50 GeV
An EMC for ALICE 11ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
L1 output data rate for 80% jet efficiency @ 50 GeV
• Details strongly dependent on models of signal + background • Qualitative conclusion nevertheless: required L1 rejection achievable with reasonable efficiency for ~50 GeV jets
requires centrality-dependent threshold
input rate
Output rate
An EMC for ALICE 12ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Trigger efficiency I
PYTHIA+HIJING
Good efficiency at 100 GeVBackground dependencies at 50 GeV
An EMC for ALICE 13ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Trigger efficiency II: patch size dependence
0.3 x 0.3 apparently larger than optimal
An EMC for ALICE 14ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Trigger efficiency III: quenching models
Strong model dependencies
But models are only first guesses prudent experimental design requires flexible patch trigger
An EMC for ALICE 15ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
PHOS trigger: Trigger Regional Unit
• each TRU: x~0.24x0.36 (depends on how we cable) • no inter-TRU communication large acceptance hit for jet patch, limited maximum patch size
An EMC for ALICE 16ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Candidate implementation: TRU hierarchy
TRU~400 towers
TRU~400 towers
TRU~400 towers
TRU~400 towers
TRU~400 towers
TRU~400 towers
……………
13K towers ~ 30 TRUs
SummaryTRU
Low-latency serial pass-through
Low-latency serial pass-through Low-latency serial
pass-through
Max aggregate input bandwidth= 3 Gbit/s
Jet patch trigger
late
ncy
1-2
s
< 10
0 M
bit/s ~100 bits/evt
An EMC for ALICE 17ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
Centrality-dependent trigger threshold? V0
Uniform jet trigger efficiency across centralities: need to account for centrality-correlated pedestal fluctuations
V0 is only fast (L1) detector with sufficient coverage
An EMC for ALICE 18ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
V0 response to Pb+Pb
Large generation of secondaries in beampipe but response is nicely linear
Forward detector TRD fig 3.6
An EMC for ALICE 19ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
V0 bits in jet patch summary TRU?
TRU~500 towers
TRU~500 towers
TRU~500 towers
TRU~500 towers
TRU~500 towers
TRU~500 towers
……………
SummaryTRU
Low-latency serial pass-through
Low-latency serial pass-through Low-latency serial
pass-throughlate
ncy
1-2
s
< 10
0 M
bit/s
V0
• System design issues: non-locality of trigger logic (i.e. not in CTP), scalars,…• PHOS TRU: Hans Muller has added 3 optical Gigabit ports for interconnectivity…
few bitsSignificant problem: no L1 cross-correlation in Central Trigger Processor (Orlando V-B)
technically feasible (H. Muller, F. Fomenti, Y. Zoccarato)
Centrality-dependent threshold?
An EMC for ALICE 20ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
To CTP
L0
in
65LVDM31
Q+
Q-
from FPGACYCLONE EP1C12
(ALTERA) Trigger 1 - 5
To LTUin
65LVDM31
Q+
Q-
from FPGACYCLONE EP1C12
(ALTERA) busy
in+
in-Q
65LVDS32B
from LTUto FPGA
CYCLONE EP1C12(ALTERA)
16
HDMP-1032
clk
data
from FPGACYCLONE EP1C12
(ALTERA)
Q+
Q-
+3.3V num
130
+3.3V num
To EMCal100nF
100nF
130
V0 Interface between CTP, LTU, EMCal and FEE
An EMC for ALICE 21ALICE Collaboration MeetingLBNL, Oct 15-16, 2005
General trigger issues
Interface with ALICE CTP for issuing and receiving L0/L1Decisions being made now, we need to be involved …