Transcript The CDF Online ilicon ertex racker
The CDF Online
S
ilicon
V
ertex
T
racker
S. Donati University and INFN Pisa
9th Topical Seminar on Innovative Particle and Radiation Detectors, May 23-26 2004, Siena, Italy
Outline
• • • • •
Overview of the CDF-II detector and trigger The Online S ilicon V ertex T racker : - Physics motivation - Working principle - Architecture - Track finding and fitting technique SVT performance in the early phase of CDF Run II SVT upgrades Conclusions
The CDF-II Detector and Trigger
detector elements
Time Of Flight
CAL COT MUON SVX CES XFT MUON PRIM.
XTRP XCES L1 CAL L1 TRACK L1 MUON GLOBAL L1 L2 • 2-D COT tracks available @Level 1 ( XFT ) • Fast SVXII readout (10 5 channels) ~10 s • SVT: 2D tracks in silicon (drops stereo info) • SVT: d = 35 mm (at 2 GeV/c) CAL GLOBAL LEVEL 2 • Parallel design (12 slices in phi) reflects SVXII design SVT TSI/CLK
Why do we need the Silicon Vertex Tracker ?
Extract the huge Tevatron beauty/charm production from the 1,000 larger QCD background (
( = 50 mb)
bb = 50
b) at trigger level Primary Vertex Secondary Vertex Decay Length
Lxy
B P T (B)
5 GeV In the pre-SVT age CDF was limited to leptonic modes (B
J/
y
X, B
lDX, suffering from low BR and acceptance) Displaced Track + Lepton (e,
)
Pt(lepton) > 4 GeV (was 8 GeV)
d = impact parameter (~100
m) Two Track Trigger
Pt(trk) > 2 GeV IP(trk) > 100 m IP(trk) > 120 m
Semileptonic modes: high statistics b-hadron lifetimes, b tagging, b mixing Fully hadronic modes: 2-body charmless B decays, B S mixing, Charm .
SVT: Silicon Vertex Tracker
(Chicago-Geneva-Pisa-Roma-Trieste)
• •
SVX II geometry: 12
-slices (30°each) “wedges” 6 modules in z (“semi-barrels”) Reflected in SVT architecture
SVT receives: -COT tracks from Level 1 (
,P t
-Digitized pulse height in
SVX
) strips and performs tracking in a two-stage process: 1. Pattern recogniton: Search “candidate” tracks (
ROADS
) @low resolution.
2. Track fitting: Associate full resolution (d,
,P t
hits to roads and fit 2-D track parameters ) using a linearized algorithm.
The SVT Algorithm (Step I) Fast Pattern Recognition
Single Hit
“XFT layer” Si Layer 4 Si Layer 3
Road
Si Layer 2 Si Layer 1
SuperStrip (bin) Hardware Implemented by AM chip (full custom - INFN Pisa) : - Receives the list of hit coordinates - Compares each hit with all the Candidate Roads in memory in parallel - Selects Roads with at least 1 hit in each SuperStrip - Outputs the list of found roads FAST: pattern rec. complete as soon as the last hit of the event is read - 32.000 roads for each 30 ° slice - ~250 micron SuperStrips - > 95% coverage for Pt >2 GeV
The SVT Algorithm (Step II) Track Fitting
When the track is confined to a road, fitting becomes easier • Linear expansion of parameters in hit positions X i
P i = F i
X i + Q i ( P i = p t ,
, d
, c 1 , c 2 , c 3
)
• then refer them to the ROAD boundary
P 0i +
d
P i = F i
( X 0i +
d
X i ) + Q i XFT layer SVX layer 4 P 0i = F i
X 0i + Q i SVX layer 3
• •
F i
and P
0 i
coefficients are calculated in advance (using detector geometry) and stored in a RAM
SVX layer 2 SVX layer 1
the task is to compute simple scalar products
TRACK Road boundary
d
X 5
d
X 4
d
X 3
d
X 2
d
X 1 P
0i
X 05 X 04 X 03 X 02 X 01
d
P i = F i
d
X i SuperStrip 250
m
COT tracks from Level 1
The SVT Boards
AM Sequencer Super Strip AM Board SVXII Data Hit Finder
L2 CPU
Roads Matching Patterns Roads + Corresponding Hits Hit Buffer Tracks + Corresponding Hits Track Fitter
SVT Performance (I)
d -
correlation
28 Aug 2001 data, c 2 <40 no Pt cut Sinusoidal shape is the effect of beam displacement from origin of nominal coordinates
SVX only d = X 0 ·sin (
) - Y 0 ·cos (
)
track (X 0 ,Y 0 ) d X 0 Y 0 = 0.0153 cm = 0.3872 cm Can find the beam consistently in all wedges even using only SVX
Online fit of X-Y Beam position
Run 128449 - October 6 2001 Can subtract beam offset online : I.P. with respect to beam position (online) : Independent fit on each SVXII z -barrel (6) ~ 45 m
Highlights of B physics (hadronic channels) B 0 s
D s
p
+ B
h + h' B 0 s
D s
p
+ resolve B 0 s golden channel to fast oscillations B
h + h' crucial to understand CP violation in the B sector (CDF-II competitive and complementary to B factories)
Highlights of B physics (semileptonic channels) B +
g
l + D 0 X 1400 Bs
l D s
l[
p
] High statistics semileptonic B samples are excellent for calibration, B + /B 0 and Bs/B 0 (for moderate lifetime measurements,tagging and B 0 x s ) and Bs mixing
4/4 – 4/5 Why is the SVT upgrade important ?
4/4 – 4/5
27 sec 1. looser matching criteria 2. Ghost roads 3. 5 layers larger Patterns 4 LVL2 buffers • fluctuations dead time depends on: • total LVL2 latency
Time (
s)
More memory for thinner patterns 5/5 Empty SS 4/5 This road share all hits with the 5/5. It’s a ghost.
First step: new Associative Memory System
Use standard cell chips to perform AM chip function Increase from 128 pattern/chip 4k pattern/chip (thinner roads, less fits, faster system, can cope with increasing Tevatron luminosity, increase coverage to forward region, lower pt threshold)
Conclusions
• • • SVT performs fast and accurate 2-D track reconstruction (is part of L2 trigger of CDF-II). Tracking is performed in two stages: - pattern recognition - high precision track fitting Taking good data since the beginning of CDF Run II, many analyses in the field of beauty/charm physics only possible thanks to this device Upgrade of the system in progress to cope with increasing Tevatron luminosity
detector elements CAL COT MUON
CDF Trigger in run II
SVX CES New for CDF run II at Level 1: • 2-D COT • Fast SVXII tracks available ( XFT ) readout (10 5 channels) : ~ 2.5 s XCES XFT MUON PRIM.
XTRP L1 CAL L1 TRACK L1 MUON GLOBAL L1 SVT d , Pt ,
Track reconstruction with “offline” resolution at Level 2
d = 35 m (at 2 GeV/c) Fast !
~10 s (50 KHz L1 accept rate) SVT L2 CAL GLOBAL LEVEL 2 TSI/CLK •
2-D tracks
(drop SVXII stereo info) •
Pt > 2 GeV/c
• parallelized design :
12
-slices (30°each)
which reflects SVX II geometry
25 20 15
Accurate deadtime model (ModSim)
4/5 simulation Ini_lum=10*10 30
M. Schmidt
RUN 168640 7% 10 4/5 sept. 2003 Ini_lum=44*10 30 4/4L2UP-L2A 300Hz 4/4L2UP-L2A 300Hz 4/5-L2A 140Hz 4/5L2UP-L2A 300Hz RUN 164308 5 0 0 5 10 15 20 25
L1A rate (Hz)
30 4/4 4/5+SVTupgrade+L2upgr Simulation: Ini_lum ~ 20*10 30 35 40 45 Low lum, Run IIa L2, no COT prob.
To be done again.
Understanding the width of d distribution
d 2 = B 2 + res 2 beam size i.p. resolution • Present (online) • Correct relative wedge misalignment • Correct for d and non-linearity • Correct internal (detector layers) alignment • Correct offline for beam z misalignment •
GOAL
: SVXII + COT offline tracks (1wedge) d ( m )
69 63 57 55 48 45 Need to have beam aligned
Effect of non-linearity :
Because of the linear approximation done by Track • Fitters, SVT measures : d SVT SVT d /cos ( tan ( ) ) Becomes important for large beam offset Can correct it making the online beam position routine fit a straight line on each wedge
Expected SVT performance
From SVT TDR (’96) : SVT simulation using RunI data Commissioning Run - Nov. 2000 : SVT simulation with SVX hits + COT Offline ~ 45 m ~ 45 m
: SVT – COT
SVT Performance (II)
correlation with offline tracks
Curvature : SVT - COT = 2 mrad = 0.33 ·10 -4 cm -1 d : SVT - COT = 21 m
Intrinsic transverse beam width
Can extract B from the correlation between impact parameters of track pairs • If the beam spot is circular :
Cut Online on chi2, P t , d, N.tracks
• L2 : • N. SVT tracks > 2 • | d | > 50 m • c 2 < 25 • Pt > 2 GeV/c L1 prerequisite : 2 XFT tracks Trigger selection successfully implemented !
About 200 nb -1 of data collected with this trigger where we can start looking for B’s !