lhcb vertex locator (velo) lars eklund
DESCRIPTION
École Polytechnique Fédérale de Lausanne. Liverpool University. Vrije Universiteit Amsterdam. LHCb Vertex Locator (VELO) Lars Eklund. Outline. Introduction to LHCb and VELO Design of the VELO Performance plots Towards a beam test in 2006 A more specific example: - PowerPoint PPT PresentationTRANSCRIPT
LHCb Vertex Locator (VELO)
Lars Eklund
École Polytechnique Fédérale de Lausanne
Liverpool University
Vrije Universiteit Amsterdam
Vertex 05, November 7, 2005Lars Eklund, CERN 2
Outline
• Introduction to LHCb and VELO
• Design of the VELO
• Performance plots
• Towards a beam test in 2006
• A more specific example:
– Digital filtering for x-talk corrections
Vertex 05, November 7, 2005Lars Eklund, CERN 3
LHCb
Tracker
ECAL HCAL Muon chambers
Magnet
Angular acceptance
15 - 300mrad
Proton interaction region
RICH 2
RICH 1
Trigger tracker
Dedicated b-physics experiment at LHC
• 1012 bb-pairs per year
• CP violation and rare b-decays
• Correlated boost of the bb-pairs
• Single forward spectrometer
The Vertex Locator
Vertex 05, November 7, 2005Lars Eklund, CERN 4
VELO – the experimental challenge
1. Trigger on the B decay of interest (Velo part of software trigger)
2. Suppress multiple interactions (Pile-up veto part of hardware trigger)
3. Reconstruct decay as a function of time
bt
Bs K
K
,K
Ds
Primary vertex
p p
Since the oscillations are fast, it requires excellent vertex resolution
Measurement of BS oscillations
Vertex 05, November 7, 2005Lars Eklund, CERN 5
The Vertex Locator (VELO)
~1m
Interaction region Downstr
eam21 Tracking stations
• Four ½ disks sensors per station
• Silicon micro-strip technology
• R-Φ geometry
Minimalist view of the VELO
Minimise material
• No conventional beam-pipe
• Sensors are operated in vacuum
• 250 μm Al foil to separate VELO from the beam vacuum
Minimise extrapolation distance
• First active element at R = 8.2 mm
• Retractable detector halves
• 30 mm at LHC filling
Pile-up veto
2 backward stations in the H/W trigger
Vertex 05, November 7, 2005Lars Eklund, CERN 6
Velo Sensors
–measuring sensor
(radial strips with an stereo angle)
Non uniform radiation environment
•1.3 * 1014 neq/cm2/year at R = 8 mm
• 5 * 1012 neq/cm2/year at R = 42 mm
• n+ in n-bulk sensors
• second metal layer for signal routing
• 2048 micro strips per sensor
• 40 – 100 μm pitch
R-measuring sensor (concentric strips)
42 mm
8 mm
FE
chip
Vertex 05, November 7, 2005Lars Eklund, CERN 7
Detector module
Cooling cookies
Pitch adaptors
Silicon sensor
TPG/carbon fibre substrate with laminated kapton circuit
Carbon fibre paddle
Front-end chip
(Beetle1.5)
PRR in December
Vertex 05, November 7, 2005Lars Eklund, CERN 8
Detector modules - performance
Operational window:
• Sufficient cluster efficiency (> 99 %)
• Acceptable noise occupancy (< 0.1 %)
• Low fake cluster rate in next time bin (< 25 %)
Cluster efficiency vs. S/N cut
Results test beam November 2004 – 300 μm thick Φ measuring sensor
Noise occupancy vs. S/N cut
Overspill vs. S/N cut
Vertex 05, November 7, 2005Lars Eklund, CERN 9
Cooling, vacuum and positioning
Beam pipe
Detector halves retractable by 30 mm
CO2 cooling manifold
Primary (beam) vacuum
Secondary (detector) vacuum
RF foil (250 μm Al)
Vacuum bellows
(to allow the retraction)
Vertex 05, November 7, 2005Lars Eklund, CERN 10
Mechanics – status
Vacuum and positioning assembly at NIKHEF
Vacuum bellow
Vertex 05, November 7, 2005Lars Eklund, CERN 11
DAQ chain
front-end ASIC
2 m low mass cable
60 m twisted
pair
ADC and pre-processing
FPGA
Radiation dose: ~ MRad
pre-compensating
cable driver
Radiation dose: ~100 kRad Radiation free area
readout network
PC farm
Radiation free area
Beetle 1.5
• analogue readout
• 128 channels
• 4 serial links
• 900 ns readout time
Kapton cables
• low mass
• flexible
• vacuum compatible
Analogue repeaters
• compensates cable response
• COTS components
• 5632 links
Cat 6 cable Gigabit Ethernet
• commercial components
TELL1
• ADC (10 bit, 40 MHz)
• digital filter
• pedestal and common mode noise subtraction
• strip re-ordering
• clustering
PC farm
• Linux cluster
• software trigger
• permanent storage
Vertex 05, November 7, 2005Lars Eklund, CERN 12
Performance
IP = 14m+35m/pT
Impact parameter resolution
Bs vtx resolution (mm)
Vertex 05, November 7, 2005Lars Eklund, CERN 13
Towards the final system…
• Components are or will soon be in production• Module production will start in December 2005• Assembly will start January 2006
– At CERN
• System test April 2006– Whole detector half powered and configured– 10 modules read out
• Beam test 2006– Whole detector half in the beam test– Fully system verification: Detector modules, control system,
DAQ, reconstruction software, alignment …
Vertex 05, November 7, 2005Lars Eklund, CERN 14
Two systems issues
1.Interaction vertices in the silicon sensors and halo tracks
2. Digital filtering of the analogue data link
Vertex 05, November 7, 2005Lars Eklund, CERN 15
Test beam set-up
scintilator trigger
proton beam
interaction in the sensor
• Detector half operated in vacuum
• Use silicon sensors as targets
• Look for interactions in the silicon
• Reconstruct tracks and align
Partially commissioned at installation
Vertex 05, November 7, 2005Lars Eklund, CERN 16
Need stand-alone tracking
Tracks from interaction point (R=0) are linear in the R-z plane
Pattern recognition assume tracks from close to R = 0
Need for a stand-alone reconstruction package
• Handles vertices anywhere
• Feeds tracks to the alignment algorithm
z [mm]
R [
mm
]Event display of an interaction in the silicon sensor.
Tracks highly non-linear in the R-z plane
Vertex 05, November 7, 2005Lars Eklund, CERN 17
Spin-off: halo tracks in LHCb
Velo Left
Velo Right
cartoon event display of the VELO
Solution: Use beam halo tracks and interactions in the silicon sensors for alignment in LHCb
• Re-use algorithms from the beam test 2006
Velo divided into four weakly coupled parts in terms of alignment:
• Backward-Left, Backward-Right, Forward-Left and Forward-Right
• very few tracks traverse more than one part
L-B
R-B
L-F
R-F
Vertex 05, November 7, 2005Lars Eklund, CERN 18
Two systems issues
1. Halo tracks and interaction vertices in the silicon sensors
2.Digital filtering of the analogue data link
Vertex 05, November 7, 2005Lars Eklund, CERN 19
Digital filtering – the problem
32 channels serial data (800 ns)
signal noiseData is transferred on four serial links per front-end ASIC:
header
25 ns
Signal travels through:
• front-end hybrid
• kapton cables (2 types)
• vacuum feed-through
• Amplifier (+ PCB)
• 60 m TP cable
Blue: no signal (pedestals)
Red: signal in two channels
Vertex 05, November 7, 2005Lars Eklund, CERN 20
Modelling the problem
Consider each output as a number series, where n corresponds to the channel number
x[n] : “True” data at the input (from sensor or test pulse)
w[n] : Raw ADC values
y[n] : Corrected values
H : Transfer function of the ASIC and the analogue link
G : Transfer function of the digital filter
Sensor, ASIC and the
analogue linkDigital filter
x[n] w[n] y[n]
H(z) G(z)
Method: (Assuming LTI: Linear Time Invariant System)
1. Determine H from the data (from beam particles or calibration pulses)
2. Find ‘G’ such that G*H = 1, implying that y[n] = x[n]
Vertex 05, November 7, 2005Lars Eklund, CERN 21
The impulse response
][][][ nkxkhnw
The transfer function H is characterised by the series h[n]:
h[n] can be determined by injecting a single pulse:
00
01][][
nif
nifnnx
Obtaining a Delta function δ:
1. Internal calibration pulses
• Feature of the Beetle chip
2. From track data
• Point a track on the sensor
• Select tracks centered on strips
• Assume no charge sharing
3. look at the signal in adjacent channels
4 ASICs, 16 serial links: 32 different impulse responses
h[+1]
h[-1]
x-talk measurements from TB November 2004
Vertex 05, November 7, 2005Lars Eklund, CERN 22
Determining the filter algorithm (1)
Sensor, ASIC and the
analogue linkDigital filter
x[n] w[n] y[n]
H(z) G(z)
Postulate:
• The impulse response h[n] = 0, except for N time-bin
• g[n] = 0, except for M time-bins (Finite Impulse Response filter = FIR).
][][][ nkwkgny
][][][ nkxkhnw
requires that
][][][ nkhkgnand
][][ nxny
][][ nnx
2
1
2
1
1 ][][][
M
M
NM nkhkgn
)(0
0
0)(
0
1
0
][1
MNnif
nif
nMNif
nMNwhere
The infinite sum is truncated to M + N – 1 conditions:
Vertex 05, November 7, 2005Lars Eklund, CERN 23
Determining the filter algorithm (2)
The truncated sum gives:
• N + M – 1 constraints
• M unknowns (the g[n])
Solve these over constrained equations with the least square method
corrected x-talk from TB November 2004
4 ASICs, 16 serial links: 32 different impulse responses
Boundary problems:
• 32 channels per link
• exceptions for channels 0 and 31
• missing information
• The data header
• specific corrections
Vertex 05, November 7, 2005Lars Eklund, CERN 24
Effects on the resolution
Track residuals:
• reversed readout order for small radii
• Asymmetric (forward/backward) x-talk
• step in residuals
Resolution:
• Smearing due to x-talk
• Odd/even channel dependence
• ~1 μm improvement
Results test beam November 2004 – 200 μm thick R measuring sensor (7 degree track angle)
Vertex 05, November 7, 2005Lars Eklund, CERN 25
Summary
• The Vertex Locator of LHCb uses silicon micro-strip sensors with R-Φ geometry
• Operated in vacuum with retractable detector halves• Shows good efficiency and noise occupancy
performance in the beam test• Is in production – assembly will start soon• System issues
– System test and beam test 2006
• Specific example– Corrections of the data link