ilc tracking r&d at scipp sid workshop thursday, june 3 2010 bruce schumm santa cruz institute...
TRANSCRIPT
ILC Tracking R&D at SCIPP
SiD Workshop
Thursday, June 3 2010
Bruce Schumm
Santa Cruz Institute for Particle Physics
Faculty/Senior
Vitaliy FadeyevBruce Schumm
Collaborators
Rich PartridgeTakashi
MaruyamaTom Markiewicz
(SLAC)
Students
Alex Bogert**
Jerome Carman*
London Chappel…Kelsey CollierSpencer KeyDonish KhanOmar Moreno
Jared NewmillerDale Owens
Sheena SchierMatt StantonDustin Stolp*
Aaron TaylorCapella Yee
The SCIPP/UCSC ILC R&D GROUP
Lead Engineer: Ned SpencerTechnical Staff: Max Wilder
Forest Martinez-McKinney
*2010 Senior thesis (graduation requirement) !Carman thesiswins Chancellor’s Award!
** CERN will support summer 2010
OUTLINE OF SCIPP ILC R&D PROGRAM
LSTFE Front-End ASIC
• Optimized for ILC (long ladders for barrel; high rates for forward tracking)
• Applicable to both SiD and ILD
SiD KPIX/Double-Metal Development
• SiD baseline; alternative to LSTFE
• SCIPP has done group’s sensor testing, now beginning critical systems tests
Fundamental (“Generic”) R&D
• Use of charge division for longitudinal coordinate (with Rich Partridge)
• Solid-state noise sources
SiD Simulation Projects (Not in talk)
• Tracking and momentum reconstruction validation
• Non-prompt signatures
OUTLINE (Continued)
The LSTFE ASIC
1-3 s shaping time (LSTFE-I is ~1.2 s); analog measurement is Time-Over-Threshold
Process: TSMC 0.25 m CMOS
The LSTFE ASIC
FIF
O (L
eadin
g and
trailin
g transition
s)Low Comparator Leading-Edge-Enable Domain
Li
Hi
Hi+4
Hi+1
Hi+2
Hi+3
Hi+5
Hi+6
Li+1
Li+2
Li+3
Li+4
Li+5
Li+6
Proposed LSTFE Back-End Architecture
Clock Period = 400 nsec
EventTime
8:1 Multi-
plexing (clock = 50 ns)
LSTFE-II PrototypeOptimized for 80cm ladder (ILD barrel application)
Institute “analog memory cells” to improve power-cycling switch-on from 30 msec to 1 msec (problems!)
Improved environmental isolation
Additional amplification stage to improve S/N, control of shaping time, and channel-to-channel matching
Improved control of return-to-baseline for < 4 mip signals (time-over-threshold resolution)
128 Channels (256 comparators) read out at 3 MHz, multiplexed onto 8 LVDS outputs
Testing underway in SCIPP lab
LSTFE-II Performance: Time-Over-Threshold vs. Injected Charge
Omar Moreno
Use of KPiX as a Tracking Chip
The SiD KPIX/Double-Metal Baseline Design
10cm2 modules tessellate the five barrel tracking layers
Traces on 2nd (surface) metal layer to two 1024-node bump-bonding arrays
SCIPP attempting to explore whetherMounting KPiX on surface of sensor leads to digital analog feed-through
It has proved difficult to mount KPiX7 onto prototype sensors (passivation between two metal layers fails during bonding)
Forest M.M.: Arranged for bonding with private firm (H&K); then cut shorted traces with laser
Connectivity After Severing Traces
Pad to Pad old R(Ohms) New R(Ohms)---------------------------------------------AVdd to CLK(-) 74.8 3.4MAVdd to CLK(+) 174.2 3.2MAVdd to DVdd 29.6 40KVref to AGND 2.7 101.1Vref to DGND 14.6 102AGND to DGND 16 12.6**DVDD to CLK(-) 75.1 1.4MDVDD to CLK(+) 173.3 1.6M
**This ohmic connection seems to be through the IC and is not fromdamage during bonding.
Next step: build enclosure for assembly and begin to read out (summer)
Triggering on Minimum-Ionizing Particles
Threshold corresponding to the average pulse height for the given injected charge.
10fC injected 9fC injected 8fC injected
4fC injected Mean Threshold:
Example channel 63
Charge in Coulombs
offset
gain
Example Channel 63
0-Charge Input Offset (mV) by Channel(x
10)
Offset in mV for no
input charge
Gain ~4 mV/fC
Next step: Interface with pulse-development simulation to confirm min-I
operability
Charge Division for Silicon Strip Sensors
Final step: practical detectors are not isolated strips. Include two nearest-neighbors in simulation:
Network effects lead to ~5% reduction in longitudinal resolution.
Readout Noise for Linear Collider Applications
Use of silicon strip sensors at the ILC tend towards different limits than for hadron collider or astrophysical applications: Long shaping time Resistive strips (narrow and/or long)
But must also achieve lowest possible noise to meet ILC resolution goals.
• How well do we understand Si strip readout noise, particularly for resistive networks?
• How can we minimize noise for resistive networks?
Standard Form for Readout Noise (Spieler)
Series Resistance
Amplifier Noise (series)Amplifier Noise (parallel)
Parallel Resistance
Fi and Fv are signal shape parameters that can be determined from average scope traces.
Expected Noise for Custom-Biased L00 Ladder
Spieler formula suggests that series noise should dominate for ladders of greater than 5 or so sensors.
CDF L00 Sensor “Snake”
CDF L00 “Snake”
LSTFE1 chip on Readout Board
Readout Noise ResultsRelative to prior results, have explored “center-tapping” (reading out from center of chain rather than end.
Naïve expectation
Expectation with measuredshape factors Fi, Fv
Measured (end readout)
Measured (center-tap)
Summary of Findings
Reading out from ladder from end:
Significantly less noise observed than expected (network effects ignored in formulation of expectation?)
Reading out from middle (“center tap”):
Noise seems further reduced (~20%) for lengths for which series noise dominates
Will explore with P-SPICE simulation…
AUGUST 2010 MILESTONES Evaluation of channel-to-channel variations in KPIX-9
Exploration of attachment of KPIX-7 to prototype sensor using cutting laser to mitigate shorts from over-glass failure
Initial characterization of LSTFE-II chip
Estimation of longitudinal resolution from resistive charge division measurement of readout noise in the series-resistance limit, including both end- and center-readout
Simulation studies supporting the design of a high-fluence electromagnetic radiation damage study of various silicon sensors
Acquisition and characterization of sensor samples to be used in damage studies
Simulation studies of SiD tracking efficiency performance in forward region
AUGUST 2011 MILESTONES
Characterization of 'analog memory' test structures and assessment of suitability for implementation of power-cycling for LSTFE
Assessment of effects of digital/analog interference effect in KPIX/double-metal assembly
PSpice simulation of series-noise readout limit and comparison with observation
Publication of charge-division results
Test-beam study of radiation damage to various silicon sensor types
Simulation studies of SiD momentum reconstruction performance in forward region
AUGUST 2012 MILESTONES
Re-optimization (for forward region) and fabrication of LSTFE ASIC.
Follow-up testbeam run to confirm radiation hardness of selected Si sensor
Determination of radial extent of sensitivity for reconstruction kinks fro stau decay in the SiD detector
KPiX/double-metal yield and reliability studies
Simulation Studies (ILC Detector Performance)
CURVATURE RECONSTRUCTION PERFORMANCE
1. Compare width of Gaussian fit to residuals with two different estimates:
• Error from square root of appropriate diagonal error matrix element
• Error from Billoir calculation (LCDTRK program)
2. Only tracks with all DOF (5 VTX and 5 CT layers) are considered.
3. Require |cos| < 0.5
Mixture of q/qbar at 500 and 1000 GeV, tau samples at 500 GeV; also use single muons
CURVATURE ERROR vs. CURVATURE
LCDTRK calculation (no beam constraint)
Gaussian fit
SiD02 DetectorModel
Results for Stiff, Central Tracks
In terms of p/p, comparing , with p=100 GeV and |cos| < 0.5 we find
LCDTRK 0.28% 0.28%
Residuals 0.37% 0.39%
LOI Result0.33% ----
Delhi group (Kirti Ranjan et al.) looking into developing Kalman Filter fitter
SiD Forward Tracking EfficiencyStart in central region (|cos| < 0.7); establish Pt cut of 2.0 GeV/c
Capella Yee, Scotts Valley High School
Pt > 2.0 0.7 < Pt < 2.0
Exploring Higher Values of cos
Note: Horizontal axis is -|cos| (learning how to move stat box!)
Capella Yee
3-Hit Tracks & Non-Prompt Signatures
Probably need 5+1 layers for prompt track
If we require 4 hits for non-prompt tracks, sensitive region for kinked tracks is very limited.
e.g.: Position-matching for isolated muons
(mm)
SCIPP algorithm for non-prompt tracks: match 3-hit seeds from tracker with stubs from calorimeter (developed by undergrads Meyer ( U. Chicago), Rice ( UC Irvine)
• Approximately 60% for 3-hit tracks in Z qq events• Being optimized for non-prompt recon-struction (GMSB signature) by undergraduates Stolp and Bogert.
0~
Non-Prompt Tracks and GMSB