View
217
Download
1
Tags:
Embed Size (px)
Citation preview
SCIPP R&D on Linear Collider Tracking – Hardware and
Simulation
DOE Site Visit
June 14 2007
Bruce Schumm
Santa Cruz Institute for Particle Physics
Support
Recently, ILC R&D has been supported at fluctuating levels by the DOE’s Linear Collider R&D (LCRD) program
2006-2007: We received $53,000 in regular funding, and have been told we will receive another $37,500 in supplemental funding
2007-2008: We have been told to expect $51,000. We might also expect some additional supplemental funding.
Virtually all funding has been used to support the hardware project.
Faculty/Senior
Vitaliy FadeyevAlex Grillo
Bruce Schumm
Post-Docs
Jurgen KrosebergLei Wang
Undergrads
Greg HornLuke Kelley
Ian HornSean Crosby
The SCIPP/UCSC SiLC/SiD GROUP (Harwdare R&D Participants)
Lead Engineer: Ned Spencer
Technical Staff: Max Wilder, Forest Martinez-McKinney
All participants are mostly working on other things (BaBar, ATLAS, biophysics, Undergraduate major…)
Students are undergraduate physics majors at UCSC
FOCUS AND MILESTONES
Goal: To develop readout generically suited to any ILC application (long or short strips, central or forward layers)
Current work focused on long ladders (more challenging!):
• Front-end electronics for long (~1 meter) ladders• Exploration of sensor requirements for long ladders• Demonstration (test-beam) of < 10 m resolution mid-2008After long-ladder proof-of-principle, will re-optimize (modest changes) for short-ladder, fast-rate application
We also hope to play an increasing role in overall system development (grounding/shielding, data transmission, module design and testing) as we have on ATLAS and GLAST
BRIEF SUMMARY OF STATUS
Testing of 8-channel (LSTFE-1) prototype fairly advanced:
•Reproducible operation (4 operating boards)•Most features working, with needed refinements understood•A number of “subtleties” (e.g. channel matching, environmental sensitivity) under control•Starting to make progress on fundamental issues confronting long-ladder/high-resolution limit.
Design of 128-channel prototype (LSTFE-2) well underway (July submission)
Now for the details…
Pulse Development Simulation
Long Shaping-Time Limit: strip sees signal if and only if hole is collected onto strip (no electrostatic coupling to neighboring strips)
Include: Landau deposition (SSSimSide; Gerry Lynch LBNL), variable geometry, Lorentz angle, carrier diffusion, electronic noise and digitization effects
Christian Flacco & Michael Young (Grads); John Mikelichand Luke Kelley (Undergrads)
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
1/4 mip
1 mip
128 mip
Operating point threshold
Readout threshold
High gain advantageousfor overall performance(channel matching)
Simulated Resolution for 167 cm Ladder
Detector Noise:
Capacitive contribution; from SPICE simulation normalized to bench tests with GLAST electronicsAnalog Measurement:
Provided by time-over-threshold; lookup table provides conversions back into analog pulse height (as for actual data)
RMS
Gaussian Fit
Detector Resolution (units of 10m)
Lower (read) threshold in fraction of min-i (High threshold is at 0.29 times min-i)
DIGITAL ARCHITECTURE: FPGADEVELOPMENT
Digital logic under development on FPGA (Wang, Kroseberg), will be included on front-end ASIC after performance verified on test bench and in test beam.
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)
Note on LSTFE Digital Architecture
Use of time-over-threshold (vs. analog-to-digital conversion) permits real-time storage of pulse-height information.
No concern about buffering
LSTFE system can operate in arbitrarily high-rate environment; is ideal for (short ladder) forward tracking systems as well as long-ladder central tracking applications.
Comparator S Curves
Vary threshold for given input charge
Read out system with FPG-based DAQ
Get
1-erf(threshold)
with 50% point giving response, and width giving noise
Stable operation toVthresh ~ 5% of min-I
Qin= 0.5 fC
Qin= 3.0 fCQin= 2.5 fC
Qin= 2.0 fCQin= 1.5 fC
Qin= 1.0 fC
Hi/Lo comparators function independently
Noise vs. Capacitance (at shape = 1.2 s)
Measured dependence is roughly(noise in equivalent electrons)
noise = 375 + 8.9*C
with C in pF.
Experience at 0.5 m had suggested that model noise parameters needed to be boosted by 20% or so; these results suggest 0.25 m model parameters are accurate
Noise performance somewhat better than anticipated.
Observed
Expected
1 meter
EQUIVALENT CAPACITANCE STUDY
Preamp Response
Power Control
Shaper Response
Power Cycling (with Small Injected Current for now)
It is essential to turn off chip for the 99% of the time that the beam is not there Must be able to turn chip back on in 1 millisecond!
Solution in hand to maintain bias levels in “off” state with low-power feedback; will eliminate need for external trickle current
Measured Noise vs. Sum of Estimated Contributions
72 cm Ladder
Estimated Johnson noise for actual 65 m strip (part of
estimate)
Projected Johnson noise for 20 m strip (not part of
estimate)
Measured noise
Sum of estimates
143 cm Ladder
Noise calculation assuming 20m strip width (actual is 60 m)
Strip Noise Idea: “Center Tapping” – half the capacitance, half the resistance?
Result: no significant change in measured noiseHowever, sensors have 237 m pitch Currently characterizing CDF L00 sensors
Measured noise
Expected noise, assuming75% reduction instrip noise
Summary/Outlook [Hardware effort]
Development of chip in relatively advanced stage
LSTFE-2 to be submitted this summer; test beam fall or winter
Version optimized for short strips shortly thereafter
Will soon open discussion with LBNL and/or UCSC engineering to incorporate digital architecture on back end
Final test-beam run in late 2009
• Tracking Validation Studies
• Non-prompt tracks with SiD
Tracking Simulation Studies at UC Santa Cruz
Tracking Performance Package
Package is C++/ROOT written by Chris Meyer (UCSC physics major)
Reads in platform-independent flat file with specific format (output by JAS, …)
Flat file includes all relevant particles (MC) and tracks, with two-way MC Truth cross-referencing, and track/particle attributes
Also reads in error-matrix information in cos/p grid (e.g. from LCDTRK)
Efficiency vs. 500 GeV uds
pT > 0.75 GeV pT > 5.0 GeV
Some examples…
= angle between jet axis and track
II. Non-Prompt Tracks with the SiD
Entries : 2505 Mean : 863.47 Rms : 394.27
Entries : 3303 Mean : 935.72 Rms : 290.46
Entries : 459 Mean : 602.79 Rms : 358.51
Hit Count
Radius (mm)
0 200 400 600 800 1,000 1,200 1,40005
101520253035404550
Entries : 445 Mean : 369.71 Rms : 358.00
rOrg good tracks
About 5% of tracks originate outside the 2nd layer of the VXD. Is the SiD able to
reconstruct these?
People and Contributions ITim Nelson (SLAC)
AxialBarrelTracker (Snowmass ’05) finds tracks using only the five central tracking layers
Begins with three track “seed” from outer layers and works inwards
Designed to find prompt tracks if VXD disabled
People and Contributions II
Tyler Rice (UCSC Physics Major)
Optimize AxialBarrelTracker for non-prompt tracks
Benchmark and enhance performance
Lori Stevens (UCSC Physics Major)
Introduce z-segmentation algorithm into AxialBarrelTracker
Study performance vs. z segmentation
“Found”: associated with a track, with at most one hit coming from a different particle.
“Fake”: Any non-associated track with pt>0.75 and DCA < 100mm.
Particles Fakes
Found 5 Hits 131 (43%) 1
Found 4 Hits 100 (33%) 270
Not Found 73 (24%) -----
AxialBarrelTracker Effieciency StudiesOut of 304 “findable” particles in Z0 bb events
• Find 43% of particles
• Four-hit tracks seem difficult
Sources of InefficiencyRestrict to particles that hit all five layers:
166 Findable MC Particles (304 before requirement)
113 Found with 5 hits (68% vs. 43%) 25 Found with 4 hits (15% vs. 33%) 28 Missed (17% vs. 24%)
Also require all three “seed” hits to be from same particle:
144 Found with 5 hits (87% vs. 43%)
15 Found with 4 hits (9% vs. 33%)
7 Missed (4% vs. 24%)
Improving AxialBarrelTracker EfficiencyFor the vast majority of particles, all hits are within /2 of one another in azimuth (). Make this restriction…
With Azimuthal Restriction
% of MCPs
Without Azimuthal Restriction
% of MCPs
# of MCPs 304 100% 304 100%
Found with 5 hits 145 48% 131 43%Found with 4 hits 112 37% 100 33%Missed 47 15% 73 24%Fake (4 hit / 5 hit) 157 / 1 270 /1
Some improvements in efficiency and reduction of fakes…
Note: Not actual spacing between modulesHit 1
Hit 2
Possible modules for following hits
Z SegmentationCan we use z-segmentation to further clean up seeds and eliminate fake tracks? Can we make 4-hit tracks usable?
For now, apply only to three-hit seeds…
AxialBarrelTracker Effieciency
Two halves (original)
30cm segments
10cm segments
5cm segments
1cm segments
# MCPs 304 302 302 302 302
Found with 5 hits 145 142 147 152 152
Found with 4 hits 112 113 114 110 101
Missed 47 47 41 40 49
4-hit fake 157 201 141 108 45
Application of segment consistency to seeds provides improvement, but only for lengths less than 10cm
AxialBarrelTracker Effieciency
Phi Restricted Z Segmentation Results
0
50
100
150
200
250
0 0.5 1 1.5
Log base 10 of segmentation length (in cm)
N u
m b
e r
o
f
t r
a c
k s
Found: 5 hits
Found: 4 hits
Missed
Fake
Summary/Outlook [Simulation effort]
UCSC Undergraduates making important contribution to development of SiD detector.
Are part of core simulation collaboration with SLAC, FNAL, Kansas State and Brown.
One student (Lori Stevens) awarded SLAC’s Pope Fellowship for this summer; will work under Norman Graf. Rice and Myers will be supported this summer with LCRD funds.
Strategizing meeting tomorrow at noon (SLAC/UCSC) to lay out summer plans.
TIME-OVER-THRESHOLD READOUT SUMMARY
The LSTFE readout system is:
• Universally applicable (long strips, short strips, central, forward, SiD, LDC, GLD, 4th…)
• Rigorously optimized for ILC tracking
• Relative simple (reliability, yield)
• In a relatively advanced stage of development
• Is now being used as an instrument to understand fundamental principles of long ladder operation, particularly for narrow strips (CDF Layer00 sensors available, being qualified)
Silicon Microstrip Readout R&D
Initial Motivation
Exploit long shaping time (low noise) and power cycling to:• Remove electronics and cabling from active area (long ladders)• Eliminate need for active cooling
SiD Tracker
c
The Gossamer Tracker
Ideas:• Low noise readout Long ladders substantially limit electronics readout and support
• Thin inner detector layers
• Exploit duty cycle eliminate need for active cooling
Competitive with gaseous tracking over full range of momentum (also: forward region)
Alternative: shorter ladders, but better point resolution
Alternative: shorter ladders, but better point resolution
The LSTFE approach would be well suited to use in short-strip applications, and would offer several potential advantages relative to other approaches
• Optimized for LC tracking (less complex)
• More efficient data flow
• No need for buffering
Would require development of2000 channel chip w/ bump bonding (should be solved by KPiX development)
LSTFE-2 DESIGN
LSTFE-1 gain rolls off at ~10 mip; are instituting log-amp design (50 mip dynamic range)
Power cycling sol’n that cancels (on-chip) leakage currents
Improved environmental isolation
Additional amplification stage (noise, shaping time, matching
Improved control of return-to-baseline for < 4 mip signals
Multi-channel (64? 128? 256?) w/ 8:1 multiplexing of output
Must still establish pad geometry (sensor choice!)
Note About LSTFE Shaping Time
Original target: shape = 3 sec, with some controlled variability (“ISHAPR”)Appropriate for long (2m) ladders
In actuality, shape ~ 1.5 sec; tests are done at 1.2 sec, closer to optimum for SLAC short-ladder approach
Difference between target and actual shaping time understood in terms of simulation (full layout)
LSTFE-2 will have ~3 sec shaping time
Power CyclingIdea: Latch operating bias points and isolate chip from outside world.
• Per-channel power consumption reduces from ~0.5 mW to ~5 W.
• Restoration to operating point should take ~ 1 msec.
Current status:
• Internal leakage (protection diodes + ?)degrades latched operating point
• Restoration takes ~40 msec (x5 power savings)
• Injection of small current (< 1 nA) to counter leakage allows for 1 msec restoration.
Future (LSTFE-2)
• Low-current feedback will maintain bias points; solution already incorporated in LSTFE-2 design
LONG LADDER EXPERIENCEA current focus of SCIPP activity
Using GLAST “cut-off” (8 channel) sensors; 237 m pitch with 65 m strip width
Have now studied modules of varying length, between 9cm and 143cm.
Measure inputs to estimate noise sources other than detector capacitance:
• Leakage current 1.0 nA/cm• Strip resistance 3.1 /cm• Bias resistance 35 M per sensor
All of these should be considered in module design!
Strip resistance for fine pitch could be an issue are doing dedicated studies and considering options feedback to detector/module design.
Simulation Result: S/N for 167 cm Ladder (capacitive noise only)
Simulation suggests that long-ladder operation is feasible
Timing Resolution Study (50 pF Load)
Nominal expectation:
SNRSNRt
1
1
where = 1.2 s is the shaping time, = 8.8 is the applied threshold in units of rms noise, and SNR = 28. This yields an expectation of
t ~ 60 ns (expected)
t was measured at a series of input charges, which were averaged together with weights from a Landau distributions, yielding
t ~ 50 ns (measured)
Total Momentum (GeV)
Number of hits per track
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
10.5
11.0
11.5
12.0
12.5
13.0
13.5
Entries : 6712 XMean : 4.5455 XRms : 11.547 YMean : 0.28509 YRms : 0.85769
Hit count vs. Total Momentum
Number of hits on track
Tra
ck M
omen
tum
“Good”
“Other” “Knock-on” (less than 10 MeV)“Looper”
Total hits: 30510 100%Good hits: 1754 5.7%Looper hits: 13546 44.4%Knock-on hits: 10821 35.5%Other hits: 4389 14.4%
Total tracks: 6712 100%Good tracks: 445 6.6%Looper tracks: 459 6.8%Knock-on tracks: 3303 49.2%Other tracks: 2505 37.3%
What’s left after “finding” (cheating!) prompt tracks?
DIGITAL ARCHITECTURE SIMULATION
ModelSim package permits realistic simulation of FPGA code (signal propagation not yet simulated)
Simulate detector background (innermost SiD layer) and noise rates for 500 GeV running, as a function of read-out threshold.
Per 128 channel chip ~ 7 kbit per spill 35 kbit/second
For entire SiD tracker ~ 0.5-5 GHz data rate, dep-ending on ladder length (x100 data rate suppression)
NominalReadoutThreshold