1. SCIPP R&D on Linear Collider Tracking – Hardware and Simulation DOE Site Visit June 10 2008 Bruce Schumm Santa Cruz Institute for Particle Physics

2. 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 received another $37,500 in supplemental funding 2007-2008: We received $67,000 in regular funding. And another $50,000 in supplemental funding. 2008-2009 No call yet for proposals.

3. Faculty/Senior Vitaliy Fadeyev Alex Grillo Bruce Schumm Post-Docs Lei Wang Undergrads Greg Horn Luke Kelley Ian Horn Sean Crosby Jerome Carmen 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

4. 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-2008 After long-ladder proof-of-principle, can re-optimize (modest changes) for short-ladder, fast-rate application We also hope to play an increasing role in complementary hardware effort (KPiX, SiD sensor development)

5. 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

6. BRIEF SUMMARY OF STATUS Testing of 8-channel (LSTFE-1) prototype completed: 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. 128-channel test-beam prototype (LSTFE-2) has been fabricated; bench tests getting underway Now for the details…

7. 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

8. 1/4 mip 1 mip 128 mip Operating point threshold Readout threshold High gain advantageous for overall performance (channel matching)

9. DIGITAL ARCHITECTURE: FPGA DEVELOPMENT 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.

10. FIFO (Leading and trailing transitions) Low Comparator Leading-Edge-Enable Domain Proposed LSTFE Back-End Architecture Clock Period  = 400 nsec Event Time 8:1 Multi- plexing (  clock = 50 ns)

11. Simulated Resolution for 167 cm Ladder Detector Noise: Capacitive contribution; from SPICE simulation normalized to bench tests with GLAST electronics Analog 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 10  m) Lower (read) threshold in fraction of min-i (High threshold is at 0.29 times min-i)

12. 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.

13. FPGA-based control and data- acquisition system INITIAL RESULTS LSTFE chip mounted on readout board

14. 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

15. 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

16. LSTFE-2 TestBeam Prototype 128 channels with 16:1 output multiplexing Full digital architecture will still be implemented on external FPGA Noise and time-over-threshold precision re- optimized (for 80cm ladder) Power-cycling re-designed Additional amplifier stage extends dynamic range Bench tests beginning in SCIPP lab.

17. LONG LADDER CONSTRUCTION

18. 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 20  m strip width (actual is 60  m)

19. Strip Noise Idea: “Center Tapping” – half the capacitance, half the resistance? Result: no significant change in measured noise Measured noise Expected noise, assuming 75% reduction in strip noise

20. Strip Noise Study: Smaller Pitch Detector Sensors used so far have 237  m pitch (GLAST test sensors Currently developing faux long-ladder with old CDF L00 sensors, with 60  m pitch  strip noise should dominate Involves microscopic alteration to biasing circutry, implementation of low-noise readout (Sean Crosby) General study of effect of strip resistance on readout noise (generic ILC R&D)

21. Summary/Outlook [Hardware effort] LSTFE-2 back in lab; testing getting underway Generic R&D study of strip noise underway Development of FPGA-based readout and Data Acquisition system (test-beam) underway (Wang, Kelley, Fadeyev) Independent project to explore charge division (longitudinal coordinate) for special SiD sensor run underway (Carmen) Hoping to visit testbeam by end of calendar year (recent ILC upheavals at Labs have introduced uncertainties)

22. Non-prompt tracks with SiD Tracking Simulation Studies at UC Santa Cruz

23. People and Contributions Work done by UCSC thesis students supervised by Schumm Lori Stevens: “Virtual” Z segmentation studies Tyler Rice, Chris Meyer: Calorimeter-assisted tracking algorithm Chris Betancourt: Application to Gauge- Mediated SUSY models

24. Non-Prompt Tracks with the SiD About 5% of tracks originate outside the 2 nd layer of the VXD. Is the SiD able to reconstruct these?

25. Prior Tools Stand-alone (no Vertex Detector) reconstruction: AxialBarrelTracker (Tim Nelson, SLAC; optimized by Rice, Meyer) Calorimeter stub-finding within Garfield package (Dima Onoprienko; Kansas State) Work within ILC “Java Analysis Studio” (JAS) framework

26. Note: Not actual spacing between modules Hit 1 Hit 2 Possible modules for following hits Z Segmentation (Lori Stevens) Can 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…

27. Non-Prompt Track Effieciency vs. Segmentation Lori Stevens

28. 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.

29. Seeds-to-Stubs Program UCSC students proposed matching precise three-hit tracker seeds to Garfield stubs Helix – Stub Matching (optimized for Z  qq) Base Difference < 2 mm Phi Difference < 100 milliradians Curvature Ratio ( (  seed -  stub )/  seed ) < 10 e.g.: Position-matching for isolated muons (mm)

30. Seed-to-Stubs Performance; Z  qq Of a total of 20 3-hit particles: 12 were reconstructed as 3-hit tracks, with only 4 fakes Two additional 4-hit particles were found BUT: Performance much worse for e + e -  qq at E cm = 500 GeV. Could optimize for this type of event, but do we want to?  Algorithm tuning dependent on signature under exploration

31. Seed-to-Stubs Performance; Z  qq Of a total of 20 3-hit particles: 12 were reconstructed as 3-hit tracks, with only 4 fakes Two additional 4-hit particles were found Chris Betancourt applying Seed-to-Stubs algorithm to physics process with kinked tracks Currently generating Stau  tau gravitino events with ~10cm mean decay length (signal) and e + e -   +  - (background)

32. Simulation effort Summary UCSC Undergraduates making important contribution to development of SiD detector. Attracts best students and propels them on: Lori Stevens: SLAC Pope Fellowship, Berkeley engineering grad Tyler Rice: Dean’s award for thesis, UC Irvine physics grad with Chancellor’s fellowship Chris Meyer: University of Chicago physics grad; will join ATLAS at CERN this summer with Merritt/Oreglia

33. RANDOM BACK-UP SLIDES

34. 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 Mikelich and Luke Kelley (Undergrads)

35. 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 to V thresh ~ 5% of min-I Q in = 0.5 fC Q in = 3.0 fC Q in = 2.5 fC Q in = 2.0 fC Q in = 1.5 fC Q in = 1.0 fC Hi/Lo comparators function independently

36. 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)

37. Efficiency vs.  500 GeV uds pT > 0.75 GeV pT > 5.0 GeV Some examples…  = angle between jet axis and track

38. “Tri-Plots” (fitting validation)

39. TIME-OVER-THRESHOLD READOUT SUMMARY The LSTFE readout system is: Universally applicable (long strips, short strips, central, forward, SiD, LDC, GLD, 4 th …) 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)

40. “Found”: associated with a track, with at most one hit coming from a different particle. “Fake”: Any non-associated track with p t >0.75 and DCA < 100mm. ParticlesFakes Found 5 Hits131(43%) 1 Found 4 Hits100(33%) 270 Not Found 73(24%) ----- AxialBarrelTracker Effieciency Studies Out of 304 “findable” particles in Z 0  bb events Find 43% of particles Four-hit tracks seem difficult

41. 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

42. 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 of 2000 channel chip w/ bump bonding (should be solved by KPiX development)

43. Sources of Inefficiency Restrict 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%)

44. Improving AxialBarrelTracker Efficiency For 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 304100%304100% Found with 5 hits 14548%13143% Found with 4 hits 11237%10033% Missed 4715%7324% Fake (4 hit / 5 hit) 157 / 1270 /1 Some improvements in efficiency and reduction of fakes…

45. AxialBarrelTracker Effieciency Two halves (original) 30cm segments 10cm segments 5cm segments 1cm segments # MCPs 304302 Found with 5 hits 145142147152 Found with 4 hits 112113114110101 Missed 47 414049 4-hit fake 15720114110845 Application of segment consistency to seeds provides improvement, but only for lengths less than 10cm

46. 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!)

47. 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

48. Power Cycling Idea: 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

49. LONG LADDER EXPERIENCE A 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 current1.0 nA/cm Strip resistance3.1  /cm Bias resistance35 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.

50. Simulation Result: S/N for 167 cm Ladder (capacitive noise only) Simulation suggests that long-ladder operation is feasible

51. Timing Resolution Study (50 pF Load) Nominal expectation: 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)

52. Number of hits on track Track Momentum “Good” “Other” “Knock-on” (less than 10 MeV) “Looper” Total hits:30510100% Good hits:17545.7% Looper hits:1354644.4% Knock-on hits:1082135.5% Other hits:438914.4% Total tracks:6712100% Good tracks:4456.6% Looper tracks:4596.8% Knock-on tracks:330349.2% Other tracks:250537.3% What’s left after “finding” (cheating!) prompt tracks?

53. 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) Nominal Readout Threshold

54. Channel-to-Channel Matching Offset: 10 mV rms Gain: 150 mV/fC <1% rms Occupancy threshold of 1.2 fC (1875 e - )  180 mV ± 2 mV (20 e - ) from gain variation ± 10 mV (100 e - ) from offset variation