1. Silicon Detector Readout 14 June 2012 Markus Friedl (HEPHY) IPM-HEPHY Detector School

2. Contents  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout2

3. Example: CMS Experiment at CERN 14 June 2012M.Friedl: Silicon Detector Readout3 Tracker (Silicon Strip & Pixel Detectors)

4. CMS Tracker 14 June 2012M.Friedl: Silicon Detector Readout4 Silicon Strip Sensor Front-End Electronics

5.  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout5

6. Various CMS Tracker Modules 14 June 2012M.Friedl: Silicon Detector Readout6 Sensors Electronics

7. Silicon Strip Detectors  Typically 300µm thick, strip pitch 50...200µm  Reverse bias voltage for full depletion 50...500V  Connection by wire bonds 14 June 2012M.Friedl: Silicon Detector Readout7 CMS Test Sensor with various geometries (1998)Belle Sensor with 45° strips (2004) Wire bond

8. Silicon Pixel Detectors  Pixels can be square (CMS) or oblong (ATLAS)  Structure size similar to strip detectors, but N 2 channels  Connection by bump bonds 14 June 2012M.Friedl: Silicon Detector Readout8 CMS Pixel Readout SchemeCMS Pixel SensorATLAS Pixel Sensor

9. Principle of Operation  p-n junction is operated at reverse bias to drain free carriers  Traversing charged particle creates electron-hole pairs  Carriers drift towards electrodes in the electric field  Moving carriers induce current in the circuit  current source 14 June 2012M.Friedl: Silicon Detector Readout9

10. Equivalent Circuit of the Detector  Applies to many types of detectors, not only silicon  Example: wire chamber  Coaxial capacitor configuration  Moving charges induce current  Example: photomultiplier tube  Small plates  Charge (current) is amplified in each stage 14 June 2012M.Friedl: Silicon Detector Readout10 Current source with capacitor in parallel

11. Comparison: Voltage vs. Current Source PropertyVoltage SourceCurrent Source Voltageconstantanything Currentanythingconstant Idle (no power)Open (I=0)Shorted (V=0) 14 June 2012M.Friedl: Silicon Detector Readout11 IDEAL PropertyVoltage SourceCurrent Source (Linear) equivalent circuit Resistor causesInternal voltage dropInternal current drop ConversionNorton-Thevenin equivalent: R V = R I ; V = I R V/I ExamplesBattery Wall plug (AC) Detector NIM module outputs REAL

12. Moving Charges 14 June 2012M.Friedl: Silicon Detector Readout12

13. Ramo’s Theorem (1939)  Moving charges between inside electric field (e.g. parallel plates) induces current in electrodes i = E q v  Current is proportional to electric field E, (moving) charge q and velocity v of the charge  It doesn’t matter if the charges eventually reach the electrodes or not, only motion counts  Fully valid for parallel plate capacitor configuration (large area diode)  A bit more complicated for strip detectors  more later 14 June 2012M.Friedl: Silicon Detector Readout13

14. A Bit of Theory 14 June 2012M.Friedl: Silicon Detector Readout14  Space charge density is given by doping  Electric field is calculated by Poisson’s equation  Potential is found by integration of field  Shown here: full depletion = space charge just extends over full detector

15. Bias Voltage and Depletion  In reality, the electric field is imposed by applied bias voltage  What happens if V bias < V depl ?  Electric field does not cover full bulk  Only part of detector contributes to charge collection  lower efficiency  Do not operate a detector like that  What happens if V bias > V depl ?  Linear offset is added to electric field  Field tends to become more flat  Faster charge collection (Ramo)  Limited by breakdown voltage 14 June 2012M.Friedl: Silicon Detector Readout15

16. Induced Currents (1) 14 June 2012M.Friedl: Silicon Detector Readout16

17. Induced Currents (2)  Typical silicon detector (D=300µm)  Very low (<1µA), very short (~20ns)  Different contributions from moving charges  Electrons have higher mobility, thus faster  Holes with lower mobility are slower  Exponential curves with (theoretically) infinite tail at V= V depl  Almost triangular shapes at V= 2 V depl due to flatter field 14 June 2012M.Friedl: Silicon Detector Readout17

18. Induced Currents (3)  Both electrons and holes contribute to overall current, but cannot be distinguished in reality  Integral over time (area under curve) is the collected charge  If all charges reach electrodes, this is identical to the number of created pairs   i dt ≈ 3.6 fC  22500 e for a 300µm thick detector  For comparison: ~10 10 e every 20ns in a 25W bulb (230VAC)  In a simple parallel plate geometry, contributions of electrons and holes are equal  However, it’s not that simple in a strip detector… 14 June 2012M.Friedl: Silicon Detector Readout18

19. Induced Current Measurement  Quite difficult due to noise constraints  Every amplifier has a limited bandwidth and thus rise time  Nonetheless, exponential decay is clearly visible 14 June 2012M.Friedl: Silicon Detector Readout19 Single shot Averaged

20. Strip Detector Case  Ramo theorem still holds, but with some modifications  Charge movement is determined by electric field (which is approximately the same as for the parallel plate case)  Induced currents are calculated by (virtual) weighting field  Why?  Now the moving charges influence a current onto several strips depending on the geometry and distance  How to calculate weighting field?  Electrode under consideration is held at unity potential, all other electrodes at zero 14 June 2012M.Friedl: Silicon Detector Readout20

21. Strip Detector Simulation (1)  300 µm thick, 50 µm pitch, n-bulk, p-strips, V bias =1.6 x V depl 14 June 2012M.Friedl: Silicon Detector Readout21 Linear color scale! Drift Potential & Field

22. Strip Detector Simulation (2)  300 µm thick, 50 µm pitch, n-bulk, p-strips, V bias =1.6 x V depl 14 June 2012M.Friedl: Silicon Detector Readout22 Nonlinear color scale! Weighting Potential & Field

23. Strip Detector Simulation (3)  300 µm thick, 50 µm pitch, n-bulk, p-strips, V bias =1.6 x V depl 14 June 2012M.Friedl: Silicon Detector Readout23 Induced currents e-e- h+h+ sum Integrated currents: Q e- = 3338 e Q h+ = 20019 e Q sum = 23356 e Measured charge is dominated by holes: 86% @ p=50µm 82% @ p=75µm 77% @ p=120µm … 53% @ p=500µm

24. Strip Detector Case  Due to the very nonlinear weighting field, the charges which drift towards the electrode largely dominate the overall induced current  Doesn’t seem very relevant, but it actually has practical implications  Lorentz shift 14 June 2012M.Friedl: Silicon Detector Readout24

25. Lorentz Shift  In a magnetic field, charge movement is deflected due to Lorentz force which depends on the carrier mobility  Resulting in an angle (approximately ~12° for e, ~4° for h at 1.8T) and spreading of signals over several strips 14 June 2012M.Friedl: Silicon Detector Readout25

26. Lorentz Shift Compensation 14 June 2012M.Friedl: Silicon Detector Readout26

27. Silicon Detector Summary  Various Geometries:  (Diode), strips, pixels  Detector is a current source || capacitance  p-n junction operated under reverse bias voltage > V depl  Charged particle creates electron-hole pairs  Carrier motion in the electric field induces current on electrodes  Signal is typically <1µA, ~20ns  Both electrons and holes contribute to induced current  In a strip detector, current is mostly generated by charges which move towards the electrode  Deflection of carriers in a magnetic field (Lorentz shift) 14 June 2012M.Friedl: Silicon Detector Readout27

28.  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout28

29. Front-End Amplifier Principle  Located close to the sensor  First stage: Integrator  Detector current  charge  Second stage: Filter  Limit bandwidth to reduce noise 14 June 2012M.Friedl: Silicon Detector Readout29

30. Shaper Bandwidth Reduction  Example: APV25 front-end amplifier (CMS) 14 June 2012M.Friedl: Silicon Detector Readout30 SimulationMeasurement

31. Example: VA2 Chip Input Stage  VA2 is a general-purpose front-end amplifier chip with 128 inputs and multiplexed output  “Slow” shaper in the µs range  low noise 14 June 2012M.Friedl: Silicon Detector Readout31

32. Shaper Output  T p …shaping time (or peaking time)  Faster shaping can be a necessity of the experiment to distinguish subsequent events, but also implies larger noise 14 June 2012M.Friedl: Silicon Detector Readout32

33. “Low-Noise” Amplifiers  Nearly all front-end chips are called “low-noise”  General feature of the integrator+shaper combination  Noise is typically given by ENC (equivalent noise charge) referred to the input  ENC = a + b  C det (a,b...const, C det...detector capacitance)  Examples:  How can noise depend on the detector capacitance? 14 June 2012M.Friedl: Silicon Detector Readout33 T p [ns]ENC [e] VA2 (~1993)200060 + 11 / pF APV25 (CMS, 1999) 50250 + 36 / pF

34. Simplified Noise Model  Amplifier noise is projected to voltage noise source and current noise source at input  Integrator measures charge (integrated current)  Superposition analysis (one by one, other voltage sources are closed, other current sources are open; very simplified):  Q n =  i p dt + C det  V s = a + b  C det = ENC 14 June 2012M.Friedl: Silicon Detector Readout34 Amplifier noise, projected to the input

35. Full Chip Example: APV25 (CMS)  Shaping time: 50ns, sampling: 40MHz  Analog pipeline (192 cells) to store data until trigger arrives, optional APSP filter, 128:1 multiplexer, differential output driver 14 June 2012M.Friedl: Silicon Detector Readout35 8.1mm 7.1mm

36. APV25 in Action 14 June 2012M.Friedl: Silicon Detector Readout36 Sensor Pitch Adapter APV25 Hybrid Bond wires

37. Shaper Output Sampling  Usually, shaper output is sampled once at the peak  Then those values are multiplexed (1:128) to the output  The timing is given by a constant offset from the particle hit (as supplied by an external trigger, e.g. scintillator)  What happens if there are several particles with different timing? 14 June 2012M.Friedl: Silicon Detector Readout37 Peak sample

38. Pile-up Events  Strip detector measurement in a high intensity beam  Trigger – hit ambiguities and non-peak sampling can occur 14 June 2012M.Friedl: Silicon Detector Readout38 “pileups” Trigger from this particle Also returns several other samples > 0!

39. How to Avoid Such Ambiguities?  Better timing information implies more data, more energy and/or a higher noise figure  Faster Shaping = narrower output pulses  Limited by noise performance  On-chip pulse shape processing (APV25)  “Deconvolution” filter which processes samples and essentially narrows down the output to a single bunch crossing  Off-chip data processing  Using multiple subsequent samples and apply a pulse shape fit 14 June 2012M.Friedl: Silicon Detector Readout39

40. Hit Time Finding  Shaper output curve is well known with two parameters  Peak amplitude, peak timing  Event-by-event fit of shaping curve determines those two  Timing resolution of ~3ns (RMS) measured with APV25 14 June 2012M.Friedl: Silicon Detector Readout40

41. Occupancy Reduction 12 November 2011Markus Friedl (HEPHY Vienna): Status of SVD41 Belle SVD2 Belle II SVD Belle II SVD with Hit time finding Belle  Belle II: 40 x increase in luminosity

42. Front-End Amplifier Summary  Integrated circuits with typically 128 channels  2 stages:  Preamplifier (integrator: current  charge)  Shaper (band-pass filter to reduce noise)  Noise is referred to input and expressed as charge:  ENC = a + b  C det (a,b...const, C det...detector capacitance)  Shaper bandwidth determined speed and noise  Fast  large noise; slow  low noise  Required speed is usually defined by the experiment  Slow shaping and pile-up can lead to ambiguities  Tricks to circumvent speed limitation, e.g. hit time finding 14 June 2012M.Friedl: Silicon Detector Readout42

43.  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout43

44. Why?  Detector front-end is usually quite crowded  Radiation environment does not allow commercial electronics  Material budget should be as low as possible  Power consumption as well (requires cooling = material)  Thus, only inevitable electronics is put at the front-end  Everything else is conveniently located in a separate room outside the detector, traditionally called “counting house”  Allows access during machine and detector operation 14 June 2012M.Friedl: Silicon Detector Readout44

45. Example: CMS Experiment  Electronics hall is almost as big as experimental cavern  Signal distance up to 100m  Huge amount of signal transmission lines 14 June 2012M.Friedl: Silicon Detector Readout45 Experimental cavern Electronics cavern

46. Generic Transmission Chain  Signal directions  Readout (large amount): front-end to back-end, analog or digital  Controls (small amount): back-end to front-end, digital (clock, trigger, settings)  Usually, the front-end chips cannot drive the full path  Repeater (driver/receiver) is needed to amplify signals 14 June 2012M.Friedl: Silicon Detector Readout46 Front-endRepeater Back-end < 2mup to 100m

47. Excursion: Electrical Signal Transmission 14 June 2012M.Friedl: Silicon Detector Readout47  Single-ended against GND  Huge ground loop  GND compensation  Single-ended in coaxial cable  No ground loop  GND compensation  Differential twisted pair (+shield)  Largely immune  

48. Cable Bandwidth  Every cable has a finite bandwidth / damping  Nonlinear attenuation with rising frequency 14 June 2012M.Friedl: Silicon Detector Readout48  Example: CAT7 network cable (shielded twisted pairs)  Significant especially for analog signal transmission

49. Alternative: Optical Fiber  Fibers have extremely high bandwidth and very little loss  Also automatically provide electrical isolation between sender and receiver sides  However: requires conversion on both ends, which makes an optical link more expensive than a cable  Best suitable for long-haul, high-speed digital data transmission such as telecom  Nonetheless also often used in HEP experiments  Optical transmission usually implies digital signals with NRZ coding (pure AC signal with only very short DC sequences) 14 June 2012M.Friedl: Silicon Detector Readout49

50. Comparison: Copper vs. Optical Fiber PropertyCopper CableOptical Fiber Cablerigiddelicate (e.g. radius) Connectorshuge varietyfew standards Size/weightlargesmall Bandwidthlimitedhigh Losshighlow Driver + receivercheapexpensive Level isolationnoyes 14 June 2012M.Friedl: Silicon Detector Readout50

51. Example: CMS Tracker (1)  Optical fiber required because of material budget  Exceptional case: analog optical transmission  Special requirements for linearity, gain stability and noise 14 June 2012M.Friedl: Silicon Detector Readout51

52. Example: CMS Tracker (2)  Several components are customized and thus expensive  O(10000) are small quantities for industry  Estimated cost per link: ~150 € (cf. ~15 € with cable) 14 June 2012M.Friedl: Silicon Detector Readout52

53. Example: Belle II Silicon Vertex Detector  Analog APV25 readout is through copper cable to FADCs  Junction box provides LV to front-end  APV25 drives 12m cables! 14 June 2012M.Friedl: Silicon Detector Readout53 1748 APV25 chips Front-end hybrids Rad-hard DC/DC converters Analog level translation, data sparsification and hit time reconstruction Unified Belle II DAQ system ~2m copper cable Junction box ~10m copper cable FADC+PROC Unified optical data link (>20m) Finesse Transmitter Board (FTB) COPPER

54. Example: Belle II Silicon Vertex Detector  Using same APV25 chip as in CMS, but much shorter distance  no optical link required  Analog signals are attenuated in long copper cable  First attempt was an analog equalizer chip (enhancing higher frequencies) with moderate success  Later tried purely digital filter after digitization  Perfect regeneration with digital signal processing (FIR filter) at the back-end inside an FPGA  Multiplication of 8 consecutive samples with 8 filter coefficients and summing in real-time (40 MHz) 14 June 2012M.Friedl: Silicon Detector Readout54

55. Example: Belle II Silicon Vertex Detector 14 June 2012M.Friedl: Silicon Detector Readout55 Raw APV25 output FIR Optimized channelNon-optimized channel FIR filter with 8 coefficients operating continuously at 40MHz Removes cable loss and reflections due to imperfect termination! withoutwith

56. Signal Transmission Summary  Signals of large number of readout channels to be transmitted to back-end for data processing  Options: copper cable or optical fiber  Copper is much cheaper, but has frequency-dependent loss  Can be compensated e.g. with digital FIR filter at back-end  Optical links are more complicated to handle  Usually digital with NRZ coding  Exception: CMS Tracker uses analog optical links 14 June 2012M.Friedl: Silicon Detector Readout56

57.  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout57

58.  Perform all the steps which can’t be done in the front-end  Readout chain:  Receiver (electrical or optical), digitization (if analog input), data processing and reduction in FPGA (field programmable gate array), output to DAQ (data acquisition)  Receiver for clock, trigger and controls (centrally distributed) Purpose of the Back-End 14 June 2012M.Friedl: Silicon Detector Readout58

59. Example: CMS-Pixel-FED  FED means “Front End Driver” (misleading)  Contains all the elements mentioned before 14 June 2012M.Friedl: Silicon Detector Readout59 Analog optical receivers ADCs FPGAs FPGA To DAQ CLK, Trigger

60. Boards and Crates  Such boards are typically built according to a certain (industrial) standard bus system  “Standard”: VME (Versa Module Eurocard), size 9U  Obsolete: CAMAC, Fastbus  Modern: µTCA  All those standards describe  Geometry of modules  Electrical interface, power supply  Bus system for communication with crate controller & PC  Organized in crates and racks 14 June 2012M.Friedl: Silicon Detector Readout60

61. VME (9U) Crates 14 June 2012M.Friedl: Silicon Detector Readout61 Empty crate as sold by industry Belle I Silicon Vertex Detector (cable input) CMS Pixel-FED (optical input)

62. What’s an FPGA?  FPGA is a huge array of logical gates which can be combined according to the user’s need  Programming by software using basic gates & library blocks  e.g. and, adder, latch, …, CPU core  Either by schematics or by VHDL programming language 14 June 2012M.Friedl: Silicon Detector Readout62

63. Comparison: FPGA vs. CPU PropertyFPGACPU Parallelismanya few cores Speed (clock)O(100MHz)O(1GHz) I/O linesO(1000)64 Best suitable forFast, simple, massive parallel processing Complex, serial programs At the back-endFirst low-level data reduction High-level data processing (DAQ) 14 June 2012M.Friedl: Silicon Detector Readout63

64. Example: APV25 Output Data Stream 14 June 2012M.Friedl: Silicon Detector Readout64 Amplitude [ADC] Time [25ns] idle header Data frame Strip data (pedestals) Hit data

65. Strip Data Composition  Analog signal output of one event is a multiplexed stream of 128 data values, but not just the actual strip signal ADC i = S i + N i + P i + CMN  i…strip number  ADC i …measured amplitude in ADC units  S i …particle signal  N i …noise (random fluctuations)  P i …pedestal (zero value; individual for each strip)  CMN…common mode noise (common to all strips in one event)  Pedestal and noise can be measured and saved for each channel, CMN is removed event-by-event 14 June 2012M.Friedl: Silicon Detector Readout65

66. How to Process Strip Data?  Data stream with individual pedestals (dots)  Dominated by pedestal variation 14 June 2012M.Friedl: Silicon Detector Readout66  Pedestals subtracted, common mode noise and individual strip noise remains  After commom mode correction, average is at zero with random noise excursions for each strip  Next: Apply hit threshold

67. Typical Tasks for Silicon Strip Detector  ADC converts data to digital  Find and extract strip data  Put the strip data in natural order (needed if entangled, e.g. APV25)  Subtract zero value for each strip  Remove common-mode noise (appears on all strips in common)  Apply hit threshold (zero suppression, sparsification) = keep only hit data  Optional post-processing (e.g. APV25) 14 June 2012M.Friedl: Silicon Detector Readout67

68. FPGA Limits  Simple state machine, but no complex programming (instruction list) as with a CPU  Typically integer arithmetic  Made for fast I/O and throughput; internal memory is limited  Ideal for first stage of data processing – O(10) times more throughput than a state-of-the-art CPU  More complex operations at a later stage with reduced data are performed on CPU farms (DAQ) 14 June 2012M.Friedl: Silicon Detector Readout68

69. Back-End Signal Processing Summary  Performs digitization, data processing (reduction) and output to subsequent DAQ (computer farm) stage  Pedestal subtraction, common mode correction, zero suppression  Boards following a bus module standard  E.g. VME (9U)  Organized in crates and racks  Typically uses FPGAs (field programmable logic arrays)  Ideal for low-level massive parallel processing  More powerful than CPUs for such tasks  Complex calculations are done in subsequent computer farm 14 June 2012M.Friedl: Silicon Detector Readout69

70.  Silicon Detector  Front-End Amplifier  Signal Transmission  Back-End Signal Processing  Summary 14 June 2012M.Friedl: Silicon Detector Readout70

71. Thank you for your attention! 14 June 2012M.Friedl: Silicon Detector Readout71