1. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Introduction to Electronics in HEP Experiments Philippe Farthouat CERN

2. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Credits and sources of information 2  I have “stolen” a lot from the previous summer student lectures from Christophe de la Taille and Jorgen Christiansen and from colleagues from the PH electronics group (PH-ESE)  Useful and more complete information can be found in the following sites:  CERN technical training ELEC 2005: http://indico.cern.ch/conferenceDisplay.py?confId=62928  LEB/LECC/TWEPP workshops from last 12 years: http://lhc-electronics-workshop.web.cern.ch/lhc%2Delectronics%2Dworkshop/  PH-ESE seminars: http://indico.cern.ch/categoryDisplay.py?categId=1591

3. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Outline 3 Detector Analog To Digital Conversion Data Acquisition & Processing Analog Processing On-detector Or Off-detector  Analog processing  Analog to digital conversion  Technology evolution  Off-detector digital electronics

4. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Outline 4 Detector Analog To Digital Conversion Data Acquisition & Processing Analog Processing On-detector Or Off-detector  Analog processing  Analog to digital conversion  Technology evolution  Off-detector digital electronics

5. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Analog Processing 5  A few basic reminders  Modelisation of the detector  Charge and current amplifiers  Noise  Example of a preamplifier design

6. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 6 The foundations of electronics  Voltage generators or source  Ideal source : constant voltage, independent of current (or load)  In reality : non-zero source impedance R S R S → 0 V RS → ∞RS → ∞ i  Current generators  Ideal source : constant current, independent of voltage (or load)  In reality : finite output source impedance R S Z V i  Ohms’ law  Z = R, 1/jωC, jωL  Note the sign convention

7. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 7 Frequency domain & time domain  Frequency domain :  V(ω,t) = A sin (ωt + φ)  Described by amplitude and phase (A, φ)  Transfer function : H(ω) [or H(s)]  The ratio of output signal to input signal in the frequency domain assuming linear electronics  V out (ω) = H(ω) V in (ω)  Time domain  Impulse response : h(t)  The output signal for an impulse (delta) input in the time domain  The output signal for any input signal v in (t) is obtained by convolution * :  V out (t) = v in (t) * h(t) = ∫ v in (u) * h(t-u) du  Correspondance through Fourier transforms  A few useful Fourier transforms in appendix below H(ω) v in (ω) v out (ω) h(t) v in (t) v out (t) F -1

8. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Appendix: Fourrier Transform 8

9. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 9 Using Ohm’s law  Example of photodiode readout  Used in high speed optical links  Signal : ~ 10 µA when illuminated  Modelisation :  Ideal current source Iin  pure capacitance C d I in C d  Simple I to V converter : R  R = 100 kΩ gives 1V output for 10 µA light volts 10 Gb/s optical receiver 100K V  Speed ?  Transfer function H(ω) = v out /i in  H has the dimension of Ω and is often called « transimpedance » and even more often (improperly) « gain »  1/RC d is called a « pole » in the transfer function

10. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 10 Frequency response  Bode plot  Magnitude (dB)  -3dB bandwidth : f -3dB = 1/2πRC d  R=10 5 Ω, C d =10pF => f -3dB =160 kHz  At f -3dB the signal is attenuated by 3dB = √2, the phase is -45°  Above f -3dB, gain rolls-off at -20dB/decade (or -6dB/octave) 100 dBΩ 80 dBΩ Magnitude Phase

11. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 11 Time response  Step response : rising exponential  Rise time : t 10-90% = 2.2 τ  « eye diagram »  Impulse response  τ (tau) = RC d = 1 µs : time constant  Speed : ~ 10 µs = 100 kb/s !  5 orders ofmagnitude away from a 10 Gb/s link ! pulse response t r 10-90% Impulse response 10Gb/s eye diagram (10 ps/div)

12. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Feedback  Y is a source linked to X  y = μ x  Open loop  x = δ e  y = µ x  s = σ y = σ μ δ e  Closed loop 12 e x y s δ µ ß σ  μ is the open loop gain  βμ is the loop gain

13. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Interest of feedback  In electronics  μ is an amplifier gain  β is the feedback loop  If μ is large enough the gain of the system is independent of the amplifier gain 13 e x y s δ µ ß σ

14. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Operational Amplifier  Gain A very large  Input impedance very high  i.e input current = 0  A(ω) as shown 14 - + -A  

15. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 How does it work? 15  Direct gain calculation  Feedback equation  Ideal Opamp - + -A   Vin Vout I R1 R2

16. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 16 Overview of readout electronics  Most front-ends follow a similar architecture Very small signals (fC) -> need amplification Measurement of amplitude and/or time (ADCs, discris, TDCs) Several thousands to millions of channels Detector Preamp Shaper ADC fC V bits DAQ & Processing V

17. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 17 Readout electronics : requirements Low cost ! (and even less) Radiation hardness High reliability High speed Large dynamic range Low power Low material Low noise

18. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 18 Detector(s)  A large variety  A similar modelization ATLAS Liquid Argon Electromagnetic calorimeter PMT for Antares CMS Pixel module 6x6 pixels,4x4 mm 2

19. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 19 Detector modelization  Detector = capacitance Cd  Silicon : 0.1-10 pF  PMs : 3-30pF  Ionization chambers 10-1000 pF  Signal : current source  Silicon : ~1fC/100µm  PMs : 1 photoelectron -> 10 5 -10 7 e -  Wire chambers : a few 10 3 e -  Modelized as an impulse (Dirac) : i(t)=Q 0 δ(t)  Missing :  High Voltage bias  Connections, grounding  Neighbours  Calibration… I in C d Detector modelisation Typical PM signal

20. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 20 Reading the signal  Signal  Signal = current source  Detector = capacitance C d  Quantity to measure  Charge => integrator needed + ADC  Time => discriminator + TDC  Integrating on Cd  Simple : V = Q/C d  « Gain » : 1/C d : 1 pF -> 1 mV/fC  Need a follower to buffer the voltage…  Input follower capacitance : Ca // Cd  Gain loss, possible non-linearities  Crosstalk  Need to empty Cd… I in C d - + Q/Cd Impulse response

21. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 21 Ideal charge preamplifier  Ideal opamp in transimpedance  Shunt-shunt feedback  Transimpedance : v out /i in  Vin-=0 =>V out (ω)/i in (ω) = - Z f = - 1/jω C f  Integrator : v out (t) = -1/C f ∫ i in (t)dt  « Gain » : 1/C f : 0.1 pF -> 10 mV/fC  C f determined by maximum signal  Integration on Cf  Simple : V = - Q/C f  Unsensitive to preamp capacitance C PA  Turns a short signal into a long one  The front-end of 90% of particle physics detectors…  But always built with custom circuits… - + Cf I in C d v out (t) = - Q/C f - Q/Cf Charge sensitive preamp Impulse response with ideal preamp

22. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 22 Non-ideal charge preamplifier Impulse response with non-ideal preamp  Finite opamp gain  Small signal loss in C d / G 0 C f << 1 (ballistic deficit)  Finite opamp bandwidth  First order open-loop gain  G(ω) = G 0 /(1 + j ω/ω 0 )  G 0 : low frequency gain  G 0 ω 0 : ω c gain bandwidth product  Preamp risetime  Due to gain variation with ω  Time constant : τ (tau) = C d / G 0 ω 0 C f  Rise-time : t 10-90% = 2.2 τ  Rise-time optimised with G 0 ω 0 (ω c ) or C f

23. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Input Impedance 23 - + -G   Vin Vout I CfCf I

24. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 24 Charge preamp seen from the input  Input impedance with ideal opamp  Zin = Zf / G+1  Zin->0 for ideal opmap  « Virtual ground » : Vin = 0  Minimizes sensitivity to detector impedance  Minimizes crostalk  Input impedance with real opamp  Resistive term :  Example : ω C = 10 9 rad/s C f = 2 pF => Rin = 500Ω  Determines the input time constant : t = R eq C d  Good stability Input impedance of charge preamp

25. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 25 Pile up  It is necessary to discharge the feedback capacitor  Successive input pulses would add up until saturation  Several ways to do it, the simpler being to put a resistor in parallel - + -A  CfCf RfRf

26. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 26 Crosstalk  Capacitive coupling between neighbours  Crosstalk signal is differentiated and with same polarity  Small contribution at signal peak  Proportionnal to Cx/Cd and preamp input impedance  Slowed derivative if RinCd ~ tp => non-zero at peak  Inductive coupling  Inductive common ground return  “Ground apertures” = inductance  Connectors : mutual inductance

27. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 27 Current preamplifiers  Transimpedance configuration  V out (  )/i in (  ) = - R f / (1+Z f /GZ d )  Gain = R f  High counting rate  Typically optical link receivers  Easily oscillatory  Unstable with capacitive detector  Inductive input impedance  Resonance at :  Quality factor :  Q > 1/2 -> ringing  Damping with capacitance C f in parallel to R f  Easier with fast amplifiers Current sensitive preamp Step response of current sensitive preamp RfRf

28. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 28 Charge vs Current preamps  Charge preamps  Best noise performance  Best with short signals  Best with small capacitance  Current preamps  Best for long signals  Best for high counting rate  Significant parallel noise  Charge preamps are not slow, they are long  Current preamps are not faster, they are shorter (but easily unstable) CurrentCharge

29. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 29 Electronics noise  Definition of Noise  Random fluctuation superimposed to interesting signal  Statistical treatment  Three types of noise  Fundamental noise (Thermal noise, shot noise)  Excess noise (1/f …)  Parasitics -> EMC/EMI (pickup noise, ground loops…)

30. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 30 Electronics noise  Modelization  Noise generators : e n, i n  Noise spectral density of e n & i n : S v (f) & S i (f)  S v (f) = | F (e n )| 2 (V 2 /Hz)  S i (f) = | F (i n ) | 2 (A 2 /Hz)  Rms noise Vn  V n 2 = ∫ e n 2 (t) dt = ∫ S v (f) df  When going through a device H(2πf)  S vout (f) = |H(2πf)| 2 S vin (f) rms Rms noise vn Noise spectral density

31. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 31 Calculating electronics noise  Fundamental noise  Thermal noise (resistors) : S v (f) = 4kTR  Shot noise (junctions) : S i (f) = 2qI  1/f noise in CMOS devices  Noise referred to the input  All noise generators can be referred to the input as 2 noise generators :  A voltage one e n in series : series noise  A current one i n in parallel : parallel noise  Two generators : no more, no less… why ? Noisy Noiseless enen Thermal noise generator Noise generators referred to the input  To take into account the source impedance  Golden rule  Always calculate the signal before the noise what counts is the signal to noise ratio  Don’t forget noise generators are V 2 /Hz => calculations in module square  Practical exercice next slide

32. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 32 Parallel noise Series noise Noise spectral density at Preamp output Noise in charge pre-amplifiers  2 noise generators at the input  Parallel noise : ( i n 2 ) (leakage currents)  Series noise : (e n 2 ) (preamp)  Output noise spectral density :  Parallel noise in 1/ω 2  Series noise is flat, with a « noise gain » of C d /C f  rms noise V n  V n 2 = ∫ S v (ω) dω/2π -> ∞ (!)  Benefit of shaping…

33. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 33 Equivalent Noise Charge (ENC) after CRRC n  Noise reduction by optimising useful bandwidth  Low-pass filters (RC n ) to cut-off high frequency noise  High-pass filter (CR) to cut-off parallel noise  -> pass-band filter CRRC n  Equivalent Noise Charge : ENC  Noise referred to the input in electrons  ENC = Ia(n) e n C t /√τ  Ib(n) i n * √τ  Series noise in 1/√τ  Paralle noise in √τ  1/f noise independant of τ  Optimum shaping time τ opt = τ c /√2n-1  Peaking time tp (5-100%)  ENC(tp) independent of n  Complex shapers are obsolete :  Power of digital filtering  Analog filter = CRRC ou CRRC 2 Step response of CR RCn shapers ENC vs tau for CR RCn shapers

34. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 34 Equivalent Noise Charge (ENC) after CRRC n  A useful formula : ENC (e- rms) after a CRRC 2 shaper :  e n in nV/ √Hz, i n in pA/ √Hz are the preamp noise spectral densities  C tot (in pF) is dominated by the detector (C d ) + input preamp capacitance (C PA )  t p (in ns) is the shaper peaking time (5-100%)  Noise minimization  Minimize source capacitance  Operate at optimum shaping time  Preamp series noise (en) best with high trans-conductance (g m ) in input transistor  => large current, optimal size

35. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 35 ENC for various technologies  ENC for Cd=1, 10 and 100 pF at I D = 500 uA  MOS transistors best between 20 ns – 2 µs Parameters Bipolar : g m = 20 mA/V R BB’ =25 Ω e n = 1 nV/ √ Hz I B =5uA i n = 1 pA/√Hz C PA =100fF PMOS 2000/0.35 g m = 10 mA/V e n = 1.4 nV/√Hz C PA = 5 pF 1/f :

36. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 36 MOS input transistor sizing  Capacitive matching : strong inversion  g m proportionnal to W/L √I D  C GS proportionnal to W*L  ENC propotionnal to (Cdet+C GS )/ √gm  Optimum W/L : C GS = 1/3 Cdet  Large transistors are easily in moderate or weak inversion at small current  Optimum size in weak inversion  g m proportionnal to I D (indep of W,L)  ENC minimal for C GS minimal, provided the transistor remains in weak inversion © P O’Connor BNL

37. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 37 Low frequency hybrid model of bipolar Design in micro-electronics  Performant design is at transistor level  Simples models  Hybrid π model  Similar for bipolar and MOS  Essential for desgin  Numerous « composites »  Darlington, Paraphase, Cascode, Mirrors… BC CECC Three basic bricks Common emitter (CE) = V to I (transconductance) Common collector (CC) = V to V (voltage buffer) Common base (BC) = I to I (current conveyor)

38. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (1) 38  From the schematic of principle  Using of a fast opamp (OP620)  Removing unnecessary components…  Similar to the traditionnal schematic «Radeka 68 »  Optimising transistors and currents - + Cf Schematic of a OP620 opamp ©BurrBrown Charge preamp ©Radeka 68

39. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (2) 39  Simplified schematic  Optimising components  What transistors (PMOS, NPN ?)  What bias current ?  What transistor size ?  What is the noise contributions of each component, how to minimize it ?  What parameters determine the stability ?  Waht is the saturation behaviour ?  How vary signal and noise with input capacitance ?  How to maximise the output voltage swing ?  What the sensitivity to power supplies, temperature… Q2 : CB I C2 =100µA Q3 : CC I C3 =100µA Q1 : CE I C1 =500µA Simplified schematic of charge preamp

40. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (3) 40  Small signal equivalent model  Transistors are reaplaced by hybrid π model  Allows to calculate open loop gain g m1  Gain (open loop) :  Ex : g m1 =20mA/V, R 0 =500kΩ, C 0 =1pF => G 0 =10 4 ω 0 =210 6 G 0 ω 0 =2 10 10 = 3 GHz ! v out /v in = - g m1 R 0 /(1 + jω R 0 C 0 ) v in v out R 0 C 0 R0 = Rout2//Rin3//r04 Small signal equivalent model of charge preamp

41. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (4) 41 Cf Input Output  Complete schematic  Adding bias elements

42. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (5) 42  Complete simulation  Checking hand calculations against 2 nd order effects  Testing extreme process parameters (« corner simulations »)  Testing robustness (to power supplies, temperature…) 1 MHz 10 ns20 ns Simulated open loop gain Saturation behaviour

43. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 Example : designing a charge preamp (6) 43  Layout  Each component is drawn  They are interconnected by metal layers  Checks  DRC : checking drawing rules (isolation, minimal dimensions…)  ERC : extracting the corresponding electrical schematic  LVS (layout vs schematic) : comparing extracted schematic and original design  Simulating extracted schematic with parasitic elements  Generating GDS2 file  Fabrication masks : « reticule » 100 µm

44. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 44 Device available  Acces to microelectronics 6 cm FET driverpreamp 100 µm ZfZf Z0Z0 Q2 Cf Q1 Q3 Cf Charge preamp in 0.8µm BiCMOS Charge preamp in SMC hybrid techno

45. philippe.farthouat@cern.chIntroduction to Electronics Summer 2009 45 Summary Charge preamplifier architecture  Charge sensitive preamplifiers  Output proportionnal to the incoming charge  « Gain » : 1/C f : C f = 1 pF -> 1 mV/fC  Transforms a short pulse into a long one  Low input impedance -> current sensitive  Virtual resistance R in -> stable with capacitive detector  The front-end of 90% of particle physics detectors…  But always built with custom circuits…  Noise minimization  Minimize source capacitance  Operate at optimum shaping time  Preamp series noise (en) better with high trans-conductance (gm) in input transistor v out (t) = - Q/C f