Development of a Pulse Shape Discrimination IC Michael Hall Southern Illinois University Edwardsville VLSI Design Research Laboratory October 20, 2006
Design Team Southern Illinois University Edwardsville: Dr. George Engel (PI) Michael Hall (graduate student) Justin Proctor (graduate student) Washington University in St. Louis: Dr. Lee Sobotka (Co-PI) Jon Elson (electronics specialist) Dr. Robert Charity Western Michigan: Dr. Mike Famiano (Co-PI)
NSF Proposal (Funded) Design, simulate, and fabricate a PSD chip suitable for use with CsI(Tl) (used for charge-particle discrimination) Liquid Scintillator (used for neutron-gamma discrimination), for example: Nuclear Enterprises (NE213) Bicron (BC501A) 8 channel prototype chip 16 channel production chip
Overview of PSD System OR A VME C F D Exte rna l logic V(t) [current through a load R] C B Detector Gate control WA WB WC DA DB DC PSD Integrator Chip A B C T Sa m ple Integration gates D E L A Y Multiplexed with other chips and sent to 4 channels of one VME Pipeline ADC Cable Detector (PMT or photodiode) External discriminators (CFDs) External delay lines so we can start integrations before arrival of pulse External control voltages determine Delay and Width of integration periods Outputs A, B, C integrator voltages and relative time, T
Channel 3 on-chip sub- channels for integrators A, B, C Delay and width of integrators set by externally supplied control voltages Timing relative to a common stop signal
Sub-Channel
Op Amp to be Used in Integrator Gain Bandwidth Product: 34 MHz Low-frequency open-loop gain: 74 db Supply Current: 1.25 ma Power Consumption: 6 mw
Simulated Input Pulse for CsI(Tl) Detector Input Pulses (V) Integrators A 0 to 600 ns B 1000 to 7000 ns C 0 to 9000 ns Integration periods at the beginning of the signal are assumed to start before the pulse (at -55 ns).
Noise Sources Poisson noise due to random arrival of discrete electrons Electronics Noise Jitter noise created by an uncertainty in the integration start time and in the width of integration period RI thermal noise from the integrating resistor sampled onto the integrating capacitor OTA thermal noise of the op amp sampled onto the integrating capacitor OTA (+) continuous additive input-referred thermal noise of the op amp 1/f 1/f noise of the op amp sampled onto the integrating capacitor 1/f (+) continuous additive input-referred 1/f noise of the op amp ADC quantization noise of a 12-bit converter
1/f Noise Model Input Referred 1/f Noise 1/f Noise + Thermal Noise 1 x 10-5 Time-Domain Plot of Signal Noise (V) Spectral Density (dbv/hz 1/2 ) 0.5 0-0.5-1 0 0.5 1 1.5 2 Time (ns) 2.5 3 3.5 4 x 10 4 Spectral Plot of Signal -140-160 -180 σ 100kHz -160dB -10dB / decade noise slope -200 10-2 10-1 10 0 10 1 10 2 10 3 Frequency (MHz) 1/f dominant Thermal dominant σ 100kHz -160dB MATLAB Equivalent Model of 1/f Noise K = 8.745e-12 (constant for 1/f model) Spectre Simulation of OTA Noise
Relative Importance of Noise Sources on Performance for CsI(Tl) Detector 70.0 60.0 50.0 40.0 30.0 20.0 10.0 Plot of Standard Deviation of Noise High Energy Detector CsI(Tl) Low Energy Proton Alpha Proton Alpha Integrator A SNR = 55.66 57.27 29.04 30.50 Energy = 100.00 100.00 1.00 1.00 db MeV Detector: CsI(Tl) Integrators A 0 to 600 ns, RI = 100kΩ B 1000 to 7000 ns, RI = 40kΩ C 0 to 9000 ns, RI = 100kΩ CI = 10pF Jitter Start: 1.00 ns Period: 0.50 ns ADC: 12 bit STD (dbuv) STD (dbuv)
Summary of Noise Analysis (CsI( CsI) Poisson noise dominates for high-energy particles, but tends to be on par with electronics noise (10 pf integrating capacitor) for low- energy particles. Jitter induced noise is not a dominant noise source, but is on par with Poisson noise for the A integrator at high energy. 1/f noise dominates for low-energy particles on the B and C integrators Electronics noise on par with quantization noise of 12-bit ADC except for 1/f noise for B and C integrators at low energy.
Pulse Shape Discrimination Plot for CsI(Tl) Detector Integrator B (mv) 600 500 400 300 200 100 PSD (Pulse Shape Discrimination) Plot Alpha Proton 0 0 50 100 150 200 250 300 350 400 450 Integrator A (mv) Detector: CsI(Tl) Integrators: A, B Energy Max: 100 MeV (for 2V at input of integrator) Energy Range: 1 100 MeV Includes all noise sources
Angular PSD Plots (CsI( CsI) Energy = 1 MeV, Perr = 2.16% Energy = 10 MeV, Perr = 0% 1200 2500 Alpha Proton Count 1000 800 600 Count 2000 1500 400 1000 200 500 0 40 45 50 55 60 65 Theta 0 50 51 52 53 54 55 56 57 58 59 60 Theta Count 4000 3500 3000 2500 2000 1500 1000 500 0 Energy = 100 MeV, Perr = 0% 51 52 53 54 55 56 57 58 59 Theta Detector: CsI(Tl) Integrators: A, B Energy Max: 100 MeV Energy Range: 1 100 MeV 5000 realizations Includes all noise sources
Simulated Input Pulse for Liquid Scintillator Detector Plot of Gamma and Neutron input pulses using a Liquid Scintillator Detector for 10 MeV incident radiation Input Pulses (V) 10 0 10-1 MAX INPUT VOLTAGE (2V) Gamma Neutron Integrator A Integrator B Integrator C 10-2 0 50 100 150 200 250 Time (ns) Integrators A 0 to 200 ns B 30 to 202 ns C 50 to 204 ns Integration periods at the beginning of the signal are assumed to start before the pulse (at -55 ns) (no jitter at the start of integration).
Relative Importance of Noise Sources on Performance for Liquid Scintillator Detector 90.0 80.0 70.0 60.0 50.0 40.0 30.0 Plot of Standard Deviation of Noise High Energy Detector LiquidScin Low Energy Integrator A 20.0 Gamma Neutron Gamma Neutron SNR = 44.76 44.76 23.70 23.70 db Energy = 10.00 10.00 0.10 0.10 MeV Detector: Liquid Scintillator Integrators A 0 to 200 ns, RI = 2kΩ B 30 to 202 ns, RI = 400Ω C 50 to 204 ns, RI = 400Ω CI = 10pF Jitter Start: 1.00 ns Period: 0.50 ns ADC: 12 bit STD (dbuv) STD (dbuv)
Summary of Noise Analysis (Liquid Scintillator) Poisson noise no longer dominates except for integrator A in which the integration begins before the start of the pulse. Jitter becomes very important for B and C integrators and dominates at high energy levels. Electronics noise (especially for B and C integrators) is significantly larger than the quantization noise of a 12-bit ADC but still on par with the Poisson noise. 1/f noise is on par with the thermal noise for low-energy particles on the B and C integrators.
Pulse Shape Discrimination Plot for Liquid Scintillator Detector Integrator B (mv) 600 500 400 300 200 100 PSD (Pulse Shape Discrimination) Plot Gamma Neutron Detector: Liquid Scintillator Integrators: A, B Energy Max: 10 MeV (for 2V at input of integrator) Energy Range: 0.1 10 MeV Includes all noise sources 0-100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Integrator A (V)
Angular PSD Plots (Liquid Scintillator) Energy = 0.1 MeV, Perr = 23.32% Energy = 1 MeV, Perr = 0.19% 1000 1200 Gamma Neutron 800 1000 Count 600 Count 800 600 400 400 200 200 0 0 5 10 15 20 25 30 35 40 45 Theta 0 10 15 20 25 30 Theta Count 2000 1500 1000 500 Energy = 10 MeV, Perr = 0% 0 8 10 12 14 16 18 20 22 24 26 28 Theta Detector: Liquid Scintillator Integrators: A, B Energy Max: 10 MeV Energy Range: 0.1 10 MeV 5000 realizations Includes all noise sources
Analytical Predictions of Variance of Angular PSD Plots var 2 sin 2θ 1 1 ( θ ) + = 2 4 2 SNR A SNR B FOM = var θ θ 1 ( θ ) + var( θ ) 1 0 0 Variance of angular PSD plot depends on the signal-to to-noise ratio of the A and B integrators. Small signal-to to-noise ratios, which correspond to low-energy particles, results in a larger variance in angle which is consistent with simulation. Figure of merit (FOM) is computed as the difference between the means divided by the square root of the sum of the variances.
Conclusions Proposed PSD IC will work very well with CSI detectors with performance limited by Poisson noise. Particles differing in energy by 40 db can be easily discriminated. Proposed PSD IC will work reasonably well with Liquid Scintillator detectors with performance limited most likely by the level of timing jitter. Particles differing in energy by more than 20 db will have high probability of misclassification. While the electronics noise dominates for the B and C integrators for both detectors at low energy, it is clearly worse for the Liquid Scintillator detector where it is significantly higher than the quantization noise of a 12-bit ADC.
Conclusions Correlated double sampling to deal with 1/f noise does not appear mandatory. Poisson noise dominates in the A integrator for both CsI(Tl) and Liquid Scintillator detectors at all energies except for the CsI(Tl) at low energies. A 10 pf integrating capacitor will be used along with a bank of 8 resistors: 400 Ω,, 1 kω, k 2 kω, k, 4 kω, k, 10 kω, k, 20 kω, 40 kω, k 100 kωk The integrating op amp will consume 6 mw of power so for an 8 channel IC, the integrating op amps will require approximately 150 mw of power (300 mw for a 16- channel IC).
Future Work For a stochastic processes course, will create an optimizer to maximize the FOM on the PSD plots. Behavioral simulations to determine performance of on-chip time-to to-voltage converters. Special attention will be given to reducing on-chip induced timing jitter. Behavioral level simulations (VerilogA) to verify functionality of one complete channel including read-out electronics
Future Work Circuit design and simulation Layout Fabrication Testing of the IC