2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH1 1 A Time Driven Readout Scheme For PET and CT… …using APDs and SiPMs F. Powolny 1), E. Auffray 1), G. Condorelli 2), S. Brunner 1), M. Despeisse 1), G. Fallica 2), H. Hillemanns 1), P. Jarron 1), A. Kluge 1), P. Lecoq 1), M. Mazzillo 2), T. C. Meyer 1,3), M. Morel 1), D. Sanfillipo 2), G. Valvo 2) 1) CERN, European Organization for Nuclear Research, 1211 Geneva 23, Switzerland, 2) STMicroelectronics, R&D DSG, Catania, Italy, 3) Presenter Fig. 5: Functional diagram of the photon detection and signal processing layout

Outline of Presentation Why a new readout technique? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH2 2005

Outline of Presentation Why a new readout technique? The Time Based approach Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH

Outline of Presentation Why a new readout technique? The Time Based approach. APD detector & TB readout Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH

Outline of Presentation Why a new readout technique? The Time Based approach. APD detector & TB readout. TOF capability? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH

Outline of Presentation Why a new readout technique? The Time Based approach. APD detector & TB readout. TOF capability? SiPMs and TB readout Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH

Outline of Presentation Why a new readout technique? The Time Based approach. APD detector & TB readout. TOF capability? SiPMs and TB readout. Summary Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH

Why a New Readout? Clinical goals: –Multimodality: CT, PET, MRI, US,... Reduce patient exposure time; Compensate for patient/organ motion; Facilitate image fusion (PET/CT); Simultaneous imaging of tumor response and responsiveness, as well as dose delivery in vivo Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH8 PHILIPS GEMINI TF

Why a New Readout? Clinical goals: –Multimodality: CT, PET, MRI, US,... Reduce patient exposure time; Compensate for patient/organ motion; Facilitate image fusion (PET/CT); Simultaneous imaging of tumor response and responsiveness, as well as dose delivery in vivo. Technical goals (From HEP to PET): –Use technologies & techniques develd for HEP State-of-the-art electronics; Compact and reliable data processing; System integration and cost Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH9 PHILIPS GEMINI TF CERN CMS

Why Time Based? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH10 Leads to simple readout architectures: Only digital pulses simple electronics No flash ADCs (running at > 500MHz, power) Simple timing circuits: Need only discriminators and TDCs; Derive time and energy information from one digital pulse. Time over threshold (T.o.T.), pulse width modulation; Build on in-house experience from large experiments.

Time Based Readout: How? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH11 i(t) V(t) t Threshold t P.A. Discr. FEDC05 NINO i(t) V(t)

Time Based Readout: How? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH12 i(t) V(t) t Threshold t Time Walk P.A. Discr. FEDC05 NINO LSO-like Test Pulse i(t) V(t)

Time Based Readout: How? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH13 i(t) V(t) t Threshold t Time Walk P.A. Discr. FEDC05 NINO LSO-like Test Pulse i(t) V(t)

Time Based Readout: How? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH14 i(t) V(t) t Threshold t Time Walk Pulse Width Timestamp from leading edge after time walk correction. Photon energy from falling edge (pulse width). Need only discriminator and TDC for both. Timestamp from leading edge after time walk correction. Photon energy from falling edge (pulse width). Need only discriminator and TDC for both. Time Walk Correction Non-Linearity Correction

APD Detector w/ LHC-Electronics 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH15 V(t)i(t) APD i(t) P.A. Discr. Hamamatsu S8550 FEDC05 (ATLAS) NINO (ALICE - TOF) FEDC05 chip: –Preamplifier developed for ATLAS Silicon Tracker –400 electrons ENC (r.m.s.) –Gain = 4mV/fC –16 channels –13ns peaking time NINO chip: –Very fast discriminator (3GHz BW) –Signal peaking time: <1ns –Output time jitter: 25ps –8 channels –Power consumption: 27mW/ch. 22 Na-Source (2 x 2 x 10 mm 3 ) SIXTH FRAMEWORK PROGRAMME PRIORITY 1 FP LIFESCIHEALTH Proposal

The NINO Chip 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH16 Out A1 + Out A1 - Out A2 + Out A2 - Out A3 + Out A3 - Out A4 + Out A4 - time [s] 5ns10ns0 time [s] 5ns10ns0 time [s] 5ns10ns0 time [s] 5ns10ns

SPICE of Readout Chain 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH17 LSO-APD Pulses FEDC05 Output NINO output (+) V(t)i(t) APD i(t) P.A. Discr. Hamamatsu S8550 FEDC05 (ATLAS) NINO (ALICE - TOF) Time Stamp Energy

Energy Resolution & Dynamic Range With APDs & NINO 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH Lu Background 122keV 57 Co 176 Lu Background 122keV 22 Na 511keV 1275keV 511keV 1275keV Raw data [Pulse Width] Corrected data [Charge/ Energy] Look-up Table Linearity after correction 57 Co 22 Na 137 Cs 22 Na Energy resolution

Timing Resolution With APDs & NINO How fast are APDs in the time encoded NINO setup? What are the limitations? What are the alternatives? The road to TOF-PET… 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH19

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH20 Timing Resolution With PMT (Reference) 470ps FWHM No corrections applied; CFD is free of Time Walk.

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH21 Timing Resolution With PMT–APD 470ps FWHM Photo-peak Selection And Time Walk Correction on APD Setup 1180ps FWHM

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH22 Timing Resolution With Dual APD 1180ps FWHM 470ps FWHM Photo-peak Selection And Time Walk Correction on both APD Setups 1600ps FWHM Raw data After photo-peak selection + Time walk correction

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH23 TOTALCrystalAPDElectronics Front end electronics noise together with the APD dark current contribute ~ 20% to the total time jitter. [ps] rms [ps] FWHM [%] Time Jitter of Detector and Readout

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH24 1.Front end electronics noise together with the APD dark current contribute ~ 20% to the total time jitter. 2.Non-uniformities in avalanche amplification in the APD account for ~ 30%. TOTALCrystalAPDElectronics [ps] rms [ps] FWHM [%] Time Jitter of Detector and Readout

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH25 1.Front end electronics noise together with the APD dark current contribute ~ 20% to the total time jitter. 2.Non-uniformities in avalanche amplification in the APD account for ~ 30%. 3.The principal jitter component of the total jitter is attributed to a)the Poisson-like photon production in LSO within the crystal decay time of 40ns, and b)the high threshold sensitivity of the APD being N ~ 20 p.e. (R = 2200 p.e., t N = 340ps, = 40ns) TOTALCrystalAPDElectronics [ps] rms [ps] FWHM [%] Time Jitter of Detector and Readout

TOF in PET: Why? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH26 From HEP we know: Event patterns congested by background; Space points help to remove confusion and improve reconstruction efficiency; Charge division, Stereo view, delay lines, cathode readout are known methods; In PET similar problems arise: Count rate contaminated with scattered and random photons; TOF reduces randoms and increases sensitivity. Data courtesy of J. S. Karp, IEEE, Trans. Med. Imag. Vol. 10 (D denotes patient diameter) D = 27cm D = 35cm Lesion Detectability APD SiPM Scatter Random True

SiPMs for TOF-PET? 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH27 SiPMs (MPPCs, etc.)… combine the advantages of conventional PMTs: fast and high gain, and those of solid state devices: compact, insensitive to magn. fields, low cost; are sensitive to single photo-electrons; are binary devices (photon counting) But have … low fill factors; high thermal (dark current-) count rates; high terminal capacitance. Pixel structure of a ST-Microelectronics 1 x 1mm 2 SiPM *) SPAD = Single Photon Avalanche Diode SiPM pulse degradation for several terminal capacitances SPAD *)

Time Response With Laser Pulses 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH28 Non-commercial 1 x 1mm 2 (400 pixels) SiPM test structure of STM – Catania: Laser pulse: 50ps, 405nm Test Setup 1 = 181ps 2 = 130ps 3 = 109ps Dark noise (1p.e.) Cross talk (2p.e.) Laser events Delay [ns] Pulse height [µA] Noise 1 SPAD 2 SPADs 3 SPADs 4 SPADs FWHM: 425ps (1p.e.) 306ps (2p.e.) 256ps (3p.e.)

Time Jitter vs. N p.e Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH29

Laser Timing With NINO 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH30 Without time walk corrections: Use look-up table for correcting data. Non-commercial 1 x 1mm 2 (400 pixels) SiPM test structure of STM – Catania:

Laser Timing With NINO 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH31 NINO has little if any influence on timing precision. FWHM: 430ps (1p.e.) 320ps (2p.e.) 275ps (3p.e.) 221ps (4p.e.) Non-commercial 1 x 1mm 2 (400 pixels) SiPM test structure of STM – Catania: With time walk corrections:

SiPM: Energy Resolution 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH32 NINO spectrum normalized from pulse width to energy (algorithm derived from SPICE simulations). LSO 3 x 3 x 20 mm 3 crystal Energy x 10 5 [eV] SiPM output FWHM: 30% NINO output FWHM: 29% Energy x 10 6 [eV] 22 Na-Source (3 x 3 x 20 mm 3 ) Hamamatsu S P: 3 x 3mm 2, 3600 pixels

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH33 22 Na SiPMHV NINO LSO 3x3x20 mm 3 PMT CFD Hamamatsu S P: 3 x 3mm 2, 3600 pixels SiPM: Timing Resolution Preliminary conclusion: SiPM timing resolution better than that of fast PMTs. SiPM resolution still contaminated by jitter from time walk (corrections to come). Time walk in energy window, expected from SPICE, to be 120ps (280ps FWHM) SiPM 71V Selection from 90ns to 110ns Width [ns] Delay [ns] After photopeak selection: FWHM SiPM = 400ps

2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH34 Summary Experience from BioCare (FP6) Work: –Validation of Time based Readout Scheme (patented). –Understand limitations in sensor performance (PMTs & APDs). – Timing resolution intrinsically limited by photon statistics and insufficient APD gain. CERN time-based electronics well adapted to SiPMs: –Simple architecture around NINO discriminator (no additional amplification needed); –Single detector time resolution of 400ps FWHM ( = 170ps) achieved. –Time walk corrections (~280ps FWHM) still pending; –Resolution intrinsically limited by photon velocity in crystal (~200µm/ps) 100ps maximum in 20mm 50ps FWHM in chosen LSO crystal.

Timing Precision vs. N p.e. (Q) 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH35 (Westcott, Knoll, Lynch, Wright) APD.34

Time Evolution of SiPM Signal 2009 Pisa Meeting on Instrumentation La Biodola, Italy – May , 2009 Thomas C. Meyer/CERN-PH36 I o = 34 μA N pe = 2000 τ crystal = 40 ns τ RC = 5 ns I SiPM (t) [uA] Time [ns] 1 p.e. response