Readout electronics and Detector Design for PET Applications

Readout electronics and Detector Design for PET Applications Ilaria Sacco Institute for Computer Engineering Heidelberg University

Positron Emission Tomography • • • Glucose-based tracer labeled with positron emitter is injected into the body • Coincident events at fix energy • Large number of Lines of Response (LOR) for image reconstruction • Resolution depends on timing, energy, spatial resolution, sensitivity, detector geometry, correction algorithm Measure the uptake rate Functional imaging technique (cell metabolism, oxygen use, blood flow, cancer detection) High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 2

Combined scanners PET/CT • • • Computed Tomography is based on x-rays transmission Data can be adjusted and used for attenuation correction High radiation dose to patient Sequential scanners Low contrast in soft tissues Now established in clinical routine PET/MR • • • High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Magnetic resonance is based on magnetic proprieties of atoms Complex attenuation correction No radiation dose to patient Needs simultaneous scanner Good contrast in soft tissues + functional imaging possible First scanners available 3

Time of Flight Measurements Time measurement difference < ns No TOF Coincidence Resolving Time CRT < 600 ps FWHM Improvement of the SNR and more accurate localization of the event along LOR High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 4

Detector Design Goals Scintillators Si. PMs • Compact MR-compatible design • Optimal pixel size • Optimal timing resolution • High sensitivity Readout electronics Diameter ~ 15 -20 cm Diameter ~ 80 cm Very small scintillators < 1 mm pitch Coincidence detection Optimum pixel size 2. 6 mm Time of Flight (TOF) Pre-clinical High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Clinical 5

DETECTOR DESIGN

Hyper. Image Scanner LYSO Scintillator array Si. PM array – area 4. 0× 4. 0 mm² Two PETA ASICs, 64 channels, wire-bonded on the PCB Data processing stage with FPGA Hyper. Image is an FP 7 EU project (2007 -2011) High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 7

Sublima vs. Hyper. Image Detector Module Sublima Project § One LTCC layer § 4 flip-chip mounted ASICs § 144 Channels § Si. PM pitch 2. 5× 2. 5 mm 2 § Internal liquid cooling High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Hyper. Image Project § Three PCB layers § 2 wire-bonded ASICs § 64 Channels § Si. PM pitch 4× 4 mm 2 § External liquid cooling with copper pipes 8

Sublima Detector Module Single layer ɣ-detection module 18 mm • Low Temperature Co-Fired Ceramic substrate (LTCC) • Pixel size 2. 5× 2. 5 mm² • Read-out electronics designed for To. F performances • 144 channels per module • MR compatible • Very compact design • Internal water-cooling High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 ca m . 3 0 m m 0 m . 3 a c 9

Detector Design • Rectangular Si. PM dies in a 2× 6 geometry • FBK HD-RGB technology – active area 2. 25 × 2. 25 mm² - cell size 25× 25 µm² 5 mm Regular 2. 5 mm pitch all over the array 15 mm Tracks resistance is adjusted depending on Si. PM position Wire-bonding pads at the external edge no gaps in the active region High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 10

Detector Design Readout electronics placed on the bottom surface • • • Four PETA 5 flip-chip mounted and controlled in parallel Almost no external components needed Inlet & outlet for liquid cooling temperature I 2 C sensors EEPROM memory for configuration storage termination resistors and decoupling capacitors High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 11

Internal Liquid Cooling 2 mm Fabricated at Micro System Engineering, Berg, Germany 11 Signal layers Liquid channel ΔT ~ 3 K § Room temperature water § Water flow rate ~ 10 ml/sec § Stabilize more than four times faster than external cooling, about 10 sec High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 12

Module Assembly Si. PM Placement Wire-bonding @ ZITI @ MSE 18 mm ca m . 3 0 m m Gluing tests 0 m 12× 12 LYSO array ( 2. 5× 10 mm³) ESR + Al + ESR layers between crystals 3 . ca High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 13

PETA

Position Energy Timing Asic Aimed to a compact module design: • • • 5× 5 mm² area, UMC 180 nm – six-metal layers technology 36 channels Bump-bonded with pads pitch ~ 270 µm Nearly no external components needed: maximal integration on chip Configuration and control scheme shared up to 4 chips with very few lines Two independent front-ends per channel: • • single-ended with low input impedance differential-ended reliable in MR environment Power: • 35 m. W analogue/ 8 m. W digital per channel High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 15

PETA History 2007 2017 Chip Name # Channels PLL Commments PETA 1 40 yes Integrator, leading edge discriminator TC_UM 12 8 no Low power latch test PETA 2 36 yes Threshold adjustment improvements PETA 3 36 yes Bandgap reference, neighbor logic, SAR ADC ISIS 1 16 no Single-ended front-end test chip PETA 4 36 yes Bump-bonded, single/diff ended FE To. T veto PETA 5 36 yes Debug PETA 4 PETA 6 36 yes Both Si. PMs polarities, high data rate, large bump pitch version Previously, other 7 test chips have been implemented (AMS 350 and UMC 180 technology) for proving the timing measurement concept High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 16

PETA Design • • • Single-Ended Front-end used on the detector module Timing/Energy Thresholds Two identical halves with separate PLL/TDC circuit TDC with 50 ps binwdith Parallel readout of L/R side High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 17

Single-Ended Front-end Channel Inject Low Impedance Front-End Charge Integrator & ADC gain DAC Input VREF IRamp Discriminator Si. PM Finite State Machine Inject - HV • • • Input impedance < 10Ω Adjustable input DC potential ± 0. 4 V Trim Threshold per channel VCM High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 x 4 Th. N Th. P 18

Charge Integrator 1. 2. 3. 4. Idle: T 1, T 2 close HIT: T 1 close, T 2 open Discharge: T 1, T 2 open Idle: T 1, T 2 close Programmable up to 200 ns VOUT Digitized energy, Tclock 3. 2 ns VREF 1 2 Gain and Ramp current can be used for tuning channel dead time Typical around 300 -500 ns for 511 Ke. V events High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 3 4 Time 19

Readout Architecture for High Data Rate Shift LEFT Read frame 33 bits Buffer Shift RIGHT 1 1 • • • Shift frequency of 311 MHz Additional buffer for 1 event word Two LVDS output lines per half Bypass readout ALL SELECTABLE FEATURES Shift RIGHT 1 Buffer 33 bits Shift LEFT 1 PETA 6: What’s new? LVDS High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 20

Three PETA 6 Versions Design of three version: • Standard front-end for negative pulses (as PETA 4/5) • Inverted front-end for positive pulses • Single-ended version with bumps pitch of 500 µm PETA 6 -SE can be flipchip mounted @ZITI on dedicated PCB PETA 6 6 mm 5 mm High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 21

RESULTS

Timing resolution Clinical configuration to prove To. F performance • • • ɣ 1: 1 coupling configuration for best light collection LYSO array with 3 reflector layers (ESR + Al+ ESR) between crystals Coincidence setup with opposite modules High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 23

Timing resolution Point-form Na 22 source, moved in different location for measuring many coincidence pairs A 1 B 1 A 8 B 8 B 1 A 8 Area of 18× 18 channels has been taken into account High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 24

Timing resolution Point-form Na 22 source, moved in different location for measuring many coincidence pairs not working channel @ 8 V OV, 22°C µ ~ 232 ps source position missing High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 25

Timing resolution Dependance on Si. PM bias voltage FBK 64 Ch 10°C FBK 1 Ch 10°C PETA 36 Ch 34°C PETA 36 Ch 22°C Si. PM 4 x 4 mm 2 LYSO 3 x 3 x 22 mm 3 Measurements with discrete electronics Si. PM 2. 5 x 2. 5 mm 2 LYSO 2. 5 x 10 mm 3 Measurements with LTCC module • at larger bias, non-linearity of the energy integration in PETA degrades the coincidence detection measurements • very close to the limit for RGB-HD technology ( not fully comparable results) High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017

Limit in Timing Resolution Si. PMs produced by FBK in Trento RGB-HD NUV-HD Peak PDE 28% at 400 -650 nm MAX Dark rate 1 MHz/mm 2 Peak PDE 40% at 390 nm MAX Dark rate 100 k. Hz/mm 2 LYSO 2 x 10 mm 3 Assembled on the LTCC modules High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 LYSO 3 x 5 mm 3 Used for measuring limit resolution 27

Limit in Timing Resolution Measurements for testing the lower limit of PETA 5: • • • Modified setup, for using other Si. PM prototypes 2 NUV-HD Si. PMs 4× 4 mm² produced at FBK, Trento, Italy Coupled with LYSO crystals 3× 3× 5 mm³ At the same conditions, FBK reaches ~105 ps CRT with discrete electronics TDC contribution binwidth 50 ps Test setup Signals to PETA via 10 cm cables with external analog pulses PETA reached ~ 80 ps – 110 ps High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Two corrections in the plot: • TDC-non linearity • time-walk correction, depending on energy value 182 ps 28

Crystal Identification Pre-clinical detectors need very high spatial resolution • • • Light spreader guide: signal on many Si. PMs mathematical reconstruction Many channels per event low data rate, low threshold needed Poor timing resolution High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 29

Module with Light Spreader Same LTCC-based module • • • 12× 12 RGB-HD Si. PMs with 2. 5 mm² pitch Glass light spreader – 1 mm thick LYSO array with 0. 8330 mm pitch 36× 36 crystals of 0. 7555× 0. 7555 mm² area ESR reflector between crystals and on lateral and top surfaces Neighbor logic per half chip used for bypassing timing/energy veto mechanism Chip 4 if one channel triggers a hit, all channels in the same neighbor group are read out Chip 3 4 neighbor groups are programmable within each half chip High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Chip 2 Chip 1 30

Module with Light Spreader • • • 36× 36 LYSO crystals 0. 833 mm pitch 12× 12 Si. PMs ~ 3× 3 crystals per Si. PM One chip broken 36× 9 crystals not detected • Neighbor logic per half chip • One crystal missing at the edges • Reconstruction at the edges between two chips works well • Low statistics at the corner, where 3 or 4 neighbor groups are necessary High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 31

Crystal Identification Pre-clinical detectors need very high spatial resolution • • • Light spreader guide to mathematical allow reconstruction Many channels per event low data rate, low threshold needed Poor timing resolution • • High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 No need of light spreader Large sensors Many channels for one sensors Less channels per detector 32

Interpolating Si. PM Intrinsically position-sensitive Si. PM Large active area More output signals per device Prototypes produced at FBK, Trento, Italy Bias voltage 8 mm • • • Signal Bias voltage High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Four output signals 33

Interpolating Si. PM Idea Assign each cell to one of the available “corner„ Local density is chosen following a defined weight function • Center of gravity of any cluster can be reconstructed from corner signals using inverse weight function Bias voltage 8 mm • • 40 x 40 cells 160 x 160 cells High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 Output signals 34

ISi. PM-based Module • • Same LTCC module Rectangular die 15× 5 mm² 2 ISi. PMs per die ONLY 96 channels necessary for the full area read-out • • Test Module 4× 4 ISi. PMs ~30 × 20 mm² active area LYSO array with 0. 8330 mm pitch High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 35

ISi. PM Module: crystal identification Each ISi. PM read out independently always only 4 channels per event Pitch 0. 833 mm Broken ISi. PM One missing wire connection per ISi. Pm Bad alignment in vertical direction High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 36

ISi. PM Module: crystal identification VERY good identification, only 0. 833 mm pitch Peak-to-valley ratio ~ 13 Energy resolution about 17% FWHM in average on an ISi. PM Still room for improvements. . High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 37

Conclusions and Outlook • PETA 5 modules have shown very good performances: o TOF of 230 ps CRT FWHM module vs module in clinical PET o PETA reaches the limit of RGB-HD Si. PM technology & crystal size o sub-millimeter crystal identification in pre-clinical PET • PETA 6 Improvements: o o higher readout rate better charge integration (larger linear range) both Si. PM polarities available additional version with 500 um bump pitch • PETA 6 modules: under test • At the moment, PETA 5/PETA 6 used by 3 groups • Tests in MR scanner for single chips and detection modules High. RR Biweekly Seminar, I. Sacco - 22. 02. 2017 38

Thank you!