Innovative Implantable and Wearable Medical Devices Enabled By

The Pennsylvania State University

The Graduate School

INNOVATIVE IMPLANTABLE AND WEARABLE MEDICAL

DEVICES ENABLED BY ULTRASONIC POWER TRANSFER

AND PIEZOELECTRIC ENERGY HARVESTING

A Dissertation in

Electrical Engineering

Miao Meng

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2019 The dissertation of Miao Meng was reviewed and approved* by the following:

Mehdi Kiani Assistant Professor of Electrical Engineering Dissertation Advisor Chair of Committee

Qiming Zhang Professor of Electrical Engineering

Weihua Guan Assistant Professor of Electrical Engineering

Susan Trolier-McKinstry Professor of Ceramic Science and Engineering

Kultegin Aydin Department Head of Electrical Engineering and Professor

*Signatures are on file in the Graduate School

Abstract

The objective of the research work enclosed in this dissertation is to develop high-performance wireless power and data transfer technologies as well as energy harvesting techniques for implantable and wearable medical devices. The first part of the research work focuses on developing wireless power transmission (WPT) to and communication with millimeter (mm)-sized implantable medical devices (IMDs). Ultrasonic and inductive techniques are developed to achieve high power transfer efficiency (PTE) and low-power pulse-based communication. The second part is to implement an ultrasonic wireless link in a real-world application of ultrasonically interrogated distributed system for gastric slow-wave (SW) recording. The third part is to develop a power management integrated circuit (PMIC) for piezoelectric energy harvesting in next generation self- powered wearables.

Wireless power and data transmission techniques have been proven to be promising solutions for IMDs considering size, weight and lifetime limitations, such as bioelectronic medicines, biosensors, and neural recording/stimulation systems. Ultrasonic links utilizing piezoelectric transducers have shown advantages over other techniques in miniaturizing the IMDs which can greatly reduce the invasiveness and increase the longevity of the IMDs while maintaining high efficiency, especially for applications requiring deep implantation.

Ultrasonic wireless links can be used in many applications. In this dissertation, an ultrasonically interrogated (power/data) distributed system (Gastric Seed) is proposed for large-scale gastric SW recording. Efficient ultrasonic power links and low-power pulse-based data communication are developed. A Gastric Seed chip is developed with emphasis on self-regulated power management and addressable pulse-based data communication. The self-regulated power management can perform rectification, regulation, and over-voltage protection in one step using only one off-chip capacitor which significantly reduces the size of the Gastric Seeds. The addressable pulse-based data communication is proposed and implemented as a proof-of-concept distributed Gastric Seeds. The pulse-based data communication consumes ultra-low power of 440 pJ/bit.

Energy harvesting has become more attractive for self-powered wearables that can enable vigilant health monitoring with 24/7 operation. Piezoelectric energy harvesters (PEHs) can be excited by

iii mechanical vibrations to convert mechanical energy into usable electrical power. PEH is in favor because of high power density and scalability. This outlines the need for an efficient energy- harvesting PMIC to extract maximum energy from PEHs that can be used for self-powered wearables.

This dissertation summarizes the contributions in research areas of ultrasonic power and data communication links and energy harvesting PMIC for PEHs. The contributions include 1) development of the theory and proposing the design methodology to optimize the PTE of ultrasonic links involving mm-sized receivers (Rx), 2) design, development, and validation of a hybrid inductive-ultrasonic WPT link for powering mm-sized implants utilizing two cascaded co- optimized inductive and ultrasonic links for WPT through media involving air/bone and tissue, 3) proposing the concept of self-image-guided ultrasonic (SIG-US) interrogation in a distributed, addressable peripheral nerve recording system to ensure high delivered power regardless of the implant’s movements by automatically tracking the location of the implant in real time, 4) development of a mm-sized Gastric Seed chip towards a distributed recording system for acquiring gastric SWs at a large scale, and 5) development of an autonomous multi-input reconfigurable power-management chip for optimal energy harvesting from weak multi-axial human motion using a multi-beam PEH.

Table of Contents List of Figures ...... viii

List of Tables ...... xiv

List of Abbreviations...... xv

Acknowledgements ...... xviii

Chapter 1 Introduction ...... 1

Wireless Power and Data Transmission to Implantable Medical Devices (IMDs) ...... 1

Ultrasonically Interrogated Gastric Wave Recording System ...... 3

Piezoelectric Energy Harvesting of Multi-Axial Human Motion ...... 3

Contributions ...... 5

1.4.1 Ultrasonic Power Transmission to Millimeter-Sized Implantable Medical Devices ...... 5 1.4.2 Ultrasonically Interrogated Distributed System for Large-Scale Gastric Slow-Wave Recording (Gastric Seed) ...... 6 1.4.3 Piezoelectric Energy Harvesting ...... 6 Chapter 2 Free-Floating Ultrasonically Interrogated (Power/Data) Gastric Recording System .....7

Significance ...... 7

Current Technologies for High-Resolution Monitoring of Slow-Waves ...... 9

Motivation ...... 10

Proposed System with Ultrasonic Free-Floating Distributed Implants ...... 11

Chapter 3 Design and Optimization of Ultrasonic Wireless Power Transmission Links for Millimeter-Sized Implantable Medical Devices ...... 13

Theory of Ultrasonic Wireless Power Transmission ...... 15

3.1.1 Design Parameters for Ultrasonic Transmitter ...... 16 3.1.2 Design Parameters for Ultrasonic Receiver ...... 17 3.1.3 Acoustic Matching ...... 18 Optimal Design of Ultrasonic Wireless Power Transmission Links ...... 18

3.2.1 Ultrasonic Link Design Procedure ...... 19

3.2.2 Ultrasonic Link Design Example for mm-sized Implants ...... 21 Measurement Results ...... 24

Comparison of Series and Parallel Resonance Based on FEM Simulation of Ultrasonic Wireless Power Transmission to Millimeter-Sized Biomedical Implants ...... 29

Ultrasonic Power Link Considering Misalignment ...... 33

Conclusions ...... 36

Chapter 4 A Hybrid Inductive-Ultrasonic Link for Wireless Power Transmission to Millimeter- Sized Biomedical Implants ...... 38

Hybrid Inductive-Ultrasonic WPT Link ...... 39

4.1.1 Design and Optimization of the Ultrasonic WPT Link ...... 39 4.1.2 Design and Optimization of the Inductive Link...... 40 4.1.3 Hybrid Link Design Example ...... 41 Simulation and Measurement Results ...... 43

Conclusions ...... 46

Chapter 5 Self-Image-Guided Ultrasonic Power Transmission to mm-Sized Implantable Devices ...... 47

Self-Image-Guided Ultrasonic Interrogation Concept ...... 47

Simulation Setup and Results ...... 49

Conclusions ...... 53

Chapter 6 Gastric Seed: Towards Distributed Ultrasonically Interrogated Millimeter-Sized Implants for Large Scale Gastric Electrical-Wave Recording ...... 54

Gastric Seed Chip Architecture ...... 54

6.1.1 Self-Regulated Power Management ...... 55 6.1.2 Addressable Pulse-Based Data Communication ...... 56 Measurement Results ...... 56

Conclusions ...... 61

Chapter 7 Ultrasonically Powered Wireless System for In Vivo Gastric Slow-Wave Recording . 62

System Architecture ...... 62

7.1.1 Gastric Seed Chip Prototype ...... 63 7.1.2 Commercial-Off-The-Shelf (COTS) Components ...... 64 Benchtop and In Vivo Experiment Results ...... 65

Conclusions ...... 68

Chapter 8 Multi-Beam Shared-Inductor Reconfigurable Voltage/SECE-Mode Piezoelectric Energy Harvesting of Multi-Axial Human Motion ...... 70

Wrist-Worn Multi-Beam Inertial Piezoelectric Energy Harvester (PEH) ...... 72

Operation and Modeling of the VM-SECE Scheme ...... 73

Multi-Beam Reconfigurable VM-SECE Chip Architecture ...... 77

Measurement Results ...... 81

8.4.1 Benchtop Measurements ...... 81 8.4.2 Measurements with a Commercial Single-Beam PEH ...... 83 8.4.3 Measurements with the Custom Multi-Beam PEH ...... 84 Conclusions ...... 88

Chapter 9 Conclusions and Future Work ...... 89

Summary of Results and Contributions ...... 89

9.1.1 Ultrasonic Power Transmission to mm-Sized Implantable Devices ...... 89 9.1.2 Ultrasonically Interrogated Distributed System for Large-Scale Gastric Slow-Wave Recording ...... 91 9.1.3 Piezoelectric Energy Harvesting ...... 91 Future Work ...... 92

References ...... 94

vii

List of Figures

Fig. 1.1 Medtronic’s pacemaker implanted under the skin under collarbone, hooked up to heart with tiny leads...... 1 Fig. 1.2 Cantilever based PEH converting mechanical vibration to electrical energy...... 4 Fig. 2.1 High Resolution Mapping [27] (a) Flexible PCB array (16 × 16 electrodes at 4 mm spacing). (b) The array placed on the corpus-antrum border...... 8 Fig. 2.2 (a) State-of-the-art wireless system for SW recording. (b) The 64-channel ASIC [33]. (c) Assembled and packaged implant for endoscopic implantation. (d) Wireless recorded SWs and a normal antegrade pattern of SW propagation in terms of an isochronal map...... 9 Fig. 2.3 Conceptual diagram of the proposed Gastric Seed technology [48]...... 12 Fig. 3.1 Generic ultrasonically-powered mm-sized biomedical implant structure with emphasis on

WPT via a pair of ultrasonic transducers (U1 and U2)...... 13 Fig. 3.2 (a) The characteristics of a sound beam generated by an unfocused disc-shaped transducer [106]. (b) Four design examples for the choice of optimal location and diameter of the Rx transducer to achieve the highest PTE...... 15 Fig. 3.3 The ultrasonic WPT link model in COMSOL to optimize the geometries of Tx and Rx transducers (U1 and U2)...... 19 Fig. 3.4 Iterative ultrasonic link optimization flowchart for efficient WPT to mm-sized biomedical implants...... 20

Fig. 3.5 Optimization of U1 and U2 geometries with matching layer at fc = 1.8 MHz to maximize

PTE by sweeping (a) t2 and Do2 for Do1 = 10.8 mm and t1 = 1.05 mm, and (b) t1 and Do1 for

Do2 = 1.2 mm and t1 =0.25 mm...... 23 Fig. 3.6 Comparison between simulated results of PTE for the links with and without matching

layer for RL = 2.5 kΩ. (a) PTE vs. fc for optimized links at each fc based on the design procedure in Fig. 3.4. (b) PTE vs. d for the optimal links at 1.4 MHz and 1.8 MH ...... 23 Fig. 3.7 (a) PTE measurement setup for the ultrasonic WPT link using a network analyzer that can accurately measure the S-parameters. (b) Ultrasonic WPT link measurement setup inside a castor oil tank...... 25

viii

Fig. 3.8 Comparison between simulated and measured real and imaginary values for the impedance

of (a) Tx transducer (U1), and (b) Rx transducer (U2), which their geometries have been stated in Table 3.1...... 26 Fig. 3.9 Simulated and measured values of PTE vs. (a) operation frequency (fc), and (b) powering

distance (d) for RL of 2.5 kΩ...... 27

Fig. 3.10 Simulated and measured values of PTE vs. RL at d = 3 cm...... 28 Fig. 3.11 The measured values of PTE at d = 3 cm vs. lateral misalignment in Y-direction as shown in the measurement setup in Fig. 3.7b...... 28 Fig. 3.12 Impedance profiles with Tx and Rx geometries listed in Table 3.3, resonant frequencies

are matched to fp of 1.4 MHz for (a) fs of Tx, (b) fs of Rx, (c) fa of Tx, and (d) fa of Rx. . 31 Fig. 3.13 PTE vs. d for all four resonance combinations with optimal load...... 32

Fig. 3.14 PTE vs. RL for all four resonance combinations at 3 cm...... 32

Fig. 3.15 PTE vs. fp for all four resonance combinations with optimal load at 3 cm...... 33 Fig. 3.16 Simulated 3D PTE surfaces of the ultrasonic link for 1.1 mm3 Rx at d = 10 mm vs. (a)

different t1 and Dou1 values in perfectly-aligned condition, and (b) Rx lateral misalignment

in Y-direction and Dou1...... 35 Fig. 3.17 Comparison of the measured values of PTE at d = 3 cm vs. lateral misalignment in Y-

direction as shown in the measurement setup in Fig. 3.7b. for two different U1s...... 36 Fig. 4.1 Hybrid inductively-ultrasonically-powered mm-sized implant structure with WPT via a

pair of coils (L2 and L3) and ultrasonic transducers (U1 and U2) for applications that involve mediums with different acoustic impedances...... 38 Fig. 4.2 Simulation Setup (a) The ultrasonic WPT link model in COMSOL to find the optimal

geometries for U1 and U2, i.e., Dou1,2, t1,2, as well as fp based on the design procedure in [42]. The soft tissue is mimicked by castor oil with similar acoustic attenuation of 0.8 dB/cm/MHz. A perfect matching layer (PML) has been considered at the boundaries of the medium to avoid acoustic reflections. (b) The 2-coil inductive link model in HFSS to find

the optimal geometries of L2 and L3, i.e., Do2,3, w2,3, n2,3, s2,3. The silver electrodes of U1 have been modeled with two parallel silver plates in HFSS...... 40

Fig. 4.3 Three different 2-coil structures in HFSS considering the location of L3 and U1 in the hybrid link. Two parallel silver plates have been considered in these strcutures to mimic

the silver electrodes of U1. (a) Structure-1: U1 is located inside L3 at the side, (b) Structure-

2: U1 is located at the center of L3, and (c) Structure-3: U1 is located next to L3...... 42 Fig. 4.4 Measurement setup (a) PTE measurement for the inductive, ultrasonic, and hybrid links using a network analyzer that can accurately measure the S-parameters. (b) Hybrid inductive-ultrasonic WPT link measurement setup, where inductive and ultrasonic links

were in air and castor oil tank, respectively. L2 and U2 were connected to a pair of SMA connectors to measure the S-parameters...... 44

Fig. 4.5 Comparison between simulated and measured PTEind vs. dind for inductive link only in air...... 45

Fig. 4.6 Comparison for fixed dind of 3 cm, (a) simulated and measured values of PTEhybrid vs.

powering distance (d = dind + dus) for the hybrid inductive-ultrasonic WPT link at fp of 1.1

MHz for RL of 2.5 kΩ, and (b) measured PTEUS vs. d for the ultrasonic link only with air- tissue medium. Coils and transducer geometries are listed in Tables 4.1 (structure-3) and 4.2, respectively...... 45 Fig. 5.1 Conceptual diagram of the proposed Self-Image-Guided Ultrasonic (SIG-US) interrogation platform used in a distributed, addressable Peripheral Nervous System (PNS)...... 48 Fig. 5.2 The ultrasonic WPT link model in COMSOL of a linear array of 11 ultrasonic transmitter (Tx) array (each 1 mm3) operating at 1 MHz to power a 1 mm3 implant located 30 mm away. A castor oil medium is considered to mimic the soft tissue, as it has similar acoustic properties...... 49 Fig. 5.3 (a) The pulse sent by Rx (in Fig. 5.2) with optimal pulse width of 450 ns for a 1 MHz ultrasonic link. (b) The received pulses by Tx transducers with relative delays and different amplitudes depends on the location of each Tx transducer...... 50 Fig. 5.4 The relative delays for each Tx transducers in Tx array...... 51 Fig. 5.5 (a) the sinusoidal signals sent by Rx with 5 V peak voltage at 1 MHz. (b) The sinusoidal signals received by the Tx transducers in the Tx array with relative delays similar to pulse excitation in Fig. 5.3...... 51 Fig. 5.6 Comparison of the voltages received by Rx with SIG-US to the voltages received by Rx without SIG-US under the same condition...... 52

Fig. 5.7 (a) Comparison of the received Rx voltages with SIG-US to the received Rx voltages with conventional beamforming with misalignment of 3 mm. (b) The improvement of normalized received power by SIG-US over conventional beamforming for misalignment of up to 6 mm...... 53 Fig. 6.1 Simplified diagram of the proposed Gastric Seed chip interfacing with a pair of stacked power/data ultrasonic transducers...... 54 Fig. 6.2 Switch controller schematic diagram and key waveforms of the self-regulated mechanism

using reverse current from VL to power Rx...... 55 Fig. 6.3 Schematic diagram and key operational waveforms of the proof-of-concept addressable data transmitter with IDs “0” and “1”...... 57 Fig. 6.4 Gastric Seed chip micrograph occupying 0.6 mm2 including pads...... 58 Fig. 6.5 Measurement setup with ultrasonic transducers perfectly aligned with 3D printed plastic holders...... 58

Fig. 6.6 (a) Measured PCE and VCE of the power management at different RL,DC. (b) PTE of the

ultrasonic link at different RL,AC...... 59

Fig. 6.7 Measured VL and VP waveforms during start-up when external VS (power Tx voltage) was

increased from zero to 8.5 Vp-p...... 59

Fig. 6.8 (a) Measured VL and VP waveforms when VS was increased from 11 VP-P to 18 VP-P. (b)

Zoomed waveforms of VL and VP (CL = 10 µF)...... 60 Fig. 6.9 Measured waveforms for the pulse-based data transmitter showing how addressable Gastric Seeds (IDs “0” and “1”) can be activated...... 61 Fig. 6.10 Measured transmitted and recovered data bits at 75 kbps...... 61 Fig. 7.1 Simplified diagram of the recording prototype interfacing with a pair of stacked power/data ultrasonic transducers...... 62 Fig. 7.2 Schematic diagram and key operational waveforms of the addressable data transmitter with ID = “1” enabled. All signals are generated inside the chip based on a 256 kHz clock from the Tx MCU...... 64 Fig. 7.3 Measured waveforms for the pulse-based data transmitter showing transmitted and recovered data bits at 16 kbps when Gastric Seed with ID “1” was activated...... 65 Fig. 7.4 In vivo experiment setup for wireless recording of gastric SWs. Wireless power and data transfer were established with ultrasound and RF links, respectively...... 66

Fig. 7.5 (a) Unfiltered direct oscilloscope probe measurement of the amplified SW signal at the amplifier output with sampling rate of 50 samples/s. (b) Filtered oscilloscope probe measured SW signal with 500 points averaging showing frequency of 1.375 cpm...... 67 Fig. 7.6 (a) Unfiltered SW signals recorded through RF link with sampling rate of 200 samples/s. (b) Filtered RF recovered SW signal with 2000 points averaging showing frequency of 1.5 cpm...... 68 Fig. 7.7 (a) Filtered oscilloscope probe measured SW signal with 500 points averaging showing frequency of 3.75 cpm. (b) Filtered RF recovered SW signal with 2000 points averaging showing frequency of 4 cpm...... 69 Fig. 8.1 Generic block diagram of a piezoelectric energy harvesting system. A power-management circuit is used to convert the AC voltage across the PEH to a usable DC voltage across a

storage capacitor CSTORE. For simplicity, the sensor node has been modeled with a DC load

RL in this paper...... 70 Fig. 8.2 Wrist-worn multi-beam piezoelectric energy harvesters (PEHs) for harvesting energy from body motion with multi-axial movements. (a) The magnetic-plucking prototype used inthe previous work in [59]. (b) The mechanical-plucking prototype with more robustness used in this chapter...... 71 Fig. 8.3 Simplified circuit schematic of the proposed reconfigurable VM-SECE power- management scheme for the single-beam operation...... 73 Fig. 8.4 Key operational waveforms of the proposed VM-SECE scheme. (a) SECE only when peak

VREC < VSTORE. (b) Reconfigurable VM-SECE when peak VREC > VSTORE...... 74

Fig. 8.5 (a) Simulated and calculated PSTORE vs. VSTORE for the FAR, conventional SECE and the

proposed VM-SECE. (b) Simulated peak VREC vs. VSTORE for the proposed VM-SECE and conventional SECE. Ideal lossless components were used in both calculations and circuit simulation...... 76

Fig. 8.6 Simulated comparison (a) PSTORE and (b) peak VREC vs. VSTORE for the proposed VM-SECE scheme and the conventional SECE with lossy circuit components listed in Table 8.1. The circuit schematic is shown in Fig. 8.3...... 78 Fig. 8.7 Block diagram of the proposed multi-beam shared-inductor reconfigurable VM-SECE chip with cold start...... 78

xii

Fig. 8.8 Key operational waveforms of the proposed VM-SECE chip for (a) asynchronous and (b) synchronous inputs (beams 1 and 6 in this example)...... 79 Fig. 8.9 Detailed schematic diagrams of the (a) NVC, (b) multi-beam sweep control, (c) peak and time zero-crossing detection, and (d) switch control...... 80 Fig. 8.10 The proposed VM-SECE chip micrograph occupying 1.9 mm2 of active area, and its key building blocks...... 81 Fig. 8.11 Measured key operational waveforms showing the VM-SECE chip operation in different modes with one beam...... 82 Fig. 8.12 Benchtop measurement results of the chip with the multi-beam PEH. (a) Measured

transient waveforms with cold start, (b) and (c) zoomed waveforms for VCHIP < 3 V and

VCHIP > 3 V, respectively, and (d) VREC waveforms across 4 beams...... 83 Fig. 8.13 Measurement setup for characterizing the chip’s performance with a commercial single- beam PEH (PPA1011) on a shaker...... 84

Fig. 8.14 Comparison of measured PSTORE vs. VSTORE between the VM-SECE chip and the on-chip 95.6%-efficient FAR using the commercial PPA1011 PEH (single beam) with the shock acceleration of 4.39 g at 1 Hz...... 85

Fig. 8.15 Measured FoM and VP,OC of the VM-SECE chip under various shock accelerations. A minimum FoM of 200% was achieved...... 85 Fig. 8.16 Measurement setup with the integrated VM-SECE chip and 5-beam mechanically plucked PEH (Fig. 8.2b) mounted on a robotic swing arm...... 86 Fig. 8.17 (a) Measured transient voltage waveforms of the VM-SECE chip interfacing the 5-beam mechanically plucked PEH in Fig. 8.2b on the robotic swing arm. (b) Zoomed waveforms showing voltage variability across beams...... 86

Fig. 8.18 Comparison of measured PSTORE vs. VSTORE between the VM-SECE chip and the on-chip FAR using the custom PEH in Fig. 8.2b on the robotic swing arm. (a) One-beam operation. (b) Five-beam operation...... 87

xiii

List of Tables

Table 3.1 Optimized Ultrasonic Transducer Geometries and Operation Frequency for Design Examples and Measurement ...... 22 Table 3.2 Benchmarking of Recent Ultrasonic WPT Links for mm-Sized Biomedical Implants 29 Table 3.3 Ultrasonic Transducer Specifications ...... 30 Table 3.4 Resonance Comparison ...... 30

Table 4.1 PTE Comparison Between Different L3-U1 Structures in Fig. 4.3 ...... 43 Table 4.2 Benchmarking of Recent WPT Links for mm-Sized Biomedical Implants ...... 46 Table 8.1 Circuit Parameters Used for VM-SECE and SECE Simulation ...... 77 Table 8.2 Benchmarking the Proposed 6-Beam Reconfigurable VM-SECE Chip Among the State- of-the-Art PEH Chips ...... 88

xiv

List of Abbreviations

Abbreviations Definition

ADC Analog to Digital Converter

AFE Analog Front-End

ASIC Application-Specific Integrated Circuit

COTS Commercial Off The Shelf

GI Current Gastrointestinal

EGG Electrogastrography

FAR Full-Wave Active Rectifier

FEM Finite Element Method

FoM Figure of Merit

FD Functional Dyspepsia

GES Gastric Electrical Stimulation

GP Gastroparesis

IMD Implantable Medical Device

LNA Low-Noise Amplifier

MCU Microcontroller

MPPT Maximum Power Point Tracking

NVC Negative Voltage Converter

OVP Over Voltage Protection

PCB Printed Circuit Board

PCE Power Conversion Efficiency

PDL Power Delivered to Load

PE Planar Expander

PEH Piezoelectric Energy Harvester

PKD Peak Detection

PMIC Power Management Integrated Circuit

PML Perfect Matching Layer

PNS Peripheral Nervous System

PTE Power Transfer Efficiency

PZT Lead Zirconium Titanate

RF Radio Frequency

Rx Receiver

SECE Synchronous Electrical Charge Extraction

SIG-US Self-Image-Guided Ultrasonic

SW Slow Wave

SSHI Synchronized Switch Harvesting on Inductor

SU Stationary Unit

TE Thickness Extensional

Tx Transmitter

US Ultrasonic

VM Voltage Mode

WPT Wireless Power Transmission

xvi

WSN Wireless Sensor Network

WU Wearable Unit

ZCD Zero Crossing Detection

xvii

Acknowledgements

Foremost, I would like to express my sincerest gratitude to my advisor, Dr. Mehdi Kiani for his guidance, patience, and encouragement, for being a mentor to me and for influencing me beyond research. I am thankful to him for letting me concentrate on research without worrying about funding. His dedication to research and teaching has greatly inspired me and helped me shape my professional career.

I would like to thank all my committee members, Dr. Susan Trolier-McKinstry, Dr. Weihua Guan and Dr. Qiming Zhang for sharing their insight and knowledge and providing comments and advice to improve my thesis.

I would like to thank all my collaborators: Dr. Susan Trolier-McKinstry, Dr. Hong Goo Yeo, and Dixiong Wang at Pennsylvania State University for designing and fabricating the high performance thin-film piezoelectric beams; Dr. Shad Roundy, Dr. Tiancheng Xue, and Binh Duc Truong at University of Utah for designing the rotor-based inertial harvester, integrating with the piezoelectric beams and setting up the robotic swing arm measurement; Dr. Aydin Farajidavar, Dr. Amir Javan Khoshkholgh, and Dr. Raddy Ramos at New York Institute of Technology for preparing the animal experiments; Dr. Shashank Priya, Dr. Rammohan Sriramdas, Dr. Christopher D. Rahn, and Dr. Tahzib Safwat at the Pennsylvania State University for offering help with some equipment. I appreciate the technical discussions we had. I could not complete my work without their efforts and help.

I would like to thank all ICSL members, Ahmed Ibrahim, Hesam Sadeghi Gougheri, Philip Graybill, Zeinab Kashani, and Sujay Houser for their discussions and generous help. It was a truly pleasure experience working with them.

This work was supported in part by the National Institutes of Health (NIH) under Grant NIBIBU18EB021789 and National Science Foundation (NSF) Nano-systems Engineering Research Center (NERC) for Advanced Self-Powered Systems of Integrated Sensors and Technologies (ASSIST), EEC-1160483.

xviii

Last, I would like to express my deepest gratitude to my parents and my wife, Sinuo Lu, for being so supportive in every way. I would not have completed this journey without their love and support. My son, Lucas, is my greatest award I could hope for during my PhD. I love you all!

xix

Chapter 1 Introduction

Wireless Power and Data Transmission to Implantable Medical Devices (IMDs)

Recently, recording and modulation of neural activities in the peripheral nervous system (PNS) has received great attention for its potential to replace prevalent pharmacological approaches with implantable electroceuticals [1], known as bioelectronics medicine. Recording and modulation of neural signals in visceral organs is required to implement the bioelectronics medicine [2]. Fig. 1.1 shows a Medtronic’s implanted pacemaker to help patient regulate the beating of the heart. Some other examples include vagus nerve stimulation for the treatment of epilepsy, inflammatory bowel disease, rheumatoid arthritis, and diabetes, as well as bladder and gastric stimulation [3]-[6]. However, most implantable devices suffer from shortcomings including bulky batteries and high volume of interconnect wirings which can potentially cause long-term tissue damage and biological responses, resulting in implantation failure over time [7]-[9]. Therefore, for future bioelectronics-medicine devices, it is key to develop efficient links for wireless power transmission to and reliable bidirectional communication with deeply implanted (several centimeters), as well as robust high-channel-count recorders and stimulators in the PNS. These battery-less devices should be miniaturized to millimeter (mm) or sub-mm scale so that they can be implanted into the PNS with minimal invasiveness and operate without frequent maintenance.

Fig. 1.1 Medtronic’s pacemaker implanted under the skin under collarbone, hooked up to heart with tiny leads.

Wireless power transmission (WPT) to biomedical implants can eliminate the need for bulky batteries, increase their longevity, and reduce their size and risks [11]-[13]. During the past few decades, inductive links have been the most attractive and efficient method for WPT and communicate to biomedical implants [14]. This technique can offer high power transmission

1 efficiency (PTE), particularly for the implant size as centimeter (cm) and sub-cm dimensions [15], [16]. By miniaturizing biomedical implants to millimeter (mm) or sub-mm dimensions, minimally- invasive biosensing and localized operations such as multisite neural recording, stimulation, and even optogenetics can be achieved with power requirements from 100’s of μW to several mWs [17], [18]. It has also been demonstrated that mm- and sub-mm-sized implants can minimize the tissue damage and increase the safety and longevity of the neural interface [7]. However, wireless power transmission to mm-sized IMDs is more challenging due to the small size of the receiver (Rx) coil. The conventional inductive power transmission at the frequency range of MHz or lower is not suitable for powering mm-sized IMDs due to its low PTE limited by the small size of the Rx coil.

For RF-based WPT to mm-sized implants, it has been suggested that by increasing the operation 3 frequency (fc) of the link to 100’s of MHz, PTE can be improved to ~0.5% for a 1 mm receiver (Rx) coil/antenna at powering distance (d) of ~10 mm [19]-[21]. Recently, an optimal design 3 procedure for inductive links is demonstrated at fc = 20 MHz for a 1 mm Rx coil [22]. This work has reported PTE and power delivered to the load (PDL) of 1.4% and 2.2 mW at d = 10 mm, respectively [22]. Although these results are promising, maximum achievable PDL within safety constraints is often limited by specific absorption rate (SAR), which is significantly increased by fc, in RF- or inductive-based WPT links [23].

For efficient WPT to mm-sized implants, ultrasound modality has recently gained great attention because 1) the acoustic loss in tissue is much smaller than that of electromagnetics, 2) low acoustic velocity allows operation at lower frequencies, and 3) allowable ultrasonic intensity for safe operation in tissue is much higher than that of electromagnetics [24]. The advantages of ultrasound suggest that potentially much higher PTE and PDL can be achieved via an ultrasonic link, especially for deeply implanted mm-sized IMDs. In addition, efficient ultrasonic data transmission link can also be established.

However, the literature lacks an application-oriented design guideline for optimal design of efficient ultrasonic WPT links involving mm-sized implants. Moreover, several challenges related to ultrasonic WPT link, such as extremely high attenuation loss in air or bone media and susceptibility to misalignment and orientation needs to be considered and addressed.

Ultrasonically Interrogated Gastric Wave Recording System

Gastric motility disorders, notably gastroparesis (GP) and functional dyspepsia (FD), are affecting a large population in the US [25]-[28]. The diagnoses of FD and GP are challenging, often requiring multiple clinical visits and investigations including radioisotope scans, endoscopy, etc. The diagnostic strategies represent a process of exclusion rather than analyzing fundamental mechanisms and can cause vast associated economic burdens.

Gastric contractions are initiated and coordinated by an underlying bioelectrical activity, termed slow waves (SWs) [29] which have been associated with the gastric dysmotility. Current technologies for acquiring gastric contractions include electrogastrography (EGG) and high- resolution EGG (HR-EGG). While these technologies are non-invasive, they have limited accuracy and are susceptible to motion artifacts. High-resolution mapping approaches provide accurate analysis, but they are highly invasive and cannot be applied in chronic studies on awake animals. Current state-of-art technology in [30]-[33] employs implants with inductive wireless powering and data communication. However, such system is large in size and has limited spatial coverage of stomach.

Therefore, the literature needs a system that employs miniaturized distributed implants to record SWs in high resolution with minimal invasiveness. The system should be able to record SWs at a large scale (whole stomach) with minimal motion artifacts.

An ultrasonically interrogated distributed system is a good candidate for SW recording because 1) it enables the miniaturization of the implant to mm- or sub-mm scales while maintaining highly efficient power transmission; this can reduce the invasiveness and enable deep implantation, and 2) the system with distributed implants can provide large-scale recording capability. The concept and implementation of such system will be discussed more in detail in Chapter 2.

Piezoelectric Energy Harvesting of Multi-Axial Human Motion

Energy harvesting has become more attractive over the past few years, because it promises either to replace or complement available batteries in wireless sensor networks (WSNs) [34]. A typical WSN sensor is often required to perform multiple functions, such as sensing, processing, and data reception and transmission, with a limited power source. The past few years have also witnessed

3 a growing demand for self-powered wearables that can enable vigilant health monitoring with 24/7 operation [35]-[37]. In all these applications, batteries are often undesirable due to their limited lifetime and cost, as well as their bulkiness.

For powering wearables, different modalities have been proposed and employed to harvest energy from the human body [38]-[40], such as motion (mechanical) [38], heat (thermal) [39], and biochemical energy (biofuel) [40]. Among these, harvesting from body motion has the potential to extract the highest available power [41]. Electromagnetic, electrostatic, and piezoelectric are three conventional methods for converting mechanical energy into usable electrical power [42]. The piezoelectric energy harvesters (PEHs) are more attractive due to their higher power density and scalability [43]-[45].

One of the most common method for PEH is to use a cantilever beam, as shown in Fig. 1.2. An applied input force on the tip will cause vibration of the beams, then the mechanical vibration is converted into electrical energy through a piezoelectric material. Although several unidirectional PEHs in the form of a single cantilever beam have been proposed and developed in the past [43]- [49], they suffer from several challenges in energy harvesting from body motion, including the presence of multi-axial motion, irregular frequencies, and unpredictable voltage levels with often low amplitudes [41]. Therefore, a new PEH needs to be developed to address those challenges. At the same time, a novel power management circuit interfacing the harvester is needed to efficiently convert the AC signal generated by the PEH into available DC power.

Fig. 1.2 Cantilever based PEH converting mechanical vibration to electrical energy.

The structure of the dissertation is as follows: In chapter 2, a system with distributed ultrasonically interrogated (power/data) mm-sized implants (Gastric Seeds) for large-scale gastric slow wave (SW)Fig. 1.2. recording Cantilever is basedproposed. PEH converting In chapter mechanical 3, the design vibration and to optimization electrical energy. of ultrasonic wireless power

4 transmission links for millimeter-sized biomedical implantable devices is presented. The concept, design and optimization of a hybrid inductive-ultrasonic link which allowed increased efficiency when air/bone-tissue medium involved, such as freely behaving animal experiment are shown in chapter 4. Chapter 5 proposes the concept of self-image-guided ultrasonic (SIG-US) interrogation in a distributed, addressable peripheral nerve recording system to ensure high delivered power regardless of the micro-motion of the implants by automatically tracking the location of implant in real time. Chapter 6 presents the design, benchtop and in vivo test results of the proof-of-concept Gastric Seeds chip with novel features of self-regulated power management and addressable pulse- based ultrasonic data communication. Chapter 7 describes the measurement and in vivo results of a gastric SW recording system utilizing the Gastric Seeds chip and commercial-off-the-shelf (COTS) components. Chapter 8 shows the multi-beam shared-inductor reconfigurable voltage/SECE-Mode (VM-SECE) piezoelectric energy harvesting of multi-axial human motion. Finally, the contributions of this dissertation and suggested future works are summarized in Chapter 9.

Contributions

1.4.1 Ultrasonic Power Transmission to Millimeter-Sized Implantable Medical Devices

A design methodology was proposed to maximize the PTE of ultrasonic links for WPT to mm- sized biomedical implants by co-optimizing the geometries of Tx and Rx transducers such as diameter and thickness, as well as the optimal operation frequency. Designers can follow the proposed design procedure to optimize the ultrasonic powering link depending on their application need. In addition, since the attenuation loss is high in air-bone medium, a hybrid inductive- ultrasonic WPT was proposed to achieve high PTE for powering mm-sized biomedical implants in applications that involve multiple mediums with different acoustic impedances (air, bone, tissue), such as neural interfacing with freely-behaving rodents inside a cage. An optimization procedure was also proposed for the hybrid inductive-ultrasonic WPT link. Moreover, to solve the issue of high sensitivity to misalignment for ultrasonic powering link, the SIG-US beamforming technique was proposed which can be used in any distributed, addressable peripheral nerve system (PNS) recording and stimulation, the location of the PNS nodes can be automatically tracked in real time to ensure high delivered power regardless of the micro-motion of the implanted nodes.

On this topic, three journal papers ([50]-[52]) and four conference papers ([53]-[56]) have been published.

1.4.2 Ultrasonically Interrogated Distributed System for Large-Scale Gastric Slow-Wave Recording (Gastric Seed)

A new paradigm for large-scale gastric interfacing with ultrasonic interrogation (Gastric Seed) with minimal invasiveness was proposed to record gastric slow-waves (SWs) which have been associated with gastric dysmotility in several significant gastric motility disorders. A proof-of- concept mm-sized Gastric Seed chip with novel features of ultrasonic self-regulated power management and pulse-based addressable data communication has been designed and fabricated. To add the recording feature, the chip was integrated with commercial-off-the-shelf (COTS) components, including an analog front-end amplifier, an analog-to-digital converter (ADC) and RF transmission links. This recording prototype was successfully tested both in benchtop settings and in in vivo experiments on anesthetized rats. One journal paper ([57]) and one conference paper ([58]) have been published.

1.4.3 Piezoelectric Energy Harvesting

A fully autonomous multi-beam reconfigurable voltage/synchronous- electrical-charge extraction- mode (VM-SECE) chip with only one shared off-chip inductor for piezoelectric energy harvesting was proposed and implemented. The VM-SECE chip successfully interfaced with a multi-beam wrist-worn inertial PEH with custom thin film beams to harvest from multi-axial body motion. The VM-SECE chip could harvest from up to 6 beams simultaneously in a modular fashion with improved extracted power regardless of the beam and frequency variations. The VM-SECE chip also has an inherent over-voltage-protection (OVP) to avoid using high-voltage complimentary- metal-oxide-semiconductor (CMOS) process which greatly reduces the design complexity and costs. The VM-SECE chip holds the promise of integrated self-powered solution for the next generation of wearables with vigilant operation capability. Regarding this work, one journal paper ([59]) has been submitted recently and one conference paper ([60]) has already been published.

Chapter 2 Free-Floating Ultrasonically Interrogated (Power/Data) Gastric Recording System

Ultrasonic wireless power transmission (WPT) links enable miniaturization of implantable devices (minimal invasiveness) and deep implantation which can potentially open up many application venues. A real-world application for the ultrasonic interrogation is a free-floating ultrasonically interrogated gastric recording system, which is proposed in this chapter. The research work has been done and the future work toward demonstrating the system will be discussed in detail in later chapters.

Significance

Gastric contractions are initiated and coordinated by an underlying bioelectrical activity, termed slow waves (SWs) [29]. Aberrant SW patterns (dysrhythmias) have been associated with gastric dysmotility in several significant gastric motility disorders, notably gastroparesis (GP) and functional dyspepsia (FD) [25]-[28]. GP is a chronic debilitating disease, with symptoms including continuous nausea and vomiting, affecting more than 1.5 million people in the US. Up to 30-50% and 16-30% of patients with type 1 and 2 diabetes suffer from GP or related symptoms, respectively. This number is increasing due to the diabetes epidemic, with hospitalizations rising 150% in the last decade with major associated costs [61], [62]. FD is a chronic disorder characterized by upper abdominal pain, bloating and early satiety. FD is highly prevalent, affecting up to 15-30% of the US population [63], with vast associated economic burden [64]. FD is also associated with gastric emptying impairments in up to 20-40% of FD patients, and among several putative mechanisms, dysrhythmic SW activity has been implicated in some 33-83% of FD patients [25], [65]. The diagnoses of FD and GP are challenging, often requiring multiple clinical visits and investigations including radioisotope scans, endoscopy, contrast studies, ultrasound and SmartPill use [66], [67]. However, these diagnostic strategies represent a process of exclusion, rather than analyzing fundamental mechanisms, and none allow real-time continuous monitoring of SW dysrhythmias that may be directly contributing to pathophysiology and symptoms.

High-resolution mapping technologies (Fig. 2.1 [27]) have become a fundamental tool for accurately defining electrophysiological properties in multiple fields, including electromyography and electrocardiography [68], [69]. It has also recently been advocated as a tool to reveal the mechanisms underlying gastric dysrhythmias, which have been poorly understood using past techniques [70], [71]. Prominent studies have recently revealed a surprising complexity underlying gastric dysrhythmias, comparable to cardiac dysrhythmias, and including complex focal activities and re-entrant patterns [72], [73]. Translating these advances to the clinic is now a research priority in this field, to clarify the role of dysrhythmias in conditions such as GP [74], [75]and FD [76]. Translational advances in high-resolution mapping are also anticipated to guide progress in therapy, which is critically needed because current therapies for GP and FD have poor efficacy [77].

(a) (b) Fig. 2.1 High Resolution Mapping [27] (a) Flexible PCB array (16 × 16 electrodes at 4 mm spacing). (b) The array placed on the corpus-antrum border. Despite 100 years of intensive investigations, the clinical significance of gastric dysrhythmia remains poorly understood. Current gastrointestinal (GI) therapies, including gastric electrical stimulation (GES) and various pharmaceuticals, are delivered and adjusted through trial and error in an open-loop fashion. Consequently, conflicting therapeutic effects have been reported by various researchers and the role of some of these therapies remains controversial [5]. As a result, novel closed-loop methods are required to enhance the therapies. For instance, synchronized closed-loop GES has been applied on anesthetized canine models to modulate gastric contractions and enhance gastric motility [6]. Real-time acquisition of SW data can be used as the feedback signal to develop a closed-loop therapy [78].

Current Technologies for High-Resolution Monitoring of Slow-Waves

To better elucidate the pathophysiological significance of gut electrical abnormalities, chronic studies in conscious subjects, in fasting and fed states, are necessary [10], [79]. Importantly, SW recordings taken directly from the stomach currently provide the only reliable and spatially descriptive data.

Cutaneous recordings also called electrogastrography (EGG) and high-resolution EGG (HR- EGG), while non-invasive, have limited accuracy and are highly prone to motion artifacts [79]- [81]. High-resolution mapping has been a critical recent advance, allowing a new era in accurate analysis of dysrhythmic onset and maintenance [70], [71], [78]. However, current high-resolution mapping approaches are highly invasive and, therefore, cannot be applied in chronic animal studies or patients. Existing serosal or mucosal recording systems currently transmit signals through wires traversing either through the abdominal wall or a natural orifice [70], [82], [83]. These wires can act as a conduit for infection, induce discomfort, become displaced, and restrict mobility due to their connection to bulky acquisition systems [79]; hence, the monitoring period is often limited to the anaesthetized state.

(a) (b) (c) (d) Fig. 2.2 (a) State-of-the-art wireless system for SW recording. (b) The 64-channel ASIC [33]. (c) Assembled and packaged implant for endoscopic implantation. (d) Wireless recorded SWs and a normal antegrade pattern of SW propagation in terms of an isochronal map.

The state-of-art technology is to use wireless and implantable technologies to acquire SWs from the GI tract [30]-[33]. The system is composed of 1) a wireless 64-channel implant (size: 30×10×6.7≈2000 mm3) that is placed in the stomach submucosa space through an endoscopic procedure, 2) a wearable unit that can power the implant through inductive coupling and receive SW signals through inductive backscattering, and 3) a stationery unit that is connected to a computer and can display SWs in real time (Fig. 2.2a). The implantable CMOS application-

9 specific integrated circuit (ASIC) (size: 25 mm2) integrates 64 low-noise amplifiers (LNAs) and filters, an 8-bit analog-to-digital-converter (ADC), a power management, and a data transmitter (Fig. 2.2b) [32], and connects to an array of flexible electrodes (Fig. 2.2c) to wirelessly record and then map SWs (Fig. 2.2d) [31].

Motivation

Although the wireless system shown above is the state-of-the-art in acquiring the gastric SW activity, it still suffers from multiple shortcomings. 1) Due to the stomach motility and the implant bulkiness, the electrodes may detach from the muscle surface, reducing the signal-acquisition quality. The recent human studies showed that these motion artifacts are minimized if the electrode is securely attached to submucosa [86]. Throughout this study SWs were recorded for ~5 days from each patient, but only through a single channel. Sanders et al. have discussed the problems with current methodologies in recording the electrical activity in GI muscle, and have emphasized the mechanical artifacts as a major role player in contaminating the signals [85]. 2) Due to the implant size and implantation procedure, the enteric nervous system under the submucosa space can be damaged, resulting into the failure of the implantation procedure or signal acquisition. 3) Due to the limitations in the endoscopic implantation method, i.e., making a pocket in the submucosa space, this system can acquire SW signals from a limited portion of the GI tract (limited spatial coverage). 4) The thin electrode pads in direct physical contact with the tissue can lead to bio-fluids penetrations, resultant leakage currents, bio-corrosion of the metal, reducing the operational lifetime [87]-[89]. These shortcomings can result in the implant failure, which is not acceptable in clinical translation.

This is the motivation to propose a new paradigm for large-scale gastric interfacing to virtually eliminate all the aforementioned shortcomings by developing a network of distributed minimally- invasive ultrasonically-powered /communicated implants (called Gastric Seeds) that will be 1) small (10 mm3), light, free-floating, and wireless to minimize motion artifacts, tissue damage, and risk of infection and expulsion, 2) modular to acquire SWs from the whole stomach by independent interrogation of each individual Gastric Seed that will have a specific address (ID), and 3), endoscopically implanted within the stomach submucosal space with minimal invasiveness similar to [32], [33].

Gastric Seeds will employ ultrasound for wireless power delivery. Ultrasound for wireless power delivery was used in an implantable passive nerve-cuff stimulator for the first time in [90]. The use of ultrasound for powering brain neural dust was then proposed and applied to the peripheral nervous system (PNS) in [91]-[92]. Due to its passive nature and weak backscattering data link, this system, however, can only work for short distances (8 mm in [92]), is only suitable for anesthetized animals, and cannot be distributed and individually interrogated. Ultrasonic power/data transmission for a single PNS implant (size: 7.8×4×1≈31 mm3) has been demonstrated in [93], [94]. But this system with bulky commercial transducers in the external unit does not have any recording capability, employs power-hungry carrier-based data communication, and more importantly cannot be distributed and individually interrogated. Thus, current ultrasonic implants are suitable for a single implant. In contrast, the proposed work is the first attempt towards the development of truly mm-scale (10 mm3) network of distributed and addressable implants for large-scale gastric mapping using stacked power/data ultrasonic transducers (to minimize the implant size) and low-power pulse-based communication.

Proposed System with Ultrasonic Free-Floating Distributed Implants

Fig. 2.3 shows the conceptual diagram of the proposed technology. It includes a network of implantable Gastric Seeds, a Wearable Unit (WU), and a Stationary Unit (SU). The WU, envisioned to be developed on a flexible printed-circuit board (PCB) in the form of a belly band worn by the subject (to cover the whole stomach from all directions), will consist of an array of stacked ultrasonic (US) power and data transducers (power transmitter, Tx, and data receiver, Rx, in Fig. 2.3). The power Tx transducers will ultrasonically power up distributed Gastric Seeds, and also modulate Gastric Seed IDs on the power carrier for continuously interrogating Gastric Seeds one at a time. The data Rx recovers data pulses (SW signals) sent by Gastric Seeds. Each Gastric Seed in Fig. 2.3 with the envisioned size of ~ 10 mm3 will integrate a CMOS chip for differential SW recording and two stacked US transducers (power Rx and data Tx) on a substrate, on the back of which two recording pads are printed. The power Rx transducer will wirelessly receive power along with amplitude-modulated interrogation data (ID) from WU. The data Tx transducer will wirelessly transmit the recorded SW signals to WU using narrow pulses. Power and data transmissions can be time-multiplexed to avoid interference on the data link. The SU receives SWs for real-time display and mapping of the stomach activity.

Fig. 2.3 Conceptual diagram of the proposed Gastric Seed technology [57]. As shown in Fig. 2.3, the proposed Gastric Seed requires research and development of several state -of-the-art core technologies. In future, the proposed system will have efficient ultrasonic wireless power and data transmission platform for distributed Gastric Seeds regardless of their locations and orientations, an implantable mm-sized Gastric Seed for recording of SWs, and a SU for real time processing and mapping. However, this dissertation is focused on the proof-of- concept development of the proposed system using a pair of stationary optimized ultrasonic transducers.

The related research work on optimization of ultrasonic WPT link, preliminary simulation of a new beamforming technique, as well as Gastric Seed chip development will be discussed in detail in the following chapters.

Chapter 3 Design and Optimization of Ultrasonic Wireless Power Transmission Links for Millimeter- Sized Implantable Medical Devices

As already discussed in chapter 1, wireless power transmission (WPT) to mm-sized implantable devices using ultrasound is favorable in terms of its size and efficiency advantages. As shown in Fig. 3.1, a pair of ultrasonic transducers is used as the transmitter (Tx) and receiver (Rx) for WPT.

The Tx transducer (U1) is driven by an efficient power driver, and the Rx transducer (U2) is followed by an efficient power-management circuitry. The implant core in Fig. 3.1 is modeled as an AC resistance (RL) in this chapter.

Fig. 3.1 Generic ultrasonically-powered mm-sized biomedical implant structure with emphasis on WPT via a pair of ultrasonic transducers (U1 and U2).

Several ultrasonic WPT links have recently been reported for mm-sized implants [50], [93], [95]- [98] . Ultrasonic and inductive WPT links have been modeled in [95] for Rx of 2-10 mm diameter.

For different distances and Rx diameters, the optimal operation frequency (fc) for the ultrasonic link has been found from the model. The simulated power transfer efficiencies (PTEs) of both links have been compared at different distances, suggesting that 1) larger Rx size results in significant improvement in PTE for both ultrasonic and inductive WPT links, and 2) the PTE of the ultrasonic link is significantly higher than that of the inductive link at large distances (10’s of mm), especially when the Rx size is in the mm range. In [96], the measured PTEs of an ultrasonic link utilizing a pair of lead zirconium titanate (PZT) disks have been compared to the simulated PTEs of a comparably-sized inductive link with identical Tx and Rx diameter of 4.4 mm. The PTEs of the ultrasonic link, which were measured with a network analyzer (E5071C, Keysight,

Santa Rosa, CA, USA), were higher than 1% in the mineral oil at separation distance (d) < 2.5 cm within the fc range of 200-400 kHz for matched-source and load conditions.

An ultrasonically-powered implant has been reported in [93], delivering regulated power of 100 µW to an ultrasonic transducer with 1 mm diameter and 1.4 mm thickness at d = 3 cm using a commercial Tx transducer (A303S-SU, Olympus, Waltham, MA, USA) operating at fc = 1 MHz.

In [97], using two different Tx transducers with fc of 2.3 MHz and 1.15 MHz, WPT to three piezoelectric transducers with the size of 5 mm3, 8 mm3 and 16 mm3 at d = 20 cm has been shown with PTEs of 0.4%, 1.7%, and 2.7% in water, respectively. Finally, it has been shown in [98] that 0.127 mm3 neural dust motes can be powered inside water at d = 3 cm with the PTE of 0.002% using a commercial Tx transducer, operating at fc = 5 MHz.

There have also been several publications that describe recharging of batteries using ultrasonic WPT links as well as intra-body communication using a pair of ultrasonic transducers [99]-[101]. However, none of these publications have discussed the design challenges for optimizing the PTE of mm-sized implants.

Although several ultrasonic WPT links have been presented in the past, the literature still lacks a detailed design methodology for finding the optimal geometries of Tx and Rx ultrasonic transducers (U1 and U2 in Fig. 3.1), as well as the optimal fc to achieve highest possible PTE. In other words, either a commercial transducer has been used in the Tx and Rx has been designed to match the Tx operation frequency, or ultrasonic transducers and fc have not been optimized based on a detailed methodology that considers the application and fabrication constraints. In this chapter, a design methodology, which has been validated by measurements, is presented for optimizing ultrasonic WPT links that helps the designers of mm-sized biomedical implants in Fig.

3.1 maximize PTE by optimizing disk-shaped piezoelectric U1 and U2 diameters (Do1 and Do2) and thicknesses (t1 and t2), as well as fc for a given RL and d. The theoretical foundation of ultrasonic WPT will be discussed in Section 3.1, followed by the design and optimization of ultrasonic WPT links for mm-sized implants in Section 3.2. The simulation and measurement results will be discussed in Section 3.3. In section 3.4, the comparison of PTEs for ultrasonic powering link operating at series and parallel resonance frequencies will be shown. In section 3.5, the optimization considering misalignment for ultrasonic power link will be discussed, followed by concluding remarks in Section 3.6.

Theory of Ultrasonic Wireless Power Transmission

Understanding the shape and behavior of a sound beam, generated by an ultrasonic transducer, is key in designing ultrasonic WPT links. In this chapter, disk-shaped piezoelectric transducers will be optimized for WPT, however, similar design methodologies can be generalized for other types of ultrasonic transducers. For both Tx and Rx transducers, similar to previous works in [93][95]- [98], PZT-5A is chosen as the piezoelectric material. Due to inherent symmetry in disc geometry, disk-shaped transducers have the advantage of less complexity in 3D modeling and computation in finite-element method (FEM) simulation tools.

It can be shown that the wireless link PTE in Fig. 3.1 highly depends on the acoustic intensity of

U1 beam (Io), which is defined as the sound beam’s power divided by its area, effective cross- section area of U2 transducer (A2), and the mechanical-electrical power conversion efficiencies

(PCEs) of U1 and U2 transducers (η1 and η2). The link PTE can be found from [93],

I  A  PTE = o 2 2 , (3.1) Pin where Pin is the delivered power to U1 from the energy source.

(a) (b) Fig. 3.2 (a) The characteristics of a sound beam generated by an unfocused disc-shaped transducer [103]. (b) Four design examples for the choice of optimal location and diameter of the Rx transducer to achieve the highest PTE.

Fig. 3.2a shows the characteristics of a sound beam, which has been generated by an unfocused disc-shaped transducer. The sound pressure filed has multiple maxima where a natural focus is defined as the last maximum of the sound pressure field [102]. The shape and operation regions of such a sound beam can be described in terms of its focus point, near- and far-field regions, focal length, and focal zone [103]. The focus point is the location where the beam is the narrowest. The beam width at the focus point is ~half the width of the beam as it leaves the transducer, i.e. Do/2, where Do is the diameter of the transducer. Thus, Io is maximum at the focus point, if the sound

15 loss in the medium is negligible. The near-field region begins from the transducer location and ends at the focus point. As shown in Fig. 3.2a, the beam converges inside the near-field region, resulting in an increase in Io. The focal length (N) is defined as the distance from the transducer to the focus point, which is also called the near-field distance. The far-field region begins at the focus point, where the beam starts to diverge, resulting in a gradual decrease in Io. Finally, the focal zone is the region around the focus point, in which the beam is relatively narrow and Io is relatively high.

Fig. 3.2b shows four different scenarios (A, B, C, and D) for small and large diameters of the Rx transducer at different distances. The larger Rx, which is the same size as Tx in this example, can collect all transmitted ultrasound energy for d < 2N if the sound loss in the medium and beam side lobes as the sound leaves Tx are neglected. Therefore, in this scenario η1, η2, and medium loss determine the PTE. If the Rx diameter is smaller than that of the Tx, which is often the case for mm-sized implants, PTE can be maximized in scenario B where d = N, in which most of transmitted ultrasound energy is received by Rx and Io is maximum, considering negligible acoustic loss of the medium. In scenarios C, D, and even A, the PTE is lower due to smaller Io.

Therefore, U1 and U2 diameters (Do1 and Do2) and N should be chosen properly to maximize PTE in (3.1). In addition, it can be seen in Fig. 3.2 that d, which is often constrained by the application, highly affects the optimal Do1 and Do2 as well.

3.1.1 Design Parameters for Ultrasonic Transmitter

According to (3.1), η1 needs to be maximized to optimize the PTE. It has been shown that operating at the resonance frequency of U1 helps to improve η1 [104], [105]. In a disc-shaped piezoelectric transducer, there are two strongly excited modes, named the thickness extensional (TE) and radial or planar expander (PE) modes, as well as several weakly-coupled vibration modes near the TE and PE modes [106]. For large aspect ratios (Do/t), where t is the thickness of the transducer, TE mode dominates and the transducer has only a piston-type displacement, leading to a simpler design and relatively high η1. As the aspect ratio is reduced, the surface displacement in TE mode becomes more non-uniform due to PE mode resonance and its harmonics, resulting in significant weakly-coupled modes and lower η1.

The thickness mode resonance frequency (fr) of a PZT transducer is determined by its frequency constant (NT) and t as [104], 16

NT fr = t (3.2)

While fr depends on t, the focal length (N) highly depends on the operation frequency (fc) and Do as,

D2  f N = o p 4 v , (3.3) where v is the sound velocity in the medium [103]. It should be noted that (3.2) and (3.3) are accurate for large aspect ratios [103].

According to (3.1)-(3.3), Do1, t1, and fc are all important in optimizing PTE as follows: 1) for a fixed fc, one should choose fr in (3.2) equal or very close to fc by selecting proper t1, 2) to achieve large aspect ratios, t1 should be reduced, increasing fr in (3.2) and adding to the medium loss, and 3) for a fixed Rx size based on the application constraints that limits the Rx transducer diameter

(Do2), Do1 should be chosen close to ~2Do2 based on Fig. 3.2b and also such that N = d based on

(3.3) to maximize I0 in (3.1) at the Rx location. However, finding optimal Do1 to satisfy such requirements is challenging. In addition, such limitations on D01 can significantly limit the aspect ratio and reduce η1. Therefore, there are clear tradeoffs between Do1, t1, and fc, parameters and they should be optimized together.

3.1.2 Design Parameters for Ultrasonic Receiver

Due to limitations on the Rx size in mm-sized implants, Do2 and t2 should both be limited to mm dimensions and below. Therefore, the Rx transducer aspect ratio (Do2/t2) is very limited, causing

(3.2) to be inaccurate in calculating fr of the Rx, and reducing η2. Equivalent circuit models for piezoelectric transducers, such as Krimholtz-Leedom-Matthaei (KLM) model [107], is widely used to design transducers for medical imaging [108], [109]. However, these models cannot be used to optimally co-design the transducer pairs in an ultrasonic powering link. This leads to the need for accurate FEM simulation tools to find fr, which should be equal or close to fc to achieve high PTE.

As shown in Fig. 3.2b and (3.1), an Rx transducer with larger Do2 or A2 can receive more acoustic power, increasing PTE. Therefore, the largest possible Do2, which is constrained by the application, can be chosen. However, as Do2 is increased, Rx transducer resistance at resonance is also changed, 17 affecting PTE for a given RL [93]. This is important, because impedance matching in Rx is needed to improve PTE [110]. In addition, when the sound wavelength in tissue is larger than Do2, which is the case in mm-sized implants operating in MHz-range, acoustic diffraction occurs and acoustic reflection reduces [111]. In this condition, fringe acoustic beams are received by the Rx transducer, which can be considered as an increase in η2 and PTE. In other words, effective A2 in (3.1) becomes larger than the physical area at low fcs. Therefore, both Do2 and t2 need to be swept in FEM simulation tools to optimize PTE, since they can affect fr, impedance matching in Rx, and η2.

3.1.3 Acoustic Matching

Due to acoustic impedance mismatch between the piezoelectric transducer and the medium, which is tissue in biomedical implants, matching layers should be used on both Tx and Rx transducers to minimize acoustic reflections and maximize PTE. Otherwise, the reflections may generate standing waves with acoustic power levels greater than the safety limits [104]. The effect of acoustic matching layer has been analyzed in previous works [50], [112]. In [50], the simulated PTE of ultrasonic link with matching layers was improved by 1.8 times (from 1.15% to 2.11%) compared to the ultrasonic link without matching layers. In [112], transducers with a matching layer achieved improved energy efficiency by 1.4 times. Several techniques have been proposed in literature for designing acoustic matching layers, including single- and multi-layer techniques [113]-[117]. In this chapter, two-layer matching technique in [116] is employed which provides more flexibility in choosing materials, and finding optimal thickness of two matching layers based on the acoustic impedance of matching materials and fc.

Optimal Design of Ultrasonic Wireless Power Transmission Links

Since the geometry of ultrasonic transducers are relatively small for WPT to mm-sized implants,

(3.1), (3.2) and (3.3) can be used to find the initial values for t1, Do1, t2, Do2, and fc, and further optimization is needed in FEM simulation tools such as COMSOL Multiphysics (COMSOL, Burlington, MA), which is the well-adopted simulation tool in industry and academia. Fig. 3.3 shows the ultrasonic WPT link model in COMSOL, where the mm-sized Rx transducer (U2) has been located inside castor oil with an attenuation coefficient of 0.8 dB/cm/MHz to mimic the soft tissue [103]. Fig. 3.3 also shows the ultrasonic Tx transducer (U1). A perfect matching layer (PML) has been considered at the boundaries of the medium to avoid acoustic reflections.

In this chapter, instead of 3-D simulation of transducers the 2-D axisymmetric geometry of transducers as shown in Fig. 3.3 was simulated that helped to significantly reduce the simulation time. The following steps were then taken in COMSOL to simulate PTE, 1) Acoustic piezoelectric interaction (frequency domain), and electrical circuit were chosen as the physics interfaces in

COMSOL, 2) design parameters such as Do, t, and fc were set as global parameters to enable sweeping them, 3) in the electrical circuit interface, an AC voltage source and a source resistor

were connected to U1 to measure the input power, while U2 was connected to a load resistor for measuring output power and consequently PTE. It should be noted that the simulator mesh size should be set at least a fifth of the sound wavelength in the medium that implant is located to improve the accuracy of the simulations. Finally, the damping factor for the piezoelectric material should be entered into the acoustic piezoelectric interaction module.

Fig. 3.3 The ultrasonic WPT link model in COMSOL to optimize the geometries of Tx and Rx transducers (U1 and U2). 3.2.1 Ultrasonic Link Design Procedure

A design procedure is presented in Fig. 3.4 to maximize the PTE of an ultrasonic link for WPT to

mm-sized implants by optimizing Tx and Rx transducer geometries and fc. The optimization flowchart starts with the design constraints imposed by the application, including the maximum

Rx size, constraining its diameter (Do2) and overall thickness (t2ov = t2 + tm2), which t2 and tm2 are the thicknesses of the piezoelectric transducer and matching layer, respectively. The nominal

values for the powering distance (d) and load resistance (RL) are also included in step-1. This procedure can offer optimal WPT links for biomedical implants located at any distance inside the body with different power requirements.

In step-2, the initial values for fc, Do1, thicknesses of the Tx piezoelectric transducer and its matching layer (t1 and tm1), and tm2 are chosen. The following guidelines can be used for selecting these parameters: 1) the minimum achievable resonance frequency for Rx determines the initial value for fc. Since (3.2) is not accurate for small aspect ratios, simulating the impedance of an Rx transducer with maximum t2 = t2ov can give the initial fc. 2) The initial value for t1 can be found from (3.2) for the Tx transducer to achieve the same resonance frequency as the selected fc. Since Tx transducer has relatively larger aspect ratio, (3.2) is fairly accurate. 3) Based on (3), the initial value for Do1 can be calculated to achieve N = d for the selected fc, and 4) tm1 and tm2 can be found from [116] based on the acoustic impedance of matching materials and selected fc.

Fig. 3.4 Iterative ultrasonic link optimization flowchart for efficient WPT to mm-sized biomedical implants.

The geometry of Rx is optimized in step-3 by sweeping both t2 and Do2 to maximize PTE. This step leads to a 3-D surface for PTE vs. t2 and Do2, and the values of t2 and Do2 that maximize PTE are chosen. It should be noted that the maximum value for t2 should be chosen as t2,max = t2ov – tm2.

In step-4, the geometry of Tx transducer is optimized by sweeping both t1 and Do1 to maximize

PTE, using the values for t2 and Do2 from step-3. This step leads to a 3-D surface for PTE vs. t1 and Do1, and the values of t1 and Do1 that maximize PTE are chosen. Steps-3 and 4 are repeated

20 iteratively until t1 and Do1 change less than 0.1% in step-4.1. This leads to optimal Tx and Rx transducer geometries for the selected fc.

In step-5, fc is slightly increased and the same procedure is repeated for the new fc. Since the initial value for fc was chosen for t2 = t2ov, adding tm2 decreases maximum value for t2 in step-3, and Rx will not operate at resonance for the initial fc. Therefore, PTE starts increasing as fc is increased, allowing Rx to operate at resonance with smaller t2. However, acoustic loss in medium and N in

(3.3) increase with fc, the result of which will be decrease in PTE at very high fcs. Therefore, the procedure in Fig. 3.4 should be repeated for several fcs until the new PTE values for higher fcs are at least two times smaller than the peak PTE. Step-5.1 determines optimal ultrasound geometries and fc to achieve the highest PTE, which can be further validated and fine-tuned through measurements.

3.2.2 Ultrasonic Link Design Example for mm-sized Implants

Based on the design procedure in Fig. 3.4, an ultrasonic link, as shown in Fig. 3.3, was optimized for WPT to mm-sized biomedical implants with high PTE. For the design example, the following assumptions were made: 1) U1 and U2 were made of PZT-5A (APC International, Mackeyville, 3 PA), and U1 was backed with air to improve η1, 2) Rx was constrained within 1 mm , 3) U2 was mounted on a silicon die with 0.3 mm thickness to consider the implant core circuitry for power management, data communication, and recording and stimulation operations, which limited the disk-shaped U2 geometry to Do2,max = 1.2 mm and t2ov = 0.7 mm, as shown in Fig. 3.3 inset, 4) Rx was located d = 3 cm inside castor oil, leading to mass density of 956 kg/m3 and sound velocity of 1474 m/s in the simulations [117], 5) the link was designed to deliver ~2 mW with the peak voltage of 3.2 V across U2, imposing RL of ~2.5 k, and 6) schott glass and araldite glue were used as the first and second layers for acoustic matching for both U1 and U2 [116].

Table 3.1 summarizes the results of optimization procedure in Fig. 3.4, including the optimal geometries of the transducers and fc to maximize PTE for three different links. The two links designated by “optimal-design” were optimized with and without matching layers for the aforementioned design assumptions, while U2 was mounted on silicon. Without matching layer, maximum t2 of t2ov = 0.7 mm was considered in the optimization. As shown in Table 3.1, the links with and without matching layer have achieved PTEs of 2.11% and 1.15% at fcs of 1.8 MHz and

Table 3.1 Optimized Ultrasonic Transducer Geometries and Operation Frequency for Design Examples and Measurement

1.4 MHz with different U1 and U2 geometries, respectively. Therefore, including matching layer in design procedure in Fig. 3.4 is key to improve PTE. The Rx transducer without matching layer operates at lower fc of 1.4 MHz to achieve high PTE, because t1 was allowed to be increased to 0.7 mm. However, the Rx transducer with matching layer has compromised higher fc with less reflection, which has resulted in higher PTE of 2.11%.

Figs. 3.5a and 3.5b show geometry optimization of U1 and U2 with matching layer to maximize

PTE at the optimal fc of 1.8 MHz for RL of 2.5 k and d of 3 cm. Fig. 3.5a shows the 3-D surface of PTE vs. t2 and Do2 with initial setting of t1 = 1.05 mm and Do1 = 10.8 mm. It can be seen that

PTE has peaked at t2 = 0.25 mm and Do2 =1.2 mm. Fig. 3.5a clearly shows that PTE tends to increase as Do2 becomes larger, which increases A2 in (3.1). Fig. 3.5b shows the 3-D surface for

PTE when t1 and Do1 were swept for optimal t2 = 0.25 mm and Do2 =1.2 mm. The maximum PTE has been achieved at t1 = 1.05 mm and Do1 =10.8 mm. As Do1 is further reduced below 10.8 mm,

N in (3.3) decreases while I0 in (3.1) increases. The latter helps to improve PTE while the former reduces PTE. This is the main reason that PTE remains relatively high in low Do1 region of Fig. 3.5b.

Fig. 3.6a shows the simulated results of PTE vs. fc for several links with and without matching layer, which were optimized based on the design procedure in Fig. 3.4 for fc range of 1-4 MHz, RL

= 2.5 kΩ, and d = 3 cm. It can be seen that high PTE values have been achieved around 1.4 MHz and 1.8 MHz for links without and with the matching layer, respectively. Since PTE at fc of 4 MHz has significantly dropped by 4.7 times compared to the peak PTE of 2.11% at the optimal fc of 1.8 MHz for the link with matching layer, the design procedure was stopped at 4 MHz. The PTE of the link without matching layer has a slight increase at 1.8 MHz, because for fc lower than 1.8

MHz, the optimal Do2 has been less than its maximum allowable 1.2 mm to achieve desired fc.

(a) (b) Fig. 3.5 Optimization of U1 and U2 geometries with matching layer at fc = 1.8 MHz to maximize PTE by sweeping (a) t2 and Do2 for Do1 = 10.8 mm and t1 = 1.05 mm, and (b) t1 and Do1 for Do2 = 1.2 mm and t1 =0.25 mm.

However, the optimal Do2 values have reached to 1.2 mm for fc of 1.8 MHz and above, which has slightly helped to improve PTE.

(a) (b) Fig. 3.6 Comparison between simulated results of PTE for the links with and without matching layer for RL = 2.5 kΩ. (a) PTE vs. fc for optimized links at each fc based on the design procedure in Fig. 3.4. (b) PTE vs. d for the optimal links at 1.4 MHz and 1.8 MH

Fig. 3.6b compares the simulated results of PTE vs. d for the optimal links with and without matching layer in Table 3.1. It can be seen that the link with matching layer has achieved higher PTE at d > 15 mm. For the link without matching layer at d < 15 mm, it is believed that more acoustic reflections at the boundary of the Tx transducer and oil medium, caused by the impedance

23 mismatch between the PZT and castor oil, have resulted in an acoustic intensity maximum at shorter distances and consequently higher PTE.

Measurement Results

In order to validate the optimization procedure in Fig. 3.4 and the accuracy of simulation results in COMSOL, an ultrasonic WPT link was optimized for measurement purposes. This is designated by “measurement link” in Table 3.1. Although a higher PTE can be achieved using matching layers, the link without matching layers was fabricated to reduce the fabrication complexity. In this link, U2 was mounted on a circular-shaped FR4 printed-circuit board (PCB) with ~5 mm diameter and 1.5 mm thickness. In addition, the following assumptions were made: 1) U1 and U2 3 were made of PZT-5A, and U1 was backed with air, 2) U2 was constrained within 1 mm , which limited the disk-shaped U2 geometry to Do2,max = 1.2 mm and t2max = 1 mm, since no matching layer was used in measurements, and 3) Rx was located at d = 3 cm inside castor oil to power RL of 2.5 k.

The design procedure in Fig. 3.4 resulted in optimal Do1, t1, Do2, t2, and fc of 15.8 mm, 1.93 mm, 1.2 mm, 1 mm, and 1.1 MHz, respectively, leading to the simulated PTE of 1% at d = 3 cm. However, for simplicity, the optimal geometries which were closest to the commercially available ones were chosen. The U1 and U2 geometries, which were used in the measurements, are shown in

Table 3.1 (measurement link). Due to increase in t1 and decrease in t2, which slightly separated frs of U1 and U2 in (3.2), the simulated PTE dropped to 0.65% at fc = 1.1 MHz for RL = 2.5 kΩ.

Fig. 3.7a shows the setup to accurately measure the S-parameters of the ultrasonic WPT link using a network analyzer, which are then converted to Z-parameters to calculate PTE from,

2 Re(Z ) Z PTE = L 21 Re(Z ) Z + Z 2 in 22 L , (3.4) where Zin = Z11 – Z12Z21 / (Z22 + ZL) and ZL = RL [118]. Fig. 3.7b shows the experimental setup for measuring PTE of the ultrasonic WPT link in Table 3.1 inside castor oil in order to mimic the soft tissue [103]. In order to minimize undesired interconnect effects on U2, its top plate was wirebonded to a PCB pad, while its bottom plate was connected to the PCB ground pad using conductive epoxy cured at the room temperature. The transducers were connected to a pair of SMA

24 connecters using two pairs of AWG28 magnet wires, and held perfectly aligned inside the oil tank (by adjusting the alignment until highest link gain was achieved), made of Plexiglas with ~20 cm height. The Rx transducer (U2) was located 3 cm above the bottom surface at the center of the large tank to minimize the sound reflections. The position of U2 was fixed by mounting its PCB on a plastic stand, while U1 was glued to an adjustable 3D-printed plastic stand, which could move in Y- and Z-directions as shown in Fig. 3.7b. In order to provide electrical isolation, U2 and top plate of U1 that was facing oil were coated by Sylgard-184, which has similar acoustic properties to castor oil, while the rest of wires were coated by epoxy.

(a)

(b) Fig. 3.7 (a) PTE measurement setup for the ultrasonic WPT link using a network analyzer that can accurately measure the S-parameters. (b) Ultrasonic WPT link measurement setup inside a castor oil tank.

As recommended by the PZT manufacturer, the impedance of U1 was first measured and matched with simulation results in COMSOL by changing the mechanical damping loss to 0.005, which was then used in all simulations. Figs. 3.8a and 3.8b show well-matched simulated and measured values for real and imaginary parts of U1 and U2 impedances, respectively.

Fig. 3.8a shows that U1 resonates at designated fc = 1.1 MHz, in which U1 resistance is ~450 Ω.

Fig. 3.8b shows that due to a fabrication error, i.e., a decrease in t2, the U2 resonance frequency has slightly increased to 1.15 MHz, resulting in some PTE loss. At fc = 1.1 MHz, the simulated and measured resistance for U2 were 3.6 kΩ and 2 kΩ, respectively. This discrepancy is due to the fabrication error and inaccuracies in modeling the complicated Rx structure in COMSOL. It should be noted that U1 and U2 impedances were measured individually, i.e., the other transducer was removed to avoid undesired reflections.

(a)

(b) Fig. 3.8 Comparison between simulated and measured real and imaginary values for the impedance of (a) Tx transducer (U1), and (b) Rx transducer (U2), which their geometries have been stated in Table 3.1.

Figs. 3.9a and 3.9b show the simulation and measurement results for PTE vs. fc and d of the measurement link in Table 3.1 for RL = 2.5 k, respectively. Fig. 3.9a shows that the maximum well-matched simulated and measured PTEs of 0.65% and 0.66% were achieved at fc of 1.1 MHz for d = 3 cm, respectively. It should be noted that to match simulation and measurement results, the castor oil attenuation coefficient was set to 0.8 dB/cm/MHz in the simulations. Since the measured U2 resistance of 2 k at 1.1 MHz is closer to the designed RL of 2.5 k than the one in

26 simulation (3.6 k), higher PTE was achieved in measurements for most distances except d = 2 cm, as shown in Fig. 3.9b. The measured PTE in Fig. 3.9b has dropped to 0.18 % at d = 6 cm, which is still significant considering WPT to a 1 mm3 Rx. A measured peak PTE of 0.79% was achieved at d = 2 cm, describing the tradeoffs of maximizing I0, N, and η1. In other words, N has been slightly below d = 3 cm in the optimization by reducing Do1, because a smaller Do1 can lead to higher I0 in Rx location as shown in Fig. 3.3. In addition, less acoustic loss in medium at d = 2 cm also contributes to the peak in PTE. The larger mismatch between simulated and measured PTEs at d ≤ 2.5 cm can be due to the acoustic reflections from the Rx PCB, as its shape was not exactly identical to the one in simulations. In order to compensate for the PTE reduction as the distance increases, a closed-loop WPT technique should be employed to monitor the received voltage and adaptively adjust the transmitted power [119], [120]. There has also been other methods that control fc to compensate for distance variations [121], [122].

(a)

(b) Fig. 3.9 Simulated and measured values of PTE vs. (a) operation frequency (fc), and (b) powering distance (d) for RL of 2.5 kΩ.

Fig. 3.10 shows the simulation and measurement results for PTE vs. RL of the measurement link

27 in Table 3.1 for d = 3 cm. Since simulated and measured U2 resistances were 3.6 kΩ and 2 kΩ in

Fig. 3.8b, PTE values were maximized for optimal RL range of 3-4 kΩ and 2-3 kΩ in simulations and measurements, respectively. Therefore, load matching in Rx is required for ultrasonic WPT links to maximize PTE.

Fig. 3.10 Simulated and measured values of PTE vs. RL at d = 3 cm. Fig. 3.11 shows the measured PTEs at d = 3 cm when Rx was misaligned for 0-10 mm in Y- direction as shown in the measurement setup in Fig. 3.7b. As shown in Section 3.1 and Fig. 3.2, it is key for Rx to be located in the focal point of Tx to maximize I0. Since Do1 was 15.9 mm in the measurements, the focal point width was Do1 / 2 = 7.95 mm. Therefore, when U2 was misaligned for 5 mm and moved away from the focal point, I0 in (3.1) was significantly decreased and reduced PTE by ~13 times from 0.66% to 0.05%. It should be noted that in a real environment inside the body, sound reflections from the nearby organs can slightly increase PTE and partially compensate the adverse effects of misalignment [97]. In addition, a larger Tx diameter can help to enhance the beam width at the focus point and mitigate the effect of misalignment at the cost of slightly less PTE in the optimal condition with alignment. In this chapter, a link was designed at the optimal

Fig. 3.11 The measured values of PTE at d = 3 cm vs. lateral misalignment in Y-direction as shown in the measurement setup in Fig. 3.7b.

28 alignment condition. However, the designer can repeat the design procedure for worst-case scenarios by misaligning Rx in COMSOL simulations.

Table 3.2 compares the key parameters of recent ultrasonic WPT links for powering mm-sized biomedical implants with the proposed link. In [93], [95]-[98], a cubic-shaped ultrasonic transducer has been utilized in Rx, and a commercial transducer was used in Tx. In [95], [97], [98], neither the nominal RL has been specified nor a matched-load condition has been considered. This chapter presents the first attempt towards co-optimizing the geometries of both Tx and Rx transducers as well as fc for realistic RL and backing materials, including both silicon and PCB.

Comparison of Series and Parallel Resonance Based on FEM Simulation of Ultrasonic

Wireless Power Transmission to Millimeter-Sized Biomedical Implants

For a PZT transducer, there are a fundamental series resonant frequency (fs) and parallel resonant frequency (fa) at which the device impedance can be purely real [93]. At these resonant frequencies, the power conversion efficiency (η) is higher than at other frequencies. Although previous works claimed that resonant frequency is supposed to be used for both Tx and Rx for all the ultrasonic WPT links, it is still not clear either series resonance or parallel resonance is better for ultrasonic WPT links to achieve higher PTE. In this Chapter, comparison of FEM simulation results will be shown to help the designers of mm-sized biomedical implants choose the type of resonance for both Tx and Rx to achieve higher PTE. Table 3.2 Benchmarking of Recent Ultrasonic WPT Links for mm-Sized Biomedical Implants

Table 3.3 Ultrasonic Transducer Specifications

Table 3.4 Resonance Comparison

In order to analyze which type of resonance is more suitable for ultrasonic WPT to mm-sized implants, COMSOL was used and the same simulation setup was used as shown in previous sections. It has been shown that ultrasonic beam intensity (I0), cross-section area of receiver (A2), power conversion efficiency of transmitter (η1) and power conversion efficiency of receiver η2 are four key parameters determining PTE of the link. In order to fairly compare the effects of different types of resonance, I0 is held constant by fixing operating frequency (fp) and diameter of transmitter

(Do1) while A2 should also be held constant by fixing diameter of receiver (Do2). Only the thickness of Tx (t1) and the thickness of Rx (t2) were tuned respectively to match either the fs or fa to fp.

Therefore, with fixed I0, A2 and input power (Pin), the difference in PTE is mainly due to the change of resonance types. It is important to note that the aspect ratio (Do/t) is also a factor of determining

η1 and η2, however, since fs and fa are not further apart, only a small change in thickness is needed to have either fs or fa matched to fp while keeping Do constant. Therefore, the difference in the aspect ratio for the two cases is not significant and can be ignored.

Table 3.3 summarizes the optimal load (RL,Opt), fp and the transducer geometries used in the simulation. The fp is chosen to be fixed at 1.4 MHz so that aspect ratio is greater than 1 and let the

30 transducers operate at thickness extensional (TE) mode at both fs and fa. Table 3.4 compares the PTEs at 3 cm for all four possible combinations of different types of resonance for Tx and Rx with optimal loads.

Fig. 3.12a and 3.12b show the Tx and Rx impedance plots with their fs matched to fc of 1.4 MHz, respectively. Fig. 3.12c and 3.12d show the Tx and Rx impedance plots with their fa matched to fp, respectively. It can be seen that at the resonant frequencies, the transducers have a purely real impedance and the impedance at fa is much higher than it is at fs.

(a) (b)

Fig. 3.12 Impedance profiles with Tx and Rx geometries listed in Table 3.3, resonant frequencies are matched to fp of 1.4 MHz for (a) fs of Tx, (b) fs of Rx, (c) fa of Tx, and (d) fa of Rx.

Fig. 3.13 shows the PTE vs. RL for all four possible resonance combinations at d = 3 cm. It can be seen that RL,Opt for the parallel resonance is ~29 times larger than it is for the series resonance which is consistent with the impedance plot in Fig. 3.12. It shows that Tx operating at fa is always better for power transmission regardless of the Rx resonance type. It also can be seen that if the RL is less than 10 k, series resonance is preferred for Rx and the parallel resonance is preferred for

Rx when RL is larger than 10 k.

Fig. 3.13 PTE vs. d for all four resonance combinations with optimal load.

Fig. 3.14 shows the PTE vs. d for all four possible resonance combinations with optimal load

assumed. It can be seen that at d = 1 cm, the PTE for parallel-series is lower than other three

combinations with similar PTE. However, at larger distances greater than 2 cm, Tx operating at fa

gives better PTE, and if Tx resonance type is fixed, Rx operating at fa is better than that operating

at fs. In addition, the rate of change in PTE is faster for Tx operating at fa, which implies that the

ultrasound beam generated by Tx (Fig. 3.2) operating at fa converges and diverges faster than the

case of Tx operating at fs.

Fig. 3.14 PTE vs. RL for all four resonance combinations at 3 cm.

Fig. 3.15 shows the PTE vs. fp for all four resonance combinations with optimal load at 3 cm. It

can be seen that for the cases of Tx operating at fs, optimal fp shifts to 1.5 MHz which is between

fs and fa, because 1) when fs of Tx is matched to fp of 1.4 MHz, fa for Tx is at ~1.55 MHz at which

can generate higher I0, and 2) at frequency of 1.5 MHz (between fs and fa ), the power dissipation

32 is the lowest (less input power is needed) [123]. In addition, for Tx operating at fa, PTE peaks at 1.4 MHz as expected.

Fig. 3.15 PTE vs. fp for all four resonance combinations with optimal load at 3 cm.

In this section, the Tx and Rx operating at fs and fa have been compared and discussed, which gives more insights about effects of different types of resonance on PTE to designers of ultrasonic WPT to mm-sized implants. The simulation results suggested that for Tx should always operate at fa to achieve efficient ultrasonic WPT links. And for Rx, the choice of types of resonance depends on the load impedance determined by the application, for RL of less than 10 kΩ, operating at fs is preferred, in contrast, operating at fa is preferred for RL is greater than 10 kΩ.

Ultrasonic Power Link Considering Misalignment

It has been shown in previous sections, although ultrasonic links achieve high PTE, particularly for WPT to deeply-implanted mm-sized devices, they are susceptible to any Rx misalignment due to their focusing nature, as shown in Fig. 3.11. This may cause ultra-low PTE and even malfunction in powering implants that involve dynamic movements, notably peripheral nervous system (PNS) interfaces for recording or stimulating actively-moving organs such as stomach and heart [33], [124]. Therefore, ultrasonic links need to be optimized for the worst-case scenario of the misalignment.

As shown in section 3.1, the ultrasound beam is the narrowest in the focal zone, which is an optimal location for the Rx when it is perfectly aligned. However, by placing Rx inside the deep far-field region, PTE can be improved for a wider range of misalignments at the cost of PTE reduction in perfectly-aligned conditions. This can be done by reducing Dou1, which decreases the focal length.

Due to the limited aspect ratio of the Tx transducer, the thickness of Tx (t1) should also be optimized along with diameter of Tx (Dou1) to maintain U1 resonance at the operating frequency

(fp). It should be noted that due to the side beams, increasing Dou1 might also improve PTE for a certain misalignment condition. However, this approach places Rx in the near-field region with several maxima and minima in the sound intensity, which can lead to drastic PTE reduction if Rx slightly moves. The proposed optimization of ultrasonic links for worst-case misalignment scenarios is as follows. Step-1: the design procedure in Fig. 3.4 with fully-aligned Rx will be followed first to find optimal U1 and U2 geometries as well as fp. Step-2: the optimal Dou1 in step-

1 is reduced for 0.1 mm and the corresponding t1 for resonating U1 at fp is found by sweeping t1 around its optimal value from step-1. Step-3: using new Dou1 and t1 values, PTE will be found for all practical misalignments. It is clear that for the new Dou1 and t1, PTE for the fully-aligned condition will be reduced. However, steps-2 and -3 will be repeated in a recursive process until the PTE at the worst-case misalignment keeps decreasing. Finally, among the U1 geometries that lead to the maximum or close-to-maximum PTEs at the worst-case misalignment, the one with relatively higher PTEs for lower misalignments, including aligned condition, will be chosen.

For a worst-case misalignment of 3 mm, the ultrasonic link was re-optimized using the same 3 procedure shown in Fig. 3.4 with 1.1 mm Rx at d = 10 mm by changing Dou1 and t1.

Fig. 3.16a. shows the simulated 3-D surface of PTE vs. t1 and Dou1 for the perfectly-aligned condition. It can be seen that the maximum PTE of 10.6% was achieved at Dou1 = 6.4 mm and t1 =

1.15 mm. In Fig. 3.16a, for each Dou1 there is an optimal t1 that corresponds to U1 resonance at optimal fp of 1.8 MHz. As Dou1 was decreased from 6.4 mm to 2 mm, the peak PTE also decreased from 10.6% to 1.45% due to widening the ultrasound beam.

Fig. 3.16b shows the simulated 3-D surface of PTE vs. Dou1 and Rx misalignment in Y-direction, using the optimal t1 in Fig. 3.16for each Dou1. It can be seen that at 3 mm of misalignment, the highest PTE of 0.36% was achieved with the small Dou1 of 2.4 mm. At similar misalignment of 3 mm, the links with Dou1 of 3 mm and 2.2 mm achieved slightly lower PTEs of 0.32% and 0.34%, respectively. Dou1 of 3 mm is preferred, because 1) at 3 mm of misalignment, its PTE is very close to the maximum achievable PTE, 2) in fully-aligned condition, it achieves higher PTE of 4% compared with the PTEs of 2.85% and 2.58% for Dou1 of 2.4 mm and 2.2 mm, respectively, and 3) as shown in Fig. 3.16b, it achieves higher PTE for a wide range of misalignments. As shown in

Figs. 3.16a and 3.16b, for the misalignment of 0 to 3 mm, the link with smaller Dou1 of 3 mm shows only 12.5 times reduction in PTE from 4% to 0.32%, as opposed to 88 times decrease in PTE for the larger Dou1 of 6.4 mm from 10.6% to 0.12%. In other words, by reducing Dou1 from 6.4 mm to 3 mm, PTE was increased by 2.7 times (0.32% vs. 0.12%) for 3 mm of misalignment at the cost of 2.6 times less PTE without misalignment (4% vs. 10.6%). The resulting smaller variation in

PTE, i.e., flatter PTE surface, for small Dou1 can also highly benefit the external power driver and battery by reducing their output power range [119].

(a)

(b) 3 Fig. 3.16 Simulated 3D PTE surfaces of the ultrasonic link for 1.1 mm Rx at d = 10 mm vs. (a) different t1 and Dou1 values in perfectly-aligned condition, and (b) Rx lateral misalignment in Y-direction and Dou1.

Fig. 3.17 compares the measured PTEs of two ultrasonic links with different U1 geometries and identical 1 mm3 Rx for d = 30 mm and Rx misalignment of 0-5 mm in Y-direction as shown in

Fig. 3.7b. The “large U1” in Fig. 3.17 refers to the optimal U1 geometry without matching layer

(Dou1 = 15.9 mm, t1 = 1.95 mm) for aligned condition, while the “small U1” refers to a transducer with reduced Dou1 of 9.5 mm (t1 = 1.95 mm) to improve robustness against misalignment. It can

35 be seen in Fig .3.17 that the PTE for large U1 is 3.1 times higher than the PTE for small U1 (0.62% vs. 0.2%) when U2 is perfectly aligned. However, with 5 mm misalignment in Y-direction, the link with small U1 outperformed the one with large U1 by 2.6 times (0.12% vs. 0.047%) due to the wider ultrasonic beam for small U1 at d = 30 mm. When U2 was misaligned for 5 mm, the PTE for the link with large U1 was significantly reduced by 13 times, while the PTE for the link with small

U1 only reduced by 1.7 times.

Fig. 3.17 Comparison of the measured values of PTE at d = 3 cm vs. lateral misalignment in Y-direction as shown in the measurement setup in Fig. 3.7b. for two different U1s. Therefore, it can be concluded that if the application involves implant movements, ultrasonic links are susceptible to misalignments and should be designed for the worst-case misalignment scenarios. Otherwise ultrasonic transducer arrays should be used to improve PTE.

Conclusions

A design methodology has been proposed to maximize the PTE of ultrasonic links for WPT to mm-sized biomedical implants. The proposed design procedure helps the designer identify the optimal geometries of Tx and Rx ultrasonic transducers such as diameter and thickness, as well as the optimal operation frequency. The optimal operation frequency of an ultrasonic WPT link with the Rx size of 1 mm3 to power a load of 2.5 kΩ at the powering distance of 3 cm was found to be 1.8 MHz using acoustic matching layer on both Tx and Rx transducers. The optimal link achieved a PTE of 2.11%, while the Rx transducer was on a silicon die with 0.3 mm thickness, mimicking the implant core circuitry. The design procedure and COMSOL models were also validated through measurements by optimizing and measuring the PTE of an ultrasonic WPT link that

36 involved a 1 mm3 Rx transducer mounted on FR4 PCB. The simulated and measured PTEs matched well at 0.65% and 0.66% to power a load of 2.5 kΩ at the distance of 3 cm, respectively. The measured PTE reduced to 0.18% at 6 cm, which is still promising for WPT to deeply- implanted mm-sized devices for local operations such as recording and stimulation. This work presents the first reported ultrasonic WPT link, in which the geometries of both Tx and Rx ultrasonic transducers as well as operation frequency have been co-optimized, with realistic design settings for mm-sized biomedical implants.

The emphasis of this work is to optimize PTE of the ultrasonic WPT link since PTE reflects the requirements on power source which is crucial especially for applications involving low-power implants where acoustic intensity is assumed to be well below the safety limits. However, power delivered to the load (PDL) under safety limits is also an important factor which needs to be taken into account. PDL can be delivered by an optimized ultrasonic WPT link under safety limit, as discussed comprehensively in [52]. The FDA safety constraints is limited by the maximum spatial- 2 peak temporal-average intensity (ISPTA) of 720 mW/cm [52]. Under this limit, the optimized ultrasonic link with 1.1-mm3 Rx can achieve maximum of 2.1 mW and 1 mW at the powering distances of 3 cm and 5 cm, respectively, whereas the expected average power consumption of most implanted devices with recording and stimulation functions ranges from few hundreds of µW to 1 mW [125], [127].

It has also been shown in [52] that ultrasonic WPT links are not only susceptible to misalignment but also to orientation. With an optimized ultrasonic WPT link with 1.1 mm3 Rx, for only 30 degrees of rotation at the powering distance of 3 cm, PTE reduces from 10.6% to 0.35% and maximum PDL under safety limit reduces from 2 mW to 62 µW. Therefore, both misalignment and orientation should be considered while designing ultrasonic WPT links especially for applications with implant movements.

Chapter 4 A Hybrid Inductive-Ultrasonic Link for Wireless Power Transmission to Millimeter-Sized Biomedical Implants

Although ultrasonic links have shown promising power transfer efficiencies (PTEs) for powering deeply-implanted mm-sized devices, its use is limited to applications that involve a homogenous medium with minimal acoustic impedance changes. In some applications that involve brain implants or implants inside the body of freely- behaving small animal subjects, such as rodents, however, power needs to be transferred through bone and air, which are highly dissipative for ultrasound [14]. At the same time, inductive link limits the PTE in powering miniaturized implants, preventing the receiver coil (Rx coil) from deep implantation into the target organ. To overcome this issue, a combination of inductive and ultrasonic links has been proposed in [128]. However, they propose to first harvest power in the inductive receiver (Rx) side and then use an additional power amplifier (PA) to drive the ultrasonic transducer, which significantly adds to the size and power loss.

Fig. 4.1 Hybrid inductively-ultrasonically-powered mm-sized implant structure with WPT via a pair of coils (L2 and L3) and ultrasonic transducers (U1 and U2) for applications that involve mediums with different acoustic impedances.

The concept, design, and optimization of a hybrid inductive-ultrasonic link, as shown in Fig. 4.1 is presented for wireless power transmission (WPT) to mm-sized implants by optimally cascading co-optimized inductive and ultrasonic links. The inductive link (L2-L3) is optimized to directly

38 drive the ultrasonic transducer as the power transmitter (Tx), U1, with no need for a PA, within air or bone mediums. The ultrasonic link (U1-U2) is optimized to drive the implant core, simplified as an AC resistance (RL). For instance, to power mm-sized implants carried by freely-moving animals, L2 is located under the animal’s cage [14], L3 and U1 are attached to the animal body inside air or under the skin, and U2 will be implanted into the target tissue. The design method will be discussed in Section 4.1, followed by the measurement results and conclusions in Sections 4.2 and 4.3, respectively.

Hybrid Inductive-Ultrasonic WPT Link

In the proposed hybrid WPT link as shown in Fig. 4.1, an inductive link, L2 and L3, is driven by the energy source and the PA to wirelessly transfer power within the first medium with high acoustic loss, e.g. air/bone. A pair of ultrasonic transducers, U1 and U2, is then followed by the Rx coil, L3, to transfer power wirelessly in the second medium with small acoustic loss, such as the tissue. In the hybrid link, the Tx transducer (U1) is directly driven by L3, while the Rx transducer

(U2) drives RL. In the actual implementation of the implant, as shown in Fig. 4.1, the AC carrier across U2 is rectified and regulated by a power management.

In order to achieve the maximum PTE in the hybrid link (PTEhybrid), the geometries of coils and transducers as well as the operation frequency (fp) should be optimized. Sweeping all these parameters together with several iterations to maximize the PTEhybrid would be a very slow process and, therefore, there is a need for a faster design approach. Considering that the PTE of the ultrasonic link (PTEUS) is much smaller than that of the inductive link due to the mm-dimension of U2, first the ultrasonic link should be optimized to find the optimal geometries for ultrasonic transducers and fp. Since the acoustic loss inside the tissue significantly increases with fp while the inductive link is less sensitive to fp within the MHz range, the optimal fp should be found along with the optimization of the ultrasonic link [50]. Then the PTE of the inductive link (PTEind) is optimized at fp, considering the input impedance of U1, i.e. RU1 in Fig. 4.1.

4.1.1 Design and Optimization of the Ultrasonic WPT Link

The optimal geometries of U1 and U2 in Fig. 4.1 as well as fp can be found from the design procedure proposed in [50]. This includes 1) building the link model in COMSOL Multiphysics (COMSOL, Burlington, MA) as shown in Fig. 4.2a, and 2) considering the requirements imposed

39 by the application such as RL, implant depth inside the tissue (dus), and implant dimensions, i.e. maximum U2 diameter (Dou2) and thickness (t2) as shown in Fig. 4.2a. The transducers are located inside castor oil with an attenuation coefficient of 0.8 dB/cm/MHz to mimic the soft tissue. In Fig. 4.2a, a perfect matching layer (PML) has been considered at the boundaries of the medium to avoid acoustic reflections. Following the design procedure in [50]. The optimal values for Dou2, t2, diameter and thickness of U1, Dou1 and t1 respectively, and fp will be found.

(a) (b)

Fig. 4.2 Simulation Setup (a) The ultrasonic WPT link model in COMSOL to find the optimal geometries for U1 and U2, i.e., Dou1,2, t1,2, as well as fp based on the design procedure in [51]. The soft tissue is mimicked by castor oil with similar acoustic attenuation of 0.8 dB/cm/MHz. A perfect matching layer (PML) has been considered at the boundaries of the medium to avoid acoustic reflections. (b) The 2-coil inductive link model in HFSS to find the optimal geometries of L2 and L3, i.e., Do2,3, w2,3, n2,3, s2,3. The silver electrodes of U1 have been modeled with two parallel silver plates in HFSS. 4.1.2 Design and Optimization of the Inductive Link

The PTEind should be maximized for the given fp and RU1 from the ultrasonic link design, and the coupling distance between Tx and Rx coils, dind, imposed by the application. Since multi-coil links in the form of 3- and 4-coil links can potentially improve PTEind, compared to conventional 2-coil links as shown in Fig. 4.1, the first decision should be to choose the optimal number of coils [16]. Then the inductive link is modeled in a full-wave electromagnetic field simulator, HFSS (Ansoft, Pittsburgh, PA), as shown in Fig. 4.2b, to find the optimal geometry of the coils, i.e., outer diameter, Do, number of turns, n, wire spacing, s, and wire width, w.

The PTEind of a 2-coil link, as shown in Fig. 4.1, can be found from,

k 2 Q Q Q  = 23 2 3L  3L 2−coil 1+ k 2 Q Q Q 23 2 3L L , (4.1) where Q2 = ωpL2/R2, Q3 = ωpL3/R3, QL = RU1/ωpL3, Q3L = Q3QL/(Q3+QL), and k23 is the mutual coupling factor [16]. In 3- and 4-coil links, an additional coil, L4, is added to the Rx side to drive

U1 and, therefore, to provide optimal load condition to achieve maximum PTEind. As shown with details in [16], the PTEind of a 3-coil link can be found from,

k 2 Q Q k 2 Q Q Q  =  = 23 2 3  34 3 4L  4L , 3−coil 23 34 1+ k 2 Q Q + k 2 Q Q 1+ k 2 Q Q Q 23 2 3 34 3 4L 34 3 4L L (4.2) where QL = ωpL4/RU1 for the series connection of L4 and RU1, Q4L = Q4QL/(Q4+QL), k34 is the new degree-of-freedom provided by 3-coil link to achieve optimal loading condition, and η23 and η34 are the PTEinds of the L2-L3 and L3-L4 links, respectively [16]. In 4-coil links, an additional coil is added to the primary side, compared to 3-coil links, when the PA resistance is quite large.

The design and optimization of 2-coil and 3- and 4-coil links has been discussed in [16], respectively. One can use such design procedures to optimize three sets of 2-, 3-, and 4-coil inductive links and compare their PTEind. However, in the following, it will be shown that for powering mm-sized implants with hybrid link, 2-coil links achieve higher PTEind.

4.1.3 Hybrid Link Design Example

As a design example, a wirelessly-powered neural interfacing system for freely-behaving rodents has been chosen [14]. The goal here is to power a 1 mm3 implant inside the body of a rodent using the proposed hybrid link in Fig. 4.1. For such an application, assumed 1) circular-shaped 3 transducers are used with maximum Dou2 = 1.1 mm and t2 = 1 mm (implant size = 1 mm ) to deliver power to RL of 2.5 k, 2) implant depth in tissue is dus = 3 cm, 3) wire-wound coils with high quality factors are used [16], 4) the Tx coil (L2) is placed under the animal cage surface while the

Rx coil (L3) is attached to the animal body and spaced from L2 by dind = 3 cm, 5) maximum values for Do2 and w2 are limited to 10 cm and 1.02 mm, respectively, and 6) the total size of L3 and U1 , attached to the animal body, are limited to 31×17 mm2 to reduce their weight.

The ultrasonic link was first designed based on the design procedure in [50], considering that U2 is mounted on a printed-circuit board (PCB) with 1.5 mm thickness for the sake of measurements, as shown in Fig. 4.2a. The optimization led to Dou1 = 15.9 mm, t1 = 1.95 mm, Dou2 = 1.1 mm, t2 =

0.95 mm, and fp = 1.1 MHz, resulting in the simulated PTEUS of 0.65% at dus = 3 cm for the ultrasonic link only. More details can be found in the previous chapter.

The optimal inductive link can be found by using RU1 = 400 Ω and fp = 1.1 MHz from the ultrasonic design. In order to find the optimal number of coils, QL in (4.2) needs to be calculated. For series load connection in a 3- or 4-coil link, QL = ωpL4/RU1 is quite low since fp is low and RU1 is high. In simulations, the highest QL of 0.033 could be achieved with n4 = 18. Such small QL, however, leads to very small Q4L < 0.033 and, therefore, η34 in (4.2) is significantly reduced. Since the PTEind of WPT links for mm-sized implants is quite small (< 1%), to achieve required delivered power the transmitted power should be increased, ruling out the use of 4-coil links [16]. Therefore, the 2-coil link with parallel resonance in Fig. 4.1 is the optimal design.

Fig. 4.2b shows the 2-coil link model in HFSS, which the geometry of the coils were optimized to maximize PTEind in (4.1) based on the design procedure in [129]. In Fig. 4.2b, the conductive parallel plates of U1 were modeled with two circular silver plates with the thickness of 25 µm and 2 separation of 2 mm. Considering the maximum area of 31×17 mm for L3 and U1 together, three geometrical structures are possible as shown in Fig. 4.3. In Figs. 4.3a and 4.3b, labeled as structures-1 and -2, U1 is located inside L3 at the side and center, respectively, while in Fig. 4.3c, labeled as structure-3, U1 is located next to L3. It should be noted that a larger coil area can be achieved in structures-1 and -2, however, the conductive plates of U1 can result in low Q3 and k23.

On the other hand, structure-3 can leave some room for more inner turns in L3, leading to higher

Q3 and k23.

(a) (b) (c)

Fig. 4.3 Three different 2-coil structures in HFSS considering the location of L3 and U1 in the hybrid link. Two parallel silver plates have been considered in these strcutures to mimic the silver electrodes of U1. (a) Structure-1: U1 is located inside L3 at the side, (b) Structure-2: U1 is located at the center of L3, and (c) Structure-3: U1 is located next to L3.

Table 4.1 PTE Comparison Between Different L3-U1 Structures in Fig. 4.3

In order to find the optimal structure in Fig. 4.3, three sets of 2-coil links as shown in Fig. 4.3 were 2 designed for the maximum area of 31×17 mm . Table 4.1 summarizes the PTEind and coil geometries of these links, suggesting that structure-3 in Fig. 4.3c results in the maximum PTEind of 50.9% at fp = 1.1 MHz and dind = 3 cm. An optimized 3-coil link with similar design constraints achieved a smaller PTEind of 13.6% due to small QL of 0.033.

Simulation and Measurement Results

The PTEs of the inductive, ultrasonic, and hybrid links were measured and compared with the simulated ones to show the accuracy of the simulations. Fig. 4.4a shows the setup to measure the S-parameters of each link using a network analyzer, which are then converted to Z-parameter to calculate PTEhybrid from

2 Re (ZL) |Z21| 푃푇퐸ℎ푦푏푟𝑖푑 = 2, (4.3) Re (Zin) |Z22+ZL| where Zin = Z11 – Z12Z21 / (Z22 + ZL) and ZL = RL [50].

Fig. 4.4b shows the experimental setup for measuring PTEhybrid, which coils’ and transducers’ geometries are listed in Tables 4.1 (structure-3) and 4.2, respectively. To mimic the soft tissue, U2 was placed inside castor oil, while the inductive link was in the air. In Fig. 4.4b, U1, L2, and L3 were held by a single adjustable plastic holder, which could move in Y- and Z-directions, i.e., dind was fixed at 3 cm. A pair of SMA connectors were attached to L2 and U2 to measure S-parameters.

(a)

(b) Fig. 4.4 Measurement setup (a) PTE measurement for the inductive, ultrasonic, and hybrid links using a network analyzer that can accurately measure the S-parameters. (b) Hybrid inductive-ultrasonic WPT link measurement setup, where inductive and ultrasonic links were in air and castor oil tank, respectively. L2 and U2 were connected to a pair of SMA connectors to measure the S-parameters.

AWG-18 and AWG-26 single-conductor wires were used to make L2 and L3, and then they were resonated out at 1.1 MHz by two capacitors of 1.22 nF and 4.4 nF, respectively. In order to provide electrical isolation inside the castor oil, U2 and the front sides of U1 and L2, which were touching the oil, were coated by Sylgard-184, which has similar acoustic properties to the castor oil. For measuring the PTEind, non-resonant L2-L3 coils were held in air and S-parameters were measured to find Q2, Q3, and k23 from,

Im(Z ) Im(Z ) Im(Z ).Im(Z ) 22 33 23 32 . Q2 = ,Q3 = , k 23 = (4.4) Re(Z 22 ) Re(Z 33 ) Im(Z 22 ).Im(Z 33 )

The PTEUS has already been measured in previous work [50].

Fig. 4.5 compares the simulated and measured PTEind vs. dind at the fp of 1.1 MHz, showing that a measured PTEind of 50.6% has been achieved at dind of 3 cm. The simulated and measured PTEUS vs. dus at the fp of 1.1 MHz are presented in [50], showing that a measured PTEUS of 0.66% was

44 achieved at dus of 3 cm. It is worth noting that simulation and measurement results were well matched for both inductive and ultrasonic links.

Fig. 4.5 Comparison between simulated and measured PTEind vs. dind for inductive link only in air.

Fig. 4.6a shows the simulated and measured values for the PTEhybrid vs. the powering distance, i.e., d = dus + dind, where dind was kept fixed at 3 cm by moving L2,3-U1 all together in Z-direction. The measured PTEhybrid at d = 6 cm was 0.16%, which was ~half of the product of individual PTEs from inductive and ultrasonic links only, i.e., 0.506×0.66% = 0.33%. This was due to the presence of the resonance capacitors for L2 that increased R2 in Fig. 4.1 from 0.6 Ω to 2.38 Ω at 1.1 MHz, decreasing Q2 from 188.5 to 47.5 and, therefore, reducing PTEind by ~ half from 50.9% to 26% at dind = 3 cm. This R2 increase is also considered in the simulated results in Fig. 4.6a.

Fig. 4.6b shows the measured PTEUS vs. d when U1 was located 3 cm away from the castor oil to mimic the same test condition as the hybrid link one. The measured PTEUS significantly dropped to 1.4×10-5% at d = 6 cm, showing the advantage of the hybrid link when two different mediums are involved.

(a) (b) Fig. 4.6 Comparison for fixed dind of 3 cm, (a) simulated and measured values of PTEhybrid vs. powering distance (d = dind + dus) for the hybrid inductive-ultrasonic WPT link at fp of 1.1 MHz for RL of 2.5 kΩ, and (b) measured PTEUS vs. d for the ultrasonic link only with air-tissue medium. Coils and transducer geometries are listed in Tables 4.1 (structure-3) and 4.2, respectively.

Table 4.2 Benchmarking of Recent WPT Links for mm-Sized Biomedical Implants

Table 4.2 benchmarks recent ultrasonic and RF methods for powering mm-sized implants against the proposed hybrid inductive-ultrasonic link. Using high fp of 1.6 GHz has resulted in the small PTE of 0.04% at d = 5 cm for the RF link in [19] with larger Rx of ~10 mm3. The inductive link in [21] has achieved higher PTE of 0.56% at lower fp of 200 MHz, but at small d of 1.2 cm. In comparison, the hybrid link has achieved a high PTEhybrid of 0.16% at large d of 6 cm with much lower and safer frequency of 1.1 MHz.

Conclusions

A hybrid inductive-ultrasonic WPT link has been proposed to achieve high PTE for powering mm- sized biomedical implants in applications that involve multiple mediums with different acoustic impedances (air, bone, tissue), such as neural interfacing with freely-behaving rodents inside a cage. A co-designed hybrid link for an air-tissue medium was optimized and operated at 1.1 MHz 3 to wirelessly power a 1 mm ultrasonic transducer, located 3 cm inside castor oil, with a PTEhybrid of 0.16%, while the Tx coil was placed 6 cm away from the Rx transducer. An input power in Watt-range will be required in the hybrid link to power mm-sized implants with mW-range of power consumption.

Chapter 5 Self-Image-Guided Ultrasonic Power Transmission to mm-Sized Implantable Devices

Although promising results have been shown in previous works using ultrasonic powering links, the location of receiver transducer (Rx) for current works has to be known to achieve claimed power transfer efficiencies (PTEs) due to the high sensitivity of ultrasonic links to the misalignment, particularly when the transmitter transducer (Tx) is highly focused. Ultrasonic links for applications involving free-floating implants or micro-motion can suffer from extremely low efficiencies. For example, for gastric-wave recording, the stomach can move up to 27 mm under free breath condition [126], while the PTE and power delivered to the load (PDL) of an ultrasonic link could reduce by 88 and 95 times from 10.6% to 0.12% and 2 mW to 210 µW for only 3 mm misalignment, respectively [52]. The literature lacks a platform to adaptively locate implants and maintain high PTEs regardless of changing in Rx locations.

This is the motivation to propose a practical ultrasonic interrogation platform, which virtually eliminates abovementioned issue by employing Self-Image-Guided ultrasonic (SIG-US) interrogation that can automatically adapt to the varying environment (micro-motion and tissue medium) without any prior knowledge of the implant’s location/medium, leading to robust, highly focused (efficient) beamforming for ultrasonic power/data transmission.

In this chapter, the SIG-US interrogation concept in a distributed peripheral nervous recording system is presented. The finite-element (FEM) simulation results will be shown to prove the feasibility of powering mm-sized implants. The SIG-US concept will be presented in Section 5.1. In Section 5.2, the simulation setup and results will be discussed, followed by conclusion remarks in Section 5.3.

Self-Image-Guided Ultrasonic Interrogation Concept

Fig. 5.1 shows the conceptual diagram of the proposed technology. It includes a network of implantable peripheral nervous system (PNS) nodes, a Wearable Unit (WU), and a Stationary Unit (SU). The WU, envisioned to be developed on a flexible printed-circuit board (PCB) in the form of bands worn by the subject (to cover the targeted area from all directions), will consist of an array of stacked ultrasonic (US) power and data transducers (power Tx, and data Rx, in Fig. 5.1). 47

Fig. 5.1 Conceptual diagram of the proposed Self-Image-Guided Ultrasonic (SIG-US) interrogation platform used in a distributed, addressable Peripheral Nervous System (PNS).

The power Tx transducers array will initially be driven to provide semi-homogeneous US power

for the PNS nodes at relative low efficiency. Then the PNS ID is modulated on the power carrier and transmitted to the PNS nodes. The PNS nodes will receive power/data, check the ID, and if matched, transmit recorded data along with 1-bit power information (indicating the supply level) back to WU using sharp pulses. Since a small (mm-sized) transducer operates as a point source, the transmitted pulses from each PNS node will be received by all external data US transducers. These received pulses with different amplitudes and delays (depending on the implant’s location and orientation) will be used to image the location of the PNS nodes. When the PNS node’s location is imaged, the external power US transducers are then driven with similar delays by a phased array in a closed-loop fashion that will steer a highly efficient and focused beam towards the PNS node regardless of its location and orientation. This beamforming process will be repeated, and the location will be updated in real time as the PNS nodes moves. The received power information is also used in WU to adaptively increase the transmitted power for distant/oriented PNS nodes with lower power efficiency. The SU receives recorded information for real-time display. The SIG-US interrogation technology can be adopted for wireless power and data transmission to any network of distributed, addressable implants with different alignment/orientation in an actively mobile environment.

As shown in Fig. 5.1, the proposed system requires several state-of-the-art core technologies. In this chapter, only the proof-of-concept FEM simulation results on the SIG-US powering for a single PNS node will be presented.

Simulation Setup and Results

Fig. 5.2 shows the simulation setup of an ultrasonic WPT link model of a linear array of 11 ultrasonic Tx array (each 1 mm3) operating at 1 MHz to power a 1 mm3 implant located 30 mm away in COMSOL Multiphysics (COMSOL, Burlington, MA), where the mm-sized Rx transducer with silicon-backing has been located inside castor oil (which has similar acoustic impedance as soft tissue) [50]. An ultrasonic Tx array with 11 transducers is chosen to simplify the design in these proof-of-concept simulations. A 2-D array should be developed and implemented in future work for practical use in a given application. The pulse-voltage source is first connected to Rx to transmit a sharp pulse to the Tx array, the pulse will be received by each Tx transducer with different delays and amplitudes depending on the Tx transducer location. These relative delays will then be used to transmit sinusoidal signals by connecting a sinusoidal voltage source to each Tx transducer in Tx array to efficiently power the Rx. The simulations results and comparisons of the cases with and without SIG-US interrogation as well as the case with conventional beamforming will be shown in detail later in this section.

Fig. 5.2 The ultrasonic WPT link model in COMSOL of a linear array of 11 ultrasonic transmitter (Tx) array (each 1 mm3) operating at 1 MHz to power a 1 mm3 implant located 30 mm away. A castor oil medium is considered to mimic the soft tissue, as it has similar acoustic properties.

Fig. 5.3a shows the pulse sent by Rx with an amplitude of 5 V and an optimal pulse width of 450 ns for a 1 MHz ultrasonic link. Fig. 5.3b shows the received pulses by each Tx transducers in the th array with their relative delays. It can be seen that the aligned 6 Tx transducer (Tx6) received the pulse first because the least distance was traveled. Tx transducers further from Rx received the pulse with larger delay and lower amplitude. Fig. 5.4 shows the relative delays of the received pulses versus the Tx transducer numbers for the Tx array in aligned condition. It can be seen that

Tx1 and Tx11 have the largest delay of 620 ns because they are the furthest from Rx.

(a)

(b) Fig. 5.3 (a) The pulse sent by Rx (in Fig. 5.2) with optimal pulse width of 450 ns for a 1 MHz ultrasonic link. (b) The received pulses by Tx transducers with relative delays and different amplitudes depends on the location of each Tx transducer. Fig. 5.5a shows the sinusoidal signals sent by the Rx with 5 V peak voltage at 1 MHz frequency, and Fig. 5.5b shows the received sinusoidal signals by different Tx transducers with relative

50 delays. It should be noted that the delays acquired by sinusoidal excitation are the same as the ones received by the pulse excitation.

Fig. 5.4 The relative delays for each Tx transducers in Tx array.

(a)

(b) Fig. 5.5 (a) the sinusoidal signals sent by Rx with 5 V peak voltage at 1 MHz. (b) The sinusoidal signals received by the Tx transducers in the Tx array with relative delays similar to pulse excitation in Fig. 5.3.

When powering the implant using SIG-US by Tx array, the transducers with larger delays will be excite first so that the signals generated by Tx transducers at different location can reach Rx at the same time with the same phase. Fig. 5.6 shows the comparison of received Rx voltages for the case with SIG-US to the case without any beamforming technique. It can be seen that the peak voltage received by SIG-US technique is 2.3 times higher than the case without beamforming, which is equivalent to 5.3 times higher received power.

Fig. 5.6 Comparison of the voltages received by Rx with SIG-US to the voltages received by Rx without SIG-US under the same condition. To mimic the condition with micro-motion of the PNS nodes, the Rx is misaligned with 1 mm step up to 6 mm. The same process was repeated to find optimal delay values. The results were compared to those of the conventional beamforming, assuming the Rx location is known only for the initial aligned condition. Fig. 5.7a shows the Rx voltages with misalignment of 3 mm for two cases. It can be seen that SIG-US beamforming achieved 11.9 times higher voltage than conventional beamforming with a misalignment of 3 mm, which is equivalent to 141.6 times in the power. Fig. 5.7b shows that with SIG-US, the normalized received power was fairly constant because the implant location is updated under each misaligned condition, while the received power reduced drastically without SIG-US. It can be seen that with SIG-US, the normalized power only reduced by 1.1 times for 6 mm misalignment while it reduced 108.7 times with the conventional beamforming. When Rx is misaligned 6 mm, the SIG-US shows 95.7 times in normalized received power.

(a)

(b) Fig. 5.7 (a) Comparison of the received Rx voltages with SIG-US to the received Rx voltages with conventional beamforming with misalignment of 3 mm. (b) The improvement of normalized received power by SIG-US over conventional beamforming for misalignment of up to 6 mm.

Conclusions

The proposed SIG-US beamforming can be used in any distributed, addressable PNS recording and stimulation system. The location of the PNS nodes will be automatically updated in real time to ensure high delivered power regardless of the micro-motion of the implanted nodes. The concept of SIG-US beamforming in an ultrasonically interrogated PNS recording platform along with the FEM simulation setup and results for powering a single implant with a linear transducer array were presented. The simulation results show a 5.3 times improvement in received power under aligned conditions compared to the case without any beamforming and a 95.7 times improvement in normalized received power compared to conventional beamforming with a misalignment of 6 mm. For misalignment of up to 6 mm, the SIG-US beamforming showed only 1.1 times decrease in normalized received power while the conventional beamforming reduced 108.7 times.

Chapter 6 Gastric Seed: Towards Distributed Ultrasonically Interrogated Millimeter-Sized Implants for Large Scale Gastric Electrical-Wave Recording

A free-floating gastric recording system with ultrasonic power and communication has been proposed in chapter 2. In this section, the recent effort towards the development of Gastric Seed is presented. A proof-of-concept chip prototype for a Gastric Seed will be presented with emphasis on its novel features of ultrasonic self-regulated integrated power management, low-power pulse- based data transfer, and addressability. In Section 6.1, the chip design will be discussed in detail. The chip measurement results will be presented in Section 6.2, followed by concluding remarks in Section 6.3.

Gastric Seed Chip Architecture

Fig. 6.1 shows the simplified diagram of the proposed Gastric Seed chip, including an analog front- end (AFE), a pulse-based data Tx, and self-regulated power management. The AFE, which is not discussed in this chapter includes a tuned low-noise amplifier (LNA) and a 10-bit successive

Fig. 6.1 Simplified diagram of the proposed Gastric Seed chip interfacing with a pair of stacked power/data ultrasonic transducers. 54 approximation register analog-to-digital converter (SAR-ADC). The chip externally requires one capacitor (CL) and a pair of stacked millimeter-sized power and data ultrasonic transducers.

6.1.1 Self-Regulated Power Management

An integrated power-management block typically requires two large capacitors for rectification and regulation [130], and often an additional large capacitor for over-voltage-protection (OVP). To implement a voltage doubler, another large capacitor is also required. However, in a millimeter- sized implant minimum number of large capacitors should be used. As shown in Fig. 6.1, the proposed power management is in the form of a voltage doubler and can perform rectification, regulation and over-voltage-protection (OVP) in one step with only one off-chip capacitor (CL). A switch controller block generates a switching signal (S) to control the pass transistor (M2) acting as an active switch. Considering the inherent internal capacitance of the power ultrasonic transducer (power Rx), the voltage doubler is implemented by connecting the diode-connected transistor (M1) in parallel with the transducer, technically eliminating the need for 3 large capacitors.

Fig. 6.2 Switch controller schematic diagram and key waveforms of the self-regulated mechanism using reverse current from VL to power Rx.

Fig. 6.2 shows the switch controller schematic diagram along with its key operational waveforms for self-regulation of VL at 3 V utilizing reverse current from CL to the ultrasonic transducer only when VL surpasses 3 V. Inside the switch controller, an amplifier (Amp) controls the bias current

(Ibias) of a comparator (Comp), comparing VL with the received voltage (VP) as in an active rectifier,

55 by amplifying the difference between 0.38×VL and VBGR = 1.15 V. If VL < 3 V, the amplifier outputs low and Ibias is maximized. Therefore, the comparator operates at its highest speed to maximize the forward current flowing from the ultrasonic transducer to CL when VP > VL and block the reverse current when VP < VL. This ensures highest power conversion efficiency (PCE) in charging

CL. When VL surpasses 3 V, the amplifier reduces Ibias that slows down the comparator during turn- off (increasing S pulse width) and allows reverse current to flow from CL to the transducer (also loading the transducer), reducing VL as shown in the Fig. 6.2 inset. This iterative process self- regulates VL at 3 V.

6.1.2 Addressable Pulse-Based Data Communication

To demonstrate the concept of addressable Gastric Seeds, the chip in Fig. 6.1 was designed to externally be assigned two IDs, “0” and “1”. To enable IDs “0” and “1”, the power carrier is modulated with a 20 s activation notch followed by nothing or a second 20 s notch after 5 s, respectively. The number of IDs can be easily scaled up by using similar amplitude modulation of the power carrier with more bits. Upon activation, the Gastric Seed transmits 15 bits in the form of sharp pulses with 140 ns width for high bits and nothing for low bits (on-off keying modulation).

These 15 bits include one bit for supply check (high: VL > 2.8 V), 3 start bits (“101”), ten digitized recorded data bits and one end bit (high).

Fig. 6.3 shows the schematic diagram and key operational waveforms of the address (ID) check block, in which VP envelope is first recovered. Then, the rising edge of the activation notch (X), triggering FF1 flip-flip, is delayed by 15 µs (Y) to trigger FF2, detecting the address (addr) command of either “1” or “0”. When addr is high, Gastric Seed with ID = “0” is enabled by setting

S0_EN low and consequently Tx_EN0 high as shown in the Fig. 4 waveforms. Similarly, Gastric

Seed with ID = “1” is enables when addr is low (Tx_EN1: high). The START and RST signals in Fig. 4 are generated internally at the beginning and end of each 15-bit data frame.

Measurement Results

Fig. 6.4 shows the die micrograph of the proof-of-concept Gastric Seed chip, fabricated in a 0.35- µm 2P4M CMOS process, occupying 0.6 mm2 and 0.4 mm2 with and without pads, respectively.

The chip operates at the power carrier frequency (fp) of 1 MHz to regulate VL at 3 V (CL = 10 µF).

Fig. 6.3 Schematic diagram and key operational waveforms of the proof-of-concept addressable data transmitter with IDs “0” and “1”.

Fig. 6.5 shows the measurement setup. The ultrasonic transducers (PZT-5A, APC International, Mackeyville, PA) with desired resonance frequency of ~1 MHz for power/data transfer were held perfectly aligned with a separation of 3.75 cm inside a water tank, made of Plexiglas. The proof- of-concept stacked transducers for data-Tx (diameter: 1.1 mm, thickness: 0.2 mm) / power-Rx (diameter: 1.1 mm, thickness: 0.9 mm, parasitic capacitance: 37.5 pF) were assembled on a PCB and located at the tank’s bottom surface with a 3D-printed holder. The power Tx (Data Rx) transducer (diameter: 15.9 mm, thickness: 2 mm, resistance at resonance: ~400 Ω) was glued to an adjustable 3D-printed plastic stand that could move in all three axes (Fig. 6). Transducers were encapsulated by biocompatible Sylgard-184 with the same acoustic properties as water.

Fig. 6.6a shows the measured PCE and voltage conversion efficiency (VCE = VL / 0.5VP,peak-peak) of the power management at different DC load resistances (RL,DC) with no reverse current. The highest PCE of 83.8% at RL,DC of 30 kΩ (~200 µW of power consumption) was achieved. For regulation, the PCE is intentionally reduced with reverse current when VL surpasses 3 V. Thanks to the proposed voltage-doubler method, a high VCE of 1.7 V/V was achieved for most RL,DC values. Fig. 6.6b shows the power transfer efficiency (PTE) of the ultrasonic link at different AC

57 load resistances (RL,AC). The highest PTE of 0.52% was achieved at RL,AC = 8 kΩ. Considering average PCE of 80% and VCE of 1.7 V/V, RL,DC is 7-8 times higher than the equivalent RL,AC.

Therefore, at RL,DC of 30 kΩ (RL,AC ≈ 4 kΩ) the total link (transducers + chip) efficiency is ~0.838×0.47% ≈ 0.4%.

Fig. 6.4 Gastric Seed chip micrograph occupying 0.6 mm2 including pads.

Fig. 6.5 Measurement setup with ultrasonic transducers perfectly aligned with 3D printed plastic holders.

Fig. 6.7 shows the key waveforms of the proposed power management during start-up when the external power Tx voltage (VS) was suddenly increased from zero to 8.5 Vp-p (peak-to-peak). As shown in Fig. 6.7, VL was first quickly charged to 1.6 V in a passive manner at which the chip became fully functional and charged VL to 3 V with self-regulation.

(a) (b) Fig. 6.6 (a) Measured PCE and VCE of the power management at different RL,DC. (b) PTE of the ultrasonic link at different RL,AC.

Fig. 6.7 Measured VL and VP waveforms during start-up when external VS (power Tx voltage) was increased from zero to 8.5 Vp-p.

Fig. 6.8 shows the measured waveforms for the self-regulated power management in response to a sudden increase in the transmitted power (proportional to power Tx voltage, VS). In Fig. 6.8a, when VS was increased from 11 Vp-p to 18 Vp-p, the chip adaptively adjusted the width of the S pulses to maintain the VL constant at 3 V with more frequent reverse currents. As shown in the zoomed waveforms in Fig. 6.8b, since the power Rx transducer has a large impedance of ~5 kΩ at

1 MHz, limiting the reverse current, at high VS of 18 Vp-p, the chip kept M2 on (S: low) for several power-carrier cycles to directly load the transducer with CL and limit the received power, regulating VL at 3 V. Fig. 6.8 also shows the proper operation of the voltage doubler by providing a DC shift across VP. At VL = 3 V, VL ripples and line regulation were measured 3.4 mV and 0.23%, respectively.

Fig. 6.9 from top to bottom shows the measured modulated VP, Tx_EN0, Tx_EN1 and transmitted data pulses (Tx Data). In this measurement, VP was modulated for both IDs “0” and “1”. When

59 addr = 1 (only activation notch) was detected, the Gastric Seed with ID “0” was enabled (Tx_EN0: high) to transmit data pulses and vice versa.

(a)

(b) Fig. 6.8 (a) Measured VL and VP waveforms when VS was increased from 11 VP-P to 18 VP-P. (b) Zoomed waveforms of VL and VP (CL = 10 µF).

Fig. 6.10 shows measured pulse-based data transmission (15 bits) at the rate of 75 kbps using a pair of ultrasonic transducers spaced by 3.75 cm inside water, mimicking tissue. The width of each transmitted pulse was ~140 ns. The chip power consumption was measured 63 µW and 30 µW with and without data transmission at 75 kbps, respectively, resulting in (63 µW - 30 µW) / 75 kbps = 440 pJ/bit energy consumption of the data Tx. Each pulse generated a ~15 µs voltage ringing at 1 MHz across the Rx data transducer. Since SWs are low frequency (10-500 mHz) signals, in the current prototype with 75-kbps transmission of 15 bits of data and additional 10 bits

60 of address ID for each Gastric Seed, at least ~1000 Gastric Seeds can be operated continuously if each is sampled at 3 Hz.

Fig. 6.9 Measured waveforms for the pulse-based data transmitter showing how addressable Gastric Seeds (IDs “0” and “1”) can be activated.

Fig. 6.10 Measured transmitted and recovered data bits at 75 kbps.

Conclusions

The proposed distributed millimeter-sized Gastric Seeds hold the promise of large-scale gastric SW recording with minimal damage. A proof-of-concept power management / data Tx chip for the Gastric Seeds was presented. The prototype chip includes a self-regulated power management that performs rectification, regulation and OVP in one step using only one off-chip capacitor as well as an addressable pulse-based data Tx with measured data rate of 75 kbps and energy consumption of 440 pJ/bit using a pair of 1 MHz ultrasonic transducers spaced by 3.75 cm in water. The future work is to extend the number of Gastric Seeds and demonstrate fully wireless recording capability with a robust ultrasonic interrogation platform.

Chapter 7 Ultrasonically Powered Wireless System for In Vivo Gastric Slow-Wave Recording

A proof-of-concept chip prototype of a Gastric Seed has been presented in chapter 6 with emphasis on its novel features of ultrasonic self-regulated integrated power management, low-power pulse- based data transfer, and addressability. In this chapter, the in vivo validation of the Gastric Seed chip prototype integrated with a COTS amplifier and microcontroller (MCU) with built-in analog- to-digital converter (ADC) and radio-frequency (RF) transceiver will be presented to record the SWs from an anesthetized rat.

In Section 7.1, the architecture of the system prototype will be discussed in detail. The in vivo results will be presented in Section 7.2, followed by concluding remarks in Section 7.3.

System Architecture

Fig. 7.1 shows a simplified diagram of the proposed recording system, including the prototype chip, a COTS two-stage amplifier, and an MCU with a built-in ADC, clock generator, and RF transmitter (Tx) to transmit the SW data for storage on a computer. The system also requires a pair of stacked power/data ultrasonic transducers with 1.2 mm diameter. The Fig. 2 inset shows a proof- of-concept of this transducer stack assembled on a printed circuit board (PCB).

Fig. 7.1 Simplified diagram of the recording prototype interfacing with a pair of stacked power/data ultrasonic transducers.

7.1.1 Gastric Seed Chip Prototype

As explained in detail in Chapter 6, the chip includes self-regulated power management and an addressable pulse-based data transmitter (Tx). As shown in Fig. 7.1, the power management is in the form of a voltage doubler and can perform rectification, regulation, and over-voltage protection

(OVP) in one step with only one off-chip capacitor (CL), eliminating the need for three external capacitors, which is crucial for a millimeter-sized implant. The chip has a self-regulated VL of 2.5 V.

To demonstrate the concept of addressable Gastric Seeds, the chip in Fig. 7.1 was designed to externally be assigned one of two IDs, “0” or “1”. To enable IDs “0” or “1”, the power carrier is modulated with a 20 s activation notch followed by either no notch or a second notch after 5 s, respectively. Upon activation, the Gastric Seed transmits 12 bits in the form of sharp pulses digitized by the ADC in the Tx MCU with a 140-ns width for high bits and nothing for low bits (on-off keying modulation). These 12 bits include 1 start bit (high), 10 recorded data bits and one end bit (high). More details on the chip can be found from prior work in Chapter 6.

In this proof-of-concept system, the data transmission is activated when an ID of “1” is detected, i.e., the Gastric Seed with ID = “1” is enabled. Fig. 7.2 shows the schematic diagram and key operational waveforms of the address (ID) check block, in which the envelope of the received ultrasonic power carrier (VP) is first recovered. Then, the rising edge of the activation notch (X), triggering flip-flop FF1, is delayed by 15 µs (Node_En) to trigger FF2, detecting the address

(Addr). When Addr is low, Gastric Seed with ID = “1” is enabled by setting S1_EN low and consequently Tx_EN1 high as shown in the Fig. 7.2 waveforms. The START and RST signals in Fig. 7.2 are generated internally at the beginning and end of each data frame

The Tx MCU generates a clock of 256 kHz, which is fed into the chip. The chip divides it into a 16 kHz clock for ultrasonic data pulse generation and an 800 Hz clock (Clock signal in Fig. 7.2) for the chip and Tx MCU synchronization. When the chip is enabled (Node_En: high), the Tx MCU sends the 12-bit digitized data (ADC + start/stop bits) to the chip at the first rising edge of the Clock for ultrasonic data transmission within t2-t3 time period in Fig. 7.2. Then the Tx MCU sends the data through its RF link at the second rising edge of Clock (t4-t5 in Fig. 7.2). As shown

63 from the X signal, after the activation notch the ultrasonic power carrier is reduced to zero by the external power Tx for ~2 ms to avoid interference between ultrasonic power and data transmission.

Fig. 7.2 Schematic diagram and key operational waveforms of the addressable data transmitter with ID = “1” enabled. All signals are generated inside the chip based on a 256 kHz clock from the Tx MCU. 7.1.2 Commercial-Off-The-Shelf (COTS) Components

Since the current chip doesn’t include an amplifier and ADC, in this proof-of-concept system the SW signals are fed to a two-stage amplifier made with COTS op-amps (OPA2333, Texas

Instruments, Dallas, TX) as shown in Fig. 7.1. The first stage provides a mid-band gain of C1/C3

= 40 dB (C1 = 47 µF, C3 = 0.47 µF) and band-pass filtering within the 10 mHz to 9.4 Hz band

(R3 = 33 MΩ, R4 = 3.6 kΩ), and the second stage provides a mid-band gain of C4/C5 = 20 dB (C4

= 4.7 µF, C5 = 0.47 µF) and band-pass filtering within the 10 mHz to 1.6 Hz band (R5 = 33 MΩ,

R6 = 1 MΩ, C6 = 0.1 µF). Therefore, a total gain of 60 dB with a 3-dB bandwidth of 18 mHz to 500 mHz is implemented.

The amplified SW signal is fed into the ADC input of the Tx MCU (nRF24LE1, Nordic Semiconductor, Norway), where it is digitized. The 2.4-GHz Tx MCU wirelessly sends the digitized data to another nRF receiving (Rx) MCU on the stationary unit. The digitized signal can be read on a PC through an RS232-USB connection.

Benchtop and In Vivo Experiment Results

The wireless system prototype in Fig. 7.1 was first validated with benchtop measurements. Fig.

7.3 from top to bottom shows the measured modulated VP, Node_En, transmitted data pulses

(Tx_Data) and recovered data (Rx_Data) directly from the Rx transducer. In this measurement, VP was modulated for a Gastric Seed ID of “1”. Power transmission was cut for 2 ms to prevent interference between the power and data links. When the Gastric Seed with ID “1” was activated, 12 bits of pulse-based data were transmitted at the rate of 16 kbps through a pair of ultrasonic transducers spaced by 6 cm inside water, mimicking tissue. The width of each transmitted pulse was ~140 ns, which resulted in 440 pJ/bit energy consumption. Each pulse generated a ~15 µs voltage ringing at 1 MHz across the Rx data transducer.

Fig. 7.3 Measured waveforms for the pulse-based data transmitter showing transmitted and recovered data bits at 16 kbps when Gastric Seed with ID “1” was activated.

It should be noted that the data Rx placed in the same tank with the high-voltage power Tx will pick up power signals both directly from the power Tx and indirectly from reflections by the water tank. In order to show the concept of time-multiplexed pulse-based ultrasonic wireless power/data transmission, in this measurement another pair of ultrasonic transducers in a separate tank was used to power the system. In the ongoing effort, data Rx circuits that disable the data Rx when power is on and only enable the data Rx when power is cut for data transmission is being developed.

To validate the system in a realistic condition, in vivo experiments have been conducted to acquire SWs in rats. For the in vivo experiment on an anesthetized rat, only ultrasonic power transmission with RF data transmission was implemented as shown in Fig. 7.4. Prior to the experiment, ethical approval for the in vivo study was granted by the NYIT Institutional Animal Care and Use Committee. A male rat of Sprague-Dawley breed weighing 350 grams was used.

Fig. 7.4 In vivo experiment setup for wireless recording of gastric SWs. Wireless power and data transfer were established with ultrasound and RF links, respectively.

The rat was anaesthetized with 4% isoflurane in 100% O2 for surgical induction and maintained supine on a heating blanket. The abdomen was wiped down with alcohol and betadine before incision to expose recording sites. Copper wire (AWG28) electrodes were secured in place at the corpus of the stomach with a fine suture. The ground wire (reference signal) was placed along the incision site in the connective tissue. Anesthesia was reduced to 1.5% isoflurane for recording of gastric activity.

As shown in Fig. 7.4, the RF communication distance was ~40 cm. The ultrasonic transducers for power transfer were held perfectly aligned with a separation of 6 cm inside a Plexiglas water tank,. The stacked data-Tx/power-Rx transducers (Fig. 7.1) were located at the bottom surface with a 3D-printed holder. The power Tx transducer was glued to an adjustable 3D-printed plastic stand. All transducers were encapsulated by Sylgard-184, which has similar acoustic properties to water. The ultrasonic link only powered the chip and the amplifier while the power-hungry Tx MCU was powered by an external supply.

In the first experiment, the output of the amplifier was recorded for approximately 8 minutes using an oscilloscope. Figs. 7.5a and 7.5b show the raw and filtered (low pass) signals, respectively. In Fig. 7.5, 11 SW events can be counted, which corresponds to a frequency of 1.375 cycle-per- minute (cpm) with maximum and minimum amplitudes of 1.79 mV and 0.46 mV peak-to-peak, respectively.

(a)

3 2 4 5 7 9 11 8 1 6

10 Voltage(mV)

Time (s) (b) Fig. 7.5 (a) Unfiltered direct oscilloscope probe measurement of the amplified SW signal at the amplifier output with sampling rate of 50 samples/s. (b) Filtered oscilloscope probe measured SW signal with 500 points averaging showing frequency of 1.375 cpm.

In the second experiment, the output of the amplifier was digitized by the 10-bit ADC and transmitted wirelessly through the RF link with a sampling rate of 200 samples/s. Figs. 7.6a and 7.6b show the raw and filtered signals. The recorded SWs showed 12 cycles during 8 minutes, which corresponds to a frequency of 1.5 cpm with maximum and minimum amplitudes of 1.03 mV and 0.28 mV peak-to-peak, respectively. The signals recorded in the two experiments represent similar wave frequency and morphology. The slight differences are due to the non- simultaneous measurements.

(a)

(b) Fig. 7.6 (a) Unfiltered SW signals recorded through RF link with sampling rate of 200 samples/s. (b) Filtered RF recovered SW signal with 2000 points averaging showing frequency of 1.5 cpm.

The same experiments on a second rat has been done. Fig. 7.7a and 7.7b show the filtered signals recorded by direct scope measurement and RF link measurement, respectively for 4 minutes. In Fig. 7.7a, the scope recorded SWs showed 15 cycles during 4 minutes, which corresponds to a frequency of 3.75 cpm with maximum and minimum amplitudes of 2.41 mV to 0.69 mV peak-to- peak, respectively. Fig. 7.7b shows the RF transmitted SWs, it showed 16 cycles during 4 minutes, which corresponds to a frequency of 4 cpm with maximum and minimum amplitude of 2.17 mV to 0.23 mV peak-to-peak, respectively.

Conclusions

A proof-of-concept ultrasonically powered system prototype implementing a power management/data Tx CMOS chip with COTS amplifiers, ADC, and RF transmission link was presented for an in vivo experiment on two anesthetized rats. The CMOS chip includes self- regulated power management with only one off-chip capacitor as well as an addressable pulse-

(a)

(b) Fig. 7.7 (a) Filtered oscilloscope probe measured SW signal with 500 points averaging showing frequency of 3.75 cpm. (b) Filtered RF recovered SW signal with 2000 points averaging showing frequency of 4 cpm. based data Tx with a measured data rate of 16 kbps and energy consumption of 440 pJ/bit using a pair of 1 MHz ultrasonic transducers spaced by 6 cm in water. SW signals were successfully recorded using ultrasonic power transfer through water to power the chip and amplifier. On the first rat, the SW recorded by oscilloscope and RF link showed similar frequencies of 1.375 cpm and 1.5 cpm, respectively, and maximum amplitudes of 1.79 mV and 1.03 mV peak-to-peak, respectively, and minimum amplitudes of 0.46 mV and 0.28 mV peak-to-peak, respectively. On the second rat, the SW recorded by oscilloscope and RF link showed faster but similar frequency of 3.75 cpm and 4 cpm, respectively, and maximum amplitudes of 2.41 mV and 2.17 mV peak-to- peak, respectively, and minimum amplitudes of 0.69 mV and 0.23 mV, respectively.

Chapter 8 Multi-Beam Shared-Inductor Reconfigurable Voltage/SECE-Mode Piezoelectric Energy Harvesting of Multi-Axial Human Motion

Other than using lead zirconate titanate (PZT) for wireless power/data transmission as shown in previous chapters, PZT can also be used for kinetic energy harvesting which allows power extraction from motion. As stated in Chapter 1, piezoelectric energy harvesters (PEHs) are attractive for energy harvesting from human motion due to the higher power density and scalability. Fig. 8.1 shows the generic block diagram of a piezoelectric energy harvesting system.

It includes a PEH modelled with its simplified electrical equivalence (electrical current IP and parallel internal capacitance/resistance CP||RP), a power-management circuit for efficient conversion of the AC signal across the PEH to a usable DC voltage across a storage capacitor

CSTORE, and the sensor node. This chapter focuses on the design of the power-management circuit to maximize the harvested power generated by the PEH and store the energy in CSTORE. The sensor node, consisted of a regulator and sensor core, can be designed depending on the application needs and is modeled as an equivalent DC load RL for simplicity in this Chapter.

Fig. 8.1 Generic block diagram of a piezoelectric energy harvesting system. A power-management circuit is used to convert the AC voltage across the PEH to a usable DC voltage across a storage capacitor CSTORE. For simplicity, the sensor node has been modeled with a DC load RL in this paper.

Although several unidirectional PEHs in the form of a single cantilever beam have been proposed and developed in the past [43]-[49], they suffer from several challenges in energy harvesting from body motion, including the presence of multi-axial motion, irregular frequencies, and unpredictable voltage levels often with low amplitudes [41]. To address these challenges, collaborators at Penn State and the University of Utah have recently developed wrist-worn eccentric rotor-based inertial PEHs specifically for energy harvesting from body motion, as shown

70 in Fig. 8.2 [41]. These PEHs utilize multiple magnetically or mechanically plucked flexible thin- film PZT/nickel (Ni)/PZT beams to significantly increase harvested energy within a small volume. These multi-beam PEHs convert multi-axial body motion into AC voltages with different phases and shock-like decaying amplitudes (up to several volts) within the frequency range of 90-160 Hz (magnetic plucking) and 230-270 Hz (mechanical plucking) across each beam [41].

(a) (b) Fig. 8.2 Wrist-worn multi-beam piezoelectric energy harvesters (PEHs) for harvesting energy from body motion with multi-axial movements. (a) The magnetic-plucking prototype used inthe previous work in [59]. (b) The mechanical-plucking prototype with more robustness used in this chapter. State-of-the-art power-management schemes and chips for PEH interfacing utilize full-wave active rectifier (FAR) [131], synchronized switch harvesting on inductor (SSHI) [132], energy investment [133], synchronous electrical charge extraction (SECE) [134]-[136], and sense-and-set (SaS) [137] methods. However, they suffer from multiple shortcomings: 1) they can only interface with one beam, 2) SSHI and SECE inherently suffer from large power loss in harvesting shock responses and large inputs, respectively, 3) SECE with passive negative voltage converters (NVCs) cannot harvest low PEH voltages, which often occur from body motion, and 4) the SaS circuit cannot tolerate frequency variations since its operation depends on predetermined timings and its input is limited to the CMOS voltage limit. Therefore, none of these techniques are optimal for interfacing with the multi-beam inertial PEHs in Fig. 8.2.

In this chapter, a new power-management scheme with either voltage-mode or SECE (VM-SECE) operation as well as its chip for the inertial PEHs is proposed (see Fig. 8.2) with the unique capabilities of 1) simultaneously harvesting energy from up to 6 piezoelectric beams in a modular fashion using a single shared off-chip inductor, 2) seamlessly reconfiguring itself between operating as an efficient FAR (voltage mode, VM) or SECE (i.e., reconfigurable VM-SECE) to improve the overall efficiency, extract maximum energy, and protect the chip against large inputs, eliminating the need for off-chip components as in [134] or high-voltage processes as in [135], 3) extending the input-voltage range to as low as 35 mV by utilizing active NVCs, 4) adaptive SECE

71 operation (peak and time zero-crossing detection) to dynamically compensate for the variations of beams’ electrical impedance, voltage, and frequency, and 5) optimal cold start (self-starting) with an active voltage doubler with a small footprint.

The multi-beam wrist-worn PEHs will be discussed in Section 8.1, followed by the VM-SECE operation and modeling in Section 8.2. The proposed chip architecture and circuit design will be described in Section 8.3, followed by the measurement setups and results in Section 8.4 and the concluding remarks in Section 8.5.

Wrist-Worn Multi-Beam Inertial Piezoelectric Energy Harvester (PEH)

Conventional translational PEHs are designed to be unidirectional, while human motion contains a high amount of multi-axial movements [41]. In addition, human motion is typically at low and irregular frequencies (~1 Hz) [138] and, therefore, resonant PEHs operating at much higher frequencies cannot directly benefit from peak dynamic magnification. In contrast, the eccentric rotor-based rotational design uses a rotor to excite higher-frequency resonance in beams via a low- frequency motion, as a rotor does not have direction or motion limits. The rotor can also enable multi-beam excitation [41]. Fig. 8.2a and 8.2b shows wrist-worn multi-beam PEHs based on this design with magnetic and mechanical plucking, respectively. Custom fabricated bimorph PZT/Ni/PZT thin-film (60 µm in thickness) beams are used to combine flexibility with strong piezoelectric response. Multi-beam design is implemented to achieve higher power density [139].

In Fig. 8.2a, the beams are magnetically plucked by placing a small magnet on the tip of each beam. The PZT beams are magnetically deflected and ring down at their natural frequency to achieve contactless actuation. Such non-contact frequency up-conversion via magnetic plucking is a promising solution. However, this technique faces several challenges. 1) The large mass at the end of a very flexible beam presents a failure risk in the presence of shocks. 2) the magnetic interaction between the two permanent magnets (placed at the beam tip and on the rotor) creates a detent torque on the rotor as it tries to move past the beam. Therefore, there exists a critical excitation strength below which the harvested power is negligibly small. And 3) the assembly of the electrical connections is constrained to a minimal area near the center of the device, which has led to low yield and unreliable electrical connections. This PEH was used in the prior work for interfacing with the VM-SECE chip [59].

To overcome these issues, a similar device with mechanical plucking was developed recently, as shown in Fig. 8.2b. The beams are excited by the pins on the rotor. This redesign removes the need for large inertia (i.e., magnet mass) at the end of the beam. This arrangement enables re-orientation of the beams such that the free end of the beam is at the center rather than the perimeter of the device, allowing the use of more standard electrical connectors and greatly simplify the device assembly. Furthermore, re-orienting the beams reduces the detent torque, which is proportional to the lever arm between the center of rotation and the point of plucking. Overall, this improves the robustness and yield of the device.

In this Chapter, the mechanically plucked PEH in Fig. 8.2b with 5 functional beams was used. Compared to the magnetically plucked PEH in Fig. 8.2a, 1) the resonance frequency of each beam increased from ~100 Hz to ~250 Hz due to the tip mass removal, and 2) the power generated by each beam for the same excitation was reduced because the pins on the rotor barely impact the PZT beams, reducing the beams displacement amplitude (i.e., a tradeoff between robustness and power generation).

Operation and Modeling of the VM-SECE Scheme

Fig. 8.3 shows the simplified circuit schematic of the proposed reconfigurable VM-SECE scheme for the single-beam operation. It includes an active NVC followed by an active voltage rectifier

(creating a FAR for VM operation) and a conventional SECE circuit in parallel to charge CSTORE to VSTORE. The voltage across the beam (VP) is first full-wave rectified by the active NVC, generating VREC. As long as peak VREC < VSTORE, the active rectifier is off (P1: OFF), and CP is

Fig. 8.3 Simplified circuit schematic of the proposed reconfigurable VM-SECE power-management scheme for the single-beam operation. 73 charged by IP to operate the circuit in the SECE mode. When peak VREC > VSTORE, the active rectifier turns on and the circuit seamlessly transitions from VM to SECE operation.

Fig. 8.4a shows key operational waveforms of the VM-SECE scheme in only conventional SECE mode (peak VREC < VSTORE). IP first charges CP until VREC reaches its peak detected by a peak- detection circuit (PKD). Then at VREC peak, when maximum energy is stored in CP, CP energy is transferred to an external inductor LEXT for a short time period via enabling SW1 (transmission gate

TG: ON) and SW2 (N1: ON) and disabling SW3 (N2: OFF) until VREC reduces to zero, which is detected by a time zero-crossing detection (ZCD) circuit. Then LEXT energy is transferred to CSTORE via a diode (diode-connected P2) by disabling SW1 and SW2 (TG, N1: OFF), and enabling SW3 (N2: ON).

(a)

(b) Fig. 8.4 Key operational waveforms of the proposed VM-SECE scheme. (a) SECE only when peak VREC < VSTORE. (b) Reconfigurable VM-SECE when peak VREC > VSTORE.

In the SECE mode (Fig. 8.4a), the energy stored in CP at peak VREC is given by,

E = 1 C (2V )2 = 2C V2 , (8.1) SECE-only 2 P P,OC P P,OC where VP,OC is the VP peak when the PEH is not loaded (PEH open circuit voltage). The VP peak in (8.1) is 2VP,OC because in the SECE mode CP is discharged by LEXT instead of IP. In other words, 74

SECE-only operation imposes very high voltages on the PEH and the interface circuit. Due to the full-wave rectification, the frequency of energy extraction is twice higher than the PEH excitation frequency fP. Therefore, the stored power in SECE-only mode can be calculated as,

2 PSTORE,SECE-only = 2fPESECE-only = 4CPVP,OC fP. (8.2)

Fig. 8.4b shows the key operational waveforms in the VM-SECE mode (peak VREC > VSTORE).

When VREC slightly surpasses VSTORE, P1 turns on and charges CSTORE with high efficiency via IP

(VM path) until IP reaches zero. This is similar to a FAR operation. As VREC goes slightly below

VSTORE, P1 is turned off, and the SECE circuit is triggered to extract the remaining stored energy in

CP as explained before in Fig. 8.4a. Therefore, peak VP always stays close to VSTORE.

The output power of the FAR can be found from [140],

PSTORE,FAR = 2fPCPVSTORE[2VP,OC - (VF - VI)], (8.3) where VF and VI are the final and initial voltages of VP (input) in a half cycle, respectively. For the FAR, VI = -VSTORE and VF = VSTORE and, therefore, the stored power can be found from,

PSTORE,FAR = 4fPCPVSTORE[VP,OC - VSTORE]. (8.4)

The VM operation is also similar to the FAR operation but due to the SECE operation following

VM in Fig. 8.4b, CP is charged by IP only from 0 to ± VSTORE, i.e., VI = 0 and VF = VSTORE. Therefore, the output power obtained by the VM operation can be calculated as,

PSTORE,VM = 2fPCPVSTORE[2VP,OC - VSTORE]. (8.5)

Right after the VM operation, the remaining energy in CP (VP ≈ VSTORE) is extracted by SECE and, therefore, the stored power in this time period can be calculated from,

P = 1 C V2 (2f ) = C V2 f . (8.6) STORE,SECE 2 P STORE P P STORE P

Finally, using (8.5) and (8.6) the total stored power during the VM-SECE operation can be calculated as,

PSTORE,VM-SECE = PSTORE,VM + PSTORE, SECE

= fPCPVSTORE[4VP,OC - VSTORE]. (8.7)

The damping loss (RP) is neglected in all these equations.

Fig. 8.5a and 8.5b shows the simulated and calculated PSTORE and simulated peak VREC vs. VSTORE, respectively, for the FAR, conventional SECE and proposed VM-SECE scheme under ideal lossless condition with CP = 10 nF, RP = ∞, LEXT = 2.2 mH, VP,OC = 4.71 V, and fP = 100 Hz. For

PSTORE calculations, (8.2), (8.4), and (8.7) were used for the conventional SECE, FAR, and VM- SECE, respectively. For simulations, the circuits were built in the Cadence Spectre circuit simulator (Cadence Technology, San Jose, CA, USA) with ideal components.

(a)

(b) Fig. 8.5 (a) Simulated and calculated PSTORE vs. VSTORE for the FAR, conventional SECE and the proposed VM-SECE. (b) Simulated peak VREC vs. VSTORE for the proposed VM-SECE and conventional SECE. Ideal lossless components were used in both calculations and circuit simulation.

Three lessons can be learned from Fig. 8.5a. 1) The obtained PSTORE by VM-SECE and FAR are close at low VSTORE < 2 V because a negligible amount of energy is wasted in charging CP. As

VSTORE increases, more energy is wasted with FAR operation while VM-SECE can extract all the remaining energy inside CP via the SECE operation. 2) When VSTORE > 2VP,OC = 9.42 V, the VM-

SECE scheme reconfigures itself to the SECE mode providing the same PSTORE as in conventional

SECE. Nonetheless, for VSTORE < 9.42 V conventional SECE achieves higher PSTORE with ideal components. And 3) the calculation and simulation results match very well, validating the accuracy of the models.

Although the conventional SECE achieves higher PSTORE for most VSTORE values with ideal components (Fig. 8.5a), as shown in Fig. 8.5b the peak VREC in conventional SECE is significantly high (9.42 V in these simulations) for all VSTORE values. However, employing VM operation in the

VM-SECE scheme limits the peak VREC at ~VSTORE that provides inherent and efficient over-voltage protection (OVP) and eliminates the need for circuit implementation in high-voltage processes.

Table 8.1 Circuit Parameters Used for VM-SECE and SECE Simulation

To provide a fair comparison between conventional SECE and the proposed VM-SECE with lossy components, circuit simulations were done for these circuits (Fig. 8.3) with the circuit parameters listed in Table 8.1 (transistors in a 0.35 µm CMOS process) for VP,OC = 2.67 V at the supply voltage

VDD of 3 V. Fig. 8.6a and 8.6b shows the simulated PSTORE and peak VREC vs. VSTORE, respectively, for the proposed VM-SECE and conventional SECE using lossy components listed in Table 8.1.

It can be seen that the VM-SECE scheme outperforms the conventional SECE for all VSTORE values in terms of achieving higher PSTORE with much lower VREC until VSTORE ≈ 3.8 V. In these simulations, the VM-SECE performance degrades at VSTORE ≥ 3.8 V because the PMOS transistor of TG in Fig. 8.3 starts to leak as VREC increases higher than VDD = 3 V by a threshold voltage of ~0.8 V. Nonetheless, the simulation results in Fig. 8.6 shows the advantage of the proposed VM-

SECE scheme compared to conventional SECE in providing both higher PSTORE and inherent OVP.

Multi-Beam Reconfigurable VM-SECE Chip Architecture

Fig. 8.7 shows the block diagram of the proposed chip. It includes 1) six identical active rectifiers with active NVCs (FAR for VM path: one per beam) connected in parallel to allow simultaneous VM operation (due to the low-frequency operation, the reverse current is negligible and, therefore, 77

(a) (b) Fig. 8.6 Simulated comparison (a) PSTORE and (b) peak VREC vs. VSTORE for the proposed VM-SECE scheme and the conventional SECE with lossy circuit components listed in Table 8.1. The circuit schematic is shown in Fig. 8.3. low -power comparators without offset compensation have been designed), 2) a shared SECE circuit for all six beams to reduce the number of off-chip inductors from six to one, 3) a control block for optimal switching in different modes, and 4) a voltage doubler interfaced to an additional th 7 beam for optimal cold start. The chip externally requires one inductor LEXT, one storage capacitor CSTORE, and one cold-start capacitor CCHIP, which with CP,7 forms a voltage doubler

(eliminating one external capacitor; small footprint) to supply the chip’s internal circuitry via VCHIP. As shown in Fig. 8.7, a conventional dynamic body-bias circuit with two PMOS transistors (similar

Fig. 8.7 Block diagram of the proposed multi-beam shared-inductor reconfigurable VM-SECE chip with cold start.

78 to [141]) in both active rectifiers and voltage doubler connects the bulk of the PMOS pass transistors to the highest potential between their source and drain, removing undesired currents.

Fig. 8.8a shows the key waveforms of the VM-SECE chip for asynchronous inputs, which is the case for the inertial PEHs in Fig. 8.2. The chip does VM operation automatically and simultaneously for all beams in parallel. It also sweeps the beams one by one every 60 µs to check their VP for generating optimal switching signals for the SECE operation whenever the condition is met as described in Fig. 8.4 (either VREC reaches its peak without VM operation or at the end of VM operation). Fig. 8.8b shows similar waveforms for synchronous inputs (worst-case scenario), in which the shared SECE circuit is time multiplexed between beams. This is possible with minimal loss because full SECE operation requires only 50 µs, which is much faster than the low- frequency inputs.

(a)

(b) Fig. 8.8 Key operational waveforms of the proposed VM-SECE chip for (a) asynchronous and (b) synchronous inputs (beams 1 and 6 in this example). Fig. 8.9 shows schematic diagrams of active NVC and control block including multi-beam sweep, peak and time zero-crossing detection (PKD & ZCD), and switching control circuits. In the NVC in Fig. 8.9a, a comparator switches pass transistors to actively full-wave rectify the input (VP,n; n = 1-6) with minimal voltage drop.

(a)

(b)

(c)

(d) Fig. 8.9 Detailed schematic diagrams of the (a) NVC, (b) multi-beam sweep control, (c) peak and time zero-crossing detection, and (d) switch control.

In the multi-beam sweep circuit in Fig. 8.9b, beams’ voltages (VP,n) are checked one by one every

60 µs in a loop by 6 cascaded beam-control blocks. There are 4 possible conditions: 1) CP,n is charging and VREC,n < VSTORE, 2) CSTORE is charging via VM and VREC,n > VSTORE, 3) VM operation ended (VVM,n in Fig. 8.7 is high) with some charges left on CP,n, and 4) VREC,n reaches its peak and is still smaller than VSTORE. Under conditions one and two (PKDn in Fig. 8.7 is low), beam-n is skipped through the flip-flop FF2,n followed by 60 µs delay to check the next beam. Under

80 conditions three and four (PKDn: high), at which SECE should immediately start, the beam control waits until SECE operation ends (SW1,n goes low) to clock FF1,n and trigger the next beam after 60 µs.

Fig. 8.9c shows the PKD & ZCD circuit [134]. A positive edge is generated at VREC,n peak and combined with VVM,n to clock a flip-flop for generating high PKDn (indicating start of SECE), which is reset to low at zero-crossing (ZCDn: high) of VREC,n. Finally, Fig. 8.9d shows the switching control that generates required six SW1,n, SW2 and SW3 signals based on the waveforms in Figs. 8.4 and 8.8.

Measurement Results

A proof-of-concept VM-SECE chip was fabricated in a 0.35-μm 2P4M standard CMOS process, occupying 1.9 mm2 active area (Fig. 8.10). The chip was extensively characterized with measurements in benchtop settings, with a commercial single-beam PEH on a shaker, and with the mechanical plucking 5-beam PEH in Fig. 8.2b on a robotic swing arm. In all measurements, LEXT = 2.2 mH (LPS5030-225MR, Coilcraft Inc., Cary, IL, USA) with measured dimensions of 4.78 mm × 4.78 mm × 2.87 mm, CSTORE = 47 µF, and CCHIP = 10 µF were used.

Fig. 8.10 The proposed VM-SECE chip micrograph occupying 1.9 mm2 of active area, and its key building blocks.

8.4.1 Benchtop Measurements

Fig. 8.11 shows the measured key operational waveforms, including VREC, PKD, and ZCD, and different operation modes of the chip with one beam. Similar to theoretical waveforms in Fig.8.4b,

81 the chip properly transitioned between VM and SECE by automatically detecting VREC peaks and zero-crossings.

Time Mode Function

t0-t1 VM Direct charging of CSTORE

t1-t2 Energy transfer from CP to LEXT

SECE Energy transfer from LEXT

charging CSTORE t2-t3

- CP charging

Fig. 8.11 Measured key operational waveforms showing the VM-SECE chip operation in different modes with one beam.

Fig. 8.12 shows the chip’s measured transient waveforms when the multi-beam inertial PEH was manually shaken gently. In Fig. 8.12a, during the cold start CCHIP was first charged via PC2 in Fig. 8.7 in a passive manner to ~1.5 V, at which the active voltage doubler started to operate and charged CCHIP efficiently to VCHIP = 3.2 V with VP,7 peak-peak voltages as large as 4 V. The DC shift of VP,7 in Fig. 8.12a is due to CP,7 and PC1 in Fig. 8.7. Similarly for VCHIP < 1.5 V, CSTORE was charged via six P1,n transistors (in VM path) in a passive manner to ~1.5 V, at which the VM active rectifiers started to operate, further charging CSTORE to ~1.9 V via VM for VCHIP < 3 V (chip not fully functional yet). As VCHIP > 3 V, the chip operated in reconfigurable VM-SECE mode (fully functional) to charge CSTORE more efficiently (faster rate in Fig. 8.12a) to VSTORE > 1.9 V at the presence of different voltages and frequency variations at each beam.

Fig. 8.12b and 8.12c shows the zoomed waveforms of Fig. 8.12a for VCHIP < 3 V (region A: chip not functional yet) and VCHIP > 3 V (region B: chip fully functional), respectively. It can be seen in Fig. 8.12b that whenever peak VREC surpasses VSTORE, CSTORE was charged via the VM operation. For most cycles, there was no energy harvesting. However, when the chip was fully functional in Fig. 8.12c, energy was extracted from each cycle via SECE-only or VM-SECE operations. Fig. 82

8.12d shows the rectified voltages across 4 different beams to highlight the voltage and frequency variabilities across multiple beams.

8.4.2 Measurements with a Commercial Single-Beam PEH

(a) (b)

(c) (d) Fig. 8.12 Benchtop measurement results of the chip with the multi-beam PEH. (a) Measured transient waveforms with cold start, (b) and (c) zoomed waveforms for VCHIP < 3 V and VCHIP > 3 V, respectively, and (d) VREC waveforms across 4 beams. To provide a fair comparison to prior arts, the performance of the VM-SECE chip was characterized with a commercial PEH with a single beam (PPA1011, MIDE Tech., Woburn, MA, USA). Fig. 8.13 shows the measurement setup for the PPA1011 with a 0.9-gram tip mass and resonance frequency (fP) of 56 Hz. A custom fabricated clamp fixed the beam on a smart shaker (K2007E01, The Modal Shop, Sharonville, OH, USA). An accelerometer (352A24, PCB Piezoelectronics, Depew, NY, USA) connected to a signal conditioner (480E09, PCB Piezoelectronics, Depew, NY, USA) was placed on the clamp to measure the shock-input accelerations. The beam outputs were connected to the chip. An RC box (RCS-502, IET Labs, Inc,

Roslyn Heights, NY, USA) provided different resistive loading (RL) across CSTORE. Shocks with different energies at the 1 Hz frequency with the optimal pulse width of 8.1 ms were applied to the smart shaker.

Fig. 8.13 Measurement setup for characterizing the chip’s performance with a commercial single-beam PEH (PPA1011) on a shaker.

Fig. 8.14 shows the VM-SECE chip improvements of PSTORE at different VSTORE for PPA1011 shock acceleration of 4.39 g at 1 Hz resulting in VP,OC = 2.82 V, compared to an on-chip FAR with measured 95.6% efficiency. At VSTORE = 1.8 V, the VM-SECE chip harvested 3.28x more power than the maximum power harvested by the FAR at VSTORE = 1 V (3.7 µW vs. 1.2 µW). This results in a shock figure of merit (FoM) of 328%, which is defined as the ratio of the maximum harvested energy by the chip to that of a full-wave rectifier [135]. Overall, the VM-SECE chip achieved much higher PSTORE for a wide range of VSTORE.

Fig. 8.15 shows the measured FoM and VP,OC vs. different shock accelerations of the PPA1011 PEH at 1 Hz. Under various shock accelerations, the VM-SECE chip maintained a high FoM of >

200% (> 2x improvement in PSTORE compared to the on-chip FAR). As the shock acceleration increased > 4.5 g, the FoM reduced because the VM-SECE chip supply was 3 V and leakage current increased for VSTORE > 3.8 V as discussed in Section III. As the shock acceleration increased from 2.85 g to 10.1 g, the VP,OC increased from 1.8 V to 7 V as shown in Fig. 8.15.

8.4.3 Measurements with the Custom Multi-Beam PEH

To verify the functionality of the proposed VM-SECE chip in a more realistic setup (similar to wearables), the chip was integrated with the mechanically plucked 5-beam PEH in Fig. 2b and tested on a robotic swing arm (one input of the chip was left open) [41]. Fig. 8.16 shows the

84 measurement setup with a motor-controlled swing arm replicating a pseudo-walking motion. The 50 cm long aluminum arm roughly mimics the human upper limb. The micro-stepping-enabled stepper motor was programmed to create varying motion profiles in a sinusoidal fashion with 25 degrees of rotation amplitude and 0.8-second period (fast pseudo walking) [41].

Fig. 8.15 Measured FoM and VP,OC of the VM-SECE chip under various shock accelerations. A minimum FoM of 200% was achieved.

Fig. 8.17 shows the measured transient waveforms of the rectified voltage of three beams (VREC1,2,3) and VSTORE. As the arm started swinging, CSTORE was charged to VSTORE ≈ 2 V with first VM and then reconfigurable VM-SECE operation within ~10 sec as shown in Fig. 8.17a. The zoomed waveforms in Fig. 8.17b show the variability of the voltage across different beams. Almost 4 s after the chip startup, the circuit operation (VSTORE) reached its steady state as shown in Fig. 8.17a.

Fig. 8.18a and 8.18b compares measured PSTORE at different VSTORE values of the VM-SECE chip and the on-chip FAR for operation with one and five beams on the robotic swing arm, respectively. The VM-SECE chip harvested 1.59x (FoM = 159%) and 2.38x (FoM = 238%) more power than the maximum energy harvested using the FAR for 1- and 5-beam operation, respectively. Utilizing

Fig. 8.16 Measurement setup with the integrated VM-SECE chip and 5-beam mechanically plucked PEH (Fig. 8.2b) mounted on a robotic swing arm.

5 beams further enhanced the FoM because the VM-SECE scheme extracted almost all energy from individual beams regardless of their voltage levels but 5 parallel FARs could not.

(a)

(b) Fig. 8.17 (a) Measured transient voltage waveforms of the VM-SECE chip interfacing the 5-beam mechanically plucked PEH in Fig. 8.2b on the robotic swing arm. (b) Zoomed waveforms showing voltage variability across beams.

Table 8.2 benchmarks the VM-SECE chip against state-of-the-art PEH chips. The VM-SECE chip offers the first integrated solution that can harvest from up to 6 beams simultaneously in a modular

86 fashion, with inputs as low as 35 mV, which makes it suitable for energy harvesting from multi- axial body motion. It can reconfigure itself between VM and SECE to improve the overall efficiency, extract maximum power, and protect itself from high voltages. With a commercial single-beam PEH, the VM-SECE chip achieved a high FoM of 328% at VP,OC = 2.8 V compared to an on-chip FAR with 95.6% efficiency. The slightly higher FoM of 330% in [135] with a SECE interface is mainly due to their FoM calculation with a passive rectifier, which is less efficient than a FAR. Also, a high-voltage (10 V) process is used in [135]. With the custom 5-beam mechanically plucked PEH, the VM-SECE chip achieved an FoM of 238% at VP,OC = 5 V.

(a)

(b) Fig. 8.18 Comparison of measured PSTORE vs. VSTORE between the VM-SECE chip and the on-chip FAR using the custom PEH in Fig. 8.2b on the robotic swing arm. (a) One-beam operation. (b) Five-beam operation. As listed in Table 8.2, the VM-SECE chip achieved highest end-to-end measured efficiencies of

95.6% and 84.6% at PSTORE of 8.5 µW and 10.2 µW for VM and VM-SECE operation, respectively. Conventional input power measurement using a current-sensing resistor was used to measure the efficiency in VM. The same method could not accurately be used for the VM-SECE, because the input current in VM and SECE were significantly different. Therefore, to measure the VM-SECE efficiency, the maximum extracted powers (PSTORE) for only a single beam (no chip) and the VM-

SECE chip both interfaced with optimal resistive loads were measured for the same input excitation. It should be noted that this leads to an optimistic value for the VM-SECE efficiency. Table 8.2 Benchmarking the Proposed 6-Beam Reconfigurable VM-SECE Chip Among the State-of-the-Art PEH Chips

Conclusions

The theory, implementation, and comprehensive measurement results of a fully autonomous multi- beam reconfigurable VM-SECE chip were shown with only one shared off-chip inductor for inertial energy harvesting, particularly from multi-axial body motion. Despite drastic variability of voltage across different beams as well as their frequency variations, the VM-SECE chip could harvest energy from up to 6 beams simultaneously in a modular fashion with improved efficiency and energy extraction, as well as inherent OVP. When interfaced with a commercial single-beam PEH, the chip could extract 3.28x more power compared to the best of an active rectifier with 95.6% efficiency. Experimental results with a custom-made mechanically plucked 5-beam inertial PEH mounted on a robotic swing arm, mimicking pseudo-walking, showed that the VM-SECE chip can operate properly in charging a storage capacitor and achieves a high FoM of 238%. The proposed multi-beam inertial PEH and VM-SECE chip hold the promise of integrated self- powered solutions for the next generation of wearables with vigilant operation capability.

Chapter 9 Conclusions and Future Work

This dissertation presents the research work on developing circuit- and system-level techniques for high-performance ultrasonic wireless power and data transmission to/from mm-sized implantable medical devices (IMDs), ultrasonically interrogated distributed system for gastric slow-waves (SWs) recording, and piezoelectric energy harvesting. For ultrasonic powering, the theory of ultrasonic wireless power transmission (WPT) to mm-sized implants has been described and an application-oriented design procedure has been proposed to optimize the ultrasonic powering link depending on the application needs; a hybrid inductive-ultrasonic powering link has been proposed to enable the powering through bone/air medium using ultrasound, which opens a variety of applications; a self-image-guided ultrasonic (SIG-US) WPT to mm-sized IMDs technique has also been proposed to adaptively update the location of implants in a distributed system with micro motions, which enables high-efficiency closed-loop powering in any distributed ultrasonically interrogated system. The ultrasonic interrogation in gastric recording system was also demonstrated. A chip with novel features of self-regulated power management and addressable communication has been fabricated and tested. A gastric recording system, consisting of the chip and commercial-off-the-shelf (COTS) components, has been tested both on benchtop and in vivo. For piezoelectric energy harvesting, a new power management scheme, reconfigurable voltage/synchronous electrical charge extraction-mode (VM-SECE) has been proposed. A multi- beam reconfigurable VM-SECE chip has been fabricated and tested to efficiently extract maximum power from a multi-beam wrist-worn PEH with only one off-chip inductor. The research work presented in this dissertation has resulted in several conference and journal publications [50]- [60]. This chapter summarizes the results and scientific contributions of this dissertation and the future works.

Summary of Results and Contributions

9.1.1 Ultrasonic Power Transmission to mm-Sized Implantable Devices

A design methodology was proposed to maximize the PTE of ultrasonic links for WPT to mm- sized biomedical implants. The proposed design procedure helps the designer identify the optimal geometries of Tx and Rx ultrasonic transducers such as diameter and thickness, as well as the optimal operation frequency. The optimal operation frequency of an ultrasonic WPT link with the 89

Rx size of 1 mm3 to power a load of 2.5 kΩ at the powering distance of 3 cm was found to be 1.8 MHz using acoustic matching layer on both Tx and Rx transducers. The optimal link achieved a PTE of 2.11%, while Rx transducer was on a silicon die with 0.3 mm thickness, mimicking the implant core circuitry. The design procedure and COMSOL models were also validated through measurements by optimizing and measuring the PTE of an ultrasonic WPT link that involved a 1 mm3 Rx transducer mounted on FR4 PCB. The simulated and measured PTEs matched well at 0.65% and 0.66% to power a load of 2.5 kΩ at the distance of 3 cm, respectively. The measured PTE reduced to 0.18% at 6 cm, which is still promising for WPT to deeply-implanted mm-sized devices for local operations such as recording and stimulation. This work presents the first reported ultrasonic WPT link, in which the geometries of both Tx and Rx ultrasonic transducers as well as operation frequency have been co-optimized, with realistic design settings for mm-sized biomedical implants.

The hybrid inductive-ultrasonic WPT can achieve high PTE for powering mm-sized biomedical implants in applications that involve multiple mediums with different acoustic impedances (air, bone, tissue), such as neural interfacing with freely-behaving rodents inside a cage. A co-designed hybrid link for an air-tissue medium was optimized and operated at 1.1 MHz to wirelessly power a 1 mm3 ultrasonic transducer, located 3 cm inside castor oil, with a considerable PTE of 0.16%, while the Tx coil was placed 6 cm away from the Rx transducer. An input power in Watt-range will be required in the hybrid link to power mm-sized implants with mW-range of power consumption.

The SIG-US beamforming can be used in any distributed, addressable PNS recording and stimulation. The location of the PNS nodes will be automatically updated in real time to ensure high delivered power regardless of the micro-motion of the implanted nodes. The concept of SIG- US beamforming was presented in an ultrasonically interrogated PNS recording platform along with the FEM simulation setup and results for powering a single implant with a linear transducer array. The simulation results show a 5.3 times improvement in received power under aligned condition comparing to the case without any beamforming technique and a 95.7 times improvement in normalized received power comparing to the conventional beamforming with a misalignment of 6 mm. For misalignment up to 6 mm, the SIG-US beamforming showed only 1.1

90 times decrease in normalized received power while the conventional beamforming reduced 108.7 times.

9.1.2 Ultrasonically Interrogated Distributed System for Large-Scale Gastric Slow-Wave Recording

Distributed millimeter-sized Gastric Seeds hold the promise of large-scale gastric SW recording with minimal damage. A proof-of-concept power management / data Tx chip for the Gastric Seeds was presented. The prototype chip includes a self-regulated power management that performs rectification, regulation and OVP in one step using only one off-chip capacitor as well as an addressable pulse-based data Tx with measured data rate of 75 kbps and energy consumption of 440 pJ/bit using a pair of 1 MHz ultrasonic transducers spaced by 3.75 cm in water.

A proof-of-concept ultrasonically powered system prototype implementing a power management/data Tx CMOS chip with COTS amplifiers, ADC, and RF transmission link was demonstrated for an in vivo experiment on two anesthetized rats. The CMOS chip includes self- regulated power management with only one off-chip capacitor as well as an addressable pulse- based data Tx with a measured data rate of 16 kbps using a pair of 1 MHz ultrasonic transducers spaced by 6 cm in water. SW signals were successfully recorded using ultrasonic power transfer through water to power the chip and amplifier. In the first in vivo experiment on a rat, the SW recorded by oscilloscope and RF link showed similar frequencies of 1.375 cpm and 1.5 cpm, respectively, and maximum amplitudes of 1.79 mV and 1.03 mV peak-to-peak, respectively, and minimum amplitudes of 0.46 mV and 0.28 mV peak-to-peak, respectively. In the second in vivo experiment on a rat, the SW recorded by oscilloscope and RF link showed faster but similar frequency of 3.75 cpm and 4 cpm, respectively, and maximum amplitudes of 2.41 mV and 2.17 mV peak-to-peak, respectively, and minimum amplitudes of 0.69 mV and 0.23 mV, respectively.

9.1.3 Piezoelectric Energy Harvesting

The theory, implementation, and comprehensive measurement results of a fully autonomous multi- beam reconfigurable VM-SECE chip were presented with only one shared off-chip inductor for inertial energy harvesting, particularly from multi-axial body motion. Despite drastic variability of voltage across different beams as well as their frequency variations, the VM-SECE chip could harvest energy from up to 6 beams simultaneously in a modular fashion with improved efficiency

91 and energy extraction, as well as inherent over-voltage-protection (OVP). When interfaced with a commercial single-beam PEH, the chip could extract 3.28x more power compared to the best of an active rectifier with 95.6% efficiency. Experimental results with a custom-made mechanically plucked 5-beam inertial PEH mounted on a robotic swing arm, mimicking pseudo-walking, showed that the VM-SECE chip can operate properly in charging a storage capacitor and achieves a high FoM of 238%. The proposed multi-beam inertial PEH and VM-SECE chip hold the promise of integrated self-powered solutions for the next generation of wearables with vigilant operation capability.

Future Work

Several groups have been working extensively on ultrasonic wireless power/communication systems. Although promising results have been shown in these works using an ultrasonic link, the location of the implant for the current works has to be known to achieve high PTEs due to the high sensitivity to misalignment of ultrasonic links, particularly when the transmitter transducer is highly focused. Thus, applications involving free-floating implants with existing micro-motions such as gastric wave recording can suffer from extremely low efficiencies since ultrasonic WPT links are susceptible to misalignment and orientation. The literature still lacks a platform to adaptively locate implants and maintain high PTEs regardless of changing in Rx locations and orientations. SIG-US beamforming provides a solution; however, this technique is still at its early stage. More work needs to be done to implement the SIG-US in a system, this includes: 1) validation of the concept through measurements, 2) optimization of ultrasonic links involving transmitter array and distributed implants, and 3) development of the implant chip and the interrogation system on a wearable unit for closed-loop powering.

Hybrid inductive-ultrasonic link provides a solution to use ultrasonic link for powering mm-sized IMDs deeply implanted in tissue involving air or bone medium. The ultrasonic Tx is directly driven by the inductive Rx which greatly simplifies the design and improves the efficiency. However, the bidirectional data communication link is not considered for the hybrid link, which is also crucial for such applications. It is unclear if the proposed hybrid link is capable of transmitting data in both directions. More research work can be done on developing the bidirectional hybrid data link.

A proof-of-concept chip with novel power management and addressable pulsed communication for Gastric Seed was demonstrated. In vivo experiments have been conducted with the chip and COTS components. Recording features need to be added to the chip to have a system-on-chip for gastric SW recording, which means a front-end and an analog-to-digital converter (ADC) should be added to the current chip. A 10-bit ADC has already been developed, however, more work needs to be done to develop the front-end amplifiers. Since the SW is at very low frequency, the front-end amplifier should be able to amplify low frequency signals (18 mHz – 100 mHz) with high gain. This is very challenging considering the size limitation of the implants. In addition, current work has been done based on a single pair of stationary ultrasonic transducers. Experiment with extended number of implants with changing location and orientation should be conducted. Adaptive closed-loop powering needs to be integrated into distributed Gastric Seed system considering the area of powering and motion of stomach. Moreover, miniaturized bio-compatible packaging needs to be investigated and implemented. More comprehensive in vivo animal experiments with freely behaving animals should be conducted for extended period of time with extended number of Gastric Seeds implanted.

A multi-beam reconfigurable VM-SECE chip interfacing a wrist-worn piezoelectric energy harvester (PEH) with custom fabricated thin-film beams has been demonstrated. However, current VM-SECE harvesting is load dependent (Fig. 8.18). Loading condition often changes during practical applications. Therefore, more work can be done to further improve the harvested power, such as maximum-power-point-tracking (MPPT), or to modify the current structure to optimize VM and SECE reconfiguration timing to compensate for load-dependent harvesting.

References

[1] K. Famm, B. Litt, K. Tracey, E. Boyden, and M. Slaoui, “Drug discovery: A jump - start for electroceuticals,” Nature, vol. 496, pp. 159-161, Apr. 2013.

[2] K. Birmingham, V. Gradinaru, P. Anikeeva, W. Grill, V. Pikov, B. McLaughlin, P. Pasricha, D. Weber, K. Ludwig, and K. Famm, “Bioelectronic medicines: a research roadmap,” Nature Rev., vol. 13, pp. 399 - 400, Jun. 2014.

[3] B. Bonaz, C. Picq, V. Sinniger, J. Mayol, and D. Clarencon, “Vagus nerve stimulation: from epilepsy to the cholinergic anti - inflammatory pathway,” Neurogast. Mot., vol. 25, pp. 208 - 221, Mar. 2013.

[4] J. Lee, D. Kim, S. Yoo, H. Lee, G. Lee, and Y. Nam, “Emerging neural stimulation technologies for bladder dysfunctions,” Int. Neurol. J., vol. 19, pp. 3 - 11, Mar. 2015.

[5] G. O’Grady, J. Egbuji, P. Du, L. Cheng, A. Pullan, and J. Windsor, “High - frequency gastric electrical stimulation for the treatment of gastroparesis: a meta - analysis,” World Journal of surgery, vol. 33, no. 8, pp .1693 – 1701, Aug. 2009.

[6] H. Zhu, H. Sallam, and J. Z. Chen, “Synchronized gastric electrical stimulation enhances gastric motility in dogs,” Neurogastroenterol. Motil., vol. 293, no. 5, pp. 1875 – 1881, Nov. 2007.

[7] G. McConnell, H. Rees, A. Levey, C. Gutekunst, R. Gross, and R. Bellamkonda, “Implanted neural electrodes cause chronic, local inflammation that is correlated with local neurodegeneration,” J. Neural Eng., vol. 6, p. 056003, Oct. 2009.

[8] H. Lee, R. Bellamkonda, W. Sun, and M. Levenston, “Biomechanical analysis of silicon microelectrode - induced strain in the brain,” J. Neural Eng., vol. 2, no. 4, pp. 81 – 89, Dec. 2005.

[9] L. Karumbaiah, T. Saxena, D. Carlson, K. Patil, R. Patkar, E. Gaupp, M. Betancur, G. Stanley, L. Carin, and R. Bellamkonda, “Relationship between intracortical electrode design and chronic recording function,” Biomate r., vol. 34, pp. 8061 – 8074, Jul. 2013.

[10] J. Song, D. Zhong, W. Qian, X. Hou, and J. Chen, “Short pulse gastric electrical stimulation for cisplatin - induced emesis in dogs,” Neurogastroenterol. Motil., vol. 23, no. 5, pp. 468 – 474, May 2011.

[11] M. Yin, D. Borton, J. Aceros, W. Patterson, and A. Nurmikko, “A100-channel hermetically sealed implantable device for chronic wireless neurosensing applications,” IEEE Trans. Biomed. Cir. Syst., vol.7, pp. 115-128, Apr. 2013.

[12] D. Zhou and E. Greenbaum, Implantable neural prostheses 1, New York, NY: Springer, 2009.

[13] K. Chen, Z. Yang, L. Hoang, J. Weiland, M. Humayun, and W. Liu, “An integrated 256- channel epiretinal prosthesis,” IEEE J. Solid State Cir., vol. 45, pp. 1946–1956, Sep. 2010.

[14] S. Lee, H. Lee, M. Kiani, U. Jow, and M. Ghovanloo, “An inductively-powered scalable 32- channel wireless neural recording system-on-a-chip for neuroscience applications,” IEEE Trans. Biomed. Cir. Syst., vol. 4, pp. 360-371, Dec. 2010.

[15] U. M. Jow and M. Ghovanloo, “Design and optimization of printed spiral coils for efficient transcutaneous inductive power transmission,” IEEE Trans. Biomed. Cir. Syst., vol. 1, pp. 193- 202, Sept. 2007.

[16] M. Kiani, U. Jow, and M. Ghovanloo, “Design and optimization of a 3-coil inductive link for efficient wireless power transmission,” IEEE Trans. Biomed. Cir. Syst., vol. 5, pp. 579-591, Dec. 2011.

[17] R. Muller, H. Le, W. Li, P. Ledochowitsch, S. Gambini, T. Bjorninen, A. Koralek, J. Carmena, M. Maharbiz, E. Alon, and J. Rabaey, “A minimally invasive 64-channel wireless µECoG implant”, IEEE J. Solid State Cir., vol. 50, pp. 344–359, Jan. 2015.

[18] K. Montgomery, A. Yeh, J. Ho, V. Tsao, S. Iyer, L. Grosenick, E. Ferenczi, Y. Tanabe, K. Deisseroth, S. Delp, and A. Poon, “Wirelessly powered, fully internal optogenetics for brain, spinal and peripheral circuits in mice,” Nature Methods, vol. 12, pp. 960-974, Oct. 2015.

[19] J. Ho, A. Yeh, E. Neofytou, S. Kim, Y. Tanabe, B. Patlolla, R. Beygui, and A. Poon, “Wireless power transfer to deep-tissue microimplants,” Proc. Natl. Acad. Sci., vol. 111, pp. 7974-7979, Jun. 2014.

[20] M. Mark, T. Bjorninen, L. Ukkonen, L. Sydanheimo, and J. Rabaey, “SAR reduction and link optimization for mm-size remotely powered wireless implants using segmented loop antennas,” Biomed. Wireless Techn. Networks Sensing Syst. (BioWireleSS), pp. 7-10, Jan. 2011.

[21] D. Ahn and M. Ghovanloo, “Optimal design of wireless power transmission links for millimeter-sized biomedical implants”, IEEE Trans. Biomed. Cir. Syst., vol. 10, no. 1, pp. 125- 137, Feb. 2016.

[22] A. Ibrahim and M. Kiani, “A figure-of-merit for design and optimization of inductive power transmission links for millimeter-sized biomedical implants,” IEEE Trans. Biomed. Cir. Syst., vol. 10, no. 6, pp. 1100-1111, Dec. 2016.

[23] “IEEE standard for the safety levels with respect to human exposure to radiofrequency electromagnetic fields, 3 KHz to 300 GHz,” IEEE Standard C95.1, 2006.

[24] D. Seo, J. Carmena, J. Rabaey, E. Alon, and M. Maharbiz, “Neural dust: an ultrasonic, low power solution for chronic brain - machine interfaces,” arXiv: 1307.2196, 2013.

[25] A. Leahy, K. Besherdas, C. Clayman, I. Mason, and O. Epstein, “Abnormalities of the electrogastrogram in functional gastrointestinal disorders,” Am. J. Gastroenterol., vol. 94, no.4, pp. 1023 – 1028, Apr. 1999.

[26] Z. Lin, I. Sarosiek, J. Forster, I. Damjanov, Q. Hou, and R. McCallum, “Association of the status of interstitial cells of Cajal and electrogastrogram parameters, gastric emptying and symptoms in patients with gastroparesis,” Neurogastroenterol. Motil. , vol. 22, no.1, pp. 56 – 61, Jan. 2010.

[27] G. O’Grady, T. Wang, P. Du, T. Angeli, W. Lammers, and L. Cheng, “Recent progress in gastric arrhythmia: Pathophysiology, clinical significance and future horizons,” Clin Exp. Pharmacol. Physiol., vol. 41, no. 10, pp. 854 – 862, Oct. 2014.

[28] G. O’Grady, T. Angeli, P. Du, C. Lahr, W. Lammers, J. Windsor, T. Abell, G. Farrugia, A. Pullan, and L. Cheng, “Abnormal initiation and conduction of slow wave activity in gastroparesis, defined by high resolution electrical mapping,” Gastroenterology, vol. 143, no.2, pp. 58 9 – 598, Sep. 2012.

[29] J. Huizinga and W. Lammers, “Gut peristalsis is coordinated by a multitude of cooperating mechanisms,” Am. J. Physiol. Gastrointest. Liver Physiol., vol. 296, no. 1, pp. G1-8, Jan. 2009.

[30] A. Farajidavar, P. McCorkle, T. Wiggins, S. Rao, C. Hagains, Y. Peng, J. Seifert, M. Romero, G. O'Grady, L. Cheng, S. Sparagana, M. Delgado, S. Tang, T. Abell, and J. - C. Chiao, “A Miniature Power - Efficient Bidirectional Telemetric Plat - form for in vivo Acquisition of Electrophysiological signals”, Conf. Proc. IEEE International Microwave Symposium (IMS), vol. 1, pp. 1 – 4, 2011.

[31] A. Farajidavar, G. O'Grady, S. Rao, L. Cheng, T. Abell, and J. C. Chiao, “A miniature bidirectional telemetry system for in vivo gastric slow wave recordings,” Physiol Meas., vol. 33, no. 6, pp. 29 – 37, Jun. 2012.

[32] N. Paskaranandavadivel, R. Wang, S. Sathar, G. O’Grady, L. Cheng, and A. Farajidavar, “A wireless multichannel system for in - vivo monitoring of gastric activity propagation,” Neurogastroenterol Motil., vol. 27, no. 4, pp. 580 - 585, Apr. 2015.

[33] A. Ibrahim, A. Farajidavar, and M. Kiani, “A 64 - channel wireless implantable system - on - chip for gastric electrical - wave recording,” IEEE Sensors Conf., pp. 1-3, Oct. 2016.

[34] R. J. M. Vullers, R. V. Schaijk, H. J. Visser, J. Penders, and C. V. Hoof, “Energy harvesting for autonomous wireless sensor networks,” IEEE Solid-State Circuit Magazine, vol. 2, no. 2, pp. 29-38, June 2010.

[35] E. Ibarra, A. Antonopoulos, E. Kartsakli, J. Rodrigues, and C. Verikoukis, “QoS-Aware energy management in body sensor nodes powered by human energy harvesting,” IEEE Sensors Journal, vol. 16, no. 2, pp. 542-549, Jan. 2016.

[36] J. Dieffenderfer, H. Goodell, S. Mills, M. McKinight, S. Yao, F. Lin, E. Beppler, B. Bent, B. Lee, V. Misra, Y. Zhu, O. Oralkan, J. Strohmaier, J. Muth, D. Peden, and A. Bozkurt, “Low- power wearable systems for continuous monitoring of environment and health for chronic respiratory disease,” IEEE J. Biomed. Health Inform., vol. 20, no. 5, pp. 1251-1264, Sept. 2016.

[37] S. Yao and Y. Zhu, “Wearable multifunctional sensors using printed stretchable conductors made of silver nanowires,” Nanoscale, vol. 6, no. 4, pp. 2345-2352, Jan. 2014.

[38] M. Renaud, P. Fiorini, R. V. Schaijk, and C. V. Hoof, “Harvesting energy from the motion of human limbs: the design and analysis of an impact-based piezoelectric generator,” Smart Mat. and Struc., vol. 21, no. 4, p. 049501, Mar. 2012.

[39] V. Leonov, “Thermoelectric energy harvesting of human body heat for wearable sensors,” IEEE Sens. J., vol. 13, pp. 2284-2291, June 2013.

[40] B. J. Hansen, Y. Liu, R. Yang, and Z. L. Wang, “Hybrid nanogenerator for concurrently harvesting biomechanical and biochemical energy,” ACS Nano, vol. 4, no. 7, pp. 3647-3652, May 2010.

[41] T. Xue, H. G. Yeo, S. Trolier-Mckinstry, and S. Roundy, “Wearable inertial energy harvester with sputtered bimorph lead zirconate titanate (PZT) thin-film beams,” Smart Mat. Struc. vol. 27, no. 8, p. 085026, July 2018.

[42] Y. K. Ramadass and A. P. Chandrakasan, “An efficient piezoelectric energy harvesting interface circuit using a bias-flip rectifier and shared inductor,” IEEE J. Solid-State Cir., vol. 45, no. 1, pp. 189-204, Jan. 2010.

[43] H. Song, P. Kumar, D. Maurya, M. Kang, W. T. Reynolds, D. Jeong, C. Kang, and S. Priya, “Ultra-low resonant piezoelectric MEMS energy harvester with high power density,” J. of Microelectromechanical Syst., vol. 26, no. 6, pp. 1226-1234, Dec. 2017.

[44] H. G. Yeo, X. Ma, C. Rahn, and S. Trolier-Mckinstry, “Efficient piezoelectric energy harvester utilizing (001) textured bimorph PZT films on flexible metal foils,” Advanced Func. Mat., vol. 26, pp. 5940-5946, Aug. 2016.

[45] R. Sriramdas and R. Pratap, “Scaling and performance analysis of MEMS piezoelectric energy harvesters,” J. of Microelectromechanical Syst., vol. 26, no. 3, pp. 679-690, June 2017.

[46] R. Sriramdas, S. Chiplunkar, R. Cuduvally, and R. Pratap, “Performance enhancement of piezoelectric energy harvesters using multilayer and multistep beam configurations,” IEEE Sens. J., vol. 15, no. 6, pp. 3338-3348, June 2015.

[47] H. Ghoddus and Z. Kordrostami, “Harvesting the ultimate electrical power from MEMS piezoelectric vibration energy harvesters: an optimization approach,” IEEE Sens. J., vol. 18, no. 21, pp. 8667-8675, Nov. 2018.

[48] W. G. Li, and S. Yu, “Improving power density of a cantilever piezoelectric power harvester through a curved L-shape proof mass,” IEEE Trans. on Ind. Elect., vol. 57, no. 3, pp. 868-876, Mar. 2010.

[49] J. C. Park, D. H. Lee, J. Y. Park, Y. S. Chang, and Y. P. Lee, “High performance piezoelectric

MEMS energy harvester based on D33 mode of PZT thin film on buffer-layer with PbTiO3 inter-layer,” Inter. Solid-State Sens. Actua. Microsyst. Conf., pp. 517-520, June 2009.

[50] M. Meng and M. Kiani, “Design and optimization of ultrasonic wireless power transmission links for millimeter-sized biomedical implants,” IEEE Trans. Biomed. Cir. Syst., vol. 11, pp. 98-107, Feb. 2017

[51] M. Meng and M. Kiani, “A hybrid inductive-ultrasonic link for wireless power transmission to millimeter-sized biomedical implants,” IEEE Trans. Cir. Syst. II, vol. 64, no. 10, pp. 1137- 1141, Oct. 2016.

[52] A. Ibrahim, M. Meng and M. Kiani, “A comprehensive comparative study on inductive and ultrasonic wireless power transmission to biomedical implants,” IEEE Sensors Jour. vol. 18, no. 9, pp. 3813-3826, May 2018.

[53] M. Meng, A. Ibrahim, and M. Kiani, “Design considerations for ultrasonic power transmission to millimeter-sized implantable microelectronics devices,” IEEE Biomed. Cir. Syst. Conf., pp. 1-4, Oct. 2015.

[54] M. Meng and M. Kiani, "Optimal resonance configuration for ultrasonic wireless power transmission to millimeter-sized biomedical implants," 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 1934-1937, 2016.

[55] A. Ibrahim, M. Meng and M. Kiani, "Inductive and ultrasonic wireless power transmission to biomedical implants," IEEE International Symposium on Circuits and Systems (ISCAS), pp. 1- 4, 2017.

[56] M. Meng and M. Kiani, “Self-image-guided ultrasonic wireless power transmission to millimeter-sized biomedical implants” 41st Annual International Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC), July 2019.

[57] M. Meng and M. Kiani, “Gastric seed: toward distributed ultrasonically interrogated millimeter-sized implants for large-scale gastric electrical-wave recording,” IEEE Trans. on Cir. and Syst. II: Expr. Brief. vol. 66, no. 5, pp. 783-787, May 2019.

[58] M. Meng, Philip Graybill, Raddy L. Ramos, A. J. Khoshkholgh, A. Farajidavar and M. Kiani, “An ultrasonically powered wireless system for in vivo gastric slow wave recording” 41st Annual International Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC), July 2019.

[59] M. Meng, D. Wang, B. D. Truong, S. Trolier-Mckinstry, S. Roundy, and M. Kiani, “A multi- beam shared-inductor reconfigurable voltage/SECE-mode piezoelectric energy harvesting interface circuit,” IEEE Trans. on Biomed. Cir. and Syst. 2019.

[60] M. Meng, A. Ibrahim, T. Xue, H. G. Yeo, D. Wang, S. Roundy, S. Trolier-McKinstry, and M. Kiani, “Multi-beam shared-inductor reconfigurable voltage/SECE-mode piezoelectric energy harvesting of multi-axial human motion,” IEEE Int. Solid-State Cir. Conf. (ISSCC), pp. 426-427, Feb. 2019.

[61] M. Camilleri, "Clinical practice. Diabetic gastroparesis," N eur. Eng. J . Med., vol. 356, no. 8, pp. 820 – 829, Feb 2007.

[62] Y. Wang, R. Fisher, and H. Parkman, “Gastroparesis - related hospitalizations in the United States: trends, characteristics and outcomes 1995 - 2004,” Am. J. Gastroenterol., vol. 103, no.2, pp. 313 – 322, Feb.2008.

[63] Y. Shai and H. El - Serag, “The prevalence and risk factors of functional dyspepsia in a multiethnic population in the United States,” Am. J. Gastroenterol, vol. 99, no. 11, pp. 2210 – 2216, Nov. 2004.

[64] M. Wallander, S. Johansson, A. Ruigomez, L. Rodriguez, and R. Jones, “Dyspepsia in general practice: incidence, risk factors, comorbidity and mortality,” Fam. Pract., vol. 24, no. 5, pp. 403 – 411, Oct. 2007.

[65] Z. Lin, E. Eaker, I. Sarosiek, and R. McCallum, “Gastric myoelectrical activity and gastric emptying in patients with functional dyspepsia,” Am. J. Gastroenterol., vol. 94, no. 9, pp. 2384 – 2389, Sep. 1999.

100

[66] H. Parkman, W. Hasler, and R. Fisher, “American gastroenterological association technical review on the diagnosis and treatment of gastroparesis,” Gastroenterology, vol. 127, no. 5, pp. 1592 – 1622, Nov. 2004.

[67] S. Waseem, B. Moshiree, and P. Draganov, “Gastroparesis: current diagnostic challenges and management consideration s,” World J. Gastroenterol., vol. 15, no. 1, pp. 25 – 37, Jan. 2009.

[68] U. Barone and R. Merletti, “Design of a portable, intrinsically safe multichannel acquisition system for high - resolution real - time processing HD - sEMG,” IEEE Trans. Biomed. Eng., vol. 60, Au g. 2013.

[69] S. Narayanaswamy, “High resolution electrocardiography,” Pacing Electr. J., vol. 2, pp. 50 - 56, 2002.

[70] P. Du, G. O'Grady, J. U. Egbuji, W. J. Lammers, D. Budgett, P. Nielsen, J. A. Windsor, A. J. Pullan, L. K. Cheng. “High - resolution mapping of in v ivo gastrointestinal slow wave activity using flexible printed circuit board electrodes: methodology and validation,” Ann. Biomed. Eng., vol. 37, no. 4, pp. 839 – 846, Apr. 2009.

[71] N. Paskaranandavadivel, G. O’Grady, and L. Cheng, “Time - delay mapping of high - resolution gastric slow - wave activity,” IEEE Trans. Biomed. Eng., vol. 64, pp. 166 - 172, Jan. 2017.

[72] W. Lammers, L. Ver Donck, B. Stephen, D. Smets, and J. Schuurkes, “Focal activities and re - entrant propagations as mechanisms of gastric tachyarrhythmias,” Gastroenterology, vol. 135, no. 5, pp. 1601 – 1611, Nov. 2008.

[73] G. O’Grady, J. Egbuji, P. Du, W. Lammers, L. Cheng, J. Windsor, and A. Pullan, “High - resolution spatial analysis of slow wave initiation and conduction in porcine gastric dysrhythmia,” Neurogastro enterol. Motil., vol. 23, no. 9, pp. 345 – 355, Sep. 2011.

[74] J. Chen, Z. Lin, J. Pan, and R. McCallum, “Abnormal gastric myoelectrical activity and delayed gastric emptying in patients with symptoms suggestive of gastroparesis,” Dig. Dis. Sci., vol. 41, no. 8, pp. 1538 – 1545, Aug. 1996.

101

[75] H. Vittal, G. Farrugia, G. Gomez, and P. Pasricha, “Mechanisms of disease: the pathological bas is of gastroparesis — a review of experimental and clinical studies,” Nat. Clin. Pract. Gastroenterol. Hepatol., vol. 4, no. 6, pp. 336 – 346, Jun. 2007.

[76] X. Lin and J. Chen, “Abnormal gastric slow waves in patients with functional dyspepsia assessed by multichannel electrogastrography,” Am. J. Physiol. Gastrointest. Liver Physiol., vol. 280, no. 8, pp. G1370 – 1375, Jun. 2001.

[77] T. Abell, R. Be rnstein, T. Cutts, G. Farrugia, J. Forster, W. Hasler, R. McCallum, K. Olden, H. Parkman, C. Parrish, P. Pasricha, C. Prather, E. Soffer, R. Twillman, and A. Vinik, “Treatment of gastroparesis: a multidisciplinary clinical review,” Neurogastroenterol. Moti l., vol. 18, no. 4, pp. 263 – 283, Apr. 2006.

[78] G. O’Grady, P. Du, W. Lammers, J. Egbuji, P. Mithraratne, J. Chen, L. Cheng, J. Windsor, and A. Pullan, “High-resolution entrainment mapping of gastric pacing: a new analytical tool,” Am. J. Physiol. Gastrointest. Liver Physiol., vol. 298, no. 2, pp.314 – 321, Feb. 2010.

[79] L. Ver Donck, W. Lammers, B. Moreaux, D. Smets, J. Voeten, J. Vekemans, J. Schuurkes, and B. Coulie, “Mapping slow waves and spikes in chronically instru mented conscious dogs: implantation techniques and recordings,” Med. Biol. Eng., vol. 44, no. 3, pp. 170 – 178, Mar. 2006.

[80] M. Bortolotti, “Electrogastrography: a seductive promise, only partially kept,” Am. J. Gastroenterol., vol. 93, no.10, pp. 1791 – 1794, O ct. 1998.

[81] M. Verhagen, L. Van Schelven, M. Samsom, and A. Smout, “Pitfalls in the analysis of electrogastrographic recordings,” Gastroenterology, vol. 177, no. 2, pp. 453 – 460, Aug. 1999.

[82] A. Gharibans, S. Kim, D. Kunkel, and T. Coleman, “High - resolution electrogastrogram: a novel, noninvasive method for determining gastric slow - wave direction and speed,” IEEE Trans. Biomed. Eng., vol. 64, pp. 807 - 815, Apr. 2017.

[83] Z. Lin, I. Sarosiek, J. Forster, R. A. Ross, J. Chen, and R. McCallum, “Two - channel gastric pacin g in patients with diabetic gastroparesis,” Neurogastroenterol. Motil., vol. 23, no. 10, pp.912 – e396. Oct. 2011.

102

[84] R. Coleski and W. L. Hasler, “Coupling and propagation of normal and dysrhythmic gastric slow waves during acute hyperglycaemia in healthy huma ns,” Neurogastroenterol. Motil., vol. 21, no. 5, pp. 492 – 499, e1 – 2, May 2009.

[85] K. Sanders, S. Ward, and G. Hennig, “Problems with extracellular recording of electrical activity in gastrointestinal muscle,” Nature Reviews Gastroenterology Hepatology, vol. 13, pp. 731 - 741, Oct. 16.

[86] R. Wang, Z. Abukhalaf, A. Javan - khoshkholgh, A. Stocker, T. Abell, and A. Farajidavar, “A novel system and methodology for continuous ambulatory monitoring of gastric slow waves”, Gastroenterology, vol. 152, no. 5, Supplement 1, Page S516, 2017.

[87] H. Fang, K. Yu, C. Gloschat, Z. Yang, E. Song, C. Chiang, J. Zhao, S. Won, S. Xu, M. Trumpis, Y. Zhong, S. Han, Y. Xue, D. Xu, S. Choi, G. Cauwenberghs, M. Kay, Y. Huang, J. Viventi, I. Efimov, and J. Rogers, “Capacitively coupled arrays of multiplexed flexible silicon transistors for long - term cardiac electrophysiology,” Nature Biomed. Eng., vol. 1, pp. 1 - 11, Mar. 2017.

[88] I. Beech and J. Sunner, “Biocorrosion: towards understanding interactions between biofilms and metals,” Curr. Opin. Biotechnol., vol. 15, pp. 181 - 186, Jun. 2004.

[89] S. Ha, A. Akinin, J. Park, C. Kim, H. Wang, C. Maier, P. Mercier, and G. Cauwenberghs, “Silicon - integrated high - density electrocortical interfaces,” Proc. IEEE, vol. 105, no. 1, pp. 11-33, Jan. 2017.

[90] P. Larson and B. Towe, “Miniature ultrasonically powered wireless nerve cuff stimulator,” IEEE Conf. Neural Eng., pp. 265-268, May 2011.

[91] D. Seo, J. Carmena, J. Rabaey, M. Maharbiz, and E. Alon, “Model validation of untethered, ultrasonic neural dust motes for cortical recording,” J. Neuroscience Methods, vol. 244, pp. 114 - 122, Apr. 2015.

[92] D. Seo, R. Neely, K. Shen, U. Singhal, E. Alon, J. Rabaey, J. Carmena, and M. Maharbiz, “Wireless recording in the peripheral nervous system with ultrasonic neural dust,” Neuron, vol. 91, pp. 529 - 539, Aug. 2016.

103

[93] J. Charthad, M. Weber, T. Chang, and A. Arbabian, “A mm - size implantable medical device (IMD) with ultrasonic power transfer and a hybrid bi - directional data link,” IEEE J. Solid - State Cir., vol. 50. pp. 1 - 13, Aug. 2015.

[94] T. Cheng, M. Wang, J. Charthad, M. Weber, and A. Arbabian, “ A 30.5mm 3 fully packaged implantable device with duplex ultrasonic data and power links achieving 95kb/s with < 10 − 4 BER at 8.5cm depth,” IEEE Int. Solid - State Conf., pp. 460-461, Feb. 2017.

[95] A. Denisov and E. Yeatman, “Ultrasonic vs. inductive power delivery for miniature biomedical implants,” International Conf. Body Sensor Networks, pp. 84-89, 2010.

[96] T. Chou, R. Subramanian, J. Park, and P. Mercier, “A miniaturized ultrasonic power delivery system,” IEEE Biomed. Cir. Syst. Conf., pp. 440-443, Oct. 2014.

[97] S. Song, A. Kim, and B. Ziaie, “Omni-directional ultrasonic powering for millimeter-scale implantable devices,” IEEE Trans. Biomed. Eng., vol. 62, no. 11, pp. 2717-2723, Nov. 2015.

[98] D. Seo, J. Carmena, J. Rabaey, M. Maharbiz, and E. Alon, “Model validation of untethered, ultrasonic neural dust motes for cortical recording” J. Neuroscience Methods, vol. 244, pp. 114-122, Apr. 2015.

[99] F. Mazzilli, M. Peisino, R. Mitouassiwou, B. Cotte, P. Thoppay, C. Lafon, P. Favre, E. Meurville, and C. Dehollain, “In-vitro platform to study ultrasound as source for wireless energy transfer and communication for implanted medical devices,” IEEE Eng. Med. Biol. Soc. Conf., pp. 3751-3754, Sept. 2010.

[100] G. Santagati, T. Melodia, L. Galluccio, and S. Palazzo, “Medium access control and rate adaption for ultrasonic intrabody sensor networks,” IEEE Trans. Net., vol. 23, pp. 1121-1134, Aug. 2015.

[101] G. Santagati and T. Melodia, “Experimental evaluation of impulsive ultrasonic intra-body communications for implantable biomedical devices,” IEEE Trans. Mobile Computing, vol. 16, no. 2, pp. 367-380, Feb. 2017.

[102] Olympus ultrasonic transducers technical notes [online]. Available: https://www.olympus- ims.com/data/File/panametrics/UT-technotes.en.pdf.

104

[103] S. K. Edelman, “Understanding ultrasound physics,” E.S.P Ultrasound, 4 edition, July, 2012.

[104] S. Ozeri and D. Shmilovitz, “Ultrasonic transcutaneous energy transfer for powering implanted devices” Ultrasonics, vol. 50, pp. 556-556, May 2010.

[105] S. Bhuyan, K. Sivanand, S. K. Panda, R. Kumar, and J. Hu, “Resonance-based wireless energizing of piezoelectric components”, IEEE Mag. Lett., vol. 2, 2011.

[106] H. Kunkel, S. Locke, and B. Pikeroen, “Finite-element analysis of vibrational modes in piezoelectric ceramic disks,” IEEE Trans. Ferroelectrics Freq. Control, vol. 37, pp. 316-328, July 1990.

[107] R. Krimholtz, D. A. Leedom, and G. L. Matthaei, “New equivalent circuit for elementary piezoelectric transducers,” Electron Letters, vol. 6, no. 13 pp. 398-399, June 1970.

[108] M. J. Zipparo, K. K. Shung, and T. S. Shrout, “Piezoceramics for high frequency (20-100 MHz) single element imaging transducers,” IEEE Trans. Ultrasonics, Ferroelectrics and Frequency Control, vol. 44, no. 5, pp. 1038-1048, Sept. 1997.

[109] Q. Zhang, P. A. Lewin, P. E. Bloomfield, “PVDF transducers- a performance comparison of single-layer and multilayer structures,” IEEE Trans. Ultrasonics, Ferroelectrics and Frequency Control, vol. 44, no. 5, pp. 1148-1156, Sept. 1997.

[110] C. Alexander and M. Sadiku, Fundamental of Electric Circuits, 2nd edition New York: McGraw-Hill, 2004.

[111] A. Daniell, “A text book of the principles of physics,” The Macmillan Company, pp.412- 476, 1984.

[112] H. S. Gougheri, A. Dangi, S. Kothapalli and M. Kiani, “A comprehensive study of ultrasonic transducer characteristics in microscopic ultrasound neuromodulation,” IEEE Trans. Biomed. Cir. and Syst., June 2019.

[113] G. Kossof, “The effects of backing and matching on the performance of piezoelectric ceramic transducers,” IEEE Trans Sonics Ultrasonics, vol. 13, pp. 20-30, Mar. 1966.

105

[114] J. Souquet, P. Defranould, and J. Desbois, “Design of low-loss wide-band ultrasonic transducers for noninvasive medical application,” IEEE Trans. Sonics Ultrasonics, vol. 26, pp. 75-80, Mar. 1979.

[115] T. Inoue, M. Ohta, and S. Takahashi, “Design of ultrasonic transducers with multiple acoustic matching layers for medical application,” IEEE Trans. Ultrason. Ferroelectr. Freq., vol. 34, pp. 8-16, Jan. 1987.

[116] D. Callens, C. Bruneel, J. Assaad, “Matching ultrasonic transducer using two matching layers where one of them is glue,” NDT&E International, vol. 37, pp. 591-596, Dec. 2004.

[117] F. Fyke, J. Greenleaf, and S. Johnson, “Continuous wave measurements of acoustic attenuation in an oil/polymer mixture,” Ultrasound Med. Biol., vol. 5, pp. 87-90. 1978.

[118] D. Pozar, Microwave Engineering. Hoboken, NJ: Wiley, 2011.

[119] M. Kiani and M. Ghovanloo, “An RFID-based closed loop wireless power transmission system for biomedical applications,” IEEE Trans. Cir. Syst.-II, vol. 57, pp. 260-264, Apr. 2010.

[120] D. Shmilovitz, S. Ozeri, C. Wang, and B. Spivak, “Noninvasive control of the power transferred to an implanted device by an ultrasonic transcutaneous energy transfer link,” IEEE Trans. Biomed. Eng., vol. 61, pp. 995-1004, Apr. 2014.

[121] H. Vihvelin, J. Leadbetter, M. Bance, J. Brown, and R. Adamson, “Compensating for tissue changes in an ultrasonic power link for implanted medical devices,” IEEE Tran. Biomed. Cir. Syst., vol. 10, pp. 404-411, Apr. 2016.

[122] K. Bocan and E. Sejdic, “Adaptive transcutaneous power transfer to implantable devices: a state of the art review,” Sensors, vol. 16, no. 3, p. E393, Mar. 2016.

[123] X. Dong, T. Yuan, M. Hu, H. Shekhani, Y. Maida, T. Tou, and K. Uchino, “Driving frequency optimization of a piezoelectric transducer and the power supply development,” Rev. Sci. Instrum. vol. 87, 105003, Oct. 2016.

[124] B. Timko, T. Cohen-Karni, G. Yu, Q. Qing, B. Tian, and C. Lieber, “Electrical recording from hearts with flexible nanowire device arrays,” Nano letters, vol. 9, pp. 914-918, Jan. 2009.

106

[125] S. Baltsavias, W. V. Treuren, M. J. Weber, J. Charthad, S. Baker, J. L. Sonnenburg, A. Arbabian, “In vivo wireless sensors for gut microbiome redox monitoring,” arXiv: 1902.07386.

[126] B. Wysocka, Z. Kassam, G. Lockwood, J. Brierley, L. A. Dawson, C. A. Buckley, D. Jaffray, B. Cummings, J. Kim, R. Wong, and J. Ringash, “Interfraction and respiratory organ motion during conformal radiotherapy in gastric cancer,” International J. Radiation Oncology*Biology*Physicas, vol. 77, no. 1, pp. 53-59, May 2010.

[127] Y. Jia, S. A. Mirbozorgi, B. Lee, W. Khan, F. Madi, O. T. Inan, A. Weber, W. Li and M. Ghovanloo, “A mm-sized free-floating wireless powered implantable optical stimulation device,” IEEE Trans. on Biomed. Cir. and Syst. vol. 13, no. 4, pp. 608-618, Aug. 2019.

[128] Sanni, A. Vilches, and C. Toumazou, “Inductive and ultrasonic multi-tier interface for low- power, deeply implantable medical devices,” IEEE Trans. Biomed. Cir. Syst., vol. 6, no. 4, pp. 297-308, Aug. 2012.

[129] M. Kiani and M. Ghovanloo, “A figure-of-merit for designing high performance inductive power transmission links,” IEEE Trans. Indus. Elect., vol. 60, pp. 5292-5305, Nov. 2013.

[130] M. Weber, Y. Yoshihara, A. Sawaby, J. Charthad, T. Chang, and A. Arbabian, “A miniaturized single-transducer implantable pressure sensor with time-multiplexed ultrasonic data and power links,” IEEE J. Solid-State Cir., vol. 53, no. 4, pp. 1089-1101, Apr. 2018.

[131] Y. Sun, N. H. Hieu, C. Jeong, and S. Lee, “An integrated high-performance active rectifier for piezoelectric vibration energy harvesting system,” IEEE Trans. Power Electr., vol. 27, pp. 623-627, Feb. 2012.

[132] D. A. Sanchez, J. Leicht, E. Jodka, E. Fazel, and Y. Manoli, “A 4μW-to-1mW parallel-SSHI rectifier for piezoelectric energy harvesting of periodic and shock excitations with inductor sharing, cold start-up and up to 681% power extraction improvement,” IEEE Int. Solid-State Cir. Conf. (ISSCC), pp. 366-367, Feb. 2016.

[133] D. Kwon and G. Rincon-Mora, “A single-inductor 0.35-μm CMOS energy investing piezoelectric harvester,” IEEE Int. Solid-State Cir. Conf. (ISSCC), pp. 78-79, Feb. 2013.

107

[134] P. Gasnier, J. Willemin, S. Boisseau, C. Despesse, C. Condemine, G. Gouvenet, and J. Chaillout, “An autonomous piezoelectric energy harvesting IC based on a synchronous multi- shot technique,” IEEE J. Solid-State Cir., vol. 49, pp. 1561-1570, July 2014.

[135] A. Morel, A. Quelen, P. Gasnier, R. Grezaud, S. Monfray, A. Badel, and G. Pillonnet, “A shock-optimized SECE integrated circuit,” IEEE J. Solid-State Cir, vol. 53, no. 12, pp. 3420- 3433, Dec. 2018.

[136] T. Hehn, F. Hagedorn, D. Maurath, D. Marinkovic, I. Kuehne, A. Frey, and Y. Manoli, “A fully autonomous integrated interface circuit for piezoelectric harvesters,” IEEE J. Solid-State Cir., vol. 47, no. 9, pp. 2185-2198, Sep. 2012.

[137] Y. Peng, D. K. Choo, S. Oh, I. Lee, T. Jang, Y. Kim, J. Lim, D. Blaauw, and D. Sylvester, “An adiabatic sense and set rectifier for improved maximum-power-point tracking in piezoelectric harvesting with 541% energy extraction gain,” IEEE Int. Solid-State Cir. Conf. (ISSCC), pp. 422-423, Feb. 2019.

[138] J. Yun, S. N. Patel, M. S. Reynolds, and G. D. Abowd, “Design and performance of an optimal inertial power harvester for human-powered devices,” IEEE Trans. Mobile Comp., vol. 10, pp. 669-683, May 2011.

[139] H. G. Yeo, T. C. Xue, S. Roundy, X. Ma, C. Rahn, and S. Trolier-McKinstry, “Strongly (001) oriented bimorph PZT film on metal foils grown by rf-sputtering for wrist-worn piezoelectric energy harvesters,” Adv. Funct. Mat. vol. 28, no. 36, p. 1801327, July 2018.

[140] Y. K. Ramadass, “Energy processing circuits for low-power applications,” PhD. Dissertation, Massachusetts Institute of Technology, Cambridge, MA, Jun. 2009.

[141] H. Lee and M. Ghovanloo, “A high frequency active voltage doubler in standard CMOS using offset-controlled comparators for inductive power transmission,” IEEE Trans. Biomed. Cir. and Syst. vol. 7, no. 3, pp. 213-224, June 2013.

108

VITA

Miao Meng

Miao Meng received his B.S. in Electrical and Computer Engineering from University of

Wisconsin Madison, Wisconsin, in May 2011, and M.S. in Electrical Engineering from Columbia

University, New York, NY in December 2012. He interned as an IC Designer and worked on level shifter optimization for I/O interface at IBM, China. In Aug. 2014, he joined Integrated Circuit and System Laboratory (ICSL) at Pennsylvania State University pursuing his doctorate degree in

Electrical Engineering under the guidance of Professor Mehdi Kiani on ultrasound-based wireless power transfer and data communication to/with miniature implantable devices and power management for piezoelectric energy harvesting.