Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

On the design of high-voltage analog front-end circuits for capacitive micromachined ultrasonic transducers… Behnamfar, Parisa 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2014_november_behnamfar_parisa.pdf [ 16.32MB ]
Metadata
JSON: 24-1.0135571.json
JSON-LD: 24-1.0135571-ld.json
RDF/XML (Pretty): 24-1.0135571-rdf.xml
RDF/JSON: 24-1.0135571-rdf.json
Turtle: 24-1.0135571-turtle.txt
N-Triples: 24-1.0135571-rdf-ntriples.txt
Original Record: 24-1.0135571-source.json
Full Text
24-1.0135571-fulltext.txt
Citation
24-1.0135571.ris

Full Text

On the Design of High-Voltage Analog Front-End Circuitsfor Capacitive Micromachined Ultrasonic Transducers(CMUT)byParisa BehnamfarB.Sc. Electrical Engineering-Electronics, Isfahan University of Technology, 2004M.Sc. Electrical Engineering-Electronics, Isfahan University of Technology, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Electrical and Computer Engineering)The University of British Columbia(Vancouver)September 2014© Parisa Behnamfar, 2014AbstractIn ultrasound imaging, capacitive micromachined ultrasonic transducer (CMUT)technology has become a promising alternative to conventionalpiezoelectric-based technology. This work focuses on various aspects ofCMUT-based imaging technologies. In the context of CMUT design andintegration with associated electronics, flexible and reliable CMUT models thatcan be seamlessly simulated with the read-out circuits and provide insights in thesystem-level performance are of great importance. This work proposes a genericVerilog-AMS model for CMUT sensors that takes into account the non-linearities,dynamic behavior and harmonic resonances of the CMUT. This model is able toprovide reliable estimations of the pull-in voltage as well as the resonancefrequency and the spring softening effect.To improve the signal-to-noise ratio (SNR), integrating the CMUT transducerwith the front-end electronics is critical. Design and implementation of acomprehensive analog front-end system in a 0.8 µm high-voltage CMOStechnology which includes high-voltage and fast-switching transmitters as well aslow-power variable-gain receivers is presented. Co-simulation of the front-endelectronics and the CMUT model demonstrates full system functionality.iiExperimental results of the system at the transmit mode confirm the reliability ofthis co-simulation. An on-chip adaptive biasing unit (ABU) is also included in thedesign which aims to improve the CMUT receive sensitivity. The ABU consists ofa DC-DC converter to generate a range of bias voltage levels and a digital controlunit to select the desired voltage. Co-simulation of the ABU with theVerilog-AMS model confirms the increase in the CMUT sensitivity in receivemode.In the context of CMUT super-resolution imaging, we present the design of atransceiver circuit in a 0.35 µm high-voltage CMOS technology that supportsboth the fundamental and asymmetric modes of operation. The transmitterprovides high-voltage pulses to the CMUT electrodes. The receiver includestransimpedance analog adders to add the fundamental mode in-phase signals aswell as differential amplifiers to combine the out-of-phase signals of theasymmetric modes. Furthermore, low-power variable-gain stages are included toamplify the resulting signals and facilitate interfacing to the ultrasound imagingmachine for additional processing and display. The design functionality isconfirmed by experimental results.iiiPrefaceI, Parisa Behnamfar, am the principal contributor of all chapters. Dr. ShahriarMirabbasi who is the research supervisor has provided technical support andediting assistance on the manuscript. Dr. Robert Rohling and Dr. Edmond Cretuhave provided technical support for the project, particularly the work presented inChapter 2 and Chapter 4. Dr. Reza Molavi has provided technical support in thedesign and test of the prototype chip presented in Chapter 6. Fabrication of theCMUT chip used in this thesis was done by Ming Cai, M.A.Sc.As mentioned below, some of the content of this thesis is written based on thefollowing published or submitted works:Conference Papers:1. P. Behnamfar and S. Mirabbasi, “Design of a high-voltage analog front-endcircuit for integration with cmut arrays,” in Biomedical Circuits and SystemsConference (BioCAS), © 2010 IEEE, 2010, pp. 298−301 [1]→ Chapter 1and Chapter 3.2. P. Behnamfar and S. Mirabbasi, “A cmut read-out circuit with improved receivesensitivity using an adaptive biasing technique,” in Biomedical Circuits andivSystems Conference (BioCAS), © 2011 IEEE, 2011, pp. 397−400 [2] →Chapter 1 and Chapter 5.3. P. Behnamfar, R. Molavi, and S. Mirabbasi, “Receiver design for cmut-basedsuper-resolution ultrasound imaging,” in International Symposium onCircuits and Systems (ISCAS), © 2014 IEEE, 2014, pp. 878−881 [3] →Chapter 6.Journal Papers:1. P. Behnamfar, S. Mirabbasi, E. Cretu, and R. Rohling “Verilog-AMSMacro-Modeling of CMUTs for System Level Simulations”, submitted. →Chapter 2 and Chapter 4.2. P. Behnamfar, R. Molavi, and S. Mirabbasi, “Transceiver design for cmut-based super-resolution ultrasound imaging,” submitted. → Chapter 3 andChapter 6.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Medical Ultrasound Imaging . . . . . . . . . . . . . . . . . . . . 21.2.1 History of Ultrasound . . . . . . . . . . . . . . . . . . . 21.2.2 Ultrasound Imaging Machines . . . . . . . . . . . . . . . 41.3 Ultrasound Transducers . . . . . . . . . . . . . . . . . . . . . . . 5vi1.3.1 Magnetostriction . . . . . . . . . . . . . . . . . . . . . . 51.3.2 Piezoelectricity . . . . . . . . . . . . . . . . . . . . . . . 61.3.3 Electrostatics . . . . . . . . . . . . . . . . . . . . . . . . 61.4 CMUT Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.4.1 CMUT Structure and Operation . . . . . . . . . . . . . . 81.4.2 CMUT Advantages and Drawbacks Compared toPiezoelectric Transducers . . . . . . . . . . . . . . . . . 101.4.3 CMUT Applications . . . . . . . . . . . . . . . . . . . . 121.5 Integration of CMUT and CMOS Electronics . . . . . . . . . . . 131.5.1 Overview of the Analog Front-End Circuitry . . . . . . . 141.5.2 High-Voltage Design Considerations . . . . . . . . . . . . 161.6 Research Objectives and Contributions . . . . . . . . . . . . . . . 181.7 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 Verilog-AMS Macro-Modeling of CMUTs for System LevelSimulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.2 Different CMUT Modeling Approaches . . . . . . . . . . . . . . 252.2.1 Finite Element Modeling . . . . . . . . . . . . . . . . . . 252.2.2 Mason’s Equivalent Circuit Model . . . . . . . . . . . . . 252.2.3 CMUT as a Single Fixed-Value Capacitor . . . . . . . . . 262.3 The Proposed Verilog-AMS Model for CMUTs . . . . . . . . . . 272.3.1 Fundamental Mode Considerations . . . . . . . . . . . . 312.3.2 Harmonic Mode Considerations . . . . . . . . . . . . . . 322.3.3 Modeling the Acoustic Interaction with the Medium . . . 34vii2.4 Simulation Results of the CMUT Verilog-AMS Model . . . . . . 352.5 CMUT Measurement Results . . . . . . . . . . . . . . . . . . . . 382.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 High-Voltage Analog Front-End Circuits for Integration withCMUT Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.2 Overview of the Analog Front-End Circuitry . . . . . . . . . . . . 453.3 High-Voltage Transmitter Architectures . . . . . . . . . . . . . . 473.3.1 HV Transmitter Circuit: The Level-Shifted Design . . . . 483.3.2 HV Transmitter Circuit: The Cross-Coupled Design . . . 503.4 Variable-Gain Receiving Amplifier Circuit . . . . . . . . . . . . . 513.5 Protection and Enable Switches . . . . . . . . . . . . . . . . . . . 553.6 Post-Layout Simulation Results of the Analog Front-End System . 573.6.1 Transmitters’ Post-Layout Simulation Results . . . . . . . 603.6.2 Receiver’s Post-Layout Simulation Results . . . . . . . . 623.7 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . 643.7.1 Level-Shifted Transmitter Characterization . . . . . . . . 643.7.2 Cross-Coupled Transmitter Characterization . . . . . . . . 663.7.3 Transimpedance Receiver Characterization . . . . . . . . 673.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674 Co-Simulation of Transceiver Circuits and CMUT Devices andOptical Measurements of the Transmitter System . . . . . . . . . . 704.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70viii4.2 Co-Simulation of the High-Voltage Level-Shifted Transmitter withthe CMUT Model . . . . . . . . . . . . . . . . . . . . . . . . . . 714.2.1 Front-End System Simulations for the Symmetric CMUT 724.2.2 Front-End System Simulations for the Asymmetric CMUT 764.3 Co-Simulation of the Full Transmit/Receive Circuit with theCMUT’s Verilog-AMS Model . . . . . . . . . . . . . . . . . . . 774.4 Optical Measurements . . . . . . . . . . . . . . . . . . . . . . . 804.4.1 Symmetric CMUT Optical Measurements . . . . . . . . . 814.4.2 Asymmetric CMUT Optical Measurements . . . . . . . . 844.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865 A CMUT Read-Out Circuit with Improved Receive Sensitivity Usingan Adaptive Biasing Technique . . . . . . . . . . . . . . . . . . . . 885.1 Introduction and Motivation . . . . . . . . . . . . . . . . . . . . 885.2 Overview of the Front-End Circuit . . . . . . . . . . . . . . . . . 915.3 Design of the Adaptive Biasing Unit . . . . . . . . . . . . . . . . 925.3.1 On-Chip SC DC-DC Converter . . . . . . . . . . . . . . 935.3.2 Digital Control Unit . . . . . . . . . . . . . . . . . . . . 945.3.3 Modifications to the Transceiver-CMUT Interface . . . . . 955.4 Post-Layout Simulation and Experimental Results . . . . . . . . . 965.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1016 Transceiver Design for CMUT-Based Super-Resolution UltrasoundImaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1036.1 Introduction and Background . . . . . . . . . . . . . . . . . . . . 1036.2 High-Voltage Transmitter Design . . . . . . . . . . . . . . . . . . 108ix6.3 Receiver Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 1096.3.1 Receiver RX2k−1 . . . . . . . . . . . . . . . . . . . . . . 1106.3.2 Receiver RX2k . . . . . . . . . . . . . . . . . . . . . . . . 1136.4 Post-Layout Simulation Results . . . . . . . . . . . . . . . . . . 1146.4.1 High-Voltage Transmitter’s Post-Layout Simulation Results 1146.4.2 RX2k−1 Post-Layout Simulation Results . . . . . . . . . . 1156.4.3 RX2k Post-Layout Simulation Results . . . . . . . . . . . 1196.5 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . 1216.5.1 High-Voltage Transmitter’s Experimental Results . . . . . 1236.5.2 Receivers RX2k−1 and RX2k Experimental Results . . . . . 1246.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1257 Conclusion and Future Work . . . . . . . . . . . . . . . . . . . . . . 1287.1 Research Contributions . . . . . . . . . . . . . . . . . . . . . . . 1287.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133xList of TablesTable 2.1 Frequently used parameters in this work. . . . . . . . . . . . . 28Table 3.1 Ultrasound machine specifications . . . . . . . . . . . . . . . 47Table 3.2 Comparison of the post-layout simulation result of thetransimpedance receiver with similar designs . . . . . . . . . . 68Table 3.3 Comparison of the experimental results of the two types offabricated transmitters with previously reported designs . . . . 68Table 6.1 Comparison of the experimental results of the transimpedancereceiver with previously reported designs . . . . . . . . . . . . 126Table 6.2 Comparison of the experimental results of the high-voltagetransmitter with previously reported designs . . . . . . . . . . 127xiList of FiguresFigure 1.1 Typical 2-D ultrasound image of fetal fingers at 20 weeks [18]and 3-D fetal head at 36 weeks [19]. . . . . . . . . . . . . . . 3Figure 1.2 View of an ultrasound imaging machine and various types oftransducers: (a) SonixRP ultrasound imaging machine [25].(b) Three different typical ultrasound transducers and theirapplications[26]. . . . . . . . . . . . . . . . . . . . . . . . . 4Figure 1.3 Cross section of a typical CMUT transducer. . . . . . . . . . 8Figure 1.4 1-D CMUT anatomy and concepts of cell, element and array. . 9Figure 1.5 Block diagram of the transceiver circuit and interfaces to theCMUT transducer and ultrasound machine. . . . . . . . . . . 15Figure 2.1 The Mason’s two port CMUT equivalent circuit model. . . . . 25Figure 2.2 Two different types of CMUT cells used in this work:asymmetric (top) and symmetric (bottom). . . . . . . . . . . . 27Figure 2.3 The typical ideal MEMS parallel-plate model: the left siderepresents the electrical port and the right side models themechanical port of the CMUT. . . . . . . . . . . . . . . . . . 30xiiFigure 2.4 Asymmetric CMUT fundamental and second harmonic modevibrations [60]. . . . . . . . . . . . . . . . . . . . . . . . . . 33Figure 2.5 CMUT simulation test setup. . . . . . . . . . . . . . . . . . . 35Figure 2.6 Simulation results of the symmetric CMUT’s input impedancein air. The peak is at 2.73 MHz. . . . . . . . . . . . . . . . . 36Figure 2.7 Simulation results of the symmetric CMUT’s input impedanceduring immersion in oil. . . . . . . . . . . . . . . . . . . . . 36Figure 2.8 Simulation results of the fundamental frequency behavior ofthe asymmetric CMUT’s input impedance in air. The peak isat 2.69 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . 37Figure 2.9 Simulation results of the fundamental frequency behavior ofthe asymmetric CMUT’s input impedance during immersionin oil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37Figure 2.10 Simulation results of the harmonic frequency behavior of theasymmetric CMUT’s input impedance in air. The peak is at5.4 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38Figure 2.11 Simulation results of the harmonic frequency behavior of theasymmetric CMUT’s input impedance during immersion inoil. The under-damped phenomenon results in a peak at4.07 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Figure 2.12 Experimental results of the symmetric CMUT inputimpedance in air medium. The peak is at 2.57 MHz. . . . . . 39Figure 2.13 Experimental results of the symmetric CMUT inputimpedance in olive oil medium. . . . . . . . . . . . . . . . . 40xiiiFigure 2.14 Experimental results of the asymmetric CMUT inputimpedance in air medium. The fundamental mode peak is at2.5 MHz, and harmonic mode peak is at 5.1 MHz. . . . . . . 41Figure 2.15 Experimental results of the asymmetric CMUT inputimpedance in olive oil medium. The fundamental mode peakis over-damped and the harmonic mode peak is shifted to4.1 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Figure 3.1 Block diagram of the transceiver circuit and interfaces to theCMUT transducer and ultrasound machine. . . . . . . . . . . 46Figure 3.2 Low voltage (LV) and high voltage (HV) PMOS and NMOStransistor symbols used in the schematics of this thesis. . . . . 48Figure 3.3 The level-shifted high-voltage transmitter schematic with itsenable switch. . . . . . . . . . . . . . . . . . . . . . . . . . . 49Figure 3.4 High-voltage transmitter architecture, the cross-coupled design. 51Figure 3.5 The receiving circuit consisting of the input variable-gaintransimpedance amplifier, op-amp and output buffers. . . . . . 53Figure 3.6 Schematic of the switch block used in the level-shiftedtransmitter to isolate the gate of transistor N2 in Figure 3.3. . . 55Figure 3.7 The layout for both transmitters, Left: level-shifted transmitterand Right: cross-coupled transmitter. . . . . . . . . . . . . . 57Figure 3.8 (a) The 16 channel transceiver layout. (b) Layout of the printedcircuit board designed to test the 16 channel transceiver chip. . 58xivFigure 3.9 (a) The transceiver chip layout. (b) Layout of the printedcircuit board designed to test the transceiver and the CMUTtogether. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Figure 3.10 (a) The cross-coupled transmitter chip layout. (b) Layout ofthe two PCBs designed to test the cross-coupled transmitterchip (the PCB on the left) and to wire-bond and test thetransmitter and CMUT chips (the PCB on the right). . . . . . 60Figure 3.11 Post-layout simulation transient response of the level-shiftedtransmitter for 20 V high voltage supply and 18 pf capacitiveload. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61Figure 3.12 Post-layout simulation transient response of the cross-coupledtransmitter for 20 V high voltage supply and 18 pf capacitiveload. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Figure 3.13 Post-layout simulation results of the variation of the gain and3 dB bandwidth versus control voltage for the variable-gaintransimpedance amplifier. . . . . . . . . . . . . . . . . . . . 63Figure 3.14 (a) The level-shifted transmitter chip micrograph. (b) Thefabricated and assembled PCB with the transceiver 24 pinpackage on the left and the CMUT 209 PGA package on theright. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.15 Experimental transient response of the level-shiftedtransmitter’s input and output voltages. . . . . . . . . . . . . 65Figure 3.16 (a) The cross-coupled transmitter chip micrograph. (b) Thefabricated and assembled PCB for testing the cross-coupledtransmitter. . . . . . . . . . . . . . . . . . . . . . . . . . . . 66xvFigure 3.17 Experimental transient response of the cross-coupledtransmitter’s input and output voltages. . . . . . . . . . . . . 67Figure 4.1 The level-shifted high-voltage transmitter connected to theCMUT’s electrical input terminal. . . . . . . . . . . . . . . . 71Figure 4.2 Symmetric CMUT normalized velocity versus frequency inair, for constant AC and different DC bias voltages. . . . . . . 72Figure 4.3 Symmetric CMUT normalized velocity versus frequency inoil, for constant AC and different DC bias voltages. . . . . . . 73Figure 4.4 Dependence of CMUT membrane displacement amplitude andresonance frequency on the applied DC bias. . . . . . . . . . 73Figure 4.5 Transient analysis simulation results of the symmetricCMUT’s input and output waveforms. The results showunder-damping behavior of the air medium. . . . . . . . . . . 74Figure 4.6 Transient analysis simulation results of the symmetricCMUT’s input and output waveforms. The results showover-damping behavior of the olive oil medium. . . . . . . . . 75Figure 4.7 Normalized velocity versus frequency, in air and oil media, forthe symmetric excitation of the asymmetric CMUT cell (at 30V DC bias). . . . . . . . . . . . . . . . . . . . . . . . . . . . 76Figure 4.8 Normalized velocity versus frequency, in air and oil media, forthe asymmetric excitation of the asymmetric CMUT cell (at 30V DC bias). . . . . . . . . . . . . . . . . . . . . . . . . . . . 77Figure 4.9 Block diagram of the test-setup for co-simulation of the fulltransmit/receive circuits and CMUT Verilog-AMS model. . . 78xviFigure 4.10 Results of the co-simulation of the transmitter and receivercircuits with the CMUT Verilog-AMS model. . . . . . . . . . 79Figure 4.11 The fabricated and assembled PCB under the laser Dopplervibrometer system, including the transceiver and CMUTpackages . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80Figure 4.12 LDV measurement of the symmetric CMUT’s absolute valuedeflections in air at different vibration modes. . . . . . . . . . 82Figure 4.13 LDV measurement results of the symmetric CMUT membranedisplacement versus frequency at transmit mode in air, for 60 VDC and 20 V AC input. . . . . . . . . . . . . . . . . . . . . . 83Figure 4.14 Experimental results for the symmetric CMUT’s displacementand resonance frequency changes versus DC bias voltage intransmit mode. . . . . . . . . . . . . . . . . . . . . . . . . . 83Figure 4.15 Measuring the cross-talk on neighbor cells by calculating theratio of the neighbor cells displacement to the actuated CMUTdisplacement. . . . . . . . . . . . . . . . . . . . . . . . . . . 84Figure 4.16 LDV measurement of the asymmetric CMUT’s absolute valuedeflections in air, at different vibration modes. . . . . . . . . . 85Figure 4.17 LDV measurement results for the asymmetric CMUT’smembrane displacement versus frequency in air, at transmitmode for 30 V DC and 5 V AC input. . . . . . . . . . . . . . 86Figure 4.18 LDV measurement of the symmetric actuation of theasymmetric CMUT’s bottom electrodes. The membranedisplacement shows no second harmonic vibration in air. . . . 87xviiFigure 4.19 An asymmetric CMUT with burnt interconnects is shown onthe right. The interconnects were not designed properly to beable to handle the current density for DC bias voltages morethan 30 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87Figure 5.1 A typical CMUT transducer’s efficiency versus DC bias [43]. . 89Figure 5.2 Block diagram of the front-end circuitry including thetransmitter, receiver and adaptive biasing unit. . . . . . . . . . 92Figure 5.3 Circuit schematic of the DC-DC converter. . . . . . . . . . . 93Figure 5.4 Block diagram of the counter and the digital logic forproducing the control and VGC signals in Figure 5.2. . . . . . 95Figure 5.5 Transceiver-CMUT interface and the protection scheme. . . . 96Figure 5.6 (a) The DC-DC converter layout and (b) Layout of the PCBwhich is designed to test the DC-DC converter’s functionality. 97Figure 5.7 Simulation results of the outputs of the on-chip SC DC-DCconverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 5.8 (a) The DC-DC converter chip micrograph and (b) Thefabricated PCB to test the 40 pin IC package which containsthe DC-DC converter chip. . . . . . . . . . . . . . . . . . . . 99Figure 5.9 Experimental results of the outputs of the on-chip SC DC-DCconverter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 5.10 Simulation setup to test the adaptive biasing concept. . . . . . 100Figure 5.11 Post-layout simulation results for CMUT2 output (Top) forvariable DC bias levels (Bottom). . . . . . . . . . . . . . . . 101xviiiFigure 6.1 Asymmetric CMUT fundamental and second harmonic modesof vibration, measured by a laser Doppler vibrometer: (a)Fundamental mode, (b) Second harmonic (asymmetric) mode. 104Figure 6.2 Block diagram of the front-end circuitry that interfaces to theodd (darker color) and even row (lighter color) CMUTelements and ultrasound machine. . . . . . . . . . . . . . . . 106Figure 6.3 Circuit schematic of one of the two TX2k−1 high-voltagetransmitters. . . . . . . . . . . . . . . . . . . . . . . . . . . . 108Figure 6.4 Block diagrams and circuit schematics of the odd and even rowreceivers: (a) Receiver RX2k−1 and (b) Receiver RX2k. . . . . . 110Figure 6.5 (a) The full transceiver chip layout and (b) Layout of the PCBwhich is designed to test the transceiver circuit and the CMUTtransducer. . . . . . . . . . . . . . . . . . . . . . . . . . . . 114Figure 6.6 Post-layout simulation results of the high voltage transmitterinput and output voltages. . . . . . . . . . . . . . . . . . . . 115Figure 6.7 Post-layout simulation results of the RX2k−1 receiver: transientresponse of the in-phase pulse input currents and the outputvoltage at 1.38 MHz. . . . . . . . . . . . . . . . . . . . . . . 116Figure 6.8 Post-layout simulation results of the RX2k−1 receiver:transient response of the in-phase sinusoidal input currentsand the output voltage at 4.35 MHz. . . . . . . . . . . . . . . 117xixFigure 6.9 Post-layout simulation results: frequency response of theRX2k−1 gain and input impedance for two settings of the gainand bias voltages. Setting 1 (minimum gain): gain controlvoltage = 0 V, bias voltage = 2.2 V. Setting 2 (maximumgain): gain control voltage = 1 V, bias voltage = 1.85 V. . . . . 118Figure 6.10 Post-layout simulation results of the RX2k receiver: transientresponse of the out-of-phase sinusoidal input currents and theoutput voltage at 1.55 MHz. . . . . . . . . . . . . . . . . . . 119Figure 6.11 Post-layout simulation results: frequency response of the RX2kgain and input impedance for two settings of the gain and biasvoltages. Setting 1 (minimum gain): gain control voltage =0 V, bias voltage = 2.2 V. Setting 2 (maximum gain): gaincontrol voltage = 1 V, bias voltage = 1.85 V. . . . . . . . . . . 120Figure 6.12 (a) Transceiver chip micrograph and (b) Fabricated andassembled PCB with both transceiver and CMUT packages. . 121Figure 6.13 Experimental transient response of the high voltage transmitterinput and output voltages, the high voltage supply is set to twovalues of 30 V and 60 V and the input frequency is 1.38 MHz. 123Figure 6.14 Experimental transient response of the RX2k−1 receiver. Thetwo input voltages are in-phase pulses with frequency of1.38 MHz and 500 mV peak-to-peak amplitude. The output isAC coupled on the oscilloscope. . . . . . . . . . . . . . . . . 124xxFigure 6.15 Experimental transient response of the RX2k receiver. The twoinput voltages are out-of-phase sinusoidal signals withfrequency of 1.55 MHz and 300 mV peak-to-peak amplitude.The output is AC coupled on the oscilloscope. . . . . . . . . . 125Figure 6.16 Comparison of the simulated and measured frequencyresponses of the two receivers’ transimpedance gain. . . . . . 126xxiAcknowledgmentsI would like to express my gratitude to my research supervisor Dr. ShahriarMirabbasi for his technical support and advice on the project. I am also verythankful to Dr. Edmond Cretu and Dr. Robert Rohling for their on-goingmotivation and guidance for this project.I am grateful to the professors who served on my Ph.D. examination committeeand provided valuable feedback on my research and thesis.Dr. Roberto Rosales from UBC SOC lab deserves great thanks for his guidanceand technical support in different aspects including sharing his vast knowledge onchip test procedures, equipment selection, test setup preparations, and running testsand debugging chips. In addition, he has been a true friend and mentor who willbe always remembered and referred to. I sincerely acknowledge him, as I havelearned a lot from him during these years.Hadi Najar, Wei You, Ming Cai and Carlos D. Gerardo from UBC MiNa laband Samuel Frew from UBC Robotics and Control lab were involved in this projectand gave valuable feedback in different stages of the project. I also thank Mr. Pui-Him Wai from Rohde & Schwarz and Dr. Karen Cheung and Nomin Oyun fromUBC MiNa lab for equipment support. Also, Roozbeh Mehrabadi from UBC ITxxiiprovided CAD tool support. I would also like to thank all my lab mates in theSystem-On-a-Chip (SOC) lab for always being supportive throughout these years.Thanks to CMC Microsystems for providing the fabrication and simulationsoftware support and to NSERC for funding the project.I can’t express enough my sincere gratitude to my family. My parents foralways believing in me and their ever increasing comprehensive support. My sisterand brothers for the inspiration they gave me by being great role models. Last butnot least, my husband, Hadi, for all his support and patience in the stressful timesof tape-out deadlines, writing the PhD dissertation and the PhD defense.xxiiiDedicationTo my parents.xxivChapter 1Introduction 11.1 MotivationUltrasound imaging systems are the primary tool world-wide for fetal anddiagnostic imaging [4, 5]. However, their cost and logistics of transportation ofsuch systems to rural areas are still a serious challenge that make themunavailable in remote areas of most developing countries. Therefore there is agenuine need for revolutionary techniques in manufacturing such equipment toovercome the aforementioned obstacles.Other circumstances in which the need for portable or hand-held ultrasoundequipment is essential are the natural disasters, the life-threatening accidents andthe battlefields. As most of the mortalities happen in the first hours of the injury,fast and reliable access of the search and rescue or emergency teams to diagnostictools can save lives. These tools should have light weight and extended battery life(i.e., minimal power consumption)[6].1 Part of Section 1.5 was published in the Proceedings of the IEEE BioCAS2010 [1]. Part ofSection 1.4 was published in the Proceedings of the IEEE BioCAS2011 [2].1In addition, there are various potential applications of ultrasound in medicinewhich are not pursued due to the shortcomings of the conventional ultrasounddevices. Harmonic imaging and super-resolution imaging are just two samples ofsuch applications.Fortunately, the last two decades of research on alternative ultrasoundtechnologies have demonstrated a promising new generation of such deviceswhich not only have lighter weights and lower costs, but also provide variety ofother diagnostic and therapeutic opportunities; some notable applications aredermal and eye cancer diagnosis [7–10].1.2 Medical Ultrasound Imaging1.2.1 History of UltrasoundUltrasound refers to sound waves with frequencies above the audible range of20 Hz to 20 kHz, i.e., in the MHz range. The most prominent applications ofultrasound are in medicine, chemistry and biology [5, 11], underwater detection[12], industrial cleaning [13], flow metering [14] and non-destructive evaluation(NDE) of materials [15].In medicine, some of the on-going research areas related to ultrasound include:surgery with ultrasound, cancer therapy using high intensity focused ultrasound(HIFU), photoacoustic imaging (PAI), intravascular ultrasound imaging systems(IVUS), image guided therapy (IGT) and real-time two, three or four dimensional(2-D, 3-D and 4-D) ultrasound medical imaging of soft tissues such as abdominalorgans, heart, and blood vessels.Although the initial diagnostic utilization of ultrasound dates back to early2Figure 1.1: Typical 2-D ultrasound image of fetal fingers at 20 weeks [18]and 3-D fetal head at 36 weeks [19].1940s [16], the first commercial real-time ultrasound scanner, Vidoson, wasmanufactured by Siemens Medical Systems in 1965 [17]. Advances in ultrasoundmedical imaging from 1965 to early 1990s include the invention of first gray-scaleB-mode and first 3-D fetal imaging. Revolutionary developments insemiconductor industry and creation of powerful computers in 1990s have helpedultrasonic medical imaging and therapy, and demonstrated many possibilities in3-D, 4-D and tissue harmonic imaging [16]. Figure 1.1 shows a typical 2-Dultrasound image of fetal fingers at 20 weeks and 3-D fetal head at 36 weeks[18, 19].Currently, almost 25% of all clinical imaging studies are performed usingultrasound equipment [20]. Compared to other imaging techniques such ascomputed tomography (CT) and magnetic resonance imaging (MRI), ultrasoundimaging has an excellent safety record [21] and lower cost, and is a real-timeprocedure [22]. In addition, the ease of portability of the recently developeddevices alleviates the previous problems for their transportation to remote areas.One notable example of these portable devices is Philips′s CX50 CompactXtreme3(a) (b)Figure 1.2: View of an ultrasound imaging machine and various types oftransducers: (a) SonixRP ultrasound imaging machine [25]. (b) Threedifferent typical ultrasound transducers and their applications[26].ultrasound system, which aims at reducing maternal and child mortality in Africa[23, 24].1.2.2 Ultrasound Imaging MachinesFigure 1.2a shows a commercial modern ultrasound imaging machine fromUltrasonix Medical Corporation [25] composed of four main parts: The hand-heldtransducer probe, the analog front-end electronics, the hardware for the back-enddigital signal processing, and the display system. Three typical probes are alsoshown in Figure 1.2b along with the detail of their frequency range and mainapplications [26].Most of the ultrasound instruments are pulse-echo systems in which thetransducer in their hand-held probe emits pulses and senses the respective echos.The probe is connected to the supporting electronic unit with a bundle of cables.4The analog front-end electronics unit is responsible for the pulse excitation andreception. In a typical probe, an electrical pulse is applied to the transducer togenerate an acoustic wave. This wave is propagated into the medium of interest(i.e., the patient body) and reflected back to the transducer from different tissueboundaries. The transducer senses the resulting echo and converts it into anelectrical signal which is then sent to the receiving amplifier of the analogfront-end unit. The received signal is further processed in the signal processingunit and is sent to the imaging console.Extensive advances in the digital signal processing and image reconstructionmethods have introduced many possibilities in ultrasound data processingalgorithms, image enhancements and analysis. However, there is still room forimprovement for transducer and analog front-end units to support higher signalquality (e.g., bandwidth and signal to noise ratio (SNR)). In the last two decades,these units and particularly the transducer unit have seen revolutionarydevelopments and improvements [8, 27, 28].1.3 Ultrasound TransducersUltrasound transducers convert electrical or magnetic energy to ultrasound energyand vice versa. The three well-known mechanisms for such conversion aremagnetostriction, piezoelectricity and electrostatics.1.3.1 MagnetostrictionIn magnetostriction, which was first observed in 1842 by James P. Joule, applyinga magnetic field to a material changes its physical dimensions and vice versa. Iron,cobalt and nickel are three commonly known materials in this category. However,5challenges such as eddy currents’ losses at high frequencies and hysteresis effectsrestrict the application of magnetostrictive materials in ultrasound transducers.1.3.2 PiezoelectricityCurrent transducers are predominantly made of piezoelectric materials.Piezoelectricity was first discovered in 1880 by the Curie brothers and have beenresearched and utilized since then. Single crystal quartz, fabricated ceramics likelead zirconate titanate, known as PZT, and piezocomposites are the most popularmaterials in this category. In piezoelectric transducers, the mechanical stressimposes bound electrical charges on the electrodes of the device. Conversely, analternating voltage with certain frequency range applied to the electrodes impliesmechanical movement of the transducer membrane, resulting in generation ofultrasound waves. However, the low electro-mechanical efficiency of thesematerials as well as the acoustic impedance mismatching between the piezomaterials and body tissues have always been a struggle for the transducermanufacturers. Moreover, these types of transducers are not amenable tointegration with the interface electronics and also there are many technicaldifficulties to expand a 2-D array of piezoelectric sensors to a 3-D imaging system[7, 29]. This difficulty results in bulky and expensive equipments which are not anaffordable choice particularly in rural areas of the low-income countries.1.3.3 ElectrostaticsIn electrostatic transduction, a variable electric field between two parallel plates,one stationary and one floating, will result in mechanical movement (vibration) ofthe floating plate. This vibration can generate acoustic waves. Furthermore,6acoustic waves from the medium apply mechanical pressure on the floating plateand vibrate it. This vibration generates a varying electrical current which can besensed and processed. Although electrostatic transduction is not a new theory andhas been studied since the early 1900s, practical implementations were notrealized until two decades ago. In order to have similar performance as thepiezoelectric transducers, electric field strengths of the order of one million voltsper centimeter (106 V/cm) are needed on the parallel plates to generate therequired acoustic pressure [30]. Therefore, implementing electrostatictransduction only became possible through advances in micromachining in early1990s [31].It was in 1994 when the first device with reasonable cavity size and electricfield requirement was fabricated in Professor Khuri-Yakub’s research group atStanford University [32]. This was the introduction of an alternative transducertechnology called the Capacitive Micromachined Ultrasonic Transducer(CMUT), which was made possible by the advances in miniaturization capabilityof silicon micromachining. Since then, extensive research on manufacturing andapplications of different types of CMUTs have resulted in many publications[1–3, 8–10, 27, 29, 31, 33–35] and recent commercialization of this type oftransducers by companies such as Hitachi (Tokyo, Japan) [36], Vermon (Tours,France) [37], and Kolo Technologies (San Jose, CA, United States) [38].7Substrate and bottom electrode CavityMembraneTop electrodeInsulation layerFigure 1.3: Cross section of a typical CMUT transducer.1.4 CMUT Basics1.4.1 CMUT Structure and OperationCMUTs are generally micromachined thin parallel plate membranes separated bya shallow air or vacuum cavity. Top membranes are suspended over a conductivesilicon wafer and are supported by insulating posts. Top and bottom electrodes areattached to the membranes to provide electrical connections to the analogfront-end circuitry. Also, an insulation layer inside the cavity prevents electricalshorting of the membranes in case of collapsing the top membrane to the bottomone. The size and thickness of the membranes depend on the imaging application;i.e. the frequency and bandwidth of operation. If technically feasible, the cavitiesare sealed to prevent any intrusion of dust or other particles into the mechanicalstructure. This will enhance the reliability and performance of the CMUT cells[39, 40].Figure 1.3 presents the cross section of a typical CMUT transducer, showingelectrodes, membranes and the cavity. As it is shown in this figure, the CMUTstructure resembles capacitors.A 1-D CMUT array comprises of multiple CMUT elements, and each element8Figure 1.4: 1-D CMUT anatomy and concepts of cell, element and array.has multiple CMUT cells connected in a one dimensional structure (a single row).In a 2-D CMUT array, each element is a two dimensional configuration of cells,and the full array is composed of multiple elements. Figure 1.4 shows a sample1-D CMUT anatomy and the concepts of cell, element and array.CMUT sensors are used for transmission and reception of ultrasound signals.In transmission, an electrical AC signal vibrates the membranes to generate anacoustical wave. In reception, the incident waves will cause the membranes tovibrate, resulting in the capacitance change of the transducer which produces aflow of current towards the adjacent analog front-end circuit. In both cases, aDC bias larger than the AC voltage is applied to the device. It has been shownthat the sensitivity and efficiency of the device depends on the amount of this DCbias, hence biasing the CMUT with an appropriate DC voltage results in betterperformance [29, 41].By applying a DC voltage to the device, an electrostatic field forms between themembrane and the substrate. This field pulls the membrane towards the substrate.9On the other hand, there exists a mechanical restoring force of the membrane whichbalances this deflection. By increasing the bias voltage, a point is reached wherethe electrostatic force overwhelms the restoring force. At this point the membranecollapses down towards the substrate [29]. The optimal DC bias voltage, called thepull-in voltage, is close to this collapse voltage [42, 43].1.4.2 CMUT Advantages and Drawbacks Compared to PiezoelectricTransducersCMUT AdvantagesCMUT technology offers various and significant advantages over piezoelectrictransducers. These advantages make this new technology a very promisingreplacement for its piezoelectric counterpart.In the desired frequency ranges, piezoelectric materials have significantmechanical impedance mismatch with air, water and body tissues which makesthem very inefficient for delivering power. To overcome this issue, matchinglayers are used on these transducers which need costly and complicated processes.CMUTs, on the other hand, have considerably better impedance matching in theintended frequency ranges, wider operating bandwidth and consequently, betteraxial resolution. Recently, the wider operating bandwidth of CMUTs has madethem a desirable choice for second-harmonic and super-resolution imagingapplications [35].CMUT fabrication is uniform and precisely reproducible as it is created withsemiconductor fabrication techniques. It is also scalable and have excellentcapability to be made in arrays, in different shapes and with variousapplication-related and dimension specifications. This also makes them10inexpensive comparing to the piezoelectric devices.Potential for integration with CMOS readout electronic circuits is anothersignificance of CMUTs. Integration will remarkably reduce the interconnectparasitics and shrinks the overall size of the probe and the front-end electronicunits mentioned in Section 1.2.2. In addition, CMUTs operate over widertemperature range than piezoelectric sensors. This makes them a reliable andlow-cost choice in applications that use or generate high power, such as HIFU.Also, different 3-D medical imaging approaches have been introduced recentlyusing CMUT technology [10, 29, 44].CMUT DrawbacksThe main drawback of CMUTs is their lower acoustic pressure which results inlow penetration depth [39]. This, consequently, reduces the receive sensitivity anddegrades the SNR. In general, further developments in fabrication processes andCMUT structural improvements are required in order to alleviate such issue. Inthe meanwhile, some solutions for this issue have been suggested which make useof the collapse mode. By applying a DC voltage more than the pull-in voltage ofthe CMUT, the device operates in its collapse mode. Then it is released from thecollapse mode to produce much higher acoustic pressure [45–47].Compared to piezoelectric transducers, CMUTs need improvements inelectrical insulation. The issue of possible trapped charges in the CMUT’sdielectric layer is a matter of safety concern. Also, packaging the CMUT togetherwith front-end electronics needs careful considerations to provide completeisolation from body tissues so that it can be used reliably in vivo [48].Another issue with CMUTs is the cross talk between the neighboring cells.11Two of the solutions proposed for this issue are the addition of a lossy top layer[49] and fabricating the CMUT elements in double periodicity [50]. Adding thelossy layer reduces the unwanted effects but increases the center frequency. Indouble periodicity fabrication, a larger distance is considered between CMUTelements than between CMUTs within an element. This technique is shown toreduce the undesired resonances close to the center frequency but increases thelower resonance frequencies. Another method to reduce the cross-talk is using acalibrated matrix. This method does not need any physical modification of thefabrication or design process. Instead, the transmit waveform matrix isprogrammed such that the cross-talk is reduced on adjacent elements [51].1.4.3 CMUT ApplicationsMany industrial and medical applications are proposed for CMUTs. Someindustrial applications include range finding in harsh environments [52], whereCMUTs that can handle high pressures are introduced and characterized. Also,CMUT transducers have applications in sonar industry, where wide-bandwidthimaging has been an obstacle using piezoelectric devices [53].Medical imaging applications of CMUTs are vast and widespread. Theyinclude the traditional 2-D ultrasound imaging and 3-D imaging using 2-D CMUTarrays. Also, one particular application of imaging with CMUTs is catheter-basedimaging [54]. Examples are the forward-looking IVUS and intracardiacechography (ICE) imaging [55]. Since CMUTs are very flexible in array designand can integrate with front-end electronics, miniaturized ICE and IVUS cathetershave been developed that can provide real-time volumetric ultrasound images ofcoronary arteries. Another application of CMUTs is in HIFU, where compared to12piezoelectric transducers, CMUTs have less self-heating and can operate undercontinuous wave conditions [56]. The reason for less self-heating feature ofCMUTs is that they have considerably less internal losses than the piezoelectrictransducers as well as the fact that they are made of highly thermally conductivesilicon that dissipates the generated heat [54]. CMUTs are also proposed for highresolution neural stimulation of the retina as they can operate over a widefrequency range and work in high intensity modes [57].1.5 Integration of CMUT and CMOS ElectronicsAt the transmit time, the ultrasound waves spread out from the transducer atdifferent angles other than the primary one. The reflection of these non-primaryangle waves causes spurious indications on the final image. These unwantedwaves interfering the main signal are called grating or side lobes [58].The hand-held probe of the ultrasound machine is composed of several sensorelements. The size and distance of these elements significantly affect theamplitude of the grating lobes and can not exceed a specific value. In CMUTs,this size limitation results in lower element sensitivity [33]. In addition, theparasitic capacitances of the interconnect lines between the sensor and thefront-end electronics unit are typically an order of magnitude larger than thetransducer capacitances. Therefore, they cause a significant loss in the outputsignal. Due to these issues, integrating the CMUT transducer elements with thefront-end electronics is critical. The integrated CMUT and analog front-endcircuits that have been developed and tested using flip-chip bonding [28, 33] andmonolithic integration [34, 59] have shown promising results with respect to thereceived SNR and signal bandwidth.13Another benefit of integration is providing the capability of channelmultiplexing. This helps to reduce the number of interconnects between the probeand the analog front-end unit in 3-D and 4-D imaging systems. This is particularlyimportant in applications such as intravascular imaging where a compact probewith minimum number of interconnects is favorable.Besides the CMUT design, careful implementation of the analog front-endunit plays an important role in acquiring the best image quality. The fundamentalpart of the electronics unit consists of a number of transceiver circuits, i.e.transmitters and receivers. Additional peripheral circuits may be necessarydepending on the application, configuration of the transceiver, the imagingmechanism and the required voltage levels. This is elaborated in the followingsection.1.5.1 Overview of the Analog Front-End CircuitryAlthough most of the data processing and image reconstruction is carried out inthe digital domain, interaction of the front-end circuit, consisting of thetransmitter and the receiver, as well as the essential switch blocks, enable circuitsand other peripheral circuits with the ultrasound transducers is in analog domain.Figure 1.5 shows the block diagram of the front-end circuit together with theCMUT transducer and the ultrasound machine. The ultrasound machine used inthis project is the SonixRP machine from Ultrasonix Medical Corporation,Richmond, BC, Canada. Using this type of ultrasound machine specifies theexcitation pulses’ amplitude and frequency, the number of input/output channels,the machine’s input impedance and voltage limitations, and the cable impedance.As shown in Figure 1.5, by proper configuration of the switch blocks and the14CMUTT/ RSwitchHV TXVG - TIAT/ RSwitchT/ RSwitchPower Management System ( PMS )Ultrasound MachineEnableAnalog Front - End CircuitFigure 1.5: Block diagram of the transceiver circuit and interfaces to theCMUT transducer and ultrasound machine.Enable signal, excitation pulses from the ultrasound machine are sent to thetransmitter which is a high voltage driver. The high voltage pulses produced at theoutput of the transmitter are applied to the CMUT elements to generate ultrasoundwaves. After propagation of the waves into the medium of interest (e.g., patientbody), the transducer elements receive the reflected echo signals and convert themto current signals sensed at the input of the receiver. These echo signals are thenamplified (using a variable gain transimpedance amplifier (VG-TIA)) and sent tothe ultrasound machine for further processing. The power management system(PMS) provides different levels of low and high DC supply and bias voltages tothe circuitry.Since this configuration shares a single input/output line on both of theinterfaces with the ultrasound machine and the CMUT array, the following designconsiderations are taken into account: Two switch blocks are used at either side of15the receiver to protect the receiver circuitry from the high voltage output pulses ofthe transmitter and the excitation pulses of the ultrasound machine. Also, enableswitches inside the transmitter and receiver blocks of Figure 1.5 (not shownexplicitly), turn on the transmitter or receiver at the appropriate time for thetransmission and reception to minimize the power consumption.This thesis covers all of the analog front-end blocks introduced in this section.1.5.2 High-Voltage Design ConsiderationsAs mentioned earlier in Section 1.4.1, the sensitivity of the CMUT is maximizedwhen it is biased close to its pull-in voltage. This voltage is typically a high DCvoltage in the range of 50 to 300 volts depending on the CMUT structure. Also,the amplitude of the AC pulse which is applied to vibrate the membrane usuallyexceeds the amplitude of the usual supply voltages in CMOS technologies. Forinstance, the AC pulse amplitude may range from 10 to 30 volts [1, 9, 33, 60].The DC bias and the AC pulse are usually provided by the transmitter circuit ofthe analog front-end interface. This analog front-end circuit is also responsible forreceiving and amplifying the current signals coming from the CMUT in the receivemode. Since the analog front-end receiver deals with the low voltage signals, theCMOS circuitry includes both the high and low voltage parts on the same die andrequires especial design considerations.In the usual CMOS technologies, the drain-source voltage of an NMOS orPMOS device is limited to its break down voltage which is not much higher thanthe recommended supply voltage for that technology. The same voltage restrictionapplies to other terminals of these transistors as well. This limitation complicatesthe design of the circuits which are intended for high voltage applications. To16overcome this limitation, some foundries provide more sophisticated transistorswhich are capable of withstanding high voltages on their terminals. Some of thesetransistors can tolerate high voltages just on their drain-source while some othersare modified to tolerate high voltages on their gate-source terminal as well.Teledyne Dalsa [61], Austria Micro Systems (AMS) [62], Atmel, Alcatel andX-Fab are examples of such foundries. Teledyne Dalsa also offers the MEMS postprocessing option which is capable of integrating the mixed low/high voltageanalog front-end circuits with some types of the MEMS devices [61].0.8 µm CMOS high voltage technology of Teledyne Dalsa has developedLateral Diffused MOS (LDMOS) and Extended Drain PMOS (EDPMOS)devices. In order to withstand the high voltage, the drain junction is separatedfrom the gate and the source by a p-diffused layer in a high voltage well. Most ofthe high voltage MOS devices in this technology can tolerate a range of voltagesfrom 100 up to 300 volts on their drain-source while their gate-source voltagecannot exceed 5 volts. Although one type of NMOS and one type of PMOSdevice in this technology can tolerate 12 volts on the gate-source as well as thedrain-source, the source and bulk of each of these two transistors are inherentlyconnected in the layout view, which restricts the flexibility of the design withthese transistors. The low voltage devices of this technology are typical NMOSand PMOS, operating with the supply voltage of 5 volts. We have used thistechnology to implement our first and second generation of analog front-endcircuits that include high-voltage transmitters and low-voltage receivers as well asDC-DC converter circuits.0.35 µm CMOS high voltage technology of Austria Micro Systems alsosupport high voltage MOS devices. This technology can provide up to 120 volts17on the drain-source while the gate-source can tolerate no more than 3.3 volts inmost devices. There are a few number of PMOS and NMOSes in this technologywhich are capable of withstanding 20 volts on the gate-source. The low-voltagedevices of this technology are divided into two categories; they operate with 5volts or 3.3 volts supply voltages. The 5 volts transistors are usually used ininput-output (I/O) stages and the 3.3 volts transistors are usually utilized in thecore circuitry. We have used this technology in the third (and last) generation ofour transceivers.1.6 Research Objectives and ContributionsThe advantages of the emerging CMUT technology have motivated a plethora ofresearch in this area [54, 56, 63–67]. At the University of British Columbia, threeresearch groups collaborate on the implementation of an integrated CMUT-CMOS-based ultrasound imaging system which is the focus of this work.In the context of CMUT design and integration with electronics, flexible andreliable CMUT models that can be simulated in conjunction with the electronicread-out circuits and provide insights on the system-level performance are of greatimportance. Towards this objective, this thesis proposes a bi-directional Verilog-AMS model for CMUT sensors that takes into account the dynamic behavior, non-linearities and harmonic resonances of the CMUT. The proposed Verilog-AMSbehavioral model predicts the pull-in voltage as well as the resonance frequencyof the transducer with a reasonable accuracy. It is also able to show the springsoftening effect.In addition, two different fast-switching, low power, high-voltage transmittersand a low-power variable-gain transimpedance amplifier are designed in Teledyne18Dalsa’s 0.8 µm high-voltage CMOS technology and are co-simulated with theCMUT Verilog-AMS model. Simulation results show a completely functionalfront-end and CMUT transducer system. Co-simulation not only increases thelikelihood of successful implementation, but also provides the opportunity foroptimization of the full system. Test results of the high-voltage transmitters andoptical measurements of the CMUT vibrations while the actuation pulses areprovided by a fabricated high-voltage transmitter are demonstrated and shown tobe reasonably matched with the co-simulation results. This confirms the reliabilityof the co-simulation of the analog front-end and the Verilog-AMS model.To increase the CMUT sensitivity, a method is developed to provide variableDC bias levels to the CMUT at the receive mode. Since the ultrasound waves whichare reflected back from body tissues are attenuated as a function of the distancetraveled, increasing the DC bias of the CMUT depending on the time of arrival ofthe signal increases the receive sensitivity so that the weaker echoes are amplifiedwith a higher gain and are not lost in the noise floor of the pre-amplifier. Anon-chip DC-DC converter is designed and fabricated in Teledyne Dalsa’s 0.8 µmhigh-voltage CMOS technology to provide such different adjustable high voltagelevels to the CMUT electrodes during the receive time.In the scope of the super-resolution ultrasound-imaging using CMUTs, anintegrated circuit is designed and implemented which provides stimulus andperforms the read-out of dual bottom electrode CMUT devices. This circuit isdesigned in Austria Micro Systems 0.35 µm high-voltage CMOS technology.All of the above-mentioned analog front-end circuit blocks have beenfabricated and successfully tested. The test results are well matched with thepost-layout simulation results.191.7 Thesis StructureChapter 2 summarizes different CMUT models and proposes a CMUTVerilog-AMS model that can be co-simulated with the analog front-end electroniccircuitry in simulation environments. The CMUT’s Verilog-AMS model ischaracterized by simulation and is compared to experimental results. Chapter 3discusses the design and fabrication of two types of high voltage transmittingcircuits and a transimpedance receiving amplifier. Simulation and experimentaltest results of these designs and comparison with similar works in literature arealso included in this chapter. In Chapter 4, a high voltage transmitter and atransimpedance receiver circuit are co-simulated with the Verilog-AMS model ofthe CMUT. Simulation results show a completely functional full analog front-endand CMUT transducer system. Optical measurements of the CMUT vibrations inair medium in transmit mode, while the CMUT actuation pulses come from afabricated high-voltage transmitter, are presented in this chapter. Chapter 5proposes the idea of boosting the CMUT receive sensitivity by controlling its DCbias at the receive mode. Design and fabrication of an on-chip DC-DC converterwhich provides multiple DC voltage levels is explained and simulation andexperimental test results are presented. Chapter 6 explores the transceiver designfor super-resolution imaging. The receiver in this design consists of multipleinput-multiple output channels and is capable of receiving the fundamental andharmonic mode vibrations of CMUT and amplifying them to the desired levels.The outputs of the receiver will be used to construct high resolution images of themedium of interest. Simulation and experimental results are also presented forthis design. Furthermore, design, simulation and experimental results of a20high-voltage transmitter is presented in this chapter. A summary of the projectand ideas for future directions of this research are discussed in Chapter 7. Thischapter concludes the thesis.21Chapter 2Verilog-AMS Macro-Modeling ofCMUTs for System LevelSimulations 22.1 IntroductionCapacitive micromachined ultrasonic transducer technology has gained anincreasing interest in the past decade. Compared to their piezoelectriccounterparts, CMUTs can provide smaller feature sizes correlated with wideroperational bandwidth which results in a better axial resolution. One of the mostprominent advantages of CMUT technology is its potential for integration withthe readout electronic circuits in a single package [7, 44].The performance of a CMUT-based ultrasound imaging system depends onthe possibility of designing and co-simulating of both the CMUT transducer and2 A version of this chapter is included in a journal paper submission.22the front-end electronics at the system level. State of the art electronic designautomation (EDA) tools (e.g. Cadencer) already emphasize on this system levelco-simulation direction by providing mixed-kernel simulators like Verilog-AMS.Nevertheless, the usual approach is still to separately design the CMUT array andthe front-end electronics; the latter sub-system is being simulated and optimizedbased on only a constant capacitive load replacing the CMUT transducer or at itsvery best, the linearized equivalent circuit model of the CMUT.CMUT models proposed in literature can be divided into two broadcategories: analytical [42, 68, 69] and numerical [70–72]. Most CMUT equivalentcircuit models are derived based on analytical methods [31, 73]. On the otherhand, finite element modeling (FEM) or finite difference modeling (FDM) usenumerical methods to model the CMUT [70, 74]. While such numerical modelsare capable of incorporating the complex behavior of CMUTs, analytical modelsin the form of equivalent circuits provide better physical insight and aresignificantly faster and simpler [69]. Equivalent circuits can also be readilyco-simulated with the front-end electronic circuits in the simulation environments.This provides the opportunity to simulate and optimize the combinedCMUT-CMOS circuits.The majority of the previous CMUT models take into account only thefundamental resonance frequency of the CMUT membrane. To model higherresonance modes of the CMUT, FEM simulations were proposed [69]. However,as FEM is a computationally expensive approach, using it within multiple designand optimization iterations is time consuming. Previous works ([73] and [75])introduced a comprehensive explanation for a general method to derive equivalentcircuit representations for the purpose of simulating the higher-order resonance23modes. This work follows a similar energy-based procedure for establishing abehavioral Verilog-AMS model for CMUTs covering both fundamental andsecond harmonic (asymmetric) resonance modes. This analytical model takes intoaccount the non-linear electro-mechanical dynamics of the CMUT devices[43, 60]. Advantages of the proposed model compared to FEM, include itssuperior speed of simulation, flexibility in adjusting the design parameters basedon experimental measurements and co-simulation with the analog front-endcircuit.To demonstrate the applicability of the model, simulation and measurementresults are presented. The reasonably good match between the simulation andmeasurement results confirms that the proposed dynamic Verilog-AMS behavioralmodel predicts the behavior of the device, as well as the pull-in voltage and themechanical resonance frequencies. As this is a full nonlinear model, it is also ableto show the spring softening effect3, which is not fully covered in approaches thatuse other analytical models [69].The focus of this chapter is on the CMUT modeling, simulations and inputimpedance measurements of the available prototype devices. Chapter 4 willpresent the CMUT and CMOS co-simulations and optical measurements. Theorganization of the current chapter is as follows: Section 2.2 provides a briefsummary of the CMUT models used in literature, as well as their features andshortcomings. Section 2.3 discusses the development of the proposedVerilog-AMS model for CMUTs. Section 2.4 presents the simulation results ofvarious CMUT structures based on the proposed model. Measurement results of3When an electric field is formed between the parallel plates of a capacitive MEMS device, thespring softening effect reduces the stiffness of the device and therefore, the natural frequency ofvibration.24C0- C0 / n2 CmL m 1 : nElectrical PortAcoustical Port ZMembraneFigure 2.1: The Mason’s two port CMUT equivalent circuit model.the CMUT input impedance are described in Section 2.5. Finally, Section 2.6concludes this chapter.2.2 Different CMUT Modeling Approaches2.2.1 Finite Element ModelingCMUT behavior can be modeled and analyzed using finite element modeling[70, 71]. In FEM, the global object geometry is divided into several smallergeometries called finite elements. Using numerical methods, FEM sets the partialdifferential equations (PDE) of each finite element. Then, it recombines theseelement equations to solve for the global geometry. Although this methodtypically has a high accuracy in predicting the CMUT static and dynamicbehaviors, it suffers from very long computation time [60].2.2.2 Mason’s Equivalent Circuit ModelFigure 2.1 shows the two port Mason’s equivalent circuit model for CMUT [76].In this figure, the left side is the electrical port and the right side represents themechano-acoustical port. C0 is the capacitance of the CMUT element. The25transformer represents the electromechanical conversion between the two sides.The negative capacitance is for considering the spring softening effect because ofthe electromechanical interaction. The inductor Lm and the capacitor Cm arerepresenting the mechanical impedance of the CMUT and the Z impedanceaccounts for the acoustical impedance of the medium.The transformer ratio is shown to be a representative of the CMUT’ssensitivity and is inversely proportional to the square of the gap’s depth. The Zimpedance defines the frequency response of the device. In air, because of the lowacoustical impedance, the quality factor of the circuit is high but liquids like wateroverdamp the resonant behavior of the device. This overdamping behavior resultsin wide bandwidth and so, high temporal resolution which translates into highaxial resolution [29]. The shortcoming of this model is its dependence on alinearized small-signal model which doesn’t take the device dynamic behaviorand non-linearities into account.2.2.3 CMUT as a Single Fixed-Value CapacitorTo simplify the CMUT model and fast analysis of the analog front-end circuitry, asingle fixed-value capacitor has been used in literature [28, 33]. Also, for newCMUT topologies, when the exact model is not derived yet, the analog front-endcircuit designers start by modeling the CMUT as a capacitive load [3]. Thisapproach does not provide any information about the CMUT behavior and issolely used for analog front-end design purposes.In addition to the above modeling methods, some researchers have proposed acombination of FEM and equivalent circuit modeling to increase the accuracy of26Figure 2.2: Two different types of CMUT cells used in this work: asymmetric(top) and symmetric (bottom).analyzing the CMUT behavior [77].2.3 The Proposed Verilog-AMS Model for CMUTsThe foundation of the proposed Verilog-AMS model for CMUTs in differentmodes of vibration is based on the MEMS parallel-plate model discussed by Cretu[78] and elaborated by Frew et al. [43] for the fundamental mode and by You [60]for the second harmonic mode. The frequently used design parameters of thismodel are listed in Table 2.1.In this work, two different types of CMUT devices are modeled inVerilog-AMS and co-simulated with the front-end electronics. These structureswere introduced by You [60] and Cai [41]. Figure 2.2 shows these two CMUTcells. The first type is a one dimensional (1-D) linear array of 71 elements. EachCMUT cell in the element is a circular structure with one top and one bottom27Table 2.1: Frequently used parameters in this work.Parameter Descriptiona Membrane radius (47 µm)N Number of cells per CMUT elementd0 Initial gap distance (0.75 µm)dm Membrane thickness (1.5 µm)ρm Membrane density (2320 kg/m3)Ym Membrane’s Young’s modulus (16×1010 N/m2)σm Membrane Poisson’s ratio (0.22)ε0 Permittivity of vacuum (8.85×10−12 F/m)z Membrane vertical displacement (deflection) (µm)vz Velocity of the membrane displacement (m/s)fz Force at the mechanical port (N)u Electrical port input voltage (V)Q Electric charge (C)S CMUT element area (µm2)m0 Effective mass of the membrane (kg)k Effective spring constant of the membrane (N/m)B Medium’s damping coefficient (Ns/m)Z0 Medium’s characteristic acoustic impedance (kg/m2s)c0 Speed of sound in the medium of interest (m/s)ρ0 Acoustic medium’s density (kg/m3)C(z) Variable CMUT capacitance (pF)λmn Roots of the Bessel function for natural frequenciesRloss Radiation loss of the medium (kg/s)C f l,mn Mass loading of the medium (kg)electrode. In this work, this type of CMUT cell is referred to as a symmetricCMUT. The second type of CMUT is a single circular cell with its bottomelectrode split into two half circles. These half circles can be excitedsymmetrically or asymmetrically with reference to the top electrode. This CMUTstructure is referred to as an asymmetric CMUT. As shown by You [60],asymmetric excitation of this type of CMUT results in the second harmonicfrequency excitation. Proposed applications of this type of CMUT include tiltable28ultrasonic transducers for improved beamforming [79], direction of arrivalestimation [80] and super-resolution imaging [35]. In our experiments, thisCMUT is used to study the accuracy of the analytical modeling of the secondharmonic resonance frequency. In the simulations and experiments of this work,the AC and DC voltage ranges, as well as the frequency range, are chosen basedon the target design specifications of these fabricated CMUTs, as explained byCai [41].From the electrical side, CMUTs are seen as variable capacitors due to thevibration of the membrane [29]. By applying a variable voltage across the CMUTelectrodes, a variable current flows into the electrical port, given by the followingequation[78]:i =dQdt=d(C(z)u)dt=C(z)dudt+udC(z)dt, (2.1)where Q is the electric charge, z is the vertical displacement of an equivalent rigidplate, u is the applied variable voltage and C(z) is the CMUT capacitance, whichvaries with the gap distance. Since:dC(z)dt=dC(z)dzdzdt=dC(z)dzvz, (2.2)in which vz is the velocity of the plate’s mechanical movement, we can write:i =C(z)dudt+udC(z)dzvz. (2.3)As shown in Figure 2.3, this electro-mechanical behavior can be modeled as atwo-port network equivalent circuit (the shaded area) where e1 and e2 are the two29R eqC( z )u+-i(in)Vz+- fz** L mCmi(out)CMUT two - port networkm 1m 2e 1e 2  -Figure 2.3: The typical ideal MEMS parallel-plate model: the left siderepresents the electrical port and the right side models the mechanicalport of the CMUT.terminals of the electrical port and m1 and m2 are the two terminals of themechanical port. From Equation 2.3, it can be seen that the electrical port ismodeled by a non-linear capacitor C(z) and a non-linear resistor Req(z,vz):Req(z,vz) =1dC(z)dz vz. (2.4)At the mechanical port (Figure 2.3) the corresponding across-through pair isgiven by the velocity (vz) and the force ( fz) respectively. The controlledgeneralized current source in the figure ( fz) reflects the electro-mechanicalcoupling in the mechanical domain. As shown by Cretu [78], its expression isgiven by:fz =−12u2dC(z)dz. (2.5)30Also, z is calculated as:z =∫ t0vz dt =∫ t0vout(t)dt, (2.6)where vout(t) is the generalized voltage across the output port which is anequivalent representation of the velocity.2.3.1 Fundamental Mode ConsiderationsAs shown by You [60],C(z) for the fundamental frequency of vibration of a circularmembrane (shown as C0(z)) is approximated by:C0(z)≈ε0Sd0− z, (2.7)where ε0 is the permittivity of free space and d0 is the nominal gap distance inthe absence of any deflection. S is the total area of the membranes of the CMUTcells defining a single CMUT element; therefore it is given by pia2N, where a isthe radius of a circular membrane cell and N is the number of cells per CMUTelement.The diport ((e1,e2),(m1,m2)) reflects only the electro-mechanical couplingpresent in a CMUT element. Therefore, the model needs to be completed byadding the inertial (effective mass) and elastic (effective spring constant) behaviorof the membrane to the mechanical port. As it is also shown by You [60], theeffective mass, m0 in Figure 2.3, is represented by a generalized capacitor Cm andis equal to:m0 =Cm = ρmdmS, (2.8)31where ρm is the density of the membrane material and dm is the membranethickness. Also in this figure, the generalized inductance, Lm, is the inverse of theeffective spring constant k, where k is [81]:k =1Lm=λmn4piYmdm312a2(1−σm2)N. (2.9)In this equation, λmn is a parameter related to the natural frequency of a clampedcircular plate in its (m,n) mode, with m and n being the number of concentriccircles and diametral nodal lines, respectively. For example, λ00 and λ01correspond to the fundamental mode and second-harmonic (or asymmetric) modeand are equal to 3.196 and 4.611 respectively [82]. Ym is the Young’s modulus andσm is the Poisson’s ratio of the membrane material.The two-port system of Figure 2.3 can be modeled in Verilog-AMS using thefollowing constitutive equations:i(in) =C(z)dudt+udC(z)dzvz, (2.10)i(out) =−12u2dC(z)dz+m0dvzdt+ k∫ t0vz dt, (2.11)in which, u = vin, vz = vout , and z and C(z) are given in Equations 2.6 and 2.7.2.3.2 Harmonic Mode ConsiderationsYou [60] and Cai [41] have implemented individual CMUT cells to study theirbehavior and explore the potential applications of using second harmonic mode ofsuch CMUTs. These CMUT cells have two half-circle bottom electrodes and one32(a) Fundamental mode vibrationfor the asymmetric CMUT.(b) Second harmonic modevibration for the asymmetric CMUT.Figure 2.4: Asymmetric CMUT fundamental and second harmonic modevibrations [60].full-circle top electrode. In such type of implementation, if the two half-circleelectrodes are connected to same signals, then the second harmonic mode iscanceled and the CMUT can be modeled as a typical piston-like CMUT, asdiscussed in Section 2.3.1. On the other hand, the second-harmonic mode appearsif different (asymmetric) AC + DC signals are applied to the two bottomelectrodes or by proper mechanical excitation. Figure 2.4 illustrates the modeshapes of the membrane for the fundamental and second-harmonic vibrations ofthis type of CMUT cell. In the second-harmonic mode, the membrane movementcan be approximated as the vibration of the two half membranes with equalamplitudes but opposite displacements (Figure 2.4b). It can be shown (e.g., [60])that the behavior of the overall CMUT cell in its second-harmonic mode can beinferred from that of its half membrane. Therefore, Cz in this case is given byC1(z):C1(z)≈ε0S/2d0− z. (2.12)332.3.3 Modeling the Acoustic Interaction with the MediumThe interaction between the vibrating membrane and the medium could berepresented by a damping effect (representing the equivalent radiation losses) anda mass loading. The radiation loss is represented by a generalized resistance andthe mass loading by a generalized capacitance, both added in parallel to themechanical port as an equivalent Zmedium (as shown in Figure 2.5). Themagnitudes of these terms depend on whether the CMUT cell is vibrating as anindividual cell or synchronized with the neighboring cells. For instance, theradiation loss term will be more significant for a CMUT cell in an array.The radiation loss, Rloss, is equal to the inverse of the damping coefficient B:Rloss =1B=1Z0S, (2.13)where Z0 is equal to ρ0×c0. ρ0 is the acoustic medium’s density and c0 is the speedof sound in the medium of interest [43]. The value of the mass loading representedby a capacitive load, C f l,mn, is defined as [60]:C f l,mn = aSρ0/λmn. (2.14)Zmedium values depend on the vibration mode as well. The symmetric modetends to radiate more than the asymmetric harmonic mode. In the asymmetricmode, the capacitive component of Zmedium will be the dominant term in a fluidmedium, leading to an under-damped vibration behavior.34V AC + V DC+-Z mediume 1e 2m 1m 2CMUT ModelFigure 2.5: CMUT simulation test setup.2.4 Simulation Results of the CMUT Verilog-AMS ModelThe CMUT Verilog-AMS model is implemented in Cadencer using the modelparameters of Table 2.1. The macro-model is used to characterize the CMUTelectrical input impedance as shown in the test setup of Figure 2.5. Bias andexcitation voltages are applied to the electrical ports, and the equivalentmechanical impedance of the medium, Zmedium, is connected to the mechanicalport. Based on the CMUT specifications in [41], the applied AC voltage rangesbetween 5 and 30 volts, with the DC bias up to 90 volts. In order to comparesimulation with the more constraining measurement setup, a 0.5 volt amplitudesinusoidal source with a DC bias of 20 volts is applied to the CMUT.Figure 2.6 illustrates the input impedance of the symmetric CMUT in air. Asit is shown in this figure, there is a peak at 2.73 MHz in both real and imaginarycomponents. Also, Figure 2.7 shows the input impedance of the symmetricCMUT during immersion in oil. As expected, the input impedance of thesymmetric CMUT has an over-damped behavior. Figure 2.8 shows the inputimpedance of the asymmetric CMUT when it is operating at its fundamental modein air. As can be seen from the figure, there is a peak at 2.69 MHz in both real andimaginary components. Also, Figure 2.9 shows the input impedance during351 1.5 2 2.5 3 3.5 4x 106024681012Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of the symmetric CMUT input impedance in air  −3−2.5−2−1.5−1−0.50x 104Impedance Imaginary (Ω)RealImaginaryFigure 2.6: Simulation results of the symmetric CMUT’s input impedance inair. The peak is at 2.73 MHz.1 2 3 4 5 6 7x 10600.511.5Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of the symmetric CMUT input impedance in oil  −2.5−1.665−0.830x 105Impedance Imaginary(Ω)ImaginaryRealFigure 2.7: Simulation results of the symmetric CMUT’s input impedanceduring immersion in oil.362.6 2.65 2.7 2.75 2.8x 1060100200300400500600700Frequency (Hz)Impedance Real (kΩ)  −3−2−101234x 107Impedance Imaginary(Ω)Real and imaginary parts of the fundamental mode asymmetric CMUT input impedance in airImaginaryRealFigure 2.8: Simulation results of the fundamental frequency behavior of theasymmetric CMUT’s input impedance in air. The peak is at 2.69 MHz.1 2 3 4 5 6 7x 10608162432Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of the fundamental mode asymmetric CMUT input impedance in oil  −10−7.7−5.4−3.1−0.80x 106Impedance Imaginary(Ω)ImaginaryRealFigure 2.9: Simulation results of the fundamental frequency behavior of theasymmetric CMUT’s input impedance during immersion in oil.374 4.5 5 5.5 6 6.5x 1060100200300400500Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of the asymmetric CMUT harmonic mode input impedance in air  −10−9−8−7−6−5x 105Impedance Imaginary (Ω)RealImaginaryFigure 2.10: Simulation results of the harmonic frequency behavior of theasymmetric CMUT’s input impedance in air. The peak is at 5.4 MHz.immersion in oil. The flat frequency response represents the over-dampedbehavior of the CMUT during immersion in oil. Figure 2.10 shows the inputimpedance of the asymmetric CMUT at its second-harmonic (asymmetric) mode,when operating in air. There is a peak at 5.4 MHz. Also, Figure 2.11 shows theinput impedance during immersion in oil. As shown, there is a peak in thefrequency response at 4.07 MHz. This result demonstrates the capability of themodel in capturing the under-damped behavior and resonance frequency shift ofthe second harmonic mode in immersion.2.5 CMUT Measurement ResultsThe CMUT used in the experimental tests of this work is designed inPolyMUMPs technology, a three-layer micro-machining process offered by383 3.5 4 4.5 5 5.5 6x 106−0.0500.050.1Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of theasymmetric CMUT harmonic mode input impedance in oil  −50510x 106Impedance Imaginary (Ω)RealImaginaryFigure 2.11: Simulation results of the harmonic frequency behavior of theasymmetric CMUT’s input impedance during immersion in oil. Theunder-damped phenomenon results in a peak at 4.07 MHz.2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3x 1060.540.560.580.60.620.64Frequency (Hz)Impedance Real (kΩ)  −1450−1380−1310−1240−1170−1100Impedance Imaginary(Ω)Real and imaginary parts of the symmetric CMUT input impedance in airImaginaryRealFigure 2.12: Experimental results of the symmetric CMUT input impedancein air medium. The peak is at 2.57 MHz.392 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3x 1061.051.11.151.21.25Frequency (Hz)Impedance Real (kΩ)  −1,350−1288−1225−1163−1100Impedance Imaginary(Ω)Real and imaginary parts of the symmetric CMUT input impedance in oilImaginaryRealFigure 2.13: Experimental results of the symmetric CMUT input impedancein olive oil medium.MEMSCAP Inc. (North Carolina, USA). To characterize the CMUT behavior, theinput impedance is measured for both types of CMUT (symmetric andasymmetric) using an Agilent 4294A precision impedance analyzer. Experimentsare carried out in air and oil media. To implement an oil medium, a droplet ofolive oil is dropped on the CMUT surface, which spreads and covers the CMUTdevice under test.Figures 2.12 and 2.13 show the real and imaginary parts of the input impedanceof the CMUT in air and olive oil media for a symmetric CMUT device. As thesefigures show, the input impedance has a resonance peak at 2.57 MHz for 20 V ofDC bias in air and no resonance peak in oil, confirming the over-damping effectof the oil medium. The AC voltage applied to the CMUT chip in this experimentis 0.5 V. The slope of the real part of the impedance at lower frequencies is dueto the parasitic series resistance and leakage current [60]. Experimental results402 2.5 3 3.5 4 4.5 5 5.5x 10600.050.10.150.20.250.30.35Frequency (Hz)Impedance Real (kΩ)  −9000−8000−7000−6000−5000−4000−3000−2000Impedance Imaginary(Ω)Real and imaginary parts of the asymmetric CMUT input impedance in airImaginaryRealFigure 2.14: Experimental results of the asymmetric CMUT input impedancein air medium. The fundamental mode peak is at 2.5 MHz, andharmonic mode peak is at 5.1 MHz.show that the Verilog-AMS model’s prediction of the resonance frequency is inclose agreement with the real data (less than 4% error between measurement andsimulation results).Figure 2.14 shows the real and imaginary parts of the input impedance of theasymmetric CMUT in the air medium. The real part of the input impedance has aresonance peak at 2.5 MHz of fundamental frequency and 5.1 MHz of harmonicfrequency for 20 V of DC bias and 0.5 V of AC. Considerably higher AC and DCvoltages would be needed to elevate these peaks, but were not observed because ofthe impedance analyzer’s voltage limits.Figure 2.15 shows that the asymmetric mode maintains a resonance peak evenin oil medium while the symmetric mode becomes over-damped. Nevertheless, theextra damping added by the oil, shifts the resonance frequency of the asymmetricmode by about 900 kHz as compared to the air medium response.411.5 2 2.5 3 3.5 4 4.5 5 5.5 6x 106−0.0200.020.040.060.080.1Frequency (Hz)Impedance Real (kΩ)Real and imaginary parts of the asymmetric CMUT input impedance in oil  −9000−8000−7000−6000−5000−4000−3000Impedance Imaginary (Ω)RealImaginaryFigure 2.15: Experimental results of the asymmetric CMUT input impedancein olive oil medium. The fundamental mode peak is over-damped andthe harmonic mode peak is shifted to 4.1 MHz.2.6 ConclusionThis chapter summarized the common CMUT modeling methods and proposeda Verilog-AMS macro-model for CMUT transducers, built from first principlesbased on energy equivalence. The model is dynamic, fully non-linear and it hasbeen used to predict both the frequency shift and the pull-in voltage of the CMUT.The model has been validated experimentally for both symmetric and asymmetricvibration modes using input impedance measurements. An advantage of the modelis providing a direct physical interpretation of the model parameters, allowing theiradjustments based on measurement results. Furthermore, other parasitics presenton the chip, like series interconnect resistances and coupling capacitors, that causedifferences between simulation and measurements, can be taken into account byincrementally adding their effect to the model.42Results of the co-simulation of the proposed CMUT Verilog-AMS model withanalog front-end circuits, as well as the related measurement results will bepresented in Chapter 4.43Chapter 3High-Voltage Analog Front-EndCircuits for Integration withCMUT Arrays 43.1 IntroductionIn this chapter design of a front-end electronic circuit, consisting of high-voltagetransmitters and low-power variable-gain receivers is presented. The circuit, whichis intended to be integrated with a CMUT array, is designed in Teledyne Dalsa’s0.8 µm high-voltage (up to 300 V) CMOS process.Different structures for the front-end electronics of the CMUT sensors havebeen reported in recent years [28, 33, 83, 84]. In the design stage of many of suchcircuits, linear small-signal CMUT models are used [28, 33, 84]. In this work, we4 Parts of Section 3.2 to Section 3.6 was published in the Proceedings of the IEEE BioCAS2010[1]. Parts of the contents of Section 3.2, Section 3.3 and Section 3.5 are used in journal papersubmissions.44use the behavioral Verilog-AMS model for CMUT sensors introduced in Chapter 2.This results in a more realistic evaluation of the front-end circuit during the designstage.Two types of high voltage transmitters are designed and fabricated and theirtest results are compared. The desired transmitting driver in this design is capableof generating narrow high voltage pulses, with shorter rise and fall times ascompared to other works [85, 86]. This fast transition time will mitigate thecommon problem of having a large short circuit current and thus results inreduced power consumption of the high voltage driver.In ultrasound systems, the amplitude of the received echo signal may changedramatically. This raises the need for a variable gain amplifier. A transimpedancevariable gain amplifier stage is presented in this work which adjusts the dynamicrange of the signal to comply with the requirements of the next processing stages.3.2 Overview of the Analog Front-End CircuitryFigure 1.5 illustrated the block diagram of the analog front-end circuit. Thisfigure is brought here again as Figure 3.1 for more convenient referencing. As thisfigure shows, the analog front-end transceiver circuit has interfaces to the CMUTtransducer and the ultrasound machine (SonixRP machine from UltrasonixMedical Corp., Richmond, BC, Canada). The transceiver circuit consists of thehigh voltage transmitters (HV TX), variable-gain transimpedance amplifierreceivers (VG-TIA), and transmit/receive switch blocks (T/R switch) and enablecircuits. A power management system (PMS) provides the required DC bias andsource voltages to the circuits. In transmit mode, the ultrasound machine provideslow voltage excitation pulses for the transmitter. The transmitter produces high45CMUTT/ RSwitchHV TXVG - TIAT/ RSwitchT/ RSwitchPower Management System ( PMS )Ultrasound MachineEnableAnalog Front - End CircuitFigure 3.1: Block diagram of the transceiver circuit and interfaces to theCMUT transducer and ultrasound machine.voltage pulses to excite the CMUT elements, resulting in generation of ultrasoundwaves. These waves propagate into the medium of interest (e.g., patient body) andreflect back from tissue boundaries. In receive mode, the reflected echoes arereceived and converted to current signals by the CMUT elements. These currentsignals are then sensed and amplified by the receiver circuit and sent to theultrasound machine. The ultrasound machine processes the received signals andconverts them into images. The SonixRP machine defines the interfacespecifications for the transceiver, such as the excitation pulses’ amplitude andfrequency, the machine’s input impedance and voltage limitations and the numberof input/output channels. These specifications are shown in Table 3.1.As it is shown in Figure 3.1, a single input/output interface is shared on bothof the ultrasound machine and the CMUT array sides of the transceiver circuit.Hence, protection switches are used to assist reliable high voltage transmission46Table 3.1: Ultrasound machine specificationsParameter ValueCable 50 Ω CoaxialInput impedance 50 ΩMin/Max frequency 2 to 10 MHzVoltage amplitude, output 5 VVoltage limitation, input ± 1 VNumber of channels 128to the CMUT input without damaging the low-voltage receiving circuitry. Also,enable switches inside the transceiver circuit turn on the transmitter or receiver atthe desired time to minimize the power consumption.3.3 High-Voltage Transmitter ArchitecturesTo design the transmitter which is a high-voltage driver in this work, differentconfigurations have been studied. High-voltage drivers introduced in literature canbe categorized into two groups: The first group uses a high-voltage level shifterand a high-voltage buffer to generate the high-voltage pulses at the output[64, 87, 88]. In this structure, the level shifter is used to limit the buffer’s PMOSdevice Vgs voltage swing. This driver suffers from high power consumptionbecause of the static current flow during the high state of the input signal. Thesecond category uses a cross-coupled transistor pair to increase the switchingspeed and help limiting the Vgs of the next stage buffer within the acceptablerange. Although drivers of this type do not dissipate any static power, they occupymore area than the first group [9, 89–93].In some of the circuit configurations of both categories, thick gate oxidetransistors, that can sustain high voltage on their gate-source as well as the47LV PMOS LV NMOS HV PMOS HV NMOSFigure 3.2: Low voltage (LV) and high voltage (HV) PMOS and NMOStransistor symbols used in the schematics of this thesis.drain-source are used [90, 91]. These configurations have more design flexibility;however, this is not applicable to the high-voltage transistors in the TeledyneDalsa 0.8 µm high-voltage technology that we have used.In this work, both types of high-voltage drivers have been designed andfabricated. The design details of these drivers come in the following sections.Figure 3.2 discerns the high voltage and low voltage transistor symbols used inthe schematics of this thesis.3.3.1 HV Transmitter Circuit: The Level-Shifted DesignThe transmitter in this design is capable of producing high voltage uni-polar pulsesat its output to drive the CMUT sensor. Figure 3.3 shows the driver circuit andits enabling switch. The driver consists of a high-voltage level shifter and a buffer.The level shifter, enclosed by dash-dot line in Figure 3.3, is needed to limit the gate-source voltage of the high-voltage transistor P2 to its maximum sustainable voltageof 5 V (dictated by the high voltage CMOS technology used). When a triggersignal from the ultrasound machine excites the driver, the high-voltage transistorN1 turns on, pulling down the voltage at the gate of P2 through the diode-connectedtransistor P1. Then P2 connects the high voltage source to the output. Therefore, by48EnableP 1 P 2N 1 N 2HV V dd HV V ddLevelShifterTo CMUTInput PulseHigh Voltage Transmitter CircuitSwitchBlockFigure 3.3: The level-shifted high-voltage transmitter schematic with itsenable switch.changing the polarity of the trigger signal, either P2 or N2 turns on. This connectsthe output to the high-voltage source or to the ground and generates high-voltageuni-polar pulses at the output of the driver.The circuit is designed based on the specifications and requirements of theavailable ultrasound machine and the CMUT transducer. The pulse-width of theinput signal from the ultrasound machine can be increased by 12.5ns steps. Wehave targeted pulse-widths of 50ns and more. The challenge of the design waskeeping the area as small as possible and decreasing the power consumption whilemaintaining the speed. Considering the design specifications of the fabricatedCMUT by Cai [41], the peak-to-peak amplitude of the AC pulses at the output ofthe transmitter ranges between 5 to 30 volts and the CMUT DC bias may rise upto 90 volts. Also, the operational frequency range of the CMUT is less than4910 MHz. If the output pulse is required to have both DC and AC components, theground potential at the source of N2 in Figure 3.3 should be replaced with therequired DC level and the trigger pulse amplitude should be adjusted accordingly.One challenge in this design is that the source-bulk voltage for most of the highvoltage transistors in Teledyne Dalsa 0.8 µm technology is limited to 5 V. Only afew of the available transistors are capable of having a floating source up to 300volts. However, the layout area for these transistors is very large(500 µm × 250 µm each) and therefore they cannot be used in designs with chiparea limitation. Because of this, we decided to bias the CMUT transducerseparately.Different techniques have been used to decrease the power consumption of thelevel-shifted transmitter, including lowering the voltage level of the actuationpulses from the ultrasound machine by using level-down shifters (shown byinverter logic symbols in Figure 3.3) and adjusting the high voltage transistorsizes while maintaining a reasonable output switching speed.3.3.2 HV Transmitter Circuit: The Cross-Coupled DesignThe cross-coupled transmitter circuit is shown in Figure 3.4. To focus on thetransmitter schematic, the enable and protection switches are not shown in thisfigure. In this circuit, when the input is low, N2 turns on and pulls down the Out2node to the ground level. At this time N4 turns on and pulls the Mid2 nodetowards the ground level. Then the Out1 node follows Mid2 to the point that theOut1 reaches the “Bias-high” voltage value plus a threshold voltage and then itstops, as P4 reaches the turn-off edge. This turns on the P5 transistor which risesthe output voltage to the high-voltage supply value. Also, P1 turns on, which50HV V ddOut 1N 1 N 2P 1 P 2N 3 N 4P 3 P 4N 5P 5To CMUTInputBias _ lowBias _ highOut 2Mid2Mid1HV V ddCrossCouplerFigure 3.4: High-voltage transmitter architecture, the cross-coupled design.results in turning off P2 and fixing the Out1 node on about the “Bias-high”voltage plus a threshold. Moreover, by turning off P2, no static current will flowinto the circuit. At the next half cycle, when the input is high, N1 turns on,resulting in turning on P2, so Out1 becomes high voltage supply value and P5turns off. At the same time, N5 turns on and the transmitter output is pulled downto the ground level.3.4 Variable-Gain Receiving Amplifier CircuitDesign of the receiving amplifier requires careful considerations. Since theCMUT output signal is in the form of current, a transimpedance amplifier is51needed. The output current of the CMUT element can have a wide dynamic range,typically from several nano-amperes to a few micro-amperes. Therefore, theamplifier should have a variable gain to not only amplify the weak signalsproperly, but also avoid saturation of its output for the larger input signals. Inaddition, the amplifier should have an appropriate bandwidth to process thesignal, while keeping the in-band noise to its minimum value. It should also havea low power consumption, in particular when the design is intended for portablemedical equipments.To design the variable-gain transimpedance amplifier, different designconfigurations are studied. Transimpedance amplifiers can be categorized into twotypes. In the first type, a resistive or capacitive feedback transforms the inputcurrent signal to the output voltage signal [9, 94]. In the second type, the currentsignal is converted into a voltage signal through a regulated cascode circuit[84, 95, 96].Different variable-gain amplifiers (VGA) are also introduced in literature [97–100]. Some VGA configurations use a resistor bank and some switches to changethe feedback loop resistance value. Others, change the gain by changing the biasvoltages which control the current sources of the circuit or the resistance of thetransistors which are biased in the triode region.In our design, it is crucial to have the input resistance of the receiver as small aspossible, since the CMUT element is also connected to the transmit driver circuitand possible current division between the receive and transmit path may causesignal power loss. Also, since the current generated by the CMUT is in the rangeof nano to micro-amperes, a high gain amplifier is required to generate detectablevoltages in the range of hundreds of milli-volts for the ultrasound machine.52Vi n + Vi n -RX _ EN M4M5M7 M6M8M9R 3M10M11M12M13RX _ ENTo Ult rasound MachineVou ti i np utF rom CMUTR 1R 2LV VddVcn trlVi n +M1M2M3LV Vdd LV Vdd LV VddVariable-gain                      Trans-impedance Amplifier Op-Amp BuffersFigure 3.5: The receiving circuit consisting of the input variable-gaintransimpedance amplifier, op-amp and output buffers.Figure 3.5 illustrates the amplifier circuit designed to fulfill the aboverequirements. This circuit consists of three stages: The input stage, the differentialamplifier circuit and the buffers. The input stage is based on the regulated cascodecircuit first introduced by Sackinger et al. in [95] and developed by Park et al. in[96]. This stage provides high gain and very low input impedance as well as anenhanced input transconductance which results in reducing the high-frequencynoise.In Figure 3.5, the current signal from the transducer is applied to the source ofM1 and the corresponding amplified voltage is generated at the drain of M1. Thevoltage at the drain of M1 (output of the input stage) is connected to the positiveinput of the differential amplifier of Figure 3.5. A dummy input stage is also neededto be connected to the negative input of the differential amplifier circuit. Thisdummy circuit is the same as the input stage, but with no input signal. TransistorM3 in Figure 3.5 is kept in the triode region by controlling its gate voltage andacts as a variable resistor. Assuming R2 is large enough so that the current signal53through it can be neglected, the input impedance of the circuit is calculated as:Rin ≈1gm1(1+gm2(R1‖ro2))≈1gm1A f b, (3.1)where A f b is defined as the (M2, R1) feedback circuit gain. The transimpedancegain of this circuit is also derived as:Vin+Iinput=R2Rongm1(A f b +1)1+gm1R2(A f b +1)≈ Ron, (3.2)where Ron is the on-resistance of transistor M3 in the triode region. Therefore,by applying a proper voltage to the gate of M3, the resistance and the gain of thecircuit can be adjusted. In addition to the gain control functionality, the controlsignal at the gate of M3 can switch off this transistor during the transmit interval toturn off the overall input stage circuit and minimize the power consumption.The differential amplifier stage (Figure 3.5) provides the rest of the requiredgain. Furthermore, it limits the bandwidth of the receiver in order to filter out thehigh-frequency noise of the circuit. In the transmit mode, the enable switch at thedrain of M9 is turned off, and therefore, the tail current sources of the differentialamplifier and the buffer stages are turned off to minimize the power consumption.In addition to the gain stages, there are two output buffer stages used in thiscircuit. We used source follower buffers because of their simple design and widebandwidth. The first buffer is to isolate the differential amplifier from the secondbuffer in order to keep the bandwidth at the required value. The second bufferprovides the required current to drive the load which is a 50 Ω coaxial cable witha maximum of 250 mV peak-to-peak 2.7 MHz signal. The overall bandwidth of54N 2T1T2Switch blockSSFigure 3.6: Schematic of the switch block used in the level-shifted transmitterto isolate the gate of transistor N2 in Figure 3.3.the circuit is designed to be around 10 MHz.3.5 Protection and Enable SwitchesProtecting the low-voltage receiving circuitry from the high-voltage pulses of thetransmitter is a common concern in most of the ultrasound systems that use asingle transmit/receive transducer array. Also, the cross-talk effects of thetransmitter and receiver are considerable in CMUT analog front-end circuits.Except for the systems that divide the transmitter and receiver array elements intotwo distinct parts [28, 101], other configurations deal with designing theprotection and enable/disable circuits to overcome the aforementioned problems.There has been noticeable efforts to remove or reduce the effects of the noise andcross-talk by using diode bridges, transformers, and high-voltage switches asprotection circuits [102–104].In this work, low voltage and high voltage transistors are used to act asswitches where appropriate. At the transmitter input, a low voltage transistor isused as shown in Figure 3.3 to isolate the transmitter input from the amplified55signals at the output of the receiver. Otherwise, as simulation has shown, thosesignals, although having lower-than-threshold voltage, might be able to excite thetransmitter circuit. This will result in unwanted pulses at the output of thetransmitter during the receive time. In addition to this problem, the inherent gatecapacitor of transistor N2 in Figure 3.3 has to be totally discharged whentransistor P2 is on and also when the transmitter circuit is off. This is done by theswitch block at the gate of N2. This switch block which is shown in detail inFigure 3.6, uses two pass transistors to strongly isolate and discharge the gate ofN2. When S is on and S¯ is off, the desired signal passes through T1 and reachesN2. When S is off and S¯ is on, N2 is isolated and the gate’s capacitor is dischargedto the ground. The post-layout simulation results of the transmitter circuit alsoverifies the importance of this switch block. If this switch block doesn’t exist, thegate capacitance will not discharge, resulting in malfunctioning of the transmittercircuit. Some other low voltage pass transistors are also used in the transmittercircuit to isolate the circuit from the low voltage supply at the receive time inorder to decrease the power consumption.At the receiver’s input, a high voltage transistor is protecting the receiver circuitfrom high voltage pulses of the transmitter. At the transmit time, the protectionswitch is off, isolating the receiver from the transmitter. At the receive time, theswitch is on, passing the echo signals to the receiver. There is also another passtransistor at the receiver’s output which protects it from the ultrasound machinesignals at the transmit time.56Figure 3.7: The layout for both transmitters, Left: level-shifted transmitterand Right: cross-coupled transmitter.3.6 Post-Layout Simulation Results of the AnalogFront-End SystemThe front-end transceiver circuit is designed and laid-out in Teledyne Dalsa 0.8 µmhigh-voltage technology. Figure 3.7 shows the layouts for both of the designedtransmitters. As the layouts show, the level-shifted transmitter (left) has occupiedapproximately 500 µm × 250µm = 125000 µm2 of the chip area while the cross-coupled transmitter (right) is 385 µm × 350 µm = 134750 µm2. Assuming a128 channel CMUT array and therefore, 128 transceiver modules, the level-shifteddesign occupies 16mm2 area while the cross-coupled design takes 17.248mm2 ofarea.Figure 3.8a shows the layout of the complete 16 channel transceiver circuit inwhich, each transceiver block includes the proposed level-shifted transmitter andvariable-gain receiver circuits. Also, Figure 3.8b shows the printed circuit board(PCB) layout which is designed to test this transceiver chip.57(a)PA02PA01COPA01PA02PA01PA02PA02PA01 PA01PA02PA01PA02PA01PA02PA02PA01PA02PA01PA02PA01PA02PA01PA01PA02PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01 PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01 PA08PA07PA06PA05PA04PA03PA02PA01PA08PA07PA06PA05PA04PA03PA02PA01PA001uf02PA001uf01CO001ufPA204pf02 PA204pf01CO204pfPA407uf02PA407uf01 CO407ufPAICKBCS9P065PAICKBCS9P066PAICKBCS9P067PAICKBCS9P068PAICKBCS9P069PAICKBCS9P070PAICKBCS9P071PAICKBCS9P072PAICKBCS9P073PAICKBCS9P074PAICKBCS9P075PAICKBCS9P076PAICKBCS9P077PAICKBCS9P078PAICKBCS9P079PAICKBCS9P080PAICKBCS9P01PAICKBCS9P02PAICKBCS9P03PAICKBCS9P04PAICKBCS9P05PAICKBCS9P06PAICKBCS9P07PAICKBCS9P08PAICKBCS9P09PAICKBCS9P010PAICKBCS9P011PAICKBCS9P012PAICKBCS9P013PAICKBCS9P014PAICKBCS9P015PAICKBCS9P016PAICKBCS9P017PAICKBCS9P018PAICKBCS9P019PAICKBCS9P020PAICKBCS9P021PAICKBCS9P022PAICKBCS9P023PAICKBCS9P024 PAICKBCS9P025 PAICKBCS9P026 PAICKBCS9P027 PAICKBCS9P028 PAICKBCS9P029 PAICKBCS9P030 PAICKBCS9P031PAICKBCS9P032 PAICKBCS9P033 PAICKBCS9P034 PAICKBCS9P035 PAICKBCS9P036 PAICKBCS9P037 PAICKBCS9P038 PAICKBCS9P039 PAICKBCS9P040 PAICKBCS9P041PAICKBCS9P042PAICKBCS9P043PAICKBCS9P044PAICKBCS9P045PAICKBCS9P046PAICKBCS9P047PAICKBCS9P048PAICKBCS9P049PAICKBCS9P050PAICKBCS9P051PAICKBCS9P052PAICKBCS9P053PAICKBCS9P054PAICKBCS9P055PAICKBCS9P056PAICKBCS9P057PAICKBCS9P058PAICKBCS9P059PAICKBCS9P060PAICKBCS9P061PAICKBCS9P062PAICKBCS9P063PAICKBCS9P064COICKBCS9PPALED02PALED01COLEDPALED0Res01 PALED0Res02COLED0ResPARes02M02PARes02M01 CORes02M(b)Figure 3.8: (a) The 16 channel transceiver layout. (b) Layout of the printedcircuit board designed to test the 16 channel transceiver chip.However, due to the technical flaws of the design kit, some DRC, LVS andEXT errors of this chip were not detected by the software package and althoughpost-layout simulations were showing perfect results, the chip did not work inexperimental setup. In our tests, we realized that the VDD and VSS of the circuitare short circuited, with 6 Ω between the package pins. We also tested theimpedance between the corresponding VDD and VSS chip pads using a probestation and the 6 Ω was shown on the multimeter. The next stage to debug theissue was using a micro-thermal imaging equipment to detect hot spots on thechip resulting from short circuits. This hot-spot detector showed that whole chipsubstrate gets hot at once when the VDD is connected. This revealed the fact thatsubstrate connections need to be revised. After several consultations with thefabrication foundry and reporting the issues, we realized that more carefulconsideration were needed in designing the layout of the high-voltage transistorsin order to avoid short circuiting the VDD and VSS through substrate.Nevertheless, the layout mistake was not detectable by the design kit. After58(a)PA02PA01COPA01PA01PA01PA01PA03PA06PA07PA08PA09PA010PA011PA012PA024PA023PA022PA021PA019PA018PA017PA016PA015PA014PA013PA05PA02PA020PA04PA02PA01PA02PA01PA02PA01PA01PA01PA01PA02PA01PA02 PA01PA02PA01 PA02PA01 PA02PA01PA02PA01 PA02PA01PA02PA01PA02PA01PA02PA01PA02PA01PA02PA01PA02PA01 PA02PA01PA001uf02PA001uf01CO001ufPA204pf02PA204pf0 CO204pfPA407uf01PA407uf02 CO407ufPAAC01PAAC02COACPACMUT Chip0A1PACMUT Chip0D1PACMUT Chip0E1PACMUT Chip0F1PACMUT Chip0G1PACMUT Chip0H1PACMUT Chip0J1PACMUT Chip0K1PACMUT Chip0L1PACMUT Chip0M1PACMUT Chip0N1PACMUT Chip0P1PACMUT Chip0Q1PACMUT Chip0R1PACMUT Chip0S1PACMUT Chip0A2PACMUT Chip0B2PACMUT Chip0C2PACMUT Chip0D2PACMUT Chip0E2PACMUT Chip0F2PACMUT Chip0G2PACMUT Chip0H2PACMUT Chip0K2PACMUT Chip0M2PACMUT Chip0N2PACMUT Chip0P2PACMUT Chip0Q2PACMUT Chip0R2PACMUT Chip0S2PACMUT Chip0A3PACMUT Chip0B3PACMUT Chip0C3PACMUT Chip0D3PACMUT Chip0E3PACMUT Chip0F3PACMUT Chip0G3PACMUT Chip0H3PACMUT Chip0J3PACMUT Chip0L3PACMUT Chip0M3PACMUT Chip0N3PACMUT Chip0P3PACMUT Chip0Q3PACMUT Chip0R3PACMUT Chip0S3PACMUT Chip0A4PACMUT Chip0B4PACMUT Chip0C4PACMUT Chip0D4PACMUT Chip0E4PACMUT Chip0F4PACMUT Chip0G4PACMUT Chip0H4PACMUT Chip0J4PACMUT Chip0K4PACMUT Chip0M4PACMUT Chip0N4PACMUT Chip0P4PACMUT Chip0Q4PACMUT Chip0R4PACMUT Chip0S4PACMUT Chip0A5PACMUT Chip0B5PACMUT Chip0C5PACMUT Chip0D5PACMUT Chip0P5PACMUT Chip0Q5PACMUT Chip0R5PACMUT Chip0S5PACMUT Chip0A6PACMUT Chip0B6PACMUT Chip0C6PACMUT Chip0D6PACMUT Chip0P6PACMUT Chip0Q6PACMUT Chip0R6PACMUT Chip0S6PACMUT Chip0A7PACMUT Chip0B7PACMUT Chip0C7PACMUT Chip0D7PACMUT Chip0P7PACMUT Chip0Q7PACMUT Chip0R7PACMUT Chip0S7PACMUT Chip0A8PACMUT Chip0B8PACMUT Chip0C8PACMUT Chip0D8PACMUT Chip0P8PACMUT Chip0Q8PACMUT Chip0R8PACMUT Chip0S8PACMUT Chip0A9PACMUT Chip0B9PACMUT Chip0C9PACMUT Chip0D9PACMUT Chip0P9PACMUT Chip0Q9PACMUT Chip0R9PACMUT Chip0S9PACMUT Chip0A10PACMUT Chip0B10PACMUT Chip0C10PACMUT Chip0D10PACMUT Chip0P10PACMUT Chip0Q10PACMUT Chip0R10PACMUT Chip0S10PACMUT Chip0A11PACMUT Chip0B11PACMUT Chip0C11PACMUT Chip0D11PACMUT Chip0P11PACMUT Chip0Q11PACMUT Chip0R11PACMUT Chip0S11PACMUT Chip0A12PACMUT Chip0B12PACMUT Chip0C12PACMUT Chip0D12PACMUT Chip0P12PACMUT Chip0Q12PACMUT Chip0R12PACMUT Chip0S12PACMUT Chip0A13PACMUT Chip0B13PACMUT Chip0C13PACMUT Chip0D13PACMUT Chip0P13PACMUT Chip0Q13PACMUT Chip0R13PACMUT Chip0S13PACMUT Chip0A14PACMUT Chip0B14PACMUT Chip0C14PACMUT Chip0D14PACMUT Chip0E14PACMUT Chip0F14PACMUT Chip0G14PACMUT Chip0H14PACMUT Chip0J14PACMUT Chip0K14PACMUT Chip0L14PACMUT Chip0M14PACMUT Chip0N14PACMUT Chip0P14PACMUT Chip0Q14PACMUT Chip0R14PACMUT Chip0S14PACMUT Chip0A15PACMUT Chip0B15PACMUT Chip0C15PACMUT Chip0D15PACMUT Chip0E15PACMUT Chip0F15PACMUT Chip0G15PACMUT Chip0H15PACMUT Chip0J15PACMUT Chip0K15PACMUT Chip0L15PACMUT Chip0M15PACMUT Chip0N15PACMUT Chip0P15PACMUT Chip0Q15PACMUT Chip0R15PACMUT Chip0S15PACMUT Chip0A16PACMUT Chip0B16PACMUT Chip0C16PACMUT Chip0D16PACMUT Chip0E16PACMUT Chip0F16PACMUT Chip0G16PACMUT Chip0H16PACMUT Chip0J16PACMUT Chip0K16PACMUT Chip0L16PACMUT Chip0M16PACMUT Chip0N16PACMUT Chip0P16PACMUT Chip0Q16PACMUT Chip0R16PACMUT Chip0S16PACMUT Chip0A17PACMUT Chip0B17PACMUT Chip0C17PACMUT Chip0D17PACMUT Chip0E17PACMUT Chip0F17PACMUT Chip0G17PACMUT Chip0H17PACMUT Chip0J17PACMUT Chip0K17PACMUT Chip0L17PACMUT Chip0M17PACMUT Chip0N17PACMUT Chip0P17PACMUT Chip0Q17PACMUT Chip0R17PACMUT Chip0S17PACMUT Chip0F5PACMUT Chip0C1PACMUT Chip0B1PACMUT Chip0L4PACMUT Chip0L2 PACMUT Chip0K3PACMUT Chip0J2COCMUT ChipPACntrl0901COCntrl09PAHDR101 PAHDR103 PAHDR104 PAHDR105PAHDR102COHDR1 PAHDR203PAHDR201 PAHDR202COHDR2PAHDR301PAHDR302COHDR3PAHDR401PAHDR402COHDR4PAHDR501PAHDR503PAHDR504PAHDR505PAHDR502COHDR5PAHDR608PAHDR607PAHDR606PAHDR605PAHDR604PAHDR603PAHDR601PAHDR602COHDR6PAHDR701PAHDR702COHDR7PAHDR801PAHDR802COHDR8PAHDR902PAHDR901COHDR9PAHDR1101PAHDR1103PAHDR1104PAHDR1105PAHDR1102COHDR11PAInverter014PAInverter013PAInverter012PAInverter011PAInverter010PAInverter09PAInverter08PAInverter07 PAInverter06 PAInverter05 PAInverter04 PAInverter03 PAInverter02 PAInverter01COInverterPALED01PALED02COLEDPALED0Res02PALED0Res01COLED0ResPALV0vdd01201COLV0vdd012PARes02M01PARes02M02CORes02MPARX0res01101 PARX0res01102CORX0res011PASwitch0170PASwitch01701PASwitch01702PASwitch01703 COSwitch017PAtst01102PAtst01101COtst011PAtst01802PAtst01801COtst018(b)Figure 3.9: (a) The transceiver chip layout. (b) Layout of the printed circuitboard designed to test the transceiver and the CMUT together.modifying the design kit’s DRC, LVS and EXT files by the provider, these newfiles were able to detect the layout issue, and the post-layout simulation whichwas perfect before, was now giving credible errors. After modifying the chiplayout, another IC with the exact same design was sent for fabrication. Figure3.9a shows the layout of the new chip. Also, Figure 3.9b shows the printedcircuit board (PCB) layout which is designed to test this transceiver chip packagetogether with a CMUT package. The transceiver package is a 24 pin CQFP andthe CMUT is a 209 pin PGA package.A sample of the cross-coupled transmitter is also designed and included inanother chip fabrication. Figure 3.10a shows the layout of this chip. Also, Figure3.10b shows two different PCB layouts designed to test this chip. The PCB onthe left is for testing the packaged chip and the one on the right is intended forwire-bonding the transmitter chip to the CMUT chip to eliminate the package andPCB parasitics when testing the two chips together. However, the wire-bondingwas not successful in practice, as the equipment available was not appropriate forour needs.59(a)PA02 PA01COPA02 PA01PA02PA01PA02PA01PA02 PA01PA02 PA01PA02PA01PA02PA01PA02PA01 PA02PA01PA01PA02 PA01PA02PA02PA01PA02PA01PA02PA01PA01PA02PA03PA01 PA02PA01 PA02PA03PA02PA01PA02PA01PA01PA02PA033 PA032 PA031 PA030 PA029 PA028 PA027 PA026 PA025 PA024 PA023PA022PA021PA020PA019PA018PA017PA016PA015PA014PA013PA012PA011PA010PA09PA08PA07PA06PA05PA04PA03PA02PA01PA044PA043PA042PA041PA040PA039PA038PA037PA036PA035PA034PA204pf02 PA204pf01CO204pfPA204pf02PA204pf01PA204pf02 PA204pf01PARes02M02PARes02M01 CORes02MPARes02M02PARes02M01PARes02M02PARes02M01PARes02M02PARes02M01(b)Figure 3.10: (a) The cross-coupled transmitter chip layout. (b) Layout of thetwo PCBs designed to test the cross-coupled transmitter chip (the PCBon the left) and to wire-bond and test the transmitter and CMUT chips(the PCB on the right).3.6.1 Transmitters’ Post-Layout Simulation ResultsIn the design stage of the transmitters, the CMUT load is modeled by Verilog-AMS, as explained in Chapter 2. However, this chapter presents the post-layoutsimulation and test results of the stand-alone high voltage transmitter circuits. Thenext chapter will present the results of the CMUT-CMOS circuit co-simulation andtests.The two transmitters are post-layout simulated considering an 18 pF capacitiveload of a voltage probe and the PCB and chip package parasitics at their outputnode. The input is defined based on the specifications and requirements of theavailable ultrasound machine, listed in Table 3.1.Post-Layout Simulation Results for the Level-Shifted TransmitterFigure 3.11 shows the input and output waveforms of the level-shifted transmitterfor a 2.7 MHZ, 5 V input pulse. This frequency and voltage selection complies601.2 1.4 1.6 1.8 2 2.2x 10−6−50510152025Time (s)Voltage Amplitude (V)Transmitter’s input and output transient response   Output InputFigure 3.11: Post-layout simulation transient response of the level-shiftedtransmitter for 20 V high voltage supply and 18 pf capacitive load.with the input requirements of the circuit as well as operational frequency range ofthe CMUT. The input duty cycle is 50%. The low voltage DC source is 5 V andthe high voltage DC source is set to 20 V. The output, as shown in this figure is a2.7 MHz, 20 V pulse with rise time of 32 ns, fall time of 52.3 ns and input-outputdelay of 23.4 ns on the rising edge. Also, the average power consumption fromlow and high voltage power supplies is 58 mW.Post-Layout Simulation Results for the Cross-Coupled TransmitterFigure 3.12 shows the input and output waveforms of the cross-coupled transmitterfor a 2.7 MHZ, 5 V input pulse. The high-voltage supply (as in Figure 3.4) is 20 V,the Bias-high voltage is 15 V and the Bias-low voltage is 5 V. The rise time is 84 ns,the fall time is 80 ns and the input-output delay is 80 ns on the rising edge. The611 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2x 10−6−505101520Time (s)Voltage (V)Transmitter’s input and output transient response for an 18pf capacitive load  InputOutputFigure 3.12: Post-layout simulation transient response of the cross-coupledtransmitter for 20 V high voltage supply and 18 pf capacitive load.main advantage of this design is its lower power consumption of 25 mW, which isless than half of the level-shifted design. The major drawbacks of this design arethe need for the multiple DC bias supply voltages, i.e., extra pads, and its lowerspeed while keeping the area comparable to the level-shifted design.3.6.2 Receiver’s Post-Layout Simulation ResultsThe receiver circuit is also laid out in Teledyne Dalsa 0.8 µm high-voltagetechnology and occupies 0.078 mm2 of area. For evaluating the receiver circuitperformance, the input to the amplifier is modeled by a current source. The outputof the amplifier is connected to a 50 Ω resistor which represents the inputimpedance of a coaxial cable that connects to the SonixRP ultrasound machine.Figure 3.13 illustrates the post-layout simulation result of the overall gain and62Gain and Bandwidth0.01.02.03.04.05.0Control Voltage ()125.0100.075.050.025.0 0Gain (dB)12.510.0 7.5 5.0 2.5 0−2.5BW (E6) Gain  Bandwidth Control Voltage ()Figure 3.13: Post-layout simulation results of the variation of the gainand 3 dB bandwidth versus control voltage for the variable-gaintransimpedance amplifier.bandwidth versus control voltage. The gain changes from 72 to 90 dBΩ when thecontrol voltage changes from 0 to 3.3 V (after which M3 in Figure 3.5 starts to gointo cut-off, resulting in turning off the circuit and zero gain). The 3 dBbandwidth changes from 11 MHz to 9 MHz for the same range of the controlvoltage and then drops dramatically.The input impedance of the circuit including the switches is less than 1 kΩwhich is significantly lower than the reported input impedances in recent relatedworks [9, 28, 33, 84]. Also, the receiver amplifier consumes 3 mW from a 5 Vsupply which rises to 23.5 mW including the buffers.63(a) (b)Figure 3.14: (a) The level-shifted transmitter chip micrograph. (b) Thefabricated and assembled PCB with the transceiver 24 pin package onthe left and the CMUT 209 PGA package on the right.3.7 Experimental Results3.7.1 Level-Shifted Transmitter CharacterizationThe transmitter IC of Figure 3.9a is fabricated using Teledyne Dalsa 0.8 µmone-poly three-metal high voltage technology. This technology supports bipolarand CMOS transistors with CMOS Vds voltages of up to 300 volts with Vgs limitedto 5 V for most of the high voltage transistors. Low voltage transistors of thistechnology can hold 5 V on the drain-source. Figure 3.14a shows the chipmicrograph. The total transmitter’s area on the chip is 125000 µm2. The packagedchip together with the CMUT chip are mounted on an FR4 Printed Circuit Board(PCB) test fixture designed for testing the full system. The assembled PCB isshown in Figure 3.14b.The high voltage level-shifted transmitter is first characterized without theCMUT chip (which otherwise is connected as it’s load). Figure 3.15 shows typicalmeasured input and output voltage waveforms of the transmitter. The input645.1374 5.1376 5.1378 5.138 5.1382 5.1384 5.1386x 10−305101520Time (s)Voltage (V)Transmitter’s input and output transient responses  InputOutputFigure 3.15: Experimental transient response of the level-shiftedtransmitter’s input and output voltages.voltage is a 5 Vp−p pulse from an Agilent 33250A function generator(representing the ultrasonix imaging machine input to the circuit) and the output isa 20 Vp−p pulse signal at the frequency of 2.7 MHz. The circuit output voltage istested for up to 60 volts at 2.7 MHz without any major distortion in the waveform.Also, the circuit stays functional when the frequency is increased to 5 MHz withpulse-widths as narrow as 62.5 ns (as indicated by ultrasonix machinespecifications). Low and high DC voltages are provided by HAMEG4040 DCpower supply. The transmitter’s output is monitored by a Rohde&SchwarzRTM1054 oscilloscope, with 13 pF probe’s input capacitance. The output signalrise and fall times are measured as 37 ns and 59 ns respectively. Also, theinput-output delay time at the rising edge is 24.5 ns. These values are very closeto the post-layout simulation values of the Section 3.6.1. The slight variation is65(a) (b)Figure 3.16: (a) The cross-coupled transmitter chip micrograph. (b)The fabricated and assembled PCB for testing the cross-coupledtransmitter.most likely because of the package and PCB parasitics. The average powerdissipation from the low and high voltage power supplies is 59 mW which is wellmatched with the post-layout simulation results.3.7.2 Cross-Coupled Transmitter CharacterizationFigure 3.16a shows the cross-coupled transmitter fabricated chip micrograph. Thefabricated PCB for testing this chip is also shown in Figure 3.16b.This high voltage cross-coupled transmitter is characterized using the samesettings as discussed in Section 3.6.1. Figure 3.17 shows typical measured inputand output voltage waveforms of this transmitter. The input voltage is a 5 Vp−ppulse and the output is a 20 Vp−p pulse signal at the frequency of 2.7 MHz. Theoutput signal rise and fall times are measured as 84 ns and 80 ns respectively.Also, the input-output delay time at the rising edge is 77 ns. These values arevery close to the post-layout simulation values of the Section 3.6.1. The slightvariation is most likely because of the package and PCB parasitics. The average665.1374 5.1376 5.1378 5.138 5.1382 5.1384 5.1386x 10−305101520Time (s)Voltage (V)Transmitter’s input and output transient responses  InputOutputFigure 3.17: Experimental transient response of the cross-coupledtransmitter’s input and output voltages.power dissipation from the low and high voltage power supplies is 26.4 mW whichis in very good proximity of the post-layout simulation results.3.7.3 Transimpedance Receiver CharacterizationAs discussed in Section 3.6, experimental tests of the first fabricated chip whichconsisted of the full transceiver block failed due to the VDD and VSS short circuit.A new type of transimpedance amplifier was designed in another chip fabricationiteration which is discussed in Chapter 6.3.8 ConclusionIn this chapter, a low-power variable-gain transimpedance amplifier as well as twohigh-voltage transmitting drivers are presented. The circuits are designed inTeledyne Dalsa 0.8 µm high-voltage (up to 300 V) CMOS process. Post-layout67Table 3.2: Comparison of the post-layout simulation result of thetransimpedance receiver with similar designsParameter [9]† [84] [33] This workTechnology (µm) 1.5 0.18 0.8 0.8Bandwidth (MHz) 10 103-256 11 9-11Gain (dBΩ) 113 84-109 75 72-90Input Impedance (kΩ) 4.3 0.14 - 1Power consumption (mW) 4 8.7 2 3Chip area (mm2) - 0.24 0.01 0.078†Measurement resultResults are per receive channelTable 3.3: Comparison of the experimental results of the two types offabricated transmitters with previously reported designsParameter [85] [86] This work (LS†) This work (CC)Technology (µm) 0.18 0.8 0.8 0.8Input voltage (V) 1.8 5 5 5Output voltage (V) 9.8-12.8 59 20 20Max. output voltage (V) 12.8 59 125 125Output load (pF) 15 20 18 18Frequency (MHz) 1.25 5 2.7 2.7Rise/fall time (ns) 40/50 69/58 37/59 84/80Input/output delay (ns) 22.5 - 24.5 77Power consumption 19.9mA dynamic0.43mA static ∗200 mAdynamic ∗59 mW 26.4 mWChip area (mm2) 0.022 0.85 0.125 0.135∗Simulation result †LS: Level-shifted CC:Cross-coupledResults are per transmit channelsimulations show that the amplifier gain varies from 72 to 90 dBΩ while its 3 dBbandwidth changes from 11 to 9 MHz. The receiver amplifier consumes 3 mWfrom a 5 V supply and its input resistance including the switch resistances is lessthan 1 kΩ.Table 3.2 compares the receiver’s post-layout simulation results with similarworks in literature. Also, Table 3.3 compares the two fabricated transmitters’performance, for 20 V output voltage amplitude at 2.7 MHz, with previous68designs. As the tables show, the receive amplifier has the convenience of gain andbandwidth variability while keeping the power consumption and input impedancereasonably low (note the technology size when comparing the input impedancevalues). For the high-voltage transmitters of this work, the maximum achievableoutput voltage provides the CMUT designer a very wide range of voltages tooptimize the performance. Considering the technology size difference with thework in [85] and so the chip area, the level-shifted design operates with two-timeshigher frequency and higher output voltage with approximately the same rise/falltimes. On the other hand, the very low power consumption of the cross-couplesdesign makes it a perfect choice for very low power applications.69Chapter 4Co-Simulation of TransceiverCircuits and CMUT Devices andOptical Measurements of theTransmitter System 54.1 IntroductionDesign, simulation and circuit functionality measurements of the transceiver’sblocks were discussed in Chapter 3. As mentioned earlier, in the design stage ofthe high-voltage transmitters and low-voltage variable-gain amplifiers, theproposed Verilog-AMS model for CMUTs (presented in Chapter 2) has beenused.5 Parts of Sections 4.2 to 4.4 of this chapter are included in a journal paper submission.70CMUT Model mediumZVdce1e2m1m2TransmitterFigure 4.1: The level-shifted high-voltage transmitter connected to theCMUT’s electrical input terminal.This chapter first presents the results of the co-simulation of the level-shiftedhigh-voltage transmitter and the CMUT devices. Then, the co-simulation resultsof a full transmit/receive system using the CMUT’s Verilog-AMS model ispresented which shows the capabilities and advantages of having a bi-directionalCMUT model. Furthermore, optical measurement results of vibration of theCMUT membrane in the transmit mode in air medium are shown. Thesemeasurements are executed while the CMUT’s actuation pulses are being appliedby the fabricated transmitter of Section 3.3.1. This confirms the functionality ofthe transmitter-CMUT system.4.2 Co-Simulation of the High-Voltage Level-ShiftedTransmitter with the CMUT ModelThe overall system test setup, consisting of the transmitter circuit block ofSection 3.3.1 and the CMUT model, is shown in Figure 4.1. A negative DC bias isconnected to the bottom electrode (e2) of the CMUT, while the transmitter’s AC711 1.5 2 2.5 3 3.5x 10600.20.40.60.81 Symmetric CMUT output frequency response in air mediumFrequency(Hz)Normalized velocity  Vbias= 15VVbias= 30VVbias= 45VVbias= 60VVbias= 75VVbias= 90VFigure 4.2: Symmetric CMUT normalized velocity versus frequency in air,for constant AC and different DC bias voltages.output is connected to the top electrode (e1). This type of connection is equivalentto connecting a mixed AC and DC voltage source to one electrode, while the otherelectrode is grounded. The impedance connected to the mechanical port of theCMUT represents the acoustic impedance of the medium.4.2.1 Front-End System Simulations for the Symmetric CMUTThe symmetric CMUT chip used in this work achieves its optimum sensitivity at20 V of AC excitation and 90 V of DC bias [41]. This CMUT device is designedfor the applications in the frequency range of up to 10 MHz. The performanceof the transmitter-CMUT integration is examined for the air and olive oil media,while the DC bias is swept from 15 V to 90 V by 15 V steps.Figure 4.2 and Figure 4.3 show the CMUT’s normalized membrane velocityversus frequency for air and olive oil media. Different normalizing factors have72104 105 106 10700.20.40.60.81 Symmetric CMUT output frequency response in oil mediumFrequency (Hz)Normalized velocity  Vbias= 15VVbias= 30VVbias= 45VVbias= 60VVbias= 75VVbias= 90VFigure 4.3: Symmetric CMUT normalized velocity versus frequency in oil,for constant AC and different DC bias voltages.15 30 45 60 75 900175200225250275300325350375400DC Bias (V)Displacement (nm)Displacement and resonance frequency changes versus DC bias  1.61.721.841.962.082.22.322.442.562.682.8Resonance Frequency (MHz)DisplacementResonance FrequencyFigure 4.4: Dependence of CMUT membrane displacement amplitude andresonance frequency on the applied DC bias.73−20246Voltage (V)Transmitter’s input01020Voltage (V)Transmitter’s output3.35 3.4 3.45 3.5 3.55x 10−5−20 2Time (s)Current (mA)CMUT outputFigure 4.5: Transient analysis simulation results of the symmetric CMUT’sinput and output waveforms. The results show under-damping behaviorof the air medium.been used for air and oil respectively, to take into account that the amplitude of thevelocities are orders of magnitude different in the two media.Figure 4.2 illustrates the resonance peaks in air, decreasing in frequency as theDC bias increases (due to the spring-softening phenomenon). Figure 4.3 showsthe over-damping effect of the oil medium, resulting in a broad band responsewithout any resonance peak. As it can be seen from this figure, the effect of theDC bias on the CMUT sensitivity is sharply increasing as the voltage gets closer740246Voltage (V)Transmitter’s input01020Voltage (V)Transmitter’s output2.02 2.04 2.06 2.08 2.1 2.12 2.14 2.16 2.18 2.2x 10−5−3.50 3.5Time (s)Current (mA)CMUT outputFigure 4.6: Transient analysis simulation results of the symmetric CMUT’sinput and output waveforms. The results show over-damping behaviorof the olive oil medium.to pull-in values (about 90 volts). This phenomena is valid as well for the airmedium, as illustrated in Figure 4.4. This figure shows the trend in increasing ofthe displacement and decreasing of the resonance frequency versus the DC bias,when the AC voltage amplitude is fixed.To more intuitively illustrate the effect of the medium on the CMUT’smechanical output signal, Figure 4.5 and Figure 4.6 show the transient responseof the transmitter’s input and output as well as the CMUT’s output. As these75104 105 106 10700.250.50.751Frequency (Hz)Normalized velocity in oilSymmetric excitation of the asymmetric CMUT in air and oil media  00.250.50.751Normalized velocity in airOilAirFigure 4.7: Normalized velocity versus frequency, in air and oil media, forthe symmetric excitation of the asymmetric CMUT cell (at 30 V DCbias).figures show, the wide bandwidth of the CMUT in immersion, has preserved theinput pulse’s shape at the output of the CMUT, while in air the narrow bandwidthhas filtered the signal and the output is a single frequency sinusoid. As it is shownin the transient responses, the CMUT output current in immersion is 3.5 mAwhile it is 1 mA in air for the same transmitter’s output (CMUT excitation) pulse.Since the CMUT output current at the mechanical port represents force, we canconclude that the CMUT output pressure in oil is much larger than air.4.2.2 Front-End System Simulations for the Asymmetric CMUTFigure 4.7 shows the frequency response of the asymmetric CMUT cell undersymmetric excitation, in both air and oil media at 30 V DC bias. Similar to thesymmetric CMUT cells, the distinct resonant behavior in air becomes abroad-band over-damped response in oil.763 3.5 4 4.5 5 5.5 6x 10600.330.661Frequency (Hz)Normalized velocity in oilAsymmetric excitation of the asymmetric CMUT in air and oil media  00.330.661Normalized velocity in airOilAirFigure 4.8: Normalized velocity versus frequency, in air and oil media, forthe asymmetric excitation of the asymmetric CMUT cell (at 30 V DCbias).Figure 4.8 illustrates the frequency response of the asymmetric CMUT cellwhen excited asymmetrically, in air and oil media at 30 V DC bias. Comparedto the symmetric actuation case (Figure 4.7) the resonant behavior is preserved inboth cases. Nevertheless, there is a significant down-shift in the resonance peakand amplitude decrease when the medium is changed to oil.The above mentioned simulation results are in good agreement with finiteelement modeling reported in [60], validating our approach towards reduced ordermacro-modeling, integrated within EDA tools.4.3 Co-Simulation of the Full Transmit/Receive Circuitwith the CMUT’s Verilog-AMS ModelFigure 4.9 shows the simulation setup diagram for co-simulation of the fullCMOS-CMUT system. As this figure shows, the transmitter’s input actuation77CMUT1Body InterfaceCMUT2Vdc1e1e2m1m2e2 m2e1 Vdc2CMUT2_outTransmitReceiveHV-TXRX m1TX_outTX_inUltrasound MachineRX_outCMUT1_outBody_outFigure 4.9: Block diagram of the test-setup for co-simulation of the fulltransmit/receive circuits and CMUT Verilog-AMS model.pulses come from the ultrasound machine. The transmitter, which is thelevel-shifted design of Chapter 3 (Section 3.3.1), provides high-voltage pulses atthe input of the transmitting CMUT (CMUT1) which sends the ultrasound wavesinto the medium of interest, i.e., patient body. The model of the body delays andattenuates the signals and then, the signals are reflected back to the receivingCMUT (CMUT2 in this figure). The current signal at the output of CMUT2 flowsinto the receiving circuit, gets converted into voltage, amplified and delivered tothe input of the ultrasound machine. The receiver circuit in this diagram is thevariable-gain transimpedance amplifier introduced in Section 3.4. Figure 4.10shows the input/output waveforms of each block of Figure 4.9. Since the bodyinterface is modeled as an “immersion” medium, the CMUT acts as a wide-band780246Voltage (V)TX Input01020Voltage (V)TX Output−0.0100.01Voltage (V)CMUT1 Output−0.00700.007Voltage (V)Body Output−202 x 10−6Current (A)CMUT2 Output  5.5 6 6.5 7 7.5 8 8.5x 10−60.370.380.39Time (s)Voltage (V)RX OutputFigure 4.10: Results of the co-simulation of the transmitter and receivercircuits with the CMUT Verilog-AMS model.79Figure 4.11: The fabricated and assembled PCB under the laser Dopplervibrometer system, including the transceiver and CMUT packagesfilter and preserves the pulse shape. The achieved transimpedance gain of thereceiver is 72 dBΩ for minimum gain setting of the variable-gain amplifier. Thisis in an excellent agreement with the minimum gain for this receiver, reported inTable 3.2.4.4 Optical MeasurementsAs mentioned in the previous chapter, the transceiver integrated circuit (IC) isfabricated in Teledyne Dalsa 0.8 µm one-poly three-metal high-voltage CMOStechnology. Also, the CMUT chip used in these experiments [41] is designed inPolyMUMPs technology, a three-layer polysilicon micro-machining processoffered by MEMSCAP Inc. (North Carolina, USA).The analog frond-end transceiver and CMUT chips are mounted on the PCBof Figure 3.14b and the complete system is tested using a Polytec Micro SystemAnalyzer (MSA-500), which uses laser Doppler vibrometry (LDV) technique formeasuring the membrane dynamics. Figure 4.11 shows the test system under the80MSA’s microscope. A chirp sinusoidal signal of 2.5 V amplitude with 2.5 V DCoffset is applied to the front-end transmitter input. This signal comes from the MSAand replicates the ultrasound machine’s output pulses to the transmitter input. Thelow and high DC voltages of the transmitter, as well as the CMUT’s DC bias areprovided by a HAMEG 4040 DC power supply. The transmitter output, togetherwith a high voltage DC bias, connects to the bottom electrode of the CMUT usinga bias-Tee (Picosecond Pulse Labs, Model 5535). The top electrode connects tothe ground. For the asymmetric CMUT, the second half of the bottom electrodeconnects to the ground as well.4.4.1 Symmetric CMUT Optical MeasurementsFor the first experiment in the air medium, the high-voltage DC power supply ofthe transmitter is set to 20 V (such that the excitation signal has a peak-to-peakamplitude of ∼20 V) and the DC bias voltage is 60 V. The PCB test fixture isplaced under the MSA microscope (as shown in Figure 4.11) and the LDV is set tomonitor and analyze the membrane movements. The MSA sweeps the input pulsefrequency from 100 kHz to 8 MHz. Figure 4.12 shows the absolute deflectionsof the fundamental and the first two harmonic modes of a symmetric CMUT cell,captured by the vibrometer at 2.6 MHz, 7.4 MHz and 7.8 MHz. The membranedisplacement versus frequency is also shown in Figure 4.13.To study the effect of the DC bias variation on the membrane displacementand its resonance frequency, the DC bias is swept from 15 V to 90 V. Figure 4.14shows the change in the membrane displacement and resonance frequency versusthe DC bias. As shown in this figure, there is a steep drop in the resonancefrequency around 90 V, suggesting that this point is close to the pull-in voltage of81(a) Symmetric CMUTfundamental resonance mode(0,0).(b) Symmetric CMUT harmonicresonance mode (0,2).(c) Symmetric CMUT harmonicresonance mode (1,2).Figure 4.12: LDV measurement of the symmetric CMUT’s absolute valuedeflections in air at different vibration modes.the membrane. This result is in close agreement with the simulation resultsderived in Section 4.2.1 for Figure 4.4. After 90 V, the transmitter circuit’shigh-voltage power supply shows a sudden surge in the current draw, which canbe attributed to the short circuit due to the membrane collapse to the substrate.Another study which is carried out using the LDV is investigating the effectof cross-talk of the vibrating CMUT cells or elements on the neighbor CMUTrows. As Figure 4.15 shows, cross talk is measured on the first row neighbors bymeasuring their peak displacement. Calculations show that the ratio of the neighbor820 1 2 3 4 5 6 7 8x 10600.511.52x 10−10Frequency (Hz)Symmetric CMUT displacement versus frequencyDisplacement (m)Figure 4.13: LDV measurement results of the symmetric CMUT membranedisplacement versus frequency at transmit mode in air, for 60 V DCand 20 V AC input.10 20 30 40 50 60 70 80 90100200300DC Bias (V)Displacement (pm)Membrane displacement and resonance frequency changes versus DC bias  22.22.42.62.8Resonance Frequency (MHz)  Displacement FrequencyFigure 4.14: Experimental results for the symmetric CMUT’s displacementand resonance frequency changes versus DC bias voltage in transmitmode.83Figure 4.15: Measuring the cross-talk on neighbor cells by calculating theratio of the neighbor cells displacement to the actuated CMUTdisplacement.cells displacement to the actuated CMUT displacement is less than 4 % for smallDC biases (<40 V), reaching 12 % for DC bias of about 90 V.4.4.2 Asymmetric CMUT Optical MeasurementsTo study the asymmetric CMUT membrane displacement in air, the top electrodeand one of the half bottom electrodes are grounded. The transmitter high-voltagesource is connected to a 5 V source and the DC bias voltage is 30 V. The outputof the bias-Tee is connected to one of the bottom electrodes and the frequency isswept from 100 kHz to 8 MHz. Figure 4.16 shows the absolute deflections of thefundamental and the first two harmonic modes of the asymmetric CMUT, capturedby the vibrometer at 2.5 MHz, 5 MHz and 7.8 MHz. The membrane displacementfrequency response is also shown in Figure 4.17.Symmetric excitation of the asymmetric CMUT is realized by grounding thetop electrode while the two bottom electrodes are connected to a same voltage,i.e., output of the bias-tee mentioned above. The asymmetric mode cancellation isconfirmed in Figure 4.18, compared to Figure 4.17.It should be noted that due to a design limitation on the size of theinterconnects of the asymmetric CMUT, we could not use a DC bias voltage84(a) Asymmetric CMUTfundamental resonance mode(0,0).(b) Asymmetric CMUTharmonic resonance mode (0,1).(c) Asymmetric CMUT harmonicresonance mode (0,2).Figure 4.16: LDV measurement of the asymmetric CMUT’s absolute valuedeflections in air, at different vibration modes.beyond 30 V, or otherwise the interconnects fail as shown in Figure 4.19. Also, asthe displacements in oil medium are smaller than in air, it is very challenging tomeasure these small displacements in the presence of the environmental noiseusing the MSA-500 equipment. To cancel the noise, complex averaging with afactor of greater than 200 is suggested [60], but that option can only be used withvelocity measurement up to 1.5 MHz of frequency range, which is below theexpected range of operational frequency of the manufactured CMUT.850 1 2 3 4 5 6 7 8x 10600.20.40.60.811.21.41.6 x 10−11Frequency (Hz)Displacement (m)Asymmetric CMUT displacement versus frequency in airFigure 4.17: LDV measurement results for the asymmetric CMUT’smembrane displacement versus frequency in air, at transmit mode for30 V DC and 5 V AC input.4.5 ConclusionThis chapter presented the results of the co-simulation of the analog front-endcircuits and the CMUT’s Verilog-AMS model. Using the Verilog-AMS model, afast transient and frequency domain analyses of the full integrated system isachieved. Implementation of the CMUT model in the mixed-kernel simulator hasallowed the design, co-simulation and optimization of the dedicated analogfront-end circuitry as well.Results of the co-simulation for transmit mode in air are proved to be reliablyclose to the experimental results. Having a single integrated design and simulationenvironment for the CMUT and front-end electronics opens the door towards asystem-level design optimization. This optimization promises to circumvent many860 1 2 3 4 5 6 7 8x 10600.511.522.533.5 x 10−11Frequency (Hz)Displacement (m)Asymmetric CMUT displacement versus frequency in airfor symmetric excitationFigure 4.18: LDV measurement of the symmetric actuation of theasymmetric CMUT’s bottom electrodes. The membrane displacementshows no second harmonic vibration in air.Figure 4.19: An asymmetric CMUT with burnt interconnects is shown onthe right. The interconnects were not designed properly to be able tohandle the current density for DC bias voltages more than 30 V.of the present limitations of the CMUT-based ultrasound imaging systems.87Chapter 5A CMUT Read-Out Circuit withImproved Receive SensitivityUsing an Adaptive BiasingTechnique 65.1 Introduction and MotivationAs mentioned earlier in Chapter 1, CMUTs are generally micromachined thinparallel-plate membranes separated by a small vacuum gap. The membranes aresuspended over a conductive silicon wafer and are supported by insulating posts.In the transmission mode, an electrical AC signal vibrates the membrane of theCMUT elements to generate an acoustic wave. In the reception mode, the incident6 Contents of this chapter except for some parts of Section 5.4 was published in Proceedings ofthe IEEE BioCAS2011 [2].88Figure 5.1: A typical CMUT transducer’s efficiency versus DC bias [43].acoustic waves will cause the membrane of the CMUT elements to vibrate,resulting in the capacitance change of the element that in turn produces a flow ofcurrent which is detected by the readout circuit. In both cases, a DC bias largerthan the AC voltage is applied to the device [105]. The sensitivity of the devicedepends on the amount of this DC bias [29, 41, 43].By applying a DC voltage to the CMUT, an electrostatic field forms betweenthe suspended membrane and the substrate of each device. This field pulls themembrane towards the substrate. On the other hand, there exists a mechanicalrestoring force of the membrane which balances this deflection. By increasingthe bias voltage, a point is reached where the electrostatic force overwhelms therestoring force. At this point, called the collapse voltage, the membrane collapsesdown towards the substrate. It has been shown that the optimal DC bias voltagewhich provides maximum sensitivity is close to this voltage and is called the pull-in voltage, VB [29, 43, 105]. Figure 5.1 illustrates a typical graph of the CMUTelement efficiency versus its DC bias voltage [43]. As it is shown in this figure, theefficiency is generally low, but it significantly increases near the pull-in voltage.In the transmission mode, the total amplitude of the AC signal plus the DC bias89should not exceedVB, while it is preferable to keep the maximum overall amplitudeof the applied voltage as close as possible to this value. In the reception mode, sinceno external AC signal is applied to the CMUT element, the amount of the DC biasis less than VB, resulting in lower efficiency. This is one of the major bottlenecksfor acquiring a better image quality with CMUT devices.Furthermore, in ultrasound imaging systems, echoes that reflect back fromtissue boundaries fade out as they travel back from deeper organs. Thus, in suchsystems, to compensate for this echo signal attenuation, a time-gain compensation(TGC) unit is typically placed after the receiver’s pre-amplifier. This unit applies avariable gain to the received echoes to compensate for the ever-decreasing signalamplitudes. Although adding the TGC helps in increasing the signal levels, still,the very low amplitude echoes of the deeper organs which are lower than thedetection threshold are eliminated by the TGC and the following post-processingunits [58]. Hence, if CMUTs are used, since their electro-mechanical conversionefficiency is increased by applying a proper DC bias in the reception mode, thesensitivity of reception can be adjusted such that fewer signals are lost in thepost-processing unit. By designing an on-chip power management system (PMS),the CMUT can be biased by a fixed high DC voltage (about 70 V for the CMUTintroduced in Chapter 2) and excited by an AC voltage (about 20 V, practically in1 to 10 MHz frequency range) during the transmit mode. In the receive mode, anadaptive DC voltage is added to the fixed high DC voltage source. The adaptivevoltage can be increased versus time so that the echoes of the deeper tissues areconverted to electrical signal with a higher efficiency. This can be implemented asan on-chip adaptive biasing technique that significantly increases the SNR andbalances the signal amplitudes before they are processed by the receive amplifier.90This chapter discusses the realization of a front-end circuit that consists of atotally on-chip CMUT transceiver together with the aforementioned adaptivebiasing unit (ABU). The circuit, which is intended to be integrated with a CMUTarray, is designed in Teledyne Dalsa 0.8 µm high-voltage (up to 300 V) CMOSprocess. In this work, the Verilog-AMS behavioral model of the CMUT,introduced in Chapter 2, is used that not only considers the dynamic behavior ofthe CMUT elements, but also provides the opportunity to evaluate the electricaloutput of the CMUT during the receive time. In this way, the effect of changingthe DC bias can be monitored. This aspect cannot be achieved with theconventional linear small-signal models that are used in literature [28, 33, 86].5.2 Overview of the Front-End CircuitFigure 5.2 shows the block diagram of the CMUT transceiver along with theadaptive biasing unit. The transmitter is a high-voltage driver and the receiver is alow-voltage variable-gain transimpedance amplifier previously discussed inChapter 3.The proposed adaptive biasing unit consists of a power management systemand a digital control unit (DCU). Since the transceiver circuit is connected to boththe ultrasound machine and the CMUT on shared input/output lines on each side,some enable and protection switches are also needed, though for the purpose ofclarity they are not shown explicitly in the figure. In a typical imaging session,after defining the target tissue depth and type, the technician sets up the DCUinputs. This unit controls the output of the PMS to provide an appropriate (high)DC voltage at different time intervals. The output of the PMS is applied to thetransmitter and functions as the high-voltage bias source during the send and91HV DCInput [ 1 : N ]Variable GainControl ( VGC )Control SignalsLV DCClkFrom / To  Ultrasound MachineReceiverCMUTPower Management System ( PMS )Digital Control Unit ( DCU )Adaptive Biasing Unit ( AB U )TransmitterFigure 5.2: Block diagram of the front-end circuitry including the transmitter,receiver and adaptive biasing unit.receive modes. The produced bias voltage is then applied to the CMUT elements.The DCU is also responsible for providing a control voltage for the variable gaininput function of the receiving amplifier. The variable gain control signal alsoturns off the receiver during the transmit time to minimize the power dissipation.The next section will explain the design of the adaptive biasing unit as well asthe modifications to the work in Section 3.3.1 which are needed to provide the highDC voltage at the output of the transmitter during the receive mode.5.3 Design of the Adaptive Biasing UnitAs mentioned in the previous section, the adaptive biasing unit consists of a powermanagement system and a control unit. The power management system in thisdesign is an on-chip switched-capacitor (SC) high-voltage DC-DC up-converterand its multiple outputs are switched into the transmitter high-voltage (HV) DC92M1M2M3M4M6M5Input Vdd ClkClkb2b1d1d2C1C2M7M8b1b2M9M10Output1M11M12Output2d2d1C3C4CL 1 CL 2M13M1 4M1 5M1 6Output3d2d1C5C6CL 3Figure 5.3: Circuit schematic of the DC-DC converter.input line (Figure 5.2) at the desired time.5.3.1 On-Chip SC DC-DC ConverterFigure 5.3 shows the core of the PMS unit which is an SC DC-DC converter. Inan ideal case, when there is no on resistance of the switches or no other parasiticeffects, the circuit provides double, triple and quadruple values of the input low DCvoltage. This design is a modified version of the ones used by Chebli et al.[86] andMaksimovic et al. [106] to meet the 0.8 µm technology requirements, and consistsof two parallel SC circuits being driven by proper gate drive signals.A brief overview of the operation of the circuit is as follows. When Clk is high,M3 is on and node d1 is at ground level. M5 is also on, thus, node d2 is at Vdd . Thisincreases the b2 voltage level to 2×Vdd to keep the voltage across the capacitor C2fixed, as it was charged to Vdd in the previous half cycle. Then, M1 turns on sinceits gate voltage is now risen up to 2×Vdd . This makes the voltage level of node b1,and capacitor C1, equal to Vdd . At this time, M8 turns on and the filter capacitor CL193is charged to 2×Vdd . In the next half cycle, M7 turns on and helps keeping the CL1voltage at 2×Vdd . Other stages can be added to this stage (as shown in Figure 5.3)and their operation principle is the same as above. The voltage conversion ratio ofeach stage, Mi with i=1, 2 or 3, is defined as:Mi =Vout(i)Vin, (5.1)where Vout(i) is the output of each stage and Vin is the input voltage of the circuit.The clock frequency and the capacitor sizes set the output ripple amplitude,and the transistor resistances and capacitor charge and discharge losses affect themaximum achievable output voltages. There is a trade-off between the capacitorand transistor sizes, and the circuit’s area and speed. After the desired voltagesare generated, the appropriate one is applied to the transmitter unit based on thecontrol signal from the DCU.5.3.2 Digital Control UnitThe DCU consists of a number of counters and some digital logic circuits whichdefine the different states of the circuit functionality. Figure 5.4 shows a genericblock diagram for the counter block and the digital logic.The input data of the ultrasound machine’s operator defines the initial values forthe counters. The enable signal is also initiated at the transmit time. The frequencyof the clock signal in Figure 5.4 depends on the depth and also the number of stepsin increasing the DC voltage from the lowest value to the highest desired value andis applied externally. The counter is reset at the end of the receive mode. Thisis when the Control Signal in Figure 5.2 sets the output of the DC-DC converter94CounterBlockResetClockOperator’sInput data To the rest of the DCU Logic for making the control and V GC signalsEnableDigital LogicFigure 5.4: Block diagram of the counter and the digital logic for producingthe control and VGC signals in Figure 5.2.to its desired value for the transmit mode. Also, at this time, the VGC signal inFigure 5.2 turns off the receiver to minimize the power consumption.5.3.3 Modifications to the Transceiver-CMUT InterfaceSince the transmitter and the receiver share a single input/output line on bothinterfaces with each CMUT element and the ultrasound machine, some switchesshould be added to protect the input and output of the low-voltage receiver fromthe high voltages of the transmitter. Also, due to the high applied voltage levelsduring the receive mode, the input of the receiver should be protected from thishigh DC voltage. Figure 5.5 shows the transceiver-CMUT interface. Theprotection has been efficiently done by placing a high-voltage capacitor and ahigh-voltage switch at the interface between the transmitter and the receiverblocks as shown in this figure. The output of the transmitter is directly connectedto the CMUT. In this way, during the transmit mode, the high-voltage pulsesexcite the CMUT. During this time, the switch M1 is off so the receiver isprotected from these pulses. In the receive mode, the high DC voltage is providingan extra bias for the CMUT. Although the switch is on to receive the echo signals95TransmitterReceiverCMUTHVM1H VC1EnableTransmit pathReceive pathFigure 5.5: Transceiver-CMUT interface and the protection scheme.from the CMUT, the capacitor blocks the DC voltage from reaching thelow-voltage transistors of the receiving amplifier while it passes the AC signals.The transmitter circuit used in this design is described previously inSection 3.3.1. As shown in Figure 3.3, in transmission, high voltage pulses aregenerated at the output of the transmitter. The high-voltage source of thetransmitter is supplied from the output of the DC-DC converter as shown in theblock diagram of Figure 5.2. In the receive mode, the N2 high-voltage transistor isturned off using the switch block of Figure 3.3 and P2 high-voltage transistoroperates as a high impedance load, so the high DC voltage appears at the CMUTterminals as an extra DC bias.5.4 Post-Layout Simulation and Experimental ResultsThe front-end circuit is designed and laid out in Teledyne Dalsa 0.8 µm high-voltage CMOS technology. Figure 5.6a shows the layout of the fabricated DC-DC converter IC. This layout consists of two similar DC-DC converter circuits96(a)PA0COPA0PA0PA0 PA0GNDPA0 PA01PA0PA0PA01PA02PA03PA04PA05PA06PA07PA08PA09PA010PA011PA012PA013PA014PA0 PA0PA01PA0PA10CO1PA204pf02PA204pf0 CO204pfPA302PA30CO3 PA702PA70CO7 PA802PA80CO8PA902PA90CO9 PA1002PA100CO10PA1102PA110CO11PA1202PA120CO12PA18 or 3802PA18 or 380CO18 or 38PA3402PA340 CO34PADIP40040 PADIP40039 PADIP40038 PADIP40037 PADIP40036 PADIP40035 PADIP40034 PADIP40033 PADIP40032 PADIP40031 PADIP40030 PADIP40029 PADIP40027 PADIP40026 PADIP40025 PADIP40024 PADIP40023 PADIP40022 PADIP40021PADIP40020PADIP40019PADIP40018PADIP40017PADIP40016PADIP40015PADIP40014PADIP40013PADIP40012PADIP40011PADIP40010PADIP4009PADIP4008PADIP4007PADIP4006PADIP4005PADIP4004PADIP4003PADIP4002PADIP4001PADIP40028CODIP40PATrim Cap0 COTrim CapPAVDD02PAVDD01COVDDPA10 PA0PATrim Cap0(b)Figure 5.6: (a) The DC-DC converter layout and (b) Layout of the PCB whichis designed to test the DC-DC converter’s functionality.and occupies 3.12mm× 2.32mm, with each DC-DC converter having an area of1.2mm× 0.47mm. This chip is to be placed on a PCB to test the functionality ofthe DC-DC converter circuit. Figure 5.6b shows the PCB layout.To verify the DC-DC converter performance, a 5 V DC voltage is applied tothe input of the circuit (the LV DC input in Figure 5.2) in the simulation setup. Thethree simulation outputs of the DC-DC converter are shown in Figure 5.7.As it can be seen in this figure, the steady-state outputs are achieved afterabout 5 µs and are equal to 8.7 V, 12.3 V and 16.1 V. This means that the voltageconversion ratio for each stage is: M1 = 1.74,M2 = 2.46,M3 = 3.22. The circuitconsumes an average of 0.5 mW from the 5 V input DC source and an average of50 µW from the clock sources. As mentioned earlier, the switches’ resistancesand capacitor charge and discharge losses are the main non-idealities that causedeviations from the ideal multiplier factors (conversion ratios) in the DC-DCconverter.970 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2x 10−54681012141618Time (s)Voltage (V)Outputs of the three stages of the DC−DC converter  InputOut1Out2Out3Figure 5.7: Simulation results of the outputs of the on-chip SC DC-DCconverter.Figure 5.8 shows the chip micrograph and the fabricated and assembled PCB.To test the circuit experimentally, a 5 V DC voltage is applied to the “vdd input”using a Xantrex LXQ 30-2 DC power supply. A 1 MHz, 5 Vp−p clock pulseis also applied to the Clk input using an Agilent 33250A function generator. Toprovide the Clk, a discrete high voltage inverter IC is mounted on the PCB. Thethree outputs of the DC-DC converter are measured using a Rohde & SchwarzRTM 1054 oscilloscope. Figure 5.9 shows these three outputs at 8.4 V, 11.9 Vand 15.3 V, with an average of 0.57 mW total power consumption which is veryclose to the simulation results. The voltage conversion ratio for each stage is:M1 = 1.68, M2 = 2.38, M3 = 3.06. The effect of the switches, capacitors andinterconnect loss is more prominent in the experimental results. One importantfactor is that the extraction tool of the technology does not take into account the98(a) (b)Figure 5.8: (a) The DC-DC converter chip micrograph and (b) The fabricatedPCB to test the 40 pin IC package which contains the DC-DC converterchip.parasitic resistances of the layout. This definitely results in discrepancies in voltagedrop measurements. Also, the input impedance of the oscilloscope’s probe (whichis considered as a pure 13 pf capacitance in simulation) can be one source of themismatch between the simulation and test results.As mentioned earlier in Chapter 4, the fabricated CMUT is not able to work inthe receive mode. Therefore, its functionality at different DC bias levels is testedjust by simulation.To run this test, we have used the proposed CMUT Verilog-AMS behavioralmodel of Chapter 2. Figure 5.10 shows the simulation test setup. CMUT1 is themodel of the CMUT during the transmit mode. Body Interface block is modelinga body tissue that attenuates and delays the signal and CMUT2 is the model of theCMUT during the receive mode. Voltage source Vdc2 in this model is thecombination of the fixed DC bias and the extra DC bias provided by the DC-DCconverter. CMUT2 generates the output current representation of the acoustic990 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1x 10−646810121416Time (s)Voltage (V)Outputs of the three stages of the DC−DC converter − Experimental results  InputOut1Out2Out3Figure 5.9: Experimental results of the outputs of the on-chip SC DC-DCconverter.CMUT 1 Body Interface CMUT 2Vdc 1e1e2m1m2e2m2e1Vdc 2CMUT 2 OutputSend ReceiveTransmitterReceiverm1Figure 5.10: Simulation setup to test the adaptive biasing concept.echoes that should be sent to the receiver for amplification. Figure 5.11 shows theexponentially increasing (shown by the dashed-line) CMUT2 output for threedifferent DC bias levels of 30 V, 70 V and 90 V. Note that the output increases100Figure 5.11: Post-layout simulation results for CMUT2 output (Top) forvariable DC bias levels (Bottom).while the amplitude of the CMUT2 input signal is not changing. This proves thegain increase with increasing the DC bias voltage on the CMUT. Also, this figureshows that the sensitivity of the CMUT is much higher near it’s pull-in voltage.For instance, the 40 V increase in the DC bias has increased the output current bya factor of two, while the 20 V increase in the DC bias closer to the pull-in voltagehas multiplied the output by about 2.5 times. The pull-in voltage is close to 90 V,the same as the model characterized in Chapter 4.5.5 ConclusionIn this chapter, an adaptive biasing unit which includes a power managementsystem and a digital control unit is proposed. This system is intended for use in101CMUT transceiver front-end circuitry. Simulation results show that the adaptivebiasing unit elevates the CMUT sensitivity. A Verilog-AMS model is used tomodel the dynamic behavior of the CMUT arrays. The circuit is designed andfabricated in Teledyne Dalsa 0.8 µm high-voltage CMOS process. The DC-DCconverter functionality is tested in simulation and experimental setups. Effect ofthe switch resistance and the capacitor parasitics and charge and discharge lossesdegrades the voltage levels in test results comparing to the simulation results. TheDC-DC converter shows very low power consumption which is perfect for ourintended low power analog front-end unit design. The effect of the adaptivebiasing in increasing the CMUT elements’ sensitivity during the receive mode isalso shown and confirmed by post-layout simulations.102Chapter 6Transceiver Design forCMUT-Based Super-ResolutionUltrasound Imaging 76.1 Introduction and BackgroundAdvantages of the capacitive micromachined ultrasonic transducer (CMUT)technology and its broad range of applications have made this technology apromising alternative for its piezoelectric counterpart. Compared to piezoelectrictransducers, CMUTs have wider operational bandwidth and can be integrated withthe microelectronic readout circuitry [44, 54, 63]. Furthermore, flexibility in arraydesign has made CMUTs a viable choice for volumetric flow measurement andforward looking intracardiac catheter-based imaging [55, 66]. The main focus of7 Parts of Sections 6.1, 6.3 and 6.4 were published in the Proceedings of the IEEE ISCAS2014[3]. A version of this chapter is submitted for a journal paper publication.103(a) (b)Figure 6.1: Asymmetric CMUT fundamental and second harmonic modes ofvibration, measured by a laser Doppler vibrometer: (a) Fundamentalmode, (b) Second harmonic (asymmetric) mode.the research on CMUTs has been on the behavior and applications of thefundamental frequency of vibration [28, 33, 40]. Although a few researchers havestudied the higher-order harmonics of vibration [27, 107, 108], the potentialapplications of asymmetric vibration modes are only proposed recently[35, 79, 80]. In these studies, a dual-bottom-electrode CMUT, termed as anadaptive or asymmetric CMUT, is used to implement the proposed applications.The fundamental and harmonic modes of operation are previously introducedin Chapter 2. Figure 6.1 shows the optically measured displacement profile of anasymmetric CMUT cell in its first two modes of vibration. The asymmetric mode(also known as the second-harmonic mode) is excited if the acoustic force orelectric potential applied to the CMUT membrane is non-symmetric. Unlike thefundamental mode of vibration with a damped behavior in immersion, theasymmetric mode allows for the exchange of kinetic energy between the twovibrating parts of the dual-electrode CMUT cell. Due to lower radiation loss ofthis mode, the asymmetric vibration dissipates smaller energy and remainsresonant even when immersed in fluids such as oil. Using the asymmetric CMUT104structure, independent signals from the two halves of the membrane can beobtained and used for super-resolution imaging [35]. Also, using the asymmetricmode of vibration, new beam-forming methods such as tiltable ultrasonictransducers and estimation of the direction of ultrasonic wave arrival in receptioncan be implemented [79, 80].The super-resolution imaging technique takes advantage of both modes ofoperation of an array of multi-electrode CMUT cells. In reception, while directbeam reflections excite the fundamental resonance mode of a given CMUTelement, the signals reflected from the off-axis targets may excite the asymmetricmode of its adjacent CMUT devices; provided that, the frequency of thefundamental vibration mode of a CMUT element is close to the resonancefrequency of the asymmetric vibration mode of an adjacent CMUT element. Theasymmetric excitation contains information that can help to more accuratelyestimate the reflector locations. It has been shown that using such informationfrom both resonance modes results in higher image resolution compared to theclassical approach which is mainly based on using the average value of thereceived signals [35].This chapter focuses on the design of a CMOS transceiver circuit forsuper-resolution imaging systems. The transmitter in this design should providehigh-voltage pulses to a given CMUT element and the receiver should be able toconcurrently process two types of signals, from both the fundamental andasymmetric modes of vibration. Different CMUT transmitters and receivers forthe fundamental mode of vibration have been discussed in literature[28, 33, 44, 55, 64]. However, to the best of the authors’ knowledge, the front-endcircuit for simultaneous receiving and processing of the signals from the105T/ RSwitchT/ RSwitchT/ RSwitchUltrasound MachineAnalog Front - End CircuitH V TXVG - TIAOp - AmpBufferEnableTX / R X ( 2 k - 1 )R X ( 2 k )V G - TIAOp - AmpBuffer100 CMUT cells64 CMUT elementsCMUT Array( 2 k ) element( 2 k - 1 ) elementFigure 6.2: Block diagram of the front-end circuitry that interfaces to theodd (darker color) and even row (lighter color) CMUT elements andultrasound machine.fundamental and harmonic mode of vibration has not been reported yet. Theproposed transceiver circuit is designed in a 0.35 µm high-voltage CMOStechnology from Austria Micro Systems (AMS).As shown in Figure 6.2, the one dimensional CMUT array used in this workhas 64 elements, each element being a row of 100 asymmetric CMUT cells. ThisCMUT array is designed in a way that the fundamental frequency of vibration ofthe odd rows (4.35 MHz in air and 1.38 MHz in immersion), is close to theasymmetric mode of the even rows (4.36 MHz in air and 1.55 MHz inimmersion). Each odd row (darker color in the figure) in this array is used to sendand receive the fundamental mode signals and each even row (lighter color in thefigure) is responsible for receiving the signals from the asymmetric modevibrations. Note that benefiting from the spring-softening concept, by applyingproper DC bias voltages to the CMUT cells of the even rows their asymmetric106mode frequency can be adjusted to further match the fundamental frequency ofthe CMUT cells in the odd rows.The ultrasound transmission is carried out in the fundamental mode due to itshigher penetration depth. In a typical setting for transmission, the top electrode ofa given CMUT element is connected to the ground and the two bottom electrodesare connected to identical DC bias and excitation pulses, so that the transmissioncan be carried out in the fundamental mode. In the receive mode, both odd andeven rows are biased by proper DC voltages. As Figure 6.1 shows, the receptionin the fundamental mode generates two in-phase signals while the asymmetricmode generates two 180◦ out-of-phase signals. These in-phase and differentialsignals which are in the form of current [1], need to be collected by odd and evenrow receivers (RX2k−1 and RX2k for k=1, 2, . . . , 32), respectively and thentransformed to two voltage signals and sent to an ultrasound imaging machine forfurther processing and display. One of the voltage signals represents thefundamental mode of vibration while the other one represents the asymmetricmode.In this chapter, the presentation of the transceiver design and the relatedsimulation and experimental results are organized as follows: Section 6.2 presentsthe detailed high-voltage transmitter design. Section 6.3 describes the receiverdesign details. Section 6.4 presents the post-layout simulation results of thefull-transceiver system. Experimental results are discussed in Section 6.5 andfinally, Section 6.6 concludes the chapter.107EnableP 1 P 2N 1 N 2HV V dd HV V ddLevelShifterTo CMUTInput PulseHigh Voltage Transmitter CircuitSwitchBlockFigure 6.3: Circuit schematic of one of the two TX2k−1 high-voltagetransmitters.6.2 High-Voltage Transmitter DesignIn order to have the flexibility of exciting each of the two bottom electrodes of anodd-row element (CMUT2k−1) individually, and providing symmetric orasymmetric excitation, two identical high voltage transmitters are designed, eachhaving the capability of generating up to 60 V pulses. The output of eachtransmitter connects to one of the CMUT bottom electrodes and shares thatinterface with an odd-row receiver (RX2k−1). The circuit schematic as well as theprotection and enable switches of this block are similar to the level-shifter-basedhigh-voltage transmitter design of Figure 3.3 in Chapter 3, shown here asFigure 6.3.The low DC voltage in this design is 3.3 V and the high DC voltage can rise upto 60 V. Also, the circuit is designed to be able to drive an 18 pF capacitive load forpulse frequencies of up to 10 MHz, with 30 V pulse amplitude. The selection of108the voltage level and frequency is due to the specifications and target applicationsof the fabricated CMUT device [41]. For the capacitive load, the input capacitanceof the test equipment and typical board parasitics are taken into account.6.3 Receiver DesignFigure 6.4 shows the block diagram and circuit schematics for the proposedfundamental and asymmetric mode receivers. The input impedance of theultrasound machine is 50 Ω. Also, the machine has an input voltage limitation of± 1 V. Since the typical input current from the CMUT sensor varies betweenhundreds of nA to a few µA, a variable gain of 100 to 120 dBΩ is targeted for thisdesign. Furthermore, since the CMUT sensor is designed for ultrasound imagingapplications with up to 10 MHz frequency range, the target receiver bandwidth is10 MHz. The power supplies used in this design are 3.3 V for VDD and 0 V forVSS.Receiver RX2k−1 (Figure 6.4a) receives the two in-phase current signals iinput1and iinput2, adds and amplifies them and converts the resulting current to a voltagesignal through a transimpedance amplifier. This voltage is then further amplifiedby an operational-amplifier (op-amp) stage followed by an output buffer to formthe fundamental signal SFund which is sent to the ultrasound imaging machine.Receiver RX2k (Figure 6.4b) receives two differential signals iinput3 and iinput4,amplifies and converts them to voltage by a transimpedance amplifier stagefollowed by a differential amplifier. The resulting voltage is buffered to form theasymmetric signal, SAsym, which is then sent to the ultrasound machine. Each unitof the two receivers is carefully designed to accommodate low-power dissipationand high linearity. Design details are discussed in the following subsections.109iinput1R 2R 1V DDV cntrlV out1M1M2M 3R L 1V SSiinput2R 4R 3M4M5M12M9M6M8 M11V out2V b_ oaM7 M10C 1 R 5 R 6C3C2M19M16M13M15 M18V b_ bufM14 M17CcompM20M21M22M23V out_ AdderVariable Gain Transimpedance Amplifier ( VG - TIA ) Op - Amp Unity Gain BufferV DD V DDV SS V SSVG - TIA Op - Amp Unity GainBufferiinput1iinput2A CMUT cell from a  (2k-1) element V out1 V out2 V out_ AdderUltrasound Machine( S F UN D )V g _ oa1V g _ oa1(a) Receiver RX2k−1M20M17M14M16 M 19V b_ bufM 15 M18CcompM 21M22M23M24V out_ DiffVariable Gain Transimpedance Amplifier ( VG - TIA ) Unity Gain BufferV DDV SSVG - TIA Op - Amp Unity GainBufferiinput3iinput4A CMUT cell from a  (2k) element V out1 V out3 V out_ DiffUltrasound Machineiinput3R 2R 1V DDV cntrlM1M 2M3R L 1V SSiinput4R 4R 3M4 M6M13M10M 7M9 M12V out3V b_ oa V out1M 8 M 11R 5 R 6C 3Op - AmpV DDV SSV out1C 1M 5R L 2V out2C2 V out2V out2 ( S Asym )V g _ oa2 V g _ oa2(b) Receiver RX2kFigure 6.4: Block diagrams and circuit schematics of the odd and even rowreceivers: (a) Receiver RX2k−1 and (b) Receiver RX2k.6.3.1 Receiver RX2k−1Transimpedance Analog AdderFigure 6.4a shows the input stage of the receiver RX2k−1 which is based onprevious designs by Behnamfar et al. [1] and Park et al. [96]. The two input110currents are applied to two identical regulated-cascode transimpedance amplifiers(RGC-TIAs) which share a single load branch. By sharing their load, they act asan analog adder for the two current signals. These RGC-TIAs, in addition toproviding robustness to the parasitic variations of the input node, offer lower inputimpedance as compared to the common-source stages with resistive feedback[96]. Since CMUT interface is shared between the transmitter and receiver (referto Figure 6.2), the low input impedance of the RGC-TIAs is particularlyadvantageous. This low impedance increases the amount of current that is sunk bythe receiver which in turn reduces the received power loss. Note that due tocertain technology limitations, pass transistor or diode switches couldn’t beimplemented on the path between the transmitter’s output and the CMUT input inFigure 6.2.As shown in [1], the input impedance of each individual circuit can be writtenas:Rin ≈1gm1(1+gm2(R2‖ro2))≈1gm1A f b, (6.1)where, A f b is defined as the (M2, R2) gm-boosting feedback circuit gain. Thetransimpedance gain of this circuit can also be derived as:Vout1iinput1 + iinput2=R1RLgm1(A f b +1)1+gm1R1(A f b +1)≈ RL, (6.2)where, RL is the total load resistance seen by M1. As shown in Figure 6.4a, theload resistance is the series combination of RL1 and the resistance of M3 which isbiased in triode:RL = RL1 +RM3,triode. (6.3)111By changing the control voltage applied to the gate of M3, the gain and thebandwidth of this stage can be adjusted. Due to the wide dynamic range of theinput current, having a variable gain is important to avoid saturation of the nextstages and/or the loss of a low amplitude signal.Op-Amp Stage of the RX2k−1As Figure 6.4a shows, the next gain stage is an op-amp. Since this stage is ac-coupled to the previous stage by capacitor C1, the gate of the input transistor M9needs to be biased properly. This biasing is realized by using the resistor R5. Tomaintain the DC balance and symmetry of the op-amp, the same DC bias is appliedto the gate of M12 using R6. Also, this gate is connected to the capacitor C3, whichis equal to C1. In the proof-of-concept prototype, the gate bias voltage ofVg oa1 andthe op-amp’s current source bias voltage of Vb oa are applied externally to provideflexibility to tune the circuit’s gain, bandwidth and operating point.Output Buffer for Receiver RX2k−1A unity gain output buffer is designed based on the one introduced in [109] and isshown in Figure 6.4a. The input transistors are kept small to avoid deteriorating thebandwidth and the output devices are designed to drive the 50Ω load (i.e., the inputimpedance of the ultrasound machine). The Vb bu f is also applied from outsideof the chip and can tune the circuit performance. Also, Ccomp is a compensationcapacitor and is designed to adjust the bandwidth of the unity gain buffer.1126.3.2 Receiver RX2kTransimpedance Amplifier StageAt the input stage of this receiver, each asymmetric current signal is first amplifiedby an RGC-TIA circuit (refer to Figure 6.4b). The two RGC-TIA circuits also havegain-bandwidth tuning capability. Since the two input currents are differential, thetwo output voltages of this stage are differential. These two voltages are applied tothe inputs of the next gain stage which is an op-amp.Differential Amplifier StageThe differential to single-ended op-amp shown in Figure 6.4b provides theamplified difference of the two out-of-phase signals from the previous stage.Equivalently, this stage provides the weighted sum of the corresponding in-phasesignals. The DC bias of the two input transistors is provided using the sameapproach described in Section 6.3.1. The bias voltage of the current mirror andthe DC bias voltage of the gates of M10 and M13 are used to adjust the gain,bandwidth and the DC operating point of the circuit.Output Buffer for Receiver RX2kA unity-gain buffer similar to that of the receiver RX2k−1 is used to drive the 50 Ωload.It is worth mentioning that the required number of transmitters and receiversfor the electronic interface of the proposed configuration is comparable to that ofthe state-of-the-art CMUT systems which do not use channel multiplexing [1, 28,33, 54].113(a)PA05PA04PA03PA02PA01COPA05PA04PA03PA02PA01PA05PA04PA03PA02PA01PA05PA04PA03PA02PA01PA0PA01PA02PA01PA02 PA0PA01PA02PA03PA04PA05PA06PA01PA02PA03PA04PA05PA06PA01PA02PA03PA04PA05PA06PA05PA04PA03PA02PA01PA05PA04PA03PA02PA01PA0PA0PA0PA0PA0PA0PA0PA0PA01PA02PA03PA04PA05PA06PA05PA04PA03PA02PA01 PA01PA02PA03PA04PA05PA06PA01PA02PA03PA04PA05PA06PA01PA02PA0PA0PA0PA0PA0PA02 PA01PA0PA02PA01 PA0PA0PA0PA0PA0PA0PA0PA0PA0PA0PA0PA0PA0PA02PA01PA0PA0PA0PA038PA039PA033PA032PA031PA030PA029PA028PA027PA026PA025PA024PA023PA022PA021PA020PA019PA018PA017PA016PA015PA014PA013PA012PA011PA010PA09PA08PA07PA06PA05PA04PA03PA02PA01PA044 PA043 PA042 PA041PA040 PA037 PA036 PA035 PA034PA0PA0PA0PA0PA0PA0PA0PA0 PA0PA0PA0PA0PA0PA0PA0PA0PA0PA0PA01PA02PA0PA02PA01PA0 PA0PA0PA0PA0 PA0PA0 PA0PA0PA0PA001uf02PA001uf01CO001ufPA001uf02PA001uf01PA001uf02PA001uf01PA001uf02PA001uf01PA001uf02PA001uf01 PA001uf02PA001uf01PA001uf02PA001uf01PA001uf02PA001uf01PA10CO1PA20CO2PA2M Res01PA2M Res02CO2M ResPA2M Res01PA2M Res02PA2M Res01PA2M Res02PA2M Res01PA2M Res02PA2M Res01PA2M Res02 PA2M Res01PA2M Res02PA2M Res01 PA2M Res02PA2M Res01 PA2M Res02PA30CO3PA40CO4PA50CO5PA60CO6PA70CO7PA80CO8PA90CO9PA100CO10 PA110CO11PA120CO12PA130CO13PA140CO14PA150CO15 PA160CO16PA170CO17 PA180CO18PA190CO19PA200CO20PA210CO21 PA220CO22PA230CO23PA240PA2401PA2402PA2403CO24PA250CO25PA260 CO26PA270 CO27PA280CO28PA290CO29PA300 CO30PA310 CO31PA320CO32PA330CO33PA340 CO34PA370CO37PA380CO38PA390PA3901PA3902PA3903CO39PA400CO40PA410CO41PA420CO42PA430CO43PA440CO44PACMUT0J2 PACMUT0K3PACMUT0L2 PACMUT0L4PACMUT0B1PACMUT0C1PACMUT0F5PACMUT0S17PACMUT0R17PACMUT0Q17PACMUT0P17PACMUT0N17PACMUT0M17PACMUT0L17PACMUT0K17PACMUT0J17PACMUT0H17PACMUT0G17PACMUT0F17PACMUT0E17PACMUT0D17PACMUT0C17PACMUT0B17PACMUT0A17PACMUT0S16PACMUT0R16PACMUT0Q16PACMUT0P16PACMUT0N16PACMUT0M16PACMUT0L16PACMUT0K16PACMUT0J16PACMUT0H16PACMUT0G16PACMUT0F16PACMUT0E16PACMUT0D16PACMUT0C16PACMUT0B16PACMUT0A16PACMUT0S15PACMUT0R15PACMUT0Q15PACMUT0P15PACMUT0N15PACMUT0M15PACMUT0L15PACMUT0K15PACMUT0J15PACMUT0H15PACMUT0G15PACMUT0F15PACMUT0E15PACMUT0D15PACMUT0C15PACMUT0B15PACMUT0A15PACMUT0S14PACMUT0R14PACMUT0Q14PACMUT0P14PACMUT0N14PACMUT0M14PACMUT0L14PACMUT0K14PACMUT0J14PACMUT0H14PACMUT0G14PACMUT0F14PACMUT0E14PACMUT0D14PACMUT0C14PACMUT0B14PACMUT0A14PACMUT0S13PACMUT0R13PACMUT0Q13PACMUT0P13PACMUT0D13PACMUT0C13PACMUT0B13PACMUT0A13PACMUT0S12PACMUT0R12PACMUT0Q12PACMUT0P12PACMUT0D12PACMUT0C12PACMUT0B12PACMUT0A12PACMUT0S11PACMUT0R11PACMUT0Q11PACMUT0P11PACMUT0D11PACMUT0C11PACMUT0B11PACMUT0A11PACMUT0S10PACMUT0R10PACMUT0Q10PACMUT0P10PACMUT0D10PACMUT0C10PACMUT0B10PACMUT0A10PACMUT0S9PACMUT0R9PACMUT0Q9PACMUT0P9PACMUT0D9PACMUT0C9PACMUT0B9PACMUT0A9PACMUT0S8PACMUT0R8PACMUT0Q8PACMUT0P8PACMUT0D8PACMUT0C8PACMUT0B8PACMUT0A8PACMUT0S7PACMUT0R7PACMUT0Q7PACMUT0P7PACMUT0D7PACMUT0C7PACMUT0B7PACMUT0A7PACMUT0S6PACMUT0R6PACMUT0Q6PACMUT0P6PACMUT0D6PACMUT0C6PACMUT0B6PACMUT0A6PACMUT0S5PACMUT0R5PACMUT0Q5PACMUT0P5PACMUT0D5PACMUT0C5PACMUT0B5PACMUT0A5PACMUT0S4PACMUT0R4PACMUT0Q4PACMUT0P4PACMUT0N4PACMUT0M4PACMUT0K4PACMUT0J4PACMUT0H4PACMUT0G4PACMUT0F4PACMUT0E4PACMUT0D4PACMUT0C4PACMUT0B4PACMUT0A4PACMUT0S3PACMUT0R3PACMUT0Q3PACMUT0P3PACMUT0N3PACMUT0M3PACMUT0L3PACMUT0J3PACMUT0H3PACMUT0G3PACMUT0F3PACMUT0E3PACMUT0D3PACMUT0C3PACMUT0B3PACMUT0A3PACMUT0S2PACMUT0R2PACMUT0Q2PACMUT0P2PACMUT0N2PACMUT0M2PACMUT0K2PACMUT0H2PACMUT0G2PACMUT0F2PACMUT0E2PACMUT0D2PACMUT0C2PACMUT0B2PACMUT0A2PACMUT0S1PACMUT0R1PACMUT0Q1PACMUT0P1PACMUT0N1PACMUT0M1PACMUT0L1PACMUT0K1PACMUT0J1PACMUT0H1PACMUT0G1PACMUT0F1PACMUT0E1PACMUT0D1PACMUT0A1COCMUTPALED01PALED02COLEDPALED01PALED02PALED0Res02 PALED0Res01COLED0ResPALED0Res02 PALED0Res01PAsh01PAsh02 COshPAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02 PAsh01PAsh02PAsh01PAsh02PAsh01PAsh02 PAsh01 PAsh02PAsh01 PAsh02PAsh01PAsh02PAVDD0SW0COVDD0SWPA10PA20PA30PA40PA50PA60PA80PA90PA100PA110PA120PA130PA140PA200 PA230PA250PA260PA270PA280PA290PA310PA320PA330PA370PA380PA400PA420PA0PA0PA0PA0PA0PAVDD0SW0(b)Figure 6.5: (a) The full transceiver chip layout and (b) Layout of thePCB which is designed to test the transceiver circuit and the CMUTtransducer.6.4 Post-Layout Simulation ResultsThe circuit is designed and laid out in the AMS 0.35 µm high-voltage CMOStechnology. Figure 6.5a shows the layout view of the chip that occupies1.86 × 2.14 mm2 and contains one complete transceiver as well as individualtransmitter and receivers for testing purposes. The complete transceiver blockoccupies 0.86× 0.38 mm2. This chip is intended to be packaged and placed on aPCB together with the CMUT package for test and measurement of the overallsystem performance. Figure 6.5b shows the layout of this PCB.6.4.1 High-Voltage Transmitter’s Post-Layout Simulation ResultsFor verifying the transmitter’s circuit performance, a 3.3 V voltage source is usedas the low-voltage VDD of the circuit. Also, another DC source is used to providethe high-voltage DC supply HV Vdd. At the input of the circuit, a 3.3 V pulsewith a frequency of 1.38 MHz is applied and the output is connected to an 18 pF1141 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6−505101520253035Time (µs)Voltage Amplitude (V)Post−layout simulation results of the transmitter’s input and output transient response  TX OutputTX InputFigure 6.6: Post-layout simulation results of the high voltage transmitterinput and output voltages.capacitive load, which represents the input impedance of the oscilloscope as wellas the package and PCB parasitics. The choice of this particular frequency, i.e.,1.38 MHz, is to show the functionality of the circuit for driving the CMUT up tothe fundamental vibration frequency during immersion in a typical fluid medium.Figure 6.6 shows the input and output of the transmitter for a 30 V high-voltagesupply. The output has a rise time of 64 ns, fall time of 51 ns and 32 ns delay (frominput rising edge to output rising edge). The input signal has 5 ns rise and fall timesto comply with the function generator which is used in the experimental tests. Theoverall power consumption of the circuit for a 30 V high-voltage DC source and a3.3 V low-voltage supply is 97.5 mW.6.4.2 RX2k−1 Post-Layout Simulation ResultsTo verify the RX2k−1 receiver circuit functionality, two 1.38 MHz in-phase pulsecurrent sources with 250 nA peak-to-peak amplitude are applied to the circuit. The1154.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6 6.2x 10−6−10123 x 10−7Input Currents (A)Transient response of the receiver (2K−1)’s inputs and output  4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6 6.2x 10−61.41.451.51.551.6Time (s)Output Voltage (V)  Input1Input2RX (2K−1)OutputFigure 6.7: Post-layout simulation results of the RX2k−1 receiver: transientresponse of the in-phase pulse input currents and the output voltage at1.38 MHz.selection of the frequency and shape of the input signal is based on the fact thatthis receiver is designed to receive in-phase current signals when the CMUT is inimmersion. The bias and gain control settings are as follows: the gain control of thefirst stage in Figure 6.4a is at 0 V, which corresponds to the smallest gain that thisstage provides. This 0 V provides the maximum Vgs for the M3 PMOS transistorin triode, which leads to a minimum resistance value. The bias voltage Vb oa of theop-amp stage is set to 2.2 V, and the Vg oa1 is 1.55 V. Vb bu f is set to 1.55 V as well.Figure 6.7 shows the transient response of this receiver with about 105 dBΩ oftransimpedance gain. The power consumption for an average gain setting is 2 mW1160 0.2 0.4 0.6 0.8 1x 10−6−2−1012 x 10−7Input Currents (A)Transient response of the receiver (2K−1)’s inputs and output  0 0.2 0.4 0.6 0.8 1x 10−61.61.651.71.751.81.85Time (s)Output Voltage (V)  Input1Input2RX (2K−1)OutputFigure 6.8: Post-layout simulation results of the RX2k−1 receiver: transientresponse of the in-phase sinusoidal input currents and the output voltageat 4.35 MHz.without considering the buffer stage.To show the functionality of the circuit for a hypothetical air mediumreception where the signals are underdamped, i.e., having a sinusoidal shape,Figure 6.8 illustrates the receiver’s transient response for the inputs withfrequency of 4.35 MHz and 125 nA amplitude.Figure 6.9 shows the receiver gain and input impedance versus frequency fortwo different gain control voltage and bias settings. The flexibility ingain-bandwidth tuning results in a minimum gain of 105 dBΩ with 14 MHz of3 dB bandwidth for the gain control voltage of 0 V and Vb oa of 2.2 V. Also, the117105 106 107 108 109−50050100Trans−impedance gain (dB Ω)Frequency response of the trans−impedance gain and input impedance of the (2K−1) receiver in two different modes  105 106 107 108 10900.511.522.5Frequency (Hz)Input impedance (kΩ)Gain = 111.2 dBΩ, BW= 8.5MHzGain= 105.1 dBΩ, BW= 14MHzFigure 6.9: Post-layout simulation results: frequency response of the RX2k−1gain and input impedance for two settings of the gain and bias voltages.Setting 1 (minimum gain): gain control voltage = 0 V, bias voltage= 2.2 V. Setting 2 (maximum gain): gain control voltage = 1 V, biasvoltage = 1.85 V.maximum gain of 111.2 dBΩ with 8.5 MHz of 3 dB bandwidth is achieved forgain control voltage of 1 V and Vb oa of 1.85 V. This figure also shows the inputimpedance versus frequency which has an average of 300 Ω at the operationfrequency of 1.38 MHz. This value is lower than those reported in the recentworks [1, 28, 33, 96].1180 0.5 1 1.5 2x 10−6−2−1012 x 10−7Input Currents (A)Transient response of the receiver (2K)’s inputs and output  0 0.5 1 1.5 2x 10−61.41.451.5Time (s)Output Voltage (V)  Input3Input4RX (2K)OutputFigure 6.10: Post-layout simulation results of the RX2k receiver: transientresponse of the out-of-phase sinusoidal input currents and the outputvoltage at 1.55 MHz.6.4.3 RX2k Post-Layout Simulation ResultsFor verification of the receiver RX2k circuit functionality, two 1.55 MHz out-of-phase sinusoidal current sources with 125 nA amplitude are applied to the inputsof the circuit. The selection of the frequency and shape of the signal is becausethis receiver is designed to receive signals from the CMUT even rows’ second-harmonic (asymmetric) mode vibrations in immersion. As the harmonic vibrationin immersion underdamps the signals, they will be received in the sinusoidal form,and they are 180◦ out-of-phase.The bias and gain control settings are selected the same as the RX2k−1 receiver.119105 106 107 108 109−50050100150Trans−impedance gain (dB Ω)Frequency response of the trans−impedance gain and input impedance of the (2K) receiver in two different modes  105 106 107 108 10900.511.52Frequency (Hz)Input impedance (kΩ)  Gain = 114.2 dBΩ, BW= 5.7MHzGain= 107.1 dBΩ, BW= 9MHzFigure 6.11: Post-layout simulation results: frequency response of the RX2kgain and input impedance for two settings of the gain and bias voltages.Setting 1 (minimum gain): gain control voltage = 0 V, bias voltage =2.2 V. Setting 2 (maximum gain): gain control voltage = 1 V, biasvoltage = 1.85 V.Figure 6.10 shows the transient response of this receiver with about 107 dBΩ oftransimpedance gain.The receiver gain and input impedance versus frequency for two different gaincontrol voltage and bias settings are shown in Figure 6.11. The minimum gain is107.1 dBΩ with 9 MHz of 3 dB bandwidth for the gain control voltage of 0 V andVb oa of 2.2 V. The maximum gain achieved for gain control voltage of 1 V andVb oaof 1.85 V is 114.2 dBΩ with 5.7 MHz of 3 dB bandwidth. This figure also shows120(a) (b)Figure 6.12: (a) Transceiver chip micrograph and (b) Fabricated andassembled PCB with both transceiver and CMUT packages.an average input impedance of 300 Ω at the operation frequency of 1.55 MHz.The power consumption for an average gain setting is 1.9 mW excluding the bufferstage.6.5 Experimental ResultsFigure 6.12a shows the micrograph of a fabricated chip with the package wire-bonds attached to the pads. Also, Figure 6.12b shows the fabricated and assembledPCB including the transceiver and CMUT packages. To test the transceiver chipfunctionality, DC bias and supplies are provided using three Xantrex LXQ 30-2DC power supplies. The current draw from the power supply is monitored usingan HP 34401A multimeter and different node voltages are probed using a Rohde& Schwarz RTM 1054 oscilloscope.To verify the functionality of the receiver circuits, a The´venin equivalent ofa current source is implemented on the PCB, using a voltage source and a largeresistance of 2 MΩ. Also, the 50 Ω load for the receivers (the nominal input121impedance of the ultrasound machine) is provided using 50 Ω terminations on thePCB.In order to measure the frequency response of the receivers, we cannot simplyuse spectrum or vector network analyzers as their 50 Ω input impedance as wellas their ac-coupling capacitor deteriorates the receiver circuit performance.Instead, the input frequency is swept and the transient response of the output ismeasured for each step of the input frequency using an oscilloscope (Rohde &Schwarz RTM 1054). To provide the required in-phase and out-of-phase inputs,two methods have been used. The first method is using a Tee connector to split theoutput of a single channel Agilent 33250A function generator into two in-phasesignals and using a Mini-Circuits ZFSCJ-2-1 splitter to generate two out-of-phasesignals. The advantage of using this function generator is that its frequency rangeis up to 80 MHz, and the frequency response can be measured for about a decadehigher than the 3 dB bandwidth of the circuit. This gives a better insight into thefrequency response of the receiver. The issue with this method is that the availableMini-Circuits splitter has a lower 3 dB bandwidth of 1 MHz which degrades theinput signals of interest (1.55 MHz inputs are close to this value). The secondmethod is using a two-channel function generator which its outputs can bephase-shifted. The available equipment is a Rigol DG1022 function/arbitrarywaveform generator with a 20 MHz bandwidth for sinusoidal outputs. We haveused this waveform generator for frequencies of up to 20 MHz and the Tee andsplitter method for frequencies above 20 MHz. However, for frequencies above25 MHz, the output signal is too low (<3 mV) and could not be reliably read onthe oscilloscope.1221 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8−10010203040506070Time (µs)Voltage Amplitude (V)Transmitter’s input and output transient response  TX Output (60 V)TX Output (30 V)TX InputFigure 6.13: Experimental transient response of the high voltage transmitterinput and output voltages, the high voltage supply is set to two valuesof 30 V and 60 V and the input frequency is 1.38 MHz.6.5.1 High-Voltage Transmitter’s Experimental ResultsFor verifying the transmitter circuit performance, the low-voltage supply is set to3.3 V, and the high-voltage supply is swept from 10 V to 60 V. A 1.38 MHz, 3.3 Vpulse is applied to the input and the output is measured with oscilloscope probes.Figure 6.13 shows the input and two sample outputs of the transmitter for high-voltage supply of 30 V and 60 V. The 30 V output pulse has a rise time of 68 ns,fall time of 65 ns and 34 ns delay (from input rising edge to output rising edge).The input signal has 5 ns rise and fall times. The overall power consumption ofthe circuit for a 30 V high DC source and a 3.3 V low DC supply is 98.1 mW. Asshown in this figure, the slew rates for the 60 V output are very close to those ofthe 30 V output.1231.2 1.4 1.6 1.8 2 2.2 2.4−0.4−0.3−0.2−0.100.10.20.30.4Time (µs)Voltage Amplitude (V)Transient response of the receiver (2K−1)’s inputs and output  Input1Input2RX(2K−1) OutputFigure 6.14: Experimental transient response of the RX2k−1 receiver. The twoinput voltages are in-phase pulses with frequency of 1.38 MHz and500 mV peak-to-peak amplitude. The output is AC coupled on theoscilloscope.6.5.2 Receivers RX2k−1 and RX2k Experimental ResultsAs discussed at the beginning of this section, we have used a The´venin equivalentcircuit to replicate the input current sources. Also, the input signal frequency isswept and the output voltage is measured using an oscilloscope in its highimpedance input mode. Figure 6.14 shows a sample input and output voltagetransient response for the RX2k−1. As it is shown, the two inputs are in-phase.Also, Figure 6.15 shows the transient response for RX2k where the inputs are outof phase.To calculate the transimpedance gain versus frequency, first the voltage gain ismeasured by manually sweeping the input frequency by 500 kHz steps and readingthe output amplitude. Then, the voltage gain graph in decibel scale, derived bythese measurements, is added to the input impedance graphs (in decibel scale) of1241 1.2 1.4 1.6 1.8 2 2.2−0.2−0.15−0.1−0.0500.050.10.15Time (µs)Voltage Amplitude (V)Transient response of the receiver (2K)’s inputs and output  Input3Input4RX(2K)OutputFigure 6.15: Experimental transient response of the RX2k receiver. The twoinput voltages are out-of-phase sinusoidal signals with frequency of1.55 MHz and 300 mV peak-to-peak amplitude. The output is ACcoupled on the oscilloscope.Figure 6.9 and Figure 6.11. The resulting graphs for maximum gain settings isshown in Figure 6.16 and compared to simulation results. As it is shown in thisfigure, the measurement results are in good agreement with the simulation results.The power consumption for RX2k−1 and RX2k is 2.1 mW and 1.99 mWrespectively, excluding the output buffers. These values are for the average gainsettings and are in good agreement with the simulation results. The powerconsumption per channel for the complete 4 input/2 output receiver circuit is1.02 mW excluding the output buffers and 58.8 mW with the buffers.6.6 ConclusionIn this chapter, a high-voltage transmitter and a multi-channel low-power,variable-gain receiver circuit are presented which are designed and laid out in the125106 10795100105110115Frequency (Hz)Trans−impedance gain (dB Ω)Trans−impedance gain of the two receivers   Receiver (2K) gain − SimulationReceiver (2K) gain − MeasurementReceiver (2K−1) gain − SimulationReceiver (2K−1) gain − MeasurementFigure 6.16: Comparison of the simulated and measured frequency responsesof the two receivers’ transimpedance gain.Table 6.1: Comparison of the experimental results of the transimpedancereceiver with previously reported designsParameter [9]† [33] [84] [96]† This work†Technology (µm) 1.5 0.8 0.18 0.6 0.35Bandwidth (MHz) 10 11 103-256 860 9Gain (dBΩ) 113 75 84-109 58 110Input Impedance (kΩ) 4.3 - 0.14 0.5-0.9 0.3Power consumption (mW) 4 2 8.7 85∗ 1.02‡Chip area (mm2) - 0.01 0.24 0.1 0.06† Measurement result ∗ Including the output buffers‡ 58.8 mW including the output buffersNote: Reported power consumption and area are per receive channelAMS 0.35 µm high-voltage technology. The transceiver is intended forsuper-resolution ultrasound imaging with CMUT sensors. The high-voltagetransmitter is capable of providing up to 60 V pulses at its output whilemaintaining a reasonably small area comparing with state-of-the-art works. Thereceiver is capable of simultaneously processing both the in- and out-of-phase126Table 6.2: Comparison of the experimental results of the high-voltagetransmitter with previously reported designsParameter [85] [86] Chapter 3 (LS†) Chapter 3 (CC) This workTechnology (µm) 0.18 0.8 0.8 0.8 0.35Input voltage (V) 1.8 5 5 5 3.3Output voltage (V) 9.8-12.8 59 20 20 30Max. output voltage (V) 12.8 59 125 125 60Output load (pF) 15 20 18 18 18Frequency (MHz) 1.25 5 2.7 2.7 1.38Rise/fall time (ns) 40/50 69/58 37/59 84/80 68/65Input/output delay (ns) 22.5 - 24.5 77 34Power consumption 19.9mA dynamic0.43mA static ∗200mAdynamic ∗59 mW 26.4 mW 98.1 mWChip area (mm2) 0.022 0.85 0.125 0.135 0.08∗Simulation result †LS: Level-shifted CC:Cross-coupledNote: Reported power consumption and area are per transmit channelcurrent signals. The receiver has an average gain of 110 dBΩ with a 3 dBbandwidth of 9 MHz and consumes an average power of 1.02 mW per receivechannel. The input impedance of each receive channel is less than 300 Ω at theoperating frequency of less than 5 MHz. The transceiver experimental test resultsare shown to compare favorably with that of the state-of-the-art.Table 6.1 summarizes the performance of the receiver circuit and compares itwith related recent works. As it is shown in the table, the low power consumptionand input impedance of the circuit is noticeable. Table 6.2 compares thetransmitter’s performance to the similar works in literature. Note that comparingwith [86], and taking into account the technology scaling, the proposed chip areais still significantly smaller. Also, in comparison with [85] which has the closestoperating frequency, this work provides about five times larger maximum outputvoltage.127Chapter 7Conclusion and Future Work7.1 Research ContributionsThe focus of this thesis is on some of the system and circuit level challengesassociated with the integration of the capacitive micromachined ultrasonictransducers and the front end actuation and readout circuits. Completetransmitter/receiver front-end system and its interfaces with both CMUT andultrasound machine are studied and different circuit blocks and TX/RX controland enable switches are designed, co-simulated with the CMUT model andfabricated. A summary of these studies as well as the simulation and experimentalresults have been published in conference proceedings [1], [2] and [3] and aresubmitted for journal publications as well.• In Chapter 2, a bi-directional Verilog-AMS CMUT model is developedwhich can be co-simulated with the analog front-end circuits. This not onlyprovides a more realistic load/source model for the transmitter and receiver,but also facilitates a fast CMUT performance analysis in conjunction with128the CMOS electronics. This fast analysis alleviates the design optimizationiterations. Furthermore, as the model is bi-directional, it provides a veryconvenient tool for measuring and optimizing the CMUT receivesensitivity. Such modeling reduces the risks of post fabrication failures andincreases the time-to-market cycle. Experimental test results confirmreasonable matching to co-simulation results.• Two different high-voltage transmitters, the so-called level-shifted andcross-coupled transmitters, have been designed and co-simulated with theVerilog-AMS model of the CMUT in Chapter 3 and Chapter 4. Differenttechniques have been used to decrease the power consumption of thelevel-shifted transmitter, including reducing the voltage level of theactuation pulses from the ultrasound machine by using level-down shiftersand adjusting the high voltage transistor sizes while maintaining areasonable output switching speed. The measurement results of the twodesigns show comparable or better overall performance comparing tosimilar works in literature. The level-shifted transmitter is integrated withthe CMUT chip on a printed circuit board and the test results in air mediummatch the simulation results. A full transmit/receive and CMUT transducersystem is also co-simulated and proved perfectly functional.• An adaptive biasing technique is proposed in Chapter 5 that improves theCMUT receive sensitivity, resulting in less signal loss. In the simulations ofthis chapter, the CMUT’s bi-directional Verilog-AMS model plays asignificant role in showing the effect of the bias adjustments in increasingthe output signal level. To provide different bias voltage levels, a multiple129stage high-voltage DC-DC up-converter is designed, fabricated and testedsuccessfully.• In Chapter 6, a complete transceiver design for super-resolution imaging isproposed. The transmitters of this design are level-shifted high-voltagedrivers and the receivers are variable-gain transimpedance amplifiers thatadd and amplify the in-phase current signals and differentially amplify theout-of-phase signals. The two amplifiers’ outputs are sent to the ultrasoundmachine for further processing and display. The circuits’ full functionalityis proved by very good matching of the post-layout simulation andmeasurement results.7.2 Future WorkFuture directions for expanding the current research include:• As mentioned in Chapter 2, the CMUT model can be modified by fitting thedata from experiments into the model. This includes the extraction of theparasitic impedances at the electrical and mechanical sides.• Testing the integrated CMOS-CMUT system while CMUT is working inimmersion needs to be considered. Currently, due to the etching holes onthe CMUT membrane, the oil may leak into the CMUT cavity. Possiblecombination of oil and air inside the cavity makes the experimentsunreliable and unpredictable. On-going effort by another student is beingcarried out to cover the etching holes using an aerosol jet printer. When theCMUT is ready to be immersed in oil, a test environment may be builtwhich consists of a set of targets inside the oil medium to reflect back the130CMUT transmitted signals. The echoes will be received by CMUT andprocessed by the receiving amplifier. The receive sensitivity improvementby bias adjustment, presented in Chapter 5, and the concept of thesuper-resolution imaging of Chapter 6, can also be tested when the receiverhas real input signals from the CMUT transducers.• To complete the adaptive biasing unit of Chapter 5, design of a high-voltageregulator and a high-voltage multiplexer is proposed. The voltage regulatorhelps in reducing the output voltage ripples of the DC-DC converter. Themultiplexer, together with a programmable input selection scheme, can beused in order to select the desired high voltage output of the aforementionedvoltage regulator.• Design details of the digital control unit of Chapter 5 needs to be elaborated.Different input conditions of the adaptive biasing unit and its correspondingoutput requirements should be verified. Then, the counter block and thedigital logic units can be designed, simulated and fabricated.• Provided that a wire-bonding machine which meets our pad sizerequirements is accessible, wire-bonding the CMUT and CMOS chips maybe considered. Currently the two chips are packaged individually andmounted on a PCB. The parasitic capacitances of the packages and the PCBtraces degrades the SNR in the receive mode and reduces the transmitter’sbandwidth in the transmit mode. Wire-bonding can eliminate these negativeeffects.• Since no Verilog-AMS model or an equivalent circuit model for the CMUT131of Chapter 6 was available, a fixed value capacitor was used as thetransmitter’s load and an ideal current source in parallel with a fixed valuecapacitor was used at the receive mode of operation in the post-layoutsimulations. Developing a Verilog-AMS model for this particular CMUTand co-simulation with the analog front-end is suggested.132Bibliography[1] P. Behnamfar and S. Mirabbasi, “Design of a high-voltage analog front-endcircuit for integration with cmut arrays,” in Biomedical Circuits andSystems Conference (BioCAS), 2010 IEEE, 2010, pp. 298–301. → pagesiv, 1, 7, 16, 44, 107, 110, 111, 113, 118, 128[2] ——, “A cmut read-out circuit with improved receive sensitivity using anadaptive biasing technique,” in Biomedical Circuits and SystemsConference (BioCAS), 2011 IEEE, 2011, pp. 397–400. → pages v, 1, 88,128[3] P. Behnamfar, R. Molavi, and S. Mirabbasi, “Receiver design forcmut-based super-resolution ultrasound imaging,” in InternationalSymposium on Circuits and Systems (ISCAS), 2014 IEEE, 2014, pp.878–881. → pages v, 7, 26, 103, 128[4] J. del Cura, P. Segui, and C. Nicolau, Learning Ultrasound Imaging.Springer, 2012. → pages 1[5] T. Szabo, Diagnostic Ultrasound Imaging: Inside Out. Elsevier Inc, 2014.→ pages 1, 2[6] J.-J. Hwang, J. Quistgaard, J. Souquet, and L. A. Crum, “Portableultrasound device for battlefield trauma,” in Proceedings of the IEEEUltrasonics Symposium, 1998, vol. 2. IEEE, 1998, pp. 1663–1667. →pages 1[7] B. Khuri-Yakub, O. Oralkan, and M. Kupnik, “Next-Gen ultrasound,” IEEESpectrum, vol. 46, no. 5, pp. 44–46, 2009. → pages 2, 6, 22[8] S. Vaithilingam, T. Ma, Y. Furukawa, I. Wygant et al., “Three-dimensionalphotoacoustic imaging using a two-dimensional CMUT array,” IEEETransactions on Ultrasonics, Ferroelectrics, and Frequency Control,vol. 56, no. 11, pp. 2411–2419, 2009. → pages 5, 7133[9] I. Wygant, X. Zhuang, D. Yeh, O. Oralkan et al., “Integration of 2D CMUTarrays with front-end electronics for volumetric ultrasound imaging,” IEEETransactions on Ultrasonics, Ferroelectrics, and Frequency Control,vol. 55, no. 2, pp. 327–342, 2008. → pages 16, 47, 52, 63, 68, 126[10] S. Wong, M. Kupnik, K. Butts-Pauly, and B. Khuri-Yakub, “Advantages ofcapacitive micromachined ultrasonics transducers (CMUTs) for highintensity focused ultrasound (HIFU),” in Proc. IEEE UltrasoundSymposium, 2007, pp. 1313–1316. → pages 2, 7, 11[11] D. Chen, S. Sharma, and A. Mudhoo, Handbook on Applications ofUltrasound: Sonochemistry for Sustainability. Taylor & Francis, 2011.[Online]. Available: http://books.google.ca/books?id=nU2eVc9SozgC →pages 2[12] J. Cheeke, Fundamentals and Applications of Ultrasonic Waves, SecondEdition. Taylor & Francis, 2012. [Online]. Available:http://books.google.ca/books?id=p5kgCN-waMsC → pages 2[13] F. Capote and D. de Castro, Analytical Applications of Ultrasound, ser.Techniques and Instrumentation in Analytical Chemistry. ElsevierScience, 2007. [Online]. Available:http://books.google.ca/books?id=IQn1rlshrYMC → pages 2[14] V. Hans and E. von Lavante, “Vortex-shedding flow metering usingultrasound,” in Fluid Mechanics of Flow Metering. Springer, 2005, pp.95–110. → pages 2[15] C. Chen, Ultrasonic and Advanced Methods for Nondestructive Testing andMaterial Characterization. World Scientific, 2007. [Online]. Available:http://books.google.ca/books?id=lKUaF6RWR5sC → pages 2[16] J. Woo, “A short history of the development of ultrasound in obstetrics andgynecology,” http://www.ob-ultrasound.net/history1.html. → pages 3[17] “Siemens’s history,”http://www.siemens.com/history/en/innovations/healthcare.htm#toc-5. →pages 3[18] “Ultrasonix medical corporation website, clinical gallery,”http://www.ultrasonix.com/applications/ob-gyn/clinical-gallery. → pagesxii, 3134[19] “Hi Mom 3D/4D ultrasound scan facility,” http://www.himom.co.uk/. →pages xii, 3[20] G. A. Alba, D. A. Kelmenson, V. E. Noble, A. F. Murray, and P. F. Currier,“Faculty staff-guided versus self-guided ultrasound training for internalmedicine residents,” Medical Education, vol. 47, no. 11, pp. 1099–1108,2013. [Online]. Available: http://dx.doi.org/10.1111/medu.12259 → pages3[21] “U.S. food and drug administration, ulrasound imaging,”http://www.fda.gov/Radiation-EmittingProducts/RadiationEmittingProductsandProcedures/MedicalImaging/ucm115357.htm#rb. → pages 3[22] F. Kuijpers, “The role of technology in future medical imaging,”Medicamundi, vol. 40, pp. 181–189, 1995. → pages 3[23] “Philips CX-50 compactxtreme ultrasound system,” http://www.healthcare.philips.com/sg en/products/ultrasound/systems/cx50/. →pages 4[24] “Low cost ultrasound,”http://www.philipsafricaroadshow.com/causes/womens-health-2/. → pages4[25] “Wikisonix, ultrasonix medical corporation wiki,”http://www.ultrasonix.com/wikisonix/index.php/Sonix RP. → pages xii, 4[26] “Ultrasonix medical corporation transducers,”http://www.ultrasonix.com/products/transducers. → pages xii, 4[27] R. Guldiken, J. McLean, and F. Degertekin, “CMUTs with dual-electrodestructure for improved transmit and receive performance,” IEEETransactions on Ultrasonics, Ferroelectrics, and Frequency Control,vol. 53, no. 2, p. 483, 2006. → pages 5, 7, 104[28] I. Wygant, N. Jamal, H. Lee, A. Nikoozadeh et al., “An integrated circuitwith transmit beamforming flip-chip bonded to a 2-D CMUT array for 3-Dultrasound imaging,” IEEE Transactions on Ultrasonics, Ferroelectrics,and Frequency Control, vol. 56, no. 10, pp. 2145–2156, 2009. → pages 5,13, 26, 44, 55, 63, 91, 104, 105, 113, 118135[29] A. Ergun, G. Yaralioglu, and B. Khuri-Yakub, “Capacitive micromachinedultrasonic transducers: Theory and technology,” Journal of AerospaceEngineering, vol. 16, p. 76, 2003. → pages 6, 7, 9, 10, 11, 26, 29, 89[30] F. Hunt, Electroacoustics: the analysis of transduction, and its historicalbackground, ser. Harvard monographs in applied science. HarvardUniversity Press, 1954. [Online]. Available:http://books.google.ca/books?id=AVvAAAAAIAAJ → pages 7[31] M. Haller and B. Khuri-Yakub, “A surface micromachined electrostaticultrasonic air transducer,” Ultrasonics, Ferroelectrics and FrequencyControl, IEEE Transactions on, vol. 43, no. 1, pp. 1–6, 1996. → pages 7,23[32] “Khuri-yakub ultrasonics group,” http://www-kyg.stanford.edu. → pages 7[33] I. Cicek, A. Bozkurt, and M. Karaman, “Design of a front-end integratedcircuit for 3D acoustic imaging using 2D CMUT arrays,” IEEETransactions on Ultrasonics, Ferroelectrics, and Frequency Control,vol. 52, no. 12, pp. 2235–2241, 2005. → pages 7, 13, 16, 26, 44, 63, 68,91, 104, 105, 113, 118, 126[34] R. Noble, R. Davies, D. King, M. Day et al., “Low-temperaturemicromachined cMUTs with fully-integrated analogue front-endelectronics,” in IEEE Ultrasonics Symposium, 2002, pp. 1045–1050. →pages 13[35] W. You, E. Cretu, and R. Rohling, “Super-resolution imaging using multi-electrode cmuts: theoretical design and simulation using point targets,”Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on,vol. 60, no. 11, pp. 2295–2309, November 2013. → pages 7, 10, 29, 104,105[36] S. Sano, A. Sako, T. Kobayashi, M. Izumi et al., “Ultrasonic probe andmethod for manufacturing the same and ultrasonic diagnostic device,”Sep. 26 2008, WO Patent 2,008,114,582. → pages 7[37] A. Nguyen-Dinh, N. Felix, A. Flesch et al., “Electrostatic membranes forsensors, ultrasonic transducers incorporating such membranes, andmanufacturing methods therefor,” Feb. 10 2009, US Patent 7,489,593. →pages 7[38] Y. Huang, “Acoustic decoupling in cmuts,” Jul. 20 2010, US Patent7,759,839. → pages 7136[39] D. Lin, Interface Engineering of Capacitive Micromachined UltrasonicTransducers for Medical Applications. Stanford University, 2011. →pages 8, 11[40] A. Caronti, G. Caliano, A. Iula, and M. Pappalardo, “An accurate model forcapacitive micromachined ultrasonic transducers,” Ultrasonics,Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 49,no. 2, pp. 159–168, Feb 2002. → pages 8, 104[41] M. Cai, A PolyMUMPs Capacitive Micromachined Ultrasonic Transducer.University of British Columbia, 2011. → pages 9, 27, 29, 32, 35, 49, 72,80, 89, 109[42] A. Nikoozadeh, B. Bayram, G. Yaralioglu, and B. Khuri-Yakub,“Analytical calculation of collapse voltage of cmut membrane [capacitivemicromachined ultrasonic transducers],” in Ultrasonics Symposium, 2004IEEE, vol. 1, Aug 2004, pp. 256–259 Vol.1. → pages 10, 23[43] S. Frew, H. Najar, and E. Cretu, “VHDL-AMS behavioural modelling of aCMUT element,” in IEEE Behavioral Modeling and Simulation Workshop,2009, pp. 19–24. → pages xviii, 10, 24, 27, 34, 89[44] S. Wong, M. Kupnik, R. Watkins, K. Butts-Pauly, and B. Khuri-Yakub,“Capacitive micromachined ultrasonic transducers for therapeuticultrasound applications,” IEEE Transactions on Biomedical Engineering,vol. 57, no. 1, pp. 114 –123, Jan. 2010. → pages 11, 22, 103, 105[45] S. Olcum, F. Y. Yamaner, A. Bozkurt, and A. Atalar, “Deep-collapseoperation of capacitive micromachined ultrasonic transducers,”Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on,vol. 58, no. 11, pp. 2475–2483, 2011. → pages 11[46] S. Olcum, F. Yamaner, A. Bozkurt, H. Koymen, and A. Atalar, “Anequivalent circuit model for transmitting capacitive micromachinedultrasonic transducers in collapse mode,” Ultrasonics, Ferroelectrics andFrequency Control, IEEE Transactions on, vol. 58, no. 7, pp. 1468–1477,2011. → pages[47] O. Oralkan, B. Bayram, G. G. Yaralioglu, A. S. Ergun, M. Kupnik, D. T.Yeh, I. O. Wygant, and B. T. Khuri-Yakub, “Experimental characterizationof collapse-mode cmut operation,” Ultrasonics, Ferroelectrics andFrequency Control, IEEE Transactions on, vol. 53, no. 8, pp. 1513–1523,2006. → pages 11137[48] S. Bhansali and A. Vasudev, Mems for Biomedical Applications, ser.Woodhead Publishing Series in Biomaterials. Elsevier Science, 2012.[Online]. Available: http://books.google.ca/books?id=8s5ZAgAAQBAJ →pages 11[49] S. Berg and A. Ronnekleiv, “Reducing fluid coupled crosstalk betweenmembranes in cmut arrays by introducing a lossy top layer,” in IEEEUltrasonics Symposium. IEEE, 2006, pp. 594–597. → pages 12[50] ——, “Reduction of crosstalk in cmut arrays by introducing doubleperiodicities,” in Ultrasonics Symposium, 2007. IEEE, Oct 2007, pp.2155–2158. → pages 12[51] S. Zhou and J. A. Hossack, “Reducing inter-element acoustic crosstalk incapacitive micromachined ultrasound transducers,” Ultrasonics,Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 54,no. 6, pp. 1217–1228, 2007. → pages 12[52] M.-C. Ho, M. Kupnik, K. K. Park, K. Eckhoff, and B. Khuri-Yakub, “Widepressure range operation of air-coupled cmuts,” in Ultrasonics Symposium(IUS), 2012 IEEE International, Oct 2012, pp. 93–96. → pages 12[53] M. S. Salim, M. A. Malek, R. Heng, K. Juni, and N. Sabri, “Capacitivemicromachined ultrasonic transducers: Technology and application,”Journal of Medical Ultrasound, vol. 20, no. 1, pp. 8 – 31, 2012. → pages12[54] B. T. Khuri-Yakub and O¨. Oralkan, “Capacitive micromachined ultrasonictransducers for medical imaging and therapy,” Journal of micromechanicsand microengineering, vol. 21, no. 5, p. 054004, 2011. → pages 12, 13, 18,103, 113[55] G. Gurun, C. Tekes, J. Zahorian, T. Xu, S. Satir, M. Karaman, J. Hasler,and F. Degertekin, “Single-chip CMUT-on-CMOS front-end system forreal-time volumetric IVUS and ICE imaging,” Ultrasonics, Ferroelectricsand Frequency Control, IEEE Transactions on, vol. 61, no. 2, pp. 239–250,February 2014. → pages 12, 103, 105[56] S. Wong, M. Kupnik, K. Butts-Pauly, and B. Khuri-Yakub, “Advantages ofcapacitive micromachined ultrasonics transducers (cmuts) for highintensity focused ultrasound (HIFU),” in Ultrasonics Symposium, 2007.IEEE, Oct 2007, pp. 1313–1316. → pages 13, 18138[57] M. D. Menz, O¨. Oralkan, P. T. Khuri-Yakub, and S. A. Baccus, “Preciseneural stimulation in the retina using focused ultrasound,” The Journal ofNeuroscience, vol. 33, no. 10, pp. 4550–4560, 2013. → pages 13[58] D. Christensen, Ultrasonic Bioinstrumentation. John Wiley & Sons Inc,1988. → pages 13, 90[59] R. Noble, R. Davies, M. Day, L. Koker et al., “Cost-effective andmanufacturable route to the fabrication of high-density 2D micromachinedultrasonic transducer arrays and (CMOS) signal conditioning electronics onthe same silicon substrate,” in IEEE Ultrasonics Symposium, vol. 2, 2001,pp. 941–944. → pages 13[60] W. You, Exploring multiple-mode vibrations of capacitive micromachinedultrasonic transducers (CMUTs). University of British Columbia, 2013.→ pages xiii, 16, 24, 25, 27, 28, 31, 32, 33, 34, 40, 77, 85[61] “Teledyne Dalsa,” http://www.teledynedalsa.com/corp/. → pages 17[62] “Austria micro systems (AMS),” http://asic.ams.com/. → pages 17[63] A. Bhuyan, J. W. Choe, B. C. Lee, I. Wygant, A. Nikoozadeh, O. Oralkan,and B. T. Khuri-Yakub, “Integrated circuits for volumetric ultrasoundimaging with 2-d cmut arrays,” Biomedical Circuits and Systems, IEEETransactions on, vol. 7, no. 6, pp. 796–804, 2013. → pages 18, 103[64] S.-J. Jung, J.-K. Song, and O.-K. Kwon, “Three-side buttable integratedultrasound chip with a 16x16 reconfigurable transceiver and capacitivemicromachined ultrasonic transducer array for 3-d ultrasound imagingsystems,” Electron Devices, IEEE Transactions on, vol. 60, no. 10, pp.3562–3569, Oct 2013. → pages 47, 105[65] S. Kothapalli, T.-J. Ma, S. Vaithilingam, O. Oralkan, B. T. Khuri-Yakub,and S. Sam Gambhir, “Deep tissue photoacoustic imaging using aminiaturized 2-d capacitive micromachined ultrasonic transducer array,”Biomedical Engineering, IEEE Transactions on, vol. 59, no. 5, pp.1199–1204, 2012. → pages[66] M. Wang and J. Chen, “Volumetric flow measurement using an implantablecmut array,” Biomedical Circuits and Systems, IEEE Transactions on,vol. 5, no. 3, pp. 214–222, 2011. → pages 103139[67] X. Cheng, J. Chen, and C. Li, “A miniature capacitive micromachinedultrasonic transducer array for minimally invasive photoacoustic imaging,”Microelectromechanical Systems, Journal of, vol. 19, no. 4, pp. 1002–1011,2010. → pages 18[68] N. Senegond, A. Boulme, C. Plag, F. Teston, and D. Certon, “Fasttime-domain modeling of fluid-coupled cmut cells: from the single cell tothe 1-d linear array element,” Ultrasonics, Ferroelectrics and FrequencyControl, IEEE Transactions on, vol. 60, no. 7, pp. 1505–1518, July 2013.→ pages 23[69] P. C. Eccardt, P. Wagner, and S. Hansen, “Analytical models formicromachined transducers - an overview (invited),” in UltrasonicsSymposium, 2006. IEEE, Oct 2006, pp. 572–581. → pages 23, 24[70] G. Yaralioglu, A. Ergun, and B. Khuri-Yakub, “Finite-element analysis ofcapacitive micromachined ultrasonic transducers,” Ultrasonics,Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 52,no. 12, pp. 2185–2198, 2005. → pages 23, 25[71] A. Lohfink, P. C. Eccardt, W. Benecke, and H. Meixner, “Derivation of a 1dcmut model from fem results for linear and nonlinear equivalent circuitsimulation,” in IEEE Symposium on Ultrasonics, vol. 1, 2003, pp. 465–468.→ pages 25[72] C. Meynier, F. Teston, and D. Certon, “A multiscale model for array ofcapacitive micromachined ultrasonic transducers,” The Journal of theAcoustical Society of America, vol. 128, no. 5, pp. 2549–2561, 2010. →pages 23[73] H. A. Tilmans, “Equivalent circuit representation of electromechanicaltransducers: I. Lumped-parameter systems,” Journal of Micromechanicsand Microengineering, vol. 6, no. 1, p. 157, 1996. → pages 23[74] D. Certon, F. Teston, and F. Patat, “A finite difference model for cmutdevices,” Ultrasonics, Ferroelectrics and Frequency Control, IEEETransactions on, vol. 52, no. 12, pp. 2199–2210, Dec 2005. → pages 23[75] H. A. Tilmans, “Equivalent circuit representation of electromechanicaltransducers: II. Distributed-parameter systems,” Journal ofMicromechanics and Microengineering, vol. 7, no. 4, p. 285, 1997. →pages 23140[76] W. Mason, Electromechanical Transducers and Wave Filters. D. VanNostrand Co, 1948. → pages 25[77] G. Yaralioglu, M. Badi, A. Sanli Ergun, and B. T. Khuri-Yakub, “Improvedequivalent circuit and finite element method modeling of capacitivemicromachined ultrasonic transducers,” in Ultrasonics, 2003 IEEESymposium on, vol. 1, 2003, pp. 469–472 Vol.1. → pages 27[78] E. Cretu, Inertial MEMS Devices: Modeling, Design and Applications.Delft University of Technology, 2003. → pages 27, 29, 30[79] W. You, E. Cretu, R. Rohling, and M. Cai, “Tiltable ultrasonic transducers:Concept, beamforming methods and simulation,” Sensors Journal, IEEE,vol. 11, no. 10, pp. 2286–2300, 2011. → pages 29, 104, 105[80] W. You, E. Cretu, and R. Rohling, “Direction of arrival estimation usingasymmetric mode components of cmuts,” in Ultrasonics Symposium (IUS),2012 IEEE International. IEEE, 2012, pp. 1818–1821. → pages 29, 104,105[81] D. Giovanni, Flat and Corrugated Diaphragm Design Handbook. Taylor& Francis, 1982. → pages 32[82] W. Soedel, Vibrations of Shells and Plates, Third Edition, ser. DekkerMechanical Engineering. Marcel Dekker Inc., 2004. → pages 32[83] J. Johansson, M. Gustafsson, and J. Delsing, “Ultra-low powertransmit/receive ASIC for battery operated ultrasound measurementsystems,” Sensors and Actuators A: Physical, vol. 125, no. 2, pp. 317–328,2006. → pages 44[84] F. Nabki and M. El-Gamal, “A high gain-bandwidth producttransimpedance amplifier for MEMS-based oscillators,” in 34th EuropeanSolid-State Circuits Conference, 2008, pp. 454–457. → pages 44, 52, 63,68, 126[85] H. K. Cha, D. Zhao, J. H. Cheong, B. Guo, H. Yu, and M. Je, “A cmoshigh-voltage transmitter ic for ultrasound medical imaging applications,”Circuits and Systems II: Express Briefs, IEEE Transactions on, vol. 60,no. 6, pp. 316–320, June 2013. → pages 45, 68, 69, 127[86] R. Chebli and M. Sawan, “Fully integrated high-voltage front-end interfacefor ultrasonic sensing applications,” IEEE Transactions on Circuits and141Systems I: Regular Papers, vol. 54, no. 1, pp. 179–190, 2007. → pages 45,68, 91, 93, 127[87] H. Ballan and M. Declercq, High Voltage Devices And Circuits In StandardCMOS Technologies. Springer-Verlag New York, LLC, 1998. → pages 47[88] J. Doutreloigne, H. De Smet, and A. Van Calster, “A versatile micropowerhigh-voltage flat-panel display driver in a 100-V 0.7-µm CMOS intelligentinterface technology,” IEEE Journal of Solid-State Circuits, vol. 36, no. 12,pp. 2039–2048, 2001. → pages 47[89] D. Pan, H. Li, and B. Wilamowski, “A low voltage to high voltage levelshifter circuit for MEMS application,” in University Government IndustryMicroelectronics Symposium, 2003, pp. 128–132. → pages 47[90] B. Serneels, M. Steyaert, and W. Dehaene, “A high speed, low voltage tohigh voltage level shifter in standard 1.2 V 0.13 µm CMOS,” AnalogIntegrated Circuits and Signal Processing, vol. 55, no. 1, pp. 85–91, 2008.→ pages 48[91] P. Saul, K. Brunson, and R. Bunyan, “Versatile high voltage level shift anddriver for MEMS applications,” Electronics Letters, vol. 39, p. 185, 2003.→ pages 48[92] M. Khorasani, M. Behnam, L. van den Berg, C. Backhouse, and D. Elliott,“High-voltage cmos controller for microfluidics,” IEEE Transactions onBiomedical Circuits and Systems, vol. 3, no. 2, pp. 89–96, Apr. 2009. →pages[93] S. Tan and X. Sun, “Low power CMOS level shifters by bootstrappingtechnique,” Electronics Letters, vol. 38, no. 16, pp. 876–878, 2002. →pages 47[94] S. Peng, M. Qureshi, P. Hasler, A. Basu, and F. Degertekin, “Acharge-based low-power high-SNR capacitive sensing interface circuit,”IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 55,no. 7, pp. 1863–1872, 2008. → pages 52[95] E. Sackinger and W. Guggenbuhl, “A high-swing, high-impedance MOScascode circuit,” IEEE Journal of Solid-State Circuits, vol. 25, no. 1, pp.289 –298, Feb. 1990. → pages 52, 53142[96] S. Park and H. Yoo, “1.25-Gb/s regulated cascode CMOS transimpedanceamplifier for gigabit ethernet applications,” IEEE Journal of Solid-StateCircuits, vol. 39, no. 1, pp. 112–121, 2004. → pages 52, 53, 110, 111, 118,126[97] B. Rahmatian and S. Mirabbasi, “A low-power 75 dB digitallyprogrammable variable-gain amplifier in 0.18µm CMOS,” CanadianJournal of Electrical and Computer Engineering, vol. 32, no. 4, pp.181–186, 2007. → pages 52[98] P. van Lieshout and R. van de Plassche, “A monolithic wideband variablegain amplifier with a high gain range and low distortion,” in Digest ofTechnical Papers of the 43rd IEEE International Solid-State CircuitsConference, Feb. 1997, pp. 358–359, 485. → pages[99] J. Rijns, “CMOS low-distortion high-frequency variable-gain amplifier,”IEEE Journal of Solid-State Circuits, vol. 31, no. 7, pp. 1029 –1034, Jul.1996. → pages[100] D. Ma, C. Zheng, H. Chao, and M. Koen, “Integrated low-powerCMFB-free variable-gain amplifier for ultrasound diagnostic applications,”Analog Integrated Circuits and Signal Processing, vol. 61, no. 2, pp.171–179, 2009. → pages 52[101] Y. Huang, “Separate cMUTs for reception and transmission,” Apr. 4 2007,US Patent App. 11/696,664. → pages 55[102] M. Fuller, T. Blalock, J. Hossack, and W. Walker, “Novel transmitprotection scheme for ultrasound systems,” IEEE Transactions onUltrasonics, Ferroelectrics, and Frequency Control, vol. 54, no. 1, pp. 79–86, Jan. 2007. → pages 55[103] J. Camacho and C. Fritsch, “Protection circuits for ultrasoundapplications,” IEEE Transactions on Ultrasonics, Ferroelectrics, andFrequency Control, vol. 55, no. 5, pp. 1160 –1164, May. 2008. → pages[104] B. Dufort, T. Letavic, and S. Mukherjee, “Digitally controlled high-voltageanalog switch array for medical ultrasound applications in thin-layersilicon-on-insulator process,” in IEEE International SOI Conference, 2002,pp. 78–79. → pages 55[105] I. Ladabaum, X. Jin, H. Soh, A. Atalar, and B. Khuri-Yakub, “Surfacemicromachined capacitive ultrasonic transducers,” Ultrasonics,143Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 45,no. 3, pp. 678–690, 1998. → pages 89[106] D. Maksimovic and S. Dhar, “Switched-capacitor DC-DC converters forlow-power on-chip applications,” in 30th Annual IEEE Power ElectronicsSpecialists Conference, vol. 1, Aug. 1999, pp. 54–59. → pages 93[107] A. Ronnekleiv, “Cmut array modeling through free acoustic cmut modesand analysis of the fluid cmut interface through fourier transform methods,”Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on,vol. 52, no. 12, pp. 2173–2184, 2005. → pages 104[108] N. A. Hall, R. O. Guldiken, J. McLean, and F. L. Degertekin, “Modelingand design of cmuts using higher order vibration modes [capacitivemicromachined ultrasonic transducers],” in Ultrasonics Symposium, 2004IEEE, vol. 1. IEEE, 2004, pp. 260–263. → pages 104[109] K. T. Z. Oo, E. Mandelli, and W. Moses, “A high-speed low-noise16-channel csa with automatic leakage compensation in 0.35-µm cmosprocess for apd-based pet detectors,” Nuclear Science, IEEE Transactionson, vol. 54, no. 3, pp. 444–453, 2007. → pages 112144

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0135571/manifest

Comment

Related Items