CMOS INSTRUMENTATION FOR
ELECTROCHEMICAL BIOSENSOR ARRAY
MICROSYSTEMS
By
Xiaowen Liu

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Electrical Engineering-Doctor of Philosophy
2014

ABSTRACT
CMOS INSTRUMENTATION FOR ELECTROCHEMICAL
BIOSENSOR ARRAY MICROSYSTEMS
By
Xiaowen Liu
Understanding the structure and function of proteins has become increasingly
important since the completion of the Human Genome Project and the sequencing of
several other important genomes. In recent years, lab-on-a-chip systems have introduced
some new capabilities for protein analyses. Rapid progress in the field of microsystems,
miniaturized devices, combining sensor and electronics, facilitates a new generation of
miniaturized biosensor arrays utilizing silicon CMOS chips that acquire and process bioelectrochemical signals. Such biosensor array microsystems permit improved sensitivity,
measurement throughput and cost per assay. This thesis research addresses the design and
development of high performance electrochemical instrumentation circuits that enable
simultaneous characterization of multiple protein interfaces in a high throughput chipscale biosensor arrays microsystem. A CMOS amperometric biosensor readout circuit
with current resolution of 1pA has been developed. Furthermore, a new compact, low
power, 100fA current resolution impedance analysis circuit that utilizes mixed-mode
signal processing to extract real and imaginary impedance components has been
developed and tested with an on-chip protein interface. Moreover, to realize a multiprotein array of sensors with different instrumentation needs, a hardware efficient CMOS
electrochemical circuit has been designed to achieve both amperometric and
impedimetric measurement while sharing hardware source. Finally, by integrating an

understanding of both protein assays challenges and microelectronics design limitations,
a novel 1000-element array instrumentation architecture that permits rapid highthroughput characterization of membrane proteins with single protein resolution has been
designed. The CMOS instrumentation circuitry developed in this thesis research could
significantly advance proteomics research and progress in characterizing newly
sequenced genomes.

Dedicated to my family

iv

ACKNOWLEDGEMENTS
The work presented in this thesis would not have been possible without the support,
contribution, and encouragement of many individuals that I have had the privilege of
interacting with during my stay at Michigan State University. Indeed this is now a unique
opportunity to acknowledge them, even though in a very small way.
First, and foremost, I would like to express my deep gratitude to my advisor,
Professor Andrew Mason, for his generous support. His broad vision and encouragement
has always been a source of inspiration for me. His enthusiasm and wise criticism have
had a tremendous influence on my professional development, which I am deeply grateful
for.
I am also grateful to Professor Tim Hogan, Professor and Professor Evangelyn
Alocilja and Professor Prem Chahal for serving on my committees. They have always
been an exceptional source of support for me at MSU. Their technical feedback and
original ideas greatly expanded the scope of my research and its contributions.
I am greatly indebted to all my colleagues and friends at MSU for their coaching and
support in our joint entrepreneurial endeavors. Special thanks to Lin Li, Haitao, Li,
Yuning, Yang, Ying Liu, Bhushan Awate, Daniel Rairigh, Dr. Yang Chao, Dr. Huang,
Yue, and Dr. Xiaoyi Mu.
My utmost gratitude, love and affection belong to my family. My parents always
unconditionally love and support me and I thank them for everything they have given me.
I would like to thank my husband, Yang Liu, who accompanied me through the past
years at Michigan State University. This thesis is dedicated to them.

v

TABLE OF CONTENTS
LIST OF TABLES ........................................................................................................... ix
LIST OF FIGURES .......................................................................................................... x
Chapter 1 Introduction..................................................................................................... 1
1.1
Electrochemical Protein Biosensor ...................................................................... 1
1.2
Miniaturized Protein Array Biosensor ................................................................. 3
1.3
Challenges for CMOS Instrumentation ................................................................ 5
1.4
Dissertation Goals ................................................................................................ 7
1.5
Outline .................................................................................................................. 8
Chapter 2 Background ................................................................................................... 10
2.1 Electrochemical Biosensor Introduction ................................................................. 10
2.1.1 Electrochemical biosensor components............................................................ 10
2.1.2 Transducer ........................................................................................................ 10
2.1.3 Immobilization.................................................................................................. 11
2.2 Electrochemical Biosensor Techniques .................................................................. 12
2.2.1 Potentiometric biosensor .................................................................................. 12
2.2.2 Amperometric biosensors ................................................................................. 13
2.2.3 Impedimetric biosensor .................................................................................... 14
2.2.4 Ion-sensitive field effect transistor biosensor ................................................... 14
2.2.5 Electrode transducer ......................................................................................... 15
2.3 Interface Circuitry for Electrochemical Biosensors ................................................ 19
2.3.1 Potentiostat ....................................................................................................... 20
2.3.2 Readout circuits ................................................................................................ 23
2.4 DC Current Readout Circuitry ................................................................................ 24
2.4.1 Resistive current readout circuits ..................................................................... 24
2.4.2 Capacitive current readout circuits ................................................................... 25
2.4.3 Current conveyor current readout circuits ........................................................ 28
2.5 AC Current Readout Circuitry ................................................................................ 31
Chapter 3 CMOS Amperometric Potentiostat and Readout Circuits ....................... 39
3.1 Amperometric Instrumentation System Structure ................................................... 40
3.2 Potentiostat Circuit .................................................................................................. 41
3.3 Readout Circuit ....................................................................................................... 45
3.4 Test Results ............................................................................................................. 52
Chapter 4 CMOS Impedance Spectroscopy Instrumentation .................................... 62

vi

4.1 Impedance Spectroscopy Methods .......................................................................... 63
4.2 CMOS Implementation ........................................................................................... 66
4.2.1 Frequency synthesizer ...................................................................................... 67
4.2.2 Impedance readout circuit CMOS design ......................................................... 69
4.3 Test Results ............................................................................................................. 74
Chapter 5 Hardware Efficiency Multi-mode Electrochemical Readout Chip .......... 83
5.1 Design Goals ........................................................................................................... 83
5.2 Chip Architecture .................................................................................................... 84
5.3 Multi-mode Current Readout Approach ................................................................. 86
5.3.1 MCR noise model ............................................................................................. 88
5.3.2 Amperometric mode ......................................................................................... 90
5.3.3 Impedimetric mode ........................................................................................... 92
5.4 MERC CMOS Implementation ............................................................................... 93
5.4.1 Potentiostat ....................................................................................................... 93
5.4.2 MCR circuit ...................................................................................................... 94
5.5 Test Results ............................................................................................................. 96
5.5.1 Circuits characterization ................................................................................... 96
5.5.2 Electrochemical test results .............................................................................. 99
5.5.2.1 Materials .................................................................................................. 100
5.5.2.2 Test setup ................................................................................................. 101
5.5.2.3 Surface immobilization ............................................................................ 102
5.5.2.4 Measurement and results.......................................................................... 103
Chapter 6 CMOS Electrochemical Instrumentation Array for On-chip High
Throughput Characterization of Single Ion Channel Proteins ................................ 107
6.1 Single Ion Channel Detection: History and Background ...................................... 108
6.2 Single-Ion-Channel Array Microsystem ............................................................... 109
6.2.1 Pore-based pBLM and Electrochemical Methods .......................................... 109
6.2.2 Microfluidic structures for self-assembled pBLM ......................................... 111
6.2.3 High throughput microsystem approach......................................................... 112
6.3. Instrumentation Circuits Requirements and Challenges ...................................... 115
6.3.1 Multi-mode instrumentation circuits .............................................................. 115
6.3.2 Instrumentation circuits requirements and challenges.................................... 116
6.4. Instrumentation Array Approach ......................................................................... 117
6.5 CMOS Implementation ......................................................................................... 120
6.5.1 Preamplifier design ......................................................................................... 120
6.5.2 CPA/EIS blocks .............................................................................................. 123
6.6 Simulation Results................................................................................................. 123
6.6.1 Simulation condition....................................................................................... 123
6.6.2 Simulation results ........................................................................................... 124

vii

6.6.3 Power consumption analysis .......................................................................... 127
6.6.4 Area estimation ............................................................................................... 128
Chapter 7 Summary and Future Work ...................................................................... 130
7.1 Summary ............................................................................................................... 130
7.2 Future Work .......................................................................................................... 132
APPENDIX .................................................................................................................... 134
BIBLIOGRAPHY ......................................................................................................... 136

viii

LIST OF TABLES
Table 2-1. Types of bio-probes used in biosensors and their common electrochemical
technique. .......................................................................................................................... 13
Table 2-2. The performance summary of amperometric current readout circuitry. ......... 37
Table 2-3. The performance summary of impedimetric readout circuitry........................ 38
Table 3-1. Potentiostat characteristics. ............................................................................. 54
Table 3-2. Amperometric readout circuit characteristics. ................................................. 55
Table 3-3. Waveform generator characteristics ................................................................ 56
Table 4-1. The performance of impedance-to-digital converter circuit. ........................... 79
Table 5-1. The performance summary of the MERC chip. .............................................. 99
Table 6-1. The performance requirement for potentiostat and readout circuits for each
mode. ............................................................................................................................... 120
Table 6-2. Simulation parameters with Candece Spectre simulator. .............................. 124
Table 6-3. The power consumption estimation of the single ion channel detection
instrumentation circuits. .................................................................................................. 128
Table 6-4. Area estimation for each components in the single ion channel detection
instrumentation circuit with AMI 0.5um process. .......................................................... 129

ix

LIST OF FIGURES
Figure 1-1. Conceptual illustration of a CMOS electrochemical biosensor array
microsystem use in a liquid environment (For interpretation of the references to color in
this and all other figures, the reader is referred to the electronic version of this
dissertation). ........................................................................................................................ 6
Figure 2-1. Common components of electrochemical biosensor...................................... 11
Figure 2-2. The structure of ISFET biosensor. ................................................................. 15
Figure 2-3. The equivalent circuit model of (a) a two electrode system and (b) a three
electrode system. ............................................................................................................... 18
Figure 2-4. The structure of the instrumentation for electrochemical biosensors. ........... 20
Figure 2-5. (a) 3-electrode potentiostat with grounded WE structure. (b) 3-electrode
potentiostat with grounded CE structure........................................................................... 22
Figure 2-6. Fully-differential potentiostat. (adapted from [50]) ....................................... 23
Figure 2-7. Resistive feedback current readout circuit. .................................................... 25
Figure 2-8. Capacitive feedback current readout circuit. .................................................. 26
Figure 2-9 The structure of CDS circuit (adapted from [49])........................................... 27
Figure 2-10. The structure of the pseduo-differential amperometry readout circuit
(adapted from [52]) ........................................................................................................... 28
Figure 2-11. The structure of current mirror amperometric readout circuit with
potentiostat. (adapted from [51]) ...................................................................................... 29
Figure 2-12. pseudo-bidirectional current conveyer with a DC offset current [72]. ........ 30
Figure 2-13 The structure of current mode sigma-delta ADC circuit. .............................. 31
Figure 2-14. Block diagram of a lock-in based impedance spectroscopy. ....................... 34
Figure 2-15. The structure of the lock-in ∑-∆ impedance extraction circuit. The circuit
extracts the imaginary portion of admittance when S=1 and real portion when S=0.
(adapted from [31]) ........................................................................................................... 35
Figure 2-16. Dual-slope multiplying ADC VLSI architecture. ........................................ 36
Figure 3-1. Block diagram of the amperometric instrumentation system. ....................... 41

x

Figure 3-2. Structure of the potentiostat designed for amperometric biosensor. .............. 42
Figure 3-3.Structure of the OPA2 used in potentiostat. .................................................... 44
Figure 3-4. The phase margin simulation result of OPA2. ............................................... 45
Figure 3-5. Schematic of amperometric readout circuit. .................................................. 47
Figure 3-6. The schematic of the clock generator to generate ph1, ph2 and ph3 signals. .. 48
Figure 3-7. The simulation results of the time sequence of ph1, ph2, and ph3 generated by
the clock generator. ........................................................................................................... 49
Figure 3-8. The schematic of the switch used on the readout circuit. .............................. 50
Figure 3-9. The simulation results of the switch clock feedthrough cancellation by
dummy NMOS transistors. ............................................................................................... 51
Figure 3-10. Schematic of OTA in amperometric readout circuit. ................................... 52
Figure 3-11. The 3×3mm2 CMOS amperometric instrumentation chip with waveform
generator and 4-channel potentiostat and amperometric readout array. ........................... 53
Figure 3-12. The deviation between input current and measured current for 1pA to 30pA
inputs at 20Hz updating clock and gain=2 setting. ........................................................... 56
Figure 3-13. Photograph of a CMOS biosensor array chip-in-package and close up views
of the post-CMOS surface electrode array........................................................................ 57
Figure 3-14. CV measurement of 0.5mM potassium ferricyanide at 100mV/s and
200mV/s for both the CHI 760 commercial instrument and the proposed circuit. ........... 58
Figure 3-15. CV measurement of potassium ferricyanide at 0.5mM and 1mM for both the
CHI 760 commercial instrument and the proposed circuit. .............................................. 59
Figure 3-16. CV measurement of cytochrome PpcA interface in 150μM and 400μM FeNTA for both CHI 760 commercial instrument and AIC chip. ........................................ 60
Figure 4-1. Illustration of the lock-in based FRA impedance component extraction
method............................................................................................................................... 65
Figure 4-2. Conceptual illustration of the impedance spectroscopy circuit with SFS and
IRC circuitry and on-chip electrode and biointerface array. The chip communicates with
a PC to upload configuration data and download digitized measurement results. ........... 67
Figure 4-3. Block diagram of the signal generator that produces stimulus signal sin(ωt)
and the reference square wave Dsin(ωt) and Dcos(ωt) (online version in color). ............ 69

xi

Figure 4.-4. Simplified structure of the impedance readout circuit presented in [31] and
the IRC (except for gray part). φ and not-φ are non-overlapping reference square waves.
........................................................................................................................................... 74
Figure 4-5. Photo of the packaged ISA chip with on-chip electrode array....................... 75
Figure 4-6. Test platform for the IRC chip with biosensor equivalent circuit model. ...... 76
Figure 4-7. The real (circles) and imaginary (squares) results from the IRC circuit for the
test setup shown in Figure 4-6 along with theoretical (expected) curve for comparision. 77
Figure 4-8. Measured DNL and INL curve of one IRC channel output when the sinusoid
input amplitude was swept from 0 to 10nA. The maximum DNL and INL are 59.8dB and
50.2dB, respectively.......................................................................................................... 78
Figure 4-9. Response of an IRC readout channel as the phase of the input signal is varied
from -180° to +180° degrees. Input amplitude (30nA) and frequency (100Hz) are kept
constant. ............................................................................................................................ 79
Figure 4-10. Schematic view and equivalent model of a two-electrode electrochemical
site with bottom electrode coated with a tBLM. In the tBLM equivalent model, Rm and
Cm represent the resistance and capacitance of the BLM assembly, respectively, while
RS is the resistance of the solution and Cdl describes the double layer capacitance. ...... 80
Figure 4-11. tBLM amplitude and phase measured by a commercial instrument (CHI 660
B) and by the IRC circuit at different frequencies. ........................................................... 82
Figure 5-1. Block diagram of the multi-mode electrochemical readout chip (MERC). ... 85
Figure 5-2. The hysteretic comparator reduces the undesired switching caused by noise.
........................................................................................................................................... 87
Figure 5-3. The structure of multi-mode current readout channel. The blue path is
implemented for the impedimetric mode. The green path is implemented for the
amperometric mode. ......................................................................................................... 87
Figure 5-4. The output of integrator when (a) the hysteresis magnitude is 0V, and when
(b) the hysteresis magnitude is 100mV............................................................................. 91
Figure 5-6. Schematic of the hysteretic comparator. ........................................................ 94
Figure 5-7. The simulation results of the hysteretic comparator. ..................................... 95
Figure 5-8. Schematic of the amplifier used in the integrator. ......................................... 96
Figure 5-9. The photograph of the multi-mode electrochemical readout chip. ................ 97
Figure 5-10. The printed circuit board for MERC characterization. ................................ 97

xii

Figure 5-11. The test result of the potentiostat. ................................................................ 98
Figure 5-12. The DNL results (represented as the least significent bit (LSB)) of the MCR
circuit with and without hysistersis................................................................................... 99
Figure 5-13. The photo of the interdigitated electrode (IDE). ........................................ 101
Figure 5-14. The test bench schematic for MERC chip with Herceptin bio-interface. .. 102
Figure 5-15. The CV measurement of the gold electrode before and after coating with
peptide CH-19. ................................................................................................................ 104
Figure 5-16. The impedance of Herceptin measured by the commercial instrument and
MERC. ............................................................................................................................ 105
Figure 5-17. The Herceptin concentration curve vs. the impedance measured by MERC
chip. A high level of linearity (R2=0.9795) was achieved. ............................................. 106
Figure 6-1. General structure of a pore-based pBLM cell with a single ion channel protein.
The opening/closing events of the ion channel can be detected by monitoring response
currents between electrodes A and B. ............................................................................. 110
Figure 6-2. PDMS microchamber with electrode fabricated in Dr. Mason’s lab. Work
completed by colleague Lin Li. ...................................................................................... 112
Figure 6-4. Example for different classes of pBLMs. .................................................... 116
Figure 6-5. Architecture of the single protein electrochemical characterization (SPEC)
high density array chip. ................................................................................................... 119
Figure 6-6. Schematic of the two stages current conveyor pixel preamplifier. .............. 122
Figure 6-7. Layout of the pixel preamplifier. ................................................................. 122
Figure 6-8. The simulation results of the first stage output (IA) and second stage ouput
(IOUT) of the preamplifier with a 5pA, 10µs width pulse current input (IWE). ................ 125
Figure 6-9. The definition of the pulse width distortion percentage (PWDP). ............... 125
Figure 6-10. Simulated results of the distortion percentage (DP) with various pulse
widths of the input current (IWE). .................................................................................... 126
Figure 6-11. Simulated IOUT pulse width distortion percentage (PWDP) as a function of
the column line parastic capacitance (Cp). ..................................................................... 127

xiii

Chapter 1 Introduction
1.1 Electrochemical Protein Biosensor
Proteins are vital entities in human cells and are involved in nearly all body
functions both in healthy and diseased individuals [1-3]. Proteins can support the skeleton,
control senses, move muscles, digest food, fight infections and process emotions.
Understanding the structure and function of gene products, that is proteins, has been
increasingly emphasized for the completion of the Human Genome Project and the
sequencing of several other important genomes. Because proteins are the targets for
virtually all pharmaceutical drugs and for most diagnostics [1], proteins can be used in
detecting disease and creating new treatments for health problems such as cancer,
cardiovascular and neurological disorders. According to American Cancer Society report,
about 1,638,910 new cancer cases are expected to be diagnosed and about 577,190
Americans are expected to die of cancer in 2012 (more than 1,500 people a day) [4]. The
deaths could potentially be prevented by better understanding the role of proteins in
biological operations.
Protein characterization (also referred to as proteomics) and sensing are two
directions in proteins study. Protein characterization is analyzing protein sequence,
structure and post-translational modifications [5]. Protein characterization has been used
widely in drug discovery, development of biosensors, clinical and forensic analysis [6-9].
In general, protein characterization is laboratory research, which requires high resolution

1

detector (or instrumentation). Sometimes, the instrumentation even needs continue record
the proteins kinetic characterization. Protein sensing is detection the concentration of
target proteins. There is a direct correlation between protein levels and disease states in
human serum, which makes protein as an attractive target for sensors and diagnostics [10].
Compared with characterization, sensing emphasizes real-time detection.
Techniques for protein sensing and characterization are generally divided into two
groups, label and label-free. The most applied label technique is enzyme-linked
immunosorbent assay (ELISA), which is complex, time-consuming and laborious. All the
optical label techniques require expensive free-space optical equipment and difficult
alignment of all the components. To overcome these drawbacks, label-free techniques
have been introduced. Surface Plasmon resonance (SPR), quartz crystal microbalance
(QCM), and electrochemical are the main techniques used in proteins interactions
measured. Compared with SPR and QCM, electrochemical biosensors provide more rapid,
simple and low-cost on-field detection [11]. Electrochemical biosensor is a molecular
sensing device that intimately couples a biological recognition element to an electrode
transducer. The principle for electrochemical biosensor transducers is that many chemical
reactions produce or consume ions or electrons, causing some change in the electrical
properties of the solution, which can be sensed as measuring parameters.
Recently, protein sensing and characterization fields are moving towards point-of
care diagnostics [12]. Point-of-care diagnostics play an important role in bridging the gap
between centralized laboratory diagnostics and peripheral healthcare service providers.
This is particularly true in infectious diseases such as HIV/AIDS and H1N5 where early
detection is imperative to improve disease outcome [13]. Electrochemical biosensors are

2

good candidates for point-of-care devices due to their easy integration and
miniaturization. Now, the trend of point-of-care devices is to combine with array
technique, which allows rapid and quantitative detection of a wide range of biomolecules
[14, 15]. These trends suggest that electrochemical biosensors could have a major impact
on future diagnostic (e.g., point-of-care devices) and therapeutic approaches (e.g.,
personalized medicine).

1.2 Miniaturized Protein Array Biosensor
A miniaturized, high throughput array takes advantage of the fact that a large
number of targets can be analyzed in parallel measurements with low sample
consumption. The reduction of sample volume is of great importance for all applications
in which only minimal amounts of samples are available. The volume of the liquids at the
micro scale is also great in molecular concentrations control and interactions [16].
Parallel measurements reduce the assay time significantly. Furthermore, it is possible to
perform comparative analyses of several different samples within a single array chip. For
example, in medical research, all relevant immune diagnostic parameters of interest can
be analyzed in parallel, which saves cost and time.
Microfluidics (or lab-on-a-chip) provide an attractive platform for array
applications because microfluidics can automatically control the liquid flow for each biocell individually. The automation of chemical processing controlled by microfluidics also
reduces errors caused by manual operation [17]. The optical DNA array associated with
microfluidics shows the advantages of less sample usage and reduced incubation time
[18]. However, the optical method still needs expensive optical instrumentation, which

3

limits its point-of-care application. Recently, microfluidic electrochemical hybridization
techniques have shown substantial successes in point-of-care diagnostic and therapeutic
approaches [19]. Electrochemical biosensors require electronic instrumentation, and most
modern instrumentation relies on complementary metal–oxide–semiconductor (CMOS)
microelectronics technology due to the success of CMOS in integrated circuits (IC)
industry. CMOS instruments are low-cost and easily miniaturized devices, which is well
suited for point-of-care diagnostic.
Traditional lab-on-a-chip devices do not include instrumentation electronics. As
such, when microfluidics technologies are integrated with sensor array approaches, a
wiring bottleneck to physically separated electronics is encountered. Thus, there are
challenges inherent to forming hundreds of individually measurable protein elements on a
single substrate.
The Advanced MicroSystems and Circuits (AMSaC) lab at Michigan State has
been developing new approaches to combine microfluidics and CMOS instrumentation
into one miniaturized system (called lab-on-CMOS), wherein no wires are required
between protein interface and instrumentation electronics. In the proposed platform, a
CMOS instrumentation chip acts as the substrate of the microsystem. An array of gold
electrodes fabricated post-CMOS directly on the surface of the chip are connected
through overglass contact openings to the underlying CMOS electronics. To insulate
surface metal routing and define size-adjustable openings over individual electrodes, a
post-CMOS package is applied after metallization. This insulation layer also planarizes
the surface and provides an interface to a variety of possible fluid handling schemes,
including microfluidics illustrated in Figure 1-1. The direct, on-chip, electrical

4

connection of electrodes to the instrumentation improves the signal to noise ratio due to
the elimination of the external wiring and immunity from environmental interference.
This noise reduction permits highly sensitive circuits to measure the responses of
miniature biosensors, allowing a high-density sensor array within the small platform of a
CMOS chip.

1.3 Challenges for CMOS Instrumentation
The challenges in forming miniaturized and high throughput array biosensors can
be organized into three main categories: biointerface formation, CMOS instrumentation,
and physical integration of all system components including electronics and microfluidics.
Each of these presents complex and challenging problems to solve. While the
biointerface formation and physical integration challenges are being explored by other
researchers, the work under this thesis will focus on the challenges highlighted below for
implementing CMOS instrumentation necessary for next-generation electrochemical
biosensor arrays.
Limited area is the first challenge of CMOS instrumentation for high throughput
application. All the electrochemical detection circuits have to be integrated in a less than
1 cm2 chip. To achieve one chip reading hundreds or even thousands of protein elements
in parallel, the detection circuits are needed to be extremely hardware efficient.

5

Figure 1-1. Conceptual illustration of a CMOS electrochemical biosensor array
microsystem use in a liquid environment (For interpretation of the references to color in
this and all other figures, the reader is referred to the electronic version of this
dissertation).

In general, multiple electrochemical techniques are required to perform protein
characterization or sensing. This creates an expectation that single chip instrumentation
for next-generation biosensor arrays must be able to achieve multi-function
electrochemical measurement. Thus, another significant challenge this thesis research
must address is how to implement highly integrated multi-function circuits in limited area.
As biosensors advance to high throughput arrays, they produces a tremendous load
of raw data. This, in turn, creates a requirement for the instrumentation that must be able

6

to acquire, process, and communicate results for many sensor elements in parallel. At the
same time, the power must be minimized to support portable applications and ensure no
local heating that could degrade on-chip biointerfaces. The handling of this high data
density while maintaining low power consumption is inherently challenging.
Miniaturization of biosensor arrays into the chip-scale platform shown in Figure 11 requires the size of on-chip electrodes to be very small to fit up to hundreds of
electrodes on the protein array chip. This miniature electrode size causes the biosensor
output signals to be extremely weak, on the order of femto to pico amperes (~fA – pA).
Furthermore, some of the most interesting and important bio-reaction responses are
inherently weak. For example, single molecule detection is considered to be the “holy
grail” of high resolution biosensing. However, responses from single molecules are
naturally very small and measuring them is beyond the capability of most existing
instruments. Therefore, acquiring single molecule reaction signals is one of the main
challenges for development of next-generation high throughput protein array
microsystems.

1.4 Dissertation Goals
The primary goal of this research is to develop CMOS instrumentation electronics
that enhance the evolution of protein biosensors toward miniaturized high throughput
array microsystems. Specifically, this thesis research seeks to:

7

•

Develop CMOS electrochemical circuits with state-of-the-art performance in
terms of hardware efficiency, power consumption and bio-reaction current
readout resolution.

•

Efficiently integrate multi-mode electrochemical techniques onto a single chip for
protein characterization and sensing under widely varied controlled conditions.

•

Develop electrochemical instrumentation circuitry that addresses the many
challenges in forming high throughput chip-scale biosensor arrays with
simultaneous characterization of multiple bio-interfaces.
These proposed CMOS instrumentations will establish new biological-to-silicon

communication pathways that enable future advances in many areas of microsensor
technology and biomedical instruments.

1.5 Outline
In this thesis, Chapter 2 covers the background of electrochemical biosensor
components, detection methods, and different structures of electrochemical CMOS
instrumentation are reviewed. A low noise amperometric instrumentation chip is
introduced in Chapter 3 and test results with a cytochrome protein interface are shown.
Chapter 4 introduces a compact, low power impedance spectroscopy circuit that was
developed to characterize tethered lipid bilayer membranes. Chapter 5 presents a novel
hybrid multifunctional readout circuit and test results from a Herceptin biosensor.
Chapter 6 presents a CMOS electrochemical instrumentation circuit for high throughput

8

membrane proteins characterization. Chapter 7 summarizes the thesis work and
contributions, and it outlines future work related to this research.

9

Chapter 2 Background
2.1 Electrochemical Biosensor Introduction

2.1.1 Electrochemical biosensor components
An electrochemical biosensor is commonly composed of four components, as
shown in Figure 2-1: analyte, bio-probe, transducer and instrumentation. The analyte is
the target of interest, such as a specific type of protein, DNA, bacteria or tissue. The bioprobe is the element that provides the selective recognition capability for a target analyte.
The transducer translates the biorecognition event into electrical signal. The typical
transducer of an electrochemical biosensor is an electrode, which translates the
chemically generated ions into a current or voltage. Instrumentation electronics serve to
record the response current or voltage.

2.1.2 Transducer
In electrochemical biosensors, electrodes are the dominate transducer. The
electrode can be made from metal, mixed metal oxide, carbon, graphite, or any other
conductive material. Because bio-reactions occur in close proximity to the electrode
surface, the study of the processes at the electrode-solution interface is crucial [20]. Two
types of processes occur at the electrode-solution interface. When electron charge is
transferred across the interface, the reactions is called as faradic processes and is

10

governed by Faraday’s law. Processes that do not transfer charges cross the electrodesolution interface are called nonfaradaic, and these can generate electrode current when
the potential, electrodes, or solution composition changes. For example, adsorption and
desorption reactions are typical nonfaradaic processes.

2.1.3 Immobilization
Bio-probe interfaces are often immobilized onto a solid surfaces, especially in
miniaturized biosensors. Adsorption, cross linking, covalent bonding, entrapment,
encapsulation, etc. have all been used for binding to solids [21]. The selection of an
appropriate immobilization method depends on the nature of the biological element, the
type of transducer used, the physicochemical properties of the analyte, and the operating
conditions for the biosensors. The most common methods for immobilization of biocomponents are adsorption and covalent bonding. Physical adsorption of the bio-probe is
based on van der Waals attractive forces. Covalent bonding is based on a reaction
between the same terminal functional groups of the protein and reactive groups on the
solid surface of the insoluble bed.

Figure 2-1. Common components of electrochemical biosensor.

11

2.2 Electrochemical Biosensor Techniques
Electrochemical biosensors can be classified according to the operating technique
used to measure electrochemical changes during a biorecognition event. These classes
have been popularly defined as potentiometric, amperometric, impedimetric and fieldeffect transistor (FET) biosensors [22]. Table 2-1 lists the typical biosensor
electrochemical techniques for different types of bio-probes. From this table, we found
that one bio-probe could be characterized by multiple electrochemical techniques. This
leads to a future research direction for multi-mode electrochemical measurement.

2.2.1 Potentiometric biosensor
In potentiometric biosensors, the analytical information is obtained by converting
the biorecognition events into a potential signal, which is proportional to the
concentration of analyte [23]. The measured potential should be made at zero-current or
quasi-equilibrium condition [24]. The most common potentiometric devices are pH
2

electrodes; several other ion (F⁻, I⁻, CN⁻, Na⁺, K⁺, Ca ⁺, NH4⁺) selective electrodes are
available. The potential differences between these indicator and reference electrodes are
proportional to the logarithm of the ion activity, as described by the Nernst-Donnan
equation [22]. Potentiometric biosensors have been used for antigen and DNA detection.
The fundamental disadvantage of the potentiometric technique is that the double-layer
charging effects are frequently larger and occur throughout the experiment [25].

12

Table 2-1. Types of bio-probes used in biosensors and their common electrochemical
technique.
Bio-probe

potentiometry

amperometry

impedance

FET

enzymes
Membrane
protein
whole cells
antigen/antibody
plant or animal
tissue

2.2.2 Amperometric biosensors
Amperometric biosensors are based on monitoring the current associated with
oxidation and/or reduction of an electro-active species involved in the recognition
process. In amperometric biosensor, a constant or ramp potential is applied on electrodeelectrolyte interface. The applied potential serves as the driving force for the
biorecognition process, and the response current directly relates to the analyte
concentration. Amperometry is a common electrochemical detection method in
immunosensors [26], enzyme biosensor [27] and pesticide monitoring [28]. Compared to
potentiometry, amperometry has higher sensitivity and wider linear range [29].

13

2.2.3 Impedimetric biosensor
Impedimetric biosensors measure the electrical impedance of the electrodeelectrolyte interface at equilibrium [30]. Impedimetric biosensors take advantage of the
bio-interfacial

characteristics,

i.e.,

capacitance

and

resistance

change,

during

biorecognition events. An impedimetric biosensor is created by imposing a small
sinusoidal stimulus voltage (or current) of variable frequency and measuring the phase
and/or amplitude change of the resulting current (or voltage) that contains information
regarding biorecognition events. This technique is generally referred to as the impedance
spectroscopy (IS) and has been employed in a variety of applications such as monitoring
bilayer lipid membranes [31, 32], detecting small molecules of biological relevance [33,
34], measuring cell concentrations [35, 36]. Compared to potentiometric and
amperometric biosensors, an important advantage of impedimetric biosensor is that the
stimulus sinusoidal voltage is small (usually 5-10 mV in amplitude) and does not damage
or disturb most bio-probe layers.

2.2.4 Ion-sensitive field effect transistor biosensor
Another method to measure ion concentrations in a solution utilizes ion-sensitive
field effect transistors (ISFET). As illustrated in Figure 2-2, an ISFET operates in a
manner similar to a MOSFET, where the current between source and drain terminals is
proportional to the gate voltage that, for an ISFET, is set by reactions between analytes
and bio-probes (immobilized on the gate) during biorecognition events [37]. ISFET
biosensors have been developed for enzyme sensing [38], measuring antigen-antibody

14

bonding reactions [39] and DNA detection [40, 41], and immobilizing nanomaterials
such as nanoparticles, nanotubes, and nanowires on the gate has been shown to improve
sensitivity [42]. Unlike potentiometric, amperometric and impedimetric biosensors that
use an electrode transducer, the ISFET’s transducer is the gate oxide layer. As a result,
ISFETs require unique instrumentation circuitry that is dissimilar to the other categories
of electrochemical sensors. The remainder of this chapter will focus on biosensors with
electrode transducers. For more information on the principles, applications and
challenges of ISFET biosensors, the reader is referred to other sources [37, 43].

Figure 2-2. The structure of ISFET biosensor.

2.2.5 Electrode transducer
Electrochemical biosensors employ multiple electrode transducers to translate
biorecognition responses into electrical signals. The electrode where electrochemical
reaction of interest takes place is called the working electrode (WE). The electrode that

15

provides a stable and known potential is called the reference electrodes (RE). The
simplest electrode system utilizes only these two electrodes. In such two-electrode
systems, a known stimulus voltage, Vst, is applied between RE and WE. The resulting
potential across the electrochemical cell’s electrode-electrolyte interface, Vcell, is, ideally,
equal Vst throughout the electrochemical reaction. However, as shown in Figure 2-3 (a),
in addition to the electrode-electrolyte interface represented by impedance Zbio,
electrochemical cells inherently possess a solution resistance, Rs, representing the ion
flow path through the electrolyte solution to an electrode. As a result, Vcell is always less
than Vst because of voltage dropped across Rs, which is commonly referred to as the “IR
drop”. Because the IR drop cannot typically be known in advance and can change with
environmental conditions, reaction kinetics, etc., it is generally considered to be an
undesirable source of noise/error.
Although Rs cannot be entirely removed from an electrochemical cell, the negative
effects of IR drop can be eliminated by adding a third electrode, called a counter
electrode (CE), to the system. As shown in Figure 2-3(b), in a three electrode system, the
stimulus voltage across Zbio is established between RE and WE, and no current is
allowed to flow through the RE terminal thus eliminating any IR drop. Instead, the
response current is read through the CE where the IR drop will have no effect on the
biosensor response current. Because of the performance advantages of eliminating the IR
drop effect, three electrode systems are widely used in electrochemical biosensors. Note

16

that potentiometric sensors are generally measured at zero current, and impedimetric
sensors generally have small signal AC currents with negligible IR drop. Thus, three
electrode systems are especially useful with amperometric techniques (typically with
redox reactions) where larger response currents are generally observed.
Biosensor interfaces with low turnover rates can result in low response currents and
low sensitivity. For redox active bio-probes, the redox recycling effect can be utilized to
electrochemically magnify the faradic current by increasing mass transport of the redox
active species of interest. To utilize redox recycling in biosensors, four electrode
electrochemical systems have been developed [25, 44-47]. Four electrode systems have
one RE, one CE and two WEs with redox probes immobilized on both. They are ideally
suited for micro-scale interdigitated electrode arrays where the close proximity of WEs
enhances redox recycling. Because redox recycling is rarely used and only available with
redox bio-probes, four electrode systems are not very common.

17

.
(a)

(b)
Figure 2-3. The equivalent circuit model of (a) a two electrode system and (b) a three
electrode system.

18

2.3 Interface Circuitry for Electrochemical
Biosensors
After electrode translating the biorecognition events to electronic domain, the
electronic signals are measured by instrumentation. Instrumentation also provides a
potential voltage for the electrode-electrolyte interface. CMOS process has been the most
popular integrated circuit technology in recent decades, so most of the instrumentation
for electrochemical biosensors is CMOS instrumentation.
CMOS instrumentation can be divided into four elements based on function: signal
generator, potentiostat, readout circuit and signal processing. Signal generator generates
various shapes of stimulus signals, such as triangle, constant potential, saw, and pulse etc.
Potentiostat controls the biasing potential of electrodes. The readout circuit is developed
to detect the response current/voltage. Signal processing block analyzes the data coming
from the readout circuits. The structure of the CMOS instrumentation is shown in Figure
2-4. The signal generator and signal processing blocks are excluded from this paper,
because these two blocks are not highly related to the electrode-electrolyte interface, and
can be implemented by general commercial circuits. The potentiostat and readout circuit
is discussed in detail in following sections.

19

Figure 2-4. The structure of the instrumentation for electrochemical biosensors.

2.3.1 Potentiostat
In an electrochemical reaction, when Vcell reaches the analyte’s redox potential, the
biochemical reaction occurs, so it is critical to have circuits to set Vcell during measuring.
The circuit served as providing accurate Vcell to the electrode-electrolyte interface is
called the potentiostat.
The first CMOS potentiostat was presented on 1987 for two electrodes
amperometric chemical sensors [48], in which one operation amplifier achieves the
potentiostat function. Due to the advantages of the three electrodes compared to two
electrodes system, most of the modern potentiostats [49-57] are designed for three
electrodes system.
There are three configurations of potentiostat in three electrodes system: ground
WE; ground RE; and ground CE [51]. The first two configurations are electrically same;

20

thus, in fact, there are really only two configurations: ground WE and ground CE. The
simplest ground WE potentiostat is implemented by one operational amplifier (OPA).
The two inputs of the OPA connect to Vst and RE, respectively and WE connects to
ground. Vcell equals Vst because of the inputs OPA having same potential. The output of
the OPA connects to CE for providing current path from CE to WE. The schematic of the
grounded WE potentiostat is shown in Figure 2-5 (a). Because the RE and WE voltages
(related to ground) vary at the same time, more circuits are needed to make sure the
voltage difference between RE and WE equaling to Vst, as shown in Figure 2-3. "A
grounded CE potentiostat circuit is presented in Figure 2-5 (b) which is more complex
than the circuit for the grounded WE, therefore nearly all potentiostats use the grounded
WE structure [50, 53-57].

21

(a)

(b)
Figure 2-5. (a) 3-electrode potentiostat with grounded WE structure. (b) 3-electrode
potentiostat with grounded CE structure

The potentiostats structures introduced above are single-ended potentiostats. The
low power supply trend of CMOS process, however, limits the functionality of
potentiostat because biosensors’ redox potentials can be higher than the power supply. To
solve this problem in low-voltage process, a fully-differential potentiostat was developed

22

to increase the output swing of potentiostat [50]. The fully-differential potentiostat as
shown in Figure 2-6, is based on a fully-differential amplifier, OP3, which dynamically
controls the RE and WE voltages. This developed fully-differential potentiostat doubles
the Vcell output range in single-ended potentiostat [50].

2.3.2 Readout circuits
In electrochemical biosensor, the electrolyte response signals are detected by
readout circuits. Because the potentiometric technique is barely used in biosensor, only
current readout circuits for amperometric and impedimetric biosensors are introduced in
this paper. Based on the characters of biosensing current, the current readout circuitry can
be classified by DC and AC current readout.

Figure 2-6. Fully-differential potentiostat. (adapted from [50])

23

2.4 DC Current Readout Circuitry
For amperometric biosensors, the biointerfaces response current is DC, so the
readout circuits are DC current readout. In general, the amperometric readout circuit
should provide wide current range, high resolution at interested bandwidth and allow
measurement of bidirectional currents (sink/source). Numerous CMOS amperometric
circuits have been introduced over the past decades with various functionality and
performance to meet different applications demands [48, 54, 55, 58-64]. Basically, there
are three structures for amperometric current readout circuits: resistive, capacitive and
current conveyor current readout.

2.4.1 Resistive current readout circuits
The simplest way to readout the response current is to use a sense resistor. The
resistive feedback current readout circuitry uses an operational amplifier with a feedback
resistor to convert the biosensor sensing current Isn to voltage, as shown in Figure 2-7.
The biosensor sensing current Isn can be calculated by

Iୱ୬ = −

୚౥౫౪
ୖ౜

(2-1)

where Rf is the feedback resistance, Vout is the output voltage of the operational
amplifier. The dynamic range of the current readout circuitry can be achieved by
changing the Rf value. This resistor feedback structure is very simple to develop,
however it does not fit for the low noise amperometric application because the feedback

24

resistor, Rf, generates significant thermal noise. That is the major drawback of this
structure [65].

Figure 2-7. Resistive feedback current readout circuit.

2.4.2 Capacitive current readout circuits
To eliminate the thermal noise generated by the feedback resistor, capacitive
feedback structure has been introduced [60, 61, 64, 66-69]. Detailed noise analysis of the
resistive feedback and capacitive feedback structure was presented [65], in which claimed
the capacitive feedback system provides better noise performance than the resistive
feedback structure.
The basic structure of the capacitive feedback current readout is an integrator
converting the input current to a voltage, as shown in Figure 2-8 [70]. The biosensor
sensing current Isn can be calculated by

25

ଵ

୘

V୭୳୲ = େ ‫׬‬଴ Iୱ୬ dt
౜

(2-2)

where Cf is the feedback capacitor, T is the integration time, Vout is the output voltage of
the operational amplifier.

Figure 2-8. Capacitive feedback current readout circuit.

The performance of this capacitive feedback current readout circuit is degraded by
the switching effect, which presents as offset errors (including clock feed through, charge
injection, and KT/C noise). To eliminate the offset error, correlated double sampling
(CDS) technique has been introduced in the capacitive feedback readout circuit [49].
CDS technique also can reduce the 1/f noise that is a significant noise source in DC
readout circuitry. The CDS circuit structure is presented in Figure 2-9, where a switchedcapacitor (SC) charge integrator converts the input current to a voltage. The voltage is
sent to a programmable gain amplifier (PGA) with a gain of C1/C2, where C is a digitally
programmable on-chip capacitor array. The output of the PGA is sampled and held and
then fed to an analog-to-digital converter (ADC) after a low pass filter. The non-

26

overlapping clock signals for switches ph1, ph2 are generated by an on-chip clock
generator block. The input current range can be adjusted to different biosensors through
programmable selection of the clock frequency and the PGA gain. These programmable
features can be used to calibrate sensors over time.
To reduce the effects of charging currents and currents due to undesired analyte,
pseudo-differential amperometric readout circuits were developed [52, 71]. Two WEs are
used instead of one. The schematic of pseudo-differential readout is presented in Figure
2-10, where the reading time controlled by switches Sn and Sp. By subtraction of the
currents from WEn and WEp, nearly all parasitic currents can be canceled while
preserving the analytical signal of interest [52].

Figure 2-9 The structure of CDS circuit (adapted from [49]).

27

Figure 2-10. The structure of the pseduo-differential amperometry readout circuit
(adapted from [52])

2.4.3 Current conveyor current readout circuits
Current conveyor structure is the third type of the amperometric current circuitry.
Current conveyor is a current mode amplifier to amplify the input current eliminating the
noise sources from the following stages [51, 61, 72, 73]. Figure 2-11 presents a schematic
of a typical current conveyor used for measuring currents. The biosensor sensing current
Isn can be calculated by

28

୍౩౤
୍∗౩౤

ሺ୛⁄୐ሻభ

= ሺ୛⁄୐ሻ

మ

(2-3)

where W/L is the transistor size. The current conveyor circuit performs decoupling and
linear operations in current mode in the same way that the operational amplifier performs
in voltage mode. One drawback of the current conveyor is a one current direction readout
circuit, which is not suitable for bi-direction current applications. To overcome the one
direction drawback, a pseudo bidirectional current conveyer with a DC offset current has
been presented [72]. Its structure is presented in Figure 2-12. Compared to the capacitive
and resistive feedback structure, the drawback of the current conveyor is the bias current
noise can directly sum with the input current, which causes larger error [65].

Figure 2-11. The structure of current mirror amperometric readout circuit with
potentiostat. (adapted from [51])

29

Figure 2-12. pseudo-bidirectional current conveyer with a DC offset current [72].

To improve the noise performance of the current conveyor, current mode sigmadelta modulator has been introduced [60, 74, 75]. Sigma-delta modulator takes advantage
of the noise shaping by oversampling, which moves the low frequency noise to high
frequency and then is removed by a low pass filter. A current mode sigma-delta structure
is shown in Figure 2-13. The duty-cycle modulation of current feedback in the ∑-∆ loop
together with variable oversampling ratio provide a programmable digital range selection
of the input current spanning [74]. The performance of several amperometric readout
circuits is listed in Table 2-2.

30

Figure 2-13 The structure of current mode sigma-delta ADC circuit.

2.5 AC Current Readout Circuitry
Impedimetric current readout circuitry is an AC current readout. In impedimetric
test, a stimulus sine wave with known frequency content is applied to the electrodeelectrolyte interface. When biorecognition events occur, the resulting output response
contains a phase and/or amplitude shift, algebraically related to real and imaginary
components. Thus, to detection the biorecognition event, the impedimetric readout circuit
should have the capability to extract the phase/amplitude or real/imaginary shift, which
makes circuits more complex than amperometric readout circuits.
The phase and amplitude shift can be extracted in the analog domain or the digital
domain. The extraction method in the digital domain is the Fast Fourier Transform (FFT)

31

algorithm [30]. The FFT algorithm utilizes a broadband stimulus, such as a pulse, and
computes the result at all frequency points simultaneously, producing an impedance
spectrum. The composite signal source and the computationally intensive Fourier
transform require FFT-based impedance spectroscopy instrumentation to use a digital
signal processor (DSP) with extensive computational resources [76]. The phase and
amplitude extraction method in analog domain is called frequency response analyzer
(FRA). The FRA processes the response of one frequency point at a time and sweeping
over the whole interested frequency range. Compared to FFT, the FRA method can be
realized with simple and compact analog circuits, making it suitable for sensor array
microsystems [76]. Therefore, most of the CMOS impedimetric current readout circuits
use the FRA method [77-83].
In the FRA method, the only reported phase/amplitude extraction technique is the
lock-in technique. Lock-in is achieved by multiplying the electrolyte response AC signal
with a reference signal with same frequency. The multiplying results provide a DC output
proportional to the real/imaginary portions of the AC signal under investigation. To
indicate how lock-in extracts the phase/amplitude or real/imaginary change, a block
diagram of the lock-in based the FRA algorithm is presented in Figure 2-14. A signal
generator generates the sinusoidal voltage sin(ωt) to stimulus the bio-cell. The biosensor
response signal is Asin(ωt+θ) with phase shift	θ and amplitude shift A. Then Asin(ωt+θ)
multiplies with reference signal, sin(ωt) or cos(ωt) respectively. The equations follow

32

୅

A sinሺωt + θሻ sinሺωtሻ = ሾcos θ − cosሺ2ωt + θሻሿ
ଶ

୐୔୊ A

୅

ଵ

= cosሺθሻ − 	 cosሺ2ωt + θሻ 	 ሳልሰ A cosሺθሻ = Reୠ୧୭
A
ଶ

ଶ

ଶ

2

(2-4)

୅

A sinሺωt + θሻ cosሺωtሻ = ሾsin θ + sinሺ2ωt + θሻሿ
ଶ

୅

୐୔୊ ଵ

୅

ଵ

= sinሺθሻ + 	 sinሺ2ωt + θሻ ሳልሰ A sinሺθሻ = Imୠ୧୭
ଶ

ଶ

ଶ

ଶ

(2-5)

where Rebio and the Imbio is the extracted real and imaginary portions of the bio-cell
impedance, respectively. ω is the frequency of stimulus signal, Acosሺ2ωt + θሻ and
Asinሺ2ωt + θሻ	are the higher order frequency components, which can be removed by
low pass filter (LPF) [31, 81]. Based on (2-4), (2-5), the amplitude A and phase θ shift
can be calculated by

A = ටReଶୠ୧୭ + Imଶୠ୧୭
θ = tanିଵ

୍୫ౘ౟౥
ୖୣౘ౟౥

(2-6)

(2-7)

The sinusoid reference signals shown in Figure 2-14, can be replaced by
rectangular waveform with the same phase of sin(ωt) and cos(ωt). This simplifies the
signal generator design. Ideally it maintains the same performance as the structure in

33

Figure 2-14 because the higher odd harmonics caused by the rectangular waveform can
be eliminated by LPF.

Figure 2-14. Block diagram of a lock-in based impedance spectroscopy.

Several EIS readout circuits have been developed based on the lock-in technique
[31, 78, 79, 81, 83, 84]. A novel structure has been reported which combines the sigmadelta technique and the lock-in multiplier to support impedance extraction and
digitization [31]. This circuit shares resources for impedance extraction and digitization
to maximize the hardware efficiency, which permits over 100 readout channels within a
3x3 mm die. Its blocks diagram was presented in Figure 2-15., where the real and
imaginary portions are extracted by the multiplier and the higher frequency noise is
removed by the integrator.
The other impedance extraction circuit uses a transimpedance amplifier to convert
the input current into voltage and then extracts impedance by the lock-in technique [81].
-8

This chip is capable of concurrently measuring admittance values as small as 10

Ω

-1

within the array with the detection dynamic range of more than 90 dB in the frequency

34

range of 10 Hz - 50 MHz [81]. An array of on-chip gold electrodes have been implanted
in the developed chip, which takes advantage of the coherent detector to measure the
impedance of the associated electrode-electrolyte interface [81, 82].
An impedance spectroscopy DNA analyzer with dual-slope multiplying ADC has
been reported [83]. The developed EIS circuit, as shown in Figure 2-16, can extract the
analyte impedance in one cycle of the interrogation. It also includes an on-chip signal
generator to generate a sine waveform as the stimulus signal in the EIS measurement.
The performance of several impedimetric readout circuits is listed in Table 2-3.

Figure 2-15. The structure of the lock-in ∑-∆ impedance extraction circuit. The circuit
extracts the imaginary portion of admittance when S=1 and real portion when S=0.
(adapted from [31])

35

Figure 2-16. Dual-slope multiplying ADC VLSI architecture.

36

Table 2-2. The performance summary of amperometric current readout circuitry.

Area/ch
annel

Power
per
chann
el

Onchip
electro
de

Onchip
bioint
erface
test

0.5μm,
5V

525μm2

1uW

Yes

None

120dB

0.5μm,
5V

400 μm2 None

Yes

Yes

50fA

60dB

0.5μm,
5V

None

11μW

None

None

Dualslope
ADC

~240pA

~50dB

0.25μm
, 2.5V

None

NA

Yes

None

Turn
er
[48]

Two
stages
OPA

10nA

32dB

5μm,
±5V

None

2mW

None

None

Naru
la
[61]

Current to
time
converter

100fA

116dB

0.5μm,
5V

None

0.13m
W

None

None

Reay
[62]

Integrator
+ dualslope
ADC

100fA

172dB

2 μm,
±5V

None

5mW

None

None

Brete
dual-slope
n
ADC
[63]

2.44pA

74dB

0.7μm,
5V

None

None

None

None

Stan
ac´ev
∑-∆ ADC
ic´[7
4]

100fA

140dB

0.5µm,
3V

None

3.4μ
W

None

None

Circuit
descriptio
n

Minimum
detection
line

Dynam
ic
range

Techno
logy

Ayer
s
[59]

Halfamplifier
structure;
CDS

110pA (at
2 kHz)

NA

Li
[49]

CDS
technique

1pA

Gore
[60]

∑-∆
ADC+

Levi
ne
[54]

37

Table 2-3. The performance summary of impedimetric readout circuitry

Yan
g
[31]
Man
icka
m
[81]
Jafar
i
[83]
Yúfe
ra
[80]
Goz
zini
[79]

Circuit
descriptio
n

Frequency
range

Lock-in
Σ-Δ ADC

10mHz10kHz

TIA +
mixer

10Hz5MHz

Dualslope
multplyin
g ADC
Mixer,
increment
al ADC
Mixer +
Σ-Δ ADC

Detection
Technol
current
ogy
range

Area/cha
Power
nnel

Onchip
electrod
es

0.5μm,
3.3V

5.4μW/c
hannel

None

330pA40uA

0.35μm, 0.001m
3.3V
m2

84.8mW
(@100k
Hz)

Yes

0.1Hz 10kHz

100fA400nA

0.13μm,
0.06mm2 42uW
1.2V

10Hz50MHz

330pA40uA

0.8μm,
3V

None

0.35μm,
0.65mm2 60mW
3V

100Hz100MHz

38

None
None

6.95mm2 2.1mW
None

Chapter 3 CMOS Amperometric
Potentiostat and Readout Circuits
In order to achieve the miniaturized CMOS electrochemical biosensor array
microsystem concept, as shown in Figure 1-1, one of the goals listed in Chapter 1 is
development of CMOS electrochemical circuits with state-of-the-art performance in
terms of hardware efficiency, power consumption and bio-reaction current readout
resolution. To achieve this thesis goal, a CMOS amperometric potentiostat and current
readout circuit was designed and characterized, as described in this chapter. The new
CMOS amperometric readout circuit achieves to detection current down to 1pA that is
the state of the art resolution for amperometry biosensors. The readout and potentiostat
circuits were implemented in a monolithic amperometric instrumentation system with onchip electrodes and a protein bio-interface, and tested in a liquid environment. The
monolithic amperometric instrumentation system has demonstrated electrochemical
performance comparable to commercial benchtop instruments. Very good agreement was
observed between the commercial instrument and the CMOS amperometric
instrumentation system, which is reported in the test results Section 3.4.

39

3.1 Amperometric Instrumentation System
Structure
For an autonomous microsystem, the CMOS instrumentation would include all
necessary instrumentation electronics and a communication interface to allow user
control of measurement operations and reporting of measurement results. In the proposed
amperometric instrumentation, a potentiostat array is needed to control individual
electrode biasing in chemical reaction, a multi-function waveform generator is needed to
produce the stimulus signals necessary for amperometric techniques, and a highly
sensitive amperometric readout circuit is required to measure the response current. Figure
3-1 illustrates this arrangement of components and serves as the system diagram for the
CMOS amperometric instrumentation chip. Furthermore, an on-chip flat gold
microelectrode array was fabricated on top of the amperometric instrumentation chip and
the amperometric chip was packaged by parylene for liquid environments and harsh
chemical cleaning. The programmable waveform generator is designed by Waqar A.
Qureshi and the post-CMOS microfabrication is designed by Lin Li. The waveform
generator and post-CMOS microfabrication have been reported in [85], so they are not
covered in this chapter.

40

Figure 3-1. Block diagram of the amperometric instrumentation system.

3.2 Potentiostat Circuit
In generally, ±2V covers most of the electrochemical reaction range, so a
5V power voltage CMOS process is selected for the instrumentation circuits. The
basic function of the potentiostat circuit is to control and maintain the biasing
voltage for electrodes. The detail analysis of the potentiostat circuit function and
requirements have been presented in Chapter 2. In this chapter, a single-end
potentiostat circuit has been developed to support 3-electrode amperometric
biosensor system. There are three operation amplifiers in the potentiostat.
Compared with one amplifier potentiostat, this structure can adjust the offset
between and the stimulus signal and the potential at RE. The schematic of the
proposed potentiostat has been shown in Figure 3-2. OP1 and OP3 are standard
one stage cascode amplifiers because there are no special requirements for them
and the one stage amplifier saves area. Vref2 is a DC offset control input to adjust
the offset between and proposed stimulus signal and VRE. Output from the

41

waveform generator (VDAC) is level shifted by OP1 to match the input range of
OP2 by summing it with Vref1. The output of OP2, which is buffered through
negative feedback using OP3, is connected to the CE. For microscale sensors, the
solution resistance between RE and CE can be assumed to be negligible so that
CE and RE are at the same potential. The input of OP3 is connected to RE and no
current flows through OP3. The output current resulting from the input stimulus
voltage at CE can flow only between CE and WE. OPA1 and OPA3 are standard
one stage amplifier.

Figure 3-2. Structure of the potentiostat designed for amperometric biosensor.

One consideration in the potentiostat design is the stability. Because the range of
the potentiostat load (including the analyte, solution, transducer, connection wires etc.) is
wide (a couple of pF to tens of pF), there is a potential stability problem for a potentiostat

42

with a large load. OPA2 is the critical amplifier because it is the driver and may cause the
instability issues for large loads. An OPA with enough phase margins is needed to avoid
the instabilities. As we know, a compensation capacitor (CC) has been widely used in
OPA for increasing the phase margin by splitting the poles and moving the second pole to
far away from the original point. Another way to realize the same pole splitting is
carefully set the value of gm2 (the transconductance of the second stage of the OPA). The
second pole frequency is determined by gm2/(CL), where CL is the load capacitor. If gm2
is large enough, the second pole can move away the first pole no matter the value of CL,
which increases the phase margin of the OPA. There are two ways to increase gm2:
increase the drain current or increase the second stage transistors size. The large drain
current in the second stage is also expected for driving large load. In summary, increasing
compensation capacitance Cc and/or the second stage transcendence always help for the
stability in large load condition. The structure of OPA2 is presented in Figure 3-3. The
Cc=1pF, the size of the second stage transistors, Ma, Mb are W/L=96µm/0.6 µm. This op
amp (OP2) was simulated in Cadence with a phase margin of 68º with a large load of 5nF.
The simulation results showed this design is stable to such large loads as shown in Figure
3-4.

43

Figure 3-3.Structure of the OPA2 used in potentiostat.

44

Figure 3-4. The phase margin simulation result of OPA2.

3.3 Readout Circuit
The important electrical parameter for the current readout circuit is the resolution.
For typical enzyme biosensors, the bio-response current range is hundreds of µA to few
2

mA with 1 cm electrode [86, 87]. For gated ion channel membrane proteins, the
2

detection limit is lower, around tens of nA with 1 cm electrode [88, 89]. It is easy to
calculate that the current for one disk electrode with 100µm diameter would be down to
the pA level in a biosensor array. Thus, the readout circuit noise significantly affects the
measurement resolution. For most amperometric biosensors, the stimulus signal

45

frequency is less than 100 Hz, and within this operating frequency, the readout circuit
noise is dominated by 1/f noise rather than thermal noise. Thus, in the amperometric
readout circuit design, the 1/f noise has to be considered and a solution to eliminate the
1/f noise must be found.
Correlated double sampling (CDS) and chopping techniques are the two methods
used to eliminate the 1/f noise. The concept of the CDS technique is that the 1/f noise is
sampled in the first phase and then subtracted in the second phase. The concept of the
chopping technique is modulation the 1/f noise into a frequency band outside the signal
band and which can be removed by a low pass filter. The inclusion of such low pass
filters increases the circuit complexity, power consumption, and area.
As an alternative to using low pass filters, a switched capacitor current readout
circuit using correlated double sampling (CDS) with reducing the 1/f noise and amplifier
DC offset is described in this section. The schematic of the CDS amperometric readout
circuit is shown in Figure 3-5, where a switched-capacitor (SC) charge integrator
converts the input current to a voltage. The voltage is sent to a programmable gain
amplifier (PGA) with a gain of C2/C3, where C2 is a digitally programmable on-chip
2

capacitor array. The area of the capacitors C2, C3 implemented for CDS is 0.019µm ,
which is significantly smaller than the typical area of a low pass filter. The output of the
PGA is sampled and held and then fed to an analog-to-digital converter (ADC) after a
low pass filter. The non-overlapping clock signals for switches ph1, ph2 and ph3 are
generated by an on-chip clock generator block. The output voltage of the current readout
circuit is given by

46

Vout =

I WE
C
• 2
f s • C1 C 3

(3-1)

where IWE is the current at the WE and fs is the frequency of switch ph2, C1 is the
integrating capacitor. Thus, the input current range can be adjusted to accommodate a
range of biosensors through programmable selection of the clock frequency and the PGA
gain. These programmable features can be used to calibrate sensors over time.

Figure 3-5. Schematic of amperometric readout circuit.

An on-chip clock generator was developed to provide precise time sequence control
for switches activation. The structure of the clock generator is presented in Figure 3-6.
The invertors used in the clock generator have large driving capability to keep a sharp
rising/falling slope. The delay cells determine the time delay between ph1 and ph2. In the
design, the time delay should be selected carefully because it limited the maximum

47

operation frequency of master clock. Figure 3-7 is the simulation results of the ph1, ph2
ph3, ph1_b and ph2_b at 200kHz master clock. The clock generator has been simulated at
different master clock frequency to ensure its function as expected.

Figure 3-6. The schematic of the clock generator to generate ph1, ph2 and ph3 signals.

48

Figure 3-7. The simulation results of the time sequence of ph1, ph2, and ph3 generated by
the clock generator.

Several improvements have been implemented in the readout circuit to increase
measurement sensitivity. In the CDS technique, the noise reduction factor is typically
determined by device matching, timing errors, charge injection and clock feedthrough. To
curtail charge injection, all switches were realized with minimum size transistors because
the charge injection error is proportional to W•L (W is the channel width, L is the
channel length). Clock feedthrough errors are the results of the coupling between the
clock signal and the analog signal passing through the switch [90]. To reduce clock
feedthrough errors, the NMOS switches were replaced by a structure with one NMOS

49

switch and two NMOS dummy transistors. The schematic of the switch is shown in
Figure 3-8. M2, M3 are the two NMOS dummy transistors with half width of transistor
M1. A delay cell is inserted to keep the control signals: Cntl and Cntl_b have exactly the
reverse phase. The clock feedthrough simulation results of the designed switch are
presented in Figure 3-9, where the error voltage caused by clock feedthrough is only
2.5mV with around 0.2pF load capacitance generated by switches’ parasitic capacitors.
This 2.5mV error can be reduced by having larger load capacitors at the switch output
point. In our design, the switches’ load capacitors are at least 2pF; Because the error
voltage is inversely proportional to the load capacitors, the voltage error is easily less
than 1mV.

Figure 3-8. The schematic of the switch used on the readout circuit.

50

Figure 3-9. The simulation results of the switch clock feedthrough cancellation by
dummy NMOS transistors.

Noise in the amplifier also directly affects the sensitivity of the readout circuit. A
folded cascode amplifier was chosen because it provides high DC gain and wide dynamic
range. According to the noise analysis in [91], the size of the input transistors is critical.
Increasing the value of WL and W/L can reduce the 1/f noise and thermal noise. However,
larger input transistors will lead to larger input bias current, which limits the sensitivity of
the readout circuit. An input bias current of hundreds of fAs at 25ºC was reported for an
input NMOS of W/L = 150 µm /1.8µm [91]. To optimize the tradeoff between noise,
speed and input bias current, this design sets W/L = 72 µm /1.8µm for the NMOS input
transistors. The structure of the operational transconductor amplifier (OTA) used in
readout circuit is shown in Figure 3-10. This is a single stage fold-cascode structure,
which can achieve wide bandwidth and high gain. To further improve noise performance,

51

all of the analog circuits were surrounded by capacitors between VDD and GND to
minimize power supply noise, and the analog and digital power routings were separated
over the whole chip to minimize digital clock noise coupling onto analog signals.
Through these improvements, the sensitivity was increased by a factor of six compared to
prior work [91].

Figure 3-10. Schematic of OTA in amperometric readout circuit.

3.4 Test Results
The amperometric potentiostat and readout circuit array was simulated by Cadence
Spectre simulator and was fabricated in AMI 0.5µm CMOS process. The 3×3mm

52

2

amperometric chip shown in Figure 3-11, includes the multi-function waveform signal
generator and a 4-channel potentiostat and amperometric readout circuit array.

2

Figure 3-11. The 3×3mm CMOS amperometric instrumentation chip with waveform
generator and 4-channel potentiostat and amperometric readout array.

A printed circuit board (PCB) was developed for the amperometric potentiostat and
readout circuit characterization. In the PCB design, elimination the noise generated by the
PCB is the most important thing for characterizing the resolution of the readout circuit.
To reduce the noise generated by the power source, a battery was used to provide the
supply voltage for the amperometric chip. Also, several decoupling capacitors were
added between Vdd, analog ground and the digital ground to reduce the noise. Moreover,
a metal electrical junction box connected to the earth ground is implemented to isolate the

53

environment noise. BNC connectors were used for connecting the input pin of the readout
circuit. The 2-layer PCB was fabricated by Advanced Circuits using an FR-4 board with
0.062" contact spacing and a lead free solder finish.
To characterize the voltage range, scan rate, and resolution of the potentiostat
circuit, a triangle waveform generated by a National Instruments data acquisition (DAQ)
card 6259 with programmable amplitude and frequency was used. The waveform from
the DAQ board was connected to the VDAC pin of the potentiostat. The output signal at
VCE pin was recorded by the DAQ 6259 card. The triangle waveform was swept from 05V and the frequency was adjusted to ensure its scan rate would cover the interested
10mV/s - 10V/s range. The test results shown in Table 3-1, presents that the potentiostat
achieve the 10mV/s - 10V/s scan rate target. However, the voltage rang is ±1.8V, which
is less than the ±2V range expectation. In Chapter 5, a rail-rail potentiostat will be
introduced, which overcomes the ±1.8V voltage limitation.

Table 3-1. Potentiostat characteristics.
2

Area

0.4mm

Voltage range

0.7-4.3V

Scan rate

1mV/s-10V/s

Resolution

10mV

To characterize the range and resolution of the amperometric readout circuit, the
input current was swept using a Keithley 6430 sub-femtoamp source meter at different

54

clock frequencies and gain settings to span the functional input current range. Within
each range, the maximum deviation between measured current and the linear input source
was recorded as the linearity error. Table 3-2 shows a set of test results spanning the
circuit’s functional input range, with range settings from ±47pA to ±47µA. Within each
setting, the input range defines the max current before the op-amp response becomes
nonlinear. The 200Hz to 2MHz data show that the resolution decreases roughly linearly
with clock frequency, and at 2kHz and above the resolution is about 0.3% of the range.
Figure 3-12 shows test results where the maximum resolution of ±500 fA is achieved
with a gain setting of 2 with a 20Hz clock; leakage currents prevent operation at lower
frequencies. The power consumption of the 4-channel instrumentation circuit could not
be measured independently of test circuits. However, the amperometric chip was
simulated to be 22.5mW at a maximum clock frequency of 200 kHz.
The signal generator was tested by my colleague, Waqar A. Qureshi [85]. The
programmable waveform generator can output a constant potential, triangle waveform,
pulse, and sinusoidal waveform. Table 3-3 summarizes the characteristics of the
waveform generator.
Table 3-2. Amperometric readout circuit characteristics.
Clock
frequency

20 Hz

200 Hz

2 KHz

20 KHz

200 KHz

2 MHz

Gain setting

2

1

1

1

1

1

Input current
range

±47 pA

±4.8 nA

±48 nA

±489 nA

±4.8 µA

±47 µA

Linearity
error

1pA

13pA

150 pA

1.5nA

15nA

190 nA

55

Figure 3-12. The deviation between input current and measured current for 1pA to 30pA
inputs at 20Hz updating clock and gain=2 setting.

Table 3-3. Waveform generator characteristics
Area

0.44mm

2

DAC Resolution 10 bit
Scan range

0- 2V pk-pk, 10-bits

Scan rate

1mV/sec-1000V/sec

Step size

2mV

The final packaged amperometric chip with electrode array is shown in Figure 3-13.
A 2×2 gold electrode array was fabricated on the surface of the amperometric chip. The
chip was packaged with parylene for compatibility with liquid test environments. The onchip electrodes and parylene packaging were designed and fabricated by Lin Li [85].

56

Figure 3-13. Photograph of a CMOS biosensor array chip-in-package and close up views
of the post-CMOS surface electrode array.

To verify electrochemical measurement capability of the integrated system, a test
setup composed of an amperometry chip, a packaged on-chip electrode array, and a PC
with a DAQ card and a LabVIEW user interface was prepared. A typical electrolyte
solution with 1M potassium chloride and 0.5mM potassium ferricyanide was prepared,
and cyclic voltammetry measurements were performed at 25ºC using an on-chip WE, a
commercial liquid junction Ag/AgCl RE and a platinum CE. Figure 3-14 shows the
results from both a commercial potentiostat (CHI760C, CH Instruments Inc.) and the AIC
at scan rates of 100mV/s and 200mV/s. In cyclic voltammetry, the peak locations give
important information for biochemical identification, and these results demonstrate that
the peak locations and amplitudes of the CMOS system compare extremely well with the
commercial system even at different scan rates. In another experiment, a second solution

57

with 1mM potassium ferricyanide was prepared and cyclic voltammetry measurements
were performed. Figure 3-15 shows results from both the commercial potentiostat and the
reported CMOS amperometric system at two different electrolyte concentrations. As
expected, the peak current increased with electrolyte concentration, and again the peak
locations and amplitudes of the CMOS system compare extremely well with the
commercial system. These experiments verify the proper operation of the AIC, the
functionality of the post-CMOS electrodes, and the suitability of the parylene-based
packaging for operation in liquid environments.

Figure 3-14. CV measurement of 0.5mM potassium ferricyanide at 100mV/s and
200mV/s for both the CHI 760 commercial instrument and the proposed circuit.

58

Figure 3-15. CV measurement of potassium ferricyanide at 0.5mM and 1mM for both the
CHI 760 commercial instrument and the proposed circuit.

To demonstrate functionality of the microsystem as an electrochemical biosensor, a
cytochrome protein biointerface was formed on the on-chip electrodes of the CMOS
microsystem. Cytochrome PpcA from Geobacter sulfurreducens was expressed
heterogeneously in E. coli and purified using cation exchange chromatography. The
biointerface was prepared by cleaning the WE with piranha solution and contacting it
with ethanolic solution of 5 mM 11-mercaptoundecanoic acid and 11-mercaptoundecanol
(1:3 ratio) to form a self-assembled monolayer. Then, cytochrome PpcA was
immobilized on the self-assembled monolayer by covalent bonding using 50mM (N-(3dimethylaminopropyl)

-N'-ethylcarbodiimide)

and

5mM

(N-hydroxysuccinimide)

solution. Unbound cytochrome PpcA was rinsed away using 20mM phosphate buffer. CV
electrochemical characterization was performed using an on-chip WE, a commercial

59

liquid junction Ag/AgCl RE and a platinum CE. The cytochrome PpcA results for both
the CHI760 instrument and the reported amperometric circuit are shown in Figure 3-16.
To determine the ability of the immobilized cytochrome biosensor interface to detect
dissolved iron, different concentrations of Fe-NTA (Ferric nitrilotriacetate) in phosphate
buffer were placed in a reservoir formed on top of the CMOS chip. A dose-dependent
increase in peak reduction current with increasing iron concentration confirmed
suitability of the PpcA interface as a biosensor. Very good agreement was observed
between the commercial instrument and the CMOS microsystem.

Figure 3-16. CV measurement of cytochrome PpcA interface in 150μM and 400μM FeNTA for both CHI 760 commercial instrument and AIC chip.

Tests were also performed to characterize the measurement variation arising from
differences in readout channels and electrodes across the array. Variation results will be
reported as µ (1±σ/µ) where µ and σ are mean and standard deviation, respectively.

60

Variations in readout channels occur due to intra-chip process variations. When four
readout channels on one chip were tested with a 20µA input current, the results were
20(1±0.8%)µA. Variations in electrodes could be caused by variations in the
photolithography process, alignment errors, or differences in the cleanliness of electrodes.
When four electrodes on the same array were tested electrochemically, oxidation and
reduction peak separations were found to be 0.09 (1±2.2%) V. For comparison, peak
separation was measured for a single electrode while sweeping the potassium
ferricyanide potential for 15 cycles, and variation in peak separation was recorded as
0.074(1±4%) V. These results indicate that the observed intra-chip variations in circuits
and electrodes are less significant than variations caused by the electrochemical solution.
The long term stability of CMOS biosensor arrays are potentially limited by a)
electronics, b) packaging, c) biointerface. Absent radiation and temperature extremes that
are not common to biosensor applications, CMOS electronics stability will be limited by
chemical interactions with the liquid environment, which must be protected by the
packaging. The reported parylene packaging was chosen specifically for its ability to
provide this protection, and a parylene coated CMOS chip has been reported to remain
stable in 77 °C saline for more than six months [92]. Thus, stability of biosensors based
on the reported CMOS array likely are limited by the performance of specific
biointerfaces chosen to implement the sensor.

61

Chapter 4 CMOS Impedance
Spectroscopy Instrumentation
One of the goals of this dissertation is the development of electrochemical circuits
with state-of-the-art performance in terms of hardware efficiency, power consumption
and current readout resolution. As discussed in Chapter 2, the primary types of
electrochemical instrumentation are amperometric and impedimetric circuits. Chapter 3
covered a new high performance amperometric circuit, and the high performance
impedimetric readout circuit developed for this thesis research is covered in this chapter.
Impedance spectroscopy is a powerful tool for bio-interfacial characterization and
the bio-interfacial impedance change caused by biorecognition events. This chapter
describes the design and implementation of an impedimetric readout circuit provides a
programmable current measurement range from 100pA to 100nA with a measured
2

resolution of ~100fA. Occupying only 0.045mm per measurement channel, the hardware
2

efficiency circuit is compact enough to include nearly 200 channels in a 3 x 3mm die
area. The power consumption of one impedimetric readout circuit is only 5.2 µW at
200kHz clock frequency and 3.3V supply voltage, which demonstrates the efficiency of
the impedimetric readout circuit. A total 26 impedimetric readout channels were
integrated in a prototype impedance spectrum array (ISA) chip that was fabricated in
0.5μm CMOS. The chip was implemented in a monolithic impedimetric instrumentation
system with on-chip electrodes and packaging for a cellular membrane interface

62

characterization in a liquid environment. Test results of the monolithic systems are shown
at the end of this chapter.

4.1 Impedance Spectroscopy Methods
For impedance biosensors, biorecognition events alter the phase and amplitude of
the stimulus signal in a manner that is algebraically related to real and imaginary
components of the sensor’s instantaneous impedance. The phase and amplitude shift
caused by biorecognition can be extracted in the digital domain using a Fast Fourier
Transform (FFT) algorithm [30] that utilizes a broadband stimulus, such as a pulse, and
computes the result at all frequency points simultaneously, producing an impedance
spectrum. The composite signal source and the computationally intensive Fourier
transform require FFT-based impedance spectroscopy instrumentation to use a digital
signal processor (DSP) with extensive computational resources [76]. Alternatively, in the
analog domain, sensor phase and amplitude can be extracted using the frequency
response analyzer (FRA) method. FRA processes the response of one frequency point at a
time as it sweeps over the frequency range of interests. Compared to FFT, the FRA
method can be realized with simple and compact analog circuits, making it more suitable
for biosensor arrays [76]. As a result, most of the reported CMOS impedance extraction
circuits have used FRA methods [77-83].
In the FRA-based impedance measurement method, a sinusoidal stimulus signal
with known frequency is applied to the sensor interface. Thus, two main functional
blocks are needed in FRA impedance spectroscopy instrumentation: a sinusoidal

63

frequency synthesizer and an impedance readout circuit. The frequency synthesizer
serves to generate a sinusoidal stimulus with programmable frequency and amplitude,
and the impedance readout circuit extracts the phase/amplitude or real/imaginary shift.
To achieve phase/amplitude extraction, all reported FRA readout circuits have
utilized the lock-in technique. Lock-in is achieved by multiplying the biosensor’s AC
response signal with a reference signal matched to the frequency of the stimulus. This
multiplication results in a DC output that is proportional to the real/imaginary portions of
the AC signal under investigation. Figure 4-1 illustrates how the lock-in technique
extracts phase/amplitude or real/imaginary response components in the lock-in based
FRA algorithm. A frequency synthesizer generates the sinusoidal voltage sin(ωt) as
stimulus signal. The bio-cell response signal is Asin(ωt+θ) with phase shift θ and
amplitude shift A. After multiplying separately with reference signals sin(ωt) and cos(ωt),
the DC component of the products are proportional to the real and imaginary parts,
respectively, of the bio-cell response signal. This function can be expressed as
୅

A sinሺωt + θሻ sinሺωtሻ = ሾcos θ − cosሺ2ωt + θሻሿ = cosሺθሻ −
A

ଶ

୅

2

୐୔୊ 1

ଵ

	 cosሺ2ωt + θሻ 	 ሳልሰ A cosሺθሻ = Reୠ୧୭
ଶ

ଶ

2

୅

(4-1)
୅

A sinሺωt + θሻ cosሺωtሻ = ሾsin θ + sinሺ2ωt + θሻሿ = sinሺθሻ +
ଶ

୅

୐୔୊ ଵ

ଶ

ଵ

	 sinሺ2ωt + θሻ ሳልሰ A sinሺθሻ = Imୠ୧୭
ଶ

ଶ

ଶ

64

(4-2)

where Rebio and the Imbio are the real and imaginary portions of the bio-cell impedance,
respectively. ω is the stimulus frequency, and Acos(2ωt+θ) and Asin(2ωt+θ) are higher
order frequency components that can be removed by a low pass filter (LPF) [31, 81].
Based on (1) and (2), the amplitude shift A and phase shift θ can be calculated by

A = ටReଶୠ୧୭ + Imଶୠ୧୭
∅ = tanିଵ

୍୫ౘ౟౥
ୖୣౘ౟౥

(4-3)

(4-4)

Figure 4-1. Illustration of the lock-in based FRA impedance component extraction
method.

The reference signals sin(ωt) and cos(ωt) can be replaced by rectangular waveform
having the same phase as sin(ωt) and cos(ωt) without loss of performance because the
higher odd harmonics caused by the rectangular waveform substitution would be
eliminated by the included LPF. As a result, no cos(ωt) generator is need within the
frequency synthesizer, which simplifies the complexity of this circuit. The ISA chip

65

described in this chapter utilizes this lock-in technique with rectangular sin(ωt) and
cos(ωt) reference signals.

4.2 CMOS Implementation
Figure 4-2 illustrates the biosensor array microsystem that serves as the conceptual
model for the work in this chapter. The ISA chip acts as the substrate for the microsystem.
An array of biosensor electrodes are fabricated post-CMOS on the surface of the chip.
The SFS circuit generates stimulus signals for the bio-cell array, and the IRC circuit
measures the bio-cells’ response signals. The ISA chip communicates with a PC to send
control commands and analyze data from the ISA chip. To support electrochemical
measurements in a liquid environment, the ISA chip is packaged using a parylene coating
process following post-CMOS deposition of an electrode array on the surface of the chip.
The ISA chip includes a digitally programmable sinusoidal frequency synthesizer (SFS)
and multiple channels of a compact impedance readout circuit (IRC). The IRC cell
includes analog domain impedance extraction and local digitization while remaining
compact enough to be instantiated for each sensor element, enabling massively parallel IS
interrogation. The ISA chip was combined with an electrode array fabricated on the
CMOS chip to form a monolithic measurement system.

66

Figure 4-2. Conceptual illustration of the impedance spectroscopy circuit with SFS and
IRC circuitry and on-chip electrode and biointerface array. The chip communicates with
a PC to upload configuration data and download digitized measurement results.

4.2.1 Frequency synthesizer
Many circuits for sinusoid frequencies have been reported, including DSP-based
solution [93], quadrate oscillators [93, 94] and triangle-to-sine generators [95]. The
drawback of the DSP-based solution is the large memory requirement for very low
frequencies. The limitations of the quadrate oscillators and triangle-to-sine generators are
their poor linearity and difficulty in precisely controlling frequency. Our group reported
an alternative method that is controlled purely by digital signals and produces excellent
frequency accuracy and linearity [96]. It provides seven orders of frequency-tuning range
(1 mHz to 10 kHz) appropriate for impedance characterization of many biosensor
interfaces. This SFS topology employs a hybrid architecture, integrating two different
structures, one for low frequencies and one for high frequencies, to accurately cover a
wide range of frequencies. For high frequency (above 100Hz), an unevenly tipped
resistor chain with tunable updating clock was employed. At lower frequency (below
100Hz), a subsampling method was utilized.

67

The SFS reported in this chapter is based on our reported hybrid structure.
However, by utilizing rectangular waveform reference sin(ωt) and cos(ωt) signals, as
described in Section 4.1, the ISA chip eliminates the need for an analog sinusoidal cosine
frequency generator. This permits significant reduction of circuit area, increasing
potential density of sensor elements on the chip, without loss of performance. The block
diagram of the SFS is shown in Figure 4-3. In the R-chain based high frequency sinusoid
generator, the resistive DAC (R-DAC) is a 1000 identical unit-value resistor chain with
sinusoidally spaced taps. These taps are selected by the output of the token ring in the
proper sequence to generate the next sine wave sample. The token ring also determines
the state of the digital sin(ωt) and cos(ωt) (square waveform in phase with sin(ωt) and
cos(ωt)). A second order Butterworth filter was implemented using a gm-C biquadratic
filter to remove high frequency quantization replicas from the R-DAC output. Its cutoff
frequency is adjustable (2kHz-200kHz) by programming both the gm and capacitor
values. The R-DAC output is also provided as an input to the subsampling signal
generator, which samples this signal at a frequency that is very close to the signal
frequency, i.e., the frequency difference is very small. The output of the sampling circuit
contains a low-frequency replica that represents the frequency difference between the
input signal and the sampling clock. A detailed analysis was presented in [96]

68

Figure 4-3. Block diagram of the signal generator that produces stimulus signal sin(ωt)
and the reference square wave Dsin(ωt) and Dcos(ωt) (online version in color).

4.2.2 Impedance readout circuit CMOS design
The IRC presented here is based on the lock-in FRA method described in above
where the main functional blocks are a multiplier and an LPF. The real and imaginary
components are represented by only the DC output. All of the high frequency harmonic
noise after the multiplier has to be removed by the LPF, so the performance of the LPF is
critical for the resolution of impedance extraction. Thus, an LPF with a low cut off
frequency and a high rate of frequency roll off is required. To relax the challenge in
achieving such difficult requirements, a hybrid method we previously reported [31] was
adopted. This method combines the lock-in and oversampling techniques to achieve a
high sensitivity impedance readout circuit. The oversampling technique moves the low
frequency noise to high frequency where it can be removed easily by an LPF with less

69

stringent design requirements. The impedance readout circuit in [31] includes an analog
multiplier, an integrator, two comparators and two bidirectional counters. The integrator
achieves the LPF function. The two comparators and two bidirectional counters were
utilized for calibration of the mismatch between two current references. In the new IRC
reported in this chapter, the current reference mismatch is reduced by increasing the WL
size of transistors within the current references. Thus, the DC offset caused by reference
mismatch can be calibrated by a digital calibration method and then removed from the
digital output [97, 98]. As a result, only one comparator and one bidirectional counter are
needed in the IRC, as shown in Figure 4-3. Furthermore, an additional block from the
prior design was eliminated, and together these changes make significant improvements
to hardware efficiency and power consumption.
The first stage of the IRC is a multiplying integrator stage that is based on previous
work in our lab [31]. It shares resources to realize both the multiplication and integration
functions shown in Figure 4-4. The multiplying integrator changes its polarity according
to the value of the square wave φ. In the IRC, if the square wave reference φ(t) is in
phase with sin(ωt), then φ(t) can be described by

φሺtሻ = sgnሺsinሺωtሻሻ = ൜

1, nT ≤ t < ሺ݊ܶ + T⁄2ሻ
(4-5)
−1, nT + T⁄2ሻ ≤ t < ሺn + 1ሻT

where T is the period of the stimulus signal and n is any integer. If the biosensor response
current Iin is multiplied by (4-5) and integrated over N continuous stimulus cycles, then
the integrator’s output is given as

70

( i +1)T



I in dt − ∫ I in dt 
∫0 I in ⋅ ϕ (t )dt = ∑
∫


i =0 
iT
( i +1 / 2 )T

2T
2T
=
⋅ N ⋅ A cos(θ ) =
⋅ N ⋅ Re bio
N −1 ( i +1 / 2 )T

NT

π

π

(4-6)

where I୧୬ = Asinሺωt + θሻ. Thus, the real component, Rebio of the biosensor response
signal can be extracted with a multiplier, an integrator, and reference square-wave φ(t) that
is in phase with sin(ωt).
Based on (6), if the ‫ ׬‬I୧୬ dt in different integration cycles can be calculated, then the
real component can be determined. To simplify analysis, it is assumed that φ(t) is in
phase with sin(ωt) and that both comparator results, D and D*, are low at all of φ’s
transition edges. If not, the multiplexing switches controlled by φ will force D and D* to
exchange their polarity of charge injection through reference current Iref1 or Iref2. From
time 0 to T/2, φ is high and the counter is counting up. Just before φ’s edge at time T/2,
according to the charge conservation theory, at the input node, we have

T/2

N2

N2

0

i =1

i =1

*
I
⋅
dt
=
CV
+
I
T
D
−
I
T
D
res1
ref1 0 ∑ i
ref2 0 ∑ i
∫ in
(4-7)

where Vres1 is the residue value at the integrator output, T0 is the updating clock period,
and N is the number of clock cycles from time 0 to T/2. From time T/2 to T, φ is low, the

71

integrator capacitor is reversed, and the counter is set to down counting mode. This mode
produces
T

∫ I in ⋅dt = C(Vres1 + Vres2 ) + I ref1T0
T/2

N

∑D

i

− I ref2 T0

N 2 +1

N

∑D

*
i

N 2 +1

(4-8)

where Vres2 is the integrator’s output voltage at time T. Simplifying the equation with
D*=1-D and combining (4-7) and (4-8), the full cycle produces

N 2
 N2
∫T/2Iin ⋅dt = −CVres2 + T0 ⋅ (I ref1 + I ref2 ) N∑2+1Di − ∑1 Di
T




 (4-9)

Thus, the real result is represented by the contents of the counter (the summations
in

(4-9)),

provided

the

circuit

1
CVres2 < min(Iref1T0, Iref 2T0)
2

parameters

are

chosen

such

that

where the residue on the integrator can be treated

as noise. If the IRC is operated for N consecutive stimulus cycles after initial reset, the
residue term in (4-9) remains in the same range while the digital part is magnified by N.
Therefore, the error due to ignoring the residual will decrease with repeated cycles. The
real component after N cycles can be expressed as

72

Rebio =
=

πT0
2T

π

NT

2TN

⋅ ∫ I in (t ) ⋅ ϕ (t )dt
0

⋅ ( I ref 1 + I ref 2 ) ⋅ (

N

N /2

∑D − ∑D )
i

N / 2+1

i

1

(4-10)

Following a similar analysis, it can be shown that the counter will contain the
imaginary portion of admittance when φ(t) is in phase with cos(ωt). The imaginary
portion can be presented by

Imbio =
=

πT0
2T

π
2TN

NT

⋅ ∫ I in (t ) ⋅ ϕ (t )dt
0
3N / 4

N /4

⋅ ( I ref 1 + I ref 2 ) ⋅ ( ∑ Di − ∑ Di −
N /4

1

N

∑D )
i

3 N / 4+1

(4-11)

Thus, by setting φ to be in or out of phase with the stimulus sinusoid, the IRC
circuit can extract both portions of the full impedance utilizing a structure that inherently
digitizes the result while sharing resources to minimize size and power. The bidirectional
counter also serves as a shift register to permit serial data upload of all channels.

73

Figure 4.-4. Simplified structure of the impedance readout circuit presented in [31] and
the IRC (except for gray part). φ and not-φ are non-overlapping reference square waves.

4.3 Test Results
The ISA chip was fabricated in a standard 0.5μm CMOS process, and photographs
of the parylene packaged chip with 10 × 10 on-chip electrode array are shown in Figure
4-5. The diameter of each on-chip electrode is 100µm. One SFS and 26 IRC channels
2

2

have been integrated on the 3 × 3mm chip. The area of each IRC channel is 0.045mm ,
which is 25% less than the area of impedance readout circuit reported in [31].

74

Figure 4-5. Photo of the packaged ISA chip with on-chip electrode array.

To characterize the performance of the IRC circuit, a DAQ 6259 was used to
generate the sinusoid stimulus voltage, reference square wave, global clock and digital
control signals and to record digital output results from the IRC on the second generation
thermal control chip. An equivalent impedance model for a biomimetic membrane sensor
interface [99] was utilized to characterize the chip. This model is shown in Figure 4-6
along with other elements of the IRC test setup.

75

Figure 4-6. Test platform for the IRC chip with biosensor equivalent circuit model.

To verify the impedance extraction capability of the IRC circuit, a sinusoidal
voltage stimulus with frequency from 0.1 to 100Hz was applied to the biosensor circuit
model. Figure 4-7 plots the results obtained from the digital output of the IRC along with
the theoretical curve for the model impedance. The IRC tracks both real and imaginary
components of the test impedance very well, with a maximum mismatch of 1% of the full
scale response.

76

Figure 4-7. The real (circles) and imaginary (squares) results from the IRC circuit for the
test setup shown in Figure 4-6 along with theoretical (expected) curve for comparision.

To characterize the circuit’s response to variable impedance, an IRC channel output
was recorded while independently changing the phase and amplitude of the stimulus
input. Ideally, the relationship between the input signal’s amplitude and the digitized real
component output is linear for a constant phase and frequency. To verify the amplitude
transfer function, a 10Hz, zero initial phase sinusoid signal was supplied with amplitude
swept from 0 to 10nA. The IRC digital output data was recorded by data acquisition card
(DAQ6259). The measured differential nonlinearity (DNL) and integral nonlinearity
(INL) characteristics of the digitized real portion output are plotted in the Figure 4-8,
which shows that the real portion dynamic range is more than 50.2dB for the 10nA input
range.

77

The relationship between the input signal’s phase and the digitized imaginary
component output is theoretically a sine wave for a constant amplitude and frequency. To
measure the phase transfer function, a cosine input with constant 100Hz frequency,
constant 3nA amplitude and variable phase was supplied. The normalized digitized
imaginary portion output is plotted in Figure 4-9. The output fits the expected sine wave
with an RMS error of only 0.02.

DNL (dB)

-40
-60
-80
-100
-120
-140
0

2

4

6

8

10

8

10

current (nA)

INL (dB)

-40
-60
-80
-100
-120
0

2

4

6

current (nA)

Figure 4-8. Measured DNL and INL curve of one IRC channel output when the sinusoid
input amplitude was swept from 0 to 10nA. The maximum DNL and INL are 59.8dB and
50.2dB, respectively.

78

Figure 4-9. Response of an IRC readout channel as the phase of the input signal is varied
from -180° to +180° degrees. Input amplitude (30nA) and frequency (100Hz) are kept
constant.

Table 4-1. The performance of impedance-to-digital converter circuit.
Power/channel

5.2 µW@ 3.3V power, fclk=200kHz

Area/channel

0.045 mm

Channel density

200 channels on a 3x3mm chip

Amplitude conversion resolution

7-8 bit

Phase conversion RMS error

<2%

Frequency range

10mHz ~ 10KHz

Maximum input current

100 nA

Maximum current sensitivity

100 fA

79

2

A tethered bilayer lipid membrane (tBLM) is useful in functional proteomics
research to characterize novel membrane proteins and can be used to develop membrane
protein biosensors [100, 101]. The equivalent circuit model of a tBLM interface is
presented in Figure 4-10, where, Rm and Cm represent the resistance and capacitance,
respectively, of the lipid bilayer and any embedded proteins, RS is the solution resistance,
and Cdl is the double layer capacitance [102]. The values of Rm and Cm can be
measured to characterize tBLM interface or interrogate any embedded membrane protein
sensing elements.

Figure 4-10. Schematic view and equivalent model of a two-electrode electrochemical
site with bottom electrode coated with a tBLM. In the tBLM equivalent model, Rm and
Cm represent the resistance and capacitance of the BLM assembly, respectively, while
RS is the resistance of the solution and Cdl describes the double layer capacitance.

80

To validate the capability of the IRC to measure a real biointerface, an example
tBLM was constructed. First, a self-assembled monolayer (SAM) of 1, 2-dipalmitoyl-snglycero-3-phosphothioethanol (DPPTE) tether lipid was formed on a clean gold electrode
that was patterned on a silicon chip. Then the upper leaflet of the bilayer membrane was
deposited by fusion of liposome vesicles made of 1, 2-dioleoyl-sn-glycero-3phosphocholine (DOPC) mobile lipids. A data acquisition card (Agilent 6259) was
connected to a computer and employed to generate the stimulus sinusoid and collect
output data from the IRC, from which the tBLM values of Rm, RS and Cm and Cdl can
be determined. Impedance measurements were conducted with the IRC circuit over the
frequency range of 10 mHz to 100 Hz after the tBLM was formed. Figure 4-11 shows the
amplitude and phase data obtained from the IRC as a function of frequency. For
comparison, the same biointerface was tested using a commercial impedance instrument
(CHI 660 B) and those results are included in Figure 4-11. These data demonstrate that
the IRC circuit generates similar results as commercial equipment for extraction of
biosensors impedances. The impedance difference between the commercial instrument
and IRC circuit measurement results is less than 5% across the frequency spectrum. The
real and imaginary impedance values extracted by the IRC were used to determine the
element values of the tBLM equivalent circuit model shown in Figure 4-10. After fitting
by ZView® software, the normalized values of each component are Rs = 95Ω, Rm =
2

2

2

211×103 Ω·cm , Cm= 0.23µF/cm , and Cdl = 2.87 µF/cm . These values are in good
agreement with reported values for tBLM [103]. Using state of the art techniques, BLMs
with impedance in the giga-ohm range have been reported [104]. The presented IRC chip
can detect impedance up to tens of giga-ohms (>1pA resolution at 10mV stimulus),

81

which is well beyond the limits of traditional instruments. Thus the IRC chip enables a
new capability to characterize high quality BLMs using impedance techniques.

Figure 4-11. tBLM amplitude and phase measured by a commercial instrument (CHI 660
B) and by the IRC circuit at different frequencies.

82

Chapter 5 Hardware Efficiency Multimode Electrochemical Readout Chip
A high performance amperometric circuit and a high performance impedimetric
readout circuit have been introduced in Chapters 3 and 4, respectively. One of the goals
listed in Chapter 1 is to efficiently integrate multi-mode electrochemical techniques onto
a single chip for protein characterization and sensing under widely varied controlled
conditions. To achieve this goal, a hardware efficient, multi-mode, electrochemical
instrumentation circuit has been developed, enabling amperometric and impedimetric
measurement

while

sharing

hardware

sources.

The

developed

multi-mode

electrochemical readout circuit significantly reduces the circuit area and the power
consumption for a miniaturized high throughput array microsystem.

5.1 Design Goals
Amperometry and impedimetry are the two dominating electrochemical techniques
used for protein characterization and sensing. These two electrochemical techniques
provide interdependent results in many assays. For example, in antigen/antibody
biosensors, the amperometric technique is firstly used to verify/confirm the probeinterface formation is as expected. Then the impedimetric technique is applied to detect
the concentration of the targets. Moreover, more than one electrochemical technique is

83

generally required for detecting multiple types of protein array. In summary, a multimode electrochemical instrumentation circuit is desired for biosensor arrays.
In general, the speed of the bio-reaction is less than kHz, so a tens (tenth or tens?)
of kHZ bandwidth of the current readout circuit would ensure that the circuit is quick
enough for monitoring the bio-reaction. The biosensor current response value is various,
depending on the type of biointerfaces and the electrodes size. For a high throughput
array microsystem, the response current could be small due to the small area of the
electrodes and the low volume of solution. The response current can be as low as pA~fA, which brings challenges for the circuit design. Another parameter that needs to be
considered during design is the area of the circuit. Area is critical for the chip-scale
biosensor array microsystem because hundreds of the multi-mode readout circuits would
be integrated into a 5x5mm2 chip. Finally, the low power consumption is always
expected for an electrochemical instrumentation.
The goals of the multi-mode electrochemical instrumentation circuits are: 1) design
a CMOS electrochemical instrumentation circuit that supports both amperometric and
impedimetric measurement techniques; 2) utilize hardware sharing to minimize hardware
resources for array implementation; 3) achieve low power and high resolution readout
performance.

5.2 Chip Architecture
To achieve the goals defined in Section 5.1, a multi-mode electrochemical readout
chip (MERC), as shown in Figure 5-1, was designed. There is a potentiostat in the chip,

84

which generates the bias potential between RE and WE. The bias generator (BiasGen)
circuit generates all the internal reference voltages/currents for the other circuits. Multimode electrochemical current detection is achieved by the multi-mode current readout
(MCR) blocks. The clock generator provides the non-overlapped clocks for switchedcapacitor circuits within the MCR. The control circuit sets the chip operation mode and
the current range.
The most important block in the MERC chip is the MCR, which is responsible for
realizing most of the critical performance goals. The MCR is also the most suitable block
for realizing hardware sharing in order to minimize chip area, power consumption, and
cost. Most of the effort was applied to improve the performance of the MCR block.

Figure 5-1. Block diagram of the multi-mode electrochemical readout chip (MERC).

85

5.3 Multi-mode Current Readout Approach
Because a multiplier and an integrator are two essential components for FRA
impedance extraction [105], they have to be included in the MCR. Different structures of
ADCs, such as ∑-∆ ADC, dual-slope ADC (integrating ADC), and pipeline ADC, etc.,
are used to readout the DC current in commercial amperometric equipment. Among these
types of ADCs, only ∑-∆ ADC and dual-slope ADC include integrator components and
both can achieve high resolution. However, the structure of the dual-slope ADC
determines that it needs a long integration time to achieve the high resolution, which
limits the speed. In contrast, ∑-∆ ADC uses an oversampling technique to achieve high
resolution without sacrificing speed.
To readout current at the femtoampere level, the current readout circuit needs
strong noise filtering capabilities. A hysteretic comparator was added in the conventional
∑-∆ ADC structure to improve the circuit noise performance. The hysteretic comparator
reduces the converter switching cycles when operating with small magnitude currents, as
shown in Figure 5-2. This leads to a reduction in substrate noise interference and
improves linearity of the converter by reducing the number of switching operations
(concept shown in Figure 5-3) on reference current sources. A hysteresis lock-in ∑-∆
ADC, as shown in Figure 5-3, has been developed to achieve amperometric and
impedimetric functions. The amperometric and impedimetric modes are determined by
the switches S0 and S1. The current readout circuit is in amperometric mode when S0 is
open and S1 is closed. The current readout circuit is in impedimetric mode when S1 is
open and S0 is closed.

86

Figure 5-2. The hysteretic comparator reduces the undesired switching caused by noise.

Figure 5-3. The structure of multi-mode current readout channel. The blue path is
implemented for the impedimetric mode. The green path is implemented for the
amperometric mode.

87

5.3.1 MCR noise model
In this section, we analyze the effect of thermal noise on the performance of the
MCR and illustrate benefits of using a hysteretic comparator for low current
measurements. For simplified analysis, we only consider the case where the input current
and the integration capacitance in Figure 5.3 are being charged and discharged by only
reference currents. The thermal noise contribution due to current references and the
operational amplifier, can be summarized as a single equation
ଶ
I୬୭୧ୱୣ
= 4γkTg ୫ BW

(5-1)

where k is Boltzmann’s constant, T is absolute temperature, and g ୫ is the equivalent
transconductance that combines the effects of input referred noise due to the current
references and the operational amplifier. BW denotes the bandwidth of the converter.

γ=

ଶ
ଷ

for the above-threshold operation and

γ=

ଵ

ଶ	

for subthreshold operation. For

zero input current, the transition points of the integrator output are marked by events
where the hysteretic comparator changes its state (comparator switching cycle). The total
integration time between two comparator events can be approximated by

T୧୬୲ = 2

େ౟౤౪ ஔା୍౨౛౜
୍౨౛౜

(5-2)

The reference current noise is integrated to obtain an equivalent voltage noise
whose signal power is given by
୘౟౤౪ ଶ
ሻ
େ୧୬୲

ଶ
ଶ
V୬୧୭ୱୣ
= I୬୭୧ୱୣ
ሺ

88

(5-3)

For the sake of simplicity, BW is approximated by 1⁄T୧୬୲ which leads to a total
noise power of
ଶ
V୬୧୭ୱୣ
=

ସ୩୘ஓ୥ౣ
େమ౟౤౪

T୧୬୲

(5-4)

The signal power is given by the square of the integrated voltage and is at the
output of an integrator
ଶ
Vୱ୧୥

=

ሺ୘౟౤౪ ୍౨౛౜ ሻଶ
େమ౟౤౪

(5-5)

From (5-7) to (5-9), the signal-to-noise ratio (SNR) per integration cycle can be
computed as

SNR =

େ౟౤౪ ஔ୍౨౛౜
଼ஓ୥ౣ ୏୘

+

୍మ౨౛౜ ୘ౙౢౡ

଼ஓ୥ౣ ୏୘

(5-6)

where δ is the hysteresis magnitude of the hysteric comparator. For small values of
reference currents, transistors in current references and amplifiers can be biased in weak
inversion, where the equivalent transconductance

g୫ =

୩

୙౐

ሺI୰ୣ୤ + Iୠ୧ୟୱ ሻ ,

with
ଵ

Iୠ୧ୟୱ being the tail current of the operational amplifier. Substitute values of k = U୘
ଶ

for weak-inversion into (5-4), then the SNR achieved during a single comparator
switching cycle is given by

SNR =

େ౟౤౪ ஔ୙౐
଼஗୏୘

+

89

୍౨౛౜ ୘ౙౢౡ ୙౐
଼஗୏୘

(5-7)

where η

୍

= ሺ ౘ౟౗౩

ା୍౨౛౜

୍౨౛౜

ሻ, and U୘ is the thermal voltage. The SNR consists of thermal

noise contribution due to integration of reference current I୰ୣ୤ and is directly proportional
to the product I୰ୣ୤ Tୡ୪୩ . Decreasing I୰ୣ୤ in (5-7) has to be compensated by an increase
in the clock period Tୡ୪୩ to maintain a constant SNR. If δ in (5-6) is 0 V, then SNR
degrades as the reference current decreases. The addition of hysteresis ensures that the
SNRs between comparator switching cycles are at least

େஔ୙౐

଼஗୏୘

.

This is weakly

dependent on the reference currents and independent of the clock frequency. Thus, for
large operational currents, the SNR in (5-7) is dominated by conventional ∑-∆ operation,
whereas for a low operational current I୰ୣ୤ , the SNR is determined by the hysteresis
magnitude. In summary, the noise performance for small input currents (~nA) can be
improved by using a hysteretic comparator.

5.3.2 Amperometric mode
When switch S1 closes, the multi-mode current readout is in the amperometric
mode. In the amperometric measurement, the bandwidth of the input current is assumed
to be much less than the sampling frequency. Therefore, the time-domain analysis can be
simplified by considering only a DC input current, which will be denoted by Iin. The
structure of the hysteresis lock-in ∑∆ ADC working as an amperometric function was
shown in Figure 5-3.

90

During each clock cycle, the integrator voltage, Vn, (n=1, …N) is updated
according to:

V୬ = V୬ିଵ + ሺ

୍౟౤

୍౨౛౜

− D୬ ሻ

୍౨౛౜ ୘ౙౢౡ
େ౟౤౪

(5-8)

D୬ = signሺV୬ିଵ + D୬ିଵ δሻ

(5-9)

where Dn (D୬ 	 ∈ ሺ−1, +1ሻ ) is the comparator output; Tclk is the clock period; Iin is the
input current; Iref is the reference current; and Cint is the integrator capacitor.
The magnitude of δ plays a key role in the noise performance of the developed ∑-∆
ADC. When the magnitude δ

≪

୍౨౛౜ ୘ౙౢౡ
େ౟౤౪

, the integrator output exhibits significantly

high switching cycles, as shown in Figure 5-4 (a). After increasing the magnitude of δ,
the number of integrator switching cycles decreases significantly, as shown in Figure 5-4
(b). The reduction in switching cycles reduces substrate noise and integrator nonlinearity,
which are important for measurement of ultra-low magnitude input currents.

(a)

(b)

Figure 5-4. The output of integrator when (a) the hysteresis magnitude is 0V, and when
(b) the hysteresis magnitude is 100mV

91

5.3.3 Impedimetric mode
The structure of the MCR working as an impedance extraction function is shown in
Figure 5-6. Based on the concept of the lock-in technique explained in Chapter 2, the
signal φ and the stimulation sinusoidal is in-quadrature. To extract the real coefficient
information, the stimulation sinusoidal, sin(ωt), is multiplied by the reference signal φ,
which is in phase with sin(ωt). According to the analysis in [106], the extracted real
portion in N continuous stimulus cycles can be presented by

portion =

real
=

πTclk
Tstim

2Tstim N

N

⋅ ( I ref ) ⋅ (

NTstim

π

∫I

⋅

in

(t ) ⋅ ϕ (t )dt

0

N /2

∑D − ∑D )
i

i

N / 2+1

1

(5-10)

where Tstim is the period of the stimulation signal.
Following a similar analysis, it can be shown that the counter will contain the
imaginary portion of admittance when φ is in phase with cos(ωt). The imaginary portion
in N continuous stimulus cycles can be presented by

imaginary portion =
=

πTclk
Tstim

NTstim

π
2Tstim N

3N / 4

N /4

N /4

1

⋅ ( I ref ) ⋅ ( ∑ Di − ∑ Di −

92

∫I

⋅

in

(t ) ⋅ ϕ (t )dt

0

N

∑D )
i

3 N / 4+1

(5-11)

5.4 MERC CMOS Implementation

5.4.1 Potentiostat
The function of the potentiostat circuit is to control and maintain the voltage
between the working electrode and the reference electrode under varying current
conditions. Its structure is shown in Figure 5-5, where the stimulus signal was applied at
the positive input pin of OP1. The CE potential follows the stimulus signal by OP2 and
there is no current going through RE because RE is the input pin of OP3. In Chapter 3, a
potentiostat was developed with 3.6V input range, which is slightly less than the 4V input
range goal. To increase the input range of the potentiostat, a rail to rail amplifier was
implemented [107], which includes three stages: an input stage, a gain stage and a buffer
stage.

Figure 5-5. Schematic of the potentiostat.

93

5.4.2 MCR circuit
The schematic of the hysteretic comparator is shown in Figure 5-6. The two
threshold values of the hysteretic comparator are controlled by Vbpr and Vbnr. When
Vbpr equals Vbnr, the hysteresis magnitude of the comparator is zero and working as a
normal comparator. The hysteretic comparator is simulated by Cadence Spector simulator
at 25ºC with a TT (typical-typical) N/PMOS process corner [108] that represents the
typical mobility of the carrier (electron/hole) in N/PMOS. The simulation of the
hysteretic comparator is shown in Figure 5-7, where the hysteresis magnitude, δ, is
100mV.

Figure 5-6. Schematic of the hysteretic comparator.

94

Figure 5-7. The simulation results of the hysteretic comparator.

The schematic of the amplifier used in the integrator is shown in Figure 5-8. This is
one stage folded cascade amplifier which provides 87dB gain. The bidirectional counter
used in the MRC is from our previous work [31], which achieves up and down counting
by one counter. The BiasGen circuit was designed to support all the reference voltage for
other analog blocks.

95

Figure 5-8. Schematic of the amplifier used in the integrator.

5.5 Test Results

5.5.1 Circuits characterization
The multi-mode electrochemical readout chip was fabricated in a 0.5μm CMOS
process, and the die photograph is shown in Figure 5-9. A rail-rail potentiostat and 4
multi-mode readout channels have been integrated on a 1.5mm × 1.5 mm chip. The
testing printed circuit board (PCB) for MERC is shown in Figure 5-10. This PCB was
fabricated by Advanced Circuits in 2-layers, using FR-4 with a 0.062” pad spacing and a
lead free solder finish.

96

Figure 5-9. The photograph of the multi-mode electrochemical readout chip.

Figure 5-10. The printed circuit board for MERC characterization.

To calculate operation and the input range of the potentiostat, a triangle waveform
from 0 to 5 V as an input signal was connected to the input node of the potentiostat. The
output voltage of the potentiostat is plotted as a function of input voltage in Figure 5-11.
The effective input range of the potentiostat is 0.1-4.8V.

97

Figure 5-11. The test result of the potentiostat.

To characterize the resolution improved by the hysteretic comparator, the MRC
was tested with both 100mV and zero hysteretic magnitude. For a ±10 pA input current
with a 100pA current reference, the differential nonlinearity (DNL) results of the MRC
with and without the hysteresis magnitude (δ = 100mV) are presented in Figure 5-12.
One bit resolution is increased by the hysteretic comparator. This shows that noise
performance at small currents is improved by using a hysteretic comparator. The power
consumption of the MRC with the hysteretic comparator is also reduced because of a
corresponding reduction in the switching cycles. The performance of the MERC is listed
in Table 5-1.

98

Figure 5-12. The DNL results (represented as the least significent bit (LSB)) of the MCR
circuit with and without hysistersis.

Table 5-1. The performance summary of the MERC chip.
CMOS process

AMI 0.5µm

VDD

5V

Current range of the ±100nA
MRC
150fA (without hysteresis)

sensitivity

60fA (with 100mV hysteresis)
Input range
potentiostat

of

the 0.1V - 4.8V

5.5.2 Electrochemical test results
To verify the amperometry and impedimetry function of the MERC, the chip was
used to measure Herceptin, a therapeutic monoclonal antibody (mAb) for the treatment of
HER2-positive breast cancer [109]. Trials have shown that Herceptin reduces the risk of
recurrence of early HER2-positive breast cancer by approximately 50% [110-113].
Herceptin influences its effectiveness, so physicians need careful monitoring of Herceptin
99

levels in the patients. The test results of cyclic voltammetry and impedimetricy measured
by MERC are presented in this section.

5.5.2.1 Materials
Peptide CH-19 as a bio-probe was used to detect Herceptin in this assay. Peptide
CH-19 (primary sequence: CGSGSGSQLGPYELWELSH) is a short synthetic peptide
that has high specificity binding to Herceptin with rapid response time and low cost
advantages [114]. Synthetic peptide, designated CH-19, was purchased from Bio. Basic,
Inc. (Ontario, Canada) and received in lyophilized condition. The quality of CH-19 was
assessed by high performance liquid chromatography (HPLC) and confirmed through
matrix-assisted laser desorption/ionization (MALDI) mass spectrometry analysis (Purity >
95%). Therapeutic monoclonal antibody, Herceptin® (Trastuzumab) was provided by
Beaumont Hospital, Royal Oak, Michigan. The UltraPureTM distilled water (Cat. No.
10977-015) and phosphate buffered saline (PBS, Cat. No. 10010-049) were obtained
from Invitrogen Corporation. These chemical materials were prepared by Dr. Yuqin
Shang, Oakland University.
Interdigitated electrodes (IDE) are commonly used for impedance sensing
applications. [115]. IDE was used as the electrode in this assay. IDE was fabricated in Dr.
Mason’s lab. Work was completed by colleague Xiaoyi Mu. First, the silicon substrate
with 1.5um thick thermal oxide was cleaned by acetone, isopropanol and DI water for
1min each respectively, followed by dehydrating on hotplate at 100°C for 10min. Then
the substrate was spin-coated with Shipley 1813 photoresist at 3000rpm and soft-baked in

100

an oven at 95°C for 10 minutes. Afterward it was exposed through a designed photomask
by UV light at 365nm wavelength for 10 seconds using a mask aligner. Next, the
photoresist was developed in AZ352 developer for 45 seconds. Then the photoresist was
hard-baked at 120°C for 1 minute. A 50Å Ti (titanium) adhesion layer and 1000Å Au
(gold) layer were deposited on the substrate using Edward 360 thermal evaporator at
1×10

-6

Torr. The photoresist was then lifted-off by rinsing with acetone, isopropanol and

DI water. The geometry of the micro-fabricated electrodes is shown in Figure 5-13. The
gap between two fingers is 15µm and each electrode consists of 12 fingers with size of
300µm x 30µm.

Figure 5-13. The photo of the interdigitated electrode (IDE).

5.5.2.2 Test setup
To verify the electrochemical measurement capability of MERC, a test bench
composed of MERC, one IDE, and a PC with a DAQ card 6259, and a LabVIEW user

101

interface was developed. A triangle waveform (for CV measurement) and a sine
waveform (for EIS measurement) working as the stimulus signals for the Herceptin
interface, were generated by the DAQ card with programmable amplitude and speed. The
MCR digital output data was also recorded by the DAQ card 6259. The schematic of the
test bench is shown in Figure 5-14.

Figure 5-14. The test bench schematic for MERC chip with Herceptin bio-interface.

5.5.2.3 Surface immobilization
CH-19 immobilization onto an IDE surface was carried out as described for gold
electrode preparation with modifications [116-118]. First, the IDE surface was cleared by
one drop of a concentrated nitric and sulfuric acid mixture (1:1 v/v). Then, the IDE
surface was washed thoroughly by biological grade water, and pure ethanol (200 proof)
three times to remove impurities. The IDE surface was then dried under pure nitrogen
stream. In order to functionalize the IDE, the freshly cleaned IDE surface was immersed
in peptide CH-19 solution (1 mM in ultrapure water) and kept at 4 oC. After an overnight
incubation at 4 oC, the surface of the CH-19 modified IDE was rinsed thoroughly with
bio-grade water and PBS buffer (pH 7.4). The IDE surface functionalized with peptide

102

CH-19 was then ready for electrochemical measurements. The CH-19 immobilization
was developed by Dr. Yuqin Shang, Oakland University.

5.5.2.4 Measurement and results
To verify the C-19 formation performance, cyclic voltammetry (CV) was applied to
the CH-19 modified IDE. A typical electrolyte solution with 1M potassium chloride and
0.5mM potassium ferricyanide was prepared. The cyclic voltammetry measurements of
the modified IDE and bared IED were performed at room temperature at scan rates of
100mV/s. The bared gold IDE was measured by a commercial potentiostat (CHI760C,
CH Instruments Inc.). There were reduction/oxidation peak currents for the bared gold
electrode in the results, as shown in Figure 5-15. MERC was used to do the CV to
characterize the modified IDE performance. The result of modified IDE is also shown in
Figure 5-15 where no reduction/oxidation peak currents were found. That proved the
whole gold electrode was coated with CH-19 as expected.

103

Figure 5-15. The CV measurement of the gold electrode before and after coating with
peptide CH-19.

Impedance measurements were conducted with the MERC chip over the frequency
range of 10 to 100 Hz after 0.001 pg/μL of Herceptin was added. For comparison, this
biointerface was tested using a commercial impedance instrument (CHI 760 C) and those
results are included in Figure 5-16. These data demonstrate that the MERC circuit
generates similar results as in commercial equipment.

104

Figure 5-16. The impedance of Herceptin measured by the commercial instrument and
MERC.

The interface impedance is proportional to the Herceptin concentration. Three
calibrators containing 0.001, 0.01, and 0.1 pg/μL of Herceptin in HBS-EP buffer were
prepared prior to each impedance assay. The test results shown in Figure 5-17 present
that the impedimetric biosensor can detect the Herceptin concentration down to
0.001pg/µL with a high linear response (R2=0.995).

105

Figure 5-17. The Herceptin concentration curve vs. the impedance measured by MERC
chip. A high level of linearity (R2=0.9795) was achieved.

106

Chapter 6 CMOS Electrochemical
Instrumentation Array for On-chip
High Throughput Characterization of
Single Ion Channel Proteins
Membrane proteins, and in particular ion channels, are main drug targets, and thus,
ion channels have received widespread recognition for their application in biosensors,
drug discovery, and protein/ligand screening. To study the characters of ion channels, a
1024-element high throughput array concept was proposed by our lab for rapid and
simultaneous ion channel characterization. An electrochemical array instrumentation
circuit for single ion channel characterization was developed to address the many
challenges in forming high throughput chip-scale biosensor arrays with simultaneous
characterization of multiple bio-interfaces, which is one of the thesis goals. A smart and
power/hardware efficient instrumentation circuit architecture was developed to address
the cross talk noise caused by simultaneous measurement, the large array parasitic
capacitors, and the multi-modes electrochemical measurements requirements. Moreover,
the instrumentation circuits provide a solution for the noise floor vs. bandwidth tradeoff
in a manner that permits both single ion channel current resolution (pA) as well as tens of
MHz bandwidth.

107

6.1 Single Ion Channel Detection: History and
Background
Approximately one-third of all proteins comprising pores, ion channels, receptors,
and enzymes are membrane associated. Information about how membrane proteins work,
their interactions with lipids and other proteins, and their dynamics and structure, is
essential for the understanding of biological membranes and cell functioning in general.
The implications of ion channels in a variety of diseases, including diabetes, epilepsy,
hypertension, cancer and even chronic pain, have signaled them as pivotal drug targets
[119]. Approximately 15% of the world’s 100 top-selling drugs are targeted at ion
channels [120]. Thus, ion channels have received widespread recognition for their
application in biosensors, drug discovery, and protein/ligand screening.
To understand the kinetic of ion channels and other membrane proteins, two
techniques, patch clamp and pore-based membranes, have been applied for single ion
channel detection (SICD). Patch clamp has been used for characterizing ion channels in
membrane proteins since 1970s. Patch clamp is a laboratory technique in
electrophysiology that allows for the study of single or multiple ion channels in cells.
Patch clamp recording uses a glass micropipette that has an open tip diameter of about
one micrometer. The open tip encloses a membrane surface area or "patch" that often
contains one or a few ion channel molecules [121]. The major drawback of patch clamp
is the lack of control over the protein and lipid composition of the membrane patch. The
pore-based membrane technique overcomes this drawback by incorporating a purified
channel protein into a bilayer lipid membrane (BLM) with a controlled lipid environment.

108

Such BLMs have a simple structure with predictable, well-defined electrical properties
[122]. The pore-based BLM method is preferred when a membrane protein needs to be
characterized in a controlled environment.
Both patch clamp and pore based membrane techniques can be measured by optical
or electrochemical instrumentation. Traditionally, ion channel detection was based on
fluorescence. In fluorescence detection methods, target molecules are labeled
fluorescently and the opening and closing rates of molecules are measured by optical
techniques such as fluorescence correlation spectroscopy and single-molecule
fluorescence resonance energy transfer. Despite the high specificity of fluorescence based
techniques , the fact that light is used as an intermediary between the biological system
and measurement electronics results in fundamental constraints in resolution and
bandwidth due to the countable number of photons emitted [123].
Electrochemical techniques are now being explored which provide higher
bandwidth ion channel measurements compared with optical techniques [124]. Moreover,
electrochemical methods have the advantage of low cost and a capability for
miniaturization [125] [126].

6.2 Single-Ion-Channel Array Microsystem

6.2.1 Pore-based pBLM and Electrochemical Methods
Planar lipid bilayer membrane (pBLM) is one type of artificial lipid bilayer. PBLM
with pore supported technique is widely used in functional studies for single ion channels

109

because the pore structure supports ions passing through the ion channel. The general
structure of a pore-based pBLM is presented in Figure 6-1. Two sides of the BLM are
isolated in fluid chambers, which are filled with electrolyte solution. One protein is
inserted in the pBLM and ions go through the ion channel under a given potential. The
ion channel opening/closing events are recorded by the constant potential amperometry
(CPA) technique that measures the response current pulse between electrode A and B at a
DC potential. CPA is also used to evaluate the pBLM formation quality by measuring the
pBLM impedance. Additionally, the electrical properties of a pBLM can be characterized
using the electrochemical impedance spectroscopy (EIS) technique, which enables
extraction of pBLM equivalent circuit model parameters.

Figure 6-1. General structure of a pore-based pBLM cell with a single ion channel protein.
The opening/closing events of the ion channel can be detected by monitoring response
currents between electrodes A and B.

110

6.2.2 Microfluidic structures for self-assembled pBLM
Recently several groups have introduced microfluidic structures using poly(methyl
methacrylate) (PMMA), polytetrafluoroethylene (PTFE), and polydimethylsiloxane
(PDMS) to flow lipid solutions across an aperture, including a device with a 500µm hole
through a glass slide [127]. Initially, it was common to construct a pBLM positioned
horizontally by flowing lipid solution across an aperture so that lipid molecules self-align
at the aperture where the organic lipid solution interfaces with an aqueous buffer.
Alternatively, in a modern adaptation of the Montal-Mueller method that eliminates the
use of heavy non-volatile solvent, an aqueous solution having a thin layer of phospholipid
stock solution at the gas/liquid interface is passed across both sides of the micropore,
forming phospholipid monolayers that merge to form a BLM across the aperture. Such
devices typically consist of microfluidic channels on one or both sides of a substrate
containing the aperture [128-130]. Aperture sizes range from 100s of micrometers in
diameter to as small as sub-micron. While these reports demonstrate that micropore
structures and microfluidics are well suited to forming stable BLMs, these new methods,
like traditional ones, suffer from a low yield in BLM formation that makes it difficult to
get consistent measurements of membrane proteins.
In contrast to horizontal devices with apertures, a new approach where the pBLM is
positioned vertically within a microfluidic channel was recently reported [137-140]. The
approach forgoes the need of an aperture-containing substrate and permits an array of
pBLMs to be self-assembled within microchambers shaped into the sidewall of the main
microfluidic channel. When a buffer and organic lipid solution are sequentially flowed
through the channel, a pBLM is constructed at the mouth of each microchamber. One

111

PDMS microchamber with electrode fabricated by Dr. Mason’s lab, as shown in Figure
6-2. The structure of the pBLM microchamber is shown in Figure 6-3(a). With one
electrode in each pBLM microchamber and one electrode is in the main channel, this
structure enables many electrochemical measurements to be made in parallel. This
structure is also suitable for high density array formation, as illustrated in Figure 6-3(b-c).
Microfluidic techniques facilitate rapid and cost-effective fabrication of BLM
microarrays, whose properties vary combinatorial across the array. The capability of
rapidly altering the solute environment around the pBLM by virtue of microfluidics
provides be a major technical advance.

Figure 6-2. PDMS microchamber with electrode fabricated in Dr. Mason’s lab. Work
completed by colleague Lin Li.

6.2.3 High throughput microsystem approach
Multiple microfluidic channels provide the capability of automatic, rapid, and high
throughput pBLM formation. One challenge for a high throughput SICD microsystem is
the wiring bottleneck inherent in connecting hundreds of individually measurable

112

microchambers to instrumentation electronics. To overcome this challenge and enable
high throughput membrane protein characterization, the pBLM microchamber arrays can
be embedded with a CMOS electrochemical instrumentation chip. When the
microfluidics are combined with CMOS instrumentation for electrical recording, the
integrated system can achieve rapid and parallel-screening membrane proteins under
multiple environmental conditions.
The SICD array microsystem being developed utilizes a carrier substrate to expand
the surface area beyond the instrumentation chip, allowing room for multiple fluidic
inlets/outlets while also connecting circuits I/O pads to external electrical contacts. The
high density electrode array is fabricated on the surface of the instrumentation chip and
directly connected to circuits for electrochemical readout. When connected to a
computer-controlled data acquisition system, the user would have full control of the
pumps and valves managing microfluidic operation as well as display and storage of
characterization data. The SICD microsystem is illustrated in Figure 6-3 (c). For a typical
2

5 × 5mm instrumentation chip, eight independent fluidic channels fit well within the
area based on our experiences. When a buffer and organic lipid solution are sequentially
flowed through the microfluidic channel, a pBLM can be constructed at the mouth of
each microchamber. The width of the pBLM microchamber is 50µm and the centercenter pitch of the microchambers is 80µm. 64 pBLM microchambers could be in each
2

microfluidic channel for a 5 × 5mm chip, making the total number of the pBLM
microchambers equal to 1024. This microsystem has been designed to enabled, for the
first time, simultaneous characterization of single-molecule events within multiple
biomembranes.

113

Figure 6-3. (a) The structure of one pBLM microchamber with working electrode
(WE) and reference electrode (RE). (b) One microfluidic channel with shared RE. (c)
Concept view of the high throughput array microsystem for single ion channel detection.

114

6.3. Instrumentation Circuits Requirements and
Challenges

6.3.1 Multi-mode instrumentation circuits
Microfluidic-driven pBLM formation and protein characterization are complicated
processes. The CMOS instrumentation circuits are designed to support multiple modes
based on the needs of different assays.
Mode 1: pBLM formation quality test by CPA technique.
Mode 2: pBLM classification by EIS technique.
Mode 3: simultaneous single ion channel event detection by CPA technique.
The pBLM quality can be evaluated by the membrane impedance because a poor
quality pBLM has a large leakage current (as noise) that considerably reduces the
membrane impedance. In mode 1, the circuit will determine if/when a high quality
membrane is formed by comparing the pBLM impedance with a programmable threshold
value selected from 100MΩ and 1GΩ. Generally, the impedance of a good quality pBLM
should be above 1GΩ [8], and when the pBLM impedance is less than 100MΩ it is
assumed to be poor. Any poor pBLMs will be skipped over during subsequent modes.
Mode 2 provides pBLM classification via EIS characterization. The different
formation shapes of pBLM, as shown in Figure 6-4, could be considered as different
classes. Different classes of pBLM have unique phase/amplitude characteristics, which

115

provide a second layer of information to describe the protein’s environment beyond the
impedance information provided in mode 1.
Mode 3 is the most challenging because it requires recording single channel
opening/closing events with MHz speed and sub-pA resolution. The instrumentation
electronics have to guarantee mode 3 performances, which require ultra-low noise and
high speed readout circuits.

Figure 6-4. Example for different classes of pBLMs.

6.3.2 Instrumentation circuits requirements and
challenges
The multi-modes measurement, single-molecule resolution, and high density array
instrumentation requirements bring a lot of challenges in the circuit implementation. One
challenge is addressing the noise floor vs. bandwidth tradeoff in a manner that permits
both pA resolution as well as tens of MHz bandwidth, as needed to measure typical single
ion channel events. Assuming it is necessary to achieve a minimum SNR of around 3, a

116

noise floor of less than 330fA is needed to record the a 1pA response of a single ion
channel. The array requirement adds even more of a challenge because of the crosstalk
noise between the tens of readout circuits working simultaneously. Further difficulty is
added by the need to keep the circuit small so that many readout blocks can be placed on
a chip for parallel array measurement. Additionally, power consumption must be closely
managed to maintain a surface temperature suitable for pBLMs (~40ºC max).

6.4. Instrumentation Array Approach
To address the requirements and realize high throughput chip-scale biosensor
arrays with simultaneous characterization of multiple bio-interfaces, a single protein
electrochemical characterization (SPEC) chip was designed. For a 1024-element BLM
array with simultaneous characterization, a straightforward approach is to have 1024
CPA/EIS readout blocks. However, the power consumption for 1024 readout blocks
would be extreme (up to hundreds of watts), which would generate local hotspots that
would degrade pBLM interface.
Because of the delicate nature of pBLM formation, this initial step of array
formation can be expected to have a low yield, around 30% based on empirical evidence.
Our experience also suggests the good quality pBLMs will occur roughly equally in three
distinct structural classes. Furthermore, we estimate that the likelihood of populating an
individual pBLM with a single membrane protein to be about 20% after experimental
optimization. In other words, we expect a very low yield of single-protein, single-pBMLclass interfaces, demanding a high density array. However, this low yield provides an

117

opportunity for several pBLM interfaces to share one CPA/EIS readout circuit. Based on
these estimates, we expect 1-3 good quality, single-class, single-protein interfaces within
each microfluidic channel (the calculation detail shown in appendix). Thus, the
instrumentation array has been designed with four CPA/EIS readout circuits per channel.
This should ensure capture of single ion channel events, simultaneously across most
functional interfaces, while also achieving hardware efficiency; a total of 32 CPA/EIS
readout circuits occupies only 3.1% of area required for an instrumentation chip with
2

1024 CPA/EIS circuits and enables implementation within a 5x5mm chip.
Tens of CPA/EIS current readout circuits working simultaneously introduce a noise
immunity challenge. The high throughput array also has a large parasitic capacitance that
distorts the electrochemical current signals. To solve these problems, an in-pixel
electrochemical current amplifier for each pBLM element was chosen in the SPEC chip.
This pixel amplifier significantly increases immunity to signal interference, solves the
signal distortion caused by the array parasitic capacitance, and loosens the resolution
requirement of the CPA/EIS readout circuit.
The architecture of the designed SPEC chip is shown in Figure 6-5. There are eight
independent potentiostats that set the independent bias potential for each microfluidic
channel during CPA and EIS measurement. Digital memory is used to buffer the results
between the CPA/EIS readout and uploading to an external host system through a serial
peripheral interface communication bus. A digital control block manages the operation
mode. Potentiostat and CPA/EIS readout circuit specifications for each mode are listed in
Table 6-1.

118

Figure 6-5. Architecture of the single protein electrochemical characterization (SPEC)
high density array chip.

119

Table 6-1. The performance requirement for potentiostat and readout circuits for each
mode.
potentiostat

CPA/EIS readout circuit

mode 1 (1) frequency: DC
(CPA)
(2) voltage range: -250mV ˗ 250mV’
(3) step resolution: 10mV

(1)current readout mode: DC
(2)current range: pA to mA
(3)bandwidth: 10kHz

mode 2 (1) frequency range: 100mHz-100Hz;
(EIS)
(2) AC amplitude: 10mV;

(1)current readout mode: AC
(2)current range: pA to mA

(3) DC bias range: -±250mV;

(3)phase range: -180° ˗ 180°
(4)bandwidth: 100kHz
mode 3 (1) frequency: DC
(CPA)
(2) voltage range: ±250mV;

(1)current readout mode: DC
(2)current range: pA to 200pA
(3)resolution: 300fA

(3) step resolution: 10mV;

(4)bandwidth: 10MHz

6.5 CMOS Implementation

6.5.1 Preamplifier design
Noise and speed are the two key considerations in the pixel amplifier design
because the input signal of the amplifier is a high speed (down to 10µs pulse width) and
weak (~pA) current pulse. A resistive feedback amplifier is simple to implement, but it
suffers from the noise generated by the feedback resistor. A capacitor feedback amplifier
is not suitable for measuring high speed low-level signals because the feedback capacitor

120

limits amplifier speed. A current conveyor was selected as the amplifier in our design
because it is good at high speed and for weak current amplification.
A cascade current mirror structure was utilized in the pixel amplifier to reduce
crosstalk noise generated by the high speed switch coupling. The schematic of the pixel
amplifier is shown in Figure 6-6. Transistor M1 is a current sink controlled by ‘Bias’ that
ensures amplification of the bi-direction ion-channel current. ‘Sel’ is the pixel selection
signal. A large, pF-level, column parasitic capacitor (Cp) is associated with each pixel
amplifier output in the high throughput array, and the total gain of the amplifier was set
as 25 to ensure capability to drive the Cp load. To eliminate the delay caused by
transistor parasitic capacitances, the minimal gate length (L=0.6µm) was selected for all
transistors. The layout size of the pixel amplifier is 27µm × 80µm, as shown in Figure 67, which is small enough to ensure that each pBLM microchamber has its own pixel
amplifier.

121

Figure 6-6. Schematic of the two stages current conveyor pixel preamplifier.

Figure 6-7. Layout of the pixel preamplifier.

122

6.5.2 CPA/EIS blocks
The multi-mode electrochemical readout circuit introduced in Chapter 5 is a great
candidate for the CPA/EIS blocks in terms of function and resolution. The Chapter 5
readout circuit is designed for an input signal less than 10KHZ (typical biosensors speed
range), which is not capable to detect the ion channel events with tens of microseconds
time scale [124]. The readout speed of the Chapter 5 circuit is limited by the large
parasitic resistance and capacitance of the 0.5µm CMOS process. This limitation could
be solved by shifting the Chapter 5 circuit to a smaller size CMOS process in order to
reduce the parasitic resistance and capacitance. Based on our experience, we believe that
shifting the Chapter 5 readout circuit to a more advanced technology is more like an
engineering project. Besides changing processes, an alternative solution is using
commercial ADCs to readout the preamplifier’s output current. No matter which method,
the CPA/EIS block is not the most challenging component in the instrumentation circuit
design.

6.6 Simulation Results

6.6.1 Simulation condition
The pixel amplifier was designed at AMI 0.5µm CMOS technology and simulated
in Cadence Spectre. The simulation parameters are listed in Table 6-2. Cp = 1.5 pF was
added at the load of the pixel amplifier in the simulation. To verify the function of the
pixel amplifier, a 5pA, 10µs width current pulse was connected to the input of the pixel

123

amplifier to mimic the fastest single ion channel opening/closing current. The parameters
used in the simulation are summarized in Table 6-3.
Table 6-2. Simulation parameters with Candece Spectre simulator.
Process

2-Poly 3-Metal, AMI 0.5µm

VDD

5V

temperature

25°C

Simulation model

ami06N, ami06P @ TT

Load capacitor

1.5pF

6.6.2 Simulation results
To verify the function of the preamplifier, a 5pA, 10µs width current pulse is
connected to the input of the preamplifier and monitors the first and second stage current
mirror output. The simulation results are presented in Figure 6-8, where the value of IA
and IOUT are as expected.
The pulse waveform distortion caused by the parasitic capacitance can be identified
as a pulse width distortion percentage (PWDP), as shown in Figure 6-9. The purple is the
original pulse waveform, and the black is the distorted waveform. To calculate the
processing speed limitation of the preamplifier, a series of various width pulse current
inputs are simulated. The simulation results are shown in Figure 6-10. The PWDP for the
10us pulse width current (the fastest ion channel speed) is only 1.04%. Thus, the pixel
amplifier is fast enough to track the high speed opening/closing events.

124

Figure 6-8. The simulation results of the first stage output (IA) and second stage ouput
(IOUT) of the preamplifier with a 5pA, 10µs width pulse current input (IWE).

Figure 6-9. The definition of the pulse width distortion percentage (PWDP).

125

Figure 6-10. Simulated results of the distortion percentage (DP) with various pulse
widths of the input current (IWE).

In order to evaluate the pixel amplifier operation in a larger scale biosensor array,
dependence of PWDP on array capacitance was simulated. Figure 6-11 shows the
simulated PWDP of IOUT under various array capacitor values for a 10µs pulse input IWE.
The PWDP is only 2% even for a large Cp = 15pF.

126

IOUT PWDP

2.5
2
1.5
1
0.5
0
1

3

5

7

9

11 13 15

Cp (pF)
Figure 6-11. Simulated IOUT pulse width distortion percentage (PWDP) as a function of
the column line parastic capacitance (Cp).

6.6.3 Power consumption analysis
To manage power consumption, the preamplifiers and readout block will be
realized with the least power needed to achieve performance goals, and power
management schemes will be implemented to regulate surface heating. The estimated
power consumption of each component is listed in Table 6-3. The total power
consumption of SICD instrumentation circuits is only 200mW. Based on our prior work
2

with CMOS heaters to control biointerface temperatures [131], ~22.75W/mm is needed
to heat a surface liquid to 50ºC, so it appears that it is easy to avoid the local overheating.

127

Table 6-3. The power consumption estimation of the single ion channel detection
instrumentation circuits.
Component

Quality

Power/element

preamplifier

1024

15.5µW

CPA/EIS readout circuit

32

5.4mW

Bias

1

50µW

Row/column selector &digital communication

1

100µW-1mW

Total power

~200mW

6.6.4 Area estimation
Based on our previous experience on the potentiostat, row/column decoder, bias,
memory and the digital control & communication circuits, an estimated area of the
designed SICD circuits is listed in Table 6-4. The estimated area of the potentiostat, bias,
CPA/EIS readout, row/column decoder, digital controller and communication blocks are
referenced [64, 106, 131], The memory size is calculated based on that the area of a
2

standard 6-transitor memory cell is 25x33µm , and the total needed number of the
8

memory is 2 x32 (assume the resolution of CPA/EIS is 8 bit). The total estimated area of
2

2

the designed SICD instrumentation circuit is 5.36mm , which is within the 5x5mm chip
size goal.

128

Table 6-4. Area estimation for each components in the single ion channel detection
instrumentation circuit with AMI 0.5um process.
Item

Area/block

Number of item

Potentiostat

200x300µm2

8

Pixel preamplifier

27x80 µm2

1024

CPA/EIS readout circuit

500x100µm2

32

Bias

300x300µm2

1

Row/column decoder

300x500µm2

1

Memory

25x33µm2
8
x2 x32

1

Digital control
&communication

~200x400µm2

1
9.96mm2

Total area

129

Chapter 7 Summary and Future Work
7.1 Summary
This thesis research developed technologies that overcome the challenges of
implementing CMOS instrumentation circuits for miniaturized high throughput biosensor
array microsystems. The requirements of the electrochemical instrumentation have been
studied, and different structures of amperometric and impedimetric instrumentation have
been evaluated. A high resolution amperometric instrumentation circuit based on
correlated double sampling (CDS) technique was developed as a low-noise solution for
readout of the weak signals in an array. Furthermore, a compact, low power, impedance
analysis circuit that utilizes frequency response analysis (FRA) method to extract real and
imaginary impedance components has been developed. A multi-mode electrochemical
instrumentation circuit was implemented for protein characterization and sensing while
sharing

hardware

instrumentation

resources.

circuit

Finally,

architecture

a

CMOS

for

high

monolithic
throughput,

electrochemical
single

protein,

characterization was designed, enabling a significance improvement in the time and
cost to acquire statistically relevant single protein measurements.
The work in this thesis provides the following significant contributions:

Developed

high

performance

electrochemical

miniaturized biosensor array microsystems

130

instrumentation

for

A CMOS amperometric biosensor readout circuit with current resolution of 1pA
has been developed. Furthermore, a new compact, low power, ~100fA current resolution
impedance analysis circuit that utilizes mixed-mode signal processing to extract real and
imaginary impedance components has been developed and tested with an on-chip protein
interface. Very good agreements were observed between the commercial instrument and
the CMOS amperometric instrumentation circuit and the impedimetric instrumentation
circuit, respectively.

Implemented a hardware efficient multi-mode CMOS electrochemical
readout circuit for biointerfaces characterization and sensing
To realize a multi-protein array of sensors with different instrumentation needs, a
hardware efficient CMOS electrochemical circuit has been designed to achieve both
amperometric and impedimetric measurement while sharing hardware source. The area of
2

one channel multi-mode readout circuit is 0.045mm , which indicates around 500 readout
2

circuits integrated in a be in a 5x5mm chip. A hysteretic comparator was applied in this
readout circuit with improved resolution and reduced power consumption. Functionality
of the multi-mode instrumentation circuit was verified by characterization of a Herceptin
biosensor useful for detecting cancer biomarkers.

Developed a novel 1000-element array instrumentation architecture that
permits rapid high-throughput characterization of membrane proteins
with single protein resolution within a chip-scale

131

By integrating an understanding of both protein assays challenges and
microelectronics design limitations, an intelligent instrumentation circuit architecture was
developed for a complex, high throughput single-protein array. The pixel preamplifier
structure overcomes the cross talk noise caused by the simultaneous measurement and the
signal distortion caused by the large array parasitic capacitors. Simulation results of the
array instrumentation circuit demonstrates capability to detect the pA-level single protein
current with more than 10MHZ bandwidth. The total area of the designed instrumentation
2

circuit easily permits a 1000-element array within a silicon chip of only 9.96mm .

7.2 Future Work
This thesis work demonstrates high performance electrochemical instrumentation
circuits that enable simultaneous characterization of multiple protein interfaces in a high
throughput chip-scale biosensor arrays microsystem. Based on the results in this research,
there are some promising directions for future research.

A. CMOS implementation for single ion channel
instrumentation circuits
The architecture of the single ion channel detection (SICD) instrumentation circuit
has been developed in Chapter 6. The next step is to implement the SICD instrumentation
circuit in an advance CMOS process. To achieve the best figure of merit (includes the
bandwidth, power consumption, noise performance and area parameters) of the

132

instrumentation circuit, various advanced CMOS processes have been analyzed. Based on
our analysis, the best CMOS process for the SICD instrumentation is 0.25µm CMOS
process. In the future, the SICD instrumentation could be implemented in this process or
alterative process to realize a functional high throughput membrane protein array
characterization system.

B. Development of a “smart” monolithic electrochemical
measurement device
Today, commercial controllers are powerful tools that can provide the hardware
necessary to control external CMOS circuits, analyze data, and communicate to an
external system such as a wireless mesh radio or a laptop PC. For example, the MSP430
is an ultra-low power controller. It supports a variety of communication standards
permitting a wide range of digital interfaces with simple software modifications. It also
includes DACs that can generate various types of waveform as stimulus signals. The
developed CMOS instrumentation associated with a commercial controller could enable a
compact monolithic electrochemical measurement device that could be widely used in
point of care applications.

133

APPENDIX

134

APPENDIX
Estimates that the percent of good quality pBLM produced will be around 30%, the
percentage of Class 1 pBLMs will be about 33%, and the percentage of membranes
including only single ion channels will be about 20%. If we assume that the pBLMs
follow a normal distribution, then there will be around 0.5~1.6 good quality, Class 1,
single-ion-channel pBLMs in one column. The calculation for this estimation is as
follows
(1) Mode 1, calculate the number of the 25% good quality pBLM. Assume pBLMs
the normal distribution: (µ=64*25%=16, σ=4 ) and [µ-2σ, µ+2σ] covers 95%
confidence interval.
Then (µ-2σ) =8, (µ+2σ) =24. [8, 24] will cover 95% range
(2) Mode 2, calculate the number of pBLM in one class. Assume 1/3 pBLMs are
one pBLM class
[8, 24]*1/3=[2.7, 8]

(6-1)

(3) Mode 3, calculate the number single ion channel pBLM
[2.7, 8]*20% = [0.5, 1.6]

(6-2)

Based on (6-2), there will be 0.5~1.6 (with 95% confidence interval) single protein
pBLMs in each column, so it is reasonable to believe that two CPA/EIS readout circuits
are enough in each column for single protein pBLM detection. Two readout circuits not
only ensures capture of the single ion channels events, but also ensures hardware
efficiency. The total number of CPA/EIS readout circuits in the instrumentation is 32.

135

BIBLIOGRAPHY

136

BIBLIOGRAPHY
[1]

S. Stovall, "Scientists Reach Midpoint of Protein Study," in The Wall Street
Journal, ed, 2010.

[2]

(2010).
Protein
[Online].
http://en.wikipedia.org/wiki/Protein#Structure_determination

[3]

J. Drews, "Drug Discovery: A Historical Perspective," Science, vol. 287, pp.
1960-1964, 2000.

[4]

(2012). Cancer Facts & Figures 2012
[Online]. Available:
http://www.cancer.org/acs/groups/content/@epidemiologysurveilance/documents/
document/acspc-031941.pdf

[5]

S. Hauck, S. Drost, E. Prohaska, H. Wolf, and a. S. Dübel, "Analysis of Protein
Interactions Using a Quartz Crystal Microbalance Biosensor."

[6]

D. J. H. A. Manz, E. Verpoorte, H. M. Widmer, "Planar chips technology for
miniaturization of separation systems: a developing perspective in chemical
monitoring," Advances in chromatography, vol. 33, pp. 1-66, 1993.

[7]

A. M. G. H. W. Sanders, "Chip-based microsystems for genomic and proteomic
analysis," Trennds in analytical chemistry, vol. 19, pp. 364-378, 2000.

[8]

E. Verpoorte, "Microfluidic Chips for Clinical and Forensic Analysis,"
Electrophoresis, vol. 23, pp. 677-712, 2002.

[9]

S. C. Yang Liu, and Evangelyn C.Alocilja, "Fundamental Building Blocks for
Molecular Bio-wire based Forward-error Correcting Biosensors," Nanotechnology,
vol. 18, p. 424017(6pp), 2007.

[10]

M. De, S. Rana, H. Akpinar, O. R. Miranda, R. R. Arvizo, U. H. F. Bunz, et al.,
"Sensing of proteins in human serum using conjugates of nanoparticles and green
fluorescent protein," Nature Chemistry, pp. 461 - 465, 2009.

[11]

M. Vestergaard, K. Kerman, and E. Tamiya, "An Overview of Label-free
Electrochemical Protein Sensors," Sensors, vol. 7, pp. 3342-3458, 2007.

[12]

V. Vermeeren and L. Michiels. (2001). Evolution Towards the Implementation of
Point-Of-Care Biosensors.

[13]

K. M. Hanafiah, M. Garcia, and D. Anderson, "Point-of-care testing and the
control of infectious diseases," Biomark Med., vol. 7, pp. 333-3347, 2013.

137

Available:

[14]

D. J. Cahill, "Protein Arrays: A High-throughput Solution for Proteomics
Research?," Proteomics: A Trends Guide, pp. 47-51, 2000.

[15]

S. J. Maerkl, "Next Generation Microfluidic Platforms for High-throughput
Protein Biochemistry," Current Opinion in Biotechnology, vol. 22, pp. 1-7, 2010.

[16]

A. G. Crevillen, M. Pumera, M. C. Gonzaleza, and A. Escarpa, "Towards lab-ona-chip approaches in real analytical domains based on microfluidic
chips/electrochemical multi-walled carbon nanotube platforms," Lab Chip, vol. 9,
pp. 346-353, 2009.

[17]

Y. H. B. Ghallab, Wael Lab-on-a-Chip : Techniques, Circuits, and Biomedical
Applications: Artech Hous, 2010.

[18]

L. Wang and P. Li, "Microfluidic DNA microarray analysis: a review," Anal
Chim Acta., vol. 687, pp. 12-27, 2011

[19]

W. B. Zimmerman, "Electrochemical microfluidics," Chemical Engineering
Science, vol. 66, pp. 1412–1425, 2011.

[20]

D. Grieshaber, R. MacKenzie, and J. V. o. o. a. E. Reimhult, "Electrochemical
Biosensors - Sensor Principles and Architectures," Sensors, vol. 8, pp. 1400-1458,
2008.

[21]

L. D. Mello and L. T. Kubota, "Review of the use of biosensors as analytical tools
in the food and drink industries," Food Chemistry, vol. 77, pp. 237–256, 2002.

[22]

D. R. Thevenot, K. Toth, R. A. Durst, and G. S. Wilsond, "Electrochemical
Biosensors: Recommended Definitions and Classification," Biosensors and
Bioelectronics, vol. 16, pp. 121-131, 2001.

[23]

M. Pohanka and P. Skládal, "Electrochemical biosensors – principles and
applications," Journal of Applied Biomedicine, vol. 6, pp. 57–64, 2008.

[24]

G. Korotcenkov, Chemical Sensors: Volume 1 General Approaches: Momentum
Press, 2010.

[25]

A. Bard, and L. Faulkner, Electrochemical Methods: Fundementals and
Applications, 2 ed. New York: John Wiley & Sons, 2000.

[26]

R. I. Stefan, J. F. van Staden, and H. Y. Aboul-Enein, " Immunosensors in clinical
analysis," Fresenius Journal of Analytical Chemistry, vol. 366, pp. 659-668, 2000.

[27]

J. Wang, "Amperometric Biosensors for Clinical and Therapeutic Drug
Monitoring: A Review," Journal of Pharmaceutical and Biomedical Analysis, vol.
19, pp. 47-53, 1999.

138

[28]

I. Palchettia, A. Cagnini, M. D. Carlo, C. Coppi, M. Mascini, and A. P. F. Turnerb,
"Determination of anticholinesterase pesticides in real samples using a disposable
biosensor," Analytica Chimica Acta, vol. 337, pp. 315–321, 1997.

[29]

A. C. Mongra, A. Kaur, and R. K. Bansal, "Review Study On ElectrochemicalBased Biosensors," International Journal of Engineering Research and
Applications, vol. 2, pp. 743-749, 2012.

[30]

E. Barsoukov and J. Macdonald, Impedance Spectroscopy: Theory, Experiment,
and Applications, 2 ed.: John Wiley and Sons, 2005.

[31]

C. Yang, S. R. Jadhav, R. M. Worden, and A. J. Mason, "Compact Low-Power
Impedance-to-Digital Converter for Sensor Array Microsystems," IEEE J. Solid
State Circuits, vol. 44, pp. 2844-2855, Oct. 2009.

[32]

M. Stelzle, G. Weissmuller, and E. Sackmann, "On the application of supported
bilayers as receptive layers for biosensors with electrical detection," Journal of
Physical Chemistry, vol. 97, pp. 2974-2981, 1993.

[33]

F. Yin, "A novel capacitive sensor based on human serum albumin-chelant
complex as heavy metal ions chelating proteins," Analytical Letters, vol. 37, pp.
1269-1284, 2004.

[34]

I. Bontidean, C. Berggren, G. Johansson, E. Csoregi, B. Mattiasson, J. A. Lloyd,
et al., "Detection of heavy metal ions at femtomolar levels using protein-based
biosensors," Analytical Chemistry, vol. 70, p. 4162, 1998.

[35]

L. J. Yang, Y. B. Li, C. L. Griffis, and M. G. Johnson, "Interdigitated
microelectrode (ime) impedance sensor for the detection of viable salmonella
typhimurium," Biosensors and Bioelectronics, vol. 19, pp. 1139-1147, 2004.

[36]

N. N. Mishra, S. Retterer, T. J. Zieziulewicz, M. Isaacson, D. Szarowski, D. E.
Mousseau, et al., "On-chip micro-biosensor for the detection of human CD4(+)
cells based on AC impedance and optical analysis," Biosens Bioelectron, vol. 21,
pp. 696-704, 2005.

[37]

C.-S. Lee, S. K. Kim, and M. Kim, "Ion-Sensitive Field-Effect Transistor for
Biological Sensing," Sensors vol. 9, pp. 7111–7131, 2009.

[38]

S.-R. Lee, K. Sawada, H. Takao, and M. Ishida, "An enhanced glucose biosensor
using charge transfer techniques," Biosensors and Bioelectronics, vol. 24, pp.
650-656, 2008.

[39]

A. B. Kharitonov, J. Wasserman, E. Katz, and I. Willner, "The Use of Impedance
Spectroscopy for the Characterization of Protein-Modified ISFET Devices: 
Application of the Method for the Analysis of Biorecognition Processes," J. Phys.
Chem. B, vol. 105, pp. 4205–4213, 2001.

139

[40]

S. Purushothaman, C. Toumazou, and C. Ou, "Protons and single nucleotide
polymorphism detection: a simple use for the ion sensitive field effect transistor,"
Sensors and Actuators B: Chemical, vol. 114, pp. 964–968, 2006.

[41]

M. Kalofonou, P. Georgiou, C. P. Ou, and C. Toumazou, " An ISFET based
translinear sensor for DNA methylation detection," Sensors and Actuators B:
Chemical, vol. 161, pp. 156-162, 2012.

[42]

G. Wang, L. Wang, Y. Han, S. Zhou, and X. Guan, "Nanopore Stochastic
Detection: Diversity, Sensitivity, and Beyond," Acc Chem Res., vol. Epub ahead
of print, 2013

[43]

J. Janata, "Historical review. Twenty years of ion-selective field-effect
transistors," Analyst, vol. 119, pp. 2275-2278, 1994.

[44]

X. Zhu and C. Ahn, "On-chip Electrochemical Analysis System Using
Nanoelectrodes and Bioelectronic CMOS Chip," IEEE J. Sensors, vol. 6, pp.
1280-1286, October 2006.

[45]

Y. Huang and A. J. Mason, "A Redox-Enzyme-Based Electrochemical Biosensor
with a CMOS Integrated Bipotentiostat," in IEEE Biomedical Circuits Systems
Conf., 2009.

[46]

A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and
Applications: Wiley, 2000.

[47]

R. Thewes, F. Hofmann, A. Frey, B. Holzapfl, M. Schienle, C. Paulus, et al.,
"Sensor arrays for fully electronic DNA detection on CMOS," IEEE ISSCC Dig.
Tech., pp. 350-351, 2002.

[48]

R. F. B. Turner, D. J. Harrison, and H. P. Baltes, "A CMOS Potentiostat for
Amperometric Chemical Sensors," IEEE J. Solid-State Circuits, vol. 22, pp. 473478, June 1987.

[49]

L. Li, X. Liu, W. A. Qureshi, and A. J. Mason, "CMOS Amperometric
Instrumentation and Packaging for Biosensor Array Applications," presented at
the IEEE Biomedical Circuits Systems Conf., 2012.

[50]

S. M. Martin, F. H. Gebara, T. D. Strong, and R. B. Brown, "A Fully Differential
Potentiostat," IEEE J. Sensors, vol. 9, pp. 135 - 142, 2009.

[51]

M. M. J. Ahmadi, G.A., "Current-Mirror-Based Potentiostats for Three-Electrode
Amperometric Electrochemical Sensors," IEEE Tran. Circuits and Systems, vol.
56, pp. 1339-1348, 2008.

[52]

S. M. Martin, F. H. Gebara, B. J. Larivee, and R. B. Brown, "A CMOS-Integrated
Microinstrument for Trace Detection of Heavy Metals," IEEE J. Solid State
Circuits, vol. 40, pp. 2777-2786, 2005.
140

[53]

E. Lauwers, J. Suls, W. Gumbrecht, D. Maes, G. Gielen, and W. Sansen, "A
CMOS Multiparameter Biochemical Microsensor with Temperature Control and
Signal Interfacing," IEEE J. Solid State Circuit, vol. 36, pp. 2030-2038,
December 2001.

[54]

P. M. Levine, P. Gong, R. Levicky, and K. L. Shepard, "Active CMOS Sensor
Array for Electrochemical Biomolecular Detection," IEEE J. Solid State Circuits,
vol. 43, pp. 1859-1971, 2008.

[55]

S. M. Martin, F. H. Gebara, T. D. Strong, and R. B. Brown, "A Low-Voltage,
Chemical Sensor Interface for Systems-On-Chip: The Fully-Differential
Potentiostat," IEEE Int. Symp. Circuits and Systems, vol. 4, pp. 892-895, May
2004.

[56]

M. Carminati, G. Ferrari, F. Guagliardo, M. Farina, and M. Sampietro, "LowNoise Single-Chip Potentiostat for Nano-Bio-Electrochemistry over a 1MHz
Bandwidth," IEEE Electronics, Circuits, and Systems, pp. 952-955, 2009.

[57]

R. J. R. S. P. Kounaves and G. T. A. Kovacs, "An integrated CMOS potentiostat
for miniaturized electroanalytical instrumentation," IEEE Int. Solid-State Circuits
Conf., pp. 162–163, 1994.

[58]

C. Yang, Y. Huang, B. L. Hassler, R. M. Worden, and A. J. Mason,
"Amperometric Electrochemical Microsystem for a Miniaturized Protein
Biosensor Array," IEEE Trans. Biomedical Circuits and Systems, vol. 3, pp. 160168, June 2009.

[59]

S. Ayers, K. D. Gillis, M. Lindau, and B. A. Minch, "Design of a CMOS
Potentiostat Circuit for Electrochemical Detector Arrays," IEEE Tran. Circuits
and Systems, vol. 54, pp. 736 - 744, 2007.

[60]

A. Gore, S. Chakrabartty, S. Pal, and E. C. Alocilja, "A Multichannel
Femtoampere-Sensitivity Potentiostat Array for Biosensing Applications," IEEE
Tran. Circuits and Systems, vol. 53, pp. 2357-2363, 2006.

[61]

H. S. Narula and J. G. Harris, "A Time-Based VLSI Potentiostat for Ion Current
Measurements," IEEE J. Sensors, vol. 6, pp. 239-247, 2006.

[62]

R. J. Reay, S.P. Kounaves, and G.T.A. Kovacs, "An Integrated CMOS
Potentiostat for Miniaturized Electroanalytical Instrumentation," IEEE Int. SolidState Circuits Conf., pp. 162 - 163, February 1994.

[63]

M. Breten, T. Lehmann, and E. Bruun, "Integrating Data Converters for
Picoampere Currents from Electrochemical Transducers," presented at the IEEE
Int. Symp. Circuits and Systems, Geneva, 2000.

141

[64]

L. Li, X. Liu, and A. Mason, "CMOS Amperometric Instrumentation and
Packaging for Biosensor Array Applications," IEEE Trans. Biomedical Circ.
Systems, vol. 5, pp. 439 - 448, 2011.

[65]

D. Kim, B. Goldstein, W. Tang, F. J. Sigworth, and E. Culurciello, "Noise
Analysis and Performance Comparison of Low Current Measurement Systems for
Biomedical Applications," IEEE Trans. Biomedical Circuits and Systems, 2012
(in press).

[66]

K. Murari, N. Thakor, M. Stanacevic, and G. Cauwenberghs, "Wide-range,
picoampere-sensitivity multichannel VLSI potentiostat for neurotransmitter
sensing," Int. Conf. IEEE Engineering in Medicine and Biology Society, vol. 6, pp.
4063-4066, 2004.

[67]

R. G. Kakerow, H. Kappert, E. Spiegel, and Y. Manoli, "Low power single chip
CMOSpotentiostat," Int. Conf. Solid-State Sensors and Actuators, pp. 142–145,
1995.

[68]

A. Bandyopadhyay, G. Mulliken, G. Cauwenberghs, and N. Thakor, "VLSI
potentiostat array for distributed electrochemical neural recording," IEEE Int.
Symp. Circuits and Systems, pp. 740-743, 2002.

[69]

G. Ferrari, F. Gozzini, A. Molari, and M. Sampietro, "Transimpedance Amplifier
for High Sensitivity Current Measurements on Nanodevices," IEEE J. Solid State
Circuits, vol. 44, pp. 1609-1616, 2009.

[70]

R. G. Kakerow, H. Kappert, E. Spiegel, and Y. Manoli, "Low-Power Single-Chip
CMOS Potentiostat," International Conference on Transducers, Solid-State
Sensors, Actuators and Microsystems, vol. 1, pp. 142-145, June 1995.

[71]

M. H. Nazari and R. Genov, "A Fully Differential CMOS Potentiostat," IEEE Int.
Symp. Circuits and Systems, pp. 2177 - 2180, 2009.

[72]

R. Genov, M. Stanacevic, M. Naware, G. Cauwenberghs, and N. Thakor, "16channel integrated potentiostat for distributed neurochemical sensing," IEEE Tran.
Circuits and Systems, vol. 53, pp. 2371–2376, 2006.

[73]

S. S. Ghoreishizadeh, I. Taurino, S. Carrara, and G. D. Micheli, " A CurrentMode Potentiostat for Multi-Target Detection Tested with Different Lactate
Biosensors," IEEE Biomedcial Circ. Sys. Conf., pp. 128-131, 2012.

[74]

M. Stanac´evic´, K. Murari, A. Rege, G. Cauwenberghs, and N. V. Thakor, "VLSI
Potentiostat Array With Oversampling Gain Modulation for Wide-Range
Neurotransmitter Sensing," IEEE Tran. Circuits and Systems, vol. 1, pp. 63-72,
2007.

142

[75]

K. Murari, M. Stanacevic, G. Cauwenberghs, and N. V. Thakor, "Integrated
Potentiostat for Neurotransmitter Sensing," IEEE Eng. Medicine Biol. Mag., vol.
24, pp. 23–29, 2005.

[76]

D. Rairigh, C. Yang, and A. Mason, "Analysis of On-Chip Impedance
Spectroscopy Methodologies for Sensor Arrays," Sensor Letters, Dec. 2006.

[77]

M. Min and T. Parve, "Improvement of Lock-in Electrical Bio-Impedance
Analyzer for Implantable Medical Devices," IEEE Trans. Instrumentation and
Measurement, vol. 56, pp. 968-974, 2007.

[78]

H. M. Jafari and R. Genov, "CMOS Impedance Spectrum Analyzer with DualSlope Multiplying ADC," presented at the IEEE Biomedical Circuits Systems
Conf., 2011.

[79]

F. Gozzini, G. Ferrari, and M. Sampietro, "An Instrumenton-chip for Impedance
Measurements on Nanobiosensors with AttoFarad Resolution," IEEE Int. SolidState Circuits Conf., pp. 346-347, 2009.

[80]

A. Yúfera, A. Rueda, J. M. Muñoz, R. Doldán, G. Leger, and E. O. RodríguezVillegas, "A Tissue Impedance Measurement Chip for Myocardial Ischemia
Detection," IEEE Tran. Circuits and Systems, vol. 52, pp. 2620-2628, 2005.

[81]

A. Manickam, A. Chevalier, M. McDermott, A. D. Ellington, and A. Hassibi, "A
CMOS Electrochemical Impedance Spectroscopy (EIS) Biosensor Array," IEEE
Trans. Biom. Circ. Systems, vol. 4, pp. 379-390, 2010.

[82]

A. Manickam, A. Chevalier, M. McDermott, A. D. Ellington, and A. Hassibi, "A
CMOS Electrochemical Impedance Spectroscopy Biosensor Array for Label-Free
Biomolecular Detection," IEEE Int. Solid-State Circuits Conf., pp. 492-493, 2010.

[83]

H. Mazhab-Jafari, L. Soleymani, and R. Genov, "16-Channel CMOS Impedance
Spectroscopy DNA Analyzer with Dual-Slope Multiplying ADCs," IEEE Trans.
Biom. Circ. Systems, vol. 6, pp. 468-478, 2012.

[84]

M. Min and T. Parve, "Improvement of lock-in bio-impedance analyzer for
implantable medical devices," IEEE Trans. Instrum. Meas., vol. 56, pp. 968-974,
2007.

[85]

L. Li, W. Qureshi, X. Liu, and A. J. Mason, "Amperometric Instrumentation
System with On-chip Electrode Array for Biosensor Application," IEEE
Biomedical Circuits Systems Conf., pp. 294-297, 2010.

[86]

D. Grieshaber, R. MacKenzie, J. Voros, and E. Reimhult, "Electrochemical
biosensors - sensor principles and architectures," Sensors and Actuators B:
Chemical, vol. 8, pp. 1400-1458, 2008.

143

[87]

M. Zhang, A. Smith, and W. Gorski, "Carbon nanotube-chitosan system for
electrochemical sensing basd on dehydrogenase enzymes," Ana. Chem., vol. 76,
pp. 5045-5050, 2004.

[88]

S. R. Jadhav, D. Sui, M. R. Garavito, and M. R. Worden, "Fabrication of highly
insulating tethered bilayer lipid membrane using yeast cell membrane fractions
for measuring ion channel activity," Journal of Colloid and Interface Science, vol.
322, pp. 465-472, 2008.

[89]

P. Krysinski, A. Zebrowska, A. Michota, J. Bukowska, L. Becucci, and M. R.
Moncelli, "Tethered mono- and bilayer lipid membranes on Au and Hg,"
Langmuir, vol. 17, pp. 3852-3857, 2001.

[90]

B. Razavi, Design of Analog CMOS Integrated Circuits: McGraw-Hill, 2001.

[91]

C. Yang, Y. Huang, B. L. Hassler, R. M. Worden, and A. J. Mason,
"Amperometric electrochemical microsystem for a miniaturized protein biosensor
array," IEEE Trans. Biomedical Circ. Systems, vol. 3, pp. 160-168, Jun. 2009.

[92]

W. Li, C. Rodger, E. Meng, J. D. Weiland, M. S. Hymayun, and Y. C. Tai,
"Wafer-level parylene packaging with integrated RF electronics for wireless
retinal prostheses," IEEE J. Microelectromechanical Systems, vol. 19, pp. 735742, Aug. 2010.

[93]

Product Specification for SPPOSC Series, Frequency Devices, Inc., Ottawa, IL.
[Online]. Available: http://www.freqdev.com/products/osc/spposc.pdf

[94]

I. A. Khan and S. Khwaja, "An integrable gm-C quadrature oscillator," Int. J. of
Electronics, vol. 87, pp. 1353-1357, 2000.

[95]

B. Linares-Barranco, T. Serrano-Gotarredona, J. Romas-Martos, J. CeballosCaceres, J. Mora, and A. Lnares-Barranco, "A precise 90◦ quadrature OTA-C
oscillator tunable in the 50–130-MHz range," IEEE Trans. Circuits Syst., vol. 51,
pp. 649–664, 2004.

[96]

Z. Tang, O. Ishizuka, and H. Matsumoto, "MOS triangle-to-sine wave convertor
based on subthreshold operation," Electron. Lett.,, vol. 26, pp. 1983-1985, 1900.

[97]

S.-C. Lee and Y. Chiu, "Digital Calibration of Capacitor Mismatch in SigmaDelta Modulators " IEEE Tran. Circuits and Systems, vol. 58, pp. 690-698, 2011.

[98]

W. Yu and G. C. Temes, "A digital calibration technique for DAC mismatches in
delta-sigma modulators," IEEE Int. Conf. Circuit and Systems, pp. 1337 - 1340,
2009.

[99]

B. Hassler, R. M. Worden, A. J. Mason, P. Kim, N. Kohli, J. G. Zeikus, et al.,
"Biomimeticinterfaces for a Multifunctional Biosensor Array Microsystem,"
presented at the IEEE Int. Conf. on Sensors, Vienna, Austria, 2004.
144

[100] S. R. Jadhav, Y. Zheng, R. M. Garavito, and R. M. Worden, "Functional
Characterization of PorB Class II Porin from Neisseria Meningitidis Using a
Tethered Bilayer Lipid Membrane," Biosensors and Bioelectronics, vol. 24, pp.
831-835, Dec 2008.
[101] S. R. Jadhav, D. Sui, R. M. Garavito, and R. M. Worden, "Fabrication of highly
insulating tethered bilayer lipid membrane using yeast cell membrane fractions
for measuring ion channel," J. Colloid Interface Sci., vol. 322, pp. 465–472, 2008.
[102] S. R. Jadhav, Y. Zheng, R. M. Garavito, and R. M. Worden, "Functional
Characterization of PorB class II Porin from Neisseria meningitidis using
Tethered Bilayer Lipid Membrane," Biosensors and Bioelectronics, vol. 24, pp.
831-835, 2008.
[103] B. Raguse, V. Braach-Maksvytis, B. A. Cornell, L. G. King, P. D. J. Osman, R. J.
Pace, et al., "Tethered lipid bilayer membranes: Formation and ionic reservoir
characterization," Langmuir, vol. 14, pp. 648-659, Feb 3 1998.
[104] S. R. Jadhav, D. X. Sui, R. M. Garavito, and R. M. Worden, "Fabrication of
highly insulating tethered bilayer lipid membrane using yeast cell membrane
fractions for measuring ion channel activity," Journal of Colloid and Interface
Science, vol. 322, pp. 465-472, 2008.
[105] C. Yang and A. J. Mason, "Fully Integrated Seven-Order Frequency-Range
Quadrature Sinusoid Signal Generator," IEEE Trans. Instrumentation and
Measurement, vol. 58, pp. 3481-3489, 2009.
[106] X. Liu, D. Rairigh, C. Yang, and A. J. Mason, "Impedance-to-Digital Converter
for Sensor Array Microsystems," IEEE Int. Symp. Circuits and Systems, pp. 353356, May 2009.
[107] R. Hogervorst, J. P. Tero, R. G. H. Eschauzier, and J. H. Huijsing, "A compact
power-efficient 3 V CMOS rail-to-rail input/output operational amplifier for
VLSI cell libraries," IEEE J. Solid State Circuits, vol. 29, pp. 1505-1513, 1994.
[108] N. Weste and D. Harris, CMOS VLSI Design: A Circuits and Systems Perspective,
3 ed.: Addison-Wesley, 2005.
[109] J. Baselga, E. A. Perez, T. Pienkowski, and R. Bell, "Adjuvant Trastuzumab: A
Milestone in the Treatment of HER-2-positive Early Breast Cancer," Oncologist,
pp. 4-12, 2006.
[110] M. J. Piccart-Gebhart, M. Procter, B. Leyland-Jones, A. Goldhirsch, M. Untch, I.
Smith, et al., "Trastuzumab after adjuvant chemotherapy in HER2-positive breast
cancer," N Engl J Med., vol. 353, pp. 1659-1672, 2005.

145

[111] E. H. Romond, E. A. Perez, J. Bryant, V. J. Suman, C. E. Geyer, N. E. Davidson,
et al., "Trastuzumab plus Adjuvant Chemotherapy for Operable HER2-Positive
Breast Cancer," N. Engl. J. Med., vol. 353, pp. 1673-1684, 2005.
[112] R. Bartsch, C. Wenzel, and G. G. Steger, "Trastuzumab in the management of
early and advanced stage breast cancer," Biologics, vol. 1, pp. 19-31, 2007.
[113] D. Slamon, W. Eierman, N. Robert, T. Pienkowski, M. Martin, J. Rolski, et al.,
"Phase III randomized trial comparing doxorubicin and cyclophosphamide
followed by docetaxel (ACT) with doxorubicin and cyclophosphamide followed
by docetaxel and trastuzumab (ACTH) with docetaxel, carboplatin and
trastuzumab (TCH) in HER2 positive early breast cancer patients: BCIRG 006
study," Breast Cancer Res Treat, vol. 96, p. S5, 2009.
[114] Y. Shang, P. R. Singh, M. M. Chisti, R. Mernaugh, and X. Zeng, "Immobilization
of a Human Epidermal Growth Factor Receptor 2 Mimotope-Derived Synthetic
Peptide on Au and Its Potential Application for Detection of Herceptin in Human
Serum by Quartz Crystal Microbalance," Anal. Chem., vol. 83, pp. 8928-8936,
2011.
[115] N. F. Sheppard, R. C. Tucker, and C. Wu, "Electrical-Conductivity Measurements
Using Microfabricated Interdigitated Electrodes," Analytical Chemistry, vol. 65,
pp. 1199-1202, 1993.
[116] Z. Shen, M. Huang, C. Xiao, Y. Zhang, X. Zeng, and P. G. Wang, "Nonlabeled
quartz crystal microbalance biosensor for bacterial detection using carbohydrate
and lectin recognitions," Anal Chem, vol. 79, pp. 2312-9, Mar 15 2007.
[117] Y. Shang, P. R. Singh, M. M. Chisti, R. Mernaugh, and X. Zeng, "Immobilization
of a human epidermal growth factor receptor 2 mimotope-derived synthetic
peptide on Au and its potential application for detection of herceptin in human
serum by quartz crystal microbalance," Anal Chem, vol. 83, pp. 8928-36, Dec 1
2011.
[118] Y. Wang, S. Chalagalla, T. Li, X. L. Sun, W. Zhao, P. G. Wang, et al.,
"Multivalent interaction-based carbohydrate biosensors for signal amplification,"
Biosens Bioelectron, vol. 26, pp. 996-1001, Nov 15 2010.
[119] G. Fernández-Ballester, A. Fernández-Carvajal, J. M. González-Ros, and A.
Ferrer-Montiel, "Ionic Channels as Targets for Drug Design: A Review on
Computational Methods," Pharmaceutics, vol. 3, pp. 932-953, 2011.
[120] P. J. England, "Discovering ion channels modulators: making the
electrophysiologists’ life more interesting," Drug Discov. Today, vol. 4, pp. 391392, 1999.
[121] Available: http://en.wikipedia.org/wiki/Patch_clamp

146

[122] M. Zagnoni, M. E. Sandison, P. Marius, A. G. Lee, and H. Morgana, "Controlled
delivery of proteins into bilayer lipid membranes on chip," Lab Chip, vol. 7(9), pp.
1176-1183, 2007.
[123] J. Rosenstein, S. Sorgenfrei, and K. L. Shepard, "Noise and Bandwidth
Performance of Single-Molecule Biosensors," IEEE Custom Integrated Circuits
Conf., pp. 19-21, 2011.
[124] J. Rosenstein, V. Ray, M. Drndic, and K. L. Shepard, "Nanopore DNA Sensors in
CMOS with On-chip Low-noise Preamplifiers," Transducers, pp. 874-877, 2011.
[125] L. Gu and J. W. Shim, "Single Molecule Sensing by Nanopores and Nanopore
Devices," Analyst, vol. 3, pp. 441-451, 2010.
[126] S. Sorgenfrei and K. Shepard, "Single-Molecule Electronic Detection Using
Nanoscale Field-Effect Devices," Design Automation Conf., pp. 712-717, 2011.
[127] T. Mach, C. Chimerel, J. Fritz, N. Fertig, M. Winterhalter, and C. Futterer,
"Miniaturzied planar lipid bilayer: increased stability, low electric noise and fast
fluid perfusion," Analytical and Bioanalytical Chemistry, vol. 390, pp. 841-846,
2007.
[128] H. Suzuki, B. L. Pioufle, and S. Takeuchi, "Ninety-six-well planar lipid bilayer
chip for ion channel recording fabricated by hybrid stereolithography," Biomed.
Microdevices, pp. 17-22, 2009.
[129] B. L. Pioufle, H. Suzuki, K. V. Tabata, H. Noji, and S. Takeuchi, "Lipid bilayer
microarray for parallel recording of transmmbrane ion currents," Anal. Chem., pp.
328-332, 2008.
[130] H. Suzuki, K. Tabata, Y. Kato-Yamada, H. Noji, and S. Takeuchi, "Planar lipid
bilayer reconstitution with a micro-fluidic system," Lab chip, pp. 502-505, 2004.
[131] X. Liu, L. Li, and A. J. Mason, "Thermal Control Microsystem for Protein
Characterization and Sensing," IEEE Biomedical Circuits Systems Conf., pp. 277280, Nov. 2009.

147