Distributed Microsystems Laboratory

Integrated Interface Circuits for Chemiresistor Arrays

Carina K. Leung* and Denise Wilson, Associate ProfessorDepartment of Electrical Engineering

University of Washington*Now with Intel, Dupont Washington

Integrated Interface Circuits for Chemiresistor Arrays

Outline• Project Description (High Density Chemiresistor Arrays)

• Chemiresistor Background

• Project Context

• Circuit Approach 1: Differential Measurement of Resistance

• Circuit Approach 2: Resistance-to-Frequency Conversion

• Comparison of Approaches

• Summary

• Acknowledgements

Project Description

• Popular approach to chemical sensing (“traditional”)

• Small number (highly selective) sensors in an

• Application targeted to 1-2 analytes

• In an “understood” background

• Another approach to chemical sensing (“olfactory”)

• Large number (broad, overlapping selective) sensors in an

• Application targeted to many analytes

• And their (many) interferents

• In a cluttered and complicated background

• Candidates for high density arrays of chemical sensors are few:

• Require small size, linear operation, broad selectivity, compatibility with integration, and

room temperature operation

Chemiresistor Background

• Composite polymer chemiresistors

• Conductive Element (such as carbon black) combined with

• Chemically sensitive element (polymer)

• Basic operation

• Polymer “swells” in response to target analytes

• Conductive particles move farther apart (conductivity increases)

• Linear response at low concentrations

• R-Ro = Ro (k) [C]

• Ro= baseline resistance (large and highly variable)

• [C] = analyte concentration

• Superposition can be applied to multiple analytes presented simultaneously

Project Context

High resolution Sensor Arrays

• Require Integration

• Circuits produced in CMOS

• Gold post-deposited electrochemically

• Sensor coating “sprayed” on gold

• 1-2 layers of metal required for sensor

• Challenge: Design processing circuits that

• Ignore large, variable baseline resistance

• Amplify very small changes in polymer

resistance on top of large baselines

• Conform to VLSI footprint that addresses:

• Electrode Geometry

• Required sensor density

• Circuit performance

Circuit Approach #1

Differential Approach• On-chip chemiresistor divided into:

– One chemically sensitive resistor

– One or (three) reference resistors

• Passivated (responsive to zero analytes) or

• Exposed, not functionalized (responsive to all

analytes)

• Resistive “Bridge” is part of sensor

• Remaining circuits are designed for maximum gain

under constrained footprint (= sensor platform)

Circuit Approach #1

Differential Approach• Resistive Bridge output transferred to:

– Differential Amplifier

– Comparator with ramping input for

serial A/D conversion

• Design constraints:

– Differential Amplifier: maximum gain

in small footprint

– Comparator: fully serial (simple) A/D

conversion acceptable because of slow

sensor response time

Vdd

Rbias

RBaseline

+-

+-

Out

Vdd

Rbias

RSensor

Circuit Approach #1

Differential Approach• Circuit Gain

– 20 (Differential Amplifier)

– -20 (Comparator)

• Sensor Performance:

– Bridge approach eliminates effect

of broad range in baseline on

circuit gain

– However, additional bias resistors

add more noise (electrical and

transduction)

Translation:– 25V detection limit

– Independent of baseline

– 0.01% (R) detection limit and resolution

Circuit Approach #2

Resistance to Frequency Conversion• Sensor platform contains three terminals:

– Outer ring terminals shorted together outside sensor

platform to enable circuits to fit underneath

– Allows a single resistor per platform for chemical

sensing

– More “active” area (fill factor) than previous

approach.

– Electrode geometry more readily optimized for best

noise performance.

Circuit Approach #1

Resistance to Frequency Conversion• Operation:

– Sensor resistance charges Co

– As the capacitor charges, it trips the Schmitt trigger, causing the feedback to discharge the capacitor

– The frequency of the charge/discharge cycle becomes smaller with increasing resistance (smaller current)

• Hysteresis reduces impact of noisy sensor response

RSensor

C’

C

SchmittTrigger

Out

Co

Circuit Approach #2

Resistance to Frequency Conversion

• Sensitivity:

– Baseline (730k) = .12%/

– Baseline (9.26k) = 4.1%/

• Resolution/Detection Limit:

– Change in resistance from baseline

– Baseline (730k) = .07%

– Baseline (9.26k) = .02%

Comparison

• Both circuits fit underneath sensor platform (.04 mm2 area)• Fill Factor:

• Approach #1: 25%• Approach #2: close to 100% (with exceptions for metal routing)

• Sensitivity: – Approach #1: 400 (V/V)

– Approach #2: between .12%/and 4.1%/

• Resolution/Detection limit: – Approach #1: .01% change in resistance

– Approach #2: between .02% and .07%

• Other:

– Approach #2: more resilience to fluctuations in response due to built in hysteresis.

Summary

We have designed and fabricated two circuits for processing the response of composite polymer chemiresistors. Performance enables sub-ppm detection of many common analytes, while having having zero impact on sensor area.

Acknowledgements

• The authors would like to thank Nathan Lewis and his

graduate group at the California Institute of

Technology for data and technical assistance, as well

as a subcontract through CalTech on ARO Grant

DAAG55-98-1-0266.