Unified HighLevel Synthesis and Module Placement for DefectTolerant Microfluidic Biochips

1
Unified High-Level Synthesis and Module Placement
for Defect-Tolerant Microfluidic Biochips

• Fei Su and Krishnendu Chakrabarty
• Electrical and Computer Engineering
• Duke University

2
Motivation
Shrink
Microfluidic Biochips
Microfluidic Lab-on-Chip
Conventional Biochemical Analyzer

• Potential applications
• Clinical diagnostics (e.g., health care for
  premature infants)
• Bio-smoke alarm
• Massively parallel DNA analysis

3
Motivation (Cont.)

• Increasing application complexity and design
  complexity

4

• Integration of microfluidics one of the
  system-level design challenges (beyond 2009)

2003 International Technology Roadmap for
Semiconductors (ITRS)
Heterogeneous SOCs -Mixed-signal -Mixed-technology
MEMS components
Digital blocks
Analog blocks
Microfluidic components
5
Outline

• Motivation
• Background
• Related prior work
• Unified synthesis methodology
• Problem formulation
• PRSA-based algorithm
• Enhancement for defect tolerance
• Evaluation example
• Summary

6
Background Microfluidic Biochip

• Integrate all necessary functions for biochemical
  analysis into one chip using microfluidics
  technology.
• Continuous-flow microfluidics vs. digital
  microfluidics

(University of Michigan) 1998
7
Background Digital Microfluidic Biochips

• Droplet actuation is achieved through
  electrowetting-on-dielectric
• Electrical modulation of the solid-liquidinterfac
  ial tension

Applied Potential The droplets surface energy
increases, which results in a reduced contact
angle. The droplet now wets the surface.
No Potential A droplet on a hydrophobic surface
originally has a large contact angle.
8
Background (Cont.)

• A droplet can be transported by removing a
  potential on the current electrode, and applying
  a potential to an adjacent electrode.

http//www.ece.duke.edu/Research/microfluidics/
9
Background (Cont.)

• Digital microfluidic biochips system level

MIXERS
TRANSPORT
DISPENSING
REACTORS
DETECTION
INTEGRATE

• Basic microfluidic functions (transport,
  splitting, merging, and mixing) have already been
  demonstrated on a 2-D array
• Microfluidic components -
  reconfigurable virtual devices -
  non-reconfigurable resources

10
Design Methodology

• State-of-the-art methodologies for VLSI design

11
Biochip Design Methodology

• Bottom-up vs. top-down design methodologies

Top-down design methodology
12
Related Prior Work

• Synthesis of integrated circuitswell-studied
  problem
• MEMS simulation synthesis tools
• Commercial CAD tools for microfluidic biochips
• Physical-level simulation CFD-ACE, FlumeCAD
• Synthesis tools for digital microfluidic biochips
• Architectural-level synthesis (Su Chakrabarty
  ICCAD04)
• Physical design automation (Su Chakrabarty
  DATE05)

13
Decoupled Synthesis Methodology

• (Su Chakrabarty, ICCAD04 DATE05)

• Scheduling of operations
• Binding to functional
• resources
• Physical design

14
Unified Synthesis Methodology

• Problem Formulation

15
Parallel recombinative simulated annealing
(PRSA)-based algorithm
16
PRSA-Based Algorithm (Cont.)

• Representation of a chromosome

Chromosome gene(1), , gene(k),
gene(k1),, gene(2k), gene(2k1),, gene(3k)
17
PRSA-Based Algorithm (Cont.)

• Construction procedure
• Phase I Resource binding
• Phase II Scheduling
• Phase III Placement
• Multi-objective optimization
• A Biochip array area
• T Bioassay completion time
• Metric (??A/Amax(1??)? T/Tmax)

18
Enhancement for Defect Tolerance

• Reconfiguration
• Modified PRSA-based algorithm
• Objective (1) minimize T (2) accommodate all
  components in the fabricated array
• Resource constraints defect-free components
• Placement phase (1) locations of defective cells
  are no longer available (2) locations of
  non-reconfigurable resources are fixed.

Fabricated biochip array
Defective cell
19
Protein Assay

• Sequencing graph model

• Maximum array area 10x10
• Maximum number of optical detectors 4
• No. of reservoirs 1 for sample 2
  for buffer 2 for reagent 1
  for waste
• Maximum bioassay time 400 s

20
Experimental Evaluation (Cont.)

• Microfluidic module library for synthesis

21
Experimental Evaluation (Cont.)

• Baseline techniques
• Full-custom design
• Architectural-level synthesis

5x8 14 lt10x10 (satisfies the resource
constraint in architectural-level synthesis)
T 560 s gt Tmax 400 s
Fail to meet the design specification!
22
Experimental Evaluation (Cont.)

• Results of the unified synthesis method

Bioassay completion time T 363 seconds lt
Tmax400 s
Biochip array 9x9 array lt 10x10 array
23
Experimental Evaluation (Cont.)

• Results of the unified synthesis method

Complete digital microfluidic biochip design
24
Experimental Evaluation (Cont.)

• Defect tolerance

Bioassay completion time T 385 seconds (6
increase)
25
Summary

• New unified synthesis methodology for digital
  microfluidic biochips
• PRSA algorithms
• Scheduling bioassay operations
• Resource binding
• Microfluidic module placement
• Real-life bioassay experimental evaluation
• Broader impact of the proposed research
• Facilitate biochip design automation
• Pave the way for the integration of biochip IP
  blocks in the next-generation SOC design