Google Glass for Pathology

Page 1

Google Glass Pathology Dr. Liron Pantanowitz, MD

1

Š Copyright AgO Inc 2011


Google Glass

2

Š Copyright AgO Inc 2011


100X Faster Simulation - Six Sigma Optimization

3

Š Copyright AgO Inc 2011


SoC Design Objectives Maximising yield • Validating greater numbers of process, voltage and temperature corners • Efficiently centering design across all PVT corners using Monte Carlo Achieving design specification • Meeting or beating performance while minimising cost of implementation • Managing greater complexity in operating and power saving modes Respin avoidance • Analog circuits are responsible for ~ 50% of IC design re-spins • Re-spins can mean missing market windows and unbudgeted costs Design porting • Moving existing circuit designs to similar technologies • Re-centering design to meet constraints of new technology 4

© Copyright AgO Inc 2011


Analog Design Flow • Design methodology has changed little over the years • Manual, iterative design with many SPICE runs Design specification & constraints

Define topology & resize devices

Spice

Physical layout & adjust Spice Extraction Layout verification 5

© Copyright AgO Inc 2011


Massively Parallel Optimization

Feasibility

Description • DC Operation • Increase Margin • Quick check • Best Design Space

Global Optimization

Description • Single Corner • Meet performance • Monte Carlo • Ready for Center

Centering

Description • All Corners • Rapid Size • Ready for P&R

AgO Optimization Strategy 6

© Copyright AgO Inc 2011

Database

Description • Conflict • Change Priority • System Analysis


AnXplorer Goals • • • • • •

7

Quickly size W/L a circuit in a given technology Explore suitability of different design options Robust design over PVT & Monte Carlo Support all types of devices Explore results using database Optimize production yield

© Copyright AgO Inc 2011


AgO Design Methodology AnXplorer automates device sizing Design specification & constraints

Define topology

Feasibility Global

Physical layout & adjust

Extraction Layout verification

8

Š Copyright AgO Inc 2011

Centering


Process Porting Common challenge • Port existing design in technology X (say 180 nm) to technology Y (in 180 nm) • Ensure that original design goals are met

AnXplorer approach • Start with original sized circuit • Define variable ranges for target circuit • “One click” command • Optimises and centers with new PVT corners

9

© Copyright AgO Inc 2011

Vendor A

Vendor Node Porting Vendor B


Prioritized Design Objectives • Most tools support weight-based prioritisation for multiple objectives – Designer often unsure of relative weights – bad design practice

• AnXplorer supports hierarchical design objectives – User defines relative priority

• AnXplorer achieves important objectives before optimizing low priority signals

10

© Copyright AgO Inc 2011


Implicit Objectives • Imposes implicit objectives on conditions of devices at DC operating point • Customizable Implicit objectives • Detects common sub-circuits and imposes constraints on their operating conditions • Ensures a robust DC • Available for MOS devices only

Examples of subcircuits:

• transistors in • • • • • • •

11

© Copyright AgO Inc 2011

saturation transistors in linear region current mirrors level shifters differential pairs voltage reference current mirror banks etc


Core Optimization Technology Early Optimisation tools • Frequently relied on traditional convex/gradient methods • These are known to have difficulty with multiple local minima AnXplorer • Based on advanced Evolutionary algorithm • Capable of finding global minimum in presence of many local minima • Successfully optimised tough tests Rastrigin’s function • logarithmic partitioning of design space 12

© Copyright AgO Inc 2011

Supports both simulation-based optimization & equation-based optimization


Multiple Local Minimum

13

Š Copyright AgO Inc 2011


Trade-off Analysis Database • Finds multiple design points satisfying design objectives • Creates exploration database for postoptimization analysis – Database stores all explored design points – Query language or GUI

• Useful for trade-off analysis with conflicting objectives • Useful for “what-if” analysis 14

© Copyright AgO Inc 2011


Industry Standard Formats Compatible with existing design flows Un-sized circuit Schematics

Definition of Design variables

Design objectives

AnXplorer

Sized and centered net list

15

Š Copyright AgO Inc 2011

Exploration database for Trade off analysis


Design Environment • Spice Simulators – – – –

Cadence Synopsys Mentor Multi-threading support

• Operating system – 64 Bit Red Hat RHEL 5

16

© Copyright AgO Inc 2011


Differentiation • Push bottom integrated technology • Robust Circuits to maximise yield –Monte Carlo • Implicit objectives for stable DC operation • Hierarchical design objectives • Trade-off analysis database • Industry standard simulators • Advanced Evolutionary Algorithm

17

© Copyright AgO Inc 2011


Contact Hillol Sarkar Hillol.Sarkar@ago-inc.com

18

Š Copyright AgO Inc 2011


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.