Numerical and symbolic MOR techniques using hierarchical circuit structure

Similar documents
Symbolic Model Reduction for Linear and Nonlinear DAEs

SyreNe System Reduction for Nanoscale IC Design

System Reduction for Nanoscale IC Design (SyreNe)

(A) Opening Problem Newton s Law of Cooling

IMPERIAL COLLEGE OF SCIENCE, TECHNOLOGY AND MEDICINE UNIVERSITY OF LONDON DEPARTMENT OF ELECTRICAL AND ELECTRONIC ENGINEERING EXAMINATIONS 2010

EE 321 Analog Electronics, Fall 2013 Homework #3 solution

Print Name : ID : ECE Test #1 9/22/2016

Switched-Capacitor Circuits David Johns and Ken Martin University of Toronto

Determining the Existence of DC Operating Points in Circuits

Chapter 2 Combinational Logic Circuits

Electrical Circuits I

Analogue Filters Design and Simulation by Carsten Kristiansen Napier University. November 2004

System Level Modeling of Microsystems using Order Reduction Methods

Industrial Technology: Electronic Technology Crosswalk to AZ Math Standards

Introduction to Linear Programming (LP) Mathematical Programming (MP) Concept (1)

x y x y 15 y is directly proportional to x. a Draw the graph of y against x.

Number of (Nested) Shopping Carts Three nested shopping carts are shown.

ELECTRONIC SYSTEMS. Basic operational amplifier circuits. Electronic Systems - C3 13/05/ DDC Storey 1

Differential Impedance finally made simple

Institute for Mathematics and its Applications

Structure preserving Krylov-subspace methods for Lyapunov equations

EE16B, Spring 2018 UC Berkeley EECS. Maharbiz and Roychowdhury. Lecture 4A: Overview Slides. State Space Representations

Parallel VLSI CAD Algorithms. Lecture 1 Introduction Zhuo Feng

Engineering *S54683A0119* Pearson BTEC Level 3 Nationals

Tennessee s State Mathematics Standards Precalculus

Switching Current Study: Hysteresis Measurement of Ferroelectric Capacitors using Current-Voltage Measurement Method

BEHAVIORAL MODELING AND TRANSIENT ANALYSIS WITH ANALOG INSYDES

Transient Sensitivity Analysis CASA Day 13th Nov 2007 Zoran Ilievski. Zoran Ilievski Transient Sensitivity Analysis

CISC - Curriculum & Instruction Steering Committee. California County Superintendents Educational Services Association

Summary Notes ALTERNATING CURRENT AND VOLTAGE

DIGITAL CIRCUIT LOGIC BOOLEAN ALGEBRA (CONT.)

The Distance Formula. The Midpoint Formula

EE100Su08 Lecture #9 (July 16 th 2008)

1. The positive zero of y = x 2 + 2x 3/5 is, to the nearest tenth, equal to

MA 123 September 8, 2016

BENCHMARKING OF PIEZOCERAMIC MATERIALS FOR GENERATOR APPLICATION

Chapter 2 Combinational Logic Circuits

Solving Equations. Pure Math 30: Explained! 255

UNAVAILABILITY CALCULATIONS WITHIN THE LIMITS OF COMPUTER ACCURACY ABSTRACT

Modeling, Simulation and Optimization of a Cross Flow Molten Carbonate Fuel Cell

THE ACOUSTIC IMPEDANCE MEASUREMNET SYSTEM USING TWO MICROPHONES

Lecture slides by Kevin Wayne

General procedure for formulation of robot dynamics STEP 1 STEP 3. Module 9 : Robot Dynamics & controls

Model Order Reduction using SPICE Simulation Traces. Technical Report

1/11/2011. Chapter 4: Variability. Overview

Contents Part I Introduction The COMSON Project Part II Partial Differential Algebraic Equations PDAE Modeling and Discretization

CHAPTER 5 : REDUCTION OF MULTIPLE SUBSYSTEMS

A-LEVEL Electronics. ELEC1 Introductory Electronics Mark scheme June Version V1: Final Mark Scheme

H(s) = 2(s+10)(s+100) (s+1)(s+1000)

Statement on. Usage of BoK at our institute

MATH 280 Multivariate Calculus Fall 2012

Simultaneous equations for circuit analysis

Course Summary. The course cannot be summarized in one lecture.

Simulating a Power System

Variational Methods for Solving Warped Multirate PDAEs

EE/ME/AE324: Dynamical Systems. Chapter 7: Transform Solutions of Linear Models

Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi

Chapter 9. Electromagnetic Waves

Modified Picard Iteration Applied to Boundary Value Problems and Volterra Integral Equations.

Feedback Control G 1+FG A

ME451 Kinematics and Dynamics of Machine Systems

XOR - XNOR Gates. The graphic symbol and truth table of XOR gate is shown in the figure.

Electrical Circuits I

Binary Decision Diagrams and Symbolic Model Checking

Algebra II B Review 5

Parallel Numerical Algorithms

Yet More On Decoupling, Part 5 When Harry Regulator Met Sally Op-Amp Kendall Castor-Perry

Prepare for this experiment!

ADI-preconditioned FGMRES for solving large generalized Lyapunov equations - A case study

Introduction to Model Order Reduction

Good representations via complete clause-learning

SNAP Centre Workshop. Solving Systems of Equations

Acoustic Research Institute ARI

PURPOSE: See suggested breadboard configuration on following page!

Relativistic Behavior Detection through Electron Acceleration

Systems of Linear Equations

Applied Linear Algebra in Geoscience Using MATLAB

Unprovability of circuit upper bounds in Cook s theory PV

AC : DEVELOPMENT OF VISUALIZATION TOOLS FOR ONE AND MULTIPLE DOF DYNAMIC SYSTEMS

Computing Phase Noise Eigenfunctions Directly from Steady-State Jacobian Matrices

XII. PROCESSING AND TRANSMISSION OF INFORMATION

Time: 1 hour 30 minutes

Robert W. Brodersen EECS140 Analog Circuit Design

AN INTRODUCTION CONTROL +

(Refer Slide Time: 0:16)

Grade 8 Big Ideas SY

0.1 Infinite Limits 2 = (2 1) =. Finally, we conclude that

Physics 1502: Lecture 25 Today s Agenda

EE 227A: Convex Optimization and Applications October 14, 2008

DRAFT. Pre-calculus Curriculum Map Quarter 1 Chapters P 2. Extraneous Critical numbers Test intervals

Q 1. Richland School District Two 8th Grade Mathematics Pacing Guide. Last Edit: 1/17/17

10/14/2009. Reading: Hambley Chapters

Physics 240 Fall 2005: Exam #3. Please print your name: Please list your discussion section number: Please list your discussion instructor:

Free download from not for resale. Apps 1.1 : Applying trigonometric skills to triangles which do not have a right angle.

D is the voltage difference = (V + - V - ).

UNIVERSITY OF CAMBRIDGE INTERNATIONAL EXAMINATIONS General Certificate of Education Advanced Level

Fairfield Public Schools

Nonlinear opamp circuits

Numeracy across the curriculum

Information in Biology

Transcription:

Numerical and symbolic MOR techniques using hierarchical circuit structure

Numerical and symbolic MOR techniques using hierarchical circuit structure COMSON Autumn School Future Developments in Model Order Reduction September 21-25, 2009, Terschelling, The Netherlands Oliver Schmidt

Slide 3

Overview Symbolic Analysis and Reduction Techniques Hierarchical Modelling, Reduction, and Examples An Algorithm for the Exploitation of the System Structure Example: Operational Amplifier Outlook Slide 4

Nonlinear Symbolic Analysis simplified DAE system original DAE system netlist description behavioral model reference simulation original simplified Slide 5

Symbolic Model Reduction specifications: error control: goal: inputs u, outputs v, numerical analysis A (AC, transient, ) error boundû, error function E find DAE system G with reduced complexity and defined accuracy DAE F R1 R 2 R k DAE G v F = A( F, u) E( v, F vg ) < ε v G = A( G, u) Slide 6

Symbolic Reduction Techniques algebraic manipulations: elimination of variables, removal of independent blocks of equations x 0 = = f ( y) g( x, y) ( f ( y y) 0= g ), term reductions: cancellation of terms, applicable to nested expressions (level concept) F j : N i= 1 t ( x) = 0 i G j N : ti ( x ɶ ) = 0 k i= 1 Slide 7

Symbolic Reduction Algorithm Slide 8

Overview Symbolic Analysis and Reduction Techniques Hierarchical Modelling, Reduction, and Examples An Algorithm for the Exploitation of the System Structure Example: Operational Amplifier Outlook Slide 9

Hierarchical Modelling system is composition of building blocks/subsystems with connection structure subsystems themselves composed of different components ( levels) hierarchical reduction: - reduce subsystems singly - replace subsystems by obtained reduced models - obtain entire reduced system entire system 1 1 1 1 2 2 2 2 f ( x, y, z ) = 0 f ( x, y, z ) = 0 f 3 f ( x 0 3 ( x, y 3 1,, z 3 y 1, x 2, y 2, x, x ) = 0 f 3, y 3 4, y 4 ( x 4 4 ) = 0 subsystem 1 subsystem 2 subsystem 3 subsystem 4, y 4, z 4 ) = 0 netlist-based model behavioral model (DAE) distributed model (PDE) Slide 10

Differential Amplifier Circuit transmission lines between sources and circuit components discretized PDE model with 20 line segments / discretization points 12 V original system: 167 equations 645 terms (level 0) -12 V Slide 11

Non-hierarchical Reduction symbolic non-hierarchical reduction: error bound 2% and 10% error function E( v, v ) = sup v ( t) v ( t) F G F G t I 2% 10% approx. 2h 10min 124 equations, 425 terms (level 0) 44 equations, 284 terms (level 0) Slide 12

A modelling approach is aspired that transmits this information into the set of equations! Intuitive Hierarchy intuitive hierarchy standard graph theoretical methods like MNA / STA lose this information DUT DUT2 Slide 13 Rleitung1

Slide 14 sub sub sub subsystem system system system a b c d test test test test bench bench bench bench sub sub sub subsystem system system system sub sub sub subsystem system system system sub sub sub subsys sys sys sys. sub sub sub subsys sys sys sys. Subsystem Reduction Workflow

Hierarchical Reduction hierarchical reduction: numerical reduction via state space form and Arnoldi s algorithm within seconds: 50 8 resp. 4 equations symbolic reduction within seconds (2% error permitted): 16 9 equations, 59 20 terms symbolic reduction within seconds (2% error permitted): 22 13 equations, 91 50 terms Almost no changes if 10% error is permitted! Slide 15

Results replace subsystems by reduced models: 167 62 equations, 645 252 terms Mostly dependent on the numerical reduction of the transmission lines! usage of more modes for transmission line 1 ( 8 12 equations): 167 66 equations, 645 396 terms Slide 16

Model Check three other inputs V1 Slide 17

Overview Symbolic Analysis and Reduction Techniques Hierarchical Modelling, Reduction, and Examples An Algorithm for the Exploitation of the System Structure Example: Operational Amplifier Outlook Slide 18

Subsystem Sensitivity entire system 1. provide k error bounds 2. choose appropriate reduction methods for all subsystems S 3. reduce subsystem S k- times allowing all of the k error bounds 4. each time check errors on the entire system sensitivity Sε 1 S { ε ε } 1,..., k... Sε k entire system 5. repeat this for all subsystems Slide 19

Algorithm exploitation of the system s hierarchy Slide 20

Overview Symbolic Analysis and Reduction Techniques Hierarchical Modelling, Reduction, and Examples An Algorithm for the Exploitation of the System Structure Example: Operational Amplifier Outlook Slide 21

Example operational amplifier op741 Slide 22

Example operational amplifier op741 CM1 CM3 DP LS CM4 CM2 DAR PP Slide 23

OpAmp741 transient input Vid: sine wave, 0.8V maximum amplitude, 1kHz frequency original system: non-hierarchical reduction: 215 equations, 1050 terms (level 0) 61 equations, 495 terms (level 0) Slide 24 error error: error 10% 10% 10% 10% not not not not always always always always feasible feasible feasible feasible! > > > > 12h 12h 12h 12h!!!!!!!!!!!! perm. error perm perm perm

Error Functions f g : = max min f ( x ) g( x) i x U ε ( x ) i i f g f g f g : = max f ( x ) g( x ) i i i ( ) Slide 25

Subsystem Sensitivities error sweep: reduction method: sw = { 1, 2,5,10, 20,30, 40,50, 60, 70,80,90,100} [%] term cancellation (level 0), algebraic manipulations eqns. eqns. eqns. 5 6 8 4 3 5 4 3 7 6 5 4 2 2 3 1 1 5 10 20 30 40 50 60 70 80 90 100 error 1 1 5 10 20 30 40 50 60 70 80 90 100 error 2 1 1 5 10 20 30 40 50 60 70 80 90 100 error terms terms terms 14 12 10 8 6 4 2 CM1 18 16 14 12 10 8 6 4 2 CM2 30 27 24 21 18 15 12 9 6 3 CM3 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error Slide 26

Subsystem Sensitivities error sweep: reduction method: sw = { 1, 2,5,10, 20,30, 40,50, 60, 70,80,90,100} [%] term cancellation (level 0), algebraic manipulations eqns. 18 16 14 12 10 8 6 4 2 eqns. 40 36 32 28 24 20 16 12 8 4 eqns. 21 18 15 12 9 6 3 eqns. 24 21 18 15 12 9 6 3 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error terms 120 100 80 DAR terms 300 250 200 DP terms 120 100 LS terms 160 140 120 PP 60 150 80 60 100 80 40 20 100 50 40 20 60 40 20 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error 1 5 10 20 30 40 50 60 70 80 90 100 error Slide 27

Results subsystem perm. error # equations CM1 5 50% 20 5 original system: 215 equations, 1050 terms (level 0) CM2 CM3 DAR 1 90% 50 60% 20 90% 21 6 30 8 33 8 DP 10 40% 72 22 (23) hierarchical reduction: 132 equations, LS PP 10 60% 40 90% 39 10 42 13 approx. 1h 48min 336-356 terms (level 0) further non-hierarchical reduction: approx. 17min perm. error: 10% 34 equations, 93 terms (level 0) Slide 28

Model Check three other inputs Vid Slide 29

Overview Symbolic Analysis and Reduction Techniques Hierarchical Modelling, Reduction, and Examples An Algorithm for the Exploitation of the System Structure Example: Operational Amplifier Outlook Slide 30

Outlook subsystem ranking 1. not clear how to spread the entire system s error bound over the subsystems 2. if error check fails: not totally clear which subsystem s error bound has to be altered error function system segmentation ranking / sensitivity analysis results depend on the choice of the error function how to get / what is a good segmentation of the entire system into subsystems? Slide 31

Thank you for your attention. The work presented in this talk has been carried out within the project SyreNe (Systemreduktion für IC Design in der Nanoelektronik, grant no. 03LAPAE6) which is supported by the German Federal Ministry of Education and Research (BMBF). Slide 32