ECE 546 Lecture 15 Circuit Synthesis

Similar documents
ECE 451 Circuit Synthesis

ECE 451 Macromodeling

ECE 451 Black Box Modeling

ECE 546 Lecture 13 Scattering Parameters

EE5900 Spring Lecture 5 IC interconnect model order reduction Zhuo Feng

ECE 451 Advanced Microwave Measurements. TL Characterization

ECE 451 Advanced Microwave Measurements. Requirements of Physical Channels

Some of the different forms of a signal, obtained by transformations, are shown in the figure. jwt e z. jwt z e

Control Systems Engineering (Chapter 2. Modeling in the Frequency Domain) Prof. Kwang-Chun Ho Tel: Fax:

A FAST MULTI-PURPOSE CIRCUIT SIMULATOR USING THE LATENCY INSERTION METHOD PATRICK KUANLYE GOH DISSERTATION

Electronic Circuits EE359A

Network Synthesis Part I

Advanced Analog Building Blocks. Prof. Dr. Peter Fischer, Dr. Wei Shen, Dr. Albert Comerma, Dr. Johannes Schemmel, etc

ECE 497 JS Lecture - 13 Projects

ECE Networks & Systems

Enforcement Passivity. Frequency Data. Wenliang Tseng, Sogo Hsu, Frank Y.C. Pai and Scott C.S. Li. Asian IBIS Summit, Taipei, Taiwan November 12, 2010

ECE 497 JS Lecture - 18 Noise in Digital Circuits

ECE 546 Lecture 16 MNA and SPICE

ECE 546 Lecture 07 Multiconductors

ECE 342 Electronic Circuits. Lecture 25 Frequency Response of CG, CB,SF and EF

Physics for Scientists & Engineers 2

12/20/2017. Lectures on Signals & systems Engineering. Designed and Presented by Dr. Ayman Elshenawy Elsefy

ECE 598 JS Lecture 06 Multiconductors

ECE 497 JS Lecture -03 Transmission Lines

MALLA REDDY ENGINEERING COLLEGE (AUTONOMOUS) B.Tech I YEAR II SEMESTER-ECAS(EEE) QUESTION BANK (OBJECTIVE)

ECE 546 Lecture 19 X Parameters

Stability and Passivity of the Super Node Algorithm for EM Modeling of IC s

Lecture 7: Laplace Transform and Its Applications Dr.-Ing. Sudchai Boonto

Circuit Theory Prof. S.C. Dutta Roy Department of Electrical Engineering Indian Institute of Technology, Delhi

Equivalent Lumped Element Network Synthesis for Distributed Passive Microwave Circuits

(amperes) = (coulombs) (3.1) (seconds) Time varying current. (volts) =

MODELING OF CONTROL SYSTEMS

ECE 546 Lecture 11 MOS Amplifiers

Model-Order Reduction of High-Speed Interconnects: Challenges and Opportunities

Qualification of tabulated scattering parameters

IBIS Modeling Using Latency Insertion Method (LIM)

OPERATIONAL AMPLIFIER APPLICATIONS

Total No. of Questions :09] [Total No. of Pages : 03

Two-Layer Network Equivalent for Electromagnetic Transients

INTEGRATION OF INTERCONNECT MODELS IN A CIRCUIT SIMULATOR

Power Transformer Transient Modeling Using Frequency Response Analysis

Source-Free RC Circuit

Graduate Diploma in Engineering Circuits and waves

ECE 342 Electronic Circuits. Lecture 6 MOS Transistors

Circuit Analysis Using Fourier and Laplace Transforms

ECEN 420 LINEAR CONTROL SYSTEMS. Lecture 6 Mathematical Representation of Physical Systems II 1/67

ECE 497 JS Lecture - 18 Impact of Scaling

Case Study: Parallel Coupled- Line Combline Filter

Basic. Theory. ircuit. Charles A. Desoer. Ernest S. Kuh. and. McGraw-Hill Book Company

Electrical Circuits I

Sophomore Physics Laboratory (PH005/105)

Network Parameters of a Two-Port Filter

Section 5 Dynamics and Control of DC-DC Converters

ECE 342 Electronic Circuits. Lecture 34 CMOS Logic

Dynamic circuits: Frequency domain analysis

ECE 497 JS Lecture -07 Planar Transmission Lines

Synthesis of passband filters with asymmetric transmission zeros

EE221 Circuits II. Chapter 14 Frequency Response

Texas A & M University Department of Mechanical Engineering MEEN 364 Dynamic Systems and Controls Dr. Alexander G. Parlos

Single-Time-Constant (STC) Circuits This lecture is given as a background that will be needed to determine the frequency response of the amplifiers.

f I d required pulse current amplitude

EE221 Circuits II. Chapter 14 Frequency Response

Networks and Systems Prof. V. G. K. Murti Department of Electrical Engineering Indian Institution of Technology, Madras

MATHEMATICAL MODELING OF CONTROL SYSTEMS

vtusolution.in Initial conditions Necessity and advantages: Initial conditions assist

Wideband Modeling of RF/Analog Circuits via Hierarchical Multi-Point Model Order Reduction

How to measure complex impedance at high frequencies where phase measurement is unreliable.

LAB MANUAL EXPERIMENT NO. 7

Chapter 6: The Laplace Transform. Chih-Wei Liu

Chapter 2 Time-Domain Representations of LTI Systems

ECE 202 Fall 2013 Final Exam

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

Lecture 6: Impedance (frequency dependent. resistance in the s- world), Admittance (frequency. dependent conductance in the s- world), and

1, I. Compensation (cont.) Operational Amplifier. Ks 2 S O L U TI OO N S

ECE-202 FINAL April 30, 2018 CIRCLE YOUR DIVISION

Study of the Electric Guitar Pickup

IEEE TRANSACTIONS ON ADVANCED PACKAGING 1

Discrete-Time State-Space Equations. M. Sami Fadali Professor of Electrical Engineering UNR

Transient Response of a Second-Order System

The norms can also be characterized in terms of Riccati inequalities.

MODELING AND SIMULATION OF EMBEDDED PASSIVES USING RATIONAL FUNCTIONS IN MULTI-LAYERED SUBSTRATES

Network Synthesis. References :

EE348L Lecture 1. EE348L Lecture 1. Complex Numbers, KCL, KVL, Impedance,Steady State Sinusoidal Analysis. Motivation

AS VERY large scale integration (VLSI) technology advances

Electric Circuit Theory

Canonical Forms of Multi-Port Dynamic Thermal Networks

INF5490 RF MEMS. LN03: Modeling, design and analysis. Spring 2008, Oddvar Søråsen Department of Informatics, UoO

QUESTION BANK SUBJECT: NETWORK ANALYSIS (10ES34)

ECE 255, Frequency Response

Chapter 2 Voltage-, Current-, and Z-source Converters

e st f (t) dt = e st tf(t) dt = L {t f(t)} s

5. Observer-based Controller Design

A Novel Tunable Dual-Band Bandstop Filter (DBBSF) Using BST Capacitors and Tuning Diode

ECE 546 Lecture 04 Resistance, Capacitance, Inductance

Network Theory 1 Analoge Netzwerke

Transfer Function and Impulse Response Synthesis using Classical Techniques

Voltage-Controlled Oscillator (VCO)

Statistical approach in complex circuits by a wavelet based Thevenin s theorem

Hurwitz Stable Reduced Order Modeling and D2M Delay Calculation Metric Motivation Review: Moments. Content. Moments: Review. Motivation.

XVI. CIRCUIT THEORY. Prof. S. J. Mason A. B. Lehman Prof. H. J. Zimmermann R. D. Thornton A. SOME LIMITATIONS OF LINEAR AMPLIFIERS

Transcription:

ECE 546 Lecture 15 Circuit Synthesis Spring 018 Jose E. Schutt-Aine Electrical & Computer Engineering University of Illinois jesa@illinois.edu ECE 546 Jose Schutt Aine 1

MOR via Vector Fitting Rational function approximation: N cn f ( s) d sh s a n 1 n Introduce an unknown function σ(s) that satisfies: Poles of f(s) = zeros of σ(s): ( sf ) ( s) N cn d sh s a n 1 n N ( s) c n f ( s) N n 1 1 s a n 1 n n1 N N cn n 1 n n 1 n N 1 cn d sh s z s a 1 s z s a n n Flip unstable poles into the left half plane. Im s-domain Re ECE 546 Jose Schutt Aine

Blackbox Formulation Transfer function is approximated H( ) d L k1 ck 1 j / Using curve fitting technique (e.g. vector fitting) In the time domain, recursive convolution is used ck where y() t dx( t T) y () t L pk k 1 () ( )1 ckt ckt y t a x tt e e y ( tt) pk k pk Recursive convolution is fast ECE 546 Jose Schutt Aine 3

Passivity Enforcement State space form: Hamiltonian matrix: x Ax Bu y Cx Du M T T T T C LC A C DKB T T A BKD C BKB T T 1 1 K I D D L I DD Passive if M has no imaginary eigenvalues. Sweep: H eig I S( j) S( j) Quadratic programming: Minimize (change in response) subject to (passivity compensation). T min vec( ΔC) H vec( ΔC) subject to = G vec( ΔC). ECE 546 Jose Schutt Aine 4

Macromodel Circuit Synthesis Use of Macromodel Time-Domain simulation using recursive convolution Frequency-domain circuit synthesis for SPICE netlist ECE 546 Jose Schutt Aine 5

Macromodel Circuit Synthesis Objective: Determine equivalent circuit from macromodel representation Motivation Circuit can be used in SPICE Goal Generate a netlist of circuit elements ECE 546 Jose Schutt Aine 6

Circuit Realization Circuit realization consists of interfacing the reduced model with a general circuit simulator such as SPICE Model order reduction gives a transfer function that can be presented in matrix form as or S s Y s s11 s s1 N s sn1 s snn s y11 s y1 N s yn1 s ynn s ECE 546 Jose Schutt Aine 7

Method 1: Y-Parameter/MOR* Each of the Y-parameters can be represented as y(s) ij d L k1 ak s p where the a k s are the residues and the p k s are the poles. d is a constant k *Giulio Antonini "SPICE Equivalent Circuits of Frequency- Domain Responses", IEEE Transactions on Electromagnetic Compatibility, pp 50-51, Vol. 45, No. 3, August 003. ECE 546 Jose Schutt Aine 8

Equivalent-Circuit Extraction Macromodel is curve-fit to take the form Constant term d Real Poles Complex Poles Y(s) d L k1 rk s p Need to find equivalent circuit associated with k ECE 546 Jose Schutt Aine 9

Y-Parameter - Circuit Realization Each of the I ij can be realized with a circuit having the following topology: The resulting current sources can then be superposed for the total current I i leaving port i All Y parameters are treated as if they were one-port Y parameters ECE 546 Jose Schutt Aine 10

Y-Parameter - Circuit Realization We try to find the circuit associated with each term: L aijk y(s) ij dij s p k1 For a given port i, the total current due to all ports with voltagesv j (j=1,,p) is given by P P P L aijk Ii yijvj dijvj V j s p j1 j1 j1 k1 ijk For each contributing port, with voltage i, the total current due to a voltage V j at port j is given by L aijk Iij dijvj V j s p k1 ijk ijk ECE 546 Jose Schutt Aine 11

Y-Parameter - Circuit Realization We try to find the circuit associated with each term: y(s) 1. Constant term d ij d L k1 s a k p k y ijd(s) d. Each pole-residue pair y (s) ijk a s k p k ECE 546 Jose Schutt Aine 1

Circuit Realization Constant Term R 1 d ECE 546 Jose Schutt Aine 13

Circuit Realization Pole/Residue In the pole-residue case, we must distinguish two cases (a) Pole is real y (s) ijk s a k p k (b) Complex conjugate pair of poles y (s) ijk k jk k jk s j s j k k k k In all cases, we must find an equivalent circuit consisting of lumped elements that will exhibit the same behavior ECE 546 Jose Schutt Aine 14

Case (a) - Real Pole Consider the circuit shown above. The input impedance Z as a function of the complex frequency s can be expressed as: Z sl R Y s L 1/ a k from circuit a s p 1/ L Ŷ(s) k s R/ L Comparing Y(s) and Ŷ(s) yields the solution R p k / ak from pole and residue k ECE 546 Jose Schutt Aine 15

Circuit Realization - Complex Poles Each term associated with a complex pole pair in the expansion gives: r1 r Yˆ s p s p 1 Where r 1, r, p 1 and p are the complex residues and poles. They satisfy: r 1 =r * and p 1 =p * It can be re-arranged as: Yˆ r r 1 1 1 / 1 s r p r p r r s s p p p p 1 1 ECE 546 Jose Schutt Aine 16

Circuit Realization - Complex Poles The pole/residue representation of the Y parameters is given by: Yˆ r r 1 1 1/ 1 s r p r p r r s s p p p p 1 1 p pp 1 product of poles DEFINE a r r 1 sum of residues g p p 1 sum of poles x rp rp 1 1 cross product ˆ s x/ a Y a s sg p ECE 546 Jose Schutt Aine 17

Circuit Realization - Complex Poles Consider the circuit shown above. The input impedance Z as a function of the complex frequency s can be expressed as: 1 Z slr slr 1 1 1/ R sc 1sCR R Z 1 R sl scr R 1 1 scr ECE 546 Jose Schutt Aine 18

Circuit Realization - Complex Poles Y LCR s s CR s 1/ CR L CRR LR C 1 R 1 R LR C Y 1 s1/ CR L L CRR R R 1 s s LRC LRC 1 ECE 546 Jose Schutt Aine 19

Circuit Realization - Complex Poles Comparing Y 1 s1/ CR L L CRR R R 1 s s LRC LRC 1 We can identify the circuit elements with ˆ s x/ a Y a s sg p L 1/ a R x a 1 g a R p x g pa 1 g C x a a x a x ECE 546 Jose Schutt Aine 0

Circuit Realization - Complex Poles L 1/ a R x a 1 g a R p x g pa 1 g C x a a x a x ECE 546 Jose Schutt Aine 1

Negative Elements In the circuit synthesis process, it is possible that some circuit elements come as negative. To prevent this situation, we add a contribution to the real parts of the residues of the system. In the case of a complex residue, for instance, assume that r1 r Yˆ s p s p 1 Yˆ r r s p1 s p s p1 s p 1 Augmented Circuit Compensation Circuit Can show that both augmented and compensation circuits will have positive elements ECE 546 Jose Schutt Aine

Why S Parameters? Y-Parameter Test line: Zc, l Y 1 e 11 l Z (1 e ) c Z c : microstrip characteristic impedance : complex propagation constant l : length of microstrip Y 11 can be unstable short Reference Line Zo S-Parameter Test Line Test line: Zc, l ( 1 e ) S 11 l 1 e Z c Z c Z o Z o Reference Line S 11 is always stable Zo ECE 546 Jose Schutt Aine 3

Y Versus S Parameters S Y Impulse Responses Observation: S-parameters decay rapidly; Y parameters do not. ECE 546 Jose Schutt Aine 4

Optimizing Reference System 1 1 1 o o S ZZ I ZZ I 1 o using Z I S I S Z 1.5 S11 - Linear Magnitude Z o = 50.0 0.0 0.0 0.0 0.0 50.0 0.0 0.0 0.0 0.0 50.0 0.0 0.0 0.0 0.0 50.0 as reference S11 1 using 0.5 S11-50 Ohm S11 - Zref Z o = 38.0 69.6 38.9 69.6 69.6 38.8 69.6 38.9 38.9 69.6 38.8 69.6 69.6 38.9 69.6 38.8 as reference 0 0 4 6 8 10 Frequency (GHz) ECE 546 Jose Schutt Aine 5

Method : S-Parameter /MOR 1 A V Z I i i o i Need equivalent circuit for S ijk 1 Bi ViZoIi All S parameters are treated as if they were one-port S parameters ECE 546 Jose Schutt Aine 6

Strategy For a given circuit, a relationship between the input admittance Y ijk (s) of the circuit and the associated one-port S-parameter representation S ijk (s) can be described by S (s) ijk Y Y o o Y (s) ijk Y (s) ijk 1 S (s) ijk Y ijk(s) Yo 1 S ijk (s) Y o is the reference admittance ECE 546 Jose Schutt Aine 7

Equivalent-Circuit Extraction Macromodel is curve-fit to take the form S( s ) d L k1 s r k p k Need to find equivalent circuit associated with Constant term d Real Poles Complex Poles ECE 546 Jose Schutt Aine 8

Constant Term 1 S ijk 1 d R Yo Y 1 S ijk 1d o S ijk R=Y o (1-d)/(1+d) ECE 546 Jose Schutt Aine 9

S- Circuit Realization Constant Term 1 d R Yo 1 d ECE 546 Jose Schutt Aine 30

S-Parameters - Poles and Residues In the pole-residue case, we must distinguish two cases (a) Pole is real s (s) ijk s a k p k (b) Complex conjugate pair of poles s (s) ijk k jk k jk s j s j k k k k In all cases, we must find an equivalent circuit consisting of lumped elements that will exhibit the same behavior ECE 546 Jose Schutt Aine 31

S-Realization Real Poles Proposed Circuit Model Admittance of proposed model is given by: 1 s R R R R C Y 1 1 RR 1 1 s RC ECE 546 Jose Schutt Aine 3

Real Poles Y from circuit 1 s R R ak 1 R R C 1 S ijk(s) RR 1 s pk 1 s RC r 1 ˆ 1 S ijk s p s a Y Yo Yo Y o 1 S r ijk 1 s a s p from pole and residue ECE 546 Jose Schutt Aine 33

Solution for Real Poles Comparing Y ˆ( s) with Y () s gives C bzo b a R 1 bc R R 1 1 ac where a p r, and b p r k k k k ECE 546 Jose Schutt Aine 34

Realization Complex Poles From the S-parameter expansion, the complex pole pair gives: Sˆ 1 1 1 1 r r s r r r p r p s p s p s s p p p p 1 1 1 which corresponds to an admittance of: sa x ˆ 1 ˆ 1 S s sg p Y Yo Y 1 Sˆ sa x 1 s sg p o ECE 546 Jose Schutt Aine 35

Realization Complex Poles The admittance expression can be re-arranged as ˆ s sg psax s s ga px Y Y o Y s sg psax s s ga px o p pp 1 product of poles WE HAD DEFINED a r r 1 sum of residues g p p 1 sum of poles x rp rp 1 1 cross product ECE 546 Jose Schutt Aine 36

Realization Complex Poles ˆ s sg psax s s ga px Y Y o Y s sg psax s s ga px This can be further rearranged as o in which ˆ s sab Y s sdf E A ( ga), B px, D ( ga), F px and E Y o 1 H ECE 546 Jose Schutt Aine 37

Realization Complex Poles There are several circuit topologies that will work Model 1 Model 9 Model 13 Model 1 ECE 546 Jose Schutt Aine 38

Realization Complex Poles More circuit topologies that will work Model 10 Model 11 Model 8 Model 1 ECE 546 Jose Schutt Aine 39

Complex Poles Model 1 In particular, if we choose model 1 Y LRR CR R C R R R s s 1 LCR LCR R o LRR 1 C R1R s s LCR LCR o 1 o o 1 Must be matched with sa x 1 1 ˆ S ijk s sg p Y Yo Y 1 S sa x ijk 1 s sg p o ECE 546 Jose Schutt Aine 40

Complex Poles Model 1 Matching the terms with like coefficients gives p x R R R o LCR 1 R o 1 Y o p x R R LCR 1 o p R R R LCR 1 R o x LCR ECE 546 Jose Schutt Aine 41

Complex Poles Model 1 Solving gives R R o p x 1 R 1 L R o a R 1 1 gro Lx a a R o C a Rx ECE 546 Jose Schutt Aine 4

Complex Poles Model 9 If we choose Model 9 Y 9 A B E L CRR3 CR1R R3 R R 3 s s R1 R3 LC R1R3 LC R1R3 RR 1 3 CR R 1RR3 LR1 1R R3 s s LCR1R3 LCR1R3 D F from which the circuit elements can be extracted ECE 546 Jose Schutt Aine 43

Complex Poles Model 9 F R 1 BH C BD AF H F R BD BF ADF F B ABD BD A F BF ADF F H R 3 F L B D A F B FH B ABD BD A F BF ADF F H ECE 546 Jose Schutt Aine 44

Special Case Model 4 A special case exists when x=0 or x rp 1 rp 1 0 ˆ s s ga px s s ga p Y Y o Y s s ga px s s ga p o in which ˆ s sa B Y s sdb E A ( ga), B px, D ( ga), F p and E Y o 1 H ECE 546 Jose Schutt Aine 45

Special Case Model 4 The proposed circuit model is Model 4 This model has one inductor, one capacitor and two resistors. ECE 546 Jose Schutt Aine 46

Special Case Model 4 The admittance seen at the input is: R 1 DE A Y 4 A B F E 1 1 s s R1 R C R 1 R LC RR 1 1 1 s s CR LC D FB The element are extracted as R R E R 1 1 E DR 1 C 1 L FC ECE 546 Jose Schutt Aine 47

Method 3 - Convolution with S Parameters In frequency domain B=SA In time domain Convolution: b(t) = s(t)*a(t) s(t)*a(t) s(t )a( )d ECE 546 Jose Schutt Aine 48

SPICE Macromodel STAMP stamp 1 o 1 Y Z 1s'(0 ) 1 s'(0 ) stamp o 1 1 I Z 1s'(0) H(t) Most of the computational effort is in the convolution calculation of the history H(t) at each time step. t1 Ht () s( ) at ( ) We make the convolution faster using a -function expansion for s(t) 1 This operation can be accelerated ECE 546 Jose Schutt Aine 49

In the frequency domain, assume that S parameter can be expanded in the form: Sl () M s( p) c ( pk) k 1 S-Parameter Expansion M j lk ce k k 1 In the time domain, this corresponds to: k Which is an impulse train of order M. Convolution will then give: M M y( p) s( p)* x( p) ck ( pk) x( p) ckx( pk) k1 k1 If we truncate the summation to the Q largest impulses: Q Q y( p) ck ( pk) x( p) ckx( pk) k1 k1 If the reference system is optimized, most of the coefficients c k will be negligibly small; therefore Q will be a relatively small number Fast Convolution is achieved by making Q small ECE 546 Jose Schutt Aine 50

Strategy s ( u) c ( u k) ijk ijk Since the S-parameter is a simple delay, the oneport circuit representation is a transmission line with characteristic impedance Z o and delay k ECE 546 Jose Schutt Aine 51

Circuit Interpretation s ( u) c ( u k) ijk ijk kt t time step d ECE 546 Jose Schutt Aine 5

Method 1: Y-Parameters/MOR Recursive convolution SPICE realization ECE 546 Jose Schutt Aine 53

Method : S-Parameters/MOR Recursive convolution SPICE realization ECE 546 Jose Schutt Aine 54

Comparisons SPICE simulation from MOR generated Netlist (Method ) Direct Convolution ECE 546 Jose Schutt Aine 55

Coupled Lines (4-port) Direct SPICE simulation Using generated netlist (Method ) ECE 546 Jose Schutt Aine 56

Comparison of S-methods SPICE Netlist from MOR with S (Method ) SPICE Netlist from Fast convolution with S (Method 3) Convolution (no SPICE netlist) ECE 546 Jose Schutt Aine 57