τ /2π S, ρ < ρ <... generate Dirac distribution δ ( τ ). Spectral density

Similar documents
Comparison between the Discrete and Continuous Time Models

Additional Exercises for Chapter What is the slope-intercept form of the equation of the line given by 3x + 5y + 2 = 0?

Elements of Stochastic Processes Lecture II Hamid R. Rabiee

Approximating the Powers with Large Exponents and Bases Close to Unit, and the Associated Sequence of Nested Limits

( ) ( ) if t = t. It must satisfy the identity. So, bulkiness of the unit impulse (hyper)function is equal to 1. The defining characteristic is

F (u) du. or f(t) = t

An Introduction to Malliavin calculus and its applications

7.3. QUANTUM MECHANICAL TREATMENT OF FLUCTUATIONS: THE ENERGY GAP HAMILTONIAN

28. Narrowband Noise Representation

The Quantum Theory of Atoms and Molecules: The Schrodinger equation. Hilary Term 2008 Dr Grant Ritchie

Stochastic Reliability Analysis of Two Identical Cold Standby Units with Geometric Failure & Repair Rates

2 int T. is the Fourier transform of f(t) which is the inverse Fourier transform of f. i t e

Physics 127b: Statistical Mechanics. Fokker-Planck Equation. Time Evolution

Vehicle Arrival Models : Headway

Block Diagram of a DCS in 411

Method For Solving Fuzzy Integro-Differential Equation By Using Fuzzy Laplace Transformation

û s L u t 0 s a ; i.e., û s 0

Laplace Transforms. Examples. Is this equation differential? y 2 2y + 1 = 0, y 2 2y + 1 = 0, (y ) 2 2y + 1 = cos x,

On two general nonlocal differential equations problems of fractional orders

THE CATCH PROCESS (continued)

Fuzzy Laplace Transforms for Derivatives of Higher Orders

Lecture 20: Riccati Equations and Least Squares Feedback Control

EE 224 Signals and Systems I Complex numbers sinusodal signals Complex exponentials e jωt phasor addition

III. Direct evolution of the density: The Liouville Operator

Chapter 14 Wiener Processes and Itô s Lemma. Options, Futures, and Other Derivatives, 9th Edition, Copyright John C. Hull

Chapter 1 Fundamental Concepts

The Optimal Stopping Time for Selling an Asset When It Is Uncertain Whether the Price Process Is Increasing or Decreasing When the Horizon Is Infinite

The Potential Effectiveness of the Detection of Pulsed Signals in the Non-Uniform Sampling

5. Response of Linear Time-Invariant Systems to Random Inputs

Mathematics Paper- II

Recursive Least-Squares Fixed-Interval Smoother Using Covariance Information based on Innovation Approach in Linear Continuous Stochastic Systems

Class Meeting # 10: Introduction to the Wave Equation

POSITIVE AND MONOTONE SYSTEMS IN A PARTIALLY ORDERED SPACE

4. Advanced Stability Theory

DYNAMIC ECONOMETRIC MODELS Vol. 7 Nicolaus Copernicus University Toruń Piotr Fiszeder Nicolaus Copernicus University in Toruń

Nature Neuroscience: doi: /nn Supplementary Figure 1. Spike-count autocorrelations in time.

SUFFICIENT CONDITIONS FOR EXISTENCE SOLUTION OF LINEAR TWO-POINT BOUNDARY PROBLEM IN MINIMIZATION OF QUADRATIC FUNCTIONAL

Delay and Its Time-Derivative Dependent Stable Criterion for Differential-Algebraic Systems

Structural Dynamics and Earthquake Engineering

Sensors, Signals and Noise

Filtering Turbulent Signals Using Gaussian and non-gaussian Filters with Model Error

Stochastic Model for Cancer Cell Growth through Single Forward Mutation

in Engineering Prof. Dr. Michael Havbro Faber ETH Zurich, Switzerland Swiss Federal Institute of Technology

Chapter 4. Truncation Errors

THE SINE INTEGRAL. x dt t

On a Fractional Stochastic Landau-Ginzburg Equation

ln y t 2 t c where c is an arbitrary real constant

Asymmetry and Leverage in Conditional Volatility Models*

CS537. Numerical Analysis

ADDITIONAL PROBLEMS (a) Find the Fourier transform of the half-cosine pulse shown in Fig. 2.40(a). Additional Problems 91

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010

Zürich. ETH Master Course: L Autonomous Mobile Robots Localization II

14 Autoregressive Moving Average Models

The Arcsine Distribution

Lecture 1 Overview. course mechanics. outline & topics. what is a linear dynamical system? why study linear systems? some examples

arxiv: v1 [physics.data-an] 14 Dec 2015

A NEW TECHNOLOGY FOR SOLVING DIFFUSION AND HEAT EQUATIONS

For example, the comb filter generated from. ( ) has a transfer function. e ) has L notches at ω = (2k+1)π/L and L peaks at ω = 2π k/l,

non-linear oscillators

Higher Order Difference Schemes for Heat Equation

R t. C t P t. + u t. C t = αp t + βr t + v t. + β + w t

Section 3.5 Nonhomogeneous Equations; Method of Undetermined Coefficients

The fundamental mass balance equation is ( 1 ) where: I = inputs P = production O = outputs L = losses A = accumulation

Navneet Saini, Mayank Goyal, Vishal Bansal (2013); Term Project AML310; Indian Institute of Technology Delhi

CHAPTER 2 Signals And Spectra

ON THE NUMBER OF FAMILIES OF BRANCHING PROCESSES WITH IMMIGRATION WITH FAMILY SIZES WITHIN RANDOM INTERVAL

Institute for Mathematical Methods in Economics. University of Technology Vienna. Singapore, May Manfred Deistler

Haar Wavelet Operational Matrix Method for Solving Fractional Partial Differential Equations

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin

12: AUTOREGRESSIVE AND MOVING AVERAGE PROCESSES IN DISCRETE TIME. Σ j =

Variational Iteration Method for Solving System of Fractional Order Ordinary Differential Equations

Solutions to Assignment 1

02. MOTION. Questions and Answers

MODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE

The electromagnetic interference in case of onboard navy ships computers - a new approach

Notes on Kalman Filtering

STRUCTURAL CHANGE IN TIME SERIES OF THE EXCHANGE RATES BETWEEN YEN-DOLLAR AND YEN-EURO IN

METHOD OF CHARACTERISTICS AND GLUON DISTRIBUTION FUNCTION

Answers to Exercises in Chapter 7 - Correlation Functions

Math 2142 Exam 1 Review Problems. x 2 + f (0) 3! for the 3rd Taylor polynomial at x = 0. To calculate the various quantities:

DEPARTMENT OF ELECTRICAL AND ELECTRONIC ENGINEERING EXAMINATIONS 2008

Guest Lectures for Dr. MacFarlane s EE3350 Part Deux

Pade and Laguerre Approximations Applied. to the Active Queue Management Model. of Internet Protocol

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

Lectures 29 and 30 BIQUADRATICS AND STATE SPACE OP AMP REALIZATIONS. I. Introduction

STATE-SPACE MODELLING. A mass balance across the tank gives:

Chapter 7 Response of First-order RL and RC Circuits

Robust estimation based on the first- and third-moment restrictions of the power transformation model

Chapter 2. First Order Scalar Equations

Ordinary Differential Equations

where the coordinate X (t) describes the system motion. X has its origin at the system static equilibrium position (SEP).

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

EECE 301 Signals & Systems Prof. Mark Fowler

Sub Module 2.6. Measurement of transient temperature

Sample Autocorrelations for Financial Time Series Models. Richard A. Davis Colorado State University Thomas Mikosch University of Copenhagen

Outline Chapter 2: Signals and Systems

Solution of Integro-Differential Equations by Using ELzaki Transform

V AK (t) I T (t) I TRM. V AK( full area) (t) t t 1 Axial turn-on. Switching losses for Phase Control and Bi- Directionally Controlled Thyristors

POSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION

Chapter One Fourier Series and Fourier Transform

Transcription:

IOM and ession of e Commission of Acousics, Bucares 5-6 May COMPARATIVE ANALYI OF THREE IMULATION METHOD FOR BAND LIMITED WHITE NOIE AMPLE Ovidiu OLOMON, Carles W. TAMMER, Tudor IRETEANU 3 *Corresponding auor: solomonovidius@yaoo.com In is paper ree simulaion meods for samples of Gaussian band limied wie noise random process are analyzed. Te relaionsips beween parameers a deermine e specral bandwid of simulaed random processes are obained by imposing e condiion o ave equal mean square values. Keywords: wie noise, Wiener process, socasic differenial equaion, Mone Carlo meod.. PRINCIPLE OF IMULATION METHOD Wie noise is a process widely used in modeling dynamic sysems subjeced o random eciaions. Tis process is no pysically realizable (i.e. energy is infinie). I is used as a useful approimaion of many random processes encounered in pracice, ofen being filered; i leads o imporan simplificaions in e sudy of socasic differenial equaions.. FILTERING METHOD We consider () a saionary random process, normal, zero mean and correlaion funcion []: τ R ( τ) = π / e,, > () ( ) For sufficienly large values of, e properies of socasic process () approimae ose of a saionary wie noise, because e correlaion funcion ( τ ) a epresses e saisical dependence beween () and ( + τ ) akes very small values, even for values very close o zero. Wen is an ineger, e series of funcions R ( ) n τ /π, < <... generae Dirac disribuion δ ( τ ). pecral densiy of random process wi correlaion funcion given above is: R iντ ( ) ( ) ν = R τ e dτ π = () + ( ν / ) Wen is obain ( ν ) =. By analogy wi wie lig a conains all frequency Z componens, is propery was given e name wie noise random process z (). If a random process specral densiy is approimaely consan unil sufficienly ig frequencies, we are dealing wi a broadband random process. In is case, e simulaed wie noise can be a useful approimaion. Romanian-American Universiy, Bvd. Epoziiei, B, RO- Universiy of Ba, Ba BA 7AY, England, 3 Insiue of olid Mecanics of e Romanian Academy, Cons. Mille 5, RO-4

355 Comparaive analysis of ree simulaion meods for band limied wie noise samples Te random process () belongs o e class of saionary random processes wi normal disribuion and raional specral densiy of e form []: were Pi ( υ) ( υ) = (3) Qi ( υ) p q p k q k k k k= k= Pi ( υ) = β ( iυ), Qi ( υ) = α ( iυ) β, k =,,.., p, α, k =,,.., q, p< q k k An imporan propery of ese processes is a ey can be modeled by Iô socasic differenial equaions. From a ecnical poin of view, is propery is equivalen o e fac a any saionary random process wi normal disribuion and raional specral densiy can be obained by filering saionary normal wie noise. Te ransfer funcion of e linear sysem corresponding o is filer and e socasic equaion are Pi ( υ) Hi ( υ) = (5) Qi ( υ) and wi e inpu z () and e oupu () p k d () d z() βk p k d q q k p αk = q k k= d k= (6). By applying (4), e parameers of equaion (6) are β =, α =, α =, p=, q=. In is case, e Iô socasic equaion becomes d () = () d + dw() (7) E dw() = π d. were ( ) (4). FORMAL DIFFERENTIATION OF WIENER PROCEE In e disribuion eory cone [3], saionary wie noise is normally defined as e derivaive of Wiener process, specifying is definiion by e noaion []: dw() = z() d (8) were dw() = w( + d) w() (9) represens e infiniesimal increases of brownian moion w(), aving e properies: Edw [ ( )] =, E ( dw() ) = cd, Edwdw [ ( ) ( )] =, () Te generalized Gaussian wie noise process is defined as e derivaive of Wiener generalized random processes, wi zero mean and auocorrelaion funcion equal wi cδ ( s), were c is a consan a can be aken by a convenien coice equal o [4]. For a fied posiive consan, is defined e random process z () as [5]:

Ovidiu OLOMON, Carles W. TAMMER, Tudor IRETEANU 356 () w ( + ) w ( ) z = () Te process defined above as a zero mean normal disribuion, e auocorrelaion funcion R z ( τ ) = ma, τ and specral densiy is: z sin( πν ) ( ν ) = πν Mean square of random process () is given by z () (3) σ = (4) Te random process defined by (), for small enoug, is a good approimaion of a Gaussian wie noise process [4], [5]..3 MONTE CARLO METHOD A discree ime isory (sample funcion) of Gaussian saionary limied band-noise zn+ = z( n Δ ), n =,,..., N (5) can be derived from a sequence of pseudo-random numbers u, n=,,..., N uniformly disribued on e n inerval [,] [6]. Mos digial compuers include pseudo-random number generaor (Mone Carlo simulaion) aving e following recursive form: = a + b(mod T), (, T], n=,,,... (6) n+ n n+ were b and T are relaively prime. For an ineger iniial value or seed, e relaion () generaes a sequence aking ineger values from o T-, e reminders wen e n+ = an + b are divided by T. Wen e coefficiens a, b and T are cosen appropriaely, e numbers: un = n / T (7) are uniformly disribued on e inerval [,]. If ( un, un), n=,,... represens a series of pairs of suc values, en e following series of values: zn = π/ Δ ln un cos πun, n=,,... (8) z = π / Δ ln u sin πu, n=,,... n n n are asympoically ( n ) independen Gaussian random variables, wi zero mean and mean square value equal o π / Δ. Te discree random process defined by: zδ () = zn, ( n ) Δ nδ (9) is mean square convergen o e wie noise process z () wen Δ..4. CALIBRATION OF AMPLING PARAMETER In order o compare e numerically simulaed samples roug e described meods, e relaions beween parameers, and Δ, are esablised by imposing e same mean square for all obained samples. Te mean square values of processes () and z () are given by

357 Comparaive analysis of ree simulaion meods for band limied wie noise samples and υ σ = d d,, d υ = = π = = dυ + υ + sinπυ sin σ z = dυ = d=, = π, d d πυ π υ = π υ Taking ino accoun e relaions () and () and e fac a e mean square value of random process zδ () is π / Δ, e following relaionsip( = ) is obained π = π = () Δ () (). NUMERICAL IMULATION REULT For e numerical simulaion of ese processes, was cosen e sampling rae oal ime inerval T = s. According o (), =, =, 59. In order o simulae e process (), is used e ieraive relaion [7]: Δ =. ( + Δ ) = ( ) μ + σ n (3) and e were μ e Δ =, σ = ( μ ) and n are independen normally disribued random numbers wi zero mean and mean square value equal o πδ. Figures -3 sows e ime isories wiin e ime inerval -s for ree samples obained by e described meods. 3 σ = 6 3 9. 8 () - - - 3... 4. 6. 8. Fig.. ample of wie noise process ( ) 3 σ = 6 3 8 8. 8 4 () z - - - 3... 4. 6. 8. Fig.. ample of wie noise process z ()

Ovidiu OLOMON, Carles W. TAMMER, Tudor IRETEANU 358 3 σ = 6 3 6 5. 9 8 () z Δ - - - 3... 4. 6. 8. Fig. 3. ample of wie noise process z ( Δ ) As one can see, e mean square values, obained for e ree numerically generaed samples, are almos equal if e sampling parameers fulfill e imposed condiions (). 3. CONCLUION Tree simulaion meods of samples of Gaussian band limied wie noise random process are analyzed, imposing e equaliy of mean square values for e numerically generaed samples. Te wellknown Mone Carlo meod is compared wi wo oer simulaion meods, based on socasic differenial equaions and on e formal derivaive of e Brownian process. Te resuls obained by e las wo meods can be viewed as a rigorous maemaical assessmen of e more pracical Mone Carlo meod. REFERENCE. Jazwinski A. ocasic Proccesses and Filering Teory, Acedemic Press, 97. ireeanu T., iseme mecanice oscilane neliniare supuse la eciaţii aleaoare, eză de docoara, 98 3. Arnold L., ocasic differenial equaions: eory and applicaions, Wiley, 974 4. Bălaş C., Conrolul semiaciv în procese socasice de ip Iô, Asab, 9 5. Kloeden P., Plaen E.,,Numerical oluion of ocasic Differenial Equaions, pringer-verlag, 99 6. Abramowiz M., egun I., Handbook of Maemaical Funcions wi Formulas, Graps, and Maemaical Tables, Dover, 964 7. Gillespie D.T., Eac numerical simulaion of e Ornsein Ulenbeck process and is inegral, Pys.Rev.E 54:84 9, 996