TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall TLT-5200/5206 COMMUNICATION THEORY, Exercise 3, Fall Problem 1.

Similar documents
ENSC327 Communications Systems 2: Fourier Representations. School of Engineering Science Simon Fraser University

Additional exercises in Stationary Stochastic Processes

Figure 3.1 Effect on frequency spectrum of increasing period T 0. Consider the amplitude spectrum of a periodic waveform as shown in Figure 3.2.

Analog Communication (10EC53)

References Ideal Nyquist Channel and Raised Cosine Spectrum Chapter 4.5, 4.11, S. Haykin, Communication Systems, Wiley.

Mixed Signal IC Design Notes set 6: Mathematics of Electrical Noise

Objectives. By the time the student is finished with this section of the workbook, he/she should be able

Introduction to Analog And Digital Communications

RATIONAL FUNCTIONS. Finding Asymptotes..347 The Domain Finding Intercepts Graphing Rational Functions

ELEG 3143 Probability & Stochastic Process Ch. 4 Multiple Random Variables

2. SPECTRAL ANALYSIS APPLIED TO STOCHASTIC PROCESSES

Signals & Linear Systems Analysis Chapter 2&3, Part II

9.1 The Square Root Function

A Fourier Transform Model in Excel #1

Stochastic Processes

2 Frequency-Domain Analysis

Chapter 6. Random Processes

Lecture 13: Applications of Fourier transforms (Recipes, Chapter 13)

IMPROVED NOISE CANCELLATION IN DISCRETE COSINE TRANSFORM DOMAIN USING ADAPTIVE BLOCK LMS FILTER

Notes on Wavelets- Sandra Chapman (MPAGS: Time series analysis) # $ ( ) = G f. y t

Analog Computing Technique

Systems & Signals 315

Probability and Statistics for Final Year Engineering Students

Research Article. Spectral Properties of Chaotic Signals Generated by the Bernoulli Map

Conference Article. Spectral Properties of Chaotic Signals Generated by the Bernoulli Map

Stochastic Processes. M. Sami Fadali Professor of Electrical Engineering University of Nevada, Reno

Longitudinal Waves. Reading: Chapter 17, Sections 17-7 to Sources of Musical Sound. Pipe. Closed end: node Open end: antinode

( x) f = where P and Q are polynomials.

Definition: Let f(x) be a function of one variable with continuous derivatives of all orders at a the point x 0, then the series.

2A1H Time-Frequency Analysis II

Two-step self-tuning phase-shifting interferometry

Scattering of Solitons of Modified KdV Equation with Self-consistent Sources

CHAPTER 8 ANALYSIS OF AVERAGE SQUARED DIFFERENCE SURFACES

10. Joint Moments and Joint Characteristic Functions

In many diverse fields physical data is collected or analysed as Fourier components.

Least-Squares Spectral Analysis Theory Summary

Quadratic Functions. The graph of the function shifts right 3. The graph of the function shifts left 3.

TSKS01 Digital Communication Lecture 1

Random Error Analysis of Inertial Sensors output Based on Allan Variance Shaochen Li1, a, Xiaojing Du2,b and Junyi Zhai3,c

8.4 Inverse Functions

A Systematic Approach to Frequency Compensation of the Voltage Loop in Boost PFC Pre- regulators.

ELEG 3143 Probability & Stochastic Process Ch. 6 Stochastic Process

Parametrization of the Local Scattering Function Estimator for Vehicular-to-Vehicular Channels

Thu June 16 Lecture Notes: Lattice Exercises I

ENSC327 Communications Systems 19: Random Processes. Jie Liang School of Engineering Science Simon Fraser University

Digital Image Processing. Lecture 6 (Enhancement) Bu-Ali Sina University Computer Engineering Dep. Fall 2009

Curve Sketching. The process of curve sketching can be performed in the following steps:

SIO 211B, Rudnick. We start with a definition of the Fourier transform! ĝ f of a time series! ( )

2A1H Time-Frequency Analysis II Bugs/queries to HT 2011 For hints and answers visit dwm/courses/2tf

A UNIFIED FRAMEWORK FOR MULTICHANNEL FAST QRD-LS ADAPTIVE FILTERS BASED ON BACKWARD PREDICTION ERRORS

Signals and Spectra - Review

Double-slit interference of biphotons generated in spontaneous parametric downconversion from a thick crystal

ECE 2100 Lecture notes Wed, 1/22/03

Chapter 2. Basic concepts of probability. Summary. 2.1 Axiomatic foundation of probability theory

Massachusetts Institute of Technology

X-ray Diffraction. Interaction of Waves Reciprocal Lattice and Diffraction X-ray Scattering by Atoms The Integrated Intensity

Probabilistic Model of Error in Fixed-Point Arithmetic Gaussian Pyramid

3. Several Random Variables

The concept of limit

Supplement To: Search for Tensor, Vector, and Scalar Polarizations in the Stochastic Gravitational-Wave Background

EDGES AND CONTOURS(1)

Fig 1: Stationary and Non Stationary Time Series

Chapter 4 Imaging. Lecture 21. d (110) Chem 793, Fall 2011, L. Ma

New Functions from Old Functions

Fundamentals of Noise

An Alternative Poincaré Section for Steady-State Responses and Bifurcations of a Duffing-Van der Pol Oscillator

Stochastic Processes. A stochastic process is a function of two variables:

Numerical Methods - Lecture 2. Numerical Methods. Lecture 2. Analysis of errors in numerical methods

Math Review and Lessons in Calculus

Physics 5153 Classical Mechanics. Solution by Quadrature-1

arxiv: v1 [gr-qc] 18 Feb 2009 Detecting the Cosmological Stochastic Background of Gravitational Waves with FastICA

Sec 3.1. lim and lim e 0. Exponential Functions. f x 9, write the equation of the graph that results from: A. Limit Rules

EECE 301 Signals & Systems Prof. Mark Fowler

Estimation and detection of a periodic signal

Review D: Potential Energy and the Conservation of Mechanical Energy

Philadelphia University Faculty of Engineering Communication and Electronics Engineering

CHAPTER 2 RANDOM PROCESSES IN DISCRETE TIME

ECE-340, Spring 2015 Review Questions

14 - Gaussian Stochastic Processes

Roberto s Notes on Differential Calculus Chapter 8: Graphical analysis Section 1. Extreme points

The achievable limits of operational modal analysis. * Siu-Kui Au 1)

Feasibility of a Multi-Pass Thomson Scattering System with Confocal Spherical Mirrors

1. Interference condition. 2. Dispersion A B. As shown in Figure 1, the path difference between interfering rays AB and A B is a(sin

Numerical Solution of Ordinary Differential Equations in Fluctuationlessness Theorem Perspective

( 1) ( 2) ( 1) nan integer, since the potential is no longer simple harmonic.

13. Power Spectrum. For a deterministic signal x(t), the spectrum is well defined: If represents its Fourier transform, i.e., if.

SOUND. Responses to Questions

Discrete-Time Fourier Transform (DTFT)

Lecture - 30 Stationary Processes

1. Definition: Order Statistics of a sample.

Standing Waves If the same type of waves move through a common region and their frequencies, f, are the same then so are their wavelengths, λ.

Lab 3: The FFT and Digital Filtering. Slides prepared by: Chun-Te (Randy) Chu

STAT 801: Mathematical Statistics. Hypothesis Testing

Physics 107 TUTORIAL ASSIGNMENT #7

APPENDIX 1 ERROR ESTIMATION

( ) ( ) ( ) + ( ) Ä ( ) Langmuir Adsorption Isotherms. dt k : rates of ad/desorption, N : totally number of adsorption sites.

9.3 Graphing Functions by Plotting Points, The Domain and Range of Functions

Intrinsic Small-Signal Equivalent Circuit of GaAs MESFET s

Statistical signal processing

Syllabus Objective: 2.9 The student will sketch the graph of a polynomial, radical, or rational function.

Transcription:

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Problem. The "random walk" was modelled as a random sequence [ n] where W[i] are binary i.i.d. random variables with P[W[i] = s] = p (orward step with probability p) P[W[i] = s] = p (backward step with probability p) n i W[ i] To illustrate, one possible sample path (realization) could look something like 3s s s -s -s [n] ut keep in mind that the above igure is just an example (one realization). We are really talking about statistical characterizations here. n TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall In order to the sequence to be wide-sense stationary (WSS), the autocovariance (or autocorrelation) unction should be shit invariant and the mean unction should be constant (see Carlson 4 th Ed. p. 357-359). This means also that the variance K (m,m) o the sequence should be constant in time as well. Now K but K ( m, m) ( m l, m l) ( m) m(s) p( p) ( m l) ( m l)(s) p( p) ( m) meaning that the variance o this random sequence is increasing in time. Thus, the process is not WSS. Furthermore, the mean unction is ( n) ns( p ) which is also time-dependent (in general) implying non-stationarity. Thus, we conclude that the sequence is NOT WSS! This signal model can also be easily examined using Matlab. In the ollowing, we generate independent realizations (o length samples) with s = and p = /. So according to the theory, the mean and variance are in this case given by [ n] E [ n] ns(p) n ( n) E [ n] [ n] E [ n] min( n, n)() (/4) n These results can now be veriied by calculating the corresponding ensemble averages over the generated independent realizations. The results are illustrated in general in the ollowing two igures. / 6 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Three Independent Realizations 4 4 6 5 5 4 5 5 4 6 8 5 5 Time Index 5 Ensemble Mean and Variance (Estimates) TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Problem. Oscillators are ubiquitous in physical systems, especially electronic and optical ones. For example, in radio requency communication systems they are used or requency translation o inormation signals and or channel selection. Oscillators are also present in digital electronic systems which require a time reerence, i.e., a clock signal, in order to synchronize operations. Noise is o major concern in oscillators, because introducing even small noise into an oscillator leads to dramatic changes in its requency spectrum and timing properties. This phenomenon is known as phase noise or timing jitter. A perect oscillator would have localized tones at discrete requencies (i.e., harmonics), but any corrupting noise spreads these perect tones, resulting in high power levels at neighbouring requencies. This eect is the major contributor to undesired phenomena such as interchannel intererence, leading to increased bit error rates (ER s) in RF communication systems. Another maniestation o the same phenomenon, jitter, is important in clocked and sampled data systems. Uncertainties in switching instants caused by noise lead to synchronization problems. Characterizing how noise aects oscillators is thereore crucial or practical applications. 5 variance In our case, we study the eects o a small phase noise present in a j( t( t)) complex oscillator signal, xt () Ae. In ideal case, () t, and we have only a discrete impulse at requency in the spectral density unction G (). Let s start by rewriting the given signal x(t) as j( t( t)) j t j () t j t x() t Ae Ae e Ae v() t. 5 mean It is important to note that v(t) is a random process and thereore x(t) is also a random process! First we study is the x(t) a WSS process and we start by taking the expected value (or ensemble average) with t held ixed at arbitrary value, as 5 5 5 Time Index 3 / 6 4 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall x t E x t E Ae v t Ae E[()] v t () [ ()] [ j t ()] j t ased on the small noise phase assumption ( () t rad), we can approximate v(t) as vt e cos t jsin t j ( t), j () t ( ) ( ( )) ( ( )) and the expected value simpliies to x t Ae E j t Ae. () j t [ ()] j t We notice that the expected value depends on time, and thereore x(t) is not a WSS process! Even though, x(t) is not a WSS process, we can still calculate the autocorrelation o the random process. When we investigate the relationship between two random variables, we use autocorrelation unction which is deined or complex random unctions as R ( t, t ) E[ x( t ) x*( t )]. This unction measures the relatedness or dependence between two random processes. Now that we have the required tools, let s start with the derivation. R ( t, t ) E[ x( t ) x*( t )] [ ( ) *( )] j t j( t) E Ae v t Ae v t j ( tt) Ae Evt [()*( v t)]. TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Recalling the small noise phase assumption ( () t rad), the expected value E[v(t )v*(t )] simpliies to E[ v( t) v*( t)] E[( j ( t))( j ( t))] E[ ( t ) ( t )]. E[ ( t)] I we now assume that the random variable () t is WSS, we are only interested in the time dierence between the unctions and thereore we mark t =t and t =t-τ. Thus, the autocorrelation unction o x(t) can be rewritten as j ( ) { E[ ( ) ( )]} j R A e t t Ae ( R( ) ) j j Ae Ae R(). We notice that the dependency orm time t disappears and that the autocorrelation unction depends only in the time dierence τ. Now, in the autocorrelation unction we have the original spectral impulse at and in addition we have requency shited version o the autocorrelation unction o () t around requency. Now it is easy to deine the Fourier transorm o R x (τ), which represents the spectral density unction G x (). G( ) F{ R ( )} A ( ) A G ( ). ecause we assumed that () t is a realization rom WSS process, there exists a Fourier transorm pair R ( ) G ( ). 5 / 6 6 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall In the spectral density unction it is even clearer that the power spectrum o the phase noise is shited around requency. The idea is even urther illustrated below. This holds when the phase noise is small. TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Problem 3. The variance o the thermal noise voltage v(t) at the open-circuit terminals o a resistor R at temperature T is given by (Carlson 4 th Ed. p.37) G ( ) Gx ( ) vt () kt v R where 3h k = oltzmann constant =.38* -3 h = Planc constant = 6.6* -34 In the book/course notes, the spectral density unction or the noise process is given as G ( ) x Rh G V ( ) h / kt e So, we immediately note that the process is not exactly white (uncorrelated) because the spectral density unction varies with requency. We could calculate the autocorrelation unction or the process by inverse Fourier transorm o G V (). This autocorrelation unction at delay tells us the value o the covariance or t i - t j =, since the mean is zero (when the mean is zero, autocovariance = autocorrelation). We can calculate the total power o the process directly with the help o the irst ormula or v or by integrating G V () rom -ininity to ininity. We do that with Maple and get 7 / 6 8 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall > k:=.37*^(-3);r:=;t:=9;h:=6.6*^(-34); - k :=.37 R := T := 9-33 h :=.66 > G:=*R*h*abs()/(exp(h*abs()/(k*T))-): > total_power_:=eval(*(pi*k*t)^*r/(3*h)); -5 total_power_ :=.5688746 > total_power_:=int(g,=-ininity...ininity); -5 total_power_ :=.5688746 To get the desired approximation or G V () we proceed as ollows. First we can write G V () as Rh ( ) h GV Rh h / kt e kt h! kt using the terms up to nd order o the series expansion or e x o the orm e x x! x... TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Using only the terms up to the nd order is justiied under the given assumption that h kt. The given orm can be simpliied to G V ( ) h Rh kt h! kt h RkT kt h kt Rh h kt Then we use the trick that (+x) - -x when x is small and we get (use x = h /kt) the desired result h G V ( ) RkT kt The approximation and true density are given in the ollowing Figure. The power in some requency range is given by the integral o G V () over the desired requencies. The integration is easy using the approximation and can be carried out by pen and paper. For example or = - P G ( ) d V symmetry h G V ( ) d 4RkT 4kT Remember, however, that the result is valid only i kt/h. 9 / 6 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall 9 x 9 8. x 6 True Approximation Spectral Density 7 6 5 4 3 Power.8.6.4 True Approximation 6 7 8 9 Frequency [Hz] Figure : Approximative and true power spectral densities (notice the logarithmic scaling o the requency axis).. Maximum Frequency (andwidth) [Hz] Figure : The power in the requency range < using both approximative and true densities. We can also calculate the wanted noise power in the requency range < GHz with Maple, using both the accurate and approximative densities: > G:=*R*k*T*(-h*abs()/(*k*T)): > power_:=int(g,=-^exponent...^exponent): > power_:=int(g,=-^exponent...^exponent): > eval(subs(exponent=9,power_)); -8.588537999 > eval(subs(exponent=9,power_)); -8.588538 Clearly, the approximation is good. In general, the power in the requency range is illustrated in the ollowing Figure using both the approximative and true densities. Finally it was asked how big a portion o the total power is within the requency range < GHz, so >raction:=*eval(subs(exponent=9,power_))/total_power_; raction :=.53485 meaning that only.% o the total noise power is within the requency range < GHz - so 99.9% o the total noise power is outside that range at higher requencies ( GHz and above)! o sounds somewhat strange (?) but this is indeed the case!! o this simply relects the act that we are really dealing with an extremely wideband (close to white) noise process here!! / 6 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Problem 4. When white noise with spectral density / is iltered with H(), the spectral density o the ilter output signal is / H(). This being the case, the power N o the output signal is given by N H ( ) d H ( ) d Noise equivalent bandwidth N is now deined as (see Carlson 4 th Ed. p.378) N g H ( ) d where g = max H(). This means that N is the bandwidth o the ideal ilter having the same maximum power gain and output power as H(). utterworth ilters are characterized with (n = ilter order and = 3d bandwidth) H ( ) / n meaning that g = or all utterworth ilters. We can then use the deinition to get N g H ( ) d / n d TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall To choose the easy way out, we can check some table o standard integrals (e.g. Carlson 4 th Ed., p. 786) and note that / k dx sin( / k) x k, k > Now, use the substitution x=/ and dx=d/. N nsin( / n) sin( / n) / n d n / x and in the limit we get lim lim n sin( / n) / n N n n / n dx sin( / n) because sin(x)/x as x (Hint: Prove with L Hospitals rule). Results are easy to check with Maple: > assume(k>);int(/(+x^k),x=..ininity); > N:=/(sin(Pi/(*n))/(Pi/(*n))); Pi ------------ Pi sin(----) k k Pi N := / --------------- Pi sin(/ ----) n n > limit(n,n=ininity); 3 / 6 4 / 6

TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall TLT-5/56 COMMUNICATION THEORY, Exercise 3, Fall Noise Equivalent andwidth N 9 8 7 6 5 4 3 4 6 8 4 6 8 Filter Order n Figure 3: Noise equivalent bandwidth N as a unction o the ilter order n or a utterworth ilter with a 3d bandwidth o Hz. H().9.8.7.6.5.4.3.. n = n = 4 n = 3 4 5 Frequency [Hz] Figure 4: Squared amplitude responses or three utterworth ilters with 3d bandwidth = Hz..9 3 d bandwidth = Hz In this case the, noise equivalent bandwidth approaches the 3d bandwidth quite rapidly when the ilter order is increased. This means that the utterworth ilters become closer and closer to "ideal" ilters in the average output power sense when the ilter order is increased. With some other ilter structures, this behaviour can be dierent (slower/aster). H().8.7.6.5.4 Noise equivalent bandwidth The ollowing igures illustrate the concept o noise equivalent bandwidth urther.3.. 3 4 5 Frequency [Hz] Figure 5: Squared amplitude response or a utterworth ilter o order n = with 3 d bandwidth = Hz, the noise equivalent bandwidth is also in the igure. 5 / 6 6 / 6