ECE 301 Fall 2010 Division 2 Homework 10 Solutions. { 1, if 2n t < 2n + 1, for any integer n, x(t) = 0, if 2n 1 t < 2n, for any integer n.

Similar documents
ECE 301 Fall 2011 Division 1 Homework 10 Solutions. { 1, for 0.5 t 0.5 x(t) = 0, for 0.5 < t 1

ECE 301 Division 1 Final Exam Solutions, 12/12/2011, 3:20-5:20pm in PHYS 114.

Chapter 5 Frequency Domain Analysis of Systems

Chap 4. Sampling of Continuous-Time Signals

Final Exam of ECE301, Section 3 (CRN ) 8 10am, Wednesday, December 13, 2017, Hiler Thtr.

Homework 4. May An LTI system has an input, x(t) and output y(t) related through the equation y(t) = t e (t t ) x(t 2)dt

X. Chen More on Sampling

ELEN 4810 Midterm Exam

Final Exam of ECE301, Prof. Wang s section 8 10am Tuesday, May 6, 2014, EE 129.

Signals and Systems Spring 2004 Lecture #9

Final Exam of ECE301, Prof. Wang s section 1 3pm Tuesday, December 11, 2012, Lily 1105.

ECE-314 Fall 2012 Review Questions for Midterm Examination II

Chapter 5 Frequency Domain Analysis of Systems

4.1 Introduction. 2πδ ω (4.2) Applications of Fourier Representations to Mixed Signal Classes = (4.1)

ECE 301 Fall 2011 Division 1. Homework 1 Solutions.

EE 224 Signals and Systems I Review 1/10

ELEN E4810: Digital Signal Processing Topic 11: Continuous Signals. 1. Sampling and Reconstruction 2. Quantization

Bridge between continuous time and discrete time signals

Homework: 4.50 & 4.51 of the attachment Tutorial Problems: 7.41, 7.44, 7.47, Signals & Systems Sampling P1

EE Homework 13 - Solutions

Review of Discrete-Time System

The Johns Hopkins University Department of Electrical and Computer Engineering Introduction to Linear Systems Fall 2002.

Overview of Sampling Topics

Lecture 8: Signal Reconstruction, DT vs CT Processing. 8.1 Reconstruction of a Band-limited Signal from its Samples

Review: Continuous Fourier Transform

Final Exam of ECE301, Section 1 (Prof. Chih-Chun Wang) 1 3pm, Friday, December 13, 2016, EE 129.

J. McNames Portland State University ECE 223 Sampling Ver

Assignment 4 Solutions Continuous-Time Fourier Transform

Fourier series for continuous and discrete time signals

George Mason University Signals and Systems I Spring 2016

x[n] = x a (nt ) x a (t)e jωt dt while the discrete time signal x[n] has the discrete-time Fourier transform x[n]e jωn

EE301 Signals and Systems In-Class Exam Exam 3 Thursday, Apr. 19, Cover Sheet

Discrete-time Signals and Systems in

Chapter 6: Applications of Fourier Representation Houshou Chen

New Mexico State University Klipsch School of Electrical Engineering EE312 - Signals and Systems I Fall 2015 Final Exam

ECE 301. Division 2, Fall 2006 Instructor: Mimi Boutin Midterm Examination 3

Solution 7 August 2015 ECE301 Signals and Systems: Final Exam. Cover Sheet

Problem Value

Question Paper Code : AEC11T02

Multirate signal processing

Final Exam 14 May LAST Name FIRST Name Lab Time

Multirate Digital Signal Processing

NAME: 11 December 2013 Digital Signal Processing I Final Exam Fall Cover Sheet

ESE 531: Digital Signal Processing

ESE 531: Digital Signal Processing

ECE 301: Signals and Systems Homework Assignment #7

Homework 6 Solutions

1.1 SPECIAL FUNCTIONS USED IN SIGNAL PROCESSING. δ(t) = for t = 0, = 0 for t 0. δ(t)dt = 1. (1.1)

Homework 7 Solution EE235, Spring Find the Fourier transform of the following signals using tables: te t u(t) h(t) = sin(2πt)e t u(t) (2)

Problem Value

Chapter 4 The Fourier Series and Fourier Transform

EE301 Signals and Systems In-Class Exam Exam 3 Thursday, Apr. 20, Cover Sheet

ECE 301: Signals and Systems Homework Assignment #3

Discrete Time Signals and Systems Time-frequency Analysis. Gloria Menegaz

Signals and Systems Profs. Byron Yu and Pulkit Grover Fall Midterm 2 Solutions

Grades will be determined by the correctness of your answers (explanations are not required).

Discrete-Time Signals: Time-Domain Representation

6.003: Signals and Systems. Sampling and Quantization

ECE 301. Division 3, Fall 2007 Instructor: Mimi Boutin Midterm Examination 3

Solutions to Problems in Chapter 4

Discrete-Time Signals & Systems

ECE 350 Signals and Systems Spring 2011 Final Exam - Solutions. Three 8 ½ x 11 sheets of notes, and a calculator are allowed during the exam.

ESE 531: Digital Signal Processing

Homework 9 Solutions

EEL3135: Homework #3 Solutions

Interchange of Filtering and Downsampling/Upsampling

Homework 3 Solutions

Problem 1. Suppose we calculate the response of an LTI system to an input signal x(n), using the convolution sum:

Chapter 2: Problem Solutions

EE 3054: Signals, Systems, and Transforms Summer It is observed of some continuous-time LTI system that the input signal.

Discrete-Time Signals: Time-Domain Representation

/ (2π) X(e jω ) dω. 4. An 8 point sequence is given by x(n) = {2,2,2,2,1,1,1,1}. Compute 8 point DFT of x(n) by

Fourier Representations of Signals & LTI Systems

Lecture 3 January 23

2.161 Signal Processing: Continuous and Discrete

MEDE2500 Tutorial Nov-7

Chapter 4 The Fourier Series and Fourier Transform

Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science. Fall Solutions for Problem Set 2

2 Background: Fourier Series Analysis and Synthesis

University Question Paper Solution

Signals & Systems. Chapter 7: Sampling. Adapted from: Lecture notes from MIT, Binghamton University, and Purdue. Dr. Hamid R.

Like bilateral Laplace transforms, ROC must be used to determine a unique inverse z-transform.

Lecture 7: Interpolation

Homework 8 Solutions

Fourier transform representation of CT aperiodic signals Section 4.1

Each problem is worth 25 points, and you may solve the problems in any order.

GEORGIA INSTITUTE OF TECHNOLOGY SCHOOL of ELECTRICAL and COMPUTER ENGINEERING

Discrete Fourier Transform

Grades will be determined by the correctness of your answers (explanations are not required).

EC Signals and Systems

QUESTION BANK SIGNALS AND SYSTEMS (4 th SEM ECE)

Signals & Systems. Lecture 5 Continuous-Time Fourier Transform. Alp Ertürk

Discrete-Time Signals and Systems. Efficient Computation of the DFT: FFT Algorithms. Analog-to-Digital Conversion. Sampling Process.

Discrete-Time David Johns and Ken Martin University of Toronto

Core Concepts Review. Orthogonality of Complex Sinusoids Consider two (possibly non-harmonic) complex sinusoids

Final Exam ECE301 Signals and Systems Friday, May 3, Cover Sheet

Discrete Fourier Transform

Lecture 13: Discrete Time Fourier Transform (DTFT)

EECE 301 Signals & Systems Prof. Mark Fowler

Transcription:

ECE 3 Fall Division Homework Solutions Problem. Reconstruction of a continuous-time signal from its samples. Consider the following periodic signal, depicted below: {, if n t < n +, for any integer n, x(t =, if n t < n, for any integer n. x(t 4 t Suppose we use the ideal impulse-train sampling model considered in class, to sample and reconstruct this signal. First, in order to avoid aliasing, we filter the signal using an ideal CT low-pass filter H with passband gain and cut-off frequency ω s /: { H, if ω ω s (jω =, otherwise. We then sample the filtered signal at the rate of ω s /(π samples per second, using an ideal impulsetrain sampler with sampling period T = π/ω s (i.e. multiplication by a periodic train of ideal continuous-time unit impulses with period T. From the continuous-time impulse train obtained through sampling, we reconstruct a signal y(t using another ideal analog low-pass filter, H: { T, if ω ω s H(jω =, otherwise. Find the reconstructed signal y(t for the following values of ω s /(π. (i.5 samples per second (i.e., one sample is taken every two seconds. (ii.5 samples per second. (iii.5 samples per second. (iv.5 samples per second. For Parts (i, (ii, (iii, give expressions for y(t without using MATLAB. For Part (iv, use MATLAB to plot y(t. Comment on your results.

x (t x'( t p (t x p (t y(t H '(f H(f Figure : Block diagram of the system in Problem. Solution. The system diagram is shown in Fig.. Let us first consider the effect of the last two stages, sampling and low-pass filter H. Suppose X, X p, and Y are the CTFT of x, x p, and y, respectively. As derived in class, we have: X p (jω = Y (jω = H(jωX p (jω = Because of the first low-pass filter H, we have: X (jω = k= T X (j(ω ω s { TXp (ω if ω ωs if ω > ωs for ω > ω s. Thus, there is no aliasing in X p. So ( can be rewritten as follows: { T Y (jω = T X (jω if ω ωs if ω > ωs = X (jω. ( This means that y(t = x (t. The last two stages together did nothing to the signal and can be omitted. Therefore, we only need to consider the effect of the first filter, to decide which frequency components of x it keeps and which ones it discards. To do that, let us compute the Fourier series expansion of x: x(t = where T = is the fundamental period of x. a k = = = = { j { j k= πk j x(te t dt x(te jkπt dt = kπ e jkπt k k = ( a k e j πk T t, (3 kπ [( k ] k k = e jkπt dt The same result can be obtained by using the time-shifting property of the CT Fourier transform, in conjunction with the result of Example from the notes on CT Fourier transform. This is because x is related to the signal s from that example as follows: x(t = s ( t + t,

..8.6.4.. 4 4 6 8 Figure : Matlab plot for Problem (iv. with A =, t =, and T =. Notice that a k = a k, to get the following from Eq. (3: x(t = + (a k e jkπt + a k e jkπt k= = + a k (e jkπt e jkπt k= = + a k j sin(kπt k= = + k= kπ [ ( k ]sin(kπt = + π sin(πt + 3π sin(3πt + sin(5πt +... (4 5π After the low-pass filter H, only those terms with angular frequency no greater than ω s / in (4 are preserved in x (t. Now let us see the results for different ω s. (i If ωs = π, only the first term in (4 gets through H : x (t =. (ii If ωs =.5π, only the first terms in (4 get through H : x (t = + π sin(πt. (iii If ωs =.5π, only the first terms in (4 get through H : x (t = + π sin(πt. 3

(iv If ωs =.5π, only the first 6 terms in (4 get through H : x (t = + π sin(πt + 3π sin(3πt + 5π sin(5πt + 7π sin(7πt + 9π sin(9πt. A plot of y(t is shown in Figure. Problem. Sampling theorem. Consider sampling the signal below (which is defined for all time, < t <. x c (t = sin(6πt πt cos(3πt (a x c (t is sampled with sampling period T = /4 seconds, to produce the discrete-time signal x d [n] = x c (nt. Plot the magnitude of the DTFT of x d, X d (e jω, over the interval π < Ω < π. Show as much detail as possible. (b What is the Nyquist sampling rate for this signal? Solution. Multiplication in the time domain corresponds to convolution in the frequency domain. Breaking up x c (t into x (t = sin(6πt πt and x (t = cos(3πt, we have: { if ω < 6π, X (jω = otherwise, X c (jω = π X X (jω = X (jω = πδ(ω + 3π + πδ(ω 3π, if ω < 3π, / if 3π ω < 9π, otherwise, see the top three plots of Fig. 3. To get X d (e jω from X c (jω, we can use the following formulas from the notes and the text: X p (jω = ( X c (j ω πk (5 T T k= ( X d (e jω = X p j Ω = ( j(ω πk X c. (6 T T T k= Note that the highest frequency in X c (jω is 9π, which means that the Nyquist sampling rate for this signal is 9 samples per second. Since we are sampling at f s = 4 Hz > 9 Hz, there will be no aliasing, i.e., the shifted replicas of X c (jω/t in the formula (5 will not overlap. Thus, in order to get X p (jω we simply put replicas of X c (jω/t at ω =, ±ω s, ±ω s,... Then, to get X(e jω from X p (jω according to the formula (6, we rescale the frequency axis by a factor T = /4, as shown in the bottom plot of Fig. 3. Problem 3. Interpolation. Consider the following continuous-time signal illustrated in Fig. 4: {, if n.5 t < n +.5, for any integer n, x(t =, if n.5 t < n.5, for any integer n. 4

X (jω.5 6kpi 6kpi 6kpi 6kpi X (jω.5 6kpi 3kpi 3kpi 6kpi X c (jω.5 6kpi 9kpi 3kpi 3kpi 9kpi 6kpi 4 X(e jω pi 3pi/8 pi/8 pi/8 3pi/8 pi Figure 3: Spectra for Problem. 5

x(t t Figure 4: Signal x(t for Problem 3. Suppose this signal is ideally sampled (with no prefiltering at the rate of Hz, to obtain a discretetime signal x d [n] = x(n. The latter is then filtered with a nonlinear discrete-time filter specified by the following input-output relationship: y d [n] = (x d [n], where y d is the response of the filter to the input x d. (a Find and plot x d [n]. (Hint. Do not do this in the frequency domain. Just recall the time-domain definition of ideal sampling and write down x d directly. { if n is even Solution. As shown in Fig. 5, x d [n] = x(n = if n is odd (b Find and plot y d [n]. x d [n] = y d [n] n Figure 5: Signals x d = y d for Problem 3. Solution. Since x(t is either zero or one for all t, we have that x d [n] is either zero or one for all n. But = and =, so y d [n] = x d [n] for this particular signal. (c Suppose we subtract.5 from y d to obtain another discrete-time signal y: y[n] = y d [n].5. y[n] y u [n] LPF y int [n] Figure 6: Interpolator for Problem 3(c. 6

Suppose further y int is the result of interpolating y using the scheme illustrated in Fig. 6: first, y is upsampled by a factor of to result in y u ; and then y u is filtered with an ideal discretetime low-pass filter. For this problem, we assume the following frequency response for the ideal low-pass filter: { H(e jω, ω π = 3 π, 3 < ω π Find y int [n]. Solution. Note that y d can be represented as follows: y d [n] =.5 +.5cos(πn, and therefore y[n] =.5cos(πn. Since y u is the result of inserting a zero after each sample of y, we have:.5 n =..., 4,,4,8,... y[n] =.5 n =..., 6,,,6,... =.5cos(πn/. if n is odd Since all frequencies below π/3 are preserved, this sinusoid of frequency π/ will be preserved exactly: y int [n] =.5cos(πn/. Problem 4. Decimation. The system depicted below is a (discrete-time decimator. Assume that the frequency response of the discrete-time low-pass filter is: { H(e jω, ω π = 3 π, 3 < ω π x[n] LPF y[n] Find the output y for each of the following inputs: (a x[n] =, for < n < ( + j n (b x[n] =, for < n < (c (d (e x[n] = j n, for < n < ( π x[n] = cos 8 n, for < n < ( π x[n] = cos n, for < n < (Hint. The inputs in (a-c are everlasting complex exponential signals, and the two remaining inputs can be represented as sums of complex exponentials. Solution. Let x denote the output signal of the low-pass filter. We know that for any complex exponential signal x[n] = e jω n, x [n] = e jω n H(e jω. Therefore, if the angular frequency ω π 3, then x [n] = x[n]; otherwise x [n] =. 7

(a (b x[n] = = e jn, < n < The angular frequency ω = π 3, so x [n] = x[n] =, and hence y[n] = x [n] =. x[n] = ( + j n = e j π 4 n, < n < The angular frequency ω = π/4 π 3, so x [n] = x[n], and the output is y[n] = x [n] = x[n] = e j π n = j n. (c x[n] = j n = e j π n, < n < The angular frequency ω = π > π 3, so x (n =, therefore y[n] =. (d (e ( π x[n] = cos 8 n = ej π 8 n + e j π 8 n, < n < Note that x is a linear combination of two complex exponentials with angular frequencies ω = π 8 and ω = π 8. Using the fact that ω = ω = π 8 π 3 and linearity of LPF, we get x [n] = x[n]. Therefore, the output is y[n] = x [n] = x[n] = cos ( π 4 n. ( π x[n] = cos n = ej π n + e j π n, < n < Again, x is a linear combination of two complex exponentials. Their angular frequencies are ω = π and ω = π. In this case, we have π > ω = ω = π > π 3. Therefore, x [n] = and y[n] =. 8