You are on page 1of 9



ECE 3084
Summer 2014
Problem Set #3
Assigned: 29-May-14
Due Date: 5-June-14
Your homework is due at the start of class on Thursday, June 5.
A sheet of Fourier transform pairs and properties has been posted to the Resources
section of the T-square website. You will want to print this out and have it handy while you are
working on the homework.
You may turn in your homework up to one day late, by 3:00 PM the following day. A
30% penalty will be assessed on late homeworks (even homeworks turned in on the day it
is due but not at the start of class, although the penalty might be slightly less at our discretion).
We understand that sometimes multiple assignments hit at once, or other life events intervene, and
hence you have to make some tough choices. Wed rather let you turn something in late, with some
points off, than have a no late assignments accepted at all policy, since the former encourages
you to still do the assignment and learn something from it, while the latter just grinds down your
soul. The somewhat aggressive late penalty is not intended to be harsh its intended to encourage
you to get things in relatively on time (or just punt if you have to and not leave it hanging over
you all semester) so that you can move on to assignments for your other classes. Also, there is the
practical matter that we cannot accept homeworks after solutions are posted, and we would like to
post solutions shortly after both sections have submitted their homework.
Please refrain from looking at backfiles of homework and exam solutions i.e.,
word in Georgia Tech parlance from previous versions of ECE2025, ECE2026, or
ECE3084, beyond your own materials assembled while taking those classes and any
old material we explicitly provide to you. If you get stuck, please come get help from your
professor or TA before consulting a backfile.


Suppose a system has an impulse response given by
h(t) = [u(t + 100) u(t 100)].
(a) Find the frequency response (or, equivalently, the Fourier transform) H(j). In addition to
giving an equation, sketch a plot of the frequency response for 0.05 < < 0.05. Be
sure to clearly label where the zeros of the sinc function are, along with the height of
the main lobe. The heights of the sidelobes need not be super-accurate; you just need to
indicate the overall trend. You may use a calculator or computer to check your work, but be
sure that you can do this sort of problem using just your brain.
(b) If the input to this system is
x(t) = 3 + cos(0.01t) +

cos(0.02t) + cos(0.03t) +

find the output y(t). Explain your reasoning by referencing the graph you made in part
(a). Be cleverthis should be easy!
(c) If the input to this system is
x(t) = cos(0.0025t) + 800(t 50),
find the output y(t). (Remember, one of the most important concepts in ECE2026 and
ECE3084 and, in fact, most of electrical engineering is to know when to use the time
domain way of thinking and when to use the frequency domain way of thinking. You may
often find yourself needing to use different domains on different parts of a problem.)

Figure 1: CLTIdemo interface for continuous-time frequency response.

The classes ECE2026 and ECE3084 have their roots in an older class called ECE2025. We developed
several cool Matlab GUIs for ECE2025 that cover various continuous-time signals and systems
topics. In ECE2025, these GUIs were explored during dedicated lab periods staffed by professors
and graduate TAs. Unfortunately, we dont have a weekly lab time devoted to ECE3084, and the
few class periods that we have dedicated as lab time are reserved for labs involving real hardware
beyond just a compute. So, we will instead have you explore these GUIs on your own as homework
problems. If you have Matlab installed on your own computer, you can download the GUIs by
downloading and installing the SP-First Toolbox by following the instructions at:
You will also find it installed on all the ECE instructional lab computers.
This homework will concentrate on the use two Matlab GUIs:
CLTIdemo: In this demo, you can select an input signal that is a sinusoid, and see the change
created by the frequency response. This demo reinforces the concept that sinusoid in gives
sinusoid out. Figure 1 shows the interface for the CLTIdemo. We know that if the input to
an LTI continuous-time system is a sinusoid of the form
x(t) = A + B cos(1 t + ),

< t < ,


then the corresponding output is also a sinusoid:

y(t) = AH(j0) + B|H(j1 )| cos (1 t + + H(j1 )) ,

< t < ,


where H(j) is the frequency response of the continuous-time LTI system. The CLTIdemo
GUI illustrates this for a variety of simple analog filters.

FseriesDemo: This GUI addresses continuous-time Fourier Series defined by

x(t) =

ak ejk0 t ,


where x(t) is a periodic signal and 0 is the fundamental frequency, 0 = 2/T .

For instance, Figure 2 shows the spectrum for a 50% duty cycle square wave. Notice that the
GUI will also show the resynthesized signal for a finite number of coefficients.

Figure 2: FSeriesDemo interface for Fourier Series synthesis.

In this problem, well explore a lowpass filter with the following frequency response:
H(j) =

40 + j


The CLTIdemo GUI can implement the LPF defined by Equation 3 if you choose the filter named
First-Order Lowpass.
(a) Use the CLTIdemo GUI to create a first-order lowpass filter by selecting First-Order Lowpass
from the menu and setting the cutoff frequency to 20 Hz.1
(b) Set the input signal to
x(t) = 1.0 + cos(40t)
Look at the output and compare its amplitudes and phases to the input amplitudes and
phases. Click the box labeled Theoretical Answer and record the result.

The CLTIdemo GUI will convert the frequencies from hertz to rad/s.

(c) Now change the input signal to x(t) = cos(80t), and record the numerical values of the
output signals amplitude and phase. Repeat for x(t) = cos(120t), again recording the
amplitude and phase of the output signal.
The results from this part and the previous part will go in the H(jk0 ) column on the
Answer sheet for Problem 3.2 page.
(d) Now consider the case where the input signal x(t) is a square wave with a period of 1/20
secs. Although the FseriesDemo GUI in cannot handle this value for the period, it can still
be used to make a plot the spectrum of this type of signal. The spectrum values plotted are
the magnitudes of {ak } for this square wave which are2
ak =

1 ejk

Use the {ak } coefficients and the frequency response of the filter to write the first few terms
of the output signal y(t) as a sum of cosines:
y(t) =

bk ejk0 t



= B0 + B1 cos(0 t + 1 ) + B2 cos(20 t + 2 ) + B3 cos(30 t + 3 ) + . . . ,


where bk = H(jk0 )ak . Use the values of the frequency response from the CLTIdemo GUI
and the {ak } coefficients above to determine the numerical values for 0 , b0 , b1 , b2 , b3 , B0 ,
B1 , B2 , B3 , and 1 , 2 , and 3 .
The calculation above amounts to an analysis of how you can filter the periodic input signal
(this square wave seen in Figure 2) through a continuous-time LTI system whose frequency response
is given in Equation 3. Since this is an analog system, we cannot do the actual filtering in Matlab;
instead, we can only calculate what the output signal would be by finding the Fourier series of the
Write your answers to this problem on the Answer sheet for Problem 3.2 table.
It is the last page of this PDF; you should print it out so you can handwrite your answers on it.
Include it in your homework submission in an appropriate spot.

Notice that this isnt quite the same as the symmetric square wave we looked at in class; its a time-shifted,
bipolar version. Section 3-6.1 of the ECE2026 textbook Signal Processing First has the derivation of the Fourier
Series coefficients for this square wave.

During their investigations into the paranormal, Agent Fox Mulder and Agent Dana Scully of the
FBI discover a continuous-time LTI system that they believe is of extraterrestrial origin. They
know that it has a one-sided decaying exponential impulse respone given by h(t) = exp(at)u(t);
the mystery lies in exactly what a is.
Undaunted, the agents take the alien LTI system to the FBI lab. We must find out what a is,
Mulder drones. Put an impulse (t) into the system so we can find out what the impulse response
The lab tech shakes his head in shame. Sorry, sir, he apologizes. Due to sequestrationrelated budget cuts, the only working piece of signal generation equipment we have is a sinusoidal
generator, and even that has a broken knobs, so its stuck generating a cosine wave with phase 0,
an amplitude of 200, and a frequency of 40 radians.
Are you sure thats 40 radians? Why not 40 radians? Scully asks.
A voice from the sky, which sounds remarkably like Aaron, says Yes, its 40 radians. Im
making it exactly 40 radians so the math in part (a) works out nicely.
The lab tech attaches his rather limited sinusoidal generator to the input of the mysterious alien
system, and exclaims, Ah ha! When given this input, the output is a cosine wave with amplitude
4, a frequency of 40 radians, and a phase of...
Ssssh! the voice from the sky says. Dont give away the phase! Thats part (b) of the
(a) Help Mulder and Scully by finding the value of a using the information given in the
(b) Find the phase of the output cosine the tech almost gave away. (I generally try to set
up my homework and quiz problems so that you can do them without a calculator, but for
this one, youll need to go ahead and pull out your calculator.)

The parts of this problem are independent of each other.
(a) In class on Thursday, May 29, we derived a derivative property for Fourier transforms, which
stated that if X(j) is the Fourier transform of x(t), then jX(j) is the Fourier transform
of dt
x(t). We proved this by writing down the inverse Fourier transform formula and taking
the derivative with respect to t of both sides.
Lets try proving this with another approach. Remember from your Freshman calculus class
that a derivative could be defined as
x(t) x(t )
x(t) = lim

where the + superscript indicates that  approaches 0 from above. (We will assume this
limit exists.)
Take the Fourier transform of [x(t) x(t )]/ using the time-shift property of Fourier
transforms, and then replace the exponential in the resulting expression with the first
two terms of its Taylor series expansion, i.e., exp(a) 1 + a. This is legitimate since
we are letting  go to zero.
(b) The derivative shown in part (a) was, technically speaking, a left derivative. Repeat the
procedure in part (a), except this time apply it to the right derivative, given by
x(t + ) x(t)
x(t) = lim

(c) In class on Thursday, May 29, we derived the Fourier transforms of cos(0 ) and sin(0 )
by rewriting both of them with their inverse Eulers formulas and applying the Fourier
transform pair
exp(jt) 2( 0 ).
Suppose we had derived the Fourier transform for cos(0 t) but didnt know the inverse
Eulers formula for sin(0 t). Derive the Fourier transform of sin(0 ) by applying the timederivative property
x(t) jX(j)
to the Fourier transform pair

cos(0 t) ( 0 ) + ( + 0 )
and remembering what the derivative of cos(0 t) is from your calculus classes.
(d) While searching the web looking for weird Fourier transform pairs, I found one that said the
Fourier transform of cos(t2 ) is


Assuming the website is correct, find the Fourier transform of cos(at2 ), where a > 0, using
the time-scaling property (also known as the time compression and expansion property)

of Fourier transforms. (Hint: at2 = ( at)2 .

(e) Lets derive a new Fourier transform property. Show that if we denote the Fourier transform
of x(t) by X(j), then the inverse Fourier transform of dX(j)
is jtx(t). This derivative-ind
frequency property is the dual of the time-derivative property we proved in class; prove it
in a manner similar in spirit to the proof we explored in class.
(f) Use Fourier transform theory to perform the integral
Z jc5
3 + jc
Simplify your answer as much as possible. This integral would be extremely difficult to solve
using standard brute-force calculus, but it may readily yield to your way-cool ninja ECE3084
knowledge. You will need to massage the expression a bit to put it into a familiar form. (Hint:
the ej5 bit is not a shift here! Think to yourself: oh, this integral is just a such-and-such
evaluated at thingamajig = something.)

Consider the Fourier transform pair

exp 2 .
Usually, the Fourier transform of a function looks nothing like the original function. For instance, rectangular windows transform into since functions and vice-versa. The above pair is
particularly interesting since both the time-domain and frequency-domain versions of this signal
have the same overall bell-like shape.3
(a) Suppose
x(t) =
and h(t) =
(i) Compute the convolution y(t) = x(t) h(t). Theoretically, you could find the result
by directly cranking on the convolution integral:


[t ]2

y(t) =

However, that way lies madness. There is an easier path!

(ii) When you convolve two similar functions, you usually dont get a function that looks
anything like the functions youre convolving. For instance, if you convolve two rectangular pulses, you get a triangle or a trapezoid. However, there are a few families of
functions for instance, the sinc function that give you a member of the same family
if you convolve two functions from that family. Such families of functions are relatively
rare and usually have quite interesting properties in many applications.
Looking at your answer to part (i), does it seem very different in form than x(t) and
h(t)? Or does it look similar?
(b) Find the inverse Fourier transform of X(j) = exp(9 2 + 7j).

You may have seen this kind of shape before in your probability and statistics class under the name Gaussian
probability density, with a corresponding to something called a variance. Here we are just treating this shape as
a time-domain function; you need not worry about any concepts from probability and statistics to do this problem.

Answer sheet for Problem 3.2

Find the {bk } Fourier series coefficients of the output signal. List the values of 0 , ak , H(jk0 ),
bk , Bk and k in the table below. Note: Bk 6= bk .

0 =

H(jk0 )

k = 0 a0 =

b0 =

B0 =

k = 1 a1 =

b1 =

B1 =

1 =

k = 2 a2 =

b2 =

B2 =

2 =

k = 3 a3 =

b3 =

B3 =

3 =