77

Chapter 4

Continuous-Time Fourier Series

4.1 Introduction
One very important tool in the study of signals and systems is the Fourier series. A very large class of signals can be
represented using Fourier series, namely most practically useful periodic signals. The Fourier series represents a signal
as a linear combination of complex sinusoids. This is often desirable since complex sinusoids are easy functions with
which to work. For example, complex sinusoids are easy to integrate and differentiate. Also, complex sinusoids have
important properties in relation to LTI systems. In particular, complex sinusoids are eigenfunctions of LTI systems.
Therefore, the response of a LTI system to a complex sinusoid is the same complex sinusoid multiplied by a complex
scaling factor.

4.2 Definition of Continuous-Time Fourier Series

Suppose that we have a set of harmonically-related complex sinusoids of the form

φk (t) = e jkω0 t = e jk(2π /T )t k = 0, ±1, ±2, . . ..

The fundamental frequency of the kth complex sinusoid φk (t) is kω0 , an integer multiple of ω0 . Since the fundamental
frequency of each of the harmonically-related complex sinusoids is an integer multiple of ω0 , a linear combination
of these complex sinusoids must be periodic. More specifically, a linear combination of these complex sinusoids is
periodic with period T = 2π /ω0 .
Suppose that we can represent a periodic complex signal x(t) as a linear combination of harmonically-related
complex sinusoids:
∞
x(t) = ∑ ck e jkω0 t . (4.1)
k=−∞
Such a representation is known as a Fourier series. More specifically, this is the complex exponential form of the
Fourier series. The terms in the summation for k = 1 and k = −1 are known as the fundamental frequency components
or first harmonic components, and have the fundamental frequency ω0 . More generally, the terms in the summation
for k = K and k = −K are called the Kth harmonic components, and have the fundamental frequency K ω0 . Since the
complex sinusoids are harmonically related, the signal x(t) is periodic with period T = 2π /ω0 (and frequency ω0 ).
Since we often work with Fourier series, it is sometimes convenient to have an abbreviated notation to indicate
that a signal is associated with particular Fourier series coefficients. If a signal x(t) has the Fourier series coefficient
sequence ck , we sometimes indicate this using the notation
FS
x(t) ←→ ck .

Consider the Fourier series representation of the periodic signal x(t) given by (4.1). In the most general case,
x(t) is a complex signal, but let us now suppose that x(t) is a real signal. In the case of real signals, an important

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

78 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

relationship exists between the Fourier series coefficients ck and c−k . To show this, we proceed as follows. Suppose
that we can represent x(t) in the form of (4.1). So, we have
∞
x(t) = ∑ ck e jkω0 t . (4.2)
k=−∞

Taking the complex conjugate of both sides of the preceding equation, we obtain
!∗
∞
∗ jkω0 t
x (t) = ∑ ck e
k=−∞
∞  ∗
= ∑ ck e jkω0 t
k=−∞
∞
= ∑ c∗k e− jkω0 t . (4.3)
k=−∞

Since x(t) is real, we know that x(t) = x∗ (t), and we can rewrite (4.3) as
∞
x(t) = ∑ c∗k e− jkω0 t .
k=−∞

Replacing k by −k in the summation, we obtain

∞
x(t) = ∑ c∗−k e jkω0 t . (4.4)
k=−∞

Comparing (4.2) and (4.4), we can see that

ck = c∗−k . (4.5)

Consequently, if x(t) is a real signal, we have that ck and c−k are complex conjugates of each other.
Using the relationship in (4.5), we can derive two alternative forms of the Fourier series for the case of real signals.
We begin by rewriting (4.1) in a slightly different form. In particular, we rearrange the summation to obtain
∞ h i
x(t) = c0 + ∑ ck e jkω0 t + c−k e− jkω0 t .
k=1

Substituting ck = c∗−k from (4.5), we obtain

∞ h i
x(t) = c0 + ∑ ck e jkω0 t + c∗k e− jkω0 t .
k=1

Now, we observe that the two terms inside the summation are complex conjugates of each other. So, we can rewrite
the equation as
∞
x(t) = c0 + ∑ 2 Re{ck e jkω0 t }. (4.6)
k=1

Let us now rewrite ck in polar form as

ck = |ck | e jθk ,

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.3. DETERMINING THE FOURIER SERIES REPRESENTATION OF A CONTINUOUS-TIME
PERIODIC SIGNAL 79

where θk is real (i.e., θk = arg ck ). Substituting this expression for ck into (4.6) yields
∞ n o
x(t) = c0 + ∑ 2 Re |ck | e j(kω0 t+θk )
k=1
∞
= c0 + ∑ 2 Re {|ck | [cos(kω0t + θk ) + j sin(kω0t + θk )]}
k=1
∞
= c0 + ∑ 2 Re {|ck | cos(kω0t + θk ) + j |ck | sin(kω0t + θk )} .
k=1

Finally, further simplification yields

∞
x(t) = c0 + 2 ∑ |ck | cos(kω0t + θk )
k=1

(where θk = arg ck ). This is known as the combined trigonometric form of a Fourier series.
A second alternative form of the Fourier series can be obtained by expressing ck in Cartesian form as

ck = 12 (ak − jbk ).

where ak and bk are real. Substituting this expression for ck into (4.6) from earlier yields
∞
x(t) = c0 + ∑ 2 Re{( 12 (ak − jbk )e jkω0 t }
k=1
∞
= c0 + ∑ Re{(ak − jbk ) (cos kω0t + j sin kω0t)}
k=1
∞
= c0 + ∑ Re {ak cos kω0t + jak sin kω0t − jbk cos kω0t + bk sin kω0t} .
k=1

Further simplification yields

∞
x(t) = c0 + ∑ [ak cos kω0t + bk sin kω0t]
k=1
(where ak = Re 2ck and bk = − Im 2ck ). This is known as the trigonometric form of a Fourier series.
By comparing the various forms of the Fourier series introduced above, we can see that the quantities ck , ak , bk ,
and θk are related as follows:

2ck = ak − jbk and ck = |ck | e jθk .

(Recall that ak , bk , and θk are real and ck is complex.)

4.3 Determining the Fourier Series Representation of a Continuous-Time

Periodic Signal
Given an arbitrary periodic signal x(t), we need some means for finding its corresponding Fourier series representa-
tion. In other words, we need a method for calculating the Fourier series coefficients ck . In what follows, we derive
a formula for the calculation of the ck . We begin with the definition of the Fourier series in (4.1). Multiplying both
sides of this equation by e− jnω0 t yields
∞
x(t)e− jnω0 t = ∑ ck e jkω0 t e− jnω0 t
k=−∞
∞
= ∑ ck e j(k−n)ω0t .
k=−∞

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

80 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

R
As a matter of notation, we use T to denote the integral over an arbitrary interval of length T (i.e., the interval
(t0 ,t0 + T ) for arbitrary t0 ). Integrating both sides of this equation over one period T of x(t), we obtain
Z Z ∞
− jnω0 t
x(t)e dt = ∑ ck e j(k−n)ω0t dt.
T T k=−∞

Reversing the order of integration and summation yields

Z ∞ Z 
x(t)e− jnω0 t dt = ∑ ck e j(k−n)ω0t dt . (4.7)
T k=−∞ T

Consider now the integral on the right-hand side of this equation. In order to evaluate this integral, we employ Euler’s
formula to write
Z Z
e j(k−n)ω0t dt = [cos ((k − n)ω0t) + j sin ((k − n)ω0t)] dt
T
ZT Z
= cos ((k − n)ω0t) dt + j sin ((k − n)ω0t) dt. (4.8)
T T

For k 6= n, cos(k − n)ω0t and sin(k − n)ω0t are both sinusoids of frequency (k − n)ω0 and period T / |k − n|. Since
we are integrating over an interval of length T , both sinusoids will be integrated over an integral number of periods,
resulting in a value of zero. For k = n, the integrand on the left-hand side of (4.8) is simply e j0 = 1, and the result of
integration is T . To summarize, we have the following:
Z
(
j(k−n)ω0 t T for k = n
e dt = (4.9)
T 0 otherwise.

Substituting (4.9) into (4.7), we obtain Z

x(t)e− jnω0 t dt = cn T. (4.10)
T
Rearranging, we obtain the following expression for cn :
Z
1
cn = x(t)e− jnω0 t dt.
T T

Thus, we can calculate the Fourier series coefficient sequence ck for an arbitrary periodic signal x(t) using the formula:
Z
1
ck = x(t)e− jkω0 t dt. (4.11)
T T

As a matter of terminology, we refer to (4.11) as the Fourier series analysis equation and (4.1) as the Fourier series
synthesis equation.
Suppose that we have a complex periodic function x(t) with period T and Fourier series coefficient sequence ck .
One can easily show that the coefficient c0 is the average value of x(t) over a single period T . The proof is trivial.
Consider the Fourier series analysis equation given by (4.11). Substituting k = 0 into this equation, we obtain
 Z 
1
c0 = x(t)e− jkω0 t dt
T T k=0
Z
1 0
= x(t)e dt
T T
Z
1
= x(t)dt.
T T
Thus, c0 is simply the average value of x(t) over a single period.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.3. DETERMINING THE FOURIER SERIES REPRESENTATION OF A CONTINUOUS-TIME
PERIODIC SIGNAL 81

Example 4.1 (Fourier series of a periodic square wave). Find the Fourier series representation of the periodic square
wave x(t) shown in Figure 4.1.
Solution. Let us consider the single period of x(t) for 0 ≤ t < T . For this range of t, we have
(
A for 0 ≤ t < T2
x(t) =
−A for T2 ≤ t < T .

We use the Fourier series analysis equation (4.11) to write (and subsequently assume that k 6= 0):
Z
1
ck = x(t)e− jkω0 t dt
T T
Z T /2 Z T 
1
= Ae− jkω0 t dt + (−A)e− jkω0 t dt
T 0 T /2
  T /2   !
1 −A − jkω0 t A − jkω0 t T
= e + e
T jkω0 0 jkω0 T /2
h i h i 
−A T /2 T
= e− jkω0 t − e− jkω0 t
j2π k 0 T /2
jA h i h i
= e− jπ k − 1 − e− j2π k − e− jπ k
2π k
jA h − jπ k i
= 2e − e− j2π k − 1
2π k
jA h − jπ k i
= 2(e ) − (e− j2π )k − 1 .
2π k
Now, we observe that e− jπ = −1 and e− j2π = 1. So, we have
jA
ck = [2(−1)k − 1k − 1]
2π k
jA
= [2(−1)k − 2]
2π k
jA
= [(−1)k − 1]
π
 k
 − j2A
for k odd
= πk
0 for k even, k 6= 0.

Now, we consider the case of c0 . We have

Z
1
c0 = T x(t)dt
T
Z T /2 Z T 
1
= T Adt − Adt
0 T /2

1 AT AT

=T 2 − 2
= 0.

Thus, the Fourier series of x(t) is given by (4.1) where the coefficient sequence ck is

 − j2A
for k odd
ck = πk
0 for k even.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

82 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

x(t)
A

t
−T − T2 T
2 T

−A

Figure 4.1: Periodic square wave.

x(t)
A A A A A A A

··· ···
t
−3T −2T −T 0 T 2T 3T

Figure 4.2: Periodic impulse train.

Example 4.2 (Fourier series of a periodic impulse train). Suppose that we have the periodic impulse train x(t) shown
in Figure 4.2. Find the Fourier series representation of x(t).

Solution. Let us consider the single period of x(t) for − T2 ≤ t < T2 . We use the Fourier series analysis equation (4.11)
to write:
Z
1
ck = x(t)e− jkω0 t dt
T T
Z
1 T /2
= Aδ (t)e− jkω0 t dt
T −T /2
Z T /2
A
= δ (t)e− jkω0 t dt.
T −T /2

Using the sifting property of the unit-impulse function, we can simplify the above result to obtain

A
ck = .
T

Thus, the Fourier series for x(t) is given by (4.1) where the coefficient sequence ck is ck = TA .

Example 4.3. Consider the periodic function x(t) with fundamental period T = 3 as shown in Figure 4.3. Find the
Fourier series representation of x(t).

2π
Solution. The signal x(t) has the fundamental frequency ω0 = 2π /T = 3 . Let us consider the single period of x(t)

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.3. DETERMINING THE FOURIER SERIES REPRESENTATION OF A CONTINUOUS-TIME
PERIODIC SIGNAL 83

x(t)

1 1 1
··· ···
t
−4 −3 −2 −1 1 2 3 4

−1 −1 −1

Figure 4.3: Periodic impulse train.

for − T2 ≤ t < T
2 (i.e., − 23 ≤ t < 32 ). Computing the Fourier series coefficients ck , we have
Z
ck = 1
T x(t)e− jkω0 t dt
T
Z 3/2
= 1
T x(t)e− jkω0 t dt
−3/2
Z 0 Z 3/2 
= 1
T −δ (t + 1)e− jkω0 t dt + δ (t − 1)e− jkω0t dt
−3/2 0
h i
= T1 −e− jkω0 (−1) + e− jkω0 (1)
h i
= T1 e− jkω0 − e jkω0
1
= T [2 j sin(−kω0 )]
= 2Tj sin(−kω0 )
= − 2Tj sin kω0
= − 23j sin 2π3 k .
Thus, x(t) has the Fourier series representation
∞
x(t) = ∑ ck e jkω0 t
k=−∞
∞
= ∑ − 23j (sin 2π3 k )e j2π kt/3 .
k=−∞

Example 4.4 (Fourier series of an even real function). Let x(t) be an arbitrary periodic real function that
R
is even. Let
ck denote the Fourier series coefficient sequence for x(t). Show that Im{ck } = 0, ck = c−k and c0 = T1 0T x(t)dt.
Proof. From the Fourier series analysis equation (4.11) and using Euler’s relation, we can write
Z
1
x(t)e− jkω0 t dt
ck =
T T
Z
1
= (x(t) [cos(−kω0t) + j sin(−kω0t)]) dt.
T T
Since cos and sin are even and odd functions, respectively, we can rewrite the above equation as
Z
1
ck = [x(t) (cos kω0t − j sin kω0t)] dt
T T
Z Z 
1
= x(t) cos kω0tdt − j x(t) sin kω0tdt .
T T T

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

84 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

Consider the first integral above (i.e., the one involving the cos function). Since x(t) is even and cos kω0t is even, we
R R T /2
have that x(t) cos kω0t is even. Thus, T x(t) cos kω0tdt = 2 0 x(t) cos kω0tdt. Consider the second integral above
(i.e., the one involving the sin function). Since x(t) is even and sin kω0t is odd, we have that x(t) sin kω0t is odd. If we
integrate an odd periodic function over one period (or an integer multiple thereof), the result is zero. Therefore, the
second integral is zero. Combining these results, we can write
 Z T /2 
1
ck = 2 x(t) cos kω0tdt
T 0
Z T /2
2
= x(t) cos kω0tdt. (4.12)
T 0

Since x(t) is real, the quantity ck must also be real. Thus, we have that Im{ck } = 0.
Consider now the expression for c−k . We substitute −k for k in (4.12) to obtain
Z T /2
2
c−k = x(t) cos(−kω0t)dt.
T 0

Since cos is an even function, we can simplify this expression to obtain

Z T /2
2
c−k = x(t) cos(kω0t)dt
T 0
= ck .

Thus, ck = c−k .
Consider now the quantity c0 . Substituting k = 0 into (4.12), we can write
Z T /2
2
c0 = x(t) cos(0)dt
T 0
Z T /2
2
= x(t)dt
T
0 Z T 
2 1
= x(t)dt
T 2 0
Z T
1
= x(t)dt.
T 0
1 RT
Thus, we have shown that c0 = T 0 x(t)dt. Therefore, in summary, we have shown that Im{ck } = 0, ck = c−k , and
R
c0 = T1 0T x(t)dt.
Example 4.5 (Fourier series of an odd real function). Let x(t) be a periodic real function that is odd. Let ck denote
the Fourier series coefficient sequence for x(t). Show that Re{ck } = 0, ck = −c−k , and c0 = 0.
Proof. From the Fourier series analysis equation (4.11) and Euler’s formula, we can write
Z
1
ck = x(t)e− jkω0 t dt
T T
Z 
1
= x(t) [cos(−kω0t) + j sin(−kω0t)] dt .
T T
Since cos and sin are even and odd functions, respectively, we can rewrite the above equation as
Z 
1
ck = x(t) [cos kω0t − j sin kω0t] dt
T T
Z Z 
1
= x(t) cos kω0tdt − j x(t) sin kω0tdt .
T T T

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.4. CONVERGENCE OF CONTINUOUS-TIME FOURIER SERIES 85

Consider the first integral above (i.e., the one involving the cos function). Since x(t) is odd and cos kω0t is even,
we have that x(t) cos kω0t is odd. If we integrate an odd periodic function over a single period (or an integer mul-
tiple thereof), the result is zero. Therefore, the first integral is zero. Consider the second integral above (i.e., the
one involving the sin function). Since x(t) is odd and sin kω0t is odd, we have that x(t) sin kω0t is even. Thus,
R R T /2
T x(t) sin kω0 tdt = 2 0 x(t) sin kω0tdt. Combining these results, we can write
Z
−j
ck = x(t) sin kω0tdt
T T
Z T /2
− j2
= x(t) sin kω0tdt. (4.13)
T 0
Since x(t) is real, the result of the integration is real, and consequently ck is purely imaginary. Thus, Re{ck } = 0.
Consider the quantity c−k . Substituting −k for k in (4.13), we obtain
Z
− j2 T /2
c−k = x(t) sin(−kω0t)dt
T 0
Z T /2
− j2
= x(t)[− sin(kω0t)]dt
T 0
Z T /2
j2
= x(t) sin kω0tdt
T 0
= −ck .

Thus, ck = −c−k .
Consider now the quantity c0 . Substituting k = 0 in the expression (4.13), we have
Z T /2
− j2
c0 = x(t) sin(0)dt
T 0
= 0.

Thus, c0 = 0. Therefore, in summary, we have shown that Re{ck } = 0, ck = −c−k , and c0 = 0.

4.4 Convergence of Continuous-Time Fourier Series

So far we have assumed that a given periodic signal x(t) can be represented by a Fourier series. Since a Fourier series
consists of an infinite number of terms, we need to more carefully consider the issue of convergence. That is, we want
to know under what circumstances the Fourier series of x(t) converges (in some sense) to x(t).
Suppose that we have an arbitrary periodic signal x(t). This signal has the Fourier series representation given
by (4.1) and (4.11). Let xN (t) denote the finite series
N
xN (t) = ∑ ck e jkω0 t .
k=−N

(i.e., xN (t) is a Fourier series truncated after the Nth harmonic components). The approximation error is given by

eN (t) = x(t) − xN (t).

Let us also define the mean-squared error (MSE) as

Z
1
EN = |eN (t)|2 dt.
T T

Before we can proceed further, we need to more precisely specify what we mean by convergence. This is necessary
because convergence can be defined in more than one way. For example, two common types of convergence are:

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

86 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

pointwise and MSE. In the case of pointwise convergence, the error goes to zero at every point. If convergence is
pointwise and the rate of convergence is the same everywhere, we call this uniform convergence. In the case of MSE
convergence, the MSE goes to zero, which does not necessarily imply that the error goes to zero at every point.
Now, we introduce a few important results regarding the convergence of Fourier series for various types of periodic
signals. The first result that we consider is for the case of continuous signals as given below.
Theorem 4.1 (Convergence of Fourier series (continuous case)). If the periodic signal x(t) is a continuous function
of t, then its Fourier series converges uniformly (i.e., converges pointwise and at the same rate everywhere).
In other words, in the above theorem, we have that if x(t) is continuous, then as N → ∞, eN (t) → 0 for all t.
Often, however, we must work with signals that are not continuous. For example, many useful periodic signals are
not continuous (e.g., the square wave). Consequently, we must consider the matter of convergence for signals with
discontinuities.
Another important result regarding convergence applies to signals that have finite energy over a single period.
Mathematically, a signal x(t) has finite energy over a single period if it satisfies
Z
|x(t)|2 dt < ∞.
T

In the case of such a signal, we have the following important result.

Theorem 4.2 (Convergence of Fourier series (finite-energy case)). If the periodic signal x(t) has finite energy in a
R
single period (i.e., T |x(t)|2 dt < ∞), the Fourier series converges in the MSE sense.
In other words, in the above theorem, we have that if x(t) is of finite energy, then as N → ∞, EN → 0.
The last important result regarding convergence that we shall consider relates to what are known as the Dirichlet
conditions. The Dirichlet1 conditions for the periodic signal x(t) are as follows:
R
1. Over a single period, x(t) is absolutely integrable (i.e., T |x(t)| dt < ∞).
2. In any finite interval of time, x(t) is of bounded variation. In other words, there must be a finite number of
maxima and minima in a single period of x(t).
3. In any finite interval of time, x(t) has a finite number of discontinuities, each of which is finite.
Theorem 4.3 (Convergence of Fourier series (Dirichlet case)). If x(t) is a periodic signal satisfying the Dirichlet
conditions, then:
1. The Fourier series converges pointwise everywhere to x(t), except at the points of discontinuity of x(t).
2. At each point t = ta of discontinuity of x(t), the Fourier series converges to 21 (x(ta− ) + x(ta+ )) where x(ta− ) and
x(ta+ ) denote the values of the signal on the left- and right-hand sides of the discontinuity, respectively.
In other words, if the Dirichlet conditions are satisfied, then as N → ∞, eN (t) → 0 for all t except at discontinuities.
Furthermore, at each discontinuity, the Fourier series converges to the average of the signal values on the left- and
right-hand side of the discontinuity.
Example 4.6. Consider the periodic function x(t) with period T = 2 as shown in Figure 4.4. Let x̂(t) denote the
Fourier series representation of x(t) (i.e., x̂(t) = ∑∞
k=−∞ ck e
jkω0 t , where ω = π ). Determine the values x̂(0) and x̂(1).
0

Solution. We begin by observing that x(t) satisfies the Dirichlet conditions. Consequently, Theorem 4.3 applies.
Thus, we have that
 
x̂(0) = 12 x(0− ) + x(0+ ) = 21 (0 + 1) = 12 and
 
x̂(1) = 21 x(1− ) + x(1+ ) = 21 (1 + 0) = 12 .

1 Pronounced Dee-ree-klay.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.5. PROPERTIES OF CONTINUOUS-TIME FOURIER SERIES 87

x(t)

··· ···

t
−1 0 1 2 3

Figure 4.4: Periodic signal x(t).

Although many signals of practical interest satisfy the Dirichlet conditions, not all signals satisfy these conditions.
For example, consider the periodic signal x(t) defined by

x(t) = 1/t for 0 < t ≤ 1 and x(t) = x(t + 1).

This signal is plotted in Figure 4.5(a). This signal violates the first condition, since the signal is not absolutely
integrable over a single period.
Consider the periodic signal x(t) defined by
 
2π
x(t) = sin for 0 < t ≤ 1 and x(t) = x(t + 1).
t

This signal is plotted in Figure 4.5(b). Since this signal has an infinite number of minima and maxima in a single
period, the second condition is violated.
The third condition is violated by the periodic signal shown in Figure 4.5(c), which has an infinite number of
discontinuities in a single period.
One might wonder how the Fourier series converges for periodic signals with discontinuities. Let us consider the
periodic square wave from Example 4.1. In Figure 4.6, we have plotted the truncated Fourier series xN (t) for the
square wave (with period T = 1 and amplitude A = 1) for several values of N. At the discontinuities of x(t), we can
see that the series appears to converge to the average of the signal values on either side of the discontinuity. In the
vicinity of a discontinuity, however, the truncated series xN (t) exhibits ripples and the peak amplitude of the ripples
does not seem to decrease with increasing N. As it turns out, as N increases, the ripples get compressed towards
discontinuity, but, for any finite N, the peak amplitude of the ripples remains constant. This behavior is known as
Gibbs phenomenon.

4.5 Properties of Continuous-Time Fourier Series

Fourier series representations possess a number of important properties. In the sections that follow, we introduce a
few of these properties. For convenience, these properties are also summarized later in Table 4.1 (on page 91).

4.5.1 Linearity
Let x(t) and y(t) denote two periodic signals with period T and frequency ω0 = 2π /T . If
FS
x(t) ←→ ak and
FS
y(t) ←→ bk ,

then
FS
Ax(t) + By(t) ←→ Aak + Bbk ,

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

88 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

t
0 1

(a)

t
1 2

··· ··· ··· ···

(b)

··· ···
1
2

1
4
··· ···
t
−1 0 1

(c)

Figure 4.5: Examples of signals that violate the Dirichlet conditions. (a) A signal that is not absolutely integrable over
a single period. (b) A signal that has an infinite number of maxima and minima over a single period. (c) A signal that
has an infinite number of discontinuities over a single period.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.5. PROPERTIES OF CONTINUOUS-TIME FOURIER SERIES 89

1.5 1.5

1 1

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

-0.5 -0.5

-1 -1

-1.5 -1.5

(a) (b)

1.5 1.5

1 1

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

-0.5 -0.5

-1 -1

-1.5 -1.5

(c) (d)

Figure 4.6: Gibbs phenomenon. The Fourier series for the periodic square wave truncated after the Nth harmonic
components for (a) N = 3, (b) N = 7, (c) N = 11, and (d) N = 101.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

90 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

where A and B are complex constants. In other words, a linear combination of signals produces the same linear
combination of their Fourier series coefficients.
To prove the above property, we proceed as follows. First, we express x(t) and y(t) in terms of their corresponding
Fourier series as
∞
x(t) = ∑ ak e jkω0 t and
k=−∞
∞
y(t) = ∑ bk e jkω0 t .
k=−∞

Now, we determine the Fourier series of Ax(t) + By(t). We have

∞ ∞
Ax(t) + By(t) = A ∑ ak e jkω0 t + B ∑ bk e jkω0 t
k=−∞ k=−∞
∞ ∞
= ∑ Aak e jkω0 t + ∑ Bbk e jkω0 t
k=−∞ k=−∞
∞
= ∑ (Aak + Bbk )e jkω0 t .
k=−∞

FS
Therefore, we have that Ax(t) + By(t) ←→ Aak + Bbk .

4.5.2 Time Shifting

Let x(t) denote a periodic signal with period T and frequency ω0 = 2π /T . If

FS
x(t) ←→ ak ,

then
FS
x(t − t0 ) ←→ e− jkω0 t0 ak = e− jk(2π /T )t0 ak ,

where t0 is a real constant. From this, we can see that time shifting a periodic signal does not change the magnitude
of its Fourier series coefficients (since e jθ = 1 for any real θ ).
To prove the time-shifting property, we proceed as follows. The Fourier series of x(t) is given by
∞
x(t) = ∑ ak e jkω0 t . (4.14)
k=−∞

We express x(t − t0 ) in terms of its Fourier series, and then use algebraic manipulation to obtain:
∞
x(t − t0 ) = ∑ ak e jkω0 (t−t0 )
k=−∞
∞
= ∑ ak e jkω0 t e− jkω0 t0
k=−∞
∞  
= ∑ ak e− jkω0 t0 e jkω0 t . (4.15)
k=−∞

FS
Comparing (4.14) and (4.15), we have that x(t − t0 ) ←→ e− jkω0 t0 ak .

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.6. FOURIER SERIES AND FREQUENCY SPECTRA 91

Table 4.1: Fourier Series Properties

Property Time Domain Fourier Domain
Linearity Ax(t) + By(t) Aak + Bbk
Time-Domain Shifting x(t − t0 ) e− jkω0 t0 ak
Time Reversal x(−t) a−k

4.5.3 Time Reversal

Let x(t) denote a periodic signal with period T and frequency ω0 = 2π /T . If

FS
x(t) ←→ ak ,

then
FS
x(−t) ←→ a−k .

In other words, the time reversal of a signal results in the time reversal of the corresponding sequence of Fourier series
FS
coefficients. Furthermore, if x(t) is a real signal, we have from (4.5) that a−k = a∗k and obtain x(−t) ←→ a∗k .
To prove the time-reversal property, we proceed in the following manner. The Fourier series of x(t) is given by
∞
x(t) = ∑ ak e jkω0 t . (4.16)
k=−∞

Now, we consider the Fourier series expansion of x(−t). The Fourier series in this case is given by
∞
x(−t) = ∑ ak e jkω0 (−t) . (4.17)
k=−∞

Now, we employ a change of variable. Let l = −k so that k = −l. Performing this change of variable, we can rewrite
(4.17) to obtain
∞
x(−t) = ∑ a−l e j(−l)ω0 (−t)
l=−∞
∞
= ∑ a−l e jl ω0 t . (4.18)
l=−∞

FS
Comparing (4.16) and (4.18), we have that x(−t) ←→ a−k .

4.6 Fourier Series and Frequency Spectra

The Fourier series represents a signal in terms of harmonically-related complex sinusoids. In this sense, the Fourier
series captures information about the frequency content of a signal. Each complex sinusoid is associated with a partic-
ular frequency (which is some integer multiple of the fundamental frequency). So, these coefficients indicate at which
frequencies the information/energy in a signal is concentrated. For example, if only the Fourier series coefficients for
the low order harmonics have large magnitudes, then the signal is mostly associated with low frequencies. On the
other hand, if a signal has many large magnitude coefficients for high order harmonics, then the signal has a consid-
erable amount of information/energy associated with high frequencies. In this way, the Fourier series representation
provides a means for measuring the frequency content of a signal. The distribution of the energy/information in a
signal over different frequencies is referred to as the frequency spectrum of the signal.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

92 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

To gain further insight into the role played by the Fourier series coefficients ck in the context of the frequency
spectrum of the signal x(t), it is helpful to write the Fourier series with the ck expressed in polar form as follows:

∞
x(t) = ∑ ck e jkω0 t
k=−∞
∞
= ∑ |ck | e j arg ck e jkω0 t
k=−∞
∞
= ∑ |ck | e j(kω0 t+arg ck ) .
k=−∞

Clearly (from the last line of the above equation), the kth term in the summation corresponds to a complex sinusoid
with fundamental frequency kω0 that has had its amplitude scaled by a factor of |ck | and has been time-shifted by an
amount that depends on arg ck . For a given k, the larger |ck | is, the larger the amplitude of its corresponding complex
sinusoid e jkω0 t , and therefore the larger the contribution the kth term (which is associated with frequency kω0 ) will
make to the overall summation. In this way, we can use |ck | as a measure of how much information a signal x(t) has
at the frequency kω0 .
Various ways exist to illustrate the frequency spectrum of a signal. Typically, we plot the Fourier series coefficients
as a function of frequency. Since, in general, the Fourier series coefficients are complex valued, we usually display
this information using two plots. One plot shows the magnitude of the coefficients as a function of frequency. This is
called the magnitude spectrum. The other plot shows the arguments of the coefficients as a function of frequency.
In this context, the argument is referred to as the phase, and the plot is called the phase spectrum of the signal.
Since the Fourier series only has frequency components at integer multiples of the fundamental frequency, we
only have values to plot for these particular frequencies. In other words, the frequency spectrum is discrete in the
independent variable (i.e., frequency). For this reason, we use a stem graph to plot such functions. Due to the general
appearance of the graph (i.e., a number of vertical lines at various frequencies) we refer to such spectra as line spectra.
Recall that, for a real signal x(t), the Fourier series coefficient sequence ck satisfies ck = c∗−k . This, however,
implies that |ck | = |c−k | and arg ck = − arg c−k . Since |ck | = |c−k |, the magnitude spectrum of a real signal is always
even. Similarly, since arg ck = − arg c−k , the phase spectrum of a real signal is always odd.

Example 4.7. The periodic square wave x(t) in Example 4.1 has fundamental period T , fundamental frequency ω0 ,
and the Fourier series coefficient sequence given by
(
− j2A
πk for k odd
ck =
0 for k even,

where A is a positive constant. Find and plot the magnitude and phase spectra of this signal. Determine at what
frequency (or frequencies) this signal has the most information.

Solution. First, we compute the magnitude spectrum of x(t), which is given by |ck |. We have
(
− j2A
πk for k odd
|ck | =
0 for k even
(
2A
π |k| for k odd
=
0 for k even.

Next, we compute the phase spectrum of x(t), which is given by arg ck . Using the fact that arg 0 = 0 and arg −πj2A
k =

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.6. FOURIER SERIES AND FREQUENCY SPECTRA 93

|ck |

2A
π

··· ···
ω
−4ω0 −3ω0 −2ω0 −ω0 0 ω0 2ω0 3ω0 4ω0
(a)

∠ck

π
2

···
ω
−4ω0 −3ω0 −2ω0 −ω0 ω0 2ω0 3ω0 4ω0

···
− π2

(b)

Figure 4.7: Frequency spectrum of the periodic square wave. (a) Magnitude spectrum and (b) phase spectrum.

− π2 sgn k, we have
(
arg −πj2A
k for k odd
arg ck =
arg 0 for k even

π

2 for k odd, k < 0
= − π2 for k odd, k > 0


0 for k even
(
− π2 sgn k for k odd
=
0 for k even.

The magnitude and phase spectra of x(t) are plotted in Figures 4.7(a) and (b), respectively. Note that the magnitude
spectrum is an even function, while the phase spectrum is an odd function. This is what we should expect, since x(t)
is real. Since ck is largest in magnitude for k = −1 and k = 1, the signal x(t) has the most information at frequencies
−ω0 and ω0 .
Example 4.8. The periodic impulse train x(t) in Example 4.2 has fundamental period T , fundamental frequency ω0 ,
and the Fourier series coefficient sequence given by
A
ck = ,
T
where A is a positive constant. Find and plot the magnitude and phase spectra of this signal.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

94 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

|ck | ∠ck

A
T

··· ···
ω ··· · · ·ω
−3ω0 −2ω0 −ω0 0 ω0 2ω0 3ω0 −3ω0 −2ω0 −ω0 0 ω0 2ω0 3ω0
(a) (b)

Figure 4.8: Frequency spectrum for the periodic impulse train. (a) Magnitude spectrum and (b) phase spectrum.

A
Solution. We have |ck | = T and arg ck = 0. The magnitude and phase spectra of x(t) are plotted in Figures 4.8(a) and
(b), respectively.

4.7 Fourier Series and LTI Systems

In Section 3.10, we showed that complex exponentials are eigenfunctions of LTI systems. More specifically, for a LTI
system defined by the operator H and having impulse response h(t), we showed that
H
est → H(s)est ,
where Z ∞
H(s) = h(τ )e−sτ d τ (4.19)
−∞
(i.e., H(s) is the system function).
Often, we are interested in the case of H(s) when Re{s} = 0 (i.e., s is purely imaginary). Let s = jω where ω is
real. Substituting s = jω for s in (4.19), we obtain
Z ∞
H( jω ) = H(s)|s= jω = h(t)e− jω t dt.
−∞

We call H( jω ) the frequency response of the system. From above, it follows that an LTI system must be such that
e jω t → H( jω )e jω t . (4.20)
Suppose now that we have a periodic signal x(t) represented in terms of a Fourier series as
∞
x(t) = ∑ ck e jkω0 t .
k=−∞

Using (4.20) and the superposition property, we can determine the system response y(t) to the input x(t) as follows:
y(t) = H {x(t)}
( )
∞
jkω0 t
=H ∑ ck e
k=−∞
∞ n o
= ∑ H ck e jkω0 t
k=−∞
∞ n o
= ∑ ck H e jkω0 t
k=−∞
∞
= ∑ ck H( jkω0 )e jkω0 t .
k=−∞

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.8. FILTERING 95

Therefore, we can view a LTI system as an entity that operates on the individual coefficients of a Fourier series. In
particular, the system forms its output by multiplying each Fourier series coefficient by the value of the frequency
response function at the frequency to which the Fourier series coefficient corresponds. In other words, if
FS
x(t) ←→ ck

then
FS
y(t) ←→ H( jkω0 )ck .

Example 4.9. Suppose that we have a LTI system with the frequency response

H( jω ) = e− jω /4 .

Find the response y(t) of the system to the input x(t) where

x(t) = 21 cos(2π t) = 41 (e j2π t + e− j2π t ).

Solution. The Fourier series for x(t) is given by

∞
x(t) = ∑ ck e jkω0 t ,
k=−∞

where ω0 = 2π , c−1 = 14 , c1 = 41 , and ck = 0 for k 6∈ {−1, 1}. Thus, we can write

∞
y(t) = ∑ ck H( jkω0 )e jkω0 t
k=−∞
= c−1 H(− jω0 )e− jω0 t + c1 H( jω0 )e jω0 t
= 14 H(− j2π )e− j2π t + 41 H( j2π )e j2π t
= 14 e jπ /2 e− j2π t + 41 e− jπ /2 e j2π t
= 14 [e− j(2π t−π /2) + e j(2π t−π /2) ]


= 14 2 cos(2π t − π2 )
= 12 cos(2π t − π2 )
 

= 12 cos 2π t − 41 .

In other words, the output y(t) is just a shifted version of the input x(t), namely x(t − 14 ). As it turns out, the frequency
response H( jω ) corresponds to an ideal delay of 14 .

4.8 Filtering
In some applications, we want to change the relative amplitude of the frequency components of a signal or possibly
eliminate some frequency components altogether. This process of modifying the frequency components of a signal is
referred to as filtering. Various types of filters exist. Frequency selective filters pass some frequencies with little or no
distortion, while significantly attenuating other frequencies. Several basic types of frequency-selective filters include:
lowpass, highpass, and bandpass.
An ideal lowpass filter eliminates all frequency components with a frequency greater than some cutoff frequency,
while leaving the remaining frequency components unaffected. Such a filter has a frequency response of the form
(
1 for |ω | ≤ ωc
H( jω ) =
0 otherwise,

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

96 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

H( jω )

ω
−ωc ωc
(a)

H( jω )

1
··· ···

ω
−ωc ωc
(b)

H( jω )

ω
−ωc2 −ωc1 ωc1 ωc2
(c)

Figure 4.9: Frequency responses of (a) ideal lowpass, (b) ideal highpass, and (c) ideal bandpass filters.

where ωc is the cutoff frequency. A plot of this frequency response is given in Figure 4.9(a).
The ideal highpass filter eliminates all frequency components with a frequency less than some cutoff frequency,
while leaving the remaining frequency components unaffected. Such a filter has a frequency response of the form
(
1 for |ω | ≥ ωc
H( jω ) =
0 otherwise,

where ωc is the cutoff frequency. A plot of this frequency response is given in Figure 4.9(b).
An ideal bandpass filter eliminates all frequency components that do not lie in its passband, while leaving the
remaining frequency components unaffected. Such a filter has a frequency response of the form
(
1 for ωc1 ≤ |ω | ≤ ωc2
H( jω ) =
0 otherwise,

where the limits of the passband are ωc1 and ωc2 . A plot of this frequency response is given in Figure 4.9(c).

Example 4.10 (Lowpass filtering). Suppose that we have a LTI system with input x(t), output y(t), and frequency

Copyright c 2013 Michael D. Adams Version: 2013-09-11

4.8. FILTERING 97

response H( jω ), where
(
1 for |ω | ≤ 3π
H( jω ) =
0 otherwise.
Further, suppose that the input x(t) is the periodic signal
x(t) = 1 + 2 cos 2π t + cos 4π t + 12 cos 6π t.
Find the Fourier series representation of x(t). Use this representation in order to find the response y(t) of the system
to the input x(t). Plot the frequency spectra of x(t) and y(t).
Solution. We begin by finding the Fourier series representation of x(t). Using Euler’s formula, we can re-express x(t)
as
x(t) = 1 + 2 cos 2π t + cos 4π t + 12 cos 6π t
    h i
= 1 + 2 21 (e j2π t + e− j2π t ) + 21 (e j4π t + e− j4π t ) + 21 21 (e j6π t + e− j6π t )
= 1 + e j2π t + e− j2π t + 21 [e j4π t + e− j4π t ] + 41 [e j6π t + e− j6π t ]
= 41 e− j6π t + 21 e− j4π t + e− j2π t + 1 + e j2π t + 21 e j4π t + 41 e j6π t
= 41 e j(−3)(2π )t + 12 e j(−2)(2π )t + e j(−1)(2π )t + e j(0)(2π )t + e j(1)(2π )t + 12 e j(2)(2π )t + 14 e j(3)(2π )t .
From the last line of the preceding equation, we deduce that ω0 = 2π , since a larger value for ω0 would imply that
some Fourier series coefficient indices are noninteger, which clearly makes no sense. Thus, we have that the Fourier
series of x(t) is given by
∞
x(t) = ∑ ak e jkω0 t ,
k=−∞

where ω0 = 2π and



1 for k = 0

1

 for k = ±1
ak = 1
2 for k = ±2

 1


 4 for k = ±3

0 otherwise.
Since the system is LTI, we know that the output y(t) has the form
∞
y(t) = ∑ bk e jkω0 t ,
k=−∞

where
bk = ak H( jkω0 ).
Using the results from above, we can calculate the bk as follows:
b0 = a0 H( j[0][2π ]) = 1(1) = 1,
b1 = a1 H( j[1][2π ]) = 1(1) = 1,
b−1 = a−1 H( j[−1][2π ]) = 1(1) = 1,
b2 = a2 H( j[2][2π ]) = 21 (0) = 0,
b−2 = a−2 H( j[−2][2π ]) = 21 (0) = 0,
b3 = a3 H( j[3][2π ]) = 41 (0) = 0, and
1
b−3 = a−3 H( j[−3][2π ]) = 4 (0) = 0.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

98 CHAPTER 4. CONTINUOUS-TIME FOURIER SERIES

ak
1 H( jω )

1
2

ω
−6π −4π −2π 0 2π 4π 6π
(a)

bk

1
2

ω
−6π −4π −2π 0 2π 4π 6π
(b)

Figure 4.10: Frequency spectra of the input and output signals.

Thus, we have

1 for k = 0

bk = 1 for k = ±1


0 otherwise.

Lastly, we plot the frequency spectra of x(t) and y(t) in Figures 4.10(a) and (b), respectively. The frequency
response H( jω ) is superimposed on the plot of the frequency spectrum of x(t) for illustrative purposes.

Copyright c 2013 Michael D. Adams Version: 2013-09-11