AE2235-II Instrumentation & Signals

The Fourier Series

Max Mulder

Version 2019-2020
Wednesday April 23, 2020

Delft
University of
Technology

Challenge the future

 Introduction

Introduction Real F-S Complex F-S Symmetry

. 2 / 66
3-1 Introduction
In this chapter (and Chapter 4) we consider procedures for resolving
certain classes of signals into superpositions of sines and cosines
or, equivalently, complex exponential signals in the form exp(jωt).
For periodic signals this resolution results in the Fourier series.
For aperiodic signals this resolution results in the Fourier integral.
The first advantage of Fourier series and Fourier transform
representations is that they immediately characterize the signal in
the frequency domain.

Note The second advantage is that when a signal can be described as the sum of sinusoidal components and since for a
linear system the superposition principle holds, the response of the system to any input signal can be computed as the sum
of the system responses to each of the sinusoidal components. This will be discussed in the next lecture.

Introduction Real F-S Complex F-S Symmetry

. 3 / 66
Fourier
Jean Baptiste Joseph Fourier (1768-1830) was a French math-
ematician and physicist.
He went with Napoleon to Egypt, then returned to France to be-
come préfect of Isère, and there, in his spare time, he worked on
the mathematical descriptions of heat transfer and vibrations.
In 1822 he presented his main work on heat flow, in which he
claims that any function of a variable, whether continuous
or discontinuous, can be expanded in a series of sines of
multiples of the variable.
Though this result is not correct, Fourier’s observation that some
discontinuous functions are the sum of infinite series was a
breakthrough.
The Fourier series and the Fourier transform have been named
in his honour.

Introduction Real F-S Complex F-S Symmetry

. 4 / 66
3-2 Trigonometric Series
A powerful technique in signal analysis is the so-called signal
de-composition method, where the signal concerned is viewed as a
linear combination of certain basis functions.
The Fourier series expansion relies on sinusoids as its basic
components and this has proven to be an extremely suitable choice.
Ziemer approaches the Fourier series somewhat heuristically, using
Examples 3-1 and 3-2, and then continues with a formal derivation.

Introduction Real F-S Complex F-S Symmetry

. 5 / 66
3-2 Trigonometric Series (2)

Example 3-1 We consider the trigonometric series given by:

x(t) = sin(ω0 t) + 13 sin(3ω0 t) + 15 sin(5ω0 t) + . . . (3-1)
where 2π/ω0 is the period. Note that since the frequencies of all sinusoidal
components are integer multiples of the fundamental frequency ω0 (their
frequencies are commensurable), the signal is periodic.
The partial sums of (3-1) can be defined as:
s1 = sin(ω0 t)
s2 = sin(ω0 t) + 31 sin(3ω0 t)
etcetera.

The first partial sum is just a sinusoid (Figure 3-1a).

Introduction Real F-S Complex F-S Symmetry

. 6 / 66
3-2 Trigonometric Series (3)

Introduction Real F-S Complex F-S Symmetry

. 7 / 66
3-2 Trigonometric Series (4)

We see that, a square wave appears to be taking shape as more terms are
included in the partial sum. As each harmonic (i.e., a sinusoid component with a
frequency that is an integer multiple of the fundamental frequency) is added, the
ripple in the ‘flat top’ becomes smaller, and possesses a frequency equal to that of
the highest harmonic included in the series.
The convergence of the series at any particular point may be examined by simply
substituting the appropriate value of ω0 t. For example, when ω0 t = π/2, we obtain:
1 − 31 + 51 − 71 + 19 + . . . = π
4
If we want the series (3-1) to have a value of one at ω0 t = π/2 then we simply
multiply the series with 4/π.
Indeed, we will see a little later that the unit-amplitude square wave can be
represented by the series:
4 1 1


xsq (t) = π sin(ω0 t) + 3 sin(3ω0 t) + 5 sin(5ω0 t) + . . . , −∞ < t < ∞
Introduction Real F-S Complex F-S Symmetry

. 8 / 66
3-2 Trigonometric Series (5)

Example 3-2 shows a similar result, where again a particular sum of sinusoidal
components is capable of describing a saw-tooth waveform (Figure 3-2).

Introduction Real F-S Complex F-S Symmetry

. 9 / 66
3-2 Trigonometric Series (6)

Hence, it is clear that apparently there exists the possibility of

building up arbitrary periodic waveforms from sums of harmonically
related sinusoidal terms.

Then: Given a specific periodic waveform, how do we

find its trigonometric series representation?

It will be shown that the resulting series, unique for each periodic
signal, is called the trigonometric Fourier series of that signal.

Introduction Real F-S Complex F-S Symmetry

. 10 / 66
 Real F-S

Introduction Real F-S Complex F-S Symmetry

. 11 / 66
3-3 Fourier Series derivation
We begin with stating the general form that the trigonometric Fourier series
representation of a periodic signal may have:
x(t) = a0 + a1 cos(ω0 t) + a2 cos(2ω0 t) + . . .
+b1 sin(ω0 t) + b2 sin(2ω0 t) + . . ., , −∞ < t < ∞ (3-5)

or, in a more compact form, in terms of summations:

n=∞
P n=∞
P
x(t) = a0 + an cos(nω0 t) + bn sin(nω0 t) (3-6)
n=1 n=1

Note that as the right-hand side of (3-6) is a sum of harmonically related

sinusoids, which is periodic, so must be the left-hand side of (3-6).

The problem now is to find values of a0 , a1 , a2 , . . . and b1 , b2 , . . . in such a

way that a given x(t) is represented (or, approximated!) best.

Introduction Real F-S Complex F-S Symmetry

. 12 / 66
3-3 Fourier Series derivation (2)

We begin with R a0 . The approach is to integrate all terms in (3-5) over one period,
indicated by T0 :
R R R R
T0
x(t)dt = T0 a0 dt + T0 a1 cos(ω0 t)dt + T0 a2 cos(2ω0 t)dt + . . .
R R
+ T0 b1 sin(ω0 t)dt + T0 b2 sin(2ω0 t)dt + . . . (3-7)
Now, all terms on the right-hand side, except the first one, are zero, as we are
integrating a sine or cosine over one period. For any sin(nω0 t) and cos(nω0 t) as
much area appears above the t-axis as below it in one period.
R
Obviously then, we get: T0 x(t)dt = a0 T0 , and therefore:

1 R
a0 = T T0 x(t)dt (3-8)
0

Hence, a0 is simply the average value of the waveform. Another interpretation of

a0 is that it represents a cosine with zero frequency (i.e., an extremely slowly
varying cosine signal).
Introduction Real F-S Complex F-S Symmetry

. 13 / 66
3-3 Fourier Series derivation (3)

Using the series form (3-6) we then derive a general expression valid for any of the
an coefficients (except a0 which is now known). The approach is to multiply both
sides of (3-6) with cos(mω0 t) and then integrate over one period T0 :
R R
T0
x(t) cos(mω0 t)dt = a0 T0 cos(mω0 t)dt
n=∞ 
R P
+ T0 an cos(nω0 t) cos(mω0 t)dt
n=1
n=∞ 
R P
+ T0 bn sin(nω0 t) cos(mω0 t)dt (3-9)
n=1
The first term on the right-hand side is zero. We then multiply each term of the
series in parentheses by the cos(mω0 t) term and integrate term-by-term (note that
essentially we interchange the order of summation and integration):
R n=∞
P R
T0
x(t) cos(mω0 t)dt = an T0 cos(nω0 t) cos(mω0 t)dt
n=1
n=∞
P R
+ bn T0
sin(nω0 t) cos(mω0 t)dt (3-10)
n=1
Introduction Real F-S Complex F-S Symmetry

. 14 / 66
3-3 Fourier Series derivation (4) VIS
Now we are going to use the orthogonality property of integrals involving
products of sines and cosines:

R 0 : m 6= n
I1 = T0 sin(mω0 t) sin(nω0 t)dt = (3-11)
T /2 : m = n 6= 0
 0
R 0 : m 6= n
I2 = T0 cos(mω0 t) cos(nω0 t)dt = (3-12)
T0 /2 : m = n 6= 0
R
I3 = T0 sin(mω0 t) cos(nω0 t)dt = 0, for all m, n (3-13)

In all these integrals, m and n are integers and T0 is the period corresponding with
the fundamental frequency ω0 .
The orthogonality of (3-11) is shown graphically in the next slides.

Introduction Real F-S Complex F-S Symmetry

. 15 / 66
3-3 Fourier Series derivation (5)

Orthogonality Here we show y1 (t) = sin(2πf0 t), y2 (t) = sin(2π2f0 t), and y3 (t) = sin(2π3f0 t) as a

function of time, for f0 = 1/20 s. We also show the products y1 (t)y1 (t), y1 (t)y2 (t) and y1 (t)y3 (t).

1
y1 (t) 1
y1 (t) (blue) and y2 (t) (green)
1
y1 (t) (blue) and y3 (t) (green)

0.5 0.5 0.5

0 0 0

−0.5 −0.5 −0.5

−1 −1 −1
0 10 20 0 10 20 0 10 20
t, s t, s t, s
y1 (t)y1 (t) y1 (t)y2 (t) y1 (t)y3 (t)
1 1 1

0.8
0.5 0.5

0.6
0 0
0.4

−0.5 −0.5
0.2

0 −1 −1
0 10 20 0 10 20 0 10 20
t, s t, s t, s

Introduction Real F-S Complex F-S Symmetry

. 16 / 66
3-3 Fourier Series derivation (6)

Orthogonality (2) Below we show y1 (t) = sin(2πf0 t), y2 (t) = sin(2π2f0 t), and their product y1 (t)y2 (t),

all as a function of time. Note that areas ‘a’ and ‘b’ are complementary: their sum is zero (a + b + b + a = 0). In fact, here b = -a.

1
y1 (t) (blue) and y2 (t) (green)

0.5

−0.5

−1
0 2 4 6 8 10 12 14 16 18 20
t, s
y1 (t)y2 (t)
1

0.5

a b b a
−0.5

−1
0 2 4 6 8 10 12 14 16 18 20
t, s

Introduction Real F-S Complex F-S Symmetry

. 17 / 66
3-3 Fourier Series derivation (7)

Orthogonality (3) Below we show y1 (t) = sin(2πf0 t), y3 (t) = sin(2π3f0 t), and their product y1 (t)y3 (t),

all as a function of time. Note that areas ‘a’ and ‘b’ are complementary: their sum is zero (a + b + a + a + b + a = 0). Here b = - 2 a!

1
y1 (t) (blue) and y3 (t) (green)

0.5

−0.5

−1
0 2 4 6 8 10 12 14 16 18 20
t, s
y1 (t)y3 (t)
1

0.5

a b a a b a
−0.5

−1
0 2 4 6 8 10 12 14 16 18 20
t, s

Introduction Real F-S Complex F-S Symmetry

. 18 / 66
3-3 Fourier Series derivation (8)

Applying these integral properties to (3-10):

R n=∞
P R
T0
x(t) cos(mω0 t)dt = an T0 cos(nω0 t) cos(mω0 t)dt
n=1
n=∞
P R
+ bn T0
sin(nω0 t) cos(mω0 t)dt, (3-10)
n=1

we see that each term in the second series on the right-hand side is zero, for all
values of m and n (integral I3 (3-13)).
We also see that each term in the first series on the right-hand side is also zero,
except for the one term where n equals m, where the integral equals T0 /2 (integral
I2 (3-12)). We therefore obtain:
 
R
x(t) cos(mω t)dt = a T0
T0 0 m 2
2 R
am = T T0 x(t) cos(mω0 t)dt , m ∈ N + (3-15)
0

Introduction Real F-S Complex F-S Symmetry

. 19 / 66
3-3 Fourier Series derivation (9) VIS
In a similar manner, we can find the values for coefficients bn , namely by
multiplying (3-6) with sin(mω0 t), integrating it over one period T0 and using the
integral properties (3-11) and (3-13). We obtain:
2 R
bm = T T0 x(t) sin(mω0 t)dt (3-16)
0

In conclusion, the coefficients of the trigonometric Fourier series

representation of a periodic signal x(t) can be computed as follows:
1 R
a0 = T T0 x(t)dt (3-8)
0

2 R
am = T T0 x(t) cos(mω0 t)dt , m ∈ N + (3-15)
0

2 R
bm = T T0 x(t) sin(mω0 t)dt , m ∈ N + (3-16)
0

Introduction Real F-S Complex F-S Symmetry

. 20 / 66
3-3 Fourier Series derivation (10)

On pages 106-107 Ziemer discusses some properties of the Fourier series.

The assumption of (3-6) was that an arbitrary periodic function can be represented
by a sum of sines and cosines. This is not always the case, however.
If x(t) has a period T0 and has continuous first and second derivatives for all t,
except possibly at a finite number of points where it may have jump discontinuities,
the Fourier series for x(t) will converge uniformly to x(t) for every t except at the
points of discontinuity.

The limit of the Fourier series sum at a point of discontinuity is simply the average
of the left- and right-hand limits (Figure 3-3).
Introduction Real F-S Complex F-S Symmetry

. 21 / 66
3-3 Fourier Series derivation (11)

Another important property is that when x(t) is known only for a particular time
interval, like T0 seconds, then the Fourier series coefficients will be computed in
such a way that the Fourier series representation converges to x(t) within T0 .
Now, outside this interval the Fourier series will describe a signal that is the
periodic extension of x(t), see Figure 3-4.

Even when x(t) is not periodic, its Fourier series approximation will always be!

Introduction Real F-S Complex F-S Symmetry

. 22 / 66
3-3 Fourier Series example
@home: Study Example 3-3
Example 3-4 We consider the square wave defined by:

A : 0 < t < T0 /2
x(t) = (3-21)
−A : T0 /2 < t < T0
and periodically extended outside this interval. The average value of x(t) is zero, so a0 = 0.
The coefficients am that belong to the cosine terms in (3-6) are computed:
T
R0
2
am = T0 x(t) cos(mω0 t)dt
0
TR
0 /2 T
R0
2 2
= T0
A cos(mω0 t)dt + T0
(−A) cos(mω0 t)dt
0 T0 /2
!
T0 /2 T0
2A sin(mω0 t) sin(mω0 t)
= T0 mω0 − mω0
0 T0 /2

Recall that ω0 = 2π/T0 such that sin(mω0 T0 /2) = sin(mπ) which equals 0, and
sin(mω0 T0 ) = sin(2mπ) which is also 0. Hence, all the am components are zero!

Introduction Real F-S Complex F-S Symmetry

. 23 / 66
3-3 Fourier Series example (2)

Example 3-4 (continued) Compute the sine terms coefficients in (3-6):

T
R0
2
bm = T0
x(t) sin(mω0 t)dt
0
TR
0 /2 T
R0
2 2
= T0
A sin(mω0 t)dt + T0
(−A) sin(mω0 t)dt
0 T0 /2
!
T0 /2 T0
cos(mω t) cos(mω0 t)
= 2A
T0
− mω0 0 + mω0
0 T0 /2

Simply filling in the integrals above we get:

bm = 2A 1
T0 mω0
(− cos(mω0 T0 /2) + 1 + cos(mω0 T0 ) − cos(mω0 T0 /2))
With ω0 = 2π/T0 , and re-arranging terms:
A
bm = mπ
(1 − 2 cos(mπ) + cos(m2π))
Then since cos(m2π) = 1 for all m, cos(mπ) = −1 for odd m and cos(mπ) = 1 for even m:
 4A
mπ
: m odd
bm =
0 : m even

Introduction Real F-S Complex F-S Symmetry

. 24 / 66
3-3 Fourier Series example (3)

Example 3-4 (continued) The Fourier series of a square wave with

amplitude A and with odd symmetry becomes:


x(t) = 4A
π
sin(ω 0 t) + 1
3
sin(3ω 0 t) + 1
5
sin(5ω 0 t) + . . . (3-25)
We note that only odd harmonic terms are present, this is (as we will see later) simply the
consequence of the fact that x(t) has a special type of symmetry, referred to as “half-wave
(odd)” symmetry.
More symmetry forms are discussed in later sections.
Note that when representing an ODD signal (i.e., a signal where x(−t) = −x(t)) only the
ODD components of the Fourier series (the ‘sines’) are used, a general property.

Similarly, when representing an EVEN signal (i.e., a signal where x(−t) = x(t) only the
EVEN components of the Fourier series (the ‘cosines’) are used.

Introduction Real F-S Complex F-S Symmetry

. 25 / 66
3-3 Fourier Series example (4)

1.5

0.5

partial sum
0

−0.5 max. harmonic = 11

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s
(Note that we took A=1 and T0 = 1 second here)
1.5

0.5
partial sum

−0.5 max. harmonic = 21

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

Introduction Real F-S Complex F-S Symmetry

. 26 / 66
3-3 Fourier Series example (5)

1.5

0.5

partial sum
0

−0.5 max. harmonic = 51

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

1.5

0.5
partial sum

−0.5 max. harmonic = 101

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

Introduction Real F-S Complex F-S Symmetry

. 27 / 66
3-3 Fourier Series example (6)

Now with 501 harmonics:

1.5

0.5
partial sum

−0.5 max. harmonic = 501

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

1.2

1.1 Gibbs’ phenomenon

partial sum

1 (discussed in 4-10)
0.9

0.8

0.7
0.465 0.47 0.475 0.48 0.485 0.49 0.495 0.5 0.505 0.51 0.515
time, s

Introduction Real F-S Complex F-S Symmetry

. 28 / 66
3-3 Fourier Series example (7)

Example 3-4 (continued) In the frequency domain, the signal ‘model’ we

derived (3-25) can be shown as well. For the real-valued trigonometric Fourier series, the
coefficients a0 , am for m = 1, 2, . . . and bm for m = 1, 2, . . . can e.g. be shown as a function
of the m’th integer multiple of the fundamental frequency ω0 .
Note that the coefficients are all real, so we only need the positive frequencies mω0 (for
m = 0, 1, 2, . . .), the harmonics of the oscillation.
We do not directly obtain an ‘amplitude’ or ‘phase’ spectrum, as we do get with the
complex-valued Fourier series. This will be discussed later.
In the plots that follow, we see that to improve the accuracy of our signal model we can add
more sine and cosine components, we include higher harmonics (m increases).
In the frequency domain we therefore get more coefficients at mω0 .

Note that when adding more components, the values of the Fourier coefficients that we
already had, do not change.

Introduction Real F-S Complex F-S Symmetry

. 29 / 66
3-3 Fourier Series example (8)

The signal model in the frequency domain, first 11 trigonometric FS coefficients

1

0.5

am 0

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

1.5

1
bm

0.5

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

Introduction Real F-S Complex F-S Symmetry

. 30 / 66
3-3 Fourier Series example (9)

The signal model in the frequency domain, first 21 trigonometric FS coefficients

1

0.5

am 0

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

1.5

1
bm

0.5

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

Introduction Real F-S Complex F-S Symmetry

. 31 / 66
3-3 Fourier Series example (10)

The signal model in the frequency domain, first 51 trigonometric FS coefficients

1

0.5

am 0

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

1.5

1
bm

0.5

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

Introduction Real F-S Complex F-S Symmetry

. 32 / 66
3-3 Fourier Series example (11)

The signal model in the frequency domain, first 101 trigonometric FS coefficients
1

0.5

am 0

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

1.5

1
bm

0.5

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m

Introduction Real F-S Complex F-S Symmetry

. 33 / 66
3-3 Fourier Series Amplitude/Phase
spectrum
The real trigonometric FS gives an array of real numbers (a0 , and {am , bm } for m = 1, 2, . . .),
which can be converted in the single-sided amplitude and phase spectra as follows.
Consider the k-th harmonic, and suppose we would like to express the cosine/sine
combination (ak , bk ) as a single cosine function with amplitude Ak and phase θk :
ak cos(kω0 t) + bk sin(kω0 t) = Ak cos(kω0 t + θk )
It can be easily shown that (use cos(u + v) = cos(u) cos(v) − sin(u) sin(v) on the right-hand side):
ak = Ak cos(θk ) and bk = −Ak sin(θk )
which allows us to obtain the amplitude and phase of the k-th harmonic:
p
Ak = a2k + b2k and θk = arctan(−bk /ak )
Then, plotting the Am ’s and θm ’s for all values of (k =)m = 1, 2, . . . we obtain the
single-sided amplitude and phase spectra, respectively. Note that A0 = a0 and θ0 = 0,
representing the signal’s average.
On the next slide, we show the amplitude/phase spectra of the signal used in Example 3-4.

Introduction Real F-S Complex F-S Symmetry

. 34 / 66
3-3 Fourier Series Amplitude/Phase
spectrum (2)
The single-sided amplitude
2
and phase spectrum, first 51 trigonometric FS coefficients
1.5

Am
0.5

−0.5

−1
0 10 20 30 40 50 60 70 80 90 100
m
2

1
θm

−1

−2
0 10 20 30 40 50 60 70 80 90 100
m

We see that, as we only needed sine components, all phases equal −π/2; the Am ’s are equal to the bm ’s. Multiplying the index m on

the horizontal axis with the fundamental frequency in Hz (f0 ) or rad/s (ω0 ) yields a frequency axis in Hz or rad/s, resp.

Introduction Real F-S Complex F-S Symmetry

. 35 / 66
 Complex F-S

Introduction Real F-S Complex F-S Symmetry

. 36 / 66
3-4 Complex Exponential Fourier Series
Another form of the Fourier series that involves complex
exponential functions can be obtained by substituting the complex
exponential forms of sin(ω0 t) and cos(ω0 t) in (3-6).



With sin(nω0 t) = 1
2j
ejnω0 t − e−jnω0 t and cos(nω0 t) = 1
2
ejnω0 t + e−jnω0 t we obtain:
n=∞
P
x(t) = a0 + an 12 (ejnω0 t + e−jnω0 t )
n=1
n=∞
P
+ bn 2j1 (ejnω0 t − e−jnω0 t )
n=1
n=∞
P n=∞
P
x(t) = a0 + 1
(a
2 n
− jbn )ejnω0 t + 1
(a
2 n
+ jbn )e−jnω0 t (3-27)
n=1 n=1

... the complex exponential form of the Fourier series (FS).

Introduction Real F-S Complex F-S Symmetry

. 37 / 66
3-4 Complex Exponential FS (2) VIS
We see in (3-27) that both exp(jnω0 t) and exp(−jnω0 t) are present in the series.
This is because the sum of two complex conjugate phasors yields a real signal.

The generic form of the complex exponential Fourier series is:

n=∞
P
x(t) = Xn ejnω0 t (3-28)
n=−∞

which is double-sided, and where Xn are complex constants known

as the Fourier series coefficients. They can be computed with:
1 R
Xn = T T0 x(t)e−jnω0 t dt (3-31)
0

@home: Study Ziemer page 113, the derivation of (3-31).

Introduction Real F-S Complex F-S Symmetry

. 38 / 66
3-4 Complex Exponential FS (3)

Equation (3-31) is known as the analysis equation of the FS, as it

allows us to analyze how the signal can be represented by complex
exponential basis functions.
Equation (3-28) is known as the synthesis equation of the FS, as it
constructs (or ‘synthesizes’) a signal using the complex exponential
basis functions.
Hence, equations (3-31) and (3-28) can be considered to be each
other’s complement (inverse?).

Introduction Real F-S Complex F-S Symmetry

. 39 / 66
3-4 Complex Exponential FS (4)

The complex FS coefficients and the real-valued FS coefficients are

related by:
 1
(a − jbn )
2 n
: n>0
Xn = 1 (3-32a)
(a + jb−n ) : n < 0
2 −n
and X0 = a0 is a real number, the signal’s average (3-32b).
So: X1 = 12 (a1 − jb1 )
X2 = 12 (a2 − jb2 )
...

X−1 = 21 (a1 + jb1 )

X−2 = 12 (a2 + jb2 )
...

Clearly then: X−n = Xn∗ , the complex conjugate.

Introduction Real F-S Complex F-S Symmetry

. 40 / 66
3-4 Complex Exponential FS (5)

Example (3-4) (extended) Remember Example 3-4 where we showed that the
trigonometric Fourier series of a square wave (amplitude A) consist of am ’s which are all
4A
zero, and bm ’s which are given by mπ for odd m and 0 for even m.
Using (3-32a) and (3-32b) we can easily compute the values of the complex exponential
FS, which is double-sided, so we need them for positive m (the positive frequencies, or the
counter-clockwise rotating phasors) and negative m (the negative frequencies, or the
clockwise rotating phasors).
−j
Since all am ’s are zero, the complex FS coefficients are purely imaginary and equal 2 m
b
and 2j bm for positive and negative frequencies, respectively.
The single-sided real-valued FS and the double-sided complex-valued FS coefficients are
shown on the next slide. Note that the complex valued FS can only be drawn like this by
virtue of the fact that it is purely imaginary (complex numbers have two dimensions, right?).

Usually, we draw the complex-valued FS coefficients using the magnitude and phase, as
will be discussed later on.

Introduction Real F-S Complex F-S Symmetry

. 41 / 66
3-4 Complex Exponential FS (6)

Example (3-4) (extended)

Single-sided 51 real FS coefficients am and bm
2 2

1.5 1.5

1 1
am

bm
0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 10 20 30 40 50 0 10 20 30 40 50
m m
Double-sided 51 complex FS coefficients Xm
1

0.5
Im(Xm )

−0.5

−1
−50 −40 −30 −20 −10 0 10 20 30 40 50
m
Introduction Real F-S Complex F-S Symmetry

. 42 / 66
3-4 Complex Exponential FS (7)

Example 3-7 In Table 3-1 the complex exponential FS coefficients of an even

square wave are given (entry 4).

Now, suppose we want to know the FS coefficients of an odd square wave. In other words,
we want to know the change in the Fourier coefficients due to a time shift in the waveform,
in this case a shift of T0 /4.

Consider the general case first. Let the Fourier coefficients of the unshifted waveform be
denoted as Xn and those for the time-shifted waveform be Yn (with time-shift τ0 seconds).

Introduction Real F-S Complex F-S Symmetry

. 43 / 66
3-4 Complex Exponential FS (8)

Example 3-7 (continued) Then:

1 R
Xn = T T0 x(t)e−jnω0 t dt, and
0

Yn = T1 T0 x(t − τ0 )e−jnω0 t dt, with ω0 = 2π/T0

R
(3-41)
0

Change the variable in the second integral t′ = t − τ0 to obtain:

h i
Yn = T1 T0 x(t′ )e−jnω0 t dt′ e−jnω0 τ0 = Xn |e−jn2πτ
R ′
0 /T0
(3-42)
0 {z }
due to time−shift

where we remember that the integration can be over any T0 -second interval. So, in our
case of shifting the even square wave with τ0 = T0 /4 we get:

−jXn : n = ±1, ±5, . . .
Yn = Xn e−jnπ/2 = (3-43)
jXn : n = ±3, ±7, . . .
with Xn the Fourier coefficients of the even square wave (Table 3-1, entry 4), which are all
real and zero for even values of n. Then Yn are the Fourier coefficients of the odd square
wave, all imaginary, and zero for even n. @home: is (3-43) correct?
Introduction Real F-S Complex F-S Symmetry

. 44 / 66
3-4 Complex Exponential FS (9)

Example 3-5 Consider the periodic sequence of asymmetrical pulses:

m=∞  
x(t) =
P
AΠ t − t 0 − mT 0 , τ < T0 (3-33)
τ
m=−∞

Let us first draw this so-called pulse train function in time (Note that it is annotated incorrectly on page 116
(Table 3-1, entry 3); further note that when τ is very small and A = 1/τ we have an impulse train).

T0
A
... τ ...

time t
t0

The function is periodic with T0 . We can use (3-31) to compute the FS coefficients:
t0 +τ
R /2
Xn = 1
T0
Ae−jnω0 t dt, (3-34)
t0 −τ /2
A t0 +τ /2
where ω0 = 2π/T0 . For n = 0: X0 = t|
T0 t0 −τ /2
= Aτ /T0 (3-35b).
Introduction Real F-S Complex F-S Symmetry

. 45 / 66
3-4 Complex Exponential FS (10)

Example 3-5 (continued)

We obtain by simple integration:
t +τ /2
Xn = −A
jnω0 T0
e−jnω0 t t0 −τ /2 , n 6= 0
0 
jnω0 τ /2 −jnω0 τ /2
= nω2A e −jnω0 t0 e −e , n 6= 0
T
0 0 2j
2A −jnω0 t0 nω0 τ


= nω0 T0 e sin 2
, n 6= 0 (3-35a)
When n = 0 we obtain X0 = Aτ
T0 (3-35b), the average of the signal.
Substituting f0 = ω0 /(2π) (where f0 is the fundamental frequency in Herz) in (3-35a), we
obtain the form:
−j2πnf0 t0 sin(πnf0 τ )
Xn = Aτ e
T0 | {z } πnf0 τ (3-35c)
time shift

Equation (3-35c) is an important result and will be used frequently.

Introduction Real F-S Complex F-S Symmetry

. 46 / 66
3-4 Intermezzo: the SINC function VIS
Equation (3-35c) contains a very important function, the so-called
sinc function, defined as:
sin(πz)
sinc(z) = πz (3-36)

The sinc function is even, and has a maximum of ‘1’ at z = 0 (use L’Hospital).
The sinc function equals ‘0’ at z = ±1, ±2, ±3, . . .
Introduction Real F-S Complex F-S Symmetry

. 47 / 66
3-4 Complex Exponential FS (11)

Example 3-5 (continued)

Recall:
−j2πnf0 t0 sin(πnf0 τ )
Xn = Aτ
T0 e (3-35c)
πnf0 τ

Using the sinc function definition in (3-35c) we obtain:

Xn = Aτ
T0 e −j2πnf0 t0
sinc(nf0 τ )

For n = 0 this function has its maximum of AτT0 (here the amplitudes of both the complex
exponential function as well as the sinc function equal one).
When t0 (the time-shift) is zero, the pulse train becomes an even function and as a result
the FS coefficients become all real (i.e., we only need the cosines...).

For t0 = 0, the FS coefficients are zero for integer values of nf0 τ , that is, when
nf0 τ = ±1, ±2, ±3, . . .

Introduction Real F-S Complex F-S Symmetry

. 48 / 66
3-4 Complex Exponential FS (12)
periodic block function with τ = 1 s and T0 = 3 s

1 Take a pulse train with A=1, τ = 1 s

t0 = 0 s and period T0 = 3 s.
0.8 An even, symmetric periodic function.

0.6

0.4

0.2

0
−20 −15 −10 −5 0 5 10 15 20
t, s
sinc function (dotted) and Fourier coefficients (circles) The Fourier series coefficients Xn are all
real and their magnitudes can be shown as a
1
nf0 = −3 nf0 = −2 nf0 = −1 T0 nf0 = 1 nf0 = 2 nf0 = 3 function of frequency f (in Hz).
0.4

0.2 Since T0 = 3 s, f0 = 3 1 Hz.

X1 X3 X5 X7
Xn = 0 when nf0 = ±1, ±2, . . .,
0
i.e., for n = ±3, ±6, ±9, . . ..
X0 X2 X4 X6
−0.2

−4 −3 −2 −1 0 1 2 3 4 To be continued later
f , Hz

Introduction Real F-S Complex F-S Symmetry

. 49 / 66
3-4 Complex Exponential FS (13)

Example 3-6 In this example the complex FS coefficients are computed for a
half-rectified sine wave:

A sin(ω0 t) : 0 ≤ t ≤ T0 /2
x(t) = (3-37)
0 : T0 /2 ≤ t ≤ T0
It can be shown (Ziemer, page 116-117) that the Fourier coefficients become:

A
 π(1 − n2 ) : n = 0, ±2, ±4, . . .

Xn = 0 : n odd and 6= ±1

 1
− 4 jnA : n = ±1
@home: Prove the second entry in Table 3-1.

The figures on the following slides (Figure 3-9) show how fast the partial sums of the FS
approach the half-rectified sine wave. This is because no discontinuities are present in the
waveform.

Introduction Real F-S Complex F-S Symmetry

. 50 / 66
3-4 Complex Exponential FS (14)

1.5

0.5

partial sum
0
max. harmonic = 2
−0.5

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

1.5

0.5
partial sum

0
max. harmonic = 8
−0.5

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

Introduction Real F-S Complex F-S Symmetry

. 51 / 66
3-4 Complex Exponential FS (15)

1.5

0.5

partial sum
0
max. harmonic = 16
−0.5

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

1.5

0.5
partial sum

0
max. harmonic = 32
−0.5

−1

−1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
time, s

Introduction Real F-S Complex F-S Symmetry

. 52 / 66
 Symmetry

Introduction Real F-S Complex F-S Symmetry

. 53 / 66
3-5 Symmetry Properties of the FS
The complex exponential Fourier series (3-31) can be written as:
1
R j
R
Xm = T0 T0 x(t) cos(mω0 t)dt − T0 T0 x(t) sin(mω0 t)dt (3-44)
For real functions x(t), the first term in (3-44) is the real part of Xm
and the second term is the imaginary part of Xm .
Comparing (3-44) with (3-15) and (3-16) we see that:
 1
(am − jbm ) : m>0
Xm = 2
1 (3-45)
(a + jb−m ) : m < 0
2 −m
which is another way to obtain (3-22a). Thus:
am = 2 Re(Xm ) and bm = −2 Im(Xm ) for m > 0, (3-46)
∗
and we already saw that Xm = X−m , the conjugates. (3-47)

Introduction Real F-S Complex F-S Symmetry

. 54 / 66
3-5 Symmetry Properties of the FS (2)

When writing Xm in polar (or phasor-) form we get:

Xm = |Xm | ejθm (3-48)
Clearly then (see (3-44) or just look at (3-47)):
|Xm | = |X−m | and θm = −θ−m , (3-49)
p
and: |Xm | = 2 a2m + b2m ; θm = arctan( −b
1
am
m
).
For real signals (as we will always be considering in this lecture):
• the magnitude of the Fourier coefficients is an even function of
the index m,
• the argument (or phase) of the coefficients is an odd function of
index m.

Introduction Real F-S Complex F-S Symmetry

. 55 / 66
3-4 Symmetry Properties of the FS (3)

Example (3-4) (extended again) Once again, remember Example 3-4 where
we showed that the trigonometric Fourier series of a square wave (amplitude A) consist of
4A
am ’s which are all zero, and bm ’s which are given by mπ for odd m and 0 for even m.
For positive frequencies (Xm ), the magnitude of the complex FS coefficient Xm , |Xm | then
√ 2
equals 2 am + b2m = 12 bm . The phase of Xm , θm equals arctan(− abm
1
m
) = arctan(−∞)
= − π2 .
For negative frequencies (X−m ) we get for the magnitude |X−m | = |Xm | = 12 bm , and for
the phase θ−m = −θm = π2 .
The magnitude and phase of the complex FS coefficients are shown in the next slide, for all
(i.e., positive and negative frequencies).
Try to imagine for yourself why this figure looks like it does. Imagine the two phasors
rotating counter-clockwise (positive frequencies) and clockwise (negative frequencies). We
need them both to construct the real signal.
The magnitude is the ‘size’ of the phasor, a complex vector, i.e., the vector length. The
phase is the initial angle of the vector for t=0.

Introduction Real F-S Complex F-S Symmetry

. 56 / 66
3-5 Symmetry Properties of the FS (4)

Example (3-4) (extended)

Magnitude of the first 51 complex FS coefficients
1

0.5
|Xm |

−0.5
−50 −40 −30 −20 −10 0 10 20 30 40 50
m
Phase of the first 51 complex FS coefficients
2

1
θm

−1

−2
−50 −40 −30 −20 −10 0 10 20 30 40 50
m
Introduction Real F-S Complex F-S Symmetry

. 57 / 66
3-5 Symmetry Properties of the FS (5)

Example Consider a signal (period T0 = 2 s, so f0 = 0.5 Hz) that is composed of the

summation of three cosines and one sine:
x(t) = 1. cos(2πf0 t) + 2. cos(2π2f0 t) + 1. cos(2π3f0 t) + 1. sin(2πf0 t)
The real and complex Fourier coefficients are shown in the figure.
5 3 3

2 2
x(t)

am

bm
0
1 1

−5 0 0
0 1 2 0 1 2 3 4 0 1 2 3 4
t m m
5 3
1

Xm
2
|Xm |
x(t)

0 0

6
1
−1
−5 0
0 1 2 −4 −3 −2 −1 0 1 2 3 4 −4 −3 −2 −1 0 1 2 3 4
t m m

Introduction Real F-S Complex F-S Symmetry

. 58 / 66
3-5 Symmetry Properties of the FS (6)

Now, consider a signal that is even, i.e.:

x(−t) = x(t).
Then in (3-44):
Z Z
1 j
Xm = x(t) cos(mω0 t)dt − x(t) sin(mω0 t)dt
T 0 T0 T 0 T0
| {z }| {z }
real part imaginary part

we can see that the imaginary part will be zero since it is an odd
function that integrates to zero over an interval symmetrically placed
about t = 0.
Hence, for an even function all bm ’s are zero and “we only use the
cosine components to describe the signal”. The Xm ’s are all real.
Introduction Real F-S Complex F-S Symmetry

. 59 / 66
3-5 Symmetry Properties of the FS (7)

Now, consider a signal that is odd, i.e.:

x(−t) = −x(t).
Then in (3-44):
Z Z
1 j
Xm = x(t) cos(mω0 t)dt − x(t) sin(mω0 t)dt
T 0 T0 T 0 T0
| {z }| {z }
real part imaginary part

we can see that the real part will be zero since it is an odd function
that integrates to zero over an interval symmetrically placed about
t = 0.
For an odd function all am ’s are zero and “we only use the sine
components to describe the signal”. The Xm ’s are all imaginary.
Introduction Real F-S Complex F-S Symmetry

. 60 / 66
3-5 Symmetry Properties of the FS (8)

Another type of symmetry is half-wave odd symmetry. Its counterpart, half-wave

even symmetry, does not exist. Both are illustrated in the figure below.

T0

... ...
x(t′ )
A

t
T
x(t + 20 )
′

T0 /2

T0 (= 2T0∗ )

... T ...
x(t′ ) x(t′ + 20 )
A

T0 /2
Introduction Real F-S Complex F-S Symmetry

. 61 / 66
3-5 Symmetry Properties of the FS (9)

For the half-wave odd symmetry:

 
x t ± T20 = −x(t), (3-50)

it turns out that:

Xn = 0 for n = 0, ±2, ±4, . . . (3-51)
Strict mathematically speaking, odd and halfwave odd signals must be zero mean.
The oscillation around the average, however, can still be considered (halfwave)
odd, which indeed shows up in the FS. For even signals, the average does not
matter.
The average only affects the a0 (of X0 ) coefficient in the FS!
Note that half-wave even symmetry does not exist, it means that the
fundamental period is T0 /2 (T0∗ in the figure above) rather than T0 !

Introduction Real F-S Complex F-S Symmetry

. 62 / 66
3-5 Symmetry Properties of the FS (10)

The most important properties of the Fourier series are summarized in Table 3-2.
Study this table well, and check whether you can prove the properties yourself.
Yet, one other property of the Fourier series coefficients that may be useful later, is
more or less ‘hidden’ in Problem 3-7.
Suppose that differentiation of a periodic signal x(t) results in a signal that has a
Fourier series Xn′ . Then it can be shown that:
Xn′ = (jnω0 )Xn
This property can be generalized as:

dk x(t) P∞
k = (jnω0 )k Xn ejnω0 t (*)
dt n=−∞

In other words, the k th derivative of a periodical signal x(t) can be expressed as a

Fourier series using (*).
Introduction Real F-S Complex F-S Symmetry

. 63 / 66
3-5 Symmetry Properties of the FS (11)

Proof Use the partial integration rule:

Rb Rb
a
(df /dx)g(x)dx = f (x)g(x)|ba − a
f (x)(dg/dx)dx
So:
−jnω0 t
T0 Xn′ =
R dx −jnω0 t
e dt
= x(t)e−jnω0 t |T0 0
− T0 x(t) deR
dt
T0 dt R dt
= (x(T0 )e−jnω0 T0 )
− (x(0) · 1) − T0 x(t)(−jnω0 )e−jnω0 t dt
Z
−jnω0 t
= x(T0 ) |e−jn2π
{z } −x(0) + (jnω 0) x(t)e dt
T0
=1 ∀ n | {z }
=T0 Xn
= x(T0 ) − x(0) +(jnω0 )T0 Xn
| {z }
=0, periodic
⇔ Xn′ = (jnω0 )Xn qed

The proof can be easily extended to higher order derivatives, e.g., by stating that when
signal g(t) = dx(t)/dt has Fourier coefficient Gn equal to Xn′ then use the same procedure
to show that the Fourier coefficient of dg(t)/dt equals G′n = (jnω0 )Gn . Etcetera etcetera
Introduction Real F-S Complex F-S Symmetry

. 64 / 66
Recommended home work
• Study 3-1, 3-2, 3-3, 3-4, 3-5
• Read the Summary points 3-1 until 3-8
• Study Table 3-2 carefully
• Derive the equations yourself, get a grip on the mathematics!
• Make Exercises 3-3, 3-4, 3-14, 3-16, 3-17, 3-18
• Use the Matlab GUI demonstration (BrightSpace)
• Look at the old exams!

Introduction Real F-S Complex F-S Symmetry

. 65 / 66
Bonus slide
Example 3-7 (continued) We found in Ziemer that:

−jXn : n = ±1, ±5, . . .
Yn = Xn e−jnπ/2 = (3-43)
jXn : n = ±3, ±7, . . .
with Xn the Fourier coefficients of the even square wave (Table 3-1, entry 4), which are all
real and zero for even values of n. Then Yn are the Fourier coefficients of the odd square
wave, which are all imaginary.
The right answer, however, 
is that:
−jXn : n = . . . , −7, −3, 1, 5, 9, . . .
Yn = Xn e−jnπ/2 =
jXn : n = . . . , −9, −5, −1, 3, 7, . . .
Basically, what you get for the odd square wave coefficients (see Example 3-4 (extended)):
 2A : n odd and n > 0

 −j nπ
Yn = +j 2A : n odd and n < 0

 |n|π
0 : even n
or: Yn = −j 2A
nπ for all odd n, and zero for all other n.

Introduction Real F-S Complex F-S Symmetry

. 66 / 66