Chapter 10

FOURIER TRANSFORMS

Fourier transforms express a given aperiodic function as a linear combination

of complex exponential functions. Unlike the Laplace transform, the function
is not restricted to be zero for negative times.
The main advantages of the Fourier transform are similar to those of the
Fourier series, namely (a) analysis of the transform is much easier than analysis
of the original function, and, (b) the transform allows us to view the signal in the
frequency domain. Unlike the Fourier series, since the function is aperiodic,
there is no fundamental frequency. The frequency spectrum of the signal will
contain continuous frequencies, not just multiples of a fundamental frequency.
The Fourier transform is most useful in characterizing the system that pro-
duces an output signal from an input signal. The system is usually defined by a
differential equation, and is never periodic. However, because of the approxi-
mation properties of the Fourier series, the input signals can be represented by
sums of periodic signals. The combination of Fourier transforms and Fourier
series is extremely powerful.

1. Introduction
1.1 Definition of the transform and spectrum
Definition: Consider a signal v(t), where t ∈ (−∞, ∞). Let ω be a real number.
The Fourier transform, V (ω), of the signal v(t) is defined by the integral
Z ∞
V (ω) = v(t)e−jωt dt (10.1)
−∞

4
This integral exists whenever

515
516 MATHEMATICS: THE LANGUAGE OF ECE

Z ∞
|v(t)|dt < ∞
−∞

Like Fourier series, evaluation of the Fourier transform in Equation 10.1

can be done by direct integration or (in a much easier fashion) by using the
properties of the transform (see Section 3).
Given the complex-valued function V (ω), the function v(t) can be found via
the inverse Fourier transform:
Z ∞
4 1
v(t) = V (ω)ejωt dω. (10.2)
2π −∞

Note that the variable of integration is real, unlike the case of the inverse
Laplace transform. We have seen how to calculate such integrals already, in
Chapter 9.
The Fourier transform can be defined to directly represent frequency in Hertz
by the transformation pair

Z ∞
V (F ) = v(t)e−j2πF t dt,
−∞
Z ∞
v(t) = V (F )ej2πF t dF.
−∞

We will tend to use radial frequency ω, for computations, since it is more

compact.
The Fourier transform, V (ω), of a signal v(t), is a complex-valued function
of the real variable ω (see Chapter 4, Section 6.1). As with the Fourier series
case, for any fixed value of ω, the complex number V (ω) can be represented via
its real and imaginary parts, or, as it is more usual in practice, via its magnitude
and phase.
Definition: The real-valued functions |V (ω)| and 6 V (ω) are called the mag-
nitude spectrum and phase spectrum of the signal v(t), respectively. Together,
these two functions are called the frequency spectrum (or simply the spectrum)
of v(t).
4
Since the parameter ω takes both positive and negative values, the plot of
the spectrum should cover both ranges. However, if the signal v(t) is real-
valued, which is almost always the case in our applications, for a fixed ω, the
complex numbers V (ω) and V (−ω) are conjugates of each other; the proof of
this property is left as an exercise. As with the Fourier series case, we have that

|V (−ω)| = |V (ω)|, 6 V (−ω) = −6 V (ω)

Fourier transforms 517

i.e., the magnitude spectrum is an even function of ω and the phase spectrum
is an odd function of ω. In the sequel, when we plot the spectrum, most of the
time we will only show positive values of ω.
The power spectrum of an aperiodic signal is defined in a manner analogous
to that of a periodic one in Chapter 9, Section 3.3.
Definition: The power spectrum, Ps (ω), of a signal s(t) with Fourier transform
S(ω), is defined as

Ps (ω) = |S(ω)|2 , ω ∈ (−∞, ∞) (10.3)

As with the Fourier series case (see Equation 9.72, page 478), the power
spectrum can also be expressed in decibels:

Ps (ω) = 20 log(|S(ω)|) (dB), ω ∈ (−∞, ∞) (10.4)

1.2 Relationship to Laplace transform and Fourier series

The Fourier transform is related to both the Laplace transform and Fourier
series.

1.2.1 Relationship to Laplace transform

Note the similarity of definition 10.1 to the Laplace transform. If v(t) = 0
for t < 0, the Laplace transform Lv (s) is also defined1 . In this case,
Z ∞ Z ∞
4
V (ω) = v(t)e−jωt dt = v(t)e−jωt = Lv (jω)
−∞ 0

Therefore,

V (ω) = Lv (s)|s=jω (10.5)

Equation 10.5 says that the Fourier transform can be found from the Laplace
transform by the substitution s = jω. Inversely, the Laplace transform can be
found from the Fourier transform by the substitution ω = s/j.

Example 10.1. The Laplace transform of the function v(t) = eat u(t) was
found to be

1 In
Chapter 8, we denoted the Laplace transform of v(t) as V (s). We change the notation here to avoid
confusion, since we use V (ω) to denote the Fourier transform of v(t).
518 MATHEMATICS: THE LANGUAGE OF ECE

1
Lv (s) =
s−a
(see Table 1, page 404 in Chapter 8). Therefore, its Fourier transform is
1
V (ω) =
jω − a
————————————————————————————- 4

1.2.2 Relationship to Fourier series

The Fourier series coefficients of a periodic signal, sp (t), with period T , are
determined by
Z
1 T /2 2π
Sp (k) = sp (t)e−j T kt dt
T −T /2
Consider an aperiodic signal s(t) that is equal to zero outside an interval
(−T /2, T /2] (and takes arbitrary values inside (−T /2, T /2]). Let S(ω) denote
the Fourier transform of s(t). Define the periodic extension of s(t) as the
periodic signal sp (t), whose period is equal to T and

sp (t) = s(t), −T /2 < t ≤ T /2

The Fourier series coefficients Sp (k) can be found from the Fourier transform
S(ω), as follows:

Z T /2
1 2π
Sp (k) = sp (t)e−j T kt
dt
T −T /2
Z
1 T /2 2π
= s(t)e−j T kt dt
T −T /2
Z
1 ∞ 2π
= s(t)e−j T kt dt
T −∞
µ ¶
1 2πk
= S (10.6)
T T
Therefore,
µ ¶
1 2πk
Sp (k) = S (10.7)
T T

Equation 10.7 says that the Fourier series Sp (k) can be found from the Fourier
transform S(ω) by dividing it by T and substituting ω = 2πk T . This is basically
Fourier transforms 519

sampling the Fourier transform at an interval of ω0 = 2π/T . As you will learn

in later courses, it is possible to reconstruct a signal from samples only under
special conditions. We cannot, in general, go from the Fourier series to the
Fourier transform by the inverse substitution k = T ω/2π.

2. Fourier transforms of common signals

Let’s see now how we can calculate the Fourier transform of some common
signals.

Example 10.2. Let

v(t) = δ(t − t0 )
where t0 is a given real number. We have

Z ∞
V (ω) = δ(t − t0 )e−jωt dt
−∞
−jωt0
= e (10.8)

Example 10.3. Let

V (ω) = πδ(ω − ω0 ) + πδ(ω + ω0 ). (10.9)

Then

Z
1 ∞
v(t) = V (ω)ejωt dω
2π −∞
Z Z
1 ∞ jωt 1 ∞
= δ(ω − ω0 )e dω + δ(ω + ω0 )ejωt dω
2 −∞ 2 −∞
1 jω0 t 1 −jω0 t
= e + e = cos(ω0 t)
2 2
From the uniqueness property (discussed in Section 3), we recognize V (ω)
as the Fourier transform of the (periodic) cos(ω0 t) signal.
————————————————————————————- 4

Example 10.4. Consider the centered unit pulse of duration τ defined by

½
1, −τ /2 ≤ t ≤ τ /2,
pτ (t) = (10.10)
0, otherwise.

We have
520 MATHEMATICS: THE LANGUAGE OF ECE

Z ∞
Pτ (ω) = pτ (t)e−jωt dt
−∞
Z τ /2
= e−jωt dt
−τ /2
1 ¯τ /2
¯
= e−jωt ¯
−jω −τ /2
1 ³ −jωτ /2 ´
= e − e−jω(−τ /2)
−jω
1 ³ jωτ /2 ´
= e − e−jωτ /2
jω
2 sin(ωτ /2)
= sin(ωτ /2) = τ (10.11)
ω ωτ /2
So, finally,

sin(ωτ /2)
Pτ (ω) = τ (10.12)
ωτ /2

A plot of the spectrum of Pτ (ω) in Equation 10.12, for τ = 2 is given in

Figure 10.1. Note that the plot is versus frequency F in Hertz, not angular
frequency ω. Only positive values of f are shown in the spectrum. Note that
the zeros in the magnitude spectrum appear at the frequencies F for which
sin(ωτ /2) = 0. Therefore, the magnitude spectrum is equal to zero for the
frequencies
1 2 3
F = , , ,···,
τ τ τ
where τ is the width of the pulse.
————————————————————————————- 4

Example 10.5. Consider the signal

½
t, t ∈ [−T /2, T /2)
s(t) = (10.13)
0, otherwise.

where T > 0 is a known parameter. This signal is equal to one period of the
periodic signal in Example 9.11.
The Fourier transform of s(t) can be found as follows:

Z ∞
S(ω) = s(t)e−jωt dt
−∞
Fourier transforms 521

0.8
Magnitude

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

3.5

2.5
Phase

1.5

0.5

0
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

Figure 10.1. The spectrum of the signal in Equation 10.12.

Z T /2
= te−jωt dt
−T /2
1 ¯T /2
−jωt ¯
= 2
e ((−jω)t − 1) ¯
(−jω) −T /2
1 −jωt ¯
¯T /2
= 2
e (jωt + 1)¯
ω −T /2
µ ¶ µ ¶
1 −jω T T 1 jω T T
= e 2 jω + 1 − 2 e 2 −jω + 1
ω2 2 ω 2
µ ¶h i h i
1 T T
−jω 2 T
jω 2 1 −jω T2 jω T2
= jω e + e + e − e
ω2 2 ω2
jT cos(ωT /2) 2j sin(ωT /2)
= − (10.14)
ω ω2
The plot of this spectrum for T = 2 is shown in Figure 10.2. Note again that
the plot is versus frequency F in Hertz, not angular frequency ω.
522 MATHEMATICS: THE LANGUAGE OF ECE

0.8
Magnitude

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

1
Phase

−1

−2
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

Figure 10.2. The spectrum of the signal in Equation 10.13.

————————————————————————————- 4

*Example 10.6. The Fourier series of the periodic ramp, one period of which
is defined as
sp (t) = t, t ∈ [−T /2, T /2)
can be evaluated now using Equation 10.7. We have, from Equation 10.14:
· ¸¯
1 jT cos(ωT /2) 2j sin(ωT /2) ¯¯
Sp (k) = − ¯ 2πk
T ω ω2 ω= T
 ³ ´ ³ ´
2πk T 2πk T
1  jT cos T 2 2j sin T 2 
=  2πk
− ³ ´2 
T T
2πk
T
jT cos(πk)
=
2πk
Fourier transforms 523

which agrees with Equation 9.57, page 466.

If we attempt to recover the transform from the series, we would start with
Equation 9.57. The substitution k = T ω/2π will give
j
Sp (k) = cos(kπ).
kω0
We could derive Equation 10.14 as

S(ω) = T Sp (k)|k= T ω
2π
¯
j ¯
= T· cos(kπ)¯¯
kω0 k= T2πω
µ ¶
Tω j
= T · T ω 2π cos π
2π T
2π
µ ¶
jT ωT
= cos
ω 2
and we see that we cannot recover the term with the sin(ωT /2) in Equation
10.14.
————————————————————————————- 4
Table 10.2 summarizes the Fourier transforms of some common signals.
Table 10.1. Fourier transforms of common signals.

v(t) V (ω)

δ(t) 1
1 1
sin(ω0 t) 2j
δ(ω − ω0 ) − 2j
δ(ω + ω0 )
1 1
cos(ω0 t) 2
δ(ω − ω0 ) + 2
δ(ω + ω0 )
at 1
e u(t) jω−a
at ω0
e sin(ω0 t)u(t) 2
(jω−a)2 +ω0
jω−a
eat cos(ω0 t)u(t) 2
(jω−a)2 +ω0

3. Properties
The Fourier transform, like the Laplace transform and Fourier series, pos-
sesses similar properties.

3.1 Uniqueness
For each given function v(t), there exists one and only one function V (ω),
defined via Equation 10.1. Inversely, given a function V (ω), there exists one
524 MATHEMATICS: THE LANGUAGE OF ECE

and only one function v(t), defined via Equation 10.2, that has the function
V (ω) as its Fourier transform.

3.2 Linearity
The following theorem is the analogue of theorems 8.1 and 9.1; it can be
proven in an entirely similar fashion:
Theorem 10.1 Consider given constants ci (real or complex valued) and
functions vi (t) with known Fourier transforms Vi (ω). Let
X
v(t) = ci vi (t) (10.15)
i

denote a linear combination of the functions vi (t). Then

X
V (ω) = ci Vi (ω) (10.16)
i

3.3 Time-shifting
Let t0 be a given real-valued constant, positive or negative (unlike the Laplace
transform). The analogue of theorems 8.2 and 9.2 is:
Theorem 10.2 Consider a signal v(t), with known Fourier transform V (ω).
Let y(t) = v(t − t0 ), where t0 is a given real-valued constant. The Fourier
transform of the time-shifted signal is given by

Y (ω) = e−jωt0 V (ω) (10.17)

Proof: We can write

Z ∞ Z ∞
4 −jωt
Y (ω) = y(t)e dt = v(t − t0 )e−jωt dt
−∞ −∞
Z ∞
= v(t − t0 )e−jωt0 ejωt0 e−jωt dt
−∞
Z ∞
−jωt0
= e v(t − t0 )e−jω(t−t0 ) dt
−∞
Z ∞
= e−jωt0 v(t)e−jωt dt = e−jωt0 V (ω)
−∞
Fourier transforms 525

Example 10.7. Let

y(t) = cos(ω0 t + θ) (10.18)

be a cosine signal with a phase shift. Find Y (ω), the Fourier transform of y(t).
4
Let v(t) = cos(ω0 t). In Example 10.3, we have already calculated the
Fourier transform V (ω). Using Theorem 10.2, we can calculate Y (ω) as fol-
lows. Rewrite Equation 10.18 as

y(t) = cos(ω0 (t + θ/ω0 ))

The time-shift is then equal to

t0 = −θ/ω0
From Equation 10.17 and Equation 10.9, page 519, we can now write

Y (ω) = e−jωt0 V (ω)

= ejωθ/ω0 [πδ(ω − ω0 ) + πδ(ω + ω0 )]
= ejθ πδ(ω − ω0 ) + e−jθ πδ(ω + ω0 ) (10.19)
In deriving Equation 10.19, we have used the property of the δ(t) function
in Equation 2.17, page 61.
————————————————————————————- 4

3.4 Time-reversal
Consider a signal v(t), with known Fourier transform V (ω).
Theorem 10.3 Let y(t) = v(−t), be the time-reversed version of v(t). The
Fourier transform of the time-reversed signal is given by

Y (ω) = V (−ω) (10.20)

Proof: We have by definition

Z ∞
4
Y (ω) = y(t)e−jωt dt
−∞
Z ∞
= v(−t)e−jωt dt
−∞
526 MATHEMATICS: THE LANGUAGE OF ECE
Z −∞
= v(u)e−jω(−u) (−du) (10.21)
∞
Z −∞
= − v(u)e−j(−ω)u du
∞
Z ∞
= v(u)e−j(−ω)u du = (10.22)
−∞
= V (−ω)
4

3.5 Time-scaling
Consider a signal v(t), with known Fourier transform V (ω). Let y(t) =
v(at), where a 6= 0 is a given constant. The Fourier transform can be calculated
as follows. Consider first the case a > 0. We have by definition
Z ∞
4
Y (ω) = y(t)e−jωt dt
−∞
Z ∞
= v(at)e−jωt dt
−∞
Z ∞ Z ∞
u du 1 ω
= v(u)e−jω a = v(u)e−j a u du (10.23)
−∞ a a −∞
1
= V (ω/a) (10.24)
a
In Equation 10.23, we have made the substitution of variables u = at.
The case where a < 0 can be derived from Equations 10.20 and 10.24, by
observing that we can write a = −|a|. Then the signal

y1 (t) = v(|a|t)
1
has Fourier transform Y1 (ω) = |a| V (ω/|a|), and the signal

y(t) = y1 (−t)
1 1
has Fourier transform Y (ω) = |a| V (−ω/|a|) = |a| V (ω/a). Therefore,

Theorem 10.4 Consider a signal v(t), with known Fourier transform V (ω).
Let y(t) = v(at), where a 6= 0 is a given constant. The Fourier transform is
given by
1
Y (ω) = V (ω/a) (10.25)
|a|
4
Fourier transforms 527

3.6 Derivatives
The Fourier transform of the n-th derivative of a signal can also be related
to the Fourier transform of the signal itself. The analogue of theorems 9.5 and
8.4 is the following.

Theorem 10.5 Consider a signal v(t) with known Fourier transform V (ω).
The Fourier transform of yk (t) = dk v(t)/dtk , the k-th order derivative of the
signal v(t), is given by

Yk (ω) = (jω)k V (ω) (10.26)

4

3.7 Examples

*Example 10.8.
Determine the Fourier transform of the signal s(t) in Equation 10.27



0, t < −1,
 −1, −1 ≤ t < 0,
s(t) = (10.27)

 2, 0 ≤ t ≤ 2,

0, t > 2.
using (a) the direct integration approach and (b) the properties of the Fourier
transform. The signal is shown in Figure 10.3.
Note that when applying the properties, we can determine this transform:
1 using the linearity/time-shifting properties and the transform of (two) cen-
tered pulses pτ (t) (of width τ ):
½
1, −τ /2 ≤ t ≤ τ /2,
pτ (t) = (10.28)
0, otherwise.

One pulse, p1 (t), of width 1, would cover the region −1 ≤ t < 0; the other,
p2 (t), of width 2, would cover the region 0 ≤ t ≤ 2.
2 using the linearity/time-scaling properties and the transform of one cen-
tered pulse p(t) = p1 (t) with unit width. The transform of the pulses was
calculated in Example 10.4.
Approach 1. From Example 10.4, we have

sin(ωτ /2)
Pτ (ω) = τ (10.29)
ωτ /2
528 MATHEMATICS: THE LANGUAGE OF ECE

The signal s(t)

3

2.5

1.5

1
s(t)

0.5

−0.5

−1

−1.5

−2
−5 −4 −3 −2 −1 0 1 2 3 4 5
time t

Figure 10.3. The signal s(t) in Example 10.8.

We can write the signal in Equation 10.27 using the two simpler signals in
Equation 10.28 as follows:

s(t) = −p1 (t + 0.5) + 2p2 (t − 1) (10.30)

Equation 10.30 simply says that the signal s(t) consists of two pulses: one
of width τ = 1, shifted to the left by 0.5 time units and another of width τ = 2,
shifted to the right by 1 time unit. A decomposition like this would avoid using
a time scaling property along with time shifting.
Therefore (applying the linearity and time-shifting properties of the Fourier
transform) and using Equation 10.29:

S(ω) = −ej0.5ω P1 (ω) + 2e−jω P2 (ω)

Fourier transforms 529

sin(ω · 1/2) sin(ω · 2/2)

= −ej0.5ω 1 · + 2e−jω 2
ω · 1/2 ω · 2/2
sin(ω/2) sin(ω)
= −ej0.5ω + 4e−jω (10.31)
ω/2 ω
Approach 2. A decomposition that uses time scaling along with time shifting
can be obtained by decomposing the signal in terms of p1 (t) only. The time
scaling

y(t) = p1 (t/2)
stretches the unit pulse from −1 to +1. We can then write

s(t) = −p1 (t + 0.5) + 2y(t − 1) = −p1 (t + 0.5) + 2p1 (1/2(t − 1)) (10.32)

Then

S(ω) = −ej0.5ω P1 (ω) + 2e−jω Y (ω) (10.33)

where

sin(ω/2)
P1 (ω) = (10.34)
ω/2
and from the time-scaling property

sin(ω)
Y (ω) = 2P1 (2ω) = 2 (10.35)
ω

Combining Equations 10.33, 10.34 and 10.35 we get

sin(ω/2) sin(ω)
S(ω) = −ej0.5ω + 2e−jω 2 (10.36)
ω/2 ω
which (of course) agrees with the function S(ω) we derived in Equation 10.31.
Approach 3. From its definition, we have

Z ∞ Z 0 Z 2
−jωt −jωt
S(ω) = s(t)e dt = (−1)e dt + (2)e−jωt dt
−∞ −1 0
Z 0 Z 2
= − e−jωt dt + 2 e−jωt
dt
−1 0
¯0 ¯2
e−jωt ¯¯ e−jωt ¯¯
= − ¯ +2 ¯
−jω ¯−1 −jω ¯0
530 MATHEMATICS: THE LANGUAGE OF ECE

1 ¯0 1 −jωt ¯¯2
¯
= e−jωt ¯ −2 e ¯
jω −1 jω 0
1 h i h
1 −j2ω i
= 1 − ejω − 2 e −1
jω jω
1 h i
= 1 − ejω + 2 − 2e−j2ω
jω
1 h i
= 3 − ejω − 2e−j2ω (10.37)
jω
Equations 10.37 and 10.31 look different. However, using Euler’s identities
in Equation 10.31, we get

1 jω/2 1 jω
j0.5ω 2j
[e − e−jω/2 ] −jω 2j
[e − e−jω ]
S(ω) = −e + 4e
ω/2 ω
1 h ³ ´i
= −ej0.5ω [ejω/2 − e−jω/2 ] + 2e−jω ejω − e−jω
jω
1 h i
= −ejω + 1 + 2 − 2e−j2ω
jω
1 h i
= 3 − ejω − 2e−j2ω (10.38)
jω
which is the same as Equation 10.37.
Figure 10.4 depicts the spectrum of this signal. Note that both negative and
positive frequencies are shown. From the graph, the zeros of the magnitude
spectrum occur at frequencies

F = 1, 2, 3, 4 Hz.

Can you prove that the frequency F = k, where k is an integer, will also yield
a zero, by using Equation 10.37?
————————————————————————————- 4

Example 10.9. The Fourier transform of a signal s(t) is given in Equation

10.39:
sin(ω)
S(ω) = e−j2ω (10.39)
ω

Write a Matlab script to plot the spectrum of the signal for a frequency range
of 0–10 Hz, with a resolution of 0.001 Hz.
The spectrum consists of two plots, one for the magnitude of S(ω) and one
for the phase of S(ω). A Matlab script to plot the spectrum is the following:
Fourier transforms 531

Magnitude spectrum
4

3
Magnitude

0
−4 −3 −2 −1 0 1 2 3 4
Frequency (Hz)

Phase spectrum
4

2
Phase

−2

−4
−4 −3 −2 −1 0 1 2 3 4
Frequency (Hz)

£ ¤
Figure 10.4. The spectrum of S(ω) = 1
jω
3 − ejω − 2e−j2ω .

Matlab script 10.1. ————————————————————-

% Define the desired frequency range;
%resolution of the range is 10 Hz
freq_range = 0:0.001:10;
omega_range = 2 * pi * freq_range;

% Calculate the Fourier transform S(omega)

% Note the .* and ./ use for element-wise operations
S = exp( -j * 2* omega_range ) .* sin( omega_range ) ...
./ omega_range ;

% Of course, Matlab will complain about

% the sin(0)/0 division,
% which occurs for omega_range(1).
% From De L’Hospital’s rule, we have the following:
532 MATHEMATICS: THE LANGUAGE OF ECE

S(1) = 1 ;

% Calculate the magnitude and phase

magnitude = abs( S );
phase = angle( S );

% Now plot the spectrum, using subplot and grid

subplot(2,1,1)
plot( freq_range, magnitude )
xlabel(’frequency (Hz)’)
ylabel(’Magnitude’)
grid on

subplot(2,1,2)
plot( freq_range, phase )
xlabel(’frequency (Hz)’)
ylabel(’Phase’)
grid on
——————————————————————————–
The spectrum produced by this script is shown in Figure 10.5. Note that
the horizontal axis is frequency in Hz, not in radians. When comparing this
spectrum to the one shown in Figure 10.1, notice that the magnitude parts are
equal. The phase parts differ by a factor of 2ω, as you can easily explain by
using the time-shifting property.
————————————————————————————- 4

Example 10.10. For the Fourier transform of the previous example, evaluate
the value of S(ω) for ω = 0 and ω = π/2. Express your results in exponential
form. 4
We have

sin(0)
S(0) = e−j2·0 =1
0
sin(π/2) 1 2
S(π/2) = e−j2π/2 = 2e−jπ = e−jπ
π/2 π π
————————————————————————————- 4

*Example 10.11. Find the Fourier transform of the signal, v(t), in Figure
10.6, by using time scaled and shifted versions of the centered unit pulse signal
p1 (t) in Equation 10.10, page 519, Example 10.4. 4
Fourier transforms 533

0.8
Magnitude

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

2
Phase

−2

−4
0 1 2 3 4 5 6 7 8 9 10
frequency (Hz)

Figure 10.5. The spectrum of S(ω) = e−j2ω sin(ω)

ω
.

Set τ = 1 in Example 10.4. The signal p1 (t) is given by


 0, t < −1/2
p1 (t) = 1, −1/2 ≤ t < 1/2

0, 1/2 ≤ t
with Fourier transform given by (see Equation 10.12 in Example 10.4):
sin(ω/2)
P1 (ω) =
ω/2
Let’s define two new signals, that are time-scaled versions of p1 (t):

y1 (t) = p1 (t/2)
y2 (t) = p1 (t/6)
y1 (t) is a centered pulse of width 2 and y2 (t) is a centered pulse of width 6.
Then, from the time-scaling property of the Fourier transform we obtain
534 MATHEMATICS: THE LANGUAGE OF ECE

The signal v(t)

0
v(t)

−1

−2
−5 −4 −3 −2 −1 0 1 2 3 4 5
t

Figure 10.6. The signal v(t) in Example 10.11.

sin(2ω/2) sin(ω)
Y1 (ω) = 2P1 (2ω) = 2 =2 (10.40)
2ω/2 ω
sin(6ω/2) sin(3ω)
Y2 (ω) = 6P1 (6ω) = 6 =6 (10.41)
6ω/2 3ω
Note that we keep the transform in the form of the sinc function sin(x)/x.
From Figure 10.6, we can write

v(t) = y2 (t + 1) − 2y1 (t − 3)
from which and the time-shifting property of the Fourier transform we get:

V (ω) = ejω Y2 (ω) − 2e−3jω Y1 (ω)

Now, from Equations 10.40 and 10.41 we have
sin(3ω) sin(ω)
V (ω) = ejω 6 − 2e−3jω 2
3ω ω
So, finally,

sin(3ω) sin(ω)
V (ω) = 6ejω − 4e−3jω (10.42)
3ω ω

Figure 10.7 depicts the spectrum of V (ω). (It looks almost artistic.) Note
that the magnitude spectrum is zero for angular frequencies ω = π, 2π, 3π, . . .,
Fourier transforms 535

or for frequencies f = 0.5, 1, 1.5, 2, . . . Note also how these frequencies are
related to the width (2) of the pulse: they are multiples of 1/pulse width. (This
relationship will be important in later classes.)

Magnitude spectrum
8

6
Magnitude

0
−1.5 −1 −0.5 0 0.5 1 1.5
Frequency f (Hz)

Phase spectrum
4

2
Phase

−2

−4
−1.5 −1 −0.5 0 0.5 1 1.5
Frequency f (Hz)

Figure 10.7. The spectrum of the transform in Equation 10.42.

————————————————————————————- 4

*Example 10.12. Find the Fourier transform of the signal, v(t), in Figure
10.6, by using time scaled and shifted versions of the centered unit pulse signal
p2 (t) in Equation 10.10, example 10.4. 4
Inspecting Figure 10.6, we can write

v(t) = p2 (t + 3) + p2 (t + 1) + p2 (t − 1) − 2p2 (t − 3)

The Fourier transform of the centered unit pulse signal p2 (t) is given by
sin(ω)
P2 (ω) = 2 .
ω
Then, from the linearity and time-shifting properties of the Fourier transform,
we can write

³ ´
V (ω) = ej2π3ω + ej2πω + e−j2πω − 2e−j2π3ω P2 (ω)
536 MATHEMATICS: THE LANGUAGE OF ECE

³ ´ sin(ω)
= ej2π3ω + ej2πω + e−j2πω − 2e−j2π3ω 2
ω
(10.43)

Equations 10.43 and 10.42 appear to be different. Can you reconcile them?
————————————————————————————- 4

4. Solving differential equations with Fourier transforms

As with the Fourier series, the most general differential equation that we will
consider in this section is Equation 9.103, page 494, in Chapter 9, i.e., a linear,
k-th order differential equation with constant coefficients, of the type:

dk v(t) dk−1 v(t) dv(t)

k
+ ak−1 k−1
+ · · · + a1 + a0 v(t) = vs (t) (10.44)
dt dt dt

Note that the driving force does not have to be a periodic function. We
are interested again in finding vp (t), the particular solution of the equation.
Therefore, throughout this section we will assume that the initial conditions are
0.
As with the Laplace transform approach, the solution methodology is to:

calculate the Fourier transform of both sides of Equation 10.44,

use Theorem 10.5, to reduce Equation 10.44 into an algebraic equation that
involves Vp (ω), the Fourier transform of the particular solution.
determine Vp (ω) from this algebraic equation
find vp (t) using Equation 10.2 or properties of the Fourier transform.

Taking the Fourier transform of both sides of Equation 10.44, and applying
Theorem 10.5, we get

(jω)k Vp (ω) + ak−1 (jω)k−1 Vp (ω) + · · · + a1 (jω)Vp (ω) + a0 Vp (ω)

= Vs (ω) (10.45)

from which the Fourier transform Vp (ω) is:

1
Vp (ω) = · Vs (ω) (10.46)
(jω)k + ak−1 (jω)k−1 + · · · + a1 (jω) + a0
Fourier transforms 537

The particular solution can now be found as the following theorem summa-
rizes:

Theorem 10.6 Consider the differential equation 10.44. Let Vs (ω) denote
the Fourier transform of the driving force. Then the particular solution is given
by

Z ∞
Vs (ω)ejωt
vp (t) = dt (10.47)
−∞ (jω)k + ak−1 (jω)k−1 + · · · + a1 (jω) + a0

Note that we will rarely compute the inverse transform but will use the
properties of the transform and tables similar to the Laplace transform to obtain
a complete description of vp (t).
As with the Fourier series case,
Definition: The function
4 1
H(ω) = (10.48)
(jω)k + ak−1 (jω)k−1 + · · · + a1 (jω) + a0
is called the transfer function of the system represented by differential equation
10.44. Note that, in Equation 10.44, there are no derivatives of the driving
force, hence the numerator in Equation 10.48 is equal to 1. Compare Equations
10.48 and 8.58.
4

*Example 10.13. Consider the differential equation

d3 v(t) d2 v(t) dv(t)

3
+ 2 2
+3 + 4v(t) = 60 cos(2π100t). (10.49)
dt dt dt

Determine the particular solution, vp (t), using Fourier transforms. 4

Note that this is the same differential equation in Equation 9.113, page 498.
We have solved the equation already using Fourier series, see Equation 9.115,
page 499.
Let vs (t) = 60 cos(2π100t) denote the driving force in Equation 10.49.
From Example 10.3, the Fourier transform, Vs (ω), of the driving force is

Vs (ω) = 60πδ(ω − 2π100) + 60πδ(ω + 2π100). (10.50)

538 MATHEMATICS: THE LANGUAGE OF ECE

Set k = 3 in Equation 10.46. From Equations 10.49 and 10.50, the Fourier
transform of the particular solution is given by:

Vs (ω)
Vp (ω) =
(jω)3 + 2(jω)2 + 3(jω) + 4
60πδ(ω − 2π100) + 60πδ(ω + 2π100)
=
−jω 3 − 2ω 2 + 3jω + 4
60πδ(ω − 2π100) + 60πδ(ω + 2π100)
=
(4 − 2ω 2 ) + jω(3 − ω 2 )
60π
= δ(ω − 2π100)
[4 − 2(2π100) ] + j(2π100)[3 − (2π100)2 ]
2

60π
+ δ(ω + 2π100)
[4 − 2(−2π100) ] + j(−2π100)[3 − (−2π100)2 ]
2

(10.51)
60π
= δ(ω − 2π100)
[4 − 2(2π100)2 ] + j[6π100 − (2π100)3 ]
60π
+ δ(ω + 2π100)
[4 − 2(2π100) ] − j[6π100 − (2π100)3 ]
2

60π
= p δ(ω − 2π100)
[4 − 2(2π100) ] + [6π100 − (2π100)3 ]2 ejθ
2 2

60π
+p δ(ω + 2π100)
[4 − 2(2π100)2 ]2 + [6π100 − (2π100)3 ]2 e−jθ
(10.52)
In deriving Equation 10.51, we have used the property of the δ(t) function
in Equation 2.17, page 61.
From Equation 10.19, page 525, the Fourier transform Vp (ω) corresponds to
a cosine time signal, with amplitude equal to
60
p
[4 − 2(2π100) ] + [6π100 − (2π100)3 ]2
2 2

and phase
à !
(2π100)[3 − (2π100)2 ]
θ = arctan ≈ −π/2
4 − 2(2π100)2
Therefore,
60
vp (t) = p cos(2π100t − π/2)
[4 − 2(2π100)2 ]2+ [6π100 − (2π100)3 ]2
60
= p sin(2π100t)
[4 − 2(2π100)2 ]2 + [6π100 − (2π100)3 ]2
Fourier transforms 539

Example 10.14. Consider the differential Equation 9.108, page 495, in

Example 9.26. What methods are there to determine the particular solution
vp (t)? 4
The methods we have to find the particular solution in general, are:
1 Guess a solution of the form

vp (t) = B cos(100t + θ1 ) + C sin(500t + θ2 )

or

vp (t) = B cos(100t) + C sin(100t) + D cos(500t) + E sin(500t)

(see Chapter 7).

2 Apply the Laplace transform to both sides of Equation 9.108
3 Apply the Fourier transform to both sides of Equation 9.108.
4 Apply the Fourier series to both sides of Equation 9.108.
Option 1 will work. Option 2 is not applicable, since the driving force
vs (t) = 20 cos(100t) + 40 sin(200t) is not 0 for t < 0. Option 3 is applicable.
Option 4 is applicable, since the driving force vs (t) is periodic, with fundamental
frequency ω0 = 100.
————————————————————————————- 4

5. Matlab commands for calculating Fourier transforms

The command

fourier(f)

where f is a symbolic variable returns the Fourier transform of the function

f (t).

Example 10.15. The script

Matlab script 10.2. ————————————————————-

syms t
f = exp(-t^2);
fourier(f)
——————————————————————————–
√ 2
returns the expression πe−1/4∗ω (try to compute this by hand!).
————————————————————————————- 4
540 MATHEMATICS: THE LANGUAGE OF ECE

The command

ifourier(F)

calculates the function f (t) whose Fourier transform is equal to F (ω).

Example 10.16. The script

Matlab script 10.3. ————————————————————-

syms w
F = 1/(1+j*w) ;
ifourier(F)

——————————————————————————–
returns the expression e−t u(t).
————————————————————————————- 4

6. Summary of main points

Representation of an aperiodic signal with the Fourier transform.

Properties of the Fourier transform.

The frequency spectrum of an aperiodic signal.

Finding the particular solution of a differential equation using Fourier trans-

forms.

Skillset to be developed
Determine analytically the Fourier transform of a signal using the direct
integration approach.

Determine analytically the Fourier transform of a signal using the properties

of the Fourier transform and the Fourier transform of simpler signals.

Determine analytically the signal v(t) in the time domain, from knowledge
of its Fourier transform V (ω).

Write Matlab scripts to plot the frequency and power spectrum of a signal.

Determine analytically the particular solution of a differential equation using

Fourier transforms.
Fourier transforms 541

Connection to other ECE courses

You will use Fourier transforms ad nauseam in ECE301, ECE402, ECE403,
ECE420, ECE421, ECE422 and ECE451. In these courses, you will go deeper
and faster; in some of these courses, you will study Fast Fourier Transforms,
the celebrated FFT.
As was the case with Laplace transforms, the concept of the transfer function
and applications in filter design will be the major topics in such courses. In
the two communications-related courses, ECE402 and ECE420, the Fourier
transform will be indispensable in understanding the fundamental notion of
bandwidth.
542 MATHEMATICS: THE LANGUAGE OF ECE

Problems
Calculation of the Fourier transform

10.1. Using direct integration, find the Fourier transform of the signal

s(t) = e−25t cos(2π100t + π/4)u(t)

You can check your result by using the Laplace transform table in Chapter
8.

10.2. Using direct integration, find the Fourier transform of the signal

s(t) = 20e−t ejπt u(t)

*10.3. Show that the Fourier transform of the triangular pulse signal
½ |t|
s(t) = 1− T , −T ≤ t ≤ T ,
0, otherwise.
is given by
µ ¶2
sin(ωT /2)
S(ω) = T
ωT /2

10.4. Consider a signal s(t) with Fourier transform S(ω). Let ω0 be a real
number. Define a new (complex-valued) signal

y(t) = ejω0 t s(t).

Show that
Y (ω) = S(ω − ω0 )
————————————————————————————–
Properties of the Fourier transform

*10.5. Consider p(t), the unit pulse signal in Equation 2.12, page 59. Let
t0 = 10.
Determine P (ω) using time shifting and scaling properties, and the results
of Example 10.4.

*10.6. Repeat the previous problem for a generic t0 . (Hint: let τ = 1 in

Example 10.4. First time-shift p1 (t) by 0.5 time units to the left, to obtain a
signal y(t) = p1 (t − 0.5). Then time scale y(t).)
Fourier transforms 543

10.7. Consider the signal h1 (t) in Equation 2.39, page 80.

(a) Calculate its Fourier transform, H1 (ω), using direct integration.
(b) Calculate its Fourier transform, H1 (ω), using time shifting and scaling
properties, and the results of Example 10.4.

10.8. Consider the signal h3 (t) shown in Figure 2.16, page 81. Calculate its
Fourier transform.

10.9. Consider the signal h4 (t), defined in Equation 2.41, page 82. Calculate
its Fourier transform.

10.10. Consider the signal h5 (t), defined in Equation 2.43, page 82. Calcu-
late its Fourier transform.

10.11. Consider the centered unit pulse p5 (t), in Example 10.4. Calculate
the Fourier transform of the signal y(t) = 3p5 (t/2).

*10.12. Consider the centered unit pulse p5 (t), in Example 10.4. Calculate
the Fourier transform of the signal y(t) = 3p5 (t/2 + 3).

10.13. Consider the signal v(t), shown in Figure 8.3, page 428.
(a) Calculate the Laplace transform V (s).
(b) Calculate the Fourier transform, V (ω) from the Laplace transform V (s).

*10.14. The Fourier transform of the signal s(t) is defined as

sin(2ω) sin(ω)
S(ω) = 8ejω + 2e−jω
2ω ω
Find the Fourier transform of the time-scaled and shifted signal y(t) = s(2t−3).

*10.15. Using the formula for the Fourier transform of the centered unit pulse
in Equation 10.10, page 519, together with time shifting and scaling properties,
find the Fourier transform of the signal defined by


 0, t<0

2, 0≤t≤1
s(t) =

 −1, 1 < t ≤ 3

0, 3<t
————————————————————————————–
544 MATHEMATICS: THE LANGUAGE OF ECE

Fourier transform and Fourier series

*10.16. The Fourier series for the periodic signal, s1 (t), whose single period
is defined by ½
sin(2πt/T ), 0 ≤ t ≤ T /2
s1 (t) =
0, T /2 < t < T
where T = 1/60 seconds, is denoted by S1 (k).
(a) Let a single period of the periodic signal s2 (t), be defined by
½
sin(2πt/T ), 0 ≤ t ≤ T /2
s2 (t) =
0, T /2 < t < T

where T = 1/120 seconds. Find the Fourier series, S2 (k), in terms of S1 (k).
Hint: Use the relationship between the Fourier series and Fourier transform to
obtain both series and look for the similarities.
(b) Let a single period of the periodic signal s3 (t), be defined by

 cos(2πt/T ), −T /4 ≤ t ≤ T /4
s3 (t) = 0, −T /2 < t < −T /4

0, T /4 < t < T /2
where T = 1/120 seconds. Find the Fourier series, S3 (k), in terms of S1 (k).

*10.17. The rectified sinusoids are very important in many ECE applications,
particularly in communications and power conversion. It is important that you
understand the relationships among the various forms, half-wave and full-wave
rectified signals, in both the time and frequency domains.
(a) Compute, by direct integration, the Fourier transform of the function,
s1 (t), defined by

 0, t ≤ −T /4
s1 (t) = cos(2πt/T ), −T /4 ≤ t ≤ T /4

0, T /4 < t

(b) Assume T = 1/60; plot the magnitude and phase spectra from zero to
10KHz. Plot the power spectrum in decibels as a function of frequency in Hertz.
Recall that the power spectrum in dB is defined by PdB (ω) = 20 log10 (|S(ω)|).
(c) The periodic extension of s1 (t) with a period of T is written

 0, −T /2 < t ≤ −T /4
s1p (t) = cos(2πt/T ), −T /4 ≤ t ≤ T /4

0, T /4 < t ≤ T /2
Use the relationship of the Fourier transform and Fourier series to obtain the
Fourier series coefficients of s1p (t).
Fourier transforms 545

(d) Use the information found above to obtain the Fourier series coefficients
of s2p (t) = | cos(2πt/T )|. This is the full-wave rectified sinusoid. Pay partic-
ular attention to the fundamental frequencies of s1p (t) and s2p (t).
(e) Use the information found above and Fourier properties (time scaling and
shifting) to obtain the Fourier series coefficients of s3p (t) = | sin(2π10t/T )|.

————————————————————————————–

Frequency and power spectrum

*10.18. Let pτ (t) be the unit centered pulse of width τ , in Equation 10.10
of Example 10.4.
(a) Find the power spectrum of a clock generator signal in Equation 2.21,
page 63. For this problem, you will periodically extend the single pulse with
1
a period T = 100 seconds, and set τ = T /2. The power spectrum does not
depend on the phase of the signal, so you may use the symmetric pulse to make
computation easy.
(b) Plot the power spectrum of the clock generator signal in part (a) in deci-
bels.

*10.19. Let s3 (t) be defined as

s3 (t) = e−αt cos(2πf t + θ)u(t)

(a) Compute the Fourier transform of s3 (t). Hint: Use Euler’s identity.
(b) For the case where α = 50, f = 200, and θ = π/4, plot s3 (t) from 0
to 125 ms. Using subplot, plot the magnitude spectrum in decibels from 0 to 2
KHz.
(c) At what frequency does the peak of the spectrum occur? Is this what you
would expect? What parameter(s) would you change to make the peak more
distinct (sharper)? less distinct (broader)?

*10.20. Two time-domain signals, s1(t) and s2(t), and their Fourier trans-
forms, SA(F ) and SB(F ), are shown in Figures 10.8 and 10.9. Using your
knowledge of the location of zeros of the spectra of rectangular pulses, match
the spectra to the corresponding time domain signal. The zeros in the spectra
can be identified as the deep negative dips. Using standard computation, it is
unlikely that an exact zero will be produced.

————————————————————————————–
546 MATHEMATICS: THE LANGUAGE OF ECE

time domain
2

1.5

1
s1(t)

0.5

−0.5

−1
0 1 2 3 4 5 6 7 8

1.5

1
s2(t)

0.5

−0.5

−1
0 1 2 3 4 5 6 7 8
time in seconds

Figure 10.8. The signals in the time domain.

Solution of differential equations

10.21. A fourth order filter (circuit) is described by the differential equation

d4 v(t) d3 v(t) d2 v(t) dv(t)

+ 26 + 341 + 2613 + 10000v(t) = 10000s(t)
dt4 dt3 dt2 dt
(10.53)

(a) Compute the transfer function of the system represented by the differential
equation.
(b) Compute the output (solution) of the system when the input is

s(t) = cos(5t)

(c) Compute the output (solution) of the system when the input is

s(t) = cos(20t)

(d) Plot the frequency response of the system from 0 ≤ ω ≤ 50.

Fourier transforms 547

Frequency domain
60

40

20
SA(F)

−20

−40
0 5 10 15 20 25

50
SB(F)

−50
0 5 10 15 20 25
frequency in hertz

Figure 10.9. The Fourier transforms.

10.22. Given a system defined by the differential equation

d2 v(t) dv(t)
2
+ 30 + 900v(t) = 1800vs (t),
dt dt
use the Fourier transform to find the particular solution vp (t), if the input, vs (t)
is defined by
vs (t) = cos(30t)

10.23. Consider the differential equation

d2 v(t) R dv(t) 1 1
2
+ + v(t) = vs (t) (10.54)
dt L dt LC LC
where L = 100µH, C = 10µF and vs (t) = 5 cos(2π5000t). Use Fourier
transforms to:
(a) find the particular solution vp (t) when R = 0.6Ω.
(b) find the particular solution vp (t) when R = 12Ω.
548 MATHEMATICS: THE LANGUAGE OF ECE

10.24. Consider the differential equation obtained from an RLC circuit

d2 v(t) R dv(t) 1 1
2
+ + v(t) = vs (t)
dt L dt LC LC
where R = 0.5Ω, L = 10µH, C = 40µF , vs (t) = u(t) − u(t − 1), v(0) = 0
and v̇(0) = 0.
Use Fourier transforms to find the particular solution vp (t).

————————————————————————————–

Problems you will see in other courses

This set of problems is intended to highlight the connection of the material
in this chapter to other ECE courses.

*10.25. (ECE301) Filter design. A Butterworth filter of order K is defined

by the transfer function
1
|H(ω)|2 = ³ ´2K (10.55)
ω
1+ ωc

In Equation 10.55, K is a positive integer and ωc is called the cutoff frequency

of the filter.

1 Plot the transfer function for K = 1, 2, . . . , 10 and ωc = 1 KHz. Vary ω

between -10 and 10 KHz. Observe how the transfer function approximates
a rectangular function. Such a function would represent an ideal low-pass
filter with cutoff frequency ωc = 1 KHz.
2 Is |H(ω)| an odd or even function?
3 Determine at what frequency, ω0 , the power is equal to 3 dB. Recall that for
this frequency, |H(ω0 )|2 = 0.5.
4 Let K = 3 and ωc = 1. Determine the poles of the denominator in Equation
10.55.
5 Consider the transfer function
1
|H(ω)| =
(jω)6 + 3.9(jω)5 + 7.5(jω)4 + 9.1(jω)3 + 7.5(jω)2 + 3.9(jω) + 1
(10.56)

Does this function represent a Butterworth filter? Hint: you must show that
Equation 10.56 can be cast in the form of Equation 10.55. To do that, find
the conjugate, H ∗ (ω), of the given transfer function and compute
Fourier transforms 549

|H(ω)|2 = |H(ω)| · |H ∗ (ω)|

6 Consider the circuit shown in Figure 10.10. The input is the current source;
the output is the voltage across the resistor. Write the transfer function.

1F
i(t) 1Ω v(t)

Figure 10.10. Implementation of a first-order Butterworth filter.

7 Show that this circuit implements a Butterworth filter of order 1. Show that
the cutoff frequency is ωc = 1 Hz.
8 Consider the circuit shown in Figure 10.11. The input is the current source;
the output is the voltage across the capacitor. Write the transfer function.
550 MATHEMATICS: THE LANGUAGE OF ECE

In ECE301, this is called a passive filter. An improved implementation that

uses operational amplifiers from ECE200 would be called an active filter.
You will study even better digital filter implementations in ECE421.

1H

1F 1F
i(t) 1Ω v(t)

Figure 10.11. Implementation of a third-order Butterworth filter.

9 Show that this circuit implements a Butterworth filter of order 3. Show that
the cutoff frequency is ωc = 1 Hz. Show that the transfer function for this
filter is given by

1
|H(ω)| =
(jω)3 + 2(jω)2 + 2(jω) + 1
Fourier transforms 551

*10.26. (ECE402) Line codes. Consider Problem 2.46.

1 Determine the magnitude spectrum of the Unipolar Non-Return-to-Zero

(UNRZ) code signal. Plot the spectrum.

2 Determine the magnitude spectrum of the Polar Non-Return-to-Zero (PNRZ)

code signal. Plot the spectrum.

3 Determine the magnitude spectrum of the Bipolar Return-to-Zero (BRZ)

code signal. Plot the spectrum.

4 Suppose that any signal power lower than -30 dB is practically zero. Which
signal consumes more bandwidth?

*10.27. (ECE402) On-Off Keying (OOK). Consider Problem 2.47.

1 Determine the magnitude spectrum of the OOK signal in equation 2.60,

page 106. Plot the spectrum.

2 Suppose that any signal power lower than -30 dB is practically zero. Is the
OOK code more expensive, in terms of bandwidth, than the signals in the
previous problem?

*10.28. (ECE402) Binary Phase Shift Keying (BPSK). Consider Problem

2.48.

1 Determine the magnitude spectra of the BPSK signals in Equations 2.62 and
2.63, page 107. Plot the two spectra.

2 Suppose that any signal power lower than -30 dB is practically zero. Is the
BPSK code more expensive, in terms of bandwidth, than the signals in the
previous two problems?

*10.29. (ECE402) Frequency Shift Keying (FSK). Consider Problem 2.49.

1 Determine the magnitude spectra of the FSK signals in equations 2.64 and
2.65, page 109. Plot the two spectra.

2 Suppose that any signal power lower than -30 dB is practically zero. Is the
FSK code more expensive, in terms of bandwidth, than the signals in the
previous three problems?
552 MATHEMATICS: THE LANGUAGE OF ECE

*10.30. (ECE402) Binary Phase Shift Keying (BPSK) using raised cosine
signals. Consider the Binary Phase Shift Keying (BPSK) code in problem 2.48.
A variation of this code uses the signal
½A
2 [1 + cos(2πfc t)], |t| < 0.5fc ,
c
c1 (t) = (10.57)
0, |t| ≥ 0.5fc .
to transmit bit 1, while bit 0 is represented by the signal

c0 (t) = −c1 (t). (10.58)

The duration of the signals c1 (t) and c0 (t) is T seconds, resulting in a rate
R = 1/T bps. Figure 10.12 shows how the ASCII letter V (with ASCII code
1010110) would be transmitted using this code, over a 1 bps channel. Note that
T = 1 sec for such a channel. Compare the signal waveforms to that of the
BPSK code, in Figure 2.26, page 108.

The letter V in ASCII code

2

1.5
1 0 1 0 1 1 0
1
m(t)

0.5

0
T 2T 3T 4T 5T 6T 7T
−0.5

−1
−1 0 1 2 3 4 5 6 7 8

The signal for letter V in raised cosine BPSK

2

1 0 1 0 1 1 0
1
s(t)

−1

−2
−1 0 1 2 3 4 5 6 7 8 9
time (sec)

Figure 10.12. The raised cosine Binary Phase Shift Keying (BPSK) code.

The reason for even considering use of a “complicated” signal like that in
Equation 10.57 is the bandwidth savings it provides, compared to other BPSK
signals (see question 6).

1 Let fc = 1/(2T ). Sketch the waveform transmitted over a 50 kbps channel,

for the letter B. Be sure to mark your abscissa accurately.
Fourier transforms 553

2 Calculate and plot the magnitude spectrum of the signal used to send bit 1.
3 Calculate and plot the magnitude spectrum of the signal used to send bit 0.
Are they the same? Why?
4 Calculate and plot the magnitude spectrum of the signal used to send bit 1,
for the signal used in the code of Problem 2.48.
5 Calculate and plot the magnitude spectrum of the signal used to send bit 0,
for the signal used in the code of Problem 2.48. Are they the same? Why?
6 Is this BPSK code more or less expensive, in terms of bandwidth, than the
code in Problem 2.48? Suppose that any signal power lower than -30 dB is
practically zero.
7 In yet another BPSK variation, bit 1 is represented by the signal
½
Ac , |t| < 0.5fc ,
c1 (t) = (10.59)
0, |t| ≥ 0.5fc .
while bit 0 is represented by the signal

c0 (t) = −c1 (t). (10.60)

Calculate and plot the magnitude spectrum of the signals c1 (t) and c0 (t).
8 Suppose that any signal power lower than -30 dB is practically zero. Using
the spectrum plots or the formulas you have calculated, show that this BPSK
code is 5 times more expensive, in terms of bandwidth, than the raised cosine
BPSK code.

————————————————————————————–