FACULTY OF SCIENCES AND TECHNOLOGY

Electrical Engineering and Automation Department

Academic Year: 2024 – 2025
Master 1 Automation and System
Signal Processing Course Mr. Dahmane Ghenam

Chapter 1
Continuous Time Fourier series (CTFS), Continuous Time
Fourier Transform (CTFT), Convolution and Correlation

1.1 Introduction
In this chapter, we focus on periodic continuous time (CT) signals and introduce the continuous
time Fourier series used to decompose such signals into their frequency components. After that,
we consider an aperiodic CT signals and develops an equivalent Fourier representation, the
continuous time Fourier transform (CTFT), for aperiodic signals. At the end of the chapter, we
will consider two signal analysis concepts namely convolution and correlation. The convolution
operation is basic to linear system analysis, where the output signal of an LTIC (Linear Time
Invariant Continuous) system is given by the convolution operation between the input signal
and the impulse response of the system. The process of correlation is useful in comparing two
deterministic signals, it provides a measure of similarity between the first signal and a time
delayed version of the second signal.

1.2 Continuous Time Fourier Series (CTFS)

There exist two Fourier series representations, the trigonometric Fourier series and the
exponential Fourier series.

1.2.1 Trigonometric Fourier series

Definition
An arbitrary periodic function 𝑥(𝑡) with fundamental period 𝑇0 can be expressed as follows:
∞

𝑥(𝑡) = 𝑎0 + ∑ 𝑎𝑛 cos 𝑛𝜔0 𝑡 + 𝑏𝑛 sin 𝑛𝜔0 𝑡

𝑛=1

Where 𝜔0 = 2𝜋⁄𝑇0 is the fundamental frequency of 𝑥(𝑡) and coefficients 𝑎0 , 𝑎𝑛 and 𝑏𝑛 are
the trigonometric Fourier coefficients. These coefficients are calculated as follows:
𝑇0 𝑇0 𝑇0
2 2 2
1 2 2
𝑎0 = ∫ 𝑥(𝑡)𝑑𝑡, 𝑎𝑛 = ∫ 𝑥(𝑡) cos 𝑛𝜔0 𝑡 𝑑𝑡 and 𝑏𝑛 = ∫ 𝑥(𝑡) sin 𝑛𝜔0 𝑡 𝑑𝑡
𝑇0 𝑇0 𝑇0
𝑇 𝑇 𝑇
− 0 − 0 − 0
2 2 2
1.2.1.1 The Fourier Coefficients for Symmetrical Signals
 If 𝑥(𝑡) is an even function then,
𝑇0 𝑇0
2 2
2 4
𝑎0 = ∫ 𝑥(𝑡)𝑑𝑡, 𝑎𝑛 = ∫ 𝑥(𝑡) cos 𝑛𝜔0 𝑡 𝑑𝑡 and 𝑏𝑛 = 0
𝑇0 𝑇0
0 0

 If 𝑥(𝑡) is an odd function then,

𝑇0
2
4
𝑎0 = 0, 𝑎𝑛 = 0 and 𝑏𝑛 = ∫ 𝑥(𝑡) sin 𝑛𝜔0 𝑡 𝑑𝑡
𝑇0
0

Example 1.1
Find the trigonometric Fourier series for the square periodic signal 𝑓(𝑡) shown in Fig. 1.1

Fig. 1.1

Here, the period is 𝑇0 = 8 and 𝜔0 = 2𝜋⁄𝑇0 = 𝜋⁄4. Therefore,

∞
𝜋 𝜋
𝑓(𝑡) = 𝑎0 + ∑ 𝑎𝑛 cos 𝑛 𝑡 + 𝑏𝑛 sin 𝑛 𝑡
4 4
𝑛=1

Because the signal 𝑓(𝑡) is odd then, 𝑎0 = 𝑎𝑛 = 0, and the 𝑏𝑛 coefficients are given as follow:
𝑇
2 4 4
4 1 𝜋 𝜋 4 𝜋 4
𝑏𝑛 = ∫ 𝑓(𝑡) sin 𝑛𝜔𝑡 𝑑𝑡 = ∫ −2 sin 𝑛 𝑡 𝑑𝑡 = − ∫ sin 𝑛 𝑡 𝑑𝑡 = − [− cos 𝑛 𝑡]
𝑇 2 4 4 𝑛𝜋 4 0
0 0 0

4
= (cos 𝑛𝜋 − 1)
𝑛𝜋
Then
4
𝑏𝑛 = (cos 𝑛𝜋 − 1), 𝑛 ≥ 1
𝑛𝜋
 If 𝑛 is an even, (𝑛 = 2𝑘, 𝑘 ≥ 1), then 𝑏𝑛 = 𝑏2𝑘 = 0.
 If 𝑛 is an odd, (𝑛 = 2𝑘 + 1, 𝑘 ≥ 0), then 𝑏𝑛 = 𝑏2𝑘+1 = − 8⁄(2𝑘 + 1)𝜋 𝑘 ≥ 0.
Finally
∞ ∞
8 1 𝜋
𝑓(𝑡) = 𝑎
⏟0 + ∑ ⏟
𝑎𝑛 cos 𝑛𝜔𝑡 + 𝑏𝑛 sin 𝑛𝜔𝑡 = − ∑ sin ((2𝑘 + 1) 𝑡)
𝜋 2𝑘 + 1 4
=0 𝑛=1 =0 𝑘=0

The fig.1.2 shows the original signal 𝑓(𝑡) and the sum of DC, first, third and fifth harmonics
(even harmonics are absent).

Fig. 1.2

1.2.2 Exponential Fourier series

Definition
An arbitrary periodic function 𝑥(𝑡) with fundamental period 𝑇0 can be expressed as follows:
∞

𝑥(𝑡) = ∑ 𝐷𝑛 𝑒 𝑗𝑛𝜔0 𝑡
𝑛=0

Where the exponential coefficients 𝐷𝑛 are calculated as

𝐷𝑛 = ∫ 𝑥(𝑡)𝑒 −𝑗𝑛𝜔0 𝑡 𝑑𝑡
𝑇0

𝜔0 being the fundamental frequency given by 𝜔0 = 2𝜋⁄𝑇0

Example 1.2
Calculate the exponential Fourier series of the signal 𝑥(𝑡) shown in fig. 1.3

The fundamental period 𝑇0 = 4 and 𝜔0 = 2𝜋⁄𝑇0 = 𝜋⁄2. The exponential Fourier coefficients
are:
 𝑇0
2 1
1 −𝑗𝑛𝜔0 𝑡
1 −𝑗𝑛𝜔0 𝑡
1 2 −𝑗𝑛𝜋 1
𝐷𝑛 = ∫ 𝑥(𝑡)𝑒 𝑑𝑡 = ∫ 2𝑒 𝑑𝑡 = [− 𝑒 2]
𝑇0 4 2 𝑗𝑛𝜋 −1
𝑇 −1
− 0
2

𝜋
sin 𝑛 2 𝑛𝜋
𝐷𝑛 = 𝜋 = sinc ( 2 )
𝑛2

 If 𝑛 is an even integer, (𝑛 = 2𝑘, 𝑘 ∈ ℤ∗ ), then 𝐷𝑛 = 𝐷2𝑘 = sin 𝑘𝜋⁄2𝑘𝜋 = 0.

 For 𝑛 = 0 (DC component)
𝑇0
2 1
1 1 1
𝐷0 = ∫ 𝑥(𝑡)𝑑𝑡 = ∫ 2𝑑𝑡 = [𝑡]1−1 = 1
𝑇0 4 4
𝑇 −1
− 0
2

 If 𝑛 is an odd integer, (𝑛 = 2𝑘 + 1, 𝑘 ∈ ℤ), then 𝐷𝑛 = 𝐷2𝑘+1 =

2(sin(2𝑘 + 1) 𝜋⁄2)⁄(2𝑘 + 1)𝜋 = 2(−1)𝑘 ⁄(2𝑘 + 1)𝜋

Then, we have:
+∞ +∞ +∞
𝜋 𝜋 2 (−1)𝑘 𝑗(2𝑘+1)𝜋𝑡
𝑥(𝑡) = ∑ 𝐷𝑛 𝑒 𝑗𝑛 2 𝑡 = 𝐷0 + ∑ 𝐷2𝑘+1 𝑒 𝑗(2𝑘+1)2 𝑡 =1+ ∑ 𝑒 2
𝜋 (2𝑘 + 1)
𝑛=−∞ 𝑘=−∞ 𝑘=−∞

Fig. 1.3

The first harmonic ℎ1 (𝑡) (for 𝑛 = −1,1, or 𝑘 = −1,0 ), is given by:

2 −𝑗𝜋𝑡 2 𝑗𝜋𝑡 4 𝜋
ℎ1 (𝑡) = 𝐷−1 𝑒 −𝑗𝜔0 𝑡 + 𝐷1 𝑒 𝑗𝜔0 𝑡 = 𝑒 2 + 𝑒 2 = cos 𝑡
𝜋 𝜋 𝜋 2

1.2.3 Fourier Spectrum

 For the trigonometric Fourier series, the amplitude spectra can be expressed as

𝑆𝒏 = √𝑎𝑛 2 + 𝑏𝑛 2
And the phase spectra is given as

𝑏𝑛
𝜃𝑛 = −tan−1
𝑎𝑛

 For the exponential Fourier series, the amplitude and phase spectra can be defined as
the magnitude of exponential Fourier coefficients |𝐷𝑛 | and the phase of exponential
Fourier coefficients ∠𝐷𝑛 respectively.

For the example 1.2, the amplitude spectrum is determine as

𝜋
sin 𝑛 2 𝑛𝜋
|𝐷𝑛 | = | 𝜋 | = |sinc ( )|
𝑛2 2

and the phase spectrum is calculated as

𝑛𝜋 0 𝑖𝑓 𝐷𝑛 ≥ 0
∠𝐷𝑛 = ∠sinc ( )={
2 180° 𝑖𝑓 𝐷𝑛 < 0

The fig. 1.4 shows the amplitude and phase spectra of the signal 𝑥(𝑡) in the example 1.2

Fig. 1.4
1.2.4 Parseval’s Identity
The power of a periodic signal 𝑥(𝑡) can be calculated from its exponential Fourier coefficients
as follows:

∞
1
𝑃𝑥 = ∫ |𝑥(𝑡)|2 𝑑𝑡 = ∑ |𝐷𝑛 |2
𝑇0
(𝑇0 ) 𝑛=−∞

For real-valued signals, |𝐷−𝑛 | = |𝐷𝑛 |, which results in the following simplified formula:
∞ ∞

𝑃𝑥 = ∑ |𝐷𝑛 |2 = |𝐷0 |2 + 2 ∑|𝐷𝑛 |2

𝑛=−∞ 𝑛=1
The power of a periodic signal 𝑥(𝑡) can also calculated from its trigonometric Fourier
coefficients as follow:

∞
1 1
𝑃𝑥 = ∫ |𝑥(𝑡)|2 𝑑𝑡 = 𝑎0 2 + ∑(𝑎𝑛 2 + 𝑏𝑛 2 )
𝑇0 2
(𝑇0 ) 𝑛=1

Example 1.3
Calculate the sum 𝑆 of the following infinite series:
∞
1
𝑆=∑
(2𝑘 + 1)2
𝑘=0

To compute the sum 𝑆, we consider the periodic signal 𝑓(𝑡) shown in Fig. 1.1. As shown in
Example 1.1, the trigonometric Fourier coefficients of 𝑓(𝑡) are given by
4
𝑎0 = 𝑎𝑛 = 0 and 𝑏𝑛 = 𝑛𝜋 (cos 𝑛𝜋 − 1), 𝑛 ≥ 1 or

0 𝑓𝑜𝑟 𝑛 = 2𝑘, 𝑘>0

𝑏𝑛 = {
− 8⁄(2𝑘 + 1)𝜋 𝑓𝑜𝑟 𝑛 = 2𝑘 + 1, 𝑘≥0

Using Parseval’s theorem, the average power of 𝑓(𝑡) is given by

∞ ∞ 2 ∞ ∞
2
1 2 2 1 2 1 −8 1 64
𝑃𝑥 = 𝑎0 + ∑(𝑎𝑛 + 𝑏𝑛 ) = ∑ 𝑏𝑛 = ∑ ( ) = ∑
2 2 2 (2𝑘 + 1)𝜋 2 (2𝑘 + 1)2 𝜋 2
𝑛=1 𝑛=1 𝑘=0 𝑘=0

∞
32 1 32
𝑃𝑥 = 2 ∑ 2
= 2 𝑆 ⋯ (1)
𝜋 (2𝑘 + 1) 𝜋
𝑘=0

The average power of 𝑓(𝑡) yields

𝑇0 𝑇0
2 2 4
1 2 1
𝑃𝑥 = ∫ |𝑓(𝑡)|2 𝑑𝑡 = ∫ |𝑓(𝑡)|2 𝑑𝑡 = ∫|−2|2 𝑑𝑡 = 4 ⋯ (2)
𝑇0 𝑇0 4
𝑇 0 0
− 0
2

Combining Eqs. (1) and (2) gives

∞
32 𝜋2 1 𝜋2
𝑆=4⇒𝑆= ⇒∑ =
𝜋2 8 (2𝑘 + 1)2 8
𝑘=0

1.3 Continuous Time Fourier Transform (CTFS)

The direct Fourier transform of an aperiodic signal 𝑥(𝑡) is given by the formula:
 ∞

𝑋(𝜔) = ℱ[𝑥(𝑡)] = ∫ 𝑥(𝑡) 𝑒 −𝑗𝜔𝑡 𝑑𝑡

−∞

And the inverse Fourier transform of 𝑋(𝜔) is calculated as follows:

∞
−1 [𝑋(𝜔)]
1
𝑥(𝑡) = ℱ = ∫ 𝑋(𝜔) 𝑒 𝑗𝜔𝑡 𝑑𝜔
2𝜋
−∞

We say that the signal 𝑥(𝑡) and its transform 𝑋(𝜔) form a Fourier transform pair and denote
their relationship by

𝑥(𝑡) ⇔ 𝑋(𝜔)

The necessary conditions required for a function to be transformable are called the Dirichlet
conditions. These are listed as:

1. 𝑥(𝑡) is bounded.
2. 𝑥(𝑡) has a finite number of maxima and minima within any finite interval.
3. 𝑥(𝑡) has a finite number of discontinuities within any finite interval.
4. 𝑥(𝑡) is absolutely integrable, that is,
∞

∫ |𝑥(𝑡)|𝑑𝑡 < ∞
−∞

We can plot the spectrum 𝑋(𝜔) as a function of 𝜔. Since 𝑋(𝜔) is complex, we have both
amplitude and angle (or phase) spectra

𝑋(𝜔) = |𝑋(𝜔)|𝑒 𝑗∠𝑋(𝜔)

in which |𝑋(𝜔)| is the amplitude and ∠𝑋(𝜔) is the angle (or phase) of 𝑋(𝜔).

Example 1.4
Determine the direct Fourier transform of the following rectangular (gate) signal 𝑥(𝑡), and
plot the corresponding amplitude and phase spectra
𝜏
𝑡 1 |𝑡| ≤
𝑥(𝑡) = Π ( ) = { 2
𝜏 𝜏
0 |𝑡| >
2
By definition
 𝜏
∞ 2 𝜏
1 2
𝑋(𝜔) = ℱ[𝑥(𝑡)] = ∫ 𝑥(𝑡) 𝑒 −𝑗𝜔𝑡 𝑑𝑡 = ∫ 𝑒 −𝑗𝜔𝑡 𝑑𝑡 = [− 𝑒 −𝑗𝜔𝑡 ]
𝑗𝜔 −
𝜏
−∞ 𝜏 2
−
2
1 𝜏 𝜏
= (𝑒 𝑗𝜔2 − 𝑒 −𝑗𝜔2 )
𝑗𝜔
𝜏 𝜏 𝜏
2 𝑒 𝑗𝜔2 − 𝑒 −𝑗𝜔2 2 𝜏 sin 2 𝜔 𝜏
𝑋(𝜔) = ( ) = sin 𝜔 = 𝜏 𝜏 = 𝜏sinc ( 𝜔)
𝜔 2𝑗 𝜔 2 2
2𝜔
The magnitude and phase of 𝑋(𝜔) are given respectively by
𝜏 𝜏 0 𝑖𝑓 𝑋(𝜔) ≥ 0
|𝑋(𝜔)| = |𝜏sinc ( 𝜔)| ∠𝑋(𝜔) = ∠𝜏sinc ( 𝜔) = {
2 2 180° 𝑖𝑓 𝑋(𝜔) < 0

The Fig. 1.5a and b illustrate the amplitude spectrum |𝑋(𝜔)| and the phase spectrum ∠𝑋(𝜔)
for 𝜏 = 1. The Fig.1.6a and b demonstrate the amplitude spectrum |𝑋(𝜔)| and the phase
spectrum ∠𝑋(𝜔) for 𝜏 = 2

(a) Fig.1.5 (b)

(a) (b)
Fig.1.6

1.3.1 Some Properties of Fourier Transform

We now study some of the important properties of the Fourier transform and their implications
as well as applications.
1.3.1.1 Linearity
The linearity property states that if 𝑥1 (𝑡) ⇔ 𝑋1 (𝜔) and 𝑥2 (𝑡) ⇔ 𝑋2 (𝜔), then 𝑎1 𝑥1 (𝑡) +
𝑎2 𝑥2 (𝑡) ⇔ 𝑎1 𝑋1 (𝜔) + 𝑎2 𝑋2 (𝜔)

1.3.1.2 Duality
The duality property states that if 𝑥(𝑡) ⇔ 𝑋(𝜔), then 𝑋(𝑡) ⇔ 2𝜋𝑥(−𝜔)

1.3.1.3 Conjugation and Conjugate Symmetry

The conjugation property states that if 𝑥(𝑡) ⇔ 𝑋(𝜔), then 𝑥 ∗ (𝑡) ⇔ 𝑋 ∗ (−𝜔). From this
property follows the conjugate symmetry property, which states that if 𝑥(𝑡) is real, then
𝑋(−𝜔) = 𝑋 ∗ (𝜔).

1.3.1.4 Time Scaling

If 𝑥(𝑡) ⇔ 𝑋(𝜔), then for any nonzero real constant 𝑎,

1 𝜔
𝑥(𝑎𝑡) ⇔ 𝑋( )
|𝑎| 𝑎

1.3.1.5 Time Shifting

If 𝑥(𝑡) ⇔ 𝑋(𝜔), then

𝑥(𝑡 − 𝑡0 ) ⇔ 𝑋(𝜔)𝑒 −𝑗𝜔𝑡0

1.3.1.6 Frequency Shifting

If 𝑥(𝑡) ⇔ 𝑋(𝜔), then

𝑥(𝑡)𝑒 𝑗𝜔0 𝑡 ⇔ 𝑋(𝜔 − 𝜔0 )

1.3.1.7 Convolution
The time-convolution property and its dual, the frequency-convolution property, state that if

𝑥1 (𝑡) ⇔ 𝑋1 (𝜔) and 𝑥2 (𝑡) ⇔ 𝑋2 (𝜔)

Then

𝑥1 (𝑡) ∗ 𝑥2 (𝑡) ⇔ 𝑋1 (𝜔)𝑋2 (𝜔) (time convolution)

And

1
𝑥1 (𝑡)𝑥2 (𝑡) ⇔ 𝑋 (𝜔) ∗ 𝑋2 (𝜔) (frequency convolution)
2𝜋 1
1.3.1.8 Time Differentiation and Time Integration
If 𝑥(𝑡) ⇔ 𝑋(𝜔), then

𝑑𝑥(𝑡)
⇔ 𝑗𝜔𝑋(𝜔) (time differentiation)
𝑑𝑡
And
𝑡
𝑋(𝜔)
∫ 𝑥(𝜏)𝑑𝜏 ⇔ + 𝜋𝑋(0)𝛿(𝜔) (time integration)
𝑗𝜔
−∞

A list of the Fourier transform properties is included in Table 1.1. And the Table 1.2 list the
some Fourier transform pairs.

1.3.1.9 Parseval’s Energy Theorem

Parseval’s theorem relates the energy of a signal in the time domain to the energy of its Fourier
transform in the frequency domain and states that if 𝑥(𝑡) ⇔ 𝑋(𝜔), then
∞ ∞
1
𝐸𝑥 = ∫ |𝑥(𝑡)|2 𝑑𝑡 = ∫ |𝑋(𝜔)|2 𝑑𝜔
2𝜋
−∞ −∞

1.4 Convolution Integral

The convolution integral of two functions 𝑥1 (𝑡) and 𝑥2 (𝑡) is denoted symbolically by 𝑥1 (𝑡) ∗
𝑥2 (𝑡) and is defined as
∞

𝑥1 (𝑡) ∗ 𝑥2 (𝑡) ≡ ∫ 𝑥1 (𝜏)𝑥2 (𝑡 − 𝜏)𝑑𝜏

−∞

Some important properties of the convolution integral follow.

Commutative Property

Convolution operation is commutative; that is, 𝑥1 (𝑡) ∗ 𝑥2 (𝑡) = 𝑥2 (𝑡) ∗ 𝑥1 (𝑡). This property
can be proved by a change of variable.
+∞

𝑥(𝑡) ∗ 𝑦(𝑡) = ∫ 𝑥(𝜏)𝑦(𝑡 − 𝜏)𝑑𝜏

−∞
−∞

= − ∫ 𝑥(𝑡 − 𝛽)𝑦(𝛽)𝑑𝛽
+∞

= 𝑦(𝑡) ∗ 𝑥(𝑡)
Table 1.1 Fourier Transform Properties
 Table 1.2 Fourier Transform Pairs

Let 𝛽 = 𝑡 − 𝜏 so that 𝜏 = 𝑡 − 𝛽 and 𝑑𝜏 = −𝑑𝛽

The Distributive Property

The distributive property, can be proved as

 +∞

𝑥(𝑡) ∗ (𝑦(𝑡) + 𝑧(𝑡)) = ∫ 𝑥(𝜏)(𝑦(𝑡 − 𝜏) + 𝑧(𝑡 − 𝜏))𝑑𝜏

−∞

+∞

= ∫ 𝑥(𝜏)(𝑦(𝑡 − 𝜏) + 𝑧(𝑡 − 𝜏))𝑑𝜏

−∞

+∞

= ∫ (𝑥(𝜏)𝑦(𝑡 − 𝜏) + 𝑥(𝜏)𝑧(𝑡 − 𝜏))𝑑𝜏

−∞

+∞ +∞

= ∫ 𝑥(𝜏)𝑦(𝑡 − 𝜏)𝑑𝜏 + ∫ 𝑥(𝜏)𝑧(𝑡 − 𝜏)𝑑𝜏

−∞ −∞

= 𝑥(𝑡) ∗ 𝑦(𝑡) + 𝑥(𝑡) ∗ 𝑧(𝑡)

The Associative Property

The associative property, can be proved as

+∞

(𝑥(𝑡) ∗ 𝑦(𝑡)) ∗ 𝑧(𝑡) = ∫ (𝑥(𝜏) ∗ 𝑦(𝜏))𝑧(𝑡 − 𝜏)𝑑𝜏

−∞
+∞ +∞

= ∫ ( ∫ 𝑥(𝛽)𝑦(𝜏 − 𝛽)𝑑𝛽) 𝑧(𝑡 − 𝜏)𝑑𝜏

−∞ −∞

+∞ +∞

= ∫ ( ∫ 𝑥(𝛽)𝑦(𝜏 − 𝛽)𝑧(𝑡 − 𝜏)𝑑𝜏) 𝑑𝛽 (Fubini's theorem)

−∞ −∞

+∞ +∞

= ∫ 𝑥(𝛽) ( ∫ 𝑦(𝜏 − 𝛽)𝑧(𝑡 − 𝜏)𝑑𝜏) 𝑑𝛽

−∞ −∞

+∞ +∞

= ∫ 𝑥(𝛽) ( ∫ 𝑦(𝜌)𝑧(𝑡 − 𝛽 − 𝜌)𝑑𝜌) 𝑑𝛽

−∞ −∞

+∞

= ∫ 𝑥(𝛽)(𝑦 ∗ 𝑧)(𝑡 − 𝛽)𝑑𝛽

−∞

= 𝑥(𝑡) ∗ (𝑦(𝑡) ∗ 𝑧(𝑡))

let 𝜌 = 𝜏 − 𝛽, so that 𝜏 = 𝜌 + 𝛽 and 𝑑𝜏 = 𝑑𝜌

The Shift Property

If
+∞

𝑦(𝑡) = 𝑥1 (𝑡) ∗ 𝑥2 (𝑡) = ∫ 𝑥1 (𝜏)𝑥2 (𝑡 − 𝜏)𝑑𝜏

−∞

Then
𝑥1 (𝑡 − 𝑡1 ) ∗ 𝑥2 (𝑡 − 𝑡2 ) = 𝑦(𝑡 − (𝑡1 + 𝑡2 ))
Proof
+∞

𝑥1 (𝑡 − 𝑡1 ) ∗ 𝑥2 (𝑡 − 𝑡2 ) = ∫ 𝑥1 (𝜏 − 𝑡1 )𝑥2 (𝑡 − 𝜏 − 𝑡2 )𝑑𝜏 (1)

−∞

let 𝜏 − 𝑡1 = 𝛼, then 𝑑𝛼 = 𝑑𝜏 and 𝜏 = 𝛼 + 𝑡1

Eq.1 become:
+∞

𝑥1 (𝑡 − 𝑡1 ) ∗ 𝑥2 (𝑡 − 𝑡2 ) = ∫ 𝑥1 (𝛼)𝑥2 (𝑡 − (𝛼 + 𝑡1 ) − 𝑡2 )𝑑𝛼
−∞
+∞

= ∫ 𝑥1 (𝛼)𝑥2 (𝑡 − (𝑡1 + 𝑡2 ) − 𝛼)𝑑𝛼 = 𝑦(𝑡 − (𝑡1 + 𝑡2 ))

−∞

Convolution with Impulse Function

Convolution of a function 𝑥(𝑡) with a unit impulse 𝛿(𝑡) results in the function 𝑥(𝑡) itself. By
definition of convolution,
+∞

𝑥(𝑡) ∗ 𝛿(𝑡) = ∫ 𝑥(𝜏)𝛿(𝑡 − 𝜏)𝑑𝜏 = 𝑥(𝑡)

−∞

1.4.1 Impulse Response of System

The impulse response ℎ(𝑡) of a Linear Time Invariant (LTI) system is the output of the system
when a unit impulse 𝛿(𝑡) is applied at the input. When an arbitrary input signal 𝑥(𝑡) is passed
through an LTI system with impulse response ℎ(𝑡) , the resulting output 𝑦(𝑡) of the system can
be calculated by convolving the input signal 𝑥(𝑡) and the impulse response ℎ(𝑡) as follows:
+∞

𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏

−∞

The Fig1.7 shows the LTI system response 𝑦(𝑡) to input signal 𝑥(𝑡).
 Input Output
LTI System

𝛿(𝑡) ℎ(𝑡)
+∞

𝑥(𝑡) 𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏

−∞
Fig.1.7

1.4.2 Graphical Method for Evaluating the convolution integral

1 Sketch the waveform for input 𝑥(𝜏) by changing the independent variable from 𝑡 to 𝜏 and
keep the waveform for 𝑥(𝜏) fixed during convolution.
2 Sketch the waveform for the impulse response ℎ(𝜏) by changing the independent variable
from 𝑡 to 𝜏.
3 Reflect ℎ(𝜏) about the vertical axis to obtain the time-inverted impulse response ℎ(−𝜏).
4 Shift the time-inverted impulse function ℎ(−𝜏) by a selected value of 𝑡. The resulting
function represents ℎ(𝑡 − 𝜏).
5 Multiply function 𝑥(𝜏) by ℎ(𝑡 − 𝜏) and plot the product function 𝑥(𝜏)ℎ(𝑡 − 𝜏).
6 Calculate the total area under the product function 𝑥(𝜏)ℎ(𝑡 − 𝜏) by integrating it over 𝜏 =
[−∞, ∞].
7 Repeat steps 4−6 for different values of 𝑡 to obtain 𝑦(𝑡) for all time, −∞ < 𝑡 < ∞.

Example 1.5
Using the graphical convolution method to calculate 𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡), where

𝑡−5 𝑡−2
𝑥(𝑡) = 2Π ( ) and ℎ(𝑡) = Π ( )
2 4
Functions 𝑥(𝜏) and ℎ(−𝜏) are plotted, in Fig.1.8

𝑥(𝜏)
ℎ(−𝜏), 𝑡 = 0

𝑡−4 𝑡
Fig.1.8
The function ℎ(𝑡 − 𝜏) is obtained by shifting ℎ(−𝜏) by time . We consider the following five
cases of .

Case 1 For 𝑡 < 4, the waveform ℎ(𝑡 − 𝜏) is on the left-hand side of the vertical axis. As is
apparent in Fig. 1.9, waveforms for ℎ(𝑡 − 𝜏) and 𝑥(𝜏) do not overlap. In other words,
𝑥(𝜏)ℎ(𝑡 − 𝜏) = 0 for all , hence 𝑦(𝑡) = 0.
 𝑥(𝜏)
ℎ(𝑡 − 𝜏), 𝑡 < 4

𝑡−4 𝑡
Fig.1.9
Case 2 For 4 ≤ 𝑡 < 6, we see from Fig. 1.10 that the ℎ(𝑡 − 𝜏) and 𝑥(𝜏) overlap over the
duration 𝜏 = [4, 𝑡]. Therefore,
+∞ 𝑡

𝑦(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏 = ∫ 2 ∙ 1𝑑𝜏 = [2𝜏]𝑡4 = 2𝑡 − 8

−∞ 4

𝑥(𝜏)
ℎ(𝑡 − 𝜏), 4 ≤ 𝑡 < 6

𝑡−4 𝑡
Fig. 1.10
Case 3 For 𝑡 ≥ 6 and 𝑡 − 4 < 4 so that 4 ≤ 𝑡 < 8, we see from Fig. 1.11 that the ℎ(𝑡 − 𝜏) and
𝑥(𝜏) overlap over the duration 𝜏 = [4,6]. Therefore,
+∞ 6

𝑦(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏 = ∫ 2 ∙ 1𝑑𝜏 = [2𝜏]64 = 4

−∞ 4

𝑥(𝜏)
ℎ(𝑡 − 𝜏), 6 ≤ 𝑡 < 8

𝑡−4 𝑡
Fig. 1.11
Case 4 For 4 ≤ 𝑡 − 4 < 6 so that 8 ≤ 𝑡 < 10, then there exist an overlap over the duration
𝜏 = [𝑡 − 4,6] (see Fig. 1.12). Therefore
 +∞ 6

𝑦(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏 = ∫ 2 ∙ 1𝑑𝜏 = [2𝜏]6𝑡−4 = −2𝑡 + 20

−∞ 𝑡−4

𝑥(𝜏)

ℎ(𝑡 − 𝜏), 8 ≤ 𝑡 < 10

𝑡−4 𝑡
Fig. 1.12

Case 5 For 𝑡 − 4 ≥ 6, so that 𝑡 ≥ 10, as in Fig. 1.13, there is no overlap between ℎ(𝑡 − 𝜏) and
𝑥(𝜏). Therefore
+∞

𝑦(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏 = 0

−∞

𝑡−4
Fig. 1.13

Combining the two cases (see Fig. 1.14), we obtain

0 𝑡<4
+∞
2𝑡 − 8 4≤𝑡<6
𝑦(𝑡) = ∫ 𝑥(𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏 = 4 6≤𝑡<8
−∞ −2𝑡 + 20 8 ≤ 𝑡 < 10
{0 𝑡 ≥ 10

Fig. 1.14
Example 1.6
Consider the RL circuit shown in Fig. 1.15a. Using the convolution, find the output voltage
𝑣𝑜 (𝑡) of the circuit if the input voltage source 𝑣𝑠 (𝑡) = 𝑢(𝑡) − 𝑢(𝑡 − 4) as represented in Fig.
1.15b. The impulse response of the circuit is given by

𝑅 −𝑅 𝑡
ℎ(𝑡) = 𝑒 𝐿 ∙ 𝑢(𝑡) = 𝑒 −𝑡 ∙ 𝑢(𝑡)
𝐿
Using the convolution, the output voltage 𝑣𝑜 (𝑡) is calculated as
+∞

𝑣𝑜 (𝑡) = 𝑣𝑠 (𝑡) ∗ ℎ(𝑡) = ∫ 𝑣𝑠 (𝜏)ℎ(𝑡 − 𝜏)𝑑𝜏

−∞

(a) Fig. 1.15 (b)

We proceed as in example 1, we find (see Fig. 1.16)

0 𝑡<0
𝑣𝑜 (𝑡) = {1 − 𝑒 −𝑡 0≤𝑡<4
𝑒 4−𝑡 − 𝑒 −𝑡 𝑡≥4

Fig. 1.16

1.5 Correlation
The process of correlation is useful in comparing two deterministic signals and it provides a
measure of similarity between the first signal and a time delayed of the second signal. The
comparison of the function 𝑥(𝑡) with its shifted version 𝑥(𝑡 + 𝜏), is called the autocorrelation
function of 𝑥(𝑡), as a function of 𝜏 and is given by
+∞

𝑅𝑥𝑥 (𝜏) = ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡

−∞
To compare two different functions 𝑥(𝑡) and ℎ(𝑡), we use the cross-correlation function
defined by
+∞

𝑅𝑥ℎ (𝜏) = 𝑥(𝜏) ∗∗ ℎ(𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡

−∞

Cross-correlation function gives the similarity between the two functions: 𝑥(𝑡) and ℎ(𝑡 + 𝜏).
Many a times the second function ℎ(𝑡) may be a corrupted version of 𝑥(𝑡), such as ℎ(𝑡) =
𝑥(𝑡) + 𝑛(𝑡), where 𝑛(𝑡) is a noise signal. In the case of 𝑥(𝑡) = ℎ(𝑡), cross-correlation reduces
to autocorrelation. In this case, at 𝜏 = 0, the autocorrelation integral gives the highest value at
𝜏 = 0. Comparison of two functions appears in many identification situations. For example, to
identify an individual based upon his speech pattern, we can store his speech segment in a
computer. When he enters, say a secure area, we can request him to speak and compute the
cross-correlation between the stored and the recorded. Then decide on the individual’s
identification based on the cross-correlation function. Generally, an individual is identified if
the peak of the cross-correlation is close to the possible peak autocorrelation value. Allowance
is necessary since the speech is a function of the individual’s physical and mental status of the
day the test is made.

1.5.1 Basic Properties of cross-correlation Functions

Folding relationship between the two cross-correlation functions is

𝑅𝑥ℎ (𝜏) = 𝑅ℎ𝑥 (−𝜏), that

+∞ +∞

𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = ∫ ℎ(𝛼)𝑥(𝛼 − 𝜏)𝑑𝛼 = 𝑅ℎ𝑥 (−𝜏)

−∞ −∞

1.5.2 Cross-correlation and Convolution

The cross-correlation function is related to the convolution as:

𝑅𝑥ℎ (𝜏) = 𝑥(𝜏) ∗∗ ℎ(𝜏) = 𝑥(−𝜏) ∗ ℎ(𝜏)

𝑅ℎ𝑥 (𝜏) = ℎ(𝜏) ∗∗ 𝑥(𝜏) = ℎ(−𝜏) ∗ 𝑥(𝜏)

So that, by changing the variable 𝑡 = −𝛼

+∞ +∞

𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑥(−𝛼)ℎ(𝜏 − 𝛼)𝑑𝛼 = 𝑥(−𝜏) ∗ ℎ(𝜏)

−∞ −∞

If 𝑥(𝑡) and ℎ(𝑡) are even functions, say 𝑥(𝑡) = 𝑥(−𝑡) and ℎ(𝑡) = ℎ(−𝑡), then

𝑅𝑥ℎ (𝜏) = 𝑥(−𝜏) ∗ ℎ(𝜏) = 𝑥(𝜏) ∗ ℎ(𝜏) ⋯ (1)

And
 𝑅ℎ𝑥 (𝜏) = ℎ(−𝜏) ∗ 𝑥(𝜏) = ℎ(𝜏) ∗ 𝑥(𝜏) ⋯ (2)

From (1) and (2), we have

𝑅𝑥ℎ (𝜏) = 𝑅ℎ𝑥 (𝜏)

Example
Derive the cross–correlation 𝑅𝑥ℎ (𝜏) for the following functions:

𝑡−1
𝑥(𝑡) = 𝑡(𝑢(𝑡) − 𝑢(𝑡 − 1)), ℎ(𝑡) = Π ( )
2
Functions 𝑥(𝑡) and ℎ(𝑡) are plotted, in Fig. 1.17a, b

(a) Fig. 1.17 (b)

The function ℎ(𝑡 + 𝜏) starts at 𝑡 = −𝜏 and ends at 𝑡 = −𝜏 + 2. As 𝜏 varies from −∞ to ∞,

there are five possible regions we need to consider.

Case 1 For −𝜏 + 2 < 0 or 𝜏 > 2, (see Fig. 1.18). There is no overlap between 𝑥(𝑡) and
ℎ(𝑡 + 𝜏). Therefore
+∞

𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = 0

−∞

−𝝉 −𝝉 + 𝟐
Fig. 1.17

Case 2 for 0 ≤ −𝜏 + 2 < 1 or 1 < 𝜏 ≤ 2, then there exist an overlap over the duration 𝑡 =
[0, −𝜏 + 2], (see Fig. 1.18) therefore
 +∞ −𝜏+2
1 2 −𝜏+2 1
𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑡 ∙ 1𝑑𝑡 = [ 𝑡 ] = (−𝜏 + 2)2
2 0 2
−∞ 0

−𝝉 −𝝉 + 𝟐

Fig. 1.18

Case 3 for −𝜏 + 2 ≥ 1 and −𝜏 < 0 or 0 < 𝜏 ≤ 1, then there exist an overlap over the
duration 𝑡 = [0,1], (see Fig. 1.19) therefore
+∞ 1
1 2 1 1
𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑡(1)𝑑𝑡 = [ 𝑡 ] =
2 0 2
−∞ 0

−𝝉 −𝝉 + 𝟐
Fig. 1.19

Case 4 for 0 ≤ −𝜏 < 1 or − 1 < 𝜏 ≤ 0, then there exist an overlap over the duration 𝑡 =
[−𝜏, 1], (see Fig. 1.20) therefore
+∞ 1
1 1 1 1
𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑡(1)𝑑𝑡 = [ 𝑡 2 ] = − 𝜏 2
2 −𝜏 2 2
−∞ −𝜏

−𝝉 −𝝉 + 𝟐
Fig. 1.20
Case 5 for −𝜏 ≥ 1 ⇒ 𝜏 ≤ −1, There is no overlap between 𝑥(𝑡) and ℎ(𝑡 + 𝜏). (See Fig. 1.21),
then
+∞

𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = 0

−∞

−𝝉 −𝝉 + 𝟐
Fig. 1.21
Finally, the cross-correlation 𝑅𝑥ℎ (𝜏) is given by (see Fig. 1.22)

0 𝜏>2
1
(−𝜏 + 2)2 1<𝜏≤2
+∞ 2
1
𝑅𝑥ℎ (𝜏) = ∫ 𝑥(𝑡)ℎ(𝑡 + 𝜏)𝑑𝑡 = 0<𝜏≤1
2
−∞ 1 1 2
− 𝜏 −1<𝜏≤0
2 2
{ 0 𝜏 ≤ −1

𝑅𝑥ℎ (𝜏)

Fig. 1.22

1.5.3 Correlation of periodic Functions

In the case of periodic functions, we assume that both are periodic with the same period. The
cross-correlation 𝑅𝑇,𝑥ℎ (𝜏) between two periodic signals with the same period is defined as
follow

1
𝑅𝑇,𝑥ℎ (𝜏) = ∫ 𝑥𝑇 (𝑡)ℎ𝑇 (𝑡 + 𝜏)
𝑇
𝑇

Where 𝑥𝑇 (𝑡) and ℎ𝑇 (𝑡 + 𝜏) are periodic signals with the same period 𝑇.
We define also the cross-correlation 𝑅𝑇,𝑥 (𝜏) between two periodic signals with the same period
as

1
𝑅𝑇,𝑥 (𝜏) = ∫ 𝑥𝑇 (𝑡)𝑥𝑇 (𝑡 + 𝜏)𝑑𝑡
𝑇
𝑇

Where 𝑥𝑇 (𝑡) is periodic signal with period 𝑇.

Problems
1.1 For each of the periodic signals shown in Fig. P1.1-3, find the trigonometric
Fourier series and sketch the amplitude and phase spectra.

Fig. P1.1.1

Fig. P1.1.2

Fig. P1.1.3

1.2 Determine the trigonometric Fourier series coefficients a,1 and bn for the following signals.
In each case, also determine the signals' fundamental radian frequency WQ. No integration is
required to solve this problem.

(a) 𝑥𝑎 (𝑡) = cos(3𝜋𝑡)

(b) 𝑥𝑏 (𝑡) = sin(7𝜋𝑡)

(c) 𝑥𝑐 (𝑡) = 2 + 4 cos(3𝜋𝑡) − 2𝑗 sin(7𝜋𝑡)

(d) 𝑥𝑑 (𝑡) = (1 + 𝑗) sin(3𝜋𝑡) + (2 − 𝑗) cos(7𝜋𝑡)

(e) 𝑥𝑒 (𝑡) = sin(3𝜋𝑡 + 1) + 2 cos(7𝜋𝑡 − 2)

(f) 𝑥𝑓 (𝑡) = sin(6𝜋𝑡) + 2 cos(14𝜋𝑡)

1.3 Find the exponential Fourier series for the signal in Fig. P1.3.1 and sketch the amplitude
and phase spectra.

Fig. P1.3.1

1.4 For each of the periodic signals in Fig. P3.4-3, find the exponential Fourier series and sketch
the corresponding spectra.

1.5 Determine graphically 𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡) for 𝑥(𝑡) = 𝑒 −𝑡 𝑢(𝑡) and ℎ(𝑡) = 𝑒 −2𝑡 𝑢(𝑡).

1.6 Find 𝑥(𝑡) ∗ ℎ(𝑡) for the functions 𝑥(𝑡) and ℎ(𝑡) shown in Figs. p1.6a and p1.6b.

Fig. P1.6

1.7 Using the direct Fourier transform Equation, find the Fourier transforms of the signals 𝑥(𝑡)
in Fig. P1.7.

Fig. P1.7

1.8 Using the inverse Fourier transform Equation, find the inverse Fourier transforms of the
spectra in Fig. P1.8.

(a) Fig. P1.8 (b)