1MAE001

Control 1

II. Signals and Systems

1
 Signals and Systems
Outline
1. Introduction
2. Discrete systems, the z-transform
3. Continuous systems, the Laplace transform
4. Laplace and z-transform, in practice
5. Frequency domain
6. Continuous and discrete systems

Signals and Systems 2

 1. Introduction

Signals and Systems 3

 Introduction

● We need a (mathematical) tool to manipulate signals and systems

– Signals 𝑢(𝑡) and 𝑦(𝑡) (input and output) 𝑢 𝑦
– System ℎ ℎ

● Everything is based on (only) two assumptions:

– System ℎ is Linear
𝑢1 𝑡 → 𝑦1 𝑡
⇒ 𝛼𝑢1 𝑡 + 𝛽𝑢2 𝑡 → 𝛼𝑦1 𝑡 + 𝛽𝑦2 𝑡
𝑢2 𝑡 → 𝑦2 𝑡
– Systems are Time Invariant
𝑢 𝑡 → 𝑦 𝑡 ⇒ 𝑢 𝑡 − 𝜏 → 𝑦(𝑡 − 𝜏)

● We will consider continuous time and discrete time signals (and

systems)

Signals and Systems 4

 Introduction

● And for real systems ?

• LTI approximation generally do not hold
• LTI approximation generally holds for small variations:
𝑈 𝑡 =𝑈+𝑢 𝑡 →𝑌 𝑡 =𝑌+𝑦 𝑡
Relationship between 𝑈 and 𝑌 is NOT LTI
Relationship between 𝑢(𝑡) and 𝑦(𝑡) is LTI (especially if 𝑢(𝑡) and 𝑦(𝑡) are small)

● Examples
• Aircraft flying at 𝑈 = 10,000𝑓𝑡 : we will consider altitude small variations
𝑢 =+/−100𝑓𝑡
• Autocruise at 𝑈 = 100𝑘𝑚/ℎ : we will consider
𝑢 =+/−1𝑘𝑚/ℎ

Signals and Systems 5

 Example of LTI systems

● System modeled by a discrete equation

𝑦 𝑛 + 𝑘𝑦 𝑛 − 1 = 𝑢 𝑛
𝑛 is “time index”
𝑘 parameter models the system
⇒ Easy to check that the system is LTI

● System modeled by a differential equation

𝑑𝑦 𝑡
+ 𝑘𝑦 𝑡 = 𝑢(𝑡)
𝑑𝑡
𝑡 is “continuous time”
𝑘 parameter models the system
⇒ Easy to check that this system is LTI

● With those two examples we cover 99,9% of interesting problems ☺

Signals and Systems 6

 2. Discrete systems: the Z-transform (1785)

Pierre-Simon Laplace (1749–1827).

Signals and Systems 7

 Impulse response of a discrete system

● Suppose we know the impulse response of the system ℎ

𝛿0 𝑛

𝑢 𝑦 0, ∀𝑛 ≠ 0
𝑢 𝑛 = 𝛿0 𝑛 = ቊ
ℎ 1, 𝑛=0 ℎ0 𝑛
⇒ 𝑦 𝑛 = ℎ0 (𝑛)

𝛿0 𝑛 𝛿1 𝑛 𝛿2 𝑛
1. « Time Invariant »
property:
ℎ0 𝑛 ℎ1 𝑛 ℎ2 𝑛

= + + + +⋯

𝑢 𝑛 = 𝑢0 𝛿0 𝑛 + 𝑢1 𝛿1 𝑛 + 𝑢2 𝛿2 𝑛 + 𝑢3 𝛿3 𝑛 +…

2. « Linear » property: = + + + +…

𝑦 𝑛 = 𝑢0 ℎ0 𝑛 + 𝑢1 ℎ1 𝑛 + 𝑢2 ℎ2 𝑛 + 𝑢3 ℎ3 𝑛 +⋯

Signals and Systems 8

 Convolution property

● Any input signal 𝑢 is a combination of impulse signals:

𝑢 = 𝑢0 , 𝑢1 , 𝑢2 , … = 𝑢0 𝛿0 + 𝑢1 𝛿1 + 𝑢2 𝛿2 + ⋯

● Output is a combination of impulse responses (superposition):

𝑦 = {𝑦0 , 𝑦1 , 𝑦2 , … } = 𝑢0 ℎ0 + 𝑢1 ℎ1 + 𝑢2 ℎ2 + ⋯
𝑦 𝑛 = 𝑢0 ℎ0 𝑛 + 𝑢1 ℎ1 𝑛 + 𝑢2 ℎ2 𝑛 + ⋯
= 𝑢0 ℎ0 𝑛 + 𝑢1 ℎ0 𝑛 − 1 + 𝑢2 ℎ0 𝑛 − 2 + ⋯ 𝑢𝑛 ℎ0 0
𝑛

𝑦 𝑛 = ෍𝑢 𝜏 ℎ 𝑛 −𝜏
𝜏=0

● ⇒ The output 𝑦 is the convolution of input 𝑢 and impulse response ℎ

● Any LTI system is fully described by its impulse response ☺
– This is why the system ℎ will be modeled by a function ℎ(𝑛)

Signals and Systems 9

 Z transform (definition and examples)

● Computing the convolution may be complicated. This is why we

introduce the “z-transform”
𝑛=∞

𝑧 ∈ ℂ, 𝑈 𝑧 = ෍ 𝑢 𝑛 𝑧 −𝑛
𝑛=−∞

● Ex. Z-transform of the discrete impulse signal

0, ∀𝑛 ≠ 0
𝛿 𝑛 =ቊ ⇒Δ 𝑧 =1
1, 𝑛 = 0

● Ex. Z-transform of a unit step signal

0, ∀𝑛 < 0 1
𝑢 𝑛 =ቊ ⇒U 𝑧 =
1, 𝑛 ≥ 0 1 − 𝑧 −1

● Ex. Z-transform of an exponential signal

0, ∀𝑛 < 0 1
𝑢 𝑛 =ቊ ⇒ U 𝑧 =
𝛼𝑛, 𝑛 ≥ 0 1 − 𝛼𝑧 −1

● Ex. Z-transform of a sinusoidal signal

Signals and Systems 10
 Z transform (main property)

● Z-transform of the impulse response

ℎ 𝑛 ⇒𝐻 𝑧

● Z-transform of the input signal

𝑢 𝑛 ⇒𝑈 𝑧

● Z-transform of the convolution

𝜏=∞

𝑦 𝑛 = ෍ 𝑢 𝜏 ℎ 𝑛−𝜏
𝜏=0

(𝑑𝑒𝑚𝑜)

𝑌(𝑧) = 𝐻 𝑧 𝑈 𝑧

● The (time domain) convolution is replaced by a (z-domain) multiplication

● Allows to compute transfer function of block-diagram

Signals and Systems 11

 Z transform (Properties 1)

● Definition
𝑛=∞

𝑧 ∈ ℂ, 𝑈 𝑧 = ෍ 𝑢 𝑛 𝑧 −𝑛
𝑛=−∞

● Convergence
– We should always give the Region of Convergence (ROC): the set of points in the complex
plane where the z-transform summation converge…

● Unilateral, bilateral, causality…

– Bilateral vs. Unilateral
𝑛=∞ 𝑛=∞

𝑈 𝑧 = ෍ 𝑢 𝑛 𝑧 −𝑛 𝑈 𝑧 = ෍ 𝑢 𝑛 𝑧 −𝑛
𝑛=−∞ 𝑛=0
– When we use the z-transform for the impulse response of a causal system ℎ(𝑛) is always
equal to 0 for every negative value of 𝑛

● Inverse z-transform
1
𝑢 𝑛 = ර 𝑈 𝑧 𝑧 𝑛−1 𝑑𝑧
2𝜋𝑗 𝐶
– Where C is a counterclockwise path encircling the origin and entirely the ROC

Signals and Systems 13

 Z transform (Properties 2)

● Linearity
𝛼𝑢 𝑛 + 𝛽𝑣 𝑛 ⇒ 𝛼𝑈 𝑧 + 𝛽𝑉 𝑧

● Delay
𝑢 𝑛 − 𝑑 ⇒ 𝑢 −𝑑 + 𝑢 −𝑑 + 1 𝑧 −1 + ⋯ 𝑢 −1 𝑧 −𝑑−1 + 𝑧 −𝑑 𝑈 𝑧

● Delay (with “causal signal” i.e. 𝑢(𝑛) = 0 for 𝑛 < 0)

𝑢 𝑛 − 𝑑 ⇒ 𝑧 −𝑑 𝑈 𝑧

● Convolution
𝑢 𝑛 ∗𝑣 𝑛 ⇒𝑈 𝑧 𝑉 𝑧

● Difference
𝑢 𝑛 − 𝑢 𝑛 − 1 ⇒ 1 − 𝑧 −1 𝑈 𝑧

● Integration
𝑛 1
෍ 𝑢(𝑘) ⇒ 𝑈 𝑧
𝑘=0 1 − 𝑧 −1

Signals and Systems 14

 Z transform (Properties 3)

● Initial value theorem

𝑢 0 = lim 𝑋(𝑧)
𝑧→∞

● Final value theorem

𝑥 ∞ = lim 𝑧 − 1 𝑋 𝑧
𝑧→1

Signals and Systems 15

 Z transform (Table of transform)

Signal 𝑢(𝑛) Z-transform 𝑈(𝑧)

𝛿(𝑛) 1

𝛿(𝑛 − 𝑑) 𝑧 −𝑑
1
𝑢(𝑛)
1 − 𝑧 −1
1
𝛼 𝑛 𝑢(𝑛)
1 − 𝛼𝑧 −1
𝑧 −1
𝑛𝑢(𝑛)
1 − 𝑧 −1 2

1 − cos 𝜔 𝑧 −1
cos 𝜔𝑛 𝑢(𝑛)
1 − 2 cos 𝜔 𝑧 −1 + 𝑧 −2
sin 𝜔 𝑧 −1
sin 𝜔𝑛 𝑢(𝑛)
1 − 2 cos 𝜔 𝑧 −1 + 𝑧 −2
1 − 𝑎 cos 𝜔 𝑧 −1
an cos 𝜔𝑛 𝑢(𝑛)
1 − 2 𝑎 cos 𝜔 𝑧 −1 + 𝑎2 𝑧 −2
𝑎 sin 𝜔 𝑧 −1
an sin 𝜔𝑛 𝑢(𝑛)
1 − 2 𝑎 cos 𝜔 𝑧 −1 + 𝑎2 𝑧 −2

Signals and Systems 16

 Discrete transfer function, poles and zeros

• If the system is LTI it can be represented by a transfer function

𝑌 𝑧 𝑏1 𝑧 −1 +. . +𝑏𝑚 𝑧 −𝑚
=𝐹 𝑧 =
𝑈(𝑧) 1 + 𝑎1 𝑧 −1 +. . +𝑎𝑛 𝑧 −𝑛
• Transfer functions are always ratio of two polynomials. They can be decomposed by partial
fraction expansion:
ς𝑚𝑖=1 1 − 𝑧𝑖 𝑧
−1
𝐹 𝑧 = 𝑛
ς𝑖=1 1 − 𝑝𝑖 𝑧 −1
• 𝑧𝑖 are the zeros, 𝑝𝑖 are the poles. They can be either distinct or multiples. In case of distinct
poles the transfer function can be decomposed:
𝑛 𝐶𝑖
𝐹 𝑧 =෍ −1
𝑖=1 1 − 𝑝𝑖 𝑧
• Poles are either real or complex conjugate by pairs. The decomposition finally gives a sum
of first and second orders:
𝐶𝑖 𝐶𝑖
𝐹 𝑠 =෍ + ෍
1 − 𝑝𝑖 𝑧 −1 1 + 𝛼𝑖 𝑧 −1 + 𝛽𝑖 𝑧 −2

Signals and Systems 17

 Discrete transfer function, stability

• Stability depends on poles

𝑛 𝐶𝑖
𝐹 𝑧 =෍ −1
𝑖=1 1 − 𝑝𝑖 𝑧
• Impulse response of 𝑓 is a sum of exponential:
𝑛 𝑛
𝑓 𝑛 =෍ 𝐶𝑖 𝑝𝑖
𝑖=1
• If one of the poles has radius greater than 1 the system is unstable!

𝐼𝑚𝑎𝑔
Stability region

𝑅𝑒𝑎𝑙
1

Signals and Systems 18

 3. Continuous systems: the Laplace transform

Signals and Systems 19

 Impulse response of a continuous system

● Suppose we know the impulse response of the system ℎ

0, ∀𝑡 ≠ 0
𝑢 𝑡 =𝛿 𝑡 =ቊ 𝑤𝑖𝑡ℎ න 𝛿 𝜏 𝑑𝜏 = 1
∞, 𝑡 = 0

𝑦 𝑡 = ℎ(𝑡)

● Any input signal 𝑢 𝑡 is a combination of impulse signals

● Output is a combination of impulse responses
𝜏=∞

𝑦 𝑡 = න 𝑢 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏
𝜏=0

● Output is the convolution of input and impulse response

● Any LTI system is fully described by its impulse response ☺
– This is why the system ℎ will be modeled by a function ℎ(𝜏)

Signals and Systems 20

 Laplace transform (1)

● Computing the convolution may be complicated. This why we introduce

the “Laplace-transform”
𝑡=∞

𝑠 ∈ ℂ, 𝑈 𝑠 = න 𝑢 𝑡 𝑒 −𝑠𝑡 𝑑𝑡
𝑡=0−

● Ex. Laplace transform of the continuous impulse signal

𝑡=∞
0, ∀𝑡 ≠ 0
𝛿 𝑡 =ቊ 𝑤𝑖𝑡ℎ න 𝛿 𝑡 𝑑𝑡 = 1 ⇒ Δ 𝑠 = 1
∞, 𝑡 = 0
𝑡=0−

● Ex. Laplace transform of a unit step signal

0, ∀𝑡 < 0 1
𝑢 𝑡 =ቊ ⇒U 𝑠 =
1, 𝑡 ≥ 0 𝑠

● Ex. Laplace transform of an exponential signal

0, ∀𝑛 < 0 𝜏
𝑢 𝑛 =ቐ 𝑡 ⇒U 𝑠 =
exp(− ), 𝑡 ≥ 0 1 + 𝜏𝑠
𝜏

Signals and Systems 21

 Laplace transform (2)

● Laplace-transform of the impulse response

ℎ 𝑡 ⇒𝐻 𝑠

● Laplace-transform of the input signal

𝑢 𝑡 ⇒𝑈 𝑠

● Laplace transform of the convolution

𝜏=∞

𝑦 𝑡 = න 𝑢 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏
𝜏=0

(𝑑𝑒𝑚𝑜)

𝑌(𝑠) = 𝐻 𝑠 𝑈 𝑠

● The (time domain) convolution is replaced by a (s-domain) multiplication

● Allows to compute transfer function of block-diagram

Signals and Systems 22

 Laplace transform (Properties 1)

● Definition
𝑡=∞

𝑠 ∈ ℂ, 𝑈 𝑠 = න 𝑢 𝑡 𝑒 −𝑠𝑡 𝑑𝑡
𝑡=0−

● Convergence
– The set of values for which U(𝑠) converges absolutely is for 𝑅𝑒 𝑠 > 𝑎 where 𝑎 is a
constant…

● Inverse Laplace transform

𝛾+𝑗𝑇
1
𝑢 𝑡 = lim න 𝑒 𝑠𝑡 𝑈 𝑠 𝑑𝑠
2𝜋𝑗 𝑇→∞ 𝛾−𝑗𝑇
– Where 𝛾 is a real number so that the contour path of integration is the ROC of 𝑈(𝑠)…

Signals and Systems 24

 Laplace transform (Properties 2)

● Linearity
𝛼𝑢 𝑡 + 𝛽𝑣 𝑡 ⇒ 𝛼𝑈 𝑠 + 𝛽𝑉 𝑠

● Delay
𝑢 𝑡 − 𝑡0 ⇒ 𝑒 −𝑡0 𝑠 𝑈 𝑠

● Convolution
𝑢 𝑡 ∗𝑣 𝑡 ⇒𝑈 𝑠 𝑉 𝑠
Derivation

● Derivation
𝑑𝑢 𝑡
⇒ 𝑠𝑈 𝑠 − 𝑢(0− )
𝑑𝑡
● Integration Pre-initial value
𝑡
1
න 𝑢 𝜏 𝑑𝜏 ⇒ 𝐹 𝑠
0 𝑠

Signals and Systems 25

 Laplace transform (Properties 3)

● Post Initial value theorem

𝑢 0+ = lim 𝑠𝑈(𝑠)
𝑠→∞

● Final value theorem (if exists)

𝑢 ∞ = lim 𝑠𝑈(𝑠)
𝑠→0

Signals and Systems 26

 Laplace transform (Table of transform)

Signal 𝑢(𝑡) Laplace-transform 𝑈(𝑠)

𝛿 𝑡 (unit impulse) 1

𝛿(𝑡 − 𝑡0 ) 𝑒 −𝑠𝑡0
1
𝑢(𝑛) (unit step)
𝑠
1
𝑡𝑢(𝑡)
𝑠2
1
𝑒 −𝑎𝑡 𝑢 𝑡
𝑠+𝑎
𝜔
sin 𝜔𝑡 𝑢 𝑡
𝑠2 + 𝜔2
𝑠
cos 𝜔𝑡 𝑢 𝑡
𝑠2 + 𝜔2
𝜔
𝑒 −𝛼𝑡 sin 𝜔𝑡
𝑠 + 𝛼 2 + 𝜔2
𝑠+𝑎
𝑒 −𝛼𝑡 cos 𝜔𝑡
𝑠 + 𝛼 2 + 𝜔2

Signals and Systems 27

 Continuous transfer function, poles and
zeros

• If the system is LTI it can be represented by a transfer function

𝑌 𝑠 𝑏𝑚 𝑠 𝑚 + ⋯ + 𝑏1 𝑠 + 𝑏0
=𝐹 𝑠 =
𝑈(𝑠) 𝑎𝑛 𝑠 𝑛 + ⋯ + 𝑎1 𝑠 + 𝑠
• Causal system: 𝑚 ≤ 𝑛 (not demonstrated)
• Transfer functions are always ratio of two polynomials. They can be decomposed by partial
fraction expansion:
ς𝑚𝑖=1 𝑠 − 𝑧𝑖
𝐹 𝑠 = 𝑛
ς𝑖=1 𝑠 − 𝑝𝑖
• 𝑧𝑖 are the zeros, 𝑝𝑖 are the poles. They can be either distinct or multiples. In case of distinct poles
the transfer function can be decomposed:
𝑛 𝐶𝑖
𝐹 𝑠 =෍
𝑖=1 𝑠 − 𝑝𝑖
• Poles are either real or complex conjugate by pairs. The decomposition finally gives a sum of first
and second orders:
𝐶𝑖 𝐶𝑖
𝐹 𝑠 =෍ +෍ 2
𝑠 − 𝑝𝑖 𝑠 + 2𝜎𝜔𝑠 + 𝜔 2

This example catches most practical examples. To be more precise read the Partial Fraction
Decomposition page of Wikipedia
https://en.wikipedia.org/wiki/Partial_fraction_decomposition

Signals and Systems 28

 Continuous transfer function, stability

• Stability depends on poles

𝑛 𝐶𝑖
𝐹 𝑠 =෍
𝑖=1 𝑠 − 𝑝𝑖

• Impulse response of 𝑓 is a sum of exponential:

𝑛
𝑓 𝑡 =෍ 𝐶𝑖 𝑒 𝑝𝑖 𝑡
𝑖=1
• If one of the poles has positive real part the system is unstable!
• If 𝑝𝑖 is small the system is slow
𝐼𝑚𝑎𝑔
• If 𝑝𝑖 is large the system is fast

Stability region
𝑅𝑒𝑎𝑙

Signals and Systems 29

 4. Laplace and Z-transform: in practice

Signals and Systems 30

 Summary

● A very high level hypothesis

– The system is Linear Time Invariant

● A very nice property

– The system is fully defined by its “impulse response”

● A not so nice property

– The output signal can be computed by convolution of input signal and impulse response

● A very good new

– Using Z-transform (discrete time) or Laplace transform (continuous time) makes everything
simple

Signals and Systems 31

 From differential equation to Laplace or Z
transform
● Many systems are modeled by differential equations
𝑑𝑛 𝑦(𝑡) 𝑑𝑛−1 𝑦(𝑡) 𝑑𝑦 𝑡 𝑑𝑚−1 𝑢(𝑡) 𝑑𝑢 𝑡
+ 𝑎𝑛−1 +. . 𝑎1 + 𝑎0 𝑦 𝑡 = 𝑏𝑛−1 +. . 𝑏0 + 𝑏0 𝑢 𝑡
𝑑𝑡 𝑛 𝑑𝑡 𝑛−1 𝑑𝑡 𝑑𝑡 𝑚−1 𝑑𝑡 (1)
● Or by discrete recurrent equation
𝑦 𝑘 + 𝑎𝑛−1 𝑦 𝑘 − 1 + ⋯ + 𝑎1 𝑦 𝑘 − 𝑛 − 1 + 𝑎𝑛 𝑦 𝑘 − 𝑛
= 𝑏1 𝑦 𝑘 − 1 + ⋯ + 𝑎𝑚+1 𝑢 𝑘 − 𝑚 − 1 + 𝑏𝑚 𝑦 𝑘 − 𝑚
(2)
● Transfer function
– Condition : coefficients 𝑎𝑖 and 𝑏𝑖 are constants (LTI system)
– Continuous transfer function (from (1))
𝑏𝑚−1 𝑠 𝑚−1 + ⋯ 𝑏1 𝑠 + 𝑏0
𝐹 𝑠 = 𝑛
𝑠 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ 𝑎1 𝑠 + 𝑎0
– Discrete transfer function (from (2))
𝑏1 𝑧 −1 +. . +𝑏𝑚 𝑧 −𝑚
𝐹 𝑧 =
1 + 𝑎1 𝑧 −1 +. . +𝑎𝑛 𝑧 −𝑛

● Stability
– Poles says if the system is stable or not (real part for 𝑠, radius for 𝑧)

Signals and Systems 32

 From differential equation to Laplace or Z
transform
● Causality
𝑚 ≤ 𝑛 (continuous case)
𝑏𝑖 = 0 𝑓𝑜𝑟 𝑖 ≤ 0 (discrete case)

● Initial conditions (example)

𝑑2 𝑦 𝑡 𝑑𝑦 𝑡 𝑑𝑢 𝑡
+ 𝑎1 + 𝑎0 𝑦 𝑡 = 𝑏1 + 𝑏0 𝑢 𝑡
𝑑𝑡 2 𝑑𝑡 𝑑𝑡
Laplace transform (derivation theorem)
𝑠 2 𝑌 𝑠 − 𝑠 𝑦 0− − 𝑦(0
ሶ − ) + 𝑎1 𝑠𝑌 𝑠 − 𝑦(0− ) + 𝑎0 𝑌 𝑠 = 𝑏1 𝑠𝑈 𝑠 − 𝑢 0− + 𝑏0 𝑈 𝑠

𝑏1 𝑠 + 𝑏0 𝑠𝑦 0− + 𝑦ሶ 0− + 𝑎1 𝑦 0− − 𝑏0 𝑢(0− )
𝑌 𝑠 = 2 𝑈 𝑠 +
𝑠 + 𝑎1 𝑠 + 𝑎0 𝑠 2 + 𝑎1 𝑠 + 𝑎0

Transfer function Transient response

(depends on initial conditions)

Transfer function is enough when initial condition are equal to zero ☺

Signals and Systems 33

 Solving differential equations thanks to
transfer function
● From differential equation to transfer function
𝑑2 𝑦 𝑡 𝑑𝑦 𝑡 𝑑𝑢 𝑡
+ 𝑎1 + 𝑎0 𝑦 𝑡 = 𝑏0 + 𝑏0 𝑢 𝑡
𝑑𝑡 2 𝑑𝑡 𝑑𝑡
𝑏1 𝑠 + 𝑏0 𝑠𝑦 0− + 𝑦ሶ 0− + 𝑎1 𝑦 0− − 𝑏0 𝑢(0− )
𝑌 𝑠 = 2 𝑈 𝑠 +
𝑠 + 𝑎1 𝑠 + 𝑎0 𝑠 2 + 𝑎1 𝑠 + 𝑎0

𝑢 𝑡 𝑦 0 , 𝑦ሶ 0 , 𝑢 0
𝑦 𝑡
𝑎1 , 𝑎0 , 𝑏1
Laplace Inverse
Transform Laplace Transform

𝑈 𝑠 𝑌 𝑠

Signals and Systems 34

 Block Diagrams

• Since systems are LTI, interconnection is easily represented by block

diagrams connections : additions, subtractions, etc.
• See exercices ☺
• Sum
• Interconnexion
• commutativity

Signals and Systems 35

 5. Frequency domain

Jean-Baptiste Joseph Fourier (1768 - 1830)

Signals and Systems 36

 Frequency response of a continuous
system

● Response of an exponential input

𝑢 𝑡 = 𝑒 𝑠𝑡
+∞
𝑦(𝑡) = න ℎ 𝜏 𝑢 𝑡 − 𝜏 𝑑𝜏
−∞
𝑌(𝑠) = 𝐻 𝑠 𝑒 𝑠𝑡

● Response of a sinusoidal input:

𝑢 𝑡 = 𝐴𝑐𝑜𝑠 𝜔𝑡
𝐴 𝑖𝜔𝑡
= 𝑒 + 𝑒 −𝑖𝜔𝑡
2
𝐴
𝑦 𝑡 = 𝐻 𝑖𝜔 𝑒 𝑖𝜔𝑡 + 𝐻 −𝑖𝜔 𝑒 −𝑖𝜔𝑡
2
Let 𝐻 𝑖𝜔 = 𝑀 𝜔 𝑒 𝑖𝜙 𝜔

𝑦 𝑡 = 𝐴 × 𝑀 𝜔 cos 𝜔𝑡 + 𝜙 𝜔
Output is also sinusoidal but:
• Amplitude amplified by magnitude of 𝐇 𝐢𝛚
• Phase shifted by phase of 𝑯 𝒊𝝎
• Input was sinusoidal forever: not transient response

Signals and Systems 37

 Fourier Transform

• Definition of the Fourier Transform

∞
𝑓 ∈ ℝ, 𝑈 𝜔 = න 𝑢 𝑡 exp −𝑖2𝜋𝑓𝑡 𝑑𝑡
−∞

• Inverse Fourier Transform

∞
𝑢 𝑡 = න 𝑈 𝜔 exp 𝑖2𝜋𝑓𝑡 𝑑𝑓
−∞

• Relationship with the Laplace transform

𝑠 replaced with 𝑗𝜔

• 𝐹(𝑖𝜔) is often called “spectrum” (variation with respect to frequency)

Signals and Systems 38

 Bode diagram

● Bode diagram
– Vertical axis:
magnitude of 𝐻(𝑖𝜔) in dB
Phase of 𝐻(𝑖𝜔) in 𝑑𝑒𝑔
– Horizontal axis: frequency 𝑓 (𝐻𝑧) or pulsation 𝜔 (𝑟𝑎𝑑. 𝑠 −1 ) in log scale

20 log10 𝐻 𝑖𝜔

arg 𝐻 𝑖𝜔

Signals and Systems 39

 Bode diagram

● Example of Bode Diagram plots

𝜔𝑐
– 𝐹 𝑠 =
𝑠
1
– 𝐹 𝑠 = Real / asymptotic
1+𝜏𝑠
1
– 𝐹 𝑠 = 2𝜎 𝑠2
Real / asymptotic
1+ +
𝜔0 `𝜔2
0

– Combinations

Signals and Systems 40

 Time domain vs frequency domain

Signals and Systems 41

 6. Continuous and discrete systems

Signals and Systems 42

 Continuous and discrete systems

• The real world

• In most application the physical system is known by continuous differential equations ⇒
Laplace transform
• In most real application the system will be controlled by a computer ⇒ discrete equations
⇒ z-transform

• A continuous signal will be sampled via a “Analog to Digital Converter”

(ADC)
• A discrete signal (computed every 𝑇𝑠 by the computer) will be converted
into a continuous signal via a “Digital to Analog Converter” (DAC)

Signals and Systems 43

 Sampling a continuous signal, time
domain

• Sampling a continuous signal is a multiplication by a train of impulses

𝑇
𝑦(𝑡) 𝑦𝑠 (𝑡)
× =

𝑘=+∞

𝑦𝑠 𝑡 = 𝑦 𝑡 × ෍ 𝛿 𝑡 − 𝑘𝑇
𝑘=−∞
𝑘=+∞

= ෍ 𝑦 𝑘𝑡 𝛿 𝑡 − 𝑘𝑇
𝑘=−∞

Signals and Systems 44

 Sampling a continuous signal, frequency
domain

• The time domain multiplication gives a frequency domain convolution

𝑘=+∞ +∞
1
𝑦𝑠 𝑡 = ෍ 𝑦 𝑘𝑡 𝛿 𝑡 − 𝑘𝑇 ⇒ 𝑌𝑠 𝑗𝜔 = ෍ 𝑌 𝑖 𝜔 − 𝑘𝜔𝑠
𝑇𝑠
𝑘=−∞ 𝑘=−∞
The sampled spectrum is made of replication (aliases) of the continuous spectrum
2𝜋
every 𝑇
𝑠 𝑌 𝑖 2𝜋𝑓 𝑌𝑠 𝑖 2𝜋𝑓

1/𝑇𝑠 ok

1/𝑇𝑠 too small

Signals and Systems 45

 Choice of sampling time

• Shannon theorem
• If the signal spectrum is limited in a band [0 𝑓𝑚𝑎𝑥 ] the sampling frequency must be chosen
such as 𝑓𝑠 > 2 × 𝑓𝑚𝑎𝑥

• If the sampling rate is large enough the continuous signal it can be fully
reconstructed by correctly filtering the sampled signal
• If the sampling rate is too small there is loss of information by aliasing

Signals and Systems 46

 Digital to Analog conversion

• Theoretically the continuous signal can be reconstructed by filtering the

discrete signal with a 0 𝑓𝑚𝑎𝑥 filter
• But this is ok only the full signal is available (infinite number of samples)

• In practical applications the continuous signal is reconstructed in real

time via a zero-holder hold (ZOH)

𝑦𝑠 (𝑡) 𝑦𝑠 (𝑡)

ZOH

Signals and Systems 47