Introduction to Signals and Systems

Lathi Chapt. 1

Gloria Menegaz 1
 Didactic material

• Textbook
– Signal Processing and Linear Systems, B.P. Lathi, CRC Press
• Other books
– Signals and Systems, Richard Baraniuk’s lecture notes, available on line
– Digital Signal Processing (4th Edition) (Hardcover), John G. Proakis, Dimitris K
Manolakis
– Teoria dei segnali analogici, M. Luise, G.M. Vitetta, A.A. D’Amico, McGraw-Hill
– Signal processing and linear systems, Schaun's outline of digital signal
processing
• All textbooks are available at the library
• Handwritten notes will be available on demand

Gloria Menegaz 2
 Signals&Systems

Input signal Output signal

System

amplitude Linear time invariant |amplitude|

systems (LTIS)

time
frequency

LTIS perform any kind

of processing on the
input data to generate
output data

Gloria Menegaz 3
 Contents

Signals Systems
• Signal classification and • Linear Time-Invariant Systems
representation – Time and frequency domain analysis
– Types of signals – Impulse response
– Sampling theory – Stability criteria
– Quantization • Digital filters
• Signal analysis – Finite Impulse Response (FIR)
– Fourier Transform
• Mathematical tools
§ Continuous time, Fourier series,
Discrete Time Fourier Transforms, – Laplace Transform
Windowed FT § Basics
– Spectral Analysis – Z-Transform
§ Basics

Applications in the domain of Bioinformatics

Gloria Menegaz 4
 What is a signal?

• A signal is a set of information of data

– Any kind of physical variable subject to variations represents a signal
– Both the independent variable and the physical variable can be either scalars or
vectors
§ Independent variable: time (t), space (x, x=[x1,x2], x=[x1,x2,x3])
§ Signal:
§ Electrochardiography signal (EEG) 1D, voice 1D, music 1D
§ Images (2D), video sequences (2D+time), volumetric data (3D)

Gloria Menegaz 5
 What is a system?

• Systems process signals to

– Extract information
– Enable transmission over channels with limited capacity (JPEG, JPEG2000,
MPEG coding)
– Improve security over networks (encryption, watermarking)
– Support the formulation of diagnosis and treatment planning (medical imaging)
– …….
closed-loop

The function linking the output

of the system with the input
signal is called transfer function
input output and it is typically indicated with
System the symbol h(•)

Gloria Menegaz 6
 Classification of signals

• Continuous time – Discrete time

• Analog – Digital (numerical)
• Periodic – Aperiodic
• Energy – Power
• Deterministic – Random (probabilistic)
• Note
– Such classes are not disjoint, so there are digital signals that are periodic of
power type and others that are aperiodic of power type etc.
– Any combination of single features from the different classes is possible

Gloria Menegaz 7
 Continuous time – discrete time

• Continuous time signal: a signal that is specified for every real value of the
independent variable
– The independent variable is continuous, that is it takes any value on the real axis
– The domain of the function representing the signal has the cardinality of real
numbers
§ Signal ↔ f=f(t)
§ Independent variable ↔ time (t), position (x)
§ For continuous-time signals: t∈

amplitude

time

Gloria Menegaz 8
 Continuous time – discrete time

• Discrete time signal: a signal that is specified only for discrete values of the
independent variable
– It is usually generated by sampling so it will only have values at equally spaced
intervals along the time axis
– The domain of the function representing the signal has the cardinality of integer
numbers
§ Signal ↔ f=f[n], also called “sequence”
§ Independent variable ↔ n amplitude
§ For discrete-time functions: t Î Z

time (discrete)

Gloria Menegaz 9
 Analog - Digital

• Analog signal: signal whose amplitude can take on any value in a

continuous range
– The amplitude of the function f(t) (or f(x)) has the cardinality of real numbers
§ The difference between analog and digital is similar to the difference between
continuous-time and discrete-time. In this case, however, the difference is with respect
to the value of the function (y-axis)
– Analog corresponds to a continuous y-axis, while digital corresponds to a
discrete y-axis

• Here we call digital what we have called quantized in the EI class

• An analog signal can be both continuous time and discrete time
Gloria Menegaz 10
 Analog - Digital

• Digital signal: a signal is one whose amplitude can take on only a finite
number of values (thus it is quantized)
– The amplitude of the function f() can take only a finite number of values
– A digital signal whose amplitude can take only M different values is said to be M-
ary
§ Binary signals are a special case for M=2
amplitude

time

Gloria Menegaz 11
 Example

– Continuous time analog

amplitude

time

– Continuous time digital (or quantized)

§ binary sequence, where the values of the function can only be one or zero.
amplitude

time

Gloria Menegaz 12
 Example

• Discrete time analog

amplitude

time
• Discrete time digital
§ binary sequence, where the values of the function can only be one or zero.
amplitude

time

Gloria Menegaz 13
 Summary

Signal amplitude/
Real Integer
Time or space

Analog Digital
Real
Continuous-time Continuous-time

Analog Digital
Integer
Discrete-time Discrete time

Gloria Menegaz 14
 Note

• In the image processing class we have defined as digital those signals that
are both quantized and discrete time. It is a more restricted definition.
• The definition used here is as in the Lathi book.

Gloria Menegaz 15
 Periodic - Aperiodic

• A signal f(t) is periodic if there exists a positive constant T0 such that

f (t + T0 ) = f (t ) "t

– The smallest value of T0 which satisfies such relation is said the period of the
function f(t)
– A periodic signal remains unchanged when time-shifted of integer multiples of
the period
– Therefore, by definition, it starts at minus infinity and lasts forever
-¥ £ t £ +¥ t ΰ
-¥ £ n £ +¥ nÎZ

– Periodic signals can be generated by periodical extension

Gloria Menegaz 16
 Examples

• Periodic signal with period T0

• Aperiodic signal

Gloria Menegaz 17
 Causal and non-Causal signals

• Causal signals are signals that are

zero for all negative time (or spatial
positions), while

• Anticausal are signals that are zero for

all positive time (or spatial positions).

• Noncausal signals are signals that

have nonzero values in both positive
and negative time

Gloria Menegaz 18
 Causal and non-causal signals

• Causal signals

f (t ) = 0 t <0

• Anticausals signals

f (t ) = 0 t ³0

• Non-causal signals

$t1 < 0 : f (t1 ) = 0

Gloria Menegaz 19
 Even and Odd signals

• An even signal is any signal f such that f (t) = f (-t). Even signals can be
easily spotted as they are symmetric around the vertical axis.

• An odd signal, on the other hand, is a signal f such that f (t)= - (f (-t))

Gloria Menegaz 20
 Decomposition in even and odd components

• Any signal can be written as a combination of an even and an odd signals

– Even and odd components

1 1
f (t ) = ( ( ) ( ) ) ( f ( t ) - f ( -t ) )
f t + f - t +
2 2
1
f e ( t ) = ( f ( t ) + f ( -t ) ) even component
2
1
f o ( t ) = ( f ( t ) - f ( -t ) ) odd component
2
f (t ) = fe (t ) + fo (t )

Gloria Menegaz 21
Example

Gloria Menegaz 22
 Example

• Proof

Gloria Menegaz 23
 Some properties of even and odd functions

• even function x odd function = odd function

• odd function x odd function = even function
• even function x even function = even function
• Area
a a

∫ f (t ) dt = 2 ∫ f (t ) dt
e e
−a 0
a

∫ f (t ) dt = 0
o
−a

Gloria Menegaz 24
 Deterministic - Probabilistic

• Deterministic signal: a signal whose • Probabilistic (or random) signals: the

physical description in known amplitude values cannot be predicted
completely precisely but are known only in terms
of probabilistic descriptors
• A deterministic signal is a signal in
which each value of the signal is fixed • The future values of a random signal
and can be determined by a cannot be accurately predicted and
mathematical expression, rule, or can usually only be guessed based on
table. the averages of sets of signals
• Because of this the future values of – They are realization of a stochastic
the signal can be calculated from past process for which a model could be
available
values with complete confidence.
– Examples: EEG, evocated potentials,
– There is no uncertainty about its noise in CCD capture devices for digital
amplitude values cameras
– Examples: signals defined through a
mathematical function or graph

Gloria Menegaz 25
 Example

• Deterministic signal
amplitude

time
• Random signal
amplitude

time

Gloria Menegaz 26
 Finite and Infinite length signals

• A finite length signal is non-zero over a finite set of values of the

independent variable

f = f ( t ) , "t : t1 £ t £ t2
t1 > -¥, t2 < +¥

• An infinite length signal is non zero over an infinite set of values of the
independent variable
– For instance, a sinusoid f(t)=sin(ωt) is an infinite length signal

Gloria Menegaz 27
 Size of a signal: Norms

• "Size" indicates largeness or strength.

• We will use the mathematical concept of the norm to quantify this notion for
both continuous-time and discrete-time signals.
• The energy is represented by the area under the curve (of the squared
signal)
amplitude

0 T time

Gloria Menegaz 28
 Energy

• Signal energy
+¥
Ef = ò
-¥
f 2 (t )dt

+¥

ò
2
Ef = f (t ) dt
-¥

• Generalized energy : Lp norm

– For p=2 we get the energy (L2 norm)

( ò ( f (t ) ) dt )
1/ p
f (t ) =
p

1 £ p < +¥

Gloria Menegaz 29
 Power

• Power
– The power is the time average (mean) of the squared signal amplitude, that is
the mean-squared value of f(t)

+T / 2
1
Pf = lim ò
2
f (t ) dt
T ®¥ T
-T / 2
+T / 2
1
T ®¥ T ò
2
Pf = lim f (t ) dt
-T / 2

Gloria Menegaz 30
 Power - Energy

• The square root of the power is the root mean square (rms) value
– This is a very important quantity as it is the most widespread measure of
similarity/dissimilarity among signals
– It is the basis for the definition of the Signal to Noise Ratio (SNR)
æ Psignal ö
SNR = 20log10 ç ÷÷
ç Pnoise
è ø
– It is such that a constant signal whose amplitude is =rms holds the same power
content of the signal itself
• There exists signals for which neither the energy nor the power are finite

f
ramp

0 t
Gloria Menegaz 31
 Energy and Power signals

• A signal with finite energy is an energy signal

– Necessary condition for a signal to be of energy type is that the amplitude goes
to zero as the independent variable tends to infinity
• A signal with finite and different from zero power is a power signal
– The mean of an entity averaged over an infinite interval exists if either the entity
is periodic or it has some statistical regularity
– A power signal has infinite energy and an energy signal has zero power
– There exist signals that are neither power nor energy, such as the ramp
• All practical signals have finite energy and thus are energy signals
– It is impossible to generate a real power signal because this would have infinite
duration and infinite energy, which is not doable.

Gloria Menegaz 32
Useful signal operations: shifting, scaling, inversion

• Shifting: consider a signal f(t) and the same signal delayed/anticipated by T

seconds f(t)
T>0

t
f(t+T)

anticipated

t
T
delayed
f(t-T)

t
Gloria Menegaz 33
Useful signal operations: shifting, scaling, inversion

• (Time) Scaling: compression or expansion of a signal in time

f(t)

f(2t) t

compression j ( t ) = f ( 2t )

f(t/2) t

expansion j (t ) = f (t / 2)

Gloria Menegaz 34
Useful signal operations: shifting, scaling, inversion

• Scaling: generalization

a >1
j ( t ) = f ( at ) ® compressed version
ætö
j ( t ) = f ç ÷ ® dilated (or expanded) version
èaø
Viceversa for a < 1

Gloria Menegaz 35
Useful signal operations: shifting, scaling, inversion

• (Time) inversion: mirror image of f(t) about the vertical axis

j ( t ) = f ( -t )
f(t)

f(-t)

Gloria Menegaz 36
Useful signal operations: shifting, scaling, inversion
qui
• Combined operations: f(t) → f(at-b)
• Two possible sequences of operations
1. Time shift f(t) by to obtain f(t-b). Now time scale the shifted signal f(t-b) by a to
obtain f(at-b).
2. Time scale f(t) by a to obtain f(at). Now time shift f(at) by b/a to obtain f(at-b).
• Note that you have to replace t by (t-b/a) to obtain f(at-b) from f(at) when replacing t by
the translated argument (namely t-b/a))

Gloria Menegaz 37
 Useful functions

• The unit step function

– Useful for representing causal signals

ì1 t ³ 0
u (t ) = í
î0 t < 0

f (t ) = u (t - 2) - u (t - 4)
Gloria Menegaz 38
 Unit step function

Gloria Menegaz 39
 Useful functions

• Continuous and discrete time unit step functions

u(t) u[k]

k=0 1 2 3…

Gloria Menegaz 40
 Useful functions

• The ramp function (continuous time)

Gloria Menegaz 41
 Useful functions

• The unit impulse function

(t) = 0, t 6= 0
Z +1
(t)dt = 1
1

δ(t) 1/ε ε→0

0 t t
-ε/2 ε/2
The function k 𝛿(t) is zero for all t different from zero and has area equal to k

Gloria Menegaz 42
 Properties of the unit impulse function

• Multiplication of a function by impulse

f (t )d (t ) = f ( 0)d (t )
f ( t ) d ( t - T ) = f (T ) d ( t - T )

• Sampling property of the unit function

+¥ +¥ +¥

ò f ( t ) d ( t ) dt = ò f ( 0 ) d ( t ) dt = f ( 0 ) ò d ( t ) dt = f ( 0 )
-¥ -¥ -¥
+¥

ò f ( t ) d ( t - T ) dt = f (T )
-¥

– The area under the curve obtained by the product of the unit impulse function
shifted by T and j(t) is the value of the function j(t) for t=T

Gloria Menegaz 43
 Properties of the unit impulse function

• The unit step function is the integral of the unit impulse function

du
= d (t )
dt
t

ò d ( t ) dt = u ( t )
-¥

– Thus

ì0 t < 0
t

ò-¥ d ( t ) dt = u ( t ) = íî1 t ³ 0

Gloria Menegaz 44
 Properties of the unit impulse function

• Discrete time impulse function

Gloria Menegaz 45
 Useful functions

• The continuous time complex exponential

f ( t ) = Ae jwt

• Euler’s relations

• Discrete time complex exponential

– k=nT

Gloria Menegaz 46
 Useful functions

• Exponential function est

– Generalization of the function ejωt

s = s + jw

Gloria Menegaz 47
 The exponential function

• Special cases
• A constant K = K ⇥ exp 0
• A monotonic exponential exp t (! = 0, s = )
• A trigonometric function cos(!t).(s = 0, = +/ j!)
• An exponentially varying trigonometric function

Gloria Menegaz 48
 The exponential function

s=σ s=jω

s=σ+jω s=σ+jω

Gloria Menegaz 49
 Complex frequency plan
signals of constant amplitude
jω

right half plan

left half plan exponentially
exponentially increasing signals
decreasing signals

monotonically
increasing/decreasing
σ exponentials

Gloria Menegaz 50
Linear Systems
 Systems

• A system is characterized by
– inputs
– outputs
– rules of operation (mathematical model of the system)

f1(t) y1(t)
f2(t) y2(t)

fn(t) yn(t)

inputs outputs

Gloria Menegaz 52
 Systems

• Study of systems: mathematical modeling, analysis, design

– Analysis: how to determine the system output given the input and the system
mathematical model
– Design or synthesis: how to design a system that will produce the desired set of
outputs for given inputs

• SISO: single input single output - MIMO: multiple input multiple output

f1(t) y1(t)
f2(t) y2(t)
f1(t) y1(t)

fn(t) yn(t)

inputs outputs inputs outputs

Gloria Menegaz 53
 Classification of systems

1. Linear and non linear

2. Constant parameters and time-varying parameters
3. Instantaneous (memoryless) and dynamic (with memory)
4. Causal and non-causal
5. Lumped-parameters and distributed-parameters
6. Continuous-time and discrete-time
7. Analog and digital

Gloria Menegaz 54
 Linear and non-linear systems

• Additivity
f1 ! y1 and f2 ! y2 then f 1 + f 2 ! y1 + y 2

• Homogeneity (scaling)
f 1 ! y1 then a 1 ⇥ f 1 ! a 1 ⇥ y1
• Superposition
a 1 ⇥ f 1 + a 2 ⇥ f 2 ! a 1 ⇥ y1 + a 2 ⇥ y 2

Gloria Menegaz 55
 Response of a linear system

• Total response = Zero-input response + Zero-state response

– The output of a system for t≥0 is the result of two independent causes: the initial
conditions of the system (or system state) at t=0 and the input f(t) for t≥0.
– Because of linearity, the total response is the sum of the responses due to
those two causes
– The zero-input response is only due to the initial conditions and the zero-state
response is only due to the input signal
– This is called decomposition property
• Real systems are locally linear
– Respond linearly to small signals and non-linearly to large signals
causal, linear causal, non linear
y y
locally linear
around f0

f f2 f1 f0 f
Gloria Menegaz 56
 Response of a linear system

• Because of linearity we can evaluate separately the two components of the

output and calculate the total response as their sum.
– The zero input component can be computed assuming the input to be zero and
– the zero-state component can be computed assuming the initial conditions to be
zero
• If the input can be expressed as the sum of simple components, the output
can be obtained as the sum of the responses to each component

f (t) = a1 ⇥ f1 (t) + a2 ⇥ f2 (t) + · · · + an ⇥ fn (t)

y(t) = a1 ⇥ y1 (t) + a2 ⇥ y2 (t) + · · · + an ⇥ yn (t)

Where yk (t) is the response to fk (t).

Gloria Menegaz 57
 Time-invariant and time-varying parameters

Gloria Menegaz 58
 Instantaneous and dynamic systems

• Instantaneous (memoryless): the output at time t only depends on the input

at time t
– Special case of dynamic systems
• Dynamic: the output does not only depend on the input at the same instant
– Finite memory: the output is completely determined by the input over the last T
seconds

Gloria Menegaz 59
 Istantaneous and dynamic systems

Gloria Menegaz 60
 Causal and non-causal systems

• Why to study non-causal systems?

– Pre-recorded data
– Other variable than time can be considered (es. Images)

Gloria Menegaz 61
 Lumped and distributed parameters

• Lumped parameters: The system parameters are assumed to be functions

of time alone. In consequence, the system equations require only one
independent variable (time) and therefore are ordinary differential equations
– This applies when the system’s size is small with respect to the wavelength of
the signal that propagates through the system
§ Electric current in a circuit

• Distributed parameters: The system parameters depends on both time and

space leading to mathematical models that are partial differential equations
– Examples: microwave and antennas where the system’s spatial dimension is not
negligible with respect to the signal wavelength
• Here we focus on lumped-parameters systems

Gloria Menegaz 62
 Continuous-time and discrete-time

Gloria Menegaz 63
 Invertible and non-invertible systems

• A signal S performs some operations on the input signal to generate the

output signal. If we can obtain the input back from the output the system is
said to be invertible
– For this to be true, a given output must correspond to a given input (one-to-one
relation)
– Example of non-invertible system: the rectifier, the differentiator. For this last
invertibility can be recovered by knowing one boundary condition

Gloria Menegaz 64
 System model: input-output description
• The system model is a mathematical expression of a rule that satisfactorily
approximates the dynamical behavior of a system
• Internal description
– The internal structure of the system is known so that we can write equations that
describe such a structure
• External description
– The system is viewed from the input and output terminals and is considered as a
black-box. In this case the system model only describes the relation between the
input and the output signals irrespectively of the internal structure
– Implicitly assumes controllability and observability

Cannot be Cannot be
controlled by observed from
the input the output

Gloria Menegaz 65
 Summary

• A signal is a set of information of data.

• A system processes input signals to produce output signals (response)
– Can be a physical system or an algorothm
• Signal can be classified in different ways that are not mutually exclusive
• The linearity property implies superposition such that the response to
different causes can be obtained as the sum of the responses to each
cause that is calculated as if the others were not present

Gloria Menegaz 66