04LSLLN Automatic Control

Automotive Engineering

Lecture 21. Discrete Signals and Systems

Signals Studied Up To Now: Continuous-Time, or Analog
• Signal (by default) is a function of continuous (physical) time
• The value at each time is also “continuous” (varies in some range).
• Correspond to processes in classical (non-quantum) physics, chemistry, biology:
dynamics of temperature, humidity, voltage, current, charge, chemical
concentration, energy, power etc.
• Typical examples: impulse, step functions, harmonic functions, solutions of various
differential equations (step/frequency/impulse response etc.)

• In this series of lectures: first acquaintance with discrete-time signals (sequences)

• This lecture:
 Motivation to introduce DT signals and systems
 Sampling and “desampling” (digital to analog transform)
 DT signals in frequency domain
 Some basic facts about DT systems
Typical DT Signals

• Unit impulse sequence: 1.5

1
0 k ≠ 0
δ(k) =  0.5
1 k = 0 0
-10 -5 0 5 10
• Unit step sequence: 1.5

1
0 k < 0
ε(k) =  0.5
1 k ≥ 0 0
-10 -5 0 5 10
• Exponential sequence: 1

y(k) = Aα k 0.5

• Delayed (shifted) sequence: 0

-10 -5 0 5 10
y(k) = x(k − k o ) k (samples)
From Continuous-Time to Discrete-Time Control Theory
Up to know, we have considered continuous-time control theory

• Analogue signals defined on some interval [0,T];

• Relations between signals = differential equations;
• Measurements (state, output, reference signal) are performed continuously;

• Frequency-domain characteristics found via the integral Laplace transform.

Another class of systems – discrete-time systems

• Signals are sequences – functions of discrete time t=0,1,…;
• Dynamics are described by difference (recurrent) equations;
• Measurements (state, output, etc.) are available at discrete time instants;
• Properties of DT systems are like those of the CT systems, except for few
essential differences (to be discussed later);
• Frequency-domain characteristics found via the Z-transform (to be introduced).

Discrete-time control theory repeats many results in continuous time; hence, we

do not study it in separate and only give a brief overview.
Fibonacci Numbers

Fibonacci numbers: one of the first discrete-time dynamical

system, available in the literature.

Fibonacci’s Puzzle: A man put a male-female pair

of newly born rabbits in a field. Rabbits take a
month to mature before mating. One month
after mating, females give birth to one male-
female pair and then mate again. No rabbits die. Leonardo Pisano
How many rabbit pairs are there after one year? (Fibonacci),
1170-1240

• Variable (number of rabbits) is integer, differential equations make no sense;

• No continuous measurements and unknown dynamics between two consecutive

measurements. We neither observe (nor are interested in observing!) when each
rabbit is being born. Only their total population is estimated once a month.

• One motivation to work with discrete-time systems:

 system’s dynamics between two observations is fully unknown and/or
 the variables have discrete number of values (differential equations make no sense).
DT Systems – Examples, Linear and Nonlinear Systems

• Moving Average (linear system):

y(k) = [u(k) + u(k − 1) + u(k − 2) + u(k − 3)]/ 4

• Maximum (nonlinear system): y(k) = max[u(k), u(k − 1), u(k − 2)]

• Ideal Delay System (linear): y(k) = u(k − k 0 )

• Memoryless (static) block - nonlinear: 𝑦𝑦(𝑘𝑘) = 𝑓𝑓(𝑢𝑢 𝑘𝑘 )

Main Source of DT Signals and Systems: Sampling (Discretization)
• More often, DT arise from discretization of CT systems. In this case, signals
of the DT system are also discretized (sampled) CT signals.
 Simplified integration (numerical solution) of the equations
 Parameter identification is much simpler than in continuous time
 Today: the necessity of software implementation on digital platforms
(PC, microcontroller, programmable logic controller…)
• Next lecture (L22): a closer look at discretization (sampling) of signals. This
lecture: only general information.
• We have already seen time sampling: L04, discretization of the pendulum!
𝐽𝐽𝑦𝑦(𝑡𝑡)
̈ = −𝐾𝐾 sin 𝑦𝑦(𝑡𝑡) − 𝛽𝛽𝑦𝑦(𝑡𝑡)
̇ + 𝑢𝑢(𝑡𝑡)

𝑦𝑦 𝑘𝑘 = −𝑎𝑎1 𝑦𝑦 𝑘𝑘 − 1 − 𝑎𝑎2 𝑦𝑦 𝑘𝑘 − 2 − 𝑎𝑎3 sin 𝑦𝑦(𝑘𝑘 − 2) + 𝑏𝑏𝑏𝑏(𝑘𝑘 − 2)

𝑇𝑇𝑇𝑇 𝑇𝑇𝑇𝑇 𝑇𝑇 2 𝐾𝐾 𝑇𝑇 2
𝑎𝑎1 = − 2, 𝑎𝑎2 = 1 − , 𝑎𝑎3 = , 𝑏𝑏 =
𝐽𝐽 𝐽𝐽 𝐽𝐽 𝐽𝐽
Analog Signal to Discrete-Time Sequence
Usually, discrete-time signals arise from a continuous-time signal via time
sampling (discretization): informally, physical time (real number) replaced by
sequence of sampling instants, separated by some fixed sampling period.

10

-10
0 20 40 60 80 100 t (s)
10

-10
0 10 20 30 40 50 k (samples)
Sampling and Key Characteristics
• Sampling is a continuous-time to discrete-time conversion:

continuous time discrete time

x t ∈ [ 0, ∞ ) t kT, =
= k 0,1,2,
x (t) x (kT )

0 1 2 3 4 5 6 7 t for simplicity of notation x (kT ) = x (k )

T is the sampling period (or sampling time) (seconds, s)

fs = 1/T is the sampling frequency (Hertz, Hz)
ωs = 2πfs = 2π/T is the sampling frequency (rad/s)
All Information Around Us is Sampled
 Continuous signals are difficult to measure (very few sensors can monitor
variables continuously, e.g., analog oscilloscopes)
 You cannot keep infinite amount of measurements in memory!
 This necessitates to keep the important information as a discrete-time sequence
of values computed at discrete (sampled) timesteps t=kT (k=0,1,2,…) integer, T
is some period): annual, monthly, hourly, minutely etc. plots.
Important Remark: Information Loss
• Sampling is irreversible: you always lose some information
• Given a sampled signal there exist infinite continuous signals
which fit the samples:
x 1
0.5

-0.5

-1
0 20 40 60 80 t 100

• Mathematically, under certain conditions an analog signal can be sampled

without loss so that it can be reconstructed perfectly (sampling theorem)
• Practically, the perfect reconstruction is never possible.
• In control theory, this is NOT a problem: we are interested in the closed-
loop system’s behavior, not ideal correspondence of signals. In Signal
Processing, the goal is to reconstruct signals in a maximally accurate way!
From Discrete-Time to Digital
• At present, most control systems use digital platforms (computers,
microcontrollers) to implement the controllers, actuators, and sensors.
• At the dawn of control theory, controller were mechanical, hydraulic,
thermal, or (most often) based on analog electronics (resistors,
capacitors, inductors). The reason to use PID: simple to implement.
• Modern semi-conductor electronic circuit consist of logic gates (diodes
etc.) Digital platforms allow one to implement logic in software rather
than in hardware. This gives relevant advantages:
 Flexibility in making modifications to the control law after the
hardware design is fixed. Rapid prototyping, simpler debugging
 Hardware and software design can proceed almost independently,
saving a large amount of time.
 Logical and nonlinear operations can be easily included in the
controller.
• At the same time, digital electronics cannot work with continuous
variables. All data are binary, which means that the accuracy is limited.
Three Types of Signals
• Analog signals: continuous in time and amplitude.
 Example: voltage, current, temperature, mass position,…
• Digital signals: discrete both in time and magnitude (finite number of
possible values!)
 Examples: how many people attend the lecture (integer number of
people), number of rabbits in a farm…

• Theory for digital signals and purely digital controllers is complicated

(finite automata, quantized controllers etc.)

• Discrete-time signals: discrete in time, continuous in amplitude (result

from time sampling). A sort of compromise!
 Example: hourly change of temperature or humidity.

• Discrete-time signals give a number of advantages:

 Most convenient to develop theory.
 Good enough approximation in most practical applications.
 The effect of discrete amplitude (finite precision) can often be modeled
as disturbances and measurement noises.
Digital Controllers for Analog Plants
• Discrete-time control system are often obtained via redesign of
continuous-time controller.
• Plants are typically continuous-time (vehicles, robots, chemical reactors
etc.), so it is much easier to perform a CT design first.

1. Design of a continuous-time controller 𝐶𝐶𝐶𝐶𝐶𝐶(𝑠𝑠). This is what the main

part of our course was about. Continuous-time design is often
simpler and more convenient.
2. Discretize of the controller: 𝐶𝐶𝐶𝐶𝐶𝐶(𝑠𝑠) � 𝐶𝐶(𝑧𝑧). Many different discrete-
time approximations, Euler’s forward method is the simplest.
3. Verify the design using the discrete-time analysis and/or numerical
simulation.

• The discrete-time controller is implemented on the real system

to be controlled as follows:
Towards Digital Control

• Analog-to-digital converter (A/D): This device samples a

physical variable and converts it into a binary code (truncation).
The sampling is performed using a given sampling period T.
• At this point, some information is always lost due to the truncation
(without it, formally lossless sampling is possible for some signals).
• Digital-to-analog converter (D/A):
1. This device transforms the binary code into an analog signal.
2. This operation is always lossy; some interpolation, e.g.
 A zero-order hold (ZOH) hold maintains the same signal
value throughout the sample period (most typical case of
D/A). The resulting u(t) is a “staircase” (piecewise-constant)
signal
Typical D/A Transform: Zero-Order Hold (ZOH) reconstruction

1 0.8
e(t) u(k)
0.8
e(k) 0.6 u(t)
0.6
0.4
0.4
0.2
0.2

0 0
e

u
-0.2
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8

-1 -0.8
0 1 2 3 4 5 0 1 2 3 4 5
t t

Discrete-time signals are often interpolated with “staircase” (piecewise-constant)

functions. Discrete instants are points where the function changes its value.
Other D/A Transforms: Impulses, First-Order Hold

“Dirac’s comb”: replacement of discrete-

time signal by sum of pulses, sampled First-order hold (FOH): linear
values = amplitudes. interpolation between consecutive
sample.

More general (and accurate) methods: spline approximation (approximation by

polynomial functions), Nyquist-Shannon method (can fully recover some signals).

In control theory, ZOH is usually enough, because ideal reconstruction of

signals is not needed – controllers should robust to small noises.
DT Signals in Frequency Domain: Z-transform.
Z-transform has almost all properties of the Laplace transform but works on
discrete-time signals (sequences). Traditionally, the variable is denoted z and
must be large enough in modulus. The inverse transform is available.

Main properties are:

• Linearity;
• Forward shift plays the role of derivative: if 𝑔𝑔(𝑘𝑘) = 𝑓𝑓(𝑘𝑘 + 1), then

• Backward shift plays the role of integral: if 𝑔𝑔(𝑘𝑘) = 𝑓𝑓(𝑘𝑘 − 1) (where f(-1)=0),
then

• Final value theorem also has a counterpart.

• Main examples:
DT Systems in Time- and Frequency-Domains

u(k) S y(k)

• Description of LTI DT system in state space form:

𝑥𝑥 𝑘𝑘 + 1 = 𝐴𝐴𝐴𝐴 𝑘𝑘 + 𝐵𝐵𝐵𝐵 𝑘𝑘
𝑦𝑦 𝑘𝑘 = 𝐶𝐶𝐶𝐶 𝑘𝑘 + 𝐷𝐷𝐷𝐷 𝑘𝑘
S : A, B, C , D

𝑥𝑥 𝑘𝑘 = 𝐴𝐴𝑘𝑘 𝑥𝑥 0 +𝐴𝐴𝑘𝑘−1 𝐵𝐵𝐵𝐵 0 +𝐴𝐴𝑘𝑘−2 𝐵𝐵𝐵𝐵 1 + ⋯ + 𝐴𝐴𝐵𝐵𝐵𝐵 𝑘𝑘 − 2 + 𝐵𝐵𝐵𝐵(𝑘𝑘 − 1)

• Description of LTI DT system in transfer function form (x(0)=0):

𝑌𝑌 𝑧𝑧 = 𝐺𝐺 𝑧𝑧 𝑈𝑈 𝑧𝑧 𝑺𝑺: 𝑮𝑮 𝒛𝒛

The matrix-valued function

𝐺𝐺 𝑧𝑧 = 𝐶𝐶 𝑧𝑧𝐼𝐼 − 𝐴𝐴 −1 𝐵𝐵 + 𝐷𝐷
is the transfer function (tf) of the DT state-space model (A,B,C,D).
L07: Finding Solutions in Time Domain – Diagonalizable Case
• Let’s first understand how it looks for a diagonal matrix:

• What if the matrix is not diagonal. A diagonalization trick:

• Don’t forget the meaning of V: columns = eigenvectors

L07: Finding Solutions in Time Domain – Jordan Form

• To find the Jordan form and matrix V, you can use the Matlab command jordan(), which has same argument and
output as eig(). This function requires Symbolic toolbox!
>> A=[2 1 0;0 2 0;1 0 2];
>> [V,J]=jordan(A)
V=
0 1 0
0 0 1
1 0 0

J=
2 1 0
0 2 1
0 0 2

Remember how the Jordan blocks look like for different dimensions:
L07: Powers of Jordan Cells
Power function (for the cell of dimension 4)

In general, it will be the upper-diagonal matrix whose j-th diagonal contains

L07: Behavior at Infinity – Entail Stability Criterion

Converges to 0 if Converges to 0 if

Bounded if Bounded if
Constant (=1) Constant (=1)

Unbounded - otherwise Unbounded - otherwise

The solution, after some transformation, is determined by these exponentials/powers!

Stability Criteria in Discrete Time

• All definitions for stability and unstability remain unchanged

• Asymptotic stability: replace 𝑅𝑅𝑅𝑅 𝜆𝜆𝑖𝑖 < 0 by |𝜆𝜆𝑖𝑖 | < 1 .

• Same for poles in the definition of stable TF:
|𝑝𝑝𝑖𝑖 | < 1 instead of 𝑅𝑅𝑅𝑅 𝑝𝑝𝑖𝑖 < 0

• In conditions of simple (marginal) stability: 𝑅𝑅𝑅𝑅 𝜆𝜆𝑖𝑖 ≤ 0 replaced

by |𝜆𝜆𝑖𝑖 | ≤ 1. Multiplicity condition remains same.

• In conditions for unstability: replace 𝑅𝑅𝑅𝑅 𝜆𝜆𝑖𝑖 > 0 by |𝜆𝜆𝑖𝑖 | > 1 and
𝑅𝑅𝑅𝑅 𝜆𝜆𝑖𝑖 ≥ 0 by |𝜆𝜆𝑖𝑖 | ≥ 1.