Digital control and

command

Gado TCHABODE
2023-2024
V1.0
 . . . Content

1. Introduction to sampled signals

2. Sampled linear servo systems

3. Analysis of sampled linear servos

4. Digital control of discrete systems

2
Gado TCHABODE
1
Introduction to Sampled Signals

3
Gado TCHABODE
 1.1 Introduction

In industrial reality, the complexity of the systems, as well as that of the processing to be carried
out, often requires the use of digital processing tools: computers, calculators, digital systems of
all kinds.
Under no circumstances can such tools accommodate continuous signals; these must be
transformed into sequences of numbers in order to be able to be processed. Similarly, these
systems deliver, at their output, sequences of digital values, in other words, digital signals.

4
Gado TCHABODE
 1.1 Introduction

To transform a continuous signal into a series of numbers compatible with a digital processing
system, two successive operations are used: sampling , which consists of taking, at regular time
intervals, discrete values from the signal, then conversion digital analog that transforms these
samples into numbers, usually encoded in binary form.

This double discretization is obviously likely to generate errors given

that the signal will only be known at given instants and that, moreover, the corresponding
numerical values will be rounded off according to the number of values available at the output.

5
Gado TCHABODE
 1.2 Definitions

➢ Signal
➢ Continuous signal
➢ Discrete signal
➢ Sampled signal
➢ digital signal
➢ Quantification
➢ Causal signal

6
Gado TCHABODE
 1.3 Reminder on the TL

➢ Laplace transform
▪ Definition
▪ Properties
▪ Elementary table of transforms
▪ Inverse transform

7
Gado TCHABODE
1.3 Reminder on the TL

8
Gado TCHABODE
 1.3 Reminder on the TL

➢ Laplace transform
▪ Definition
▪ Properties (linearity, derivative, integration…)
▪ Elementary table of transforms
▪ Inverse transform
▪ Usual signals

9
Gado TCHABODE
 1.4 TL and block diagram

➢ toilet bowl
We consider a parallelepipedic tank of section S =800 cm 2 provided with a float
connected to a valve. The valve ensures the filling of the tank with a flow rate q
proportional to the difference in height of water with respect to a set height h 0 = 24 cm.
The valve is therefore closed if h = h 0 and open to the maximum when the tank is
empty. This proportionality coefficient will be denoted by k and given: k = 0.002 m 2 / s.

Establish the functional diagram of the system by showing a control loop. h0 corresponds
to the setpoint, h(t) corresponds to the system response.
10
Gado TCHABODE
1.5 Open and closed loop system

Confer second year course

11
Gado TCHABODE
2
Sampled linear servo systems

12
Gado TCHABODE
 2.1 Basic principles of sampling

❑ Definition
Sampling a time signal s ( t ) consists in transforming it into a discrete sequence s ( nTe ) of
values taken at instants nTe . Te is called the sampling period. nTe moments are called the
sampling instants. Practically, sampling a signal amounts to multiplying it by a sampling
function p ( t ), zero everywhere, except in the neighborhood of the instants nTe . This function is
often called the Dirac comb.
+∞

𝒔∗ = 𝒔 𝒕 ∙ 𝒑(𝒕) 𝑠 ∗ = ෍ 𝑠(𝑘𝑇) ∙ 𝛿(𝑡 − 𝑘𝑇)

𝑘=0

13
Gado TCHABODE
 2.1 Basic principles of sampling

❑ Definition
The result of a sampling operation, visible in the figure below.

Sampling therefore produces, from a signal s(t), the

sequence:
s(0), s(Te), s(2Te), . . . , s( nTe )
What is generally noted 𝑠 ∗ 𝑡 = {𝑠0 , 𝑠1 , 𝑠2 , 𝑠3 , … , 𝑠𝑛 }

Or 𝑠(𝑘) = {𝑠0 , 𝑠1 , 𝑠2 , 𝑠3 , … , 𝑠𝑛 }

14
Gado TCHABODE
 2.1 Basic principles of sampling

❑ Exercise : graphically sample the signals below.

𝑓(𝑡) with Te=0.5s and g(𝑡) with Te=𝜋/6

15
Gado TCHABODE
 2.2 Basic principles of sampling

❑ Shanon
The condition of non-recovery is therefore, obviously:
𝐵 < 𝑓𝑒 − 𝐵
2𝐵 < 𝑓𝑒

This inequality constitutes Shannon's theorem which can also be stated as follows:

To preserve, during its sampling, the information contained in a signal, the sampling frequency fe
must be greater than twice the spectral width of the signal.

𝒇𝒆 > 𝟐𝒇𝒔 16
Gado TCHABODE
 2.2 Basic principles of sampling

❑ Shanon
In fact, in practice:
▪ the sampling period 𝑇 must respect
Shannon's theorem.
▪ too small a sampling period will have the
disadvantage of reducing the
effectiveness of feedback in the face of
disturbances.
▪ too high a sampling period unnecessarily
overloads the computer's memory.
17
Gado TCHABODE
 2.2 Simple sampled signals

❑ Dirac pulse
We define the unit pulse sampled by the signal:
𝛿 ∗ (𝑡) = {1,0,0,0,0, … }

In other words :

18
Gado TCHABODE
 2.2 Simple sampled signals

❑ Unit step
The unit step sampled by the signal is defined:
𝑢∗ (𝑡) = {1,1,1,1,1, … }

In other words :

19
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.1 Definition
Let s(t) be any continuous signal that is sampled at a frequency fe (ie a period Te), respecting, of
course, Shannon's theorem.

Let s(t) be this sequence is nothing other than the sum of pulses shifted in time and each
multiplied by the coefficient𝑠𝑘

20
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.1 Definition
From a mathematical point of view, the Laplace transform is a means of processing signals
and systems described in continuous time; while the 𝑧-transform is the way to deal with signals
and systems described in discrete time. The '𝑧' transform of a sampled signal 𝑓∗(𝑡), denoted
𝐹(𝑧) or 𝒵[𝑓∗(𝑡)], is defined by the following series of negative powers, where '𝑧' is a complex
variable :

21
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.2 Usefulness of the z-transform

We will find with the z transform, the same interest as that which we had found in the Laplace
transform for continuous time signals. When we know how to model the operation of a digital
system by a transfer function in z, we will have a complete tool allowing us to describe and
therefore study their operation.

Just as we wrote:

we will then write:

22
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.3 Properties
❑ Linearity
For all time signals e1(t) and e2(t) and real numbers a and b, we have:

❑ Temporal Translation
For any time signal e(t), natural integer n and real number T > 0, we have:

❑ Multiply by time
For any time signal e(t) we have: 23
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.3 Properties
❑ Initial Value Theorem
Let e(t) be any time signal and E(z) its z-transform. So we have

❑ Final value theorem

Let e(t) be any time signal and E(z) its z-transform. Then, if all the poles in z of E(z) are in the
trigonometric circle,

24
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.4 Z-transform of usual signals

❑ Pulse unit
The unit pulse being defined by:

We
have :

25
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.4 Z-transform of usual signals

❑ Unit step
The unit scale interval being
defined by:

We
have :

26
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.4 Z-transform of usual signals

❑ Ramp unit
The unit ramp in continuous time is defined by:

27
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.4 Z-transform of usual signals

❑ Decreasing exponential
Let s(t) be the signal defined by 𝑠(𝑡) = 𝑒 −𝑎𝑡 for 𝑡 ≥ 0. The z-transform of this signal has the
expression:

28
Gado TCHABODE
 2.3 Z-transform of sampled signals

Let f(t) be a causal function and f*(t) the same function sampled at the frequency fe .

The Laplace transform of a Dirac being equal to

Its Laplace transform is written :
unity and that of a Dirac delayed by τ. By setting
𝑧 = 𝑒 𝑝𝑇𝑒 , we obtain the z-transform of the
function f(t):

29
Gado TCHABODE
 2.3 Z-transform of sampled signals

2.3.5 Transfer function in z

Let E (z) be the z-transform of the input signal and S (z) that of the output signal.
we define the transfer function in z, G(z) of the system by:

30
Gado TCHABODE
 31
Gado TCHABODE
 3
Stability and performance of
sampled servo systems

32
Gado TCHABODE
 3.1 Equation

Just like continuous systems, sampled systems can be

controlled according to the same closed loop principle.

The direct chain and the return chain are modeled by their transfer functions in z and the input
and output signals are of course sampled at a frequency fe and each have a transform in z: E(z)
and S(z ). The gap 𝜀(t) is no exception to the rule. Let 𝜀(z) be its z-transform. Just as in the case
of continuous time systems, we define the transfer functions in open loop G(z) and in closed
loop H(z) by:

33
Gado TCHABODE
 3.2 Stability

For discrete-time systems, the definition of stability remains the same: a finite input must
correspond to a finite output . Consider a sampled system defined by the following transfer
function:

The zi and the pj are respectively the zeros and the poles of the transfer function.

A sampled system is stable if and only if all the poles pj of its transfer
function are such that | pj | < 1.
34
Gado TCHABODE
 3.2 Stability

Example 1: open-loop stability of a first-order system

We consider a system governed, in open loop, by a recurrence equation of order 1:

The unique pole of the transfer function is a. Therefore, the stability condition is:

35
Gado TCHABODE
 3.2 Stability

Example 2: closed-loop stability of a first-order system

We consider a sampled open-loop transfer function system G(z) placed in a unitary feedback loop
with:

Parameters b and a are positive. Moreover, we will assume that a < 1.

The system is stable in closed loop if the single pole of

this transfer function is less than 1:

36
Gado TCHABODE
 3.3 Accuracy

Sampled systems, like continuous-time systems, must generally meet specifications which
impose, in closed loop, a certain number of performances (which, moreover, are the same as in
continuous time): precision, speed, stability margin and limitation of overshoot.

The idea consists, here again, in introducing into the direct chain, upstream of the system A(z), an
additional transfer function device C(z), called a digital corrector and whose essential role must
consist in modifying the performance of the original system.

𝐶 𝑧 𝐴(𝑧)
𝐺𝑐 𝑧 =
1 + 𝐶 𝑧 𝐴 𝑧 𝐵(𝑧)
37
Gado TCHABODE
 3.3 Accuracy

The whole art of correcting sampled systems consists in choosing the right transfer function C(z)
for this digital corrector so as to adjust each performance to its required value, without disturbing,
of course, the operation of the system. These corrections are generally provided by a computer.
Very often, one chooses, for questions of flexibility and precision, to digitally correct a continuous
time system.

Things are not so simple when it comes to sampled servos. Indeed, the diverse and varied forms
of the recurrence equations of systems sometimes pose a problem when it comes to concluding
general results.

38
Gado TCHABODE
 3.1 General principles

Admittedly, one can always presuppose a principle of equivalence between the elementary
corrective actions in continuous time and the corresponding form in z:

39
Gado TCHABODE
 3.2 Attempts at corrective actions

❑ Accuracy improvement
Integral action improves the precision of the system, but, be careful, it can upset the other
performances, in an often unpredictable way (in the sense of the continuous time perception that
we may have of this corrective action).
Consequently, to simply improve the precision, in closed loop, of a discrete time system, one can
choose a corrector of transfer function equal to:

We will choose n=1 if the specifications only impose a condition of zero position error and n=2 if
the speed error must also be zero.
40
Gado TCHABODE
 3.2 Attempts at corrective actions

❑ Accuracy improvement (example application)

Consider a discrete-time system with an open-loop transfer function G(z) placed in a unitary
feedback loop, with:

1. Calculate the closed loop transfer function

2. Calculate the poles of this transfer function and conclude on the stability
This system is closed loop stable since the single pole of the closed loop transfer function is less
than 1.
3. Plot the first 10 samples following a rung type input.
41
Gado TCHABODE
 3.3 Accuracy

❑ Accuracy improvement (example application)

4. Calculate the position error.

5. We introduce in the direct chain a corrector given by

Calculate FTBF, its poles, position error and samples

We can therefore deduce that the margin of stability is slightly reduced by the addition of the
corrector (it nevertheless remains very comfortable).
This is confirmed by the representation of the samples.
We note the presence of a slight overshoot (approximately 6%) which 42
corroborates the slight loss of margin of stability and speed . Gado TCHABODE
 3.3 Accuracy

❑ Compensation for loss of stability by pole placement

Let's go back to the system that we have just studied this time for a gain K different from 1 in the
direct chain in addition to the integrator. So we now have:
or, in closed loop:

To increase the stability margin, one must seek to reduce

the modulus of the poles. The discriminant remaining
negative as long as K > 0.0625, we can assume that the
poles will remain complex conjugate:
It is enough to choose a value of K which corresponds to a desired value for
the module of each pole, for example if one wishes that the module is equal 43
to 0.25, it is necessary K=3.5 Gado TCHABODE
 3.3 Accuracy

❑ Compensation for loss of stability by pole placement

Let's go back to the system that we have just studied this time for a gain K different from 1 in the
direct chain in addition to the integrator. So we now have:
or, in closed loop:

To increase the stability margin, one must seek to reduce

the modulus of the poles. The discriminant remaining
negative as long as K > 0.0625, we can assume that the
poles will remain complex conjugate:
It is enough to choose a value of K which corresponds to a desired value for
the module of each pole, for example if one wishes that the module is equal 44
to 0.25, it is necessary K=3.5 Gado TCHABODE