1

EEE 188: Digital Control Systems

Lecture summary #1

the controlled variable.

Example: cruise control.
• In feedback control, sensors and measurements play an
important role.
• In discrete time systems, the control is updated at discrete
time instants.
• Terminology
– Plant: system/process to be controlled
– Reference input: desired value for the output.
– Control error: the difference between the reference
input and the measured value of the output
– Disturbance: undesired input.
– Controller: a system used to regulate, correct or
change the behavior of the controlled system.
– Measured output: Measurable characteristics of the
plant
– Control input: variable allowing to regulate or change
the behavior of the plant
– Sensor: a transducer allowing to obtain the numerical
value of the output so that it is compared with the
reference input.

Fig. 1. Open loop vs closed loop control systems C ONTROL OBJECTIVES

The most common objectives of a control system are:
• This class is about digital controls: we want to regulate • Regulatory control: ensure that the measured output is

the characteristics or behavior of a given system using equal (or near) to the reference input. Stability, steady
a feedback controller. The main idea behind feedback state error, and transients are among the most important
control is to use measurements of the system’s output factors to consider, as illustrated in figures 2, 3, and 4.
to achieve the desired goals. Control systems are an The goal is to obtain a reasonable settling time, reduce
important part of our daily life, they are used every where: overshoot and steady state error, and guarantee stability.
• Disturbance rejection: ensure disturbance does not affect
– In cars
the system output.
– In industrial processes
• Optimization: achieve the best response with respect to
– In homes ...
some predefined criteria. Example: minimize the settling
• Examples: time, minimize fuel, etc ...
Cruise control achieves the desired speed by adjusting This class deals mainly with regulatory control.
the acceleration based on the speed measured from the Most controllers use negative feedback where the controlled
speedometer. A thermostat achieves the desired tempera- variable is compared to the reference value, that is
ture by adjusting the furnace cycle and fan.
• An important part of control systems is about understand- error= desired value- actual value (1)
ing the effects of the controller and disturbance on the
This error plays an important role in controller design.
output of the system.
• There are two formulations to solve control problems:
E XAMPLES OF CONTROL SYSTEMS
– Open loop control systems: no feedback loop is used.
Car-following problem
An open-loop controller is usually used to control
simple processes because of its simplicity and low Consider a car-following problem where we want vehicle 2
cost. to follow vehicle 1 at constant distance assuming that vehicle
Example: conventional washing machine. 1 moves in a straight line. The block diagram is shown in
– Closed loop control systems (feedback): the system figure 5 where
is self adjusting based on the measured change of • Reference is r0 , the desired constant distance
Digital Control Systems, spring 2018 Summary 1

Step Response
0.8

0.7

0.6

0.5
Amplitude

Step response 2
0.4

0.3 Step response 1

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3 3.5
Time (seconds)

Fig. 2. Time response illustrating problems in regulatory control: different

settling times. Ideally, we want to get a reasonable value for the settling time.
That “reasonable value” is dictated by the application.

Step Response

Fig. 5. A simple control problem

1

Steady state error

0.8
• Output variable (measured output): r(t)
• Control input: speed of vehicle 2.
Amplitude

0.6

DC motor angle control

0.4
We want to control the angle of a DC motor; the control
input is the voltage and the output variable is the angle. A
0.2
simple proportional controller is used. The block diagram is
shown in figure 3 bottom
0
o
0 0.5 1 1.5 2 2.5 3 3.5 • Desired angle position: 42 .
Time (seconds)
• Encoder or potentiometer: sensors that can be used to

Fig. 3. Time response illustrating problems in regulatory control: non-zero

measure the angle.
steady state error. Ideally, we want to achieve zero steady state error • Proportional controller: v = K(42 − θ). As θ → 42, v →
0
The plant (motor) has a transfer function:
output Θ(s)
Step Response transfer function = = (2)
input V (s)
2
How can we obtain the transfer function of a DC motor?
• The transfer function is a mathematical model that can
1.5
be obtained using extensive experiments. Several methods
exist to approximate the transfer function of a system.
Amplitude

• Match the mathematical models with the motor’s param-

1
eters obtained from the manufacturer.
• Apply a step input and obtain the time response. The type
0.5 and the characteristics of the response are used to get the
frequency domain open loop transfer function.

0
0 0.5 1 1.5 2 2.5 3 3.5 I. C OMPARISON BETWEEN DIGITAL AND ANALOG
Time (seconds) CONTROLLERS :

Fig. 4. Time response illustrating problems in regulatory control: high percent In a digital control problem, the controller and the controlled
overshoot. Ideally, damping has to be reduced to a reasonable value process to not speak the same language, for this reason we
have the DAC (D/A) and the ADC (A/D) to translate.

2
Digital Control Systems, spring 2018 Summary 1

Fig. 6. Top: digital controllers and bottom: DC motor example block diagram Fig. 7. Illustration of digital controllers

Frequency Domain Time Domain

Transfer function State space models
Laplace transform differential equations
z–transform difference equations

• Two formulations to study control systems: time domain

and frequency domain as shown in the table.
• The z–transform is plays important role in digital control Fig. 8. Illustration of the ADC
and digital signal processing.

S AMPLING Tc is very small and therefore is neglected. This leads to the

• Continuous controllers are built using analog electronics ideal sampler.
such as resistor, capacitors, and operational amplifiers. The sampling theorem: The sampling frequency should be
• Digital controllers use digital computers (including at least twice the highest frequency contained in the signal,
micro–controllers) that is:
• There exist two representations for signals: ωs > 2ωm (4)
– Time domain: the system is represented by difference for band limited signals. In this case, the spectrum of the
or differential equations
– Frequency domain: the system is represented by the
z-transform or the Laplace transform.

Continuous Digital
x(t) x(k) (3)
Laplace transform z–transform
Analog to digital conversion (ADC) allows to convert the
analog signal x(t) to a discrete signal x(k) as illustrated in
figures 8 and 9. The conversion occurs at discrete times where:
• T is the sampling period
1
• T is the sampling frequency
A sampler is basically a switch that closes every T seconds,
where T is the sampling period. The switch closure time is Tc .
In practice and control applications, the switch closure time Fig. 9. Sampling process

3
Digital Control Systems, spring 2018 Summary 1

continuous time waveform can be recovered using an ideal T HE z- TRANSFORM

low pass filter. A rule of thumb is to choose ωs as
The z-transform is a mathematical operator used for digital
ωs = Kωm (5) systems and signals. It can be seen as the discrete version of
the Laplace transform.
with 5 ≤ K ≤ 10. The choice of K depends on the application
and is constrained by the hardware. The samples need to • Difference equations can be solved using the z-transform.
provide good representation of the continuous time signal The z-transform is an important tool in the analysis and
design of discrete time control systems.
• What is the advantage of the z-transform from system’s
A PPROXIMATION OF THE DERIVATIVE
point of view? In the frequency domain, the convolution
This method uses the definition of the derivative to obtain becomes a multiplication.
discrete time models from the continuous models.
What is the derivative?
δx Definition
ẋ = lim as δt → 0 (6)
δt Causal system: signal has zero value for negative time.
which can be written as:
x(k + 1) − x(k)
ẋ = (7) Definition
T
where Given a causal sequence {u(0), u(1), u(2), ..., u(k), ...}, its
z-transform is defined as
• k is an integer representing the discrete time.
• x(k) is the value of x(t) at time kT U (z) = u(0) + u(1)z −1 + u(2)z −2 + ... + u(k)z −k (15)
• x(k + 1) is the value of x(t) at time (k + 1)T
k=∞
X
This approximation allows to transform a differential equation U (z) = u(k)z −k (16)
to a recursive algebraic equation. k=0

In this case, z −1 can be seen as some kind of delay.

Example
Convert the following systems to discrete time systems
Example 1
ẋ(t) = −3x(t) + 3u(t) (8)
Obtain the z-transform of the sequence:
ẋ(t) = − sin(x(t)) (9)
u(k) = {1, 3, 2, 0, 4} (17)

Solution
1) For the first system: Solution
Recall that: Using the definition of the z-transform, it is possible to write
x(k + 1) − x(k)
ẋ(t) = (10) U (z) = 1 + 3z −1 + 2z −2 + 4z −4 (18)
T
from which we can write:
x(k + 1) − x(k)
= −3x(k) + 3u(k) (11)
T Example 2: Unit impulse
x(k + 1) = T [−3x(k) + 3u(k)] + x(k) (12)
The unit impulse is given by
2) For the second system: 
1 for k=0
x(k + 1) − x(k) u(k) = δ(k) = (19)
= − sin(x(k)) (13) 0 for k 6= 0
T
x(k + 1) = T [− sin(x(k))] + x(k) (14)
Solution
Equations (12) and (14) are called difference equations. Using the definition again, we get
Can you solve these equations? For example, find x(10)? No,
you need an initial condition. This is an initial value problem.
U (z) = 1 (20)
• Difference equations can be easily solved numerically:
calculate the current sample using the previous ones.
• How to algebraically solve linear differential equations? Example 3: Sampled step
One way is to use Laplace transform.
The sampled step is given by
• In a similar way, linear difference equations can be solved
algebraically using the z-transform. u(k) = {1, 1, 1, 1, ...} (21)

4
Digital Control Systems, spring 2018 Summary 1

F INAL VALUE THEOREM

We are interested in the final value of the sequence, that is
f (∞).

Theorem
If a sequence approaches a constant limit as k approaches
infinity, then the limit is given by
z−1
f (∞) = lim f (k) = lim F (z) = lim (z − 1)F (z)
k→∞ z→1 z z→1
(28)

A. Questions
• What is the keyword in the theorem? The answer is: “the
limit needs to exist”.
• Does the theorem apply to an oscillatory signal (sine wave
for example)? No, because the limit does not exist.

R EMARK
Fig. 10. Some common discrete time signals How did we get equation (26) from equation (25)?
1
= 1 + x + x2 + x3 + ... (29)
Solution 1−x
with |x| < 1
Using the definition again, we get
X
U (z) = z −k (22) Proof
which can be simplified to

U (z) =
1
(23) S = 1 + x + x2 + x3 + ... (30)
1 − z −1 S = 2 3
1 + x(1 + x + x + x + ...) (31)
S = 1 + xS (32)
Example 4: Exponential 1
S = (33)
1−x
ak

for k≥0
u(k) = (24)
0 for k<0
with |a| < 1.

Solution
We have
U (z) = 1 + az −1 + a2 z −2 + a3 z −3 + ... + ak z −k + ... (25)
which can be simplified to
1 z
U (z) = a = (26)
1− z z−a

S OME PROPERTIES OF THE Z – TRANSFORM

Let
Z{f (k)} = F (z) (27)
then
• Linearity: Z{αf1 (k) + βf2 (k)} = αF1 (z) + βF2 (z)
• Time delay: Z{f (k − n)} = z −n F (z)
• Time advance: Z{f (k + 1)} = zF (z) − zf (0)
• Multiplication by an exponential: Z{a−k f (k)} = F (a·z)