Chapter I

Introduction to classical Linear control

systems

2
Chapter II

Input/Output system models

Definition 1. An input/output mathematical model, or shortly a model of a physical system is a

mathematical description that relates its output to its input.
Essentially, a (mathematical) model is any mathematical description that allows you to compute the
output of the a system given it’s input. Models are generally obtained using two methods:
• System modeling (theoretical): combining physical laws governing the system’s internals.
• System identification (experimental): Using input/output data collected from an experi-
ment on the system.
Ideally, a model captures exactly the input/output relatioship of the physical system. In the
diagram below, the difference e(t) between the real physical output and the theoretical computed
output would be zero for every input r(t)

applied measured
physically physically
Physical system

r(t) b
e(t)
−
Mathematical model computed
mathematically

Practically, however, a model can never be exact, it is only approximate. In general :

• There are many (approximate) models for the same system
• The more accurate(realistic) a model, the more complicated it is, and the more complicated
the analysis/design based on this model would be. and vice versa.
• A model is only valid for certain range of input signals.
For example, a standard model for the resistor with the input and output being the current i(t)
and voltage v(t) is Ohm law: v(t) = Ri(t), this model is simple but approximate since it ignores
other effects like temperature, skin effect...etc. It is also only valid for "small enough" currents and
voltages (up to the rated maximum power Pmax )

3
EE352 Lectures:2 Spring 2023

1 Differential equations models

• Differential equations are equations that involve the input and output signals and any func-
tion of them.

• They are general enough to describe all practical systems(linear, nonlinear, time varying,
stable....).

• They are usually obtained when performing modeling.

1.1 Modeling examples of electrical systems

Example 1: series RLC circuit

i(t) L R

+ +
vi(t) − C vc (t)
−

vi (t) RLC circuit vc (t) .

Functional block diagram Schematics

Replacing i(t) = C dvdt

c (t)
in the KVL equation of the loop, we get:

d2 vc (t) dvc (t)

LC + RC + vc (t) = vi (t) (2nd order Linear constant coefficients differential equation)
dt2 dt

Example 2: A Diode: A silicon diode with input voltage Vd (t) and the output current Id (t):
+ −
Vd (t)
+ −

Vd (t) a Diode Id (t) Id (t)

Id (t) = Is (eαVd (t) − 1) nonlinear DE

1.2 Modeling examples of Mechanical systems

Example 1: Translational Spring-Mass damper

x(t)
b

M f (t)
Mass, spring, k
f (t) x(t)
damper system

Dr.Guernane 4
EE352 Lectures:2 Spring 2023

Where, f(t) is the input force, x(t) is the output linear position. Assuming the spring is linear
(fs (t) = kx(t)) and the friction force is viscous (ff (t) = bẋ(t)):
X
Forces = Mẍ(t) Newton’s 2nd Law ⇒ f(t) − fs (t) − ff (t) = Mẍ(t) ⇒
b k 1
f(t) − kx(t) − bẋ(t) = Mẍ(t) ⇐⇒ ẍ(t) + ẋ(t) + x(t) = f(t) (2nd order LCCDE)
M M M

Example 2: Rotational Spring-Inertia-damper

θ(t) T (t)
k b

Inertia, spring, J
T (t) θ(t)
damper system

Where, T (t) is the input torque, θ(t) is the output angular position. assuming that the torsional
spring is linear(Ts (t) = kθ(t)) and the friction torque is viscous(Tf (t) = bθ̇(t)).
X
Torques = Jθ̈(t) Newton’s 2nd Law ⇒ T (t) − Ts (t) − Tf (t) = Jθ̈(t) ⇒
b k 1
T (t) − kθ(t) − bθ̇(t) = Jθ̈(t) ⇐⇒ θ̈(t) + θ̇(t) + θ(t) = T (t) (2nd order LCCDE)
J J J

Example 3: Mechanical Pendulum Consider a mechanical friction-less pendulum where T (t) is

the input torque about the pivot (CCW), and θ(t) is the output angular position(CCW). the ball
has mass M and the rod is mass-less and of length L.
X
Torques = Jθ̈(t) Newton’s 2nd Law ⇒ T (t) − Tg (t) = Jθ̈(t) ⇒

g 1
T (t) − gML sin(θ(t)) = ML2 θ̈(t) ⇒ θ̈(t) + sin(θ(t)) = T (t) (NLDE)
L ML2

T (t) Pendulum θ(t)

This model can be simplified -linearized- by considering small displacement around the θ = 0
(thus only valid for small θ):

g 1
|θ(t)| <  ⇒ sin(θ(t)) ' θ(t) ⇒ θ̈(t) + θ(t) = T (t) (2nd order LCCDE)
L ML2

Dr.Guernane 5
EE352 Lectures:2 Spring 2023

Important: :A Linear constant coefficients Differential equation (LCCDE) represents a linear and time-
invariant (LTI) system only if the initial conditions are zero

2 Impulse response model for LTI systems

If a system is LTI, then its input and output can be related by the convolution integral:
Zt
y(t) = h(t) ∗ u(t) = h(t − τ)u(τ) dτ
−∞

Where h(t) is the impulse response (IR) of the system. Thus, the IR of an LTI system is a model
for that system

u(t) h(t) y(t)

Important: LTI systems are very special because, they are the only kind of systems that can be represented
by a signal (their IR h(t)):

• given an input signal u(t), the system’s output y(t) can be readily computed using h(t):

y(t) = u(t) ∗ h(t)

• The system is completely characterized by h(t), i,e all other system’s properties can be determined from
h(t): causality, memoryless, stability,...

Important: if a system is not LTI, then its IR, h(t) is meaningless.

Example: a capacitor this system is LTI: input is current ic (t), the output is voltage vc (t):
Zt
1 1
vc (t) = ic (τ) dτ ⇒ h(t) = us (t)
C −∞ C

by examining h(t), we can see that the capacitor is a memory, causal, unstable system.

Dr.Guernane 6
EE352 Lectures:2 Spring 2023

3 Laplace Transform Review

Definition 2. Laplace transform The Laplace transform F(s) of a function f(t) is defined as:
Z +∞
F(s) = L(f(t)) = f(t)e−st dt
0−
s = σ + jω ∈ C

Region of convergence
• The LT function definition must also include the region of convergence. e,i the values of s for which
the above integral converges. Even though in general we do not specify it, one must know that it exists.

• It sufficient in general that for the above integral to converge:

Z +∞
|f(t)|e−σ0 t dt < ∞
0

for some real number σ0 . if for instance |f(t)| < Meγt t ≥ 0, then LT integral converges.

Examples:
1. The unit step signal us (t):
 Z +∞
1 ,t ≥ 0 1  −st +∞ 1
f(t) = us (t) = ⇒= L(f(t)) = 1.e−st dt = − e 0
=
0 ,t < 0 0− s s

2. Another common basic signal is the right exponential

 Z +∞
e−at ,t ≥ 0 1  −st +∞ 1
f(t) = ⇒ F(s) = e−(s+a)t dt = − e 0
=
0 ,t < 0 0− s+a s+a

3.1 Common Laplace Transform properties

In all what follow, Let F(s) = L(f(t)), Fi (s) = L(fi (t)),

3.1.1 Linearity

L(αf1 (t) + βf2 (t)) = αL(f1 (t)) + βL(f2 (t)) = αF1 (s) + βF2 (s)

Example Knowing the LT of the exponential and using Linearity, we can also compute the LT of
the sine and cosine signals:
 jωt   
e − e−jωt 1 1 1 ω
L (sin(ωt)) = L = − = 2
2j 2j s − jω s + jω s + ω2
   
ejωt + e−jωt 1 1 1 s
L (cos(ωt)) = L = + =
2 2 s − jω s + jω s2 + ω2

Dr.Guernane 7
EE352 Lectures:2 Spring 2023

3.1.2 Time Delay

L(f(t − T )us (t − T ))] = e−Ts F(s)

Example:
1 e−3s
L(2e3(t−3) us (t − 3)) = 2 .e−3s = 2
s−3 s−3
The t-domain s-domain equivalence is:

f (t) T sec Delay f (t − T )us (t − T ) F (s) e−T s e−T s F (s)

t-domain s-domain

3.1.3 Differentiation
The simple derivative property:
 
df(t)
L = sF(s) − f(0− )
dt
It can be generalized to any order

dn f(t)
L( n
) = sn F(s) − sn−1 f(0− ) − sn−2 f(1) (0− ) · · · − f(n−1) (0− )
dt

Example Using this property, we can compute the LT of the impulse signal δ(t):
 
dus (t) 1
L δ(t) = = s. − 0 = 1
dt s

Example Using this property, we can compute the LT of the cosine signal cos(ωt):
 
1 d sin(ωt) 1 s×ω s
L cos(ωt) = = − sin(0) = 2
ω dt ω s2 + ω2 s + ω2

The t-domain s-domain equivalence is:

f (0)

f (t) d./dt f˙(t) F (s) s −

sF (s) − f (0− )

t-domain s-domain

Dr.Guernane 8