Topic 04: Laplace and z-transforms

Advanced System Theory

Winter semester 2021/22

Dr.-Ing. Oliver Wallscheid

Automatic Control
Faculty of Electrical Engineering, Computer Science, and Mathematics

Oliver Wallscheid AST Topic 04 1

Table of contents

1 Laplace transform
2 Transfer function calculus (in the Laplace domain)
Transfer functions of interconnected systems
System response in the complex frequency domain
3 z-transform
4 Conversion between the two domains

Learning objectives
Revise basic calculus of the continuous and discrete-time frequency domain: How can we
describe the system behavior using transfer functions? How to get these transfer functions?
How to use them in order to solve problems in the time domain?

Oliver Wallscheid AST Topic 04 2

Course outline

No. Topic
1 Introduction (admin, system classes, motivating example)
2 Linear algebra basics
3 Response of linear systems incl. discrete-time representations
4 Laplace and z-transforms (complex frequency domain)
5 Frequency response
6 Stability
7 Controllability and observability
8 State transformation and realizations
9 State feedback and state observers

Oliver Wallscheid AST Topic 04 3

Motivation
The analysis of higher-order ODEs

an y (n) (t) + · · · + a1 ẏ(t) + a0 y(t) = bq u(q) (t) + · · · + b1 u̇(t) + b0 u(t)

or first-order state-space ODEs

d
x(t) = Ax(t) + Bu(t),
dt
y(t) = Cx(t) + Du(t)

can be cumbersome since the input-output (I/O) behavior of a given dynamic system can be
not directly evaluated in an ODE. The consideration of interconnected systems (e.g., parallel
or series connection) is also not straightforward.

Accordingly, it would be desirable to transform the above ODEs into some algebraic equations,
which then reflect the I/O behavior in a more direct way.
Oliver Wallscheid AST Topic 04 4
Laplace transform
Definition 4.1: Laplace transform
The unilateral (one-sided) Laplace transform of a continuous-time signal x(t) is defined as
Z ∞
X(s) = L{x(t)} = x(t)e−st dt (4.1)
0

with s = σ + jω, where σ and ω are the real and imaginary parts of s.

The inverse Laplace transform is:

Z σ+jω
1
x(t) = L−1 {X(s)} = X(s)est ds. (4.2)
2πj σ−jω
For s = jω, X(jω) is the unilateral Fourier transform of x(t). With s = σ + jω, the Laplace
transform can be interpreted as the unilateral Fourier transform of x(t) exp(−σt):
Z ∞ Z ∞
−(σ+jω)t
X(σ + jω) = x(t)e dt = [x(t)e−σt ]e−jωt dt . (4.3)
0 0
Oliver Wallscheid AST Topic 04 5
Region of convergence (ROC) example
For the Laplace transform to exist, certain convergence conditions must be fulfilled. See the
following example signal:
t
x(t) = e− τ σ(t), τ ∈ R. (4.4)
Applying the Laplace integral yields:
Z ∞ Z ∞
− τt −st 1
X(s) = e σ(t)e dt = e−t(s+ τ ) dt
"0 #∞ 0
(4.5)
−1 −t(s+ 1 )
= e τ .
s + τ1
0

Obviously the above integral and the resulting Laplace transform only exists if
1
Re {s} > − . (4.6)
τ

Oliver Wallscheid AST Topic 04 6

Important Laplace transform pairs
x(t) X(s) Region of convergence (ROC)
δ(t) 1 all s
1
σ(t) σ>0
s
1
e−αt σ(t) σ > −α
s+α
s
cos(ω0 t)σ(t) σ>0
s + ω02
2
ω0
sin(ω0 t)σ(t) σ>0
s + ω02
2
s+α
e−αt cos(ω0 t)σ(t) σ > −α
(s + α)2 + ω02
ω0
e−αt sin(ω0 t)σ(t) σ > −α
(s + α)2 + ω02
Extended list see: https://en.wikipedia.org/wiki/List_of_Laplace_transforms
Oliver Wallscheid AST Topic 04 7
Selected Laplace transform properties
Time domain Laplace domain
Linearity ax(t) + by(t) aX(s) + bY (s)

Convolution (for causal x and y) x(t) ∗ y(t) X(s)Y (s)

Laplace-domain shift x(t)es0 t X(s − s0 )

Time scaling, a > 0 x(at) X (s/a) /a

Conjugation x∗ (t) X ∗ (s∗ )

d
Time-domain differentiation dt x(t) sX(s) − x(0)
d
Laplace-domain differentiation −tx(t) ds X(s)
Rt 1
Time-domain integration 0 x(τ ) dτ s X(s)

Note: Most operations change the ROC!

Oliver Wallscheid AST Topic 04 8
Example: solution of an initial value problems (1)
Consider an exemplary system described by
d2 d
2
y(t) + 3 y(t) + 2y(t) = u(t)
dt dt
d
with initial conditions y(0) = β and dt y(0) = θ. Let the input be u(t) = ασ(t) (scaled unit
step). Taking the Laplace transform:
s2 Y (s) − βs − θ + 3sY (s) − 3β + 2Y (s) = α/s

β(s + 3) θ α
⇔ Y (s) = + + .
(s + 1)(s + 2) (s + 1)(s + 2) s(s + 1)(s + 2)
| {z } | {z }
L of natural response L of forced response
Choose, e.g., α = 2, β = 3, θ = −5. Then a partial fraction expansion yields
Y (s) = 1/s − 1/(s + 1) + 3/(s + 2)
and thus y(t) = [1 − e−t + 3e−2t ]σ(t) using the Laplace transform pairs table.
Oliver Wallscheid AST Topic 04 9
Example: solution of an initial value problems (2)

Total reponse
Only natural response
2 Only forced reponse

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Fig. 4.1: Time-domain solution to previous ODE with nonzero initial conditions

Oliver Wallscheid AST Topic 04 10

Initial and final value theorems
The following theorems hold true if the sbusequent conditions apply:
I x(t) is a causal signal, i.e., x(t) = 0, ∀t < 0 and
I x(t) contains no impulses or higher-order singularities at t = 0 and
I X(s) does not contain poles in the right s-half plane (only for the final value theorem).

Theorem 4.1: Initial value theorem for Laplace transform

If X(s) is the one-sided Laplace transform of x(t) its initial value can be found by:

x(0) = lim sX(s). (4.7)

s→∞

Theorem 4.2: Final value theorem for Laplace transform

If X(s) is the one-sided Laplace transform of x(t) its final value can be found by:

lim x(t) = lim sX(s). (4.8)

t→∞ s→0

Oliver Wallscheid AST Topic 04 11

Table of contents

1 Laplace transform

2 Transfer function calculus (in the Laplace domain)

Transfer functions of interconnected systems
System response in the complex frequency domain

3 z-transform

4 Conversion between the two domains

Oliver Wallscheid AST Topic 04 12

Description by differential equations
Consider a SISO linear constant-coefficient differential equation
n q
X dk y(t) X dk u(t)
ak = bk . (4.9)
(dt)k (dt)k
k=0 k=0
Taking the Laplace transform of both sides, assuming zero!initial conditions:
n q
!
X X
k
ak s Y (s) = bk sk U (s). (4.10)
k=0 k=0
This leads to a rational transfer function q
X
bk sk
Y (s)
H(s) = = k=0
n . (4.11)
U (s) X
k
ak s
k=0
The roots of the numerator are called the zeros, and the roots of the denominator are called
the poles of the transfer function.
Oliver Wallscheid AST Topic 04 13
Description of LTI systems

I Response of a LTI system with input u(t):

y(t) = h(t) ∗ u(t),

where h(t) is the impulse response.

I For causal systems and input signals, we obtain the equivalent description in the Laplace
domain:
Y (s) = H(s)U (s),
where H(s) is the system transfer function.

I The ROC of the transfer function H(s) of a causal system is the right-half plane 1 .
1
We will utilize this property later in the course when discussing system stability.
Oliver Wallscheid AST Topic 04 14
Series connection of systems
If two systems with their transfer functions
Y1 (s) Y (s)
H1 (s) = , H2 (s) = (4.12)
U (s) Y1 (s)
are connected in series, the transfer function of the overall system is:
Y (s) Y (s) Y1 (s)
H(s) = = = H2 (s)H1 (s) = H1 (s)H2 (s). (4.13)
U (s) Y1 (s) U (s)

Fig. 4.2: Both transfer function depictions lead to identical input-output behavior
Oliver Wallscheid AST Topic 04 15
Parallel connection of systems
If two systems with their transfer functions H1 (s) and H2 (s) are connected in parallel, the
transfer function of the overall system follows directly from the Laplace linearity:

Y (s) = H1 (s)U (s) + H2 (s)U (s) = [(H1 (s) + H2 (s)] U (s) = H(s)U (s). (4.14)

Fig. 4.3: Both transfer function depictions lead to identical input-output behavior

Oliver Wallscheid AST Topic 04 16

Systems with feedback loops
The output signal of systems connected within a feedback loop (cf. figure below) is
Y (s) = F (s) [U (s) + G(s)Y (s)] . (4.15)
From that the overall system transfer function follows:
Y (s) F (s)
H(s) = = . (4.16)
U (s) 1 − F (s)G(s)

Fig. 4.4: Both transfer function depictions lead to identical input-output behavior
Oliver Wallscheid AST Topic 04 17
Multiple-input-multiple-output (MIMO) systems (1)
So far we only considered transfer functions for systems with single-input-single-output
(SISO) characteristics. However, many dynamic systems

d
x(t) = Ax(t) + Bu(t),
dt
y(t) = Cx(t) + Du(t)

have more than m > 1 inputs u ∈ Rm and p > 1 outputs y ∈ Rp . Hence, for
multiple-input-multiple-output (MIMO) systems we require a transfer function matrix
 
H11 (s) H12 (s) . . . H1m (s)
 .. .. 
H21 (s) . . 
Y (S) = H(s)U (s) with H(s) =  . 
. ..  .
 (4.17)
 . . . . . 
Hp1 (s) Hp2 (s) . . . Hpm (s)

Oliver Wallscheid AST Topic 04 18

Multiple-input-multiple-output (MIMO) systems (2)

The transfer function characteristics discussed for SISO system also apply to the MIMO case.

I Series connection:
Y (s) = H2 (s)Y1 (s) = H2 (s)H1 (s)U (s) = H(s)U (s) (4.18)

I Parallel connection:
Y (s) = H1 (s)U (s) + H2 (s)U (s) = [(H1 (s) + H2 (s)] U (s) = H(s)U (s) (4.19)

I Feedback loop:
Y (s) = [I − F (s)G(s)]−1 F U (s) = H(s)U (s) (4.20)

Oliver Wallscheid AST Topic 04 19

LTI transfer function matrix
Applying the calculation rules for connected transfer functions on the LTI state-space
representation
d
x = Ax + Bu, y = Cx + Du (4.21)
dt
and assuming zero initial conditions we receive
Y (s)
H(s) = = C [sI − A]−1 B + D. (4.22)
U (s)

Oliver Wallscheid AST Topic 04 20

Variants for the representation of continuous-time, dynamic systems

State-space representation Transfer function

Ordinary differential equation

Fig. 4.5: Interchangeable representations to describe the (SISO) input-output system behavior

Oliver Wallscheid AST Topic 04 21

Table of contents

1 Laplace transform

2 Transfer function calculus (in the Laplace domain)

Transfer functions of interconnected systems
System response in the complex frequency domain

3 z-transform

4 Conversion between the two domains

Oliver Wallscheid AST Topic 04 22

General idea
We have seen in (3.33) that the system response calculation in the time domain can be quite
elaborate. However, in the complex frequency domain we can make use of the simple linear
mapping between inputs and outputs via the transfer function

Y (s) = H(s)U (s). (4.23)

This more indirect solution path requires the following steps:

1 Obtain the transfer function (from ODE or state-space model; only valid for x0 = 0).
2 Transform u(t) towards U (s) (using a correspondence table).
3 Calculate Y (s) = H(s)U (s), i.e., obtain the result in the frequency domain.
4 Apply inverse Laplace transform to obtain y(t) (using a correspondence table).

Since ready-made Laplace correspondences exist for most standard signals, the (inverse)
Laplace transform never actually needs to be calculated.

Oliver Wallscheid AST Topic 04 23

RLC circuit example (1)
Lets reuse the (slightly modified) RLC circuit state-space model from the first lecture
      
d x1 (t) 0 −1/C x1 (t) 1/C
= + u(t),
dt x2 (t) 1/L −R/L x2 (t) 0
 
  x1 (t)
y(t) = 0 1
x2 (t)
with x1 = uC (capacitor voltage), x2 = iL (inductor current), u = i (input current) and
y = x2 = iL . Using (4.22) we receive the transfer function
Y (s) IL (s) 1
= = H(s) = 2 .
U (s) I(s) s LC + sRC + 1
Comparing to the standard form of a second-order system
K
H(s) = 2 1
s ω2 + s 2D
ω0 + 1
o
q
we can identify the gain K = 1, the eigenfrequency ω0 = √1 and the damping D = R C
LC 2 L.
Oliver Wallscheid AST Topic 04 24
RLC circuit example (2)
We want to calculate the step response given a unit input step

u(t) = σ(t) c s1 = U (s) .

s
Hence, the output signal in the Laplace domain yields
1 K
Y (s) = H(s)U (s) = = .
s(s2 LC + sRC + 1) s(s2 ω2 + s 2D
1
ω0 + 1)
o

For choosing the appropriate correspondence we need to specify D beforehand, since it will
indicate if the system can oscillate or not. For C = 1 F and L = 1 H and R < 2 Ω we receive
0 < D < 1, i.e., a stable but oscillating system and the correspondence is:
 
K s c 1 −Dω0 t
 p
2
Y (s) = K 1− √ e sin ω0 1 − D t + arccos D = y(t).
s(s2 ω12 + s 2D
ω0 + 1) 1 − D2
o

Oliver Wallscheid AST Topic 04 25

RLC circuit example (3)

Fig. 4.6: Step response of inductor current within the RLC circuit (C = 1 F and L = 1 H)

Oliver Wallscheid AST Topic 04 26

Determining exp(At): Laplace transform method
If we are only interested in the transition matrix as an intermediate step for the system
response calculation we can also make use of the Laplace transform. Applying it to ẋ = Ax
with x(0) = x0 we receive
sX(s) − x0 = AX(s). (4.24)
Therefore, (sI − A)X(s) = x0 and thus

X(s) = (sI − A)−1 x0 . (4.25)

This inverse exists for all s except at the eigenvalues of A. Taking the inverse Laplace
transform:
!
x(t) = L−1 {(sI − A)−1 }(t)x0 = Φ(t, 0)x0 = exp(At)x0 . (4.26)
As this must hold for arbitrary x0 , we may therefore compute

exp(At) = L−1 {(sI − A)−1 }(t). (4.27)

The inverse Laplace transform is usually accomplished by correspondence tables.

Oliver Wallscheid AST Topic 04 27
Example
Consider the following autonomous LTI state-space system
 
2 1
ẋ(t) = x(t).
1 2
Applying the Laplace approach to it results in
 −1  
−1 s − 2 −1 s−2 1 x0
X(s) = (sI − A) x0 = x0 = .
−1 s − 2 1 s − 2 (s − 1)(s − 3)
Partial fraction decomposition of
   
s−2 1 1 1 1 1 −1 1
= + , = +
(s − 1)(s − 3) 2 s−1 s−3 (s − 1)(s − 3) 2 s−1 s−3
and using the transform pair x(t) = σ(t)e−αt c ss+α
1
= X(s) leads to
 1 1 −1 1
σ(t) et + e3t −et + e3t
  
1 s−1 + s−3 s−1 + s−3 s c
X(s) = −1 1 1 1 x0 x0 = Φ(t, t0 = 0)x0 .
2 s−1 + s−3 s−1 + s−3 2 −et + e3t et + e3t
Oliver Wallscheid AST Topic 04 28
Table of contents

1 Laplace transform

2 Transfer function calculus (in the Laplace domain)

Transfer functions of interconnected systems
System response in the complex frequency domain

3 z-transform

4 Conversion between the two domains

Oliver Wallscheid AST Topic 04 29

Recap: discrete-time systems
For many real-world applications discrete-time systems play an important role, e.g., for the
controller design on embedded hardware. Those systems can be represented by higher-order
difference equations

an y[k − n] + · · · + a1 y[k − 1] + a0 y[k] = bq u[k − q] + · · · + b1 u[k − 1] + b0 u[k]

or by first-order state-space difference equations

x[k + 1] = Ax[k] + Bu[k],

(4.28)
y[k] = Cx[k] + Du[k].

Likewise to the Laplace transform for continuous-time systems, we want to introduce a

discrete-time complex frequency counter part. This will be done using the so-called
z-transform in the following.

Oliver Wallscheid AST Topic 04 30

z-transform

Definition 4.2: z-transform

The unilateral (one-sided) z-transform of a discrete-time signal x[k] is defined as
∞
X
X(z) = Z{x[k]} = x[k]z −k (4.29)
k=0

with z = rejφ is complex, with r = |z| and φ = ∠ {z}.

The inverse z-transform is:

I
−1 1
x[k] = Z {X(z)} = X(z)z n−1 dz (4.30)
2πj c

where c is a counterclockwise closed path encircling the origin and entirely in the ROC.

Oliver Wallscheid AST Topic 04 31

Discrete-time Fourier transform
For z = ejφ , X(ejφ ) is the unilateral discrete-time Fourier transform (DFT) of x[k]. With
z = rejφ , the z-transform can be interpreted as the unilateral DFT of r−k x[k]:
∞
X ∞
X
jφ jφ −k
X(re ) = x[k](re ) = (rk x[k])ejφk . (4.31)
k=0 k=0

Fig. 4.7: Illustrative representation of the Fourier transformation (not specifically in discrete time)
Oliver Wallscheid AST Topic 04 32
ROC of the z-transform
Analogous to the Laplace transform, certain convergence conditions must be fulfilled, such
that the z-transform converges. See the following discrete-time example signal:
 k  
1 1 1
x[k] = = 1, , , . . . , k ≥ 0. (4.32)
2 2 4
Applying the z-transform summation yields:
∞ ∞  k ∞ 
0.5 k

X X 1 X
X(z) = x[k]z −k = z −k =
2 z
0 0 0 (4.33)
1
= .
1 − 0.5z −1
Obviously the above geometric series and the resulting z-transform only exists if

|0.5z −1 | < 1 ⇔ |z| > 0.5. (4.34)

Oliver Wallscheid AST Topic 04 33

Important z-transform pairs
x[k] X(z) ROC
δ[k] 1 all z
1
σ[k] |z| > 1
1 − z −1
1
αk σ[k] |z| > |α|
1 − αz −1
1 − cos(ε0 )z −1
cos(ε0 k)σ[k] |z| > 1
1 − (2 cos ε0 )z −1 + z −2
sin(ε0 )z −1
sin(ε0 k)σ[k] |z| > 1
1 − (2 cos ε0 )z −1 + z −2
1 − a cos(ε0 )z −1
ak cos(ε0 k)σ[k] |z| > a, a real
1 − (2a cos ε0 )z −1 + a2 z −2
a sin(ε0 )z −1
ak sin(ε0 k)σ[k] |z| > a, a real
1 − (2a cos ε0 )z −1 + a2 z −2
Extended list see: https://en.wikipedia.org/wiki/Z-transform
Oliver Wallscheid AST Topic 04 34
Selected z-transform properties
Time domain z-domain
Linearity ax[k] + by[k] aX(z) + bY (z)
Convolution (for causal x, y) x[k] ∗ y[k] X(z)Y (z)
Time delay x[k − 1] z −1 X(z) + x[−1]
Time advance x[k + 1] zX(z) − zx[0]
ak x[k] X a−1 z


z-domain scaling
Conjugation x∗ [k] X ∗ (z ∗ )
Time-domain difference x[k] − x[k − 1] (1 − z −1 )X(z) − x[−1]
d
z-domain differentiation kx[k] −z dz X(z)
Pk 1
Time-domain accumulation n=0 x[n] X(z)
1 − z −1

Note: Most operations change the ROC!

Oliver Wallscheid AST Topic 04 35
Example: solution of an initial value problems (1)
Consider the system described by

y[k] = u[k] − γy[k − 1]

with initial condition y[−1] = β, input u[k] = ασ[k] and constants {α, β, γ} ∈ R > 0. Taking
the z-transform:
Y (z) = α/(1 − z −1 ) − γ z −1 Y (z) + β

−γβ α
⇔ Y (z) = + .
1 + γz −1 (1 + γz −1 )(1 − z −1 )
| {z } | {z }
Z of natural response Z of forced response

Then a partial fraction expansion yields

   
α 1 α 1
Y (z) = γ −β −1
+ .
1+γ 1 + γz 1 + γ 1 − z −1

Oliver Wallscheid AST Topic 04 36

Example: solution of an initial value problems (2)
Utilizing the z-transform pair table we receive
   
α α
y[k] = γ − β (−γ)k + σ[k].
1+γ 1+γ

0 Total reponse
Only natural response
-2 Only forced reponse
-4
0 5 10 15 20 25 30

Fig. 4.8: Discrete-time solution for α = 8, β = 2 and γ = 10/12

Oliver Wallscheid AST Topic 04 37
Initial and final value theorems for z-transform
Assume that
I x[k] is a causal signal, i.e,. x[k] = 0, ∀k < 0 and
I X(z) only has poles within the unit circle (only for final value theorem).
Then, the following theorems hold true.

Theorem 4.3: Initial value theorem for z-transform

If X(z) is the one-sided z-transform of x[k] its initial value can be found by:

x[0] = lim X(z). (4.35)

z→∞

Theorem 4.4: Final value theorem for z-transform

If X(z) is the one-sided z-transform of x[k] its final value can be found by:

lim x[k] = lim (z − 1)X(z). (4.36)

k→∞ z→1

Oliver Wallscheid AST Topic 04 38

Description by difference equations

Consider a general linear constant-coefficient difference equation

n
X q
X
ai y[k − i] = bi u[k − i] (4.37)
i=0 i=0

with zero initial conditions. Taking the z-transform of both sides gives:
n
X q
X
−i
ai z Y (z) = bi z −i U (z). (4.38)
i=0 i=0

This leads to a rational transfer function

Pq
Y (z) bi z −i
H(z) = = Pni=0 −i
. (4.39)
U (z) i=0 ai z

Oliver Wallscheid AST Topic 04 39

Description of discrete-time LTI systems

I Response of a discrete-time LTI system with input u[k]:

y[k] = h[k] ∗ u[k],

where h[k] is the impulse response.

I For causal systems and input signals, we obtain the equivalent description in the
z-domain:
Y (z) = H(z)U (z),
where H(z) is the system transfer function2 .

I The ROC of the transfer function H(z) of a discrete-time causal system is the is the
exterior of a circle, including infinity3 .
2
Simplifications of combined systems and the transfer from the state space are similar to the Laplace domain.
3
We will utilize this property later in the course when discussing system stability.
Oliver Wallscheid AST Topic 04 40
Variants for the representation of discrete-time, dynamic systems

State-space representation Transfer function

Ordinary difference equation

Fig. 4.9: Interchangeable representations to describe the (SISO) input-output system behavior

Oliver Wallscheid AST Topic 04 41

Table of contents

1 Laplace transform

2 Transfer function calculus (in the Laplace domain)

Transfer functions of interconnected systems
System response in the complex frequency domain

3 z-transform

4 Conversion between the two domains

Oliver Wallscheid AST Topic 04 42

Relationship between Laplace and z-domain
Consider the Laplace transform of a sampled signal x(t) = x(kTs ) (using zero-order hold) with
constant sampling interval Ts :
Z ∞ ∞
X
X(s) = x(t)e−st dt = x(kTs )e−skTs . (4.40)
0 k=0

1
If we substitute s = Ts ln z, we receive
∞
X
x(kTs )z k = X(z), (4.41)
k=0

i.e., the z-transform of the given signal. Thus, the mapping between the Laplace domain and
z-domain is given by:
1
z = esTs ⇐⇒ s = ln z. (4.42)
Ts
Oliver Wallscheid AST Topic 04 43
First-order discrete-time approximation
A typical problem is to convert a continuous-time transfer function to discrete time, e.g., to
perform a controller tuning. Using s = ln z/Ts is often cumbersome and it might be very
difficult to obtain the poles and zeros of the discrete-time transfer function.

Possible workaround: first-order approximation of the mapping using

∞
z − 1 2i+1
 
1 1 X 2
s= ln z =
Ts Ts 2i + 1 z + 1
i=0 (4.43)
2 z−1
≈ .
Ts z + 1
also known as the bilinear transform.
Please note that this mapping is easier to handle but that comes at the cost of approximations
errors, i.e., the discrete-time transfer function will differ from its continuous-time origin.
Oliver Wallscheid AST Topic 04 44
Bilinear transform example (1)

Lets reuse the RLC transfer function

R/LC
H(s) = . (4.44)
s2 + (R/L)s + 1/LC

Apply bilinear transform

2 z−1
s← (4.45)
Ts z + 1
and receive the approximated transfer function in the z-domain:

R z 2 + 2z + 1
H(z) ≈     . (4.46)
LC z 2 4
+ 2R
+ 1
+ 2z LC 1
− T42 + 4
− 2R
+ 1
Ts2 LTs LC s Ts2 LTs LC

Oliver Wallscheid AST Topic 04 45

Bilinear transform example (2)
0

Magnitude (dB) -20

-40

-60
0
Phase (deg)

-90

-180
10 -2 10 -1 10 0 10 1

Frequency (rad/s)
Fig. 4.10: Bode plots of transfer functions (4.44) and (4.46) for parameter values C = 1 F, L = 1 H,
R = 0.05 Ω and Ts = 1 s
Above Bode plot is a teaser for the upcoming section where its background will be explained.
Oliver Wallscheid AST Topic 04 46
Concluding overview

Signals in Frequency domain

continuous time (continuous time)
Laplace transform

Sample / discretize Map (e.g., bilinear)

Signals in Frequency domain

discrete time (discrete time)
z-transform

Fig. 4.11: Summary of the previously considered relationships

Oliver Wallscheid AST Topic 04 47

Important Matlab / GNU Octave commands

1 sys = tf(numerator,denominator); % Introduces a Laplace transfer function (matrix)

2 =
sys = tf(numerator,denominator,ts); % Introduces a z domain transfer function (matrix)
3 sys = tf(ltiSys); % Converts the dynamic system model ltiSys to a transfer function model
4
5 sys = parallel(sys1,sys2); % Parallel connection of two models
6 sys = series(sys1,sys2) ; % Series connection of two models
7 sys = feedback(sys1,sys2); % Feedback connection of two models
8 sys = append(sys1,sys2); % Group models by appending their inputs and outputs
9
10 % The already mentioned discretization commands can be also applied to TF models
11 sysd = c2d(sysc,Ts); % Converts (a TF) model from continuous to discrete time
12 sysc = d2c(sysd); % Converts (a TF) model from discrete to continuous time
13 =
sys1 = d2d(sys, Ts); % Resamples discrete time model

Note that above commands are from Matlab; GNU Octave commands might slightly differ.
And some commands require special toolboxes (e.g., Control System Toolbox).

Oliver Wallscheid AST Topic 04 48

Summary: what you’ve learned in this section

I The Laplace transform enables a system description in the frequency domain for
continuous-time systems.
I Certain technical criteria have to be fulfilled to find a converging Laplace transform.
I In particular, we require a LTI system representation to apply the frequency transform.
I It is only describing the input-output behavior (no knowledge about internal states).
I We can use the Laplace transform to solve initial values problems in the time domain.
I Due to the linearity of the transform we can perform certain transfer function
simplifications for interconnected systems.
I The z-transform is the discrete-time counterpart assuming ideal sampling of a system.
I Most findings of the Laplace domain can be also applied to the z-domain.
I A transfer between the two frequency domains is possible but only often approximated.

Oliver Wallscheid AST Topic 04 49

End of this section

Thanks for your attention!

Oliver Wallscheid AST Topic 04 50