Course of “Automatic Control Systems”

2022/23

Harmonic response function

Prof. Francesco Montefusco

Department of Economics, Law, Cybersecurity, and Sports Sciences
Università degli Studi di Napoli Parthenope
[email protected]
Team code: uxbsz19

Prof. Francesco Montefusco Automatic Control Systems 2022/23

 Transient and Steady-state

ñ Let us consider an asymptotically stable LTI system.

ñ Given an input signal 𝑢 𝑡 and an initial condition 𝑥(0) ,we define

ò steady-state response 𝒚𝒔𝒔 (𝒕), the regular behavior of the total response
y(𝑡) (if exist) after an infinite time from the application of the input.

ò transient response 𝒚𝒕 (𝒕), the difference between the total response of

the system and the steady-state response 𝒚𝒕 𝒕 = 𝒚 𝒕 − 𝒚𝒔𝒔 𝒕 .

Prof. Francesco Montefusco 2 Automatic Control Systems 2022/23

 Transient and Steady-state

ñ The steady-state response of asymptotically stable system is independent

from the initial condition.

ñ It depends on the particular input applied to the system

Prof. Francesco Montefusco 3 Automatic Control Systems 2022/23

 Step response: Transient and Steady-state

ñ The step response is characterized by "decaying" exponential functions related

to the system evolution modes and a constant value

u(t)=1(t)
ñ The "decaying" exponential functions determine the transient part of the
response while the constant term is the steady-state value.

Prof. Francesco Montefusco 4 Automatic Control Systems 2022/23

 Step response: Transient and Steady-state

ñ Different evolution modes determine different values of the transient.

t [s]
5
 LTI system response to exponential inputs

Let us consider input signal LTI: 𝒚̇ 𝒕 + 𝟐 𝒚 𝒕 = 𝟑𝒖 𝒕

belonging to the class of complex
exponential functions:
• 𝑢 𝑡 = 𝑒 #$ , 𝑠 = 𝛼 + 𝑗𝜔
Recall that
• 𝑒 #$ = 𝑒 (&'()) $
= 𝑒 &$ cos 𝜔𝑡 + 𝑗 sin 𝜔𝑡

Many signals may be written as

a linear combination of complex
exponential functions.

After an initial transient, the

LTI response is proportional to
the input (i.e. exhibits the same
form of the input). 6
 LTI system response to exponential input

• A SISO system of n-th order, t0=0, x(0)=x0:

𝑥̇ 𝑡 = 𝐴 𝑥 𝑡 + 𝐵𝑢 𝑡 ℒ 𝑠𝑋 𝑠 − 𝑥* = 𝐴 𝑋(𝑠) + 𝐵𝑈(𝑠)
𝑦 𝑡 =𝐶𝑥 𝑡 +𝐷𝑢 𝑡 𝑌 𝑠 = 𝐶 𝑋(𝑠) + 𝐷𝑈(𝑠)

𝑋 𝑠 = (𝑠𝐼 − 𝐴)+, 𝑥* +(𝑠𝐼 − 𝐴)+, 𝐵 𝑈(𝑠)

𝑊 𝑠 = 𝐶(𝑠𝐼 − 𝐴)+, 𝐵 + 𝐷

𝑌 𝑠 = 𝐶(𝑠𝐼 − 𝐴)+, 𝑥* + 𝐶(𝑠𝐼 − 𝐴)+, 𝐵 + 𝐷 𝑈(𝑠)

Yl
Yf
• For exponential input:
ℒ 1 𝑘
𝑢 𝑡 = 𝑒 -$ ,𝑡 ≥ 0 𝑌. 𝑠 = 𝑾 𝑠 𝑈 𝑠 = 𝑊 𝑠 = ⋯+
𝑠−𝜆 𝑠−𝜆
ℒ +,
⋯ + 𝑘𝑒 -$

7
 LTI system response to exponential input

• Initial condition x(0) that nullifies the evolution modes

𝑥 𝑡 = 𝑥(0)𝑒 -$

By using the state equation

𝑥̇ 𝑡 = 𝐴 𝑥 𝑡 + 𝐵𝑢 𝑡 𝜆𝑥(0)𝑒 -$ = 𝐴𝑥(0)𝑒 -$ + 𝐵𝑒 -$

𝜆𝐼 − 𝐴 𝑥 0 = 𝐵 𝑥 0 = 𝜆𝐼 − 𝐴 !" 𝐵, if 𝜆 is not an eigenvalue of A

Then,
𝑥 𝑡 = 𝑥(0)𝑒 #$ = 𝜆𝐼 − 𝐴 !"
𝐵𝑒 #$ , 𝑡 > 0
𝑦 𝑡 = 𝐶 𝑥 𝑡 + 𝐷 𝑢 𝑡 =𝐶 𝜆𝐼 − 𝐴 !" 𝐵𝑒 #$ + 𝐷𝑒 #$ = 𝐶 𝜆𝐼 − 𝐴 !" 𝐵 + 𝐷 𝑒 #$= 𝑾(𝝀)𝑒 #$ .

If the system is asymptotically stable, 𝑥 𝑡 = 𝜆𝐼 − 𝐴 !"

𝐵𝑒 #$ , 𝑦(𝑡) = 𝑾(𝝀)𝑒 #$ ,

these functions represent the asymptotic movements of the state and the
output of the system, for any initial condition x(0).
8
 Steady state response at sinusoidal inputs

ñ Let us consider an asymptotically stable LTI system with a transfer function

𝑊 𝑠 subject to a sinusoidal input signal

u (t ) = U 0 sin (w0t + j ) y(t)

W(s)

ñ The evaluation of the steady state response of LTI system to sinusoidal

inputs is very interest taking into account that any signal can be decomposed
in the sum of a finite (periodic signal) and infinite number (aperiodic
signal) of sinusoids by means of the Fourier series.

Prof. Francesco Montefusco 9 Automatic Control Systems 2022/23

 Steady state response at sinusoidal inputs

ñ It is possible to prove that the steady state response of an LTI system with
transfer function 𝑊 𝑠 to a sinusoidal inputs 𝑢 𝑡 = U* sin(𝜔* 𝑡 + 𝜙) can
be written in the time domain as

(
yss (t ) = U 0 W (s ) s = jw sin w0t + j + ÐW ( s ) s = jw0
0
)
where
ò |𝑊 𝑠 |#/()! is the magnitude of the Laplace transform of 𝑊 𝑠
evaluated in 𝑠 = 𝑗𝜔* .
ò ∠𝑊 𝑠 |#/()! is the phase of the Laplace transform of 𝑊 𝑠 evaluated
in 𝑠 = 𝑗𝜔* .

Prof. Francesco Montefusco 10 Automatic Control Systems 2022/23

 LTI system response to sinusoidal input

For a sinusoidal input,

45
𝑢 𝑡 = sin(𝜔𝑡) , 𝑡 ≥ 0, 𝜔 = ,
6
we exploit the results achieved for an exponential input.

Indeed, sin 𝜔𝑡 = Im 𝑒 %&$ . R𝐞𝐜𝐚𝐥𝐥 𝐭𝐡𝐚𝐭 𝑒 ()$ = cos 𝜔𝑡 + 𝑗 sin 𝜔𝑡 .

F𝐨𝐫 𝑢N 𝑡 = 𝑒 ()$ , if A without eigenvalues in ±𝑗𝜔,then there is an initial state

𝑥 0 = 𝑗𝜔𝐼 − 𝐴 !" 𝐵, such that the movements of the state and the output:

• 𝑥N 𝑡 = 𝑥(0)𝑒 ()$ = 𝑗𝜔𝐼 − 𝐴 +, 𝐵𝑒 ()$ , 𝑡 >0

• 𝑦N 𝑡 = 𝐶 𝑥 𝑡 + 𝐷 𝑢 𝑡 = 𝐶 𝑗𝜔𝐼 − 𝐴 +, 𝐵𝑒 ()$ + 𝐷𝑒 ()$
= 𝐶 𝑗𝜔𝐼 − 𝐴 +, 𝐵 + 𝐷 𝑒 ()$ = 𝑊(𝑗𝜔)𝑒 ()$
( )$'012 𝑾 ()
= 𝑊 𝑗𝜔 𝑒 (012 𝑾 () 𝑒 ()$ = 𝑾 𝑗𝜔 𝑒
(
Recall that z= 𝑎 + 𝑖𝑏 = 𝑟 cos 𝜃 + 𝑖sin 𝜃 = 𝑟𝑒 %' with 𝜃 = arg 𝑧 = tan!" ) + 2𝑘𝜋
11
 LTI system response to sinusoidal input

+, 𝐵𝑒 ()$ , ( )$'012 𝑾 ()
𝑥N 𝑡 = 𝑗𝜔𝐼 − 𝐴 𝑦N 𝑡 = 𝑾 𝑗𝜔 𝑒 ,𝑡 > 0

These functions represent the asymptotic movements, for LTI asymptotically

Y 𝒕 = 𝒆𝒋𝝎𝒕 .
stable with 𝒖

For 𝑢 𝑡 = sin(𝜔𝑡) = Im 𝑒 ()$ , then

𝑥 𝑡 = Im 𝑥N 𝑡 , 𝑦 𝑡 = Im 𝑦N 𝑡 = 𝑾 𝑗𝜔 sin 𝜔𝑡 + arg 𝑾 𝑗𝜔 ,𝑡 > 0

In general for 𝑢 𝑡 = 𝑈 sin 𝜔* 𝑡 + 𝜑 , 𝑡 > 0,

there is an initial state such that the output is a sinusoidal signal:
𝑦 𝑡 = 𝑌 sin 𝜔* 𝑡 + 𝜓 , 𝑡 > 0
with 𝑌 = |𝑊 𝑗𝜔* |𝑈,
where 𝑾 𝑗𝜔* = 𝐶 𝑗𝜔* 𝐼 − 𝐴 +, 𝐵 + 𝐷, and 𝜓 = 𝜑 + arg 𝑾 𝑗𝜔* .

If the system is a.s. y(t) (and x(t)) represents the asymptotic movement of the
output (state).
12
 Steady state response at sinusoidal inputs

Total response of system 𝑊(𝑠) = 1⁄(𝑠 4 +𝑠 + 1) to the input 𝑢 𝑡 = sin 2𝑡 ⋅ 1 𝑡 .

Prof. Francesco Montefusco 13 Automatic Control Systems 2022/23

 Filters

ñ The proposed result can be summarized as follows:

ò The magnitude of a sinusoidal input signal 𝑢 𝑡 = sin(𝜔* 𝑡 + 𝜙) is

amplified or reduced by a linear system depending on the value of
|𝑊 𝑠 |#/()! .

ò An input signal 𝑢 𝑡 = sin(𝜔* 𝑡 + 𝜙) is phase shifted by a linear

system depending on the value of ∠𝑊 𝑠 |#/()! .

ñ In other terms, a linear system can be designed as a filter able to amplify

without distortion a certain set of input signals Ω, and reduce or eliminate
the another signals.

ñ Possible structures of filters will be discussed in the following lessons.

Prof. Francesco Montefusco 14 Automatic Control Systems 2022/23

 Harmonic response function

ñ This result underlines the importance of the function 𝑊(𝑗𝜔) for the
analysis of the forced response of LTI systems.

ñ The function 𝑊(𝑗𝜔) is called harmonic response function of the system.

ñ In the following we present a method able to rapidly evaluate the magnitude

and the phase 𝑊(𝑗𝜔) as a function of 𝜔.

Prof. Francesco Montefusco 15 Automatic Control Systems 2022/23

 W(jw) general form

ñ Given an asymptotically stable LTI system, the harmonic response function

𝑊(𝑗𝜔) is given by the ratio of polynomial with real and complex conjugate
roots

hq
æ 2x q s ö2
s Õ (1 + s i s ) Õ ç1 +
n
s+ 2 ÷
mi

q è
ç w wnq ÷ø
W ( jw ) = W (s ) s = jw
i nq
=K kp
æ 2z p s ö÷
Õj (1 + t j s ) Õp ç1 + w s + w 2 ÷
2
nj
ç
è np np ø
s = jw

Prof. Francesco Montefusco 16 Automatic Control Systems 2022/23

 W(jw) general form

ñ Bode diagrams allows to extract the magnitude and the phase of 𝑊(𝑗𝜔) as
a function of 𝜔

ñ Bode diagrams are a main tool for the closed loop control design

ñ For the closed loop control problems we will be interested to analyze

magnitude and the phase of transfer functions 𝑊(𝑠) also in case of stable
and unstable systems

ñ In that cases, 𝑊 𝑠 |#/() will be not the harmonic function.

Prof. Francesco Montefusco 17 Automatic Control Systems 2022/23