SIGNALS AND SYSTEMS USING MATLAB

Chapter 6 Application of Laplace Analysis to Control

Luis F. Chaparro

Classical and modern control

Classical control: uses frequencydomain methods

Modern control: uses timedomain methods

LTI system connections and block diagrams
Cascade: isolated systems with overall transfer function
H(s) = H1(s)H2(s)

Parallel: same input into systems, overall transfer function

H(s) = H1(s) + H2(s)

Negative Feedback: output is fed back into input and subtracted from it. Overall
transfer function
H(s) =

H1(s)
1 + H2(s)H1(s)

Open-loop transfer function: Ho`(s) = H1(s)

Closed-loop transfer function: Hc`(s) = H(s)

2 / 22

Y (s)

X(s)
H1 (s)

H2 (s)

x(t)

y(t)

H1 (s)

X(s)

Y (s)
y(t)

x(t)
H2 (s)

Cascade (top) and parallel (bottom) connections of systems with transfer function H1(s)
and H2(s). The input/output are given in the time or frequency domains
x(t)

e(t)

H1 (s)

y(t)

H2 (s)

Negative feedback connection of systems with transfer function H1(s) and H2(s). The
input and the output are x(t) and y (t), e(t) is the error signal

3 / 22

Application to classical control Feedback systems

(t)
x(t)

e(t)

Hc (s)

G(s)

(t)
y(t)

x(t)

Hc (s)

G(s)

y(t)

Closed-loop system

Open-loop system

Closed- and open-loop control of systems. The transfer function of the plant is G (s) and
the transfer function of the controller is Hc (s)
Openloop control: Controller cascaded with plant. So output y (t) follows reference

signal x(t), minimize error signal

e(t) = y (t) x(t)

No disturbance, (t) = 0
Y (s) = L[y (t)] = Hc (s)G (s)X (s)
E (s) = Y (s) X (s) = [Hc (s)G (s) 1]X (s) 0 Hc (s) = 1/G (s)
Disturbance (t)
Y (s) = Hc (s)G (s)X (s) + (s), (s) = L[(t)]
E (s) = [Hc (s)G (s) 1]X (s) + (s) cannot be made zero

4 / 22

Closedloop control: Assume y (t) and x(t) same type of signals,

No disturbance (t) = 0
E (s) = X (s) Y (s), Y (s) = Hc (s)G (s)E (s)
X (s)
E (s) =
1 + G (s)Hc (s)
steadystate e(t) 0, poles E (s) in open lefthand splane
Disturbance (t)
E (s) = X (s) Y (s), Y (s) = Hc (s)G (s)E (s) + (s)
(s)
X (s)
E (s) =

= E1(s) + E2(s)
1 + G (s)Hc (s) 1 + G (s)Hc (s)
steadystate e(t) 0, poles E1(s), E2(s) in open lefthand splane

5 / 22

Example: Cruise control of speed of car using controller Hc (s) = 1 + 1/s

PI controller

x(t)
+

e(t)

Plant

c(t)

1
Hc (s) = 1 +
s

v(t)
Hp (s)

Cruise control system: reference speed x(t) = V0u(t), output speed of car v (t). Car
model Hp (s) = /(s + ), mass > 0 and friction coefficient > 0
Hc (s)Hp (s)
V0
X (s) =
1 + Hc (s)Hp (s)
s(s + 1)
B
V0
V (s) =
+
steadystate lim v (t) = V0
t
s +1
s

V (s) =

Error in steadystate

1
V0
1
finalvalue theorem E (s) = X (s) V (s) =
s
s +1


1
lim e(t) = lim sE (s) = lim V0 1
=0
t
s0
s0
s +1
Controlling signal c(t)
c(t) = e(t) +

e( )d

even if e(t) = 0 at some point, the value of c(t) 6= 0

6 / 22

Internal and external stability

Causal LTI system with transfer function H(s) = B(s)/A(s) exhibiting no polezero
cancellation is
Asymptotically stable if all-pole transfer function H1 (s) = 1/A(s), used to determine

the zeroinput response, has all its poles in the open lefthand splane (the jaxis
excluded), or
A(s) 6= 0 for Re[s] 0

BIBO stable if all the poles of H(s) are in the open lefthand s-plane (the j-axis

excluded), or equivalently

A(s) 6= 0 for Re[s] 0

If H(s) exhibits polezero cancellations, the system may be BIBO stable but not

asymptotically stable.

7 / 22

Stabilization
x(t)

y(t)

Ha (s)

G(s)

x(t)

y(t)
K

G(s)

Stabilization of an unstable plant G (s) = 1/(s 2) using (top) an all-pass filter, and
(bottom) a proportional controller of gain K .
s 2
s +2
Ha (j)Ha (j) = Ha (j)Ha(j) = |Ha (j)|2 = 1
Allpass system Ha (s) =

1
s +2
same magnitude as G (s) : |H(j)| = |G (j)||Ha (j)| = |G (j)|
KG (s)
K
=
Negative feedback system H(s) =
1 + KG (s) s + (K 2)
gain K > 2, stable feedback system
stabilized system H(s) = G (s)Ha (s) =

8 / 22

Transient analysis
vi (t)
+

e(t)

vc (t)

1
G(s) =
RCs

vi (t) _

vC (t)
+
C

Feedback modeling of RC circuit as firstorder system with transfer function

H(s) = Vc (s)/Vi (s) = 1/(1 + RCs) = G (s)/(1 + G (s)), G (s) = 1/RCs
Unitstep response
vi (t) = u(t) Vi (s) = 1/s, Vc (s) =

1
1
1
=
s(sRC + 1) s s + 1/RC

vc (t) = (1 e t/RC )u(t)

0.5
RC=1

1
RC=1

vc(t)

0.8

0.6
RC=10

0.4
RC=10

0.2

0.5

0.8

0.6

0.4

0.2

0
0

10

20

30

40

50

Clustering of poles (left) and transient responses vc (t) for 1 RC 10

9 / 22

x(t)

y(t)

e(t)
G(s)

RLC circuits modeled as secondorder feedback system

Vc (s)
1/LC
G (s)
2n
=
=
, G (s) =
Vs (s) s 2 + (R/L)s + 1/LC
1 + G (s
s(s + 2n )
p

where n = 1/ CL natural frequency = 0.5R C /L damping ratio

1
y(t)

0.8

0.4

0.8

0.2

0.6

0
0

0.4
0.2

0.6

10

20

30

40

50

30

40

50

0
0.2

0.4

1.5

0.6

y(t)

0.8
1
1

0.5

1
0.5
0
0

10

20
t

Clustering
of poles (left) and time responses

of second-order feedback system for

2/2 1 (top right) and 0 2/2 (bottom right)

10 / 22

State variable representation of LTI systems

Statevariable representation nonunique internal representation of system. State

variables are memory of system

Transfer function representation external representation of system
vc (0)

vi (t)

v c (t)

vc (t)

RC circuit represented by vC (t) + dvC (t)/dt = vi (s), t 0, and initial condition vC (0)
statevariable vC (t)
for t1 > t0 > 0, vC (t1) = e (t1t0)vC (t0) +

t1

e (t1 )vi ( )d

t0

i.e., given the state at t0 and the input vi (t), we can compute future value vC (t1) for
t1 > t0 independent of how vC (t0) is attained


dvC (t)
Stateequation:
= [1][vC (t)] + [1][vi (t)]
dt

11 / 22

State {xk (t)}, k = 1, , N, of LTI system: the smallest set of variables that if known
at a certain time t0 allows us to compute the response of the system at times t > t0 for
specified inputs {wi (t)}, i = 1, , M
Multiple-input multiple-output state equation
x (t) = Ax(t) + Bw(t)
xT (t) = [x1(t) x2(t) xN (t)],
x T (t) = [x 1(t) x 2(t) x N (t)]
A = [aij ] N N matrix
B = [bij ] N M matrix
wT (t) = [w1(t) w2(t) wM (t)]

state vector

input vector

Output equation
y(t) = Cx(t) + Dw(t)
y(t) = [y1(t) y2(t) yL(t)],
C = [cij ] L N matrix
D = [dij ] L M matrix

12 / 22

output vector

Example:
R=6

+
vs (t)
_

L = 1H

iL (t)

+
vC (t)
_

w(t)

C = 1/8 F

x2 (t)

6
1

Initial conditions

w(t) = vs (t)

x2 (t) = iL (t)

x1 (t) = vC (t)

y(t) = vC (t)

1
dvC (t)
= iL(t)
dt
C
1
R
1
diL(t)
= vC (t) iL(t) + vs (t)
dt
L
L
L

 

  
x 1(t)
0 8
x1(t)
0
=
+
w (t)
x 2(t)
1 6
x2(t)
1



 x1(t)
+ [0] w (t)
y (t) = 1 0
x2(t)
y (t) = x1(t)
y (t) = x 1(t) = 8x2(t)

13 / 22

x1 (t)
y(t)

Another set of state equations

VC (s) =

8
4Vs (s) 4Vs (s)
V
(s)
=
+
s
s 2 + 6s + 8
s + 2) | s {z
+4}
| {z }
2 (s)
X

1 (s)
X

Letting x1(t), x2(t) state variables, w (t) = vs (t) input, and y (t) = vC (t) output:


x 1(t)
x 2(t)



x1(t)
2 0
x2(t)
0 4



 x1(t)
y (t) = 1 1
x2(t)
=

4
w (t)
4

Connection between the first and second set of state equations

x(t) = F
x(t), F invertible

x (t) = F1AF
x(t) + F1bw (t)
y (t) = cT F
x(t)




1
1
2
4
where F =
,
F1 =
so that
0.25 0.5
1 4






2
0
4
F1AF =
,
F1b =
,
cT F = 1 1
0 4
4

14 / 22

Nonuniqueness of states

State variables of a system are not unique

Given state and output equations
x (t) = Ax(t) + bw (t)
y (t) = cT x(t) + dw (t)
New set of state variables {zi (t)} obtained using invertible transformation matrix F
x(t) = Fz(t)
New state variable representation is
z (t) = A1z(t) + b1w (t)
y (t) = c1T z(t) + d1w (t)
A1 = F1AF, b1 = F1b, cT1 = cT F, d1 = d

15 / 22

Canonical realizations

Direct minimal realization

x1 (t)
x2 (t)

w(t)

y(t)

y(t)

y(t)

a1
a0

Minimal direct realization of d 2y (t)/dt 2 + a1 dy (t)/dt + a0y (t) = w (t) (all-pole

system) with state variables x1(t) = y (t) and x2(t) = y (t)

 

  
x 1(t)
a1 a0
x1(t)
1
=
+
w (t)
x 2(t)
1
0
x2(t)
0



 x1(t)
y (t) = 0 1
x2(t)

Minimal realization: Number of integrators equals order of the system

16 / 22

General transfer function: input w (t), output y (t)

Y (s)
bm s m + bm1s m1 + + b0
H(s) =
=
, m<n
W (s)
s n + an1s n1 + + a0
Y (s) =

W (s)
N(s)
D(s)
| {z }
Z (s)

allow us to define all-pole and onlyzeros transfer functions

(i)

Z (s)
1
=
,
W (s) D(s)

(ii)

Y (s)
= N(s)
Z (s)

From which
d n1z(t)
d n z(t)
+ an1
+ + a0z(t) = w (t)
D(s)Z (s) = W (s)
dt n
dt n1
requiring n integrators (minimal realization)
d m z(t)
d m1z(t)
N(s)Z (s) = Y (s) bm
+ bm1
+ + b0z(t) = y (t)
m
m1
dt
dt
requiring n integrators (minimal realization)

17 / 22

Example:
d 2y (t)
dy (t)
dw (t)
+
a
+
a
y
(t)
=
b
+ b0w (t)
1
0
1
dt 2 
dt 
dt



Y (s)
Z (s)
1
H(s) =
= [b0 + b1s]
Z (s) W (s)
a0 + a1 s + a2 s 2
Realizations of
dz(t) d 2z(t)
+
w (t) = a0z(t) + a1
dt
dt 2
dz(t)
y (t) = b0z(t) + b1
dt

and of

b1

w(t)

x1 (t)

x2 (t)

b0

y(t)

a1
a0

State variables x1(t) = z(t)

and x2(t) = z(t)


 

  



 x1(t)
x 1(t)
a1 a0
x1(t)
1
=
+
w (t), y (t) = b1 b0
x 2(t)
1
0
x2(t)
0
| {z } x2(t)
|
{z
}
| {z }
cT
c
Ac

bc

18 / 22

 Parallel and cascade realizations

N(s)
proper rational
H(s) =
D(s)
N
X
parallel realization H(s) =
Hi (s)
i=1

Hi (s) proper rational with real coefficients

Example:
H(s) =

1 + 2s
1 + 2s
1
3
=
=
+
2 + 3s + s 2 (s + 1)(s + 2) s + 1 s + 2
1

x1 (t)

w(t)

19 / 22

x2 (t)

y(t)



1 0
x1(t)
0 2
x2(t)



 x1(t)
y (t) = 1 1
x2(t)
x 1(t)
x 2(t)

1
w (t)
3

Example: Transfer function with repeated poles

s +2
G (s) =
(s + 1)2
parallel realization:

1
1
G (s) =
+
not minimal
s + 1 (s + 1)2
cascade and parallel realization:


1
1
G (s) =
1+
minimal
s +1
s +1




1 0
x1(t)
1 1
x2(t)



 x1(t)
y (t) = 0 1
x2(t)
x 1(t)
x 2(t)

20 / 22

1
w (t)
1

Complete solution

State and output equations

x(t)
= Ax(t) + bw(t)
y (t) = cT x(t),
t>0
Complete solution
T A(t)

y (t) = c e

x(0) + c

exponential matrix e

At

e A(t )bw( )d t 0

= [I + At + A

Impulse response
h(t) = cT e At b

21 / 22

2t

2!

+A

3t

3!

Cramers rule solution

Example: Input w (t) = u(t), IC: x1(0) = 1 and x2(0) = 0.


 

  
x 1(t)
0 1
x1(t)
0
=
+
w(t)
x 2(t)
8 6
x2(t)
8



 x1(t)
y (t) = 1 0
x2(t)
Laplace transform of the state equation, non-zero initial conditions:


 

s 1
X1(s)
x1(0)
=
8 s +6
X2(s)
x2(0) + 8W (s)
Y (s) = X1(s)
Cramers rule

x1(0)
1
x2(0) + 8W (s) s + 6
Y (s) = X1(s) =
s(s + 6) + 8
s +6
8
x1(0)(s + 6) + x2(0) + 8W (s)
=
+
=
2
s 2 + 6s + 8
6s + 8} s(s 2 + 6s + 8)
|s + {z
|
{z
}
det

Yzi (s)

22 / 22

Yzs (s)