T W E L V E

Design via State Space

SOLUTION TO CASE STUDY CHALLENGE
Antenna Control: Design of Controller and Observer
a. We first draw the signal-flow diagram of the plant using the physical variables of the system
as state variables.

Writing the state equations for the physical variables shown in the signal-flow diagram, we
obtain

0 1 0   0 
 
z = 0 −1.32 0.8  z +  0  u ; y = [0.2 0 0] z
0 0 −100  2000 

The characteristic polynomial for this system is s3 + 101.32s2 + 132s + 0. Hence, the A and B
matrices of the phase-variable form are

Ax Bx
0 1 0 0
0 0 1 0
0 −132 −101.32 1

Writing the controllability matrices and their determinants for both systems yields
12-2 Chapter 12: Design via State Space

CMz Controllability Matrix of z CMx Controllability Matrix of x

0 0 1600 0 0 1
0 1600 −162112 0 1 −101.32
2000 −200000 20000000 1 −101.32 10133.7424
Det(CMz) −5.12 E + 09 Det(CMx) −1

where the system is controllable. Using Eq. (12.39), we find the transformation matrix and its inverse
to be

P Transformation Matrix z = Px PINV

1600 0 0 0.000625 0 0
0 1600 0 0 0.000625 0
0 2640 2000 0 −0.000825 0.0005

The characteristic polynomial of the phase-variable system with state feedback is

s3 + (k3 + 101.32)s2 + (k2 + 132)s + (k1 + 0)
For 15% overshoot, Ts = 2 seconds, and a third pole 10 times further from the imaginary axis than
the dominant poles, the characteristic polynomial is
(s + 20)(s2 + 4s + 14.969) = s3 + 24s2 + 94.969s + 299.38
Equating coefficients, the controller for the phase-variable system is

Kx Controller for x
299.38 −37.031 −77.32

Using Eq. (12.42), the controller for the original system is

Kz Controller for z
0.1871125 0.04064463 −0.03866

b. Using Kz, gain from m = - 0.1871125 (including gear train, pot, and operational amplifier); gain
from tachometer = - 0.04064463; and gain from power amplifier output = 0.03866.
 Solution to Case Study Challenge 12-3

c. Using the original system from part (a) and its characteristic polynomial, we find the observer
canonical form which has the following A and C matrices:

Ax
−101.32 1 0
−132 0 1
0 0 0
Cx
1 0 0

Writing the observability matrices and their determinants for both systems yields

OMz Observability Matrix of z OMx Observability Matrix of x

0.2 0 0 1 0 0
0 0.2 0 −101.32 1 0
0 −0.264 0.16 10133.7424 −101.32 1
Det(OMz) 0.0064 Det(OMx) 1
where the system is observable. Using Eq. (12.89), we find the transformation matrix and its inverse to be

P Transformation Matrix z = Px PINV

5 0 0 0.20 0.00 0.00
−506.6 5 0 20.26 0.20 0.00
62500 −625 6.25 26.40 20.00 0.16
12-4 Chapter 12: Design via State Space

The characteristic polynomial of the dual phase-variable system with state feedback is
s3 + (l1 + 101.32)s2 + (l2 + 132)s + (l3 + 0)

For 10% overshoot, n = 10 14.969 = 38.69 rad/s, and a third pole 10 times further from the
imaginary axis than the dominant observer poles, the characteristic polynomial is
(s + 228.72)(s2 + 45.743s + 1496.916) = s3 + 274.46s2 + 11959s + 3.4237x105
Equating coefficients, the observer for the observer canonical system is

Lx Observer for x
173.14
11827
342370

Using Eq. (12.92), the observer for the original system is

Lz Observer for z
865.7
−28577.724
5569187.5
 Solution to Case Study Challenge 12-5

d.

e.
Program:
'Controller'
A=[0 1 0;0 -1.32 0.8;0 0 -100];
B=[0;0;2000];
C=[0.2 0 0];
D=0;
pos=input('Type desired %OS ');
Ts=input('Type desired settling time ');
z=(-log(pos/100))/(sqrt(pi^2+log(pos/100)^2));
wn=4/(z*Ts); %Calculate required natural
%frequency.
[num,den]=ord2(wn,z); %Produce a second-order system that
%meets the transient response
%requirements.
r=roots(den); %Use denominator to specify dominant
%poles.
poles=[r(1) r(2) 10*real(r(1))];
%Specify pole placement for all
%poles.
K=acker(A,B,poles)
'Observer'
pos=input('Type desired %OS ');
z=(-log(pos/100))/(sqrt(pi^2+log(pos/100)^2));
wn=10*wn %Calculate required natural
%frequency.
12-6 Chapter 12: Design via State Space

[num,den]=ord2(wn,z); %Produce a second-order system that

%meets the transient response
%requirements.
r=roots(den); %Use denominator to specify dominant
%poles.
poles=[r(1) r(2) 10*real(r(1))];%Specify pole placement for all
%poles.
l=acker(A',C',poles)'

Computer response:
ans =

Controller

Type desired %OS 15

Type desired settling time 2

K =

0.1871 0.0406 -0.0387

ans =

Observer

Type desired %OS 10

wn =

38.6899

l =

1.0e+006 *

0.0009
-0.0286
5.5691

ANSWERS TO REVIEW QUESTIONS

1. Both dominant and non-dominant poles can be specified with state-space design techniques.
2. Feedback all state variables to the plant's input through a variable gain for each. Decide upon a closed-
loop characteristic equation that has a pole configuration to yield a desired response. Write the
characteristic equation of the actual system. Match coefficients and solve for the values of the variable
gains.
3. Phase-variable form
4. The control signal developed by the controller must be able to affect every state variable.
5. If the signal-flow diagram is in the parallel form, which leads to a diagonal system matrix, controllability
can be determined by inspection by seeing that all state variables are fed by the control signal.
6. The system is controllable if the determinant of the controllability matrix is non-zero.
7. An observer is a system that estimates the state variables using information from the output of the actual
plant.
 Answers to Review Questions 7

8. If the plant's state-variables are not accessible, or too expensive to monitor

9. An observer is a copy of the plant. The difference between the plant's output and the observer's output is
fed back to each of the derivatives of the observer's state variables through separate variable gains.
10. Dual phase-variable
11. The characteristic equation of the observer is derived and compared to a desired characteristic equation
whose roots are poles that represent the desired transient response. The variable gains of each feedback
path are evaluated to make the coefficients of the observer's characteristic equation equal the coefficients of
the desired characteristic equation.
12. Typically, the transient response of the observer is designed to be much faster than that of the
controller. Since the observer emulates the plant, we want the observer to estimate the plant's states rapidly.
13. Det[A-BK], where A is the system matrix, B is the input coupling matrix, and K is the controller.
14. Det[A-LC], where A is the system matrix, C is the output coupling matrix, and L is the observer.
15. The output signal of the system must be controlled by every state variable.
16. If the signal-flow diagram is in the parallel form, which leads to a diagonal system matrix, observability
can be determined by inspection by seeing that all state variables feed the output.
17. The system is observable if the determinant of the observability matrix is non-zero.
12-8 Chapter 12: Design via State Space

SOLUTIONS TO PROBLEMS
1.

(s + 3) 1
i. G(s ) = = 2 * (s + 3)
(s + 4) 2
s + 8s + 16

a.

b.

c.
•  0 1  0 
x=  x +   r ; y =  3 1 x
 −(k1 + 16) −(k2 + 8)  1 
d.
s+3
T (s) =
s + (k2 + 8)s + (k1 + 16)
2
 Solutions to Design Problems 12-9

•
x1 = (−20 − 71.25k1 ) x1 − 71.25k2 x2 − 71.25k3 x3 + 71.25r
•
x 2 = 27.5k1 x1 + (−10 x2 + 27.5k2 ) x2 + 27.5k3 x3 − 27.5r
•
x 3 = −6.25k1 x1 − 6.25k2 x2 − 6.25k3 x3 + 6.25r
(−20 − 71.25k1 ) −71.25k2 −71.25k3   71.25 
A=  27.5k1 
(−10 x2 + 27.5k2 ) 27.5k3  ; B =  −27.5 ; C = 1 1 1
 −6.25k1 −6.25k2 −6.25k3   6.25 
200(s 2 + 7s + 25)
T (s) = 3
4 s + (120 + 285k1 − 110k2 + 25k3 )s 2 + (800 + 2850k1 − 2200k2 + 750k3 )s + 5000k3

e.

Part d. yields same result as i(d).

s 1
ii. G(s ) = = 2 *s
(s + 5)(s + 7) s + 12s + 35
a.

b.
12-10 Chapter 12: Design via State Space

c.
•  0 1  0 
x=  x +   r ; y =  0 1 x
 −(k1 + 35) −(k2 + 12)  1 
d.

s
T (s) =
s + (k2 + 12)s + (k1 + 35)
2

e.

 0 1  0 
T (s) = C(sI − A)−1 B; A =   ; B =   ; C =  0 1
 −(k1 + 35) −(k2 + 12) 1 
which yields the same result as ii(d).

20s(s + 7) 1
iii. G(s ) = = 3 * (20s2 +140s)
(s + 3)(s + 7)(s + 9) s + 19s + 111s + 189
2

a.

b.
 Solutions to Design Problems 12-11

c.
 0 1 0  0 
•
   
x= 0 0 1  x + 0  r ; y =  0 140 20  x
 −(k1 + 189) −(k2 + 111) −( k3 + 19)  1 

d.
20s(s + 7)
T (s) =
s + (k3 + 19)s + (k2 + 111)s + (k1 + 189)
3 2

e.
 0 1 0  0 
   
 ; B = 0  ; C =  0 140 20 
−1
T (s) = C(sI − A) B; A =  0 0 1
 −(k1 + 189) −(k2 + 111) −( k3 + 19)  1 

which yields the same result as iii(d).

30(s + 2)(s + 3) 1
iv. G(s ) = = 3 * (30s2 + 150 s + 180)
(s + 4)(s + 5)(s + 6) s + 15s + 74s + 120
2

a.
12-12 Chapter 12: Design via State Space

b.

c.

 0 1 0  0 
•
x =  0 0 1  x + 0  r ; y = 180 150 30 x
    
 −(k1 + 120) −(k2 + 74) −(k3 + 15)  1 

d.

30s 2 + 150s + 180

T (s) =
s 3 + (k3 + 15)s 2 + (k2 + 74)s + (k1 + 120)
e.

 0 1 0  0 
   
 ; B = 0  ; C = 180 150 30 
−1
T (s) = C(sI − A) B; A =  0 0 1
 −(k1 + 120) −(k2 + 74) −(k3 + 15) 1 

which yields the same result as iv(d).

s2 + 8s + 15 1
v. G(s ) = = 4 * (s2 + 8s + 15)
(s + 4s + 10)(s + 3s + 12) s + 7s + 34s 2 + 78s + 120
2 2 3
 Solutions to Design Problems 12-13

a.

b.

c.

 0 1 0 0  0 
•
 0 0 1 0  0 
x=   x +   r ; y = 15 8 1 0  x
 0 0 0 1  0 
   
 −(k1 + 120) −(k2 + 78) −( k3 + 34) −( k4 + 7)  1 
d.

s 2 + 8s + 15
T (s) = 4
s + (k4 + 7)s 3 + (k3 + 34)s 2 + (k2 + 78)s + (k1 + 120)
12-14 Chapter 12: Design via State Space

e.

 0 1 0 0  0 
 0 0 1 0  0 
−1
T (s) = C(sI − A) B; A =   ; B =   ; C = 15 8 1 0
 0 0 0 1  0 
   
 −(k1 + 120) −(k2 + 78) −(k3 + 34) −( k4 + 7)  1 

which yields the same result as v(d).

2.
i
a. The output is

Since,

b.

ii
 Solutions to Design Problems 12-15

a.

b.

3.
i
a.
12-16 Chapter 12: Design via State Space

b. Writing the state equations:

But, u = -k1x1 - k2x2 - k3x3 + r . Substituting into the state equations,

 Solutions to Design Problems 12-17

Therefore, T(s) = C(sI - A)-1B, where

Hence,

ii
a.

b. Writing the state equations:

12-18 Chapter 12: Design via State Space

But,
u = −k1 x1 − k2 x2 − k3 x3 + r
Substituting into the state equations, collecting terms, and converting to vector-matrix form yields

Therefore, T(s) = C(sI - A)-1B, or

4.

Using Eqs. (4.39) and (4.34), we find ζ= 0.5192 and n = = 7.352 , respectively.
TP (1 −  2 )1/ 2

Factoring the denominator of Eq. (4.22), the required poles are – 3.676 ± j6.2832. We place the third

pole at -2 to cancel the open loop zero. Multiplying the three closed-loop pole terms yields the desired

characteristic equation:

s3 + 9.6343 s2 + 69.3186 s + 108.1 = 0.

100s 2 + 2700s + 5000

Since G( s ) = , the controller canonical form is:
s 3 + 9s 2 + 23s + 15
 Solutions to Design Problems 12-19

 −9 −23 −15 1 
 0  ;  
A= 1
 0 B= 0
 
;***** 
C = 100 2700 5000  ;
 0 1 0  0 

The first row of A contains the coefficients of the characteristic equation. Thus comparing the first

row of A to the desired characteristic equation and using the results of Problem 5, we have:

k1 = - (9.6343 - 9) = 0.6343; k2 = - (69.3186 - 23) = 46.3186; and k3 = - (108.1 - 15) = 93.1.

5.
The plant is given by

The characteristic polynomial for the plant with phase-variable state feedback is

The desired characteristic equation is

based upon 15% overshoot, Ts =1 seconds, and a third pole ten times further from the imaginary axis
than the dominant poles. Comparing the two characteristic equations,

6.
Expand G(s) by partial fractions and obtain

Drawing the signal-flow diagram with state feedback

12-20 Chapter 12: Design via State Space

Writing the state equations yields the following system matrix:

Evaluating the characteristic polynomial yields,

From Problem 7, the desired characteristic polynomial is

Equating coefficients and solving simultaneously yields

7.

a.

Writing the state equation and the controllability matrix for the system yields
 −7 1  b 
x=  x +  1 u
 −1 −2  b2 

b −7b1 + b2 
CM =  B AB =  1
b2 −b1 − 2b2 
 Solutions to Design Problems 12-21

The controllability matrix has a zero determinant if

b12 − 5b1b2 + b22 = 0

b.

If b1 = 1 the condition becomes b22 − 5b2 + 1 = 0 so there are two possible solutions
b2 = 0.2087 and b2 = 4.7913 .

8.
The controllability matrix is given by Eq. (12.26) for each of the following solutions:
a.

 −2 0 1  0  0 1 −5
   
A =  0 −2 0  ; B = 1  ; CM = 1 −2 4  ;det CM = 0 ; system is uncontrollable
 0 0 −3 1  1 −3 9 

b.

 −2 1 0  0  0 1 −4 
   
A =  0 −2 0  ; B = 1  ; CM = 1 −2 4  ;det CM = −1; system is controllable
 0 0 −3 1  1 −3 9 

c.

 −4 1 0  0  0 2 −7
   
A =  0 0 1  ; B = 2  ; CM = 2 1 −3 ;det CM = 7 ; system is controllable
 0 0 −3 1  1 −3 9 

d.

 −4 1 0  1  1 −4 17 
A =  0 0 1  ; B = 0  ; CM = 0 1 −8 ;det CM = −5 ; system is controllable
   
 −5 0 −3 1  1 −8 44 

e.

0 1 1  1 −2 
A=  ; B =   ; CM =   ;det CM = 0 ; system is uncontrollable
 −6 −5  −2   −2 4 

f.
12-22 Chapter 12: Design via State Space

 −4 0 0  1  1 −4 16 
   
A =  0 −5 0  ; B = 0  ; CM = 0 0 0  ;det CM = 0 ; system is uncontrollable
 0 0 −6  1  1 −6 36 

This system can also be determined uncontrollable by inspection.

9.
Program:
'(d)'
A=[-4 1 0;0 0 1;-5 0 -3]
B=[1;0;1]
Cm=ctrb(A,B)
Rank=rank(Cm)
pause
'(f)'
A=[-4 0 0;0 -5 0;0 0 -6]
B=[1;0;1]
Cm=ctrb(A,B)
Rank=rank(Cm)

Computer response:
ans =

(d)

A =

-4 1 0
0 0 1
-5 0 -3

B =

1
0
1

Cm =

1 -4 17
0 1 -8
1 -8 44

Rank =

ans =

(f)

A =

-4 0 0
0 -5 0
0 0 -6
B =

1
0
 Solutions to Design Problems 12-23

Cm =

1 -4 16
0 0 0
1 -6 36

Rank =

2
10.

Drawing the signal-flow diagram for the plant in cascade form yields
1 1 1
1
u 1 s z3 1 s z2 1 z1 6
s y
-10 -8 -3

Writing the A and B matrices for the z system,

Az Bz
-3 1 0 0
0 -8 1 0
0 0 -10 1

Writing the A and B matrices for the x (phase-variable) system,

Ax Bx
0 1 0 0
0 0 1 0 Phase-Variable Form
-240 -134 -21 1

From the phase variable from, the characteristic polynomial is s3 + 21s2 + 134s + 240.
Finding the controllability matrices and their determinants for the z and x systems shows that there is
controllability,

CMz Controllability Matrix of z CMx Controllability Matrix of x

0 0 1 0 0 1
0 1 -18 0 1 -21
1 -10 100 1 -21 307

Det(CMz) -1 Det(CMx) -1

Using Eq. (12.39), the transformation matrix P and its inverse are found to be
12-24 Chapter 12: Design via State Space

P Transformation Matrix z=Px PINV

1 0 0 1.00 0.00 0.00
3 1 0 -3.00 1.00 0.00
24 11 1 9.00 -11.00 1.00
Using the given transient requirements, and placing the third closed-loop pole over the zero at -6
yields the following desired closed-loop characteristic polynomial:
(s2 + 8s + 45.78)(s + 6) = s3 + 14s2 + 93.78s + 274.68
Using the phase-variable system with state feedback the characteristic polynomial is
s3 + (k3 + 21)s2 + (k2 + 134)s + (k1 + 240)
Equating the two characteristic polynomials yields the state feedback vector for the x system as

Kx Controller for x
34.68 -40.22 -7
Using Eq. (12.42),
Kz Controller for z
92.34 36.78 -7
11.
Program:
A=[-3 1 0;0 -8 1;0 0 -10]; %Generate system matrix A
B=[0;0;1]; %Generate input coupling matrix B
C=[3 1 0]; %Generate output coupling matrix C
D=0; %Generate matrix D
Po=10; %Input desired percent overshoot
Ts=1; %Input desired settling time
z=(-log(Po/100))/(sqrt(pi^2+log(Po/100)^2));
%Calculate required damping ratio
wn=4/(z*Ts); %Calculate required natural
%frequency
[num,den]=ord2(wn,z); %Produce a second-order system that
%meets transient requirements
r=roots(den); %Use denominator to specify
%dominant poles
poles=[r(1) r(2) -6]; %Specify pole placement for all
%poles.
%A few tries at the the third-pole
%value shows T(s) with a closed-
%loop zero at -7.
%Thus, choose the third pole to
%cancel this zero.
K=acker(A,B,poles) %Calculate controller gains in z-
%system
Anew=A-B*K; %Form compensated A matrix
Bnew=B; %Form compensated B matrix
Cnew=C; %Form compensated C matrix
Dnew=D; %Form compensated D matrix
[numt,dent]=ss2tf(Anew,Bnew,Cnew,Dnew);
%Form T(s)
'T(s)' %Display label
T=tf(numt,dent) %Display T(s)
poles=pole(T) %Display poles of T(s)

Computer response:
 Solutions to Design Problems 12-25

K =

92.3531 36.7844 -7.0000

ans =

T(s)

Transfer function:
-3.553e-015 s^2 + s + 6
------------------------------
s^3 + 14 s^2 + 93.78 s + 274.7

poles =

-4.0000 + 5.4575i
-4.0000 - 5.4575i
-6.0000

12.
Using Eqs. (4.39) and (4.34), we find ζ= 0.456 and ωn =17.6498. Factoring the
denominator of Eq. (4.22), the required poles are – 8.0483 ± j15.708. We place the third
pole 10 times further at – 80.483. Multiplying the three closed-loop pole terms yields the
desired characteristic equation: s3 + 96.6 s 2 + 1607 s + 25072 = 0 . Representing the
plant in parallel form:

0 0 0   2.5 
Apar  
= 0 −4 0  ; Bpar =  −4.1667  ; Cpar = 1 1 1 ;
0 0 −10   1.667 

 2.5 0 0 
Using Eq. (12.26), CM par 
=  −4.1667 16.667 −66.667 , which is controllable since
 1.667 −16.667 166.7 
the determinant is 4168.1.

100
Since, G(s ) = the controller canonical form is:
s + 14s 2 + 40s
3

 −14 −40 0  1 

Acc =  1 0 0  ; Bcc = 0  ; C =  0 0 100  . Using Eq. (12.26),

 0 1 0  0 
12-26 Chapter 12: Design via State Space

1 −14 156 
CMcc = 0 1 −14  , which is controllable since the determinant is 1.
0 0 1 

 2.5 0 0 
CM par 
=  −4.1667 16.667 −66.667 
 1.667 −16.667 166.7 

The first row of ACC contains the coefficients of the characteristic equation. Comparing the
first row of ACC to the desired characteristic equation and using the results of Problem 5,
we have:

(14 + k1) = 96.6; (40 + k2) = 1607; and (0 + k3) = 25072. Hence, KCC = [25072 1567 82.6].

 2.500 35.00 100.0 


=  −4.1667 −41.6668 0.0030 
−1
The transformation matrix is: P = CM par CMcc
 1.6670 6.6710 0.0420 

Transforming back to the original system: K par = K CC P =  −10.0 3942 2490.8

−1

14.

The corresponding A and C matrices in observer canonical form are

 Solutions to Design Problems 12-27

The characteristic polynomial is . We check observability by calculating the

observability matrix and its determinant

| OM |= 1

Using the given transient requirements, and placing the third closed-loop pole 10 times farther
from the imaginary axis than the dominant pole yield the following desired characteristic
polynomial

Equating this polynomial to the one in Equation (12.67) yields the observer gains

15.
Using Eqs. (4.39) and (4.34) to find  = 0.5912 and n = 19.4753 respectively. Factoring the
denominator of Eq. (4.22), the required poles are -11.513 ± j15.708. We place the third pole 20 times
further at –230.26. Multiplying the three closed-loop pole terms yields the desired characteristic
equation: s3 + 253.28s2 + 5681.19s + 87334.19 = 0.

 −25 1 0  0
 
Representing the plant in observer canonical form: Aoc = −171 0 1 ; Boc =  0  ; *****
 
 −315 0 0  10 
Coc = [1 0 0]. The first column of Aoc contains the coefficients of the characteristic equation.
Comparing the first column of Aoc to the desired characteristic equation and using Eq. (12.67), l1 =
253.28 - 25 = 228.28; l2 = 5681.19 - 171 = 5510.19; and l3 = 87334.19 - 315 = 87019.19. Hence,

Loc = [228.28 5510.19 87019.19]T.

16.
The A, L, and C matrices for the phase-variable system are:
 0 1  l 
A=  ; C =  2 1 ; L=  1 
 −45 −14  l2 
Hence,
12-28 Chapter 12: Design via State Space

  + 2l1 l1 − 1 
 − ( A − LC) =  
2l2 + 45 l2 +  + 14 
or
 2 + (2l1 + l2 + 14) + (2l2 − 17l1 + 45)
From the problem statement, the desired characteristic polynomial is 2 + 144 + 14400.
Equating coefficients yields,

(2l1 + l2 + 14) = 144; (2l2 − 17l1 + 45) = 14400

Solving simultaneously,

l1 = −671.2; l2 = 1472.4
17.
The A matrix for each part is given in the solution to Problem 11. Each observability matrix is
calculated from Eq. (12.79).
a.
 −2 0 1   5 5 5 
A =  0 −2 0  ; C = (5,5,5) ; OM =  −10 −10 −10 ; | OM |= 0 ; unobservable
   
 0 0 −3  20 20 20 

b.
 −2 1 0   5 0 5 
  
A = 0 −2 0 ; C = (5,0,5) ; OM = −10 5 −15 ; | OM |= 125 ; observable
   
 0 0 −3  20 −20 45 

c.
 −4 1 0   1 0 0
A =  0 0 1  ; C = (1,0,0) ; OM =  −4 1 0 ; | OM |= 1; observable
   
 0 0 −3  16 −4 1
d.
 −4 1 0   1 0 0
A = 0 0 1 ; C = (1,0,0) ; OM =  −4 1 0 ; | OM |= 1; observable
 
   
 −5 0 −3  16 −4 1
e.
 0 1  1 0
A=  ; C = (1,0) ; OM =  ; | OM |= 1; observable
 −6 −5  0 1
 Solutions to Design Problems 12-29

f.
 −4 0 0   1 1 1
 
A = 0 −5 0 ; C = (1,1,1) ; OM =  −4 −5 −6 ; | OM |= −2 ; observable
   
 0 0 −6  16 25 36 

18.
Program:
'(a)'
A=[-2 0 1;0 -2 0;0 0 -3] %Form compensated A matrix
C=[5 5 5] %Form compensated C matrix
Om=obsv(A,C) %Form observability matrix
Rank=rank(Om) %Find rank of observability
%matrix
'(f)'
A=[-4 0 0;0 -5 0;0 0 -6] %Form compensated A matrix
C=[1 1 1] %Form compensated C matrix
Om=obsv(A,C) %Form observability matrix
Rank=rank(Om) %Find rank of observability

Computer response:
ans =

(a)

A =

-2 0 1
0 -2 0
0 0 -3

C =

5 5 5

Om =

5 5 5
-10 -10 -10
20 20 20

Rank =

ans =

(f)

A =

-4 0 0
0 -5 0
0 0 -6

C =

1 1 1

Om =
12-30 Chapter 12: Design via State Space

1 1 1
-4 -5 -6
16 25 36

Rank =

19.
The A and C matrices for the system represented in cascade form is

Az
-20 1 0
0 -13 1
0 0 -5

Cz
1 0 0

The characteristic polynomial found from the transfer function of the plant is
s3 + 38s2 + 425s + 1300
From this characteristic polynomial, we can write observer canonical form of the state equations. The
A and C matrices of the observer canonical form are given below as

Ax
-38 1 0
-425 0 1
-1300 0 0

Cx
1 0 0
To test observability, we write the observability matrices for both systems and show that both
observability matrices have non zero determinants. Using Eq. (12.79),

OMz Observability Matrix of z OMx Observability Matrix of x

1 0 0 1 0 0
-20 1 0 -38 1 0
400 -33 1 1019 -38 1

Det(OMz) 1 Det(OMx) 1

Using Eq. (12.89), we obtain the transformation matrix, P, and its inverse as

P Transformation Matrix z=Px PINV

1 0 0 1.00 0.00 0.00
-18 1 0 18.00 1.00 0.00
25 -5 1 65.00 5.00 1.00
 Solutions to Design Problems 12-31

Using the characteristic polynomial given in the problem statement, the plant’s characteristic
equation, and Eq. (12.67), the observer for the observer canonical system is
Lx Observer for x
562
39575
1498700

Using Eq. (12.92), the observer for the cascade system is found to be

Lz Observer for z
562
29459
1314875
20.
Program:
A=[-20 1 0;0 -13 1;0 0 -5]
B=[0;0;1]
C=[1 0 0]
D=0
poles=roots([1 600 40000 1500000])
L=acker(A',C',poles);
'L'
L'

Computer response:
A =

-20 1 0
0 -13 1
0 0 -5

B =

0
0
1

C =

1 0 0

D =

poles =

1.0e+002 *

-5.2985
-0.3508 + 0.4001i
-0.3508 - 0.4001i

ans =

L
12-32 Chapter 12: Design via State Space

ans =

1.0e+006 *

0.0006
0.0295
1.3149

21.

Use Equations (4.39) and (4.42) to find ζ= 0.5912 and ωn = 135.328. Factoring the denominator

of Eq. (4.22), the required poles are − 80 ± j109.15. We place the third pole 10 times further at −

800. Multiplying the three closed-loop pole terms yields the desired characteristic equation:

s3 + 960 s2 + 146313.746 s + 14650996.915 = 0

45
Since G(s ) = , the plant in observer canonical form is:
s + 18 s + 95 s + 150
3 2

 −18 1 0  0
 
Aoc =  −95 0 1  ; Boc =  0  ; Coc = 1 0 0  ;
 −150 0 0   45

 1 0 0
Using Eq. (12.79), OMoc =  −18 1 0  , which is observable since the determinant is 1.

 229 −18 1 

45
Since (s ) = , the phase-variable form is:
s + 18 s + 95 s + 150
3 2

 0 1 0  0 

Apv =  0 0 1  ; Bpv = 0  ; C =  45 0 0  ;

 −150 −95 −18 1 
 Solutions to Design Problems 12-33

 45 0 0 
Using Eq. (12.79), OMpv =  0 45 0  , which is observable since the determinant is 91125.
 0 0 45

The first column of Aoc contains the negative values of the coefficients of the characteristic

equation. Comparing the first column of Aoc to the desired characteristic equation and using Eq.

(12.67), l1 = 960 ˗ 18 = 942; l2 = 146313.746 ˗ 99 = 146214.746; and l3 = 14650996.915 ˗ 162 =

14650834.915.

Hence, Loc = [942 146214.746 14650834.915]. The transformation matrix is,

 0.0222 0 0 
P=O −1
OMoc 
=  −0.4000 0.0222 0 
Mpv

 5.0889 −0.4000 0.0222 

Transforming back to the original system, Lpv = PLoc = [20.91 2869.17 271556.38] T.

22.

a.

Since both systems are in phase variable form it readily follows that:

det(sI − A1 ) = s2 + 3s + 2

and

det(sI − A2 ) = s3 + 6s2 + 11s + 6

 s + 3 1
 −2 s  0 
G1 (s) = C1 (sI − A1 ) B1 =  2 0  2 
−1 2
  = 2
s + 3s + 2 1  s + 3s + 2
12-34 Chapter 12: Design via State Space

 s 2 + 6s + 11 s+6 1
 
 −6 s(s + 6) s
 −6s 0 
 −(11s + 6) s 2   
G2 (s) = C2 (sI − A2 ) B2 =  6 2 0 
−1
0 
s 3 + 6s 2 + 11s + 6
1 
2(s + 3) 2
= 3 = 2
s + 6s + 11s + 6 s + 3s + 2
2

b.

 C  2 0 
OM1 =  1  =   ; rank(OM1 ) = 2 System 1 is observable
 C 1 A1   0 2 
 C2   6 2 0
  
O M 2 = C 2 A 2  =  0 6 2  ; rank(OM2 ) = 2 System 2 is not observable
C2 A22   −12 −22 −6 

23.
s+2
The open-loop transfer function of the plant is T(s) = C (sI-A)-1 B = .
s −s−2
2

Using Eqs. (12.115), the closed-loop state equations with integral control are:

• 
 x1   −1 1 0   x1  0   x1 
•  
 x 2  =  − k1 − k2 + 2 ke   x2  + 0  r; y = 1 1 0   x2 
 •   −1 −1 0   xN  1   xN 
xN  
 

The characteristic polynomial is: s 3 + (k2 − 1)s 2 + (k2 + k1 + ke − 2)s + 2 ke

The desired characteristic polynomial is calculated from the desired transient

response stated in the problem. Also, the third pole will be placed to cancel the
zero at -2. Hence, the desired characteristic polynomial is:
(s + 2)(s2 + 13.335 s + 166.325) = s3 + 15.335 s2 + 192.995 s + 332.65
 Solutions to Design Problems 12-35

Equating coefficients of the characteristic polynomials yields,

ke = 166.325, k2 = 16.335, k1 = 12.335

SOLUTIONS TO DESIGN PROBLEMS

24.
a. By inspection the phase variable form of this system results in
 0 1 0  0 

A= 0 0 1  ; B = 0  ;

 −0.000078 −0.01265 −0.5250  1 

C = [−0.01197  10−4 −0.958  10−4 0] ; D = 0

b. The characteristic equation of the closed loop system will have 3 poles. Two of them will be
4 1
placed at − = sec −1 to satisfy the settling time spec. The third pole will be placed
300 sec 75
farther to the left of the latter, say at -0.05. The desired characteristic polynomial is
(s + 0.0133)2 (s + 0.05) = s3 + 0.0766s 2 + 0.00150689s + 0.0000088445 = 0

The closed loop matrix is

 0 1 0 

A − BK =  0 0 1 

 −(0.000078 + k1 ) −(0.01265 + k2 ) −(0.525 + k3 ) 

Which will result in a characteristic polynomial

s 3 + (0.525 + k3 )s 2 + (0.01265 + k2 )s + (0.000078 + k3 ) = 0

Equating coefficients for both polynomials we get k3 = −0.4484 , k2 = −0.01114311 and

k1 = −6.91555  10−5

c.

>> A=[0 1 0; 0 0 1; -0.000078 -0.01265 -0.525];

>> B=[0;0;1];

>> C=[-0.01197e-4 -0.9580e-4 0];

>> D=0;
12-36 Chapter 12: Design via State Space

>> K=[-6.91555e-5 -0.01114311 -0.4484];

>> eig(A-B*K) %check

ans =

-0.0133

-0.0133

-0.0500

>> step(A-B*K,B,C,D)

Step Response
0

-0.02

-0.04

-0.06
Amplitude

-0.08

-0.1

-0.12

-0.14
0 100 200 300 400 500 600
Time (sec)

27.

a. From the solution of Problem III the system can be written in phase variable form as:
 Solutions to Design Problems 12-37

 0 1 0  0 

Az =  0 0 1  ; Bz = 0  ;

 −0.000078 −0.01265 −0.5250  1 

Cz = [−0.01197  10−4 −0.958  10−4 0] ; Dz = 0

The corresponding observability matrix is:

 Cz   −0.012  10 −4 −0.958  10 −4 0 
   −4 
OMz =  Cz Az = 0 −0.019  10 −4 −0.958  10 
Cz Az2   0.0001  10 −4 0.0121  10 −4 0.491  10 −4 

It is readily verified that the rank of OMz is 3 and the system is observable.

Next we write the system in observable canonical form, for this let
−0.958  10 −4 0.01197  10 −4
−4 −4 −
C (s ) −0.958  10 s − 0.01197  10 s2 s3
G(s) = = 3 =
R(s ) s + 0.525s + 0.01265s + 0.000078
2
0.525 0.01265 0.000078
1+ + +
s s2 s3

From which we can write:

1 1
s s
1
s
(
C (s) =  −0.525C (s) +  −0.958  10 −4 R(s) − 0.01265C (s) + −0.01197  10 −4 R(s) − 0.000078C (s

Giving the following state flow diagram.

12-38 Chapter 12: Design via State Space

The system in observer canonical form is

 −0.5250 1 0   0 
   −4 
A x =  −0.01265 0 1  ; B x =  −0.958  10  ;
 −0.000078 0 0   −0.01197  10 −4 

C x = [1 0 0] ; D x = 0

And
 −(0.5250 + l1 ) 1 0 
A x − L xCx =  −(0.01265 + l2 ) 0 1 
 −(0.000078 + l3 ) 0 0 

With a characteristic polynomial

s 3 + (0.525 + l1 )s 2 + (0.01265 + l2 ) + (0.000078 + l3 ) = 0

The plant has poles at -0.5, -0.0128 and -0.0122. To obtain the tenfold speed increase in the observer

arbitrarily we will place the observer poles 10 times to the left of the plant poles, namely -5, -0.128

and -0.122. The resulting desired polynomial is

s3 + 0.75s2 + 0.1406s + 0.007808 = 0 Comparing the observer and the desired polynomials

we obtain:

l1 = 0.75 − 0.525 = 0.225

 Solutions to Design Problems 12-39

l2 = 0.1406 − 0.01265 = 0.128

l3 = 0.007808 − 0.000078 = 0.0077
 0.225 

Or L x = 0.128

 
0.0077 

The observability matrix for the observable canonical system is:

 Cx   1 0 0
  
OMx =  Cx Ax  =  −0.525 1 0 
C x A 2x   0.263 −0.525 1 

The transformation matrix is

14  10 4 −688  10 4 33480  10 4 
 
P = O−Mz1 Omx =  −1  10 4 14  10 −4 −688  10 4 
 1  10 4 −1  10 4 14  10 4 


Finally
1.7286  106 
 2 
****** L z = OL x =  −379  10 
 700 
 

b.
12-40 Chapter 12: Design via State Space

-
x
Gain9

1 1 1
-
s s K- s
Integrator Integrator Integrator Gain8 3
x

Gain7
K-
-
5 4 3
Gain6 2
x

Gain5
K-
-
1
1 K-

Constant

- K-

Gain4

1 1 1 l1
-
s s K- s
-
Integrator Integrator Integrato Gain3

l2
Gain2
K-
-
- K-
2 1 r
Gain1
l3
-
-
Gain K-
K-
-

K-
K-

K-
Solutions to Design Problems 12-41
12-42 Chapter 12: Design via State Space

28.
a. As per equation (12.115) in the text the augmented system is:
 x1   x1  0 
 x  =  A − BK Bke   x  + 0  r
 2   −C 0     
2

 x N   xN  1 
  0 −83.33 166.67  166.67    x1  0 

=  500
−
−10   0 
 k1 k2   0  ke   x  + 0  r
   2   

  0 −1 0   xN  1 

 −166.67k1 −(83.33 + 166.67k2 ) 166.67ke   x1  0 
=  500 −10 0   x  + 0  r
 2   
 0 −1 0   xN  1 
 x1 
y =  C 0   x2 
 xN 

The characteristic equation for the system above is:

 Solutions to Design Problems 12-43

s + 166.67k1 83.33 + 166.67k2 −166.67ke

det(sI − Aaug ) = −500 s + 10 0
0 1 s
= s 3 + (10 + 166.67k1 )s 2 + (1666.7k1 + 41665 + 83335k2 ) s + 83335ke

The desired polynomial for Ts = 0.5sec and %OS=20% with an extra far away pole is:

(s2 + 16s + 130.65)(s + 50) = s3 + 66s 2 + 930.6s + 6532.5

Equating coefficients in both polynomials we have:

10 + 166.67k1 = 66  k1 = 0.336
1666.7k1 + 41665 + 83335k2 = 930.6  k2 = −0.4955
83335ke = 6532.5  ke = 0.0783884

b.

1 1 1
1 -K- -K- -K-
s s s
r ke Integrator2 Gain1 Integrator1 Gain2 Integrator c

k1

-K- 10
-83.33

-K-

k2

-K-
12-44 Chapter 12: Design via State Space

29.
a. We start by verifying that the system is controllable. For a system of this order it is best
to use MATLAB or other computing tools:

>>A=[-5 0 0 0 0; 0 0 1 0 0; -10.5229 -1066.67 -3.38028 23.5107 0; 0 993.804 3.125 -23.5107 0; 0 0 0

10 -10];
>>B = [5;0;0;0;0];
>>C = [0 0 0 1.2331e5 0];
>>D=0;
>> rank(ctrb(A,B))

ans =

So the system is controllable.

We also use MATLAB to find the corresponding transfer function:

>> [n,d]=ss2tf(A,B,C,D)

n=

1.0e+010 *

0 0.000000000000000 0.000000000000000 -0.002027466873438 -

0.665044169115073 -6.447695003806981
 Solutions to Design Problems 12-45

d=

1.0e+004 *

0.000100000000000 0.004189098000000 0.152603651149600 1.914775683863999

7.933055056779999 8.565653331000014

We use this information to express the system in state variable form:

 0 1 0 0 0  0 
 0 0 1 0 0   0 
  
xx =  0 0 0 1 0  x + 0  u
   
 0 0 0 0 1  0 
 −85650 −79330 −19150 −1526 −41.89  1 

y =  −64476950038.1 −6650441691.2 −20274668.73 0 0 x x

Now we find the desired characteristic polynomial using a dominant pole approximation.
4
Assuming a pair of complex conjugate dominant poles Ts = = 2 s, or  n = 2 . The
 n
10% OS requirement corresponds to a  = 0.6 damping factor. Therefore the dominant
poles can be included in a s + 2  n s +  n2 = s 2 + 4s + 11.11 . The must be a total of 5
2

poles so arbitrarily we will add three poles 10 times to the left of the dominant poles, namely
( s + 20)3 . Thus the desired polynomial is given by:
D(s) = (s 2 + 4s + 11.11)(s + 20)3
= s 5 + 64s 4 + 1451.11s 3 + 13466.77s 2 + 45335.47s + 88903.11

Under state feedback, the phase variable system will become

 0 1 0 0 0  0 
 0 0 1 0 0  0 
   
xx =  0 0 0 1 0  x + 0  u
   
 0 0 0 0 1  0 
 −(85650 + k1x ) −(79330 + k2 x ) −(19150 + k3 x ) −(1526 + k4 x ) −(41.89 + k5 x ) 1 

With a characteristic equation given by:

12-46 Chapter 12: Design via State Space

det(sI − ( Ax − Bx K x ))
= s 5 + (41.89 + k5 x )s 4 + (1526 + k4 x )s 3 + (19150 + k3 x )s 2
+ (79330 + k2 x )s + (85650 + k1x )

We now equate the coefficients of this equation with those of the desired polynomial to get:

K x =  k1x k2 x k3 x k4 x k5 x  =  3253.11 −33994.53 −5683.28 −74.89 22.11

This vector is now transformed to the original coordinate system using MATLAB:

>> Ax=[0 1 0 0 0; 0 0 1 0 0; 0 0 0 1 0; 0 0 0 0 1;-8.565653331000014e4 -

7.933055056779999E4 -1.914775683863999e4 -0.152603651149600E4 -

0.004189098000000E4];

>> Bx = [0;0;0;0;1];

>> P=ctrb(A,B)*inv(ctrb(Ax,Bx));

>>Kz=Kx*inv(P);

K =  4.422 110.04 16.93 −1.75 −0.14 

b. The simulation is performed with the following commands:

>>Af=(A-B*Kz)
>>step(Af,B,C,D)
 Solutions to Design Problems 12-47

5
x 10 Step Response
0

-1

-2

-3
Amplitude

-4

-5

-6

System: sys
-7 Settling Time (sec): 1.83

-8
0 System:0.5
sys 1 1.5 2 2.5 3
Peak amplitude: -7.96e+005 Time (sec)
Overshoot (%): 9.7
At time (sec): 1.23
30.

a. The controllability matrix for the original system is

 0 61.7325 −2170.9283 34794.0908 

 0 −61.7325 2170.9283 13187.7158 
C MO =  B AB A 2 B A 3 B  = 
 61.7325 −2170.9283 34794.0908 237593.3651 
 
 −61.7325 2170.9283 −13187.7158 −997418.2728 

C MO = −1.779  1012  0 so the system is controllable

b. The original systems characteristic equation is

det(sI - A) = s4 + 35.1667s3 + 1023.07s2 + 12308s + 0

It follows by inspection, that the phase variable representation of the system is:
0 1 0 0  0 
0 0 1 0   
xP =   x + 0  u
0 0 0 1  0 
   
0 −12308.345 −1023.07 −35.1667  1 
12-48 Chapter 12: Design via State Space

c. The controllability matrix of the phase variable representation is

0 0 0 1 
0 0 1 −35.1667 
C MP =  BP A P BP AP2 BP A3P BP  = 
0 1 −35.1667 213.6268 
 
1 −35.1667 213.6268 16157.1015 

The transformation matrix is

21606.375 0 61.7325 0 
 0 0 −61.7325 0 
P = C MOC-1MP =  
 0 21606.375 0 61.7325 
 
 0 0 0 = 61.7325

d. Using state feedback the closed loop phase state representation has an ACL matrix

 0 1 0 0 
 0 0 1 0 
ACL = (AP - BP K P ) =  
 0 0 0 1 
 
 − k1P −(k2 P + 12308.345) −(k 3 P + 1023.07) −( k4 P + 35.1667) 

which results in a characteristic equation

det(sI − ACL ) = s 4 + (k4 P + 35.1667)s 3 + (k3 P + 1023.07)s 2 + (k2 P + 12308.345)s + k1P

Equating coefficients with the desired pole positions

D(s) = (s + 10)4 = s 4 + 40s3 + 600s2 + 4000s + 10000

We get K P = k1P k2 P k3 P k4 P  = 10000 −8308.34 −423.07 4.8333

e.
K O = K P P -1 =  0.4628 7.3161 −0.3845 −0.4628

31.
a. The observability matrix is obtained from

 C   1 0 0 
OMZ  
=  CΑ  =  0 1 0 

CΑ 2   −52.6532 −4.9353 −2768.1557

OMZ = −2768.1557  0
 Solutions to Design Problems 12-49

so the system is observable.

b. The original system’s characteristic equation is

det(sI − A) = s3 + 4.9964s2 + 52.9546s − 0.1428 . The observer canonical form of
the system can be written by inspection, resulting in

 −4.9964 1 0 
AX =  −52.9546 0 1 
 0.1428 0 0 

C x = 1 0 0 

c. The observability matrix for the observer canonical form representation is

 Cx   1 0 0
OMX   
= Cx Ax  =  −4.9964 1 0 
Cx Ax2   −27.9906 −4.9964 1 

 −1 0 0 
-1
P = O OMX 
=  −4.9964 1 0 
MZ 
 −1.4017  10 −6 2.2073  10 −5 −3.6125  10 −4 

d. We write

 −(l1x + 4.9964) 1 0 
Ax - L xCx =  −(l2 x + 52.9546) 0 1 
 −(l3 x − 0.1428) 0 0 

From which it follows that

det(sI − ( A x - L x C x )) = s3 + (4.9964 + l1x )s 2 + (52.9546 + l2 x )s + (l3 x − 0.1428)

Comparing with the specified polynomial D(s) = s3 + 30s2 + 316s + 1160 we get
 25.0036 
L x =  263.0454 
1160.1428
12-50 Chapter 12: Design via State Space

e.

 25.0036 
Lz = PL x = 138.1174 
 −0.4133 

32.

The modified Simulink model is shown below. As could be seen, two state feedback amplifiers

were added: an angle feedback amplifier with a gain of 12 and a speed feedback amplifier with a

gain of 0.1. The settings of the rate feedback amplifier and the PID block (configured as a PD

controller) were kept the same.

As could be seen from the scope graph, shown below, using state feedback amplifiers with

appropriate gains improved the impulse response of the angle control loop if compared the

response obtained in problem IX-z: The amplitude of the pulse here has been reduced to 0.17

radians (less than half of that in IX-z) and it goes down to zero in 0.05 rather than in 0.2 seconds.

l
 Solutions to Design Problems 12-51

33.

Dividing all terms in the numerator and denominator of GP (s) by s3, we get:

1 1.2 12500
+ +
C (s ) 250 * Y (s) s s2 s3
GP ( s ) = = = 250 *
M (s) M (s) 8.1 62003 31250
1+ + +
s s2 s3

Cross-multiplying, and combining terms of like powers gives:

1 1 1
Yˆ (s) =  M (s) − 8.1Yˆ (s) + 2 1.2 M (s) − 62003Yˆ (s)  + 3 12500M (s) − 31250Yˆ (s) 
s s s

Thus, the plant may be represented as shown below in Figure 1a. Identifying the state variables as

the outputs of the integrators, we have:

 −8.1 1 0  1 
xˆ = A xˆ + B m =  −62003 0 1 
 xˆ +  1.2  m ;
 −31250 0 0  12500 

yˆ = 1 0 0  xˆ
12-52 Chapter 12: Design via State Space

Now form the difference between the plant’s actual output, y, and the observer’s estimated

output, ŷ , and add the feedback paths from this difference to the derivative of each state variable.

The result is shown in Figure 1b

(a)

(b)

Figure 1

Next find the characteristic polynomial. From Eqs. (12.64) and (12.66), the observer
error is:
 Solutions to Design Problems 12-53

 −(8.1 + l1 ) 1 0
e x = ( A − LC) e x =  −(62003 + l2 ) 0 1  e x ;

•

 −(31250 + l3 ) 0 0 

Using equation 12.65, we obtain the characteristic polynomial.

s 3 + (8.1 + l1 )s 2 + (62003 + l2 )s + (31250 + l3 )

The dominant poles of the output, c(t), in the proportionally controlled loop of problem
IX-z were – 34.14 ± j111.44. For the observer to respond 10 times faster, its poles
should be at:

– 341.4 ± j1114.4. Select the third pole to be 10 times the real part of the dominant
observer poles, or – 3414. Hence, the desired characteristic polynomial is:

(s + 3414)(s2 + 68.28s + 12418.9) = s3 + 3482.3s2 + 245526.8s + 42398034

Equating the two polynomials gives:

l1 = 3473.9; l2 = 183523.8; l3 = 42366784;

Interested students can use MATLAB (or any other software) to simulate the observer
and check its response to a step and/or a ramp input.

34.

a.

In chapter 4 we found that the open loop transfer function of this system is type 0. Thus for zero

steady state error, integral control is required. Also as the open loop transfer function has a zero at -

0.02, and this zero will appear as a zero of the closed loop transfer function if not eliminated, it will

be cancelled with a closed loop pole. The 10% overshoot requirement corresponds to a

 = 0.6 damping factor. The settling time requirement correspond to a second order term with
4
 n = = 0.04 or n = 0.0667 . The desired closed loop polynomial is:
Ts

(s2 + 0.08s + 0.0044)(s + 0.02)(s + 0.5) = s 4 + 0.6s3 + 0.056s2 + 0.003088s + 0.000044

where a far away pole was arbitrarily added at -0.5.

12-54 Chapter 12: Design via State Space

Following equation (12.115a) in the text, the closed loop system matrix A is given by

 A − BK Bke 
ACL = 
 −C 0 
  −0.04167 0 −0.0058  5.2   5.2  
 
 
  
0.0217 −0.24 0.0058  −  −5.2   k1 k2 k3   −5.2  ke 
=
  0 100 −2.4   0   0  
 
  0 0 −1 0 
 −(0.04167 + 5.2k1 ) −5.2k2 −(0.0058 + 5.2k3 ) 5.2ke 
 0.0217 + 5.2k −0.24 + 5.2k2 0.0058 + 5.2k3 −5.2ke 
=  1

 0 100 −2.4 0 
 
 0 0 −1 0 

The corresponding characteristic polynomial is:

s 4 + (2.68167 + 5.2k1 − 5.2k2 )s 3 + (0.1060088 + 13.728k1 − 12.583844k2 − 520k3 )s 2 +
(0.01241932 + 2.9952k1 − 0.2492256 k2 − 10.3844 k3 − 520 ke )s − 10.3844ke

Equating the coefficients of this polynomial with those of the desired polynomial and solving

simultaneous equations one gets

K =  k1 k2 k3  =  −0.004174 0.39615 −0.0096 and ke = −4.237125  10 −6

b.

-K-

-5.2

1 1 1 1
1 -K- 5.2 -K- -K-
s s s s
Constant Integrator3 ke Gain Integrator 0.0217 Integrator1 100 Integrator2
Scope

-2.4
-0.04167 -0.24
-K-
-K- -K-

k1
0.0058
-K-
-K-

k2

-K-
-0.0058

k3 -K-

-K-
 Solutions to Design Problems 12-55

35.

We’ll start by finding the key parameters needed to meet specifications:

• The damping ratio,  for a 4.32% overshoot is:

− ln(%OS /100) − ln(4.32 /100)

= = = 0.707
 + ln (%OS /100)
2 2
 2 + ln 2 (4.32 /100)
• The natural frequency, n, for a settling time = 4 sec at the above  is:

4 4
Ts = 4=   n = 1.414 rad/sec
 n 0.707 n
• The coordinates of the required dominant poles, – n ± jd = – 1 ± j 1.
• The open-loop transfer-function of the “plant” is given by equation (3.73):

Y (s )
G(s ) = = C(sI − A)-1B + D +++++
U (s)

Here:
12-56 Chapter 12: Design via State Space

 s 0  −20 −40   s + 20 40 
(sI − A) =   −   = 
0 s  0.2491 −0.0191  −0.2491 s + 0.0191

 s + 20 40   s + 0.0191 0.2491
adj   
adj(sI − A)  −0.2491 s + 0.0191 =  −40 s + 20 
(sI − A)-1 = =
det(sI − A) s 2 + 20.0191s + 10.346 s 2 + 20.0191s + 10.346

Hence:
 s + 0.0191 0.2491  0 
0 0.06154 
s + 20  1000 
G(s ) =
Y (s)
=  −40 = 2
61.54 (s + 20)
U (s) s + 20.0191s + 10.346
2
s + 20.0191s + 10.346

Now, we write the closed-loop state equations for the system with integral control in a form similar to

that given in the text by equations 12.115a and 12.115b (Figure 12.21):

 Ia   −20 −40 0   I a  0 
      + 0  r *****
   = 0.2491 − 1000k1 −0.0191 − 1000 k2 1000 K e     
 x N   0 −0.06154 0   x N  1 
 Ia 
And the output equation is given by: y (t ) = v (t ) =  0 0.06154 0   
 
 xN 

 s 0 0  −20 −40 0 
  
(sI − A) = 0 s 0  − 0.2491 − 1000 K1 −0.0191 − 1000 K 2 1000 K e  =
0 0 s   0 −0.06154 0 

 s + 20 40 0 
 −0.2491 + 1000 K s + 0.0191 + 1000 K 2 −1000 K e 
 1

 0 0.06154 s 

The characteristic polynomial for that system is:

s 3 + (20.0191 + 1000 K 2 ) s 2 + (−40000 K1 + 20000 K 2 + 61.54 K e + 10.346) s + 1230.8K e

 Solutions to Design Problems 12-57

The desired characteristic polynomial (with the third pole placed to cancel the zero at -20) is:

s3 + 22 s2 + 42 s + 40

Equating coefficients of the characteristic polynomials yields:

K1 = 0.000249, K2 = 0.001981, Ke = 0.0325.

Substituting these values into the state equations for the system yields:

 I a   −20 −40 0   I a  0 
  
   = 0.0001 −2 32.5    + 0  r
 x N   0 −0.06154 0   x N  1 

 Ia 
y (t ) = v (t ) =  0 0.06154 0   
 
 xN 

*****

To check our assumptions, we use Eq. (3.73) and MATLAB to find the closed-loop transfer function

of the system to be:

Y (s) 2s + 40 2(s + 20) 2

T (s) = = 3 = 2 = 2
R(s) s + 22s + 42s + 40 (s + 2s + 2)(s + 20) s + 2s + 2
2

The following MATLAB file was written to plot the step response.

A=[-20 -40 0; 0.0001 -2.000 32.5; 0 -0.06154 0];

B= [0; 0; 1];

C = [ 0 0.06154 0];

D = 0;

[num, den]= ss2tf(A,B,C,D,1);

 12-58 Chapter 12: Design via State Space

T = tf(num, den); %T is the closed-loop TF

step (T);

The characteristics displayed on the step response shown below indicate that the desired transient and

steady-state response requirements are met.

To find, analytically, the steady-state error for a unit step input, we apply Equation (7.96) to the state

equations to obtain:

−1
 −20 −40 0  0 
e() = 1 +  0 0.06154 0  0.0001 −2 32.5 0  = 0

 0 −0.06154 0  1 

Thus, the system behaves like a Type 1 system.

Step Response
1.4

System: T
1.2 Peak amplitude: 1.04
Overshoot (%): 4.32
At time (sec): 3.18 System: T
Final Value: 1
1
System: T
Settling Time (sec): 4.22
P. U. Output Speed

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6
Time (sec)
 Solutions to Design Problems 12-59

36.
The phase-variable form representation of the system is

 x1   0 1   x1  0 
 x  =  −196  10 −6    +   137.2  10 −6 u(t )
 2  −0.0024   x2  1 

x 
y = 1 0   1 
 x2 

4
The desired closed-loop poles are obtained from Ts = = 200 from which we get
 n
 n = 0.04 , so s 2 + 2 n s +  n2 = s 2 + 0.04s + 0.0016 the roots of which are
−0.02  j 0.03464 . Since integral control will be used a far-away pole is added to the
polynomial, resulting in
(s2 + 0.04s + 0.0016)(s + 0.1) = s3 + 0.14s2 + 0.0056 + 0.00016 .

Following the analysis in this section, the system with feedback can be represented by

 x1    0 1  0   0    x1  0 
 x  =   −196  10 −6 −0.0024  − 1   k1 k2  1  K e   x  + 0  r
 2           2   

 x N   − 1 0  0   x N  1 
 0 1 0   x1  0 
 
( )
=  − 196  10 + k1 − ( 0.0024 + k2 ) K e   x2  + 0  r
−6

 −1 0   xN  1 
 0

 x1 
y = 1 0 0   x2 
 x N 

The characteristic equation for this system is found by inspection to be

s 3 + (0.0024 + k2 )s 2 + (196  10 −6 + k1 )s + K e . Equation coefficients with the desired
polynomial we get k1 = 0.005404 , k2 = 0.1376 and K e = 0.00016 . We substitute
these values into the equation above to get:
12-60 Chapter 12: Design via State Space

 x1   0 1 0   x1  0 
 x  =  −0.0056 −0.14 0.00016   x  + 0  r
 2   2   
 xN   −1 0 0   xN  1 

 x1 
y = 1 0 0   x2 
 x N 

The step response of this system results in:

Step Response
1.4 System: sys
Peak amplitude: 1.15
Overshoot (%): 14.8
1.2 At time (seconds): 102
System: sys
Settling time (seconds): 210
1

0.8
Amplitude

0.6

0.4

0.2

0
0 50 100 150 200 250 300 350
Time (seconds)