1 DIGITAL CONTROL SYSTEM

DESIGN 1.

Pole Assignment Method

 Pole assignment method is a method to adjust the poles of a

given system by adjusting the feeback gains from state
variables.

 This method can be applied to such a system that is

described by the state equations.

1
 x1
+ z
-1

-1 x2 Y
R Q + z c
+
+
x3
+ -1
z

state feed back

K

 Discrete time state equation is

x(k + 1) = Px(k) + qu(k)

 Since a linear combination of the state variables is fed back

to the
2
 differential node, the input u(k) is replaced by

u(k) = r(k) − Kx(k).

 Where, K = [K 1 , K 2 , · · · K n ], a row vector, assuming the

system is of nth order. x(k) is a column vector of size n.
Substituing u(k) in the state equation yields,

x(k + 1) = Px(k) + q[r(k) −

Kx(k)]
= [P − qK]x(k) + qr(k)
 Writing the z-transform of this equation, we have

zx(z) = [P − qK]x(z) + qr(z)

[zI − P + qK]x(z) = qr(z)

x(z) = [zI − P + qK] − 1 qr(z)
 Therefore, the poles of the closed loop
system with the state
3
 feedback is found from

det [zI − P + qK] = 0.

 The solutions of this equation, z = λ 1 , z = λ 2 , · · · , z = λ n

are obtained from the expanded form of the determinant,
i.e. characteristic equation,

(z − λ 1 )(z − λ 2 ) · · · (z − λ n ) = 0.

 Pole assignment method determines the state feedback

gain vector
 K for a given set of poles at z = λ 1 , z = λ 2 , · · · , z = λ n .

4
z-plane poles of 2nd order systems

 Referring to the standard 2nd order transfer

function,
G 2 (s) = 2 ω2n
s + 2ζω n s + ω2n ,
 the poles of this system are located at

,
s = −ζω n ± j ω n 1 − ζ 2 .

 This equation defines damping ratio ζ and damped

natural frequency ωd as

,
ζ = cos θ and ωd = ω n 1 − ζ 2 .

 In the z-plane, poles are

,
z=e sT
=e − ζ ωn T ƒ
± ωn T 1 − ζ2 = rƒ ± θ
5
 from e− ζ ωn T
=r

Thus,

damping ratio

natural freq.

time const.
1 −T
τ = =
ζω n ln
r
6
Pole Assignment for 2nd Order Systems
 Consider the second order case of

x(k + 1) = [P − qK]x(k) +
qr(k).
Let

K = [K 1 , K 2 ]

7
8
The characteristic equation of the 2nd order system takes either
of
two real roots

(z − γ 1 )(z − γ 2 ) = z2 − (γ1 + γ 2 )z + γ1γ2 = 0

complex roots

(z − α − jβ)(z − α + j β ) = z2 + 2αz + α2 + β2 = 0

Equating the characteristic equation of unknown K and

that defined by two desired poles, we have a set of
simultaneous equations.

For the case of real roots,

p11 − q1K1 + p22 − q2K2 = γ1 + γ2
(q2p12 − q1p22)K1 + (q1p21 − q2p11)K2 + (p11p22 − p12p21) = γ1γ2

9
10
11
The characteristic equation jzI − Acj = 0 becomes,

z 2 +(0.00480K 1 +0.0952K 2 −1.905)z+0.00468K 1 −0.0952K 2 +0.905 =

0

When the desired characteristic equation is,

α c (z) = (z − λ 1 )(z − λ 2 ) = z2 − (λ 1 + λ 2 )z + λ 1 λ 2 = 0
K 1 = 105[λ 1 λ 2 − (λ 1 + λ 2 ) + 1.0]
K2 = 14.67 − 5.34λ 1 λ 2 − 5.17(λ 1 + λ 2 )

14
The unity gain feedback gives satisfactory ζ but the time
constant
2.12 sec. is too large. Try to reduce the time constant to 1
sec.
−T
ln r = τ = −0.1/1 = −0.1 =⇒ r = 0.905
2 2
ln r
θ2 = 2
ζ2 − ln r = (0.193)

16
17
2 Ackermann’s Formula
Since it is difficult to extend the approach demonstrated in
the second order case to a higher order system, the pole
assignment must rely on a more practical procedure called
Ackermann’s formula. When we know the desired poles,
z = λ 1 , z = λ 2 , · · · z = λ n for a n-th order system, write
the characteristic equation,

18
 ×

19
MATLAB Solution

23
in z-plane. The following MATLAB program uses Ackermann’s
method to move the poles of this system to an assigned
location. This program has four preassigned pole locations:

1. Case (1) [1, 0.97+j*0.05, 0.97-j*0.05] under damped

2. Case (2) [1, 0.98+j*0.05, 0.98-j*0.05] under damped

3. Case (3) [1, 0.97, 0.97] critical damping

4. Case (4) [1, 0.95, 0.95] critical damping

Pay attention to the gain (final position) for a pulse input that
drive the motor to turn for a limited time duration. If the real
part of the two poles are moved away from z = 1, more damping
slows the motor and the number of turns (rotational angle) gets
less.

26
 Open loop response to pulse
1.5

0.5

0
0 5 10 15 25
20

Response without state

feedback

27
 State Feedback by Pole Assignment
1.4

1.2

0.8

0.6

0.4

0.2

0
0 5 10 20 25

15

Response for Case

1

28
 State Feedback by Pole Assignment
1.4

1.2

0.8

0.6

0.4

0.2

0
0 5 10 20 25

15

Response for Case

2

29
 State Feedback by Pole Assignment
3.5

2.5

1.5

0.5

0
0 5 10 20 25

15

Response for Case

3

30
 State Feedback by Pole Assignment
1.4

1.2

0.8

0.6

0.4

0.2

0
0 5 10 20 25

15

Response for Case

4

31
Assignment No. 6

1. Design a state feedback control system that makes the

following 3rd order underdamped system to be critically
damped by the pole assignment method. (this system is
similar to Figure 9-1 in page 338 in textbook, but 3rd order
instead of 2nd order)

- - > s am pl e r - - > ZOH - - > 1 / ( s ( s ^ 2 + s + 1 ) ) - - > sa m p le r -->

First convert this given transfer function to a state equation by

tf2ss in analog domain, then use c2d to convert the obtained
continuous time state equation into a discrete time state
equation. Then, apply Ackermann’s design formula. Use
T=0.05 second. MATLAB program for this problem is
provided by the name PoleAssignment2008.m The specific
assignment is, therefore, to run this program and interpret
the 37
 results. In particular, calculate the steady state DC value
for the rectangular pulse of 25 ones for each case of the 4
state feedback systems resulted from the pole assignment
method.

2. Do Problem 9-11 in page 378. Try both manual

calculations and MATLAB solution.

38
STATE ESTIMATION
• The state feedback by the pole assignment method designs
an effective controller to satisfy the desired transient
characteristics related to the pole locations.
• This method, however, requires that all the state variables
are measurable and available for feedback.
• Since not all state variables are necessarily available, we
need to use an observer that estimates (or predicts) the state
variables.
• The technique to design the prediction observer also
uses Ackermann’s formula.
• The prediction observer leads to a controller that
implements the pole assignment together with the state
estimation. 39
4 Observer Model

Our plant is described

by
x(k + 1) = Ax(k) + Bu(k)
y(k) = Cx(k)

Using the z-
transform
zX( z) = AX(z) + BU(z)
X( z) = (zI − A) − 1 BU(z)

40
The observer is a system to observe (estimate) the state x(k +
1) from input u(k) and output y(k) . Thus, the equation of the
observer is

q(k + 1) = Fq(k) + Gy(k) + Hu(k)

Q( z) = (zI − F) − 1 [GY(z) +
HU(z)]
From
Y(z) = CX(z), X(z) = (zI − A) − 1 BU(z)

Q( z) = (zI − F) − 1 [GY(z) +
HU(z)]
41
 Q(z) = (zI − F) −1 [GCX(z) + HU(z)]
Q(z) = (zI − F) −1 [GC(zI − A) − 1 B + H]U(z)

The state estimation is to make Q(z) = X(z) for the same

input
U(z).
Q(z) = X(z) = (zI − A) − 1 BU(z)
Thus,

(zI − A) − 1 B = (zI − F) −1 [GC(zI − A) − 1 B + H]

42
 (zI − A) − 1 B = (zI − F) −1 [GC(zI − A) − 1 B + H]
(zI − A) − 1 B = (zI − F) −1 GC(zI − A) − 1 B + (zI − F) − 1 H

[I − (zI − F) −1 GC](zI − A) − 1 B = (zI − F) − 1 H

(zI − F) − 1 [zI − (F + GC](zI − A) − 1 B = (zI − F)
−1
H
[zI − (F + GC](zI − A) − 1 B = H
(zI − A) − 1 B = [zI − (F + GC)]−1 H
If we choose the input matrix of the plant B equal to that of
the estimator, H, then
A = F + GC
The observer state equation is rewritten as

q(k + 1) =43 Fq(k) + Gy(k) + Hu(k)

 q(k + 1) = (A − GC)q(k) + Gy(k) + Bu(k)

This is the equation of a prediction observer. q(k + 1) is the

predicted estimate of x(k). The prediction observer must
satisfy the characteristic equation,

jzI − A + GCj = 0

44
5 Errors in Estimation

The error in the state estimation process

is

e(k) = x(k) − q(k)

e(k + 1) = x(k + 1) − q(k + 1)
= Ax(k) + Bu(k) − [(A − GC)q(k) + Gy(k) + Hu(k)]
= Ax(k) + Bu(k) − [(A − GC)q(k) + GCx(k) + Bu(k)]
= (A − GC)[x(k) − q(k)]
= (A − GC)e(k)

Thus, the error dynamics have the same chracteristic equation

as that of the observer,

jzI − (A − GC)j = 0

45
All the matrices in the observer state equation except G are from
the plant state equation. G can be independently determined
from the characteristic equation of the error dynamics,

α e (z) = jzI − (A − GC)j = 0

The characteristic polynomial is written as

α e (z) = z n + α n 1 z n − 1
+α zN − 2 + · · · + α z +
α =0
−
n−2
1
0

46
Poles of
α e (z)
An approach usually suggested for choosing α e (z) is to make
the
observer two to four times faster than the closed loop control
system designed by the pole assignment due to α c (z) = 0. The
fastest time constant of the closed-loop system, determined
from the system characteristic equation,

det [zI − P + qK] = 0,

is calculated first. The time constants of the observer are then

set at a value equal to from one-forth to one-half this fastest
time constant.

49
q(k + 1) = Fq(k) + Gy(k) + Hu(k)
q(k + 1) = (A − GC)q(k) + Gy(k) + Bu(k)
q(k + 1) = Aq(k) + G[y(k) − Cq(k)] + Bu(k)

50
51
52
53
54