State Variable Analysis of Digital Control Systems 367

From Eqns (6.39), the transfer function model is given as follows:

Y (z)
G(z) = = c(zI – F)–1 g
U(z)
LMz -0.3679 -0.2387 -0.3935 OP -1
LM0OP
-1 MM0PP
= [1 0 0]
MM 00 z
PP
N 0 z Q N1Q
z2 LM 0.2387z 0.2387 + 0.3935z OP L0O
= 2
1
z ( z - 0.3679)
[1 0 0] 0 MM z( z - 0.3679) z - 0.3679 PP MM0PP
0 N 0 z( z - 0.3679) Q MN1PQ
0.2387 + 0.3935z 0.3935(z + 0.6066)
= = (6.40)
z 2 ( z - 0.3679) z 2 ( z - 0.3679)
Note that the same result was obtained earlier in Example 3.3.

6.5 SOLUTION OF STATE DIFFERENCE EQUATIONS

6.5.1 Solution by Recursion

In this section, we investigate the solution of the state equation
x(k + 1) = Fx(k) + gu(k); x(0) = D x0 (6.41)
where x is n ´ 1 state vector, u is a scalar input, F is n ´ n real constant matrix, and g is n ´ 1 real constant
vector.
In general, discrete-time equations are easier to solve than differential equations because the former can
be solved easily by means of a recursion procedure. The recursion procedure is quite simple and convenient
for digital computations.
The solution of Eqn. (6.41) for any positive integer k may be obtained directly by recursion as follows.
From x(0) and u(0), x(1) can be calculated:
x(1) = Fx(0) + gu(0) (6.42a)
Then using x(1) and u(1):
x(2) = Fx(1) + gu(1) (6.42b)
From x(2) and u(2):
x(3) = Fx(2) + gu(2) (6.42c)
M
From x(k – 1) and u(k – 1):
x(k) = Fx(k – 1) + gu(k – 1) (6.42d)

6.5.2 Closed-Form Solution

In the following, we obtain the closed-form solution of state equation (6.41).
From Eqns (6.42a)–(6.42b), we obtain
x(2) = F[Fx(0) + gu(0)] + gu(1)
= F2x(0) + Fgu(0) + gu(1) (6.43)
From Eqns (6.43) and (6.42c), we get
x(3) = F[F2x(0) + Fgu(0) + gu(1)] + gu(2)
= F3x(0) + F2 gu(0) + Fgu(1) + gu(2)
 368 Digital Control and State Variable Methods
By repeating this procedure, we obtain
x(k) = Fkx(0) + Fk – 1 gu(0) + Fk – 2 gu(1) + L + F0 gu(k – 1); F0 = I
k -1
= Fkx(0) + å Fk–1– i gu(i) (6.44)
i = 0
Clearly x(k) consists of two parts, one representing the contribution of the initial state x(0), and the other
the contribution of the input u(i); i = 0, 1, 2, ..., (k – 1).
State Transition Matrix Notice that it is possible to write the solution of the homogeneous state
equations
D x0
x(k + 1) = Fx(k); x(0) = (6.45a)
as x(k) = Fkx(0) (6.45b)
From Eqn. (6.45b) it is observed that the initial state x(0) at k = 0 is driven to the state x(k) at the sampling
instant k. This transition in state is carried out by the matrix Fk. Due to this property, Fk is known as the state
transition matrix, and is denoted by f (k):
f (k) = Fk; f (0) = I (Identity matrix) (6.46)
In the following, we discuss commonly used methods for evaluating state transition matrix in closed form.
Evaluation Using Inverse z-Transforms Taking the z-transform on both sides of Eqn. (6.45a), yields
zX(z) – zx(0) = FX(z)

where D Z [x(k)]
X(z) =
Solving for X(z), we get
X(z) = (zI – F)–1zx(0)
The state vector x(k) can be obtained by inverse transforming X(z):
x(k) = Z –1[(zI – F)–1z] x(0)
Comparing this equation with Eqn. (6.45b), we get
Fk = f (k) = Z –1[(zI – F)–1z] (6.47)

Example 6.3 Consider the matrix F =

LM 0 1OP
N-0.16 -1Q
-1
For this F, (zI – F)–1 =
LM z -1 OP
N0.16 z + 1Q
LM z + 1 1 OP
( z + 0.2)( z + 0.8) ( z + 0.2)( z + 0.8)
=M PP
MM (z + 0.-20)(.16z + 0.8) z
PQ
N ( z + 0.2)( z + 0.8)

LM 4 / 3 + -1/ 3 5/3
+
-5 / 3 OP
z + 0.2 z + 0.8 z + 0.2 z + 0.8
=M
MN z-0+.80/.32 + z0+.8/03.8 -1/ 3 4/3 PP
+
z + 0.2 z + 0.8 Q
 State Variable Analysis of Digital Control Systems 369
k –1 –1
Therefore f (k) = F = Z [(zI – F) z]
LM 4
3
z
- 13
z 5
3
z
- 53
z OP
–1 z + 0.2 z + 0.8 z + 0.2 z + 0.8
=Z MM -0.8 z
+ 3 0.8 z -1 z
+ 3 4 z PP
N 3
z + 0.2 z + 0.8 3
z + 0.2 z + 0.8 Q
=
LM 4
3 ( -0.2 )
k
- 1
3 ( -0.8)
k 5
3 ( -0.2 )
k
- 5
3 ( -0.8)
k
OP
MN -0.8
3 ( -0.2)
k
+ 0.8
3 ( -0.8)
k -1
3 ( -0.2 )
k
+ 4
3 ( -0.8)
k
PQ
Evaluation Using Similarity Transformation Suppose that F is an n ´ n non-diagonal matrix with distinct
eigenvalues l1, l2, ..., ln. We define the diagonal matrix
0 L 0 LMl 1 OP
l2 L 0 0
L=M PP
M M MM M PQ
0 L ln N0
F and L are similar matrices; there exists a nonsingular matrix P such that (refer Eqns (5.22))
L = P–1 FP
Now P F P = P–1 [FF L F]P
–1 k

= P–1 [(P L P–1) (P L P–1) L (P L P–1)]P = L k

Thus the matrices F and L are similar. Since L is diagonal, L k is given by
k k

LMl k
1 0 L 0 OP
lk2
=M
0 L 0P
Lk
MM M M M P
P
0 L kn MN 0 l PQ
The state transition matrix Fk of matrix F with distinct eigenvalues l1, l2, ..., ln may, therefore, be evaluated
using the following relation:
LMl k
1 0 L 0 OP
l k2
=P M
0 L 0P
Fk = P L k P–1 P –1
(6.48)
MM M M M P
P
0 L kn MN 0 l PQ
where P is a transformation matrix that transforms F into the diagonal form (For the general case where
matrix F has multiple eigenvalues, refer [105]; also refer Review Example 6.5 given at the end of this
chapter).

Example 6.4 Consider the matrix

F=
LM 0 1OP
N-2 -3Q
The characteristic equation of the system is
|lI – F| = l2 + 3l + 2 = 0
which yields l1 = – 1 and l2 = – 2 as the eigenvalues of F.
 370 Digital Control and State Variable Methods
Since the matrix F is in companion form, the eigenvectors3, and hence the transformation matrix, can
easily be obtained (refer Problem 5.21).

P=
1 1
=
LMOP LM OP
1 1
l1 l 2 N Q N Q
-1 -2
gives the diagonalized matrix

L = P–1 FP =
-1 0 LM OP
0 -2 N Q
From Eqn. (6.48), we may write

Fk
L(-1) 0 OP P
=P M
k
–1
k
N 0 (-2) Q
L 1 1OP LM(-1) 0 OP LM 2 1OP
=M
k

N-1 -2Q N 0 (-2) Q N-1 -1Q k

L 2(-1) - (-2)
=M
k
( -1) - ( -2 ) O
k k k

N-2(-1) + 2(-2) -(-1) + 2(-2) PQ

k k k k

Evaluation Using Cayley-Hamilton Technique The Cayley-Hamilton technique has already been explained
in the earlier chapter. We illustrate the use of this technique for evaluation of Fk by an example.

Example 6.5 Consider the matrix

F=
LM 0 1OP
N-1 -2Q
Let us evaluate f(F) = Fk.
Matrix F has two eigenvalues at l1 = l2 = –1.
Since F is of second order, the polynomial g(l) will be of the form (refer Eqns (5.98)):
g(l) = b0 + b1l
The coefficients b0 and b1 are evaluated from the following equations:
f (–1) = (–1)k = b0 – b1
d d
f (l ) = k(–1)k –1 = g (l ) = b1
dl l = -1 dl l = -1

The result is b0 = (1 – k) (–1)k

b1 = – k(–1)k
Hence f(F) = Fk = b0I + b1F

= (1 – k) (–1)k
LM1 0OP – k(–1) LM 0 1OP k
N 0 1Q N-1 -2Q
= (–1)k
LM(1 - k) -k OP
N k (1 + k)Q
3
Refer Section 5.6 for methods of determination of eigenvectors for a given general matrix F.
 State Variable Analysis of Digital Control Systems 371
State Transition Equation The solution of the non-homogeneous state difference equation (6.41) is
given by Eqn. (6.44). In terms of the state transition matrix f (k), Eqn. (6.44) can be written in the form
k -1
x(k) = f (k) x(0) + å f (k – 1 – i) gu(i) (6.49)
i = 0
This equation is called the state transition equation; it describes the change of state relative to the initial
conditions x(0) and the input u(k).

Example 6.6 Consider the system

LM x (k + 1) OP = LM 0 1OP LM x (k) OP + LM0OP (– 1)

1 1 k
N x (k + 1)Q N-2 -3Q N x (k)Q N1Q
2 2

x1(0) = 1 = x2(0)
y(k) = x1(k)
Find y(k) for k ³ 1.
Solution For the given state equation, we have

F=
LM 0 1OP
N-2 -3Q
For this F, f (k) = Fk was evaluated in Example 6.4:

f (k) = Fk =
LM 2(-1) k
- ( -2 ) k ( -1) k - ( -2 ) k OP
k k k k
N-2(-1) + 2 ( -2 ) -( -1) + 2 ( -2 ) Q
k -1
The state x(k) = Fk x(0) + å Fk – 1 – i gu(i)
i = 0

With
L0O L1O
g = M P , x(0) = M P , and u(k)= (–1)k,
N1Q N1Q
we get
k -1
y(k) = x1(k) = 3(–1)k – 2(– 2)k + å [(–1)k – 1 – i – (–2)k – 1 – i] (–1)i
i = 0
k -1
1 i
= 3(–1)k – 2(–2)k + k(– 1)k – 1 – (–2)k – 1 å
i = 0
ch2

Since4
1 k
k -1
å 1 i
ch =
1- ch
2
=–2[ 1 k
ch - 1] , we have
2 1 - 12 2
i = 0

y(k) = 3(–1)k – 2(–2)k – k(– 1)k + (–2)k [1 - 1 k

di]
2
= 3(–1)k – 2(–2)k – k(–1)k + (– 2)k – (–1) = (2 – k) (– 1)k – (–2)k
k

k
1 - ak +1
4
å aj =
1- a
; a ¹ 1.
j = 0