Imthias Ahamed

State Space Approach

In this chapter we learn

 About state space models
 State space models of electrical and mechanical systems
 Concept of state variables, state vector, state trajectory and state
space
 Different forms of state space models: physical variable, phase
variable and diagonal forms
 Canonical and Jordan canonical form
 Conversion of state space model to transfer function
 Conversion of transfer function to state space
 Eigen values
 Eigen vectors
 Diagonalisation
 State diagram
 Solution of state equation

11.1 Introduction
In early part of the nineteenth century, control system analysis and design
were based on transfer function approach or classical approach. In transfer function
approach, systems are modeled using transfer function. As we have seen, Transfer
function is obtained by taking Laplace transform of differential equations representing
the system. Finding time response involves using inverse Laplace transform which is
a tedious process. In this approach, stability analyses are carried out mainly using
graphical methods such as Evans root locus method and Bode Plot. Even though these
methods are easy to use, they give only approximate results. Controller design using
root locus and Bode plot is an iterative procedure. In classical approach, initial
conditions are assumed zero. Hence analysis of system with non-zero initial
conditions are not possible. Moreover, analysis of Multi Input Multi Output (MIMO)
is difficult.
In the later half of nineteenth century, there were tremendous developments in
the field of digital computers, linear algebra and numerical methods. These
developments resulted in the emergence of modern approach or state space approach
to control system analysis and design.
In this approach systems are modeled using two equations:
(i) A matrix differential equation, which is known as the state equation

(ii) An algebraic equation, the output equation.

These two equations together is known as the state space model. The structure of the
state space model does not change as the complexity of the system increases. Hence,
once the model is obtained, the method of analysis is similar for all the systems. In
particular analysis of MIMO system is as easy analysis of single input single output
Control Systems
1
 Ch11_may2008 Imthias Ahamed

(SISO) system. Moreover, increase in the order of the system does not increase the
difficulty proportionally (A computer program which can solve two differential
equations can solve ten first order differential equation with little additional effort.) In
state space approach we can easily take into consideration, initial conditions.
In this chapter, first we study how to obtain state space model of electrical,
mechanical and electromechanical systems. Even though state space models have
many advantages, there are many systems which were modeled using transfer
function approach. Hence we learn how to obtain a state space model from a given
transfer function. We also learn how to obtain transfer function from a given state
space model. For a given system, we can obtain many state space models. However,
all the models for a given system will have the same eigen values. Eigen values and
eigen vectors have an important role in state space approach. Hence, we learn about
eigen values and eigen vectors.
Finding the response of a system involves solving state equation. In practice,
software tools are used to solve state equations. However, to get insight, we explain
how to solve state equation analytically.

11.2 State Space Model of an RLC Network

R L

ei(t) vc(t) C e0(t)

Figure 11.1: RLC Network

Consider the RLC network shown in fig 11.1. We can view it as a single input, single
output system. The input of the system is ei(t) and the output is e0(t). We can
represent the system by the following differential equations.

di L (t )
ei (t )  R i L (t )  L  v C (t )  (11.1)
dt
1
vC (t )  i L (t ) dt  (11.2)
C ò

We have seen that we can represent this system using the transfer function.
1
Y (s) LC

U (s) 2
s  R s 1
L LC

State Space Approach

2
 Imthias Ahamed

It may be recalled that transfer function approach involves converting differential

equations to algebraic equations by taking Laplace transforms and the transfer
Y ( s)
function G ( s )  is obtained by rearranging the terms. To find the output,
U ( s)
first Y(s)=G(s)U(S) is found. Then the output is found by taking Laplace inverse of
Y(s). One limitation of transfer function approach is that it assumes initial conditions
to be zero. Moreover for complex system finding Laplace inverse is difficult.
Another approach to model the system in a compact fashion is state space
approach. In this approach we define a set of variables known as state variables1 and
represent the system by a set of first order differential equations.
For the system given in fig 11.1 define
x1(t) = iL(t)
x2(t) = vc(t)

given input u(t) = ei(t)

y(t) = vc(t) = eo(t)
dx(t ) . d 2 x(t ) ..
A convenient notation for is x(t ) and is x (t )
dt dt 2
We rewrite equation (11.1) in terms of x1(t), x2(t), u(t) and y(t)
.
u(t) = R x1(t) + L x1 (t) + x2(t) … (11.3)

x2(t) = 1/C ò x1 (t ) dt … (11.4)

.
(11.3)  L x1 (t )   Rx1 (t )  x 2 (t )  u (t )  (11.5)
. R 1 1
 x1 (t )   x1 (t )  x 2 (t )  u (t )  (11.6)
L L L
. 1
(11.4)  x 2 (t )  x1 (t )  (11.7)
C

We can write (11.6) and (11.7) in a compact form using matrix notation2

 R 1 
 .     1
x (t)
 1  L L   x1 (t) 
   L  u(t)  (11.8)
 .   1    
 x 2 (t)   0   x 2 (t)  0
 C 

We can also write the output in terms of state variables. In this example

y(t) = vc (t) = x2(t)

1
A formal definition of state variables will be given in 11.6
2
Refer Appendix for a revision on Matrix Algebra
Control Systems
3
Ch11_may2008 Imthias Ahamed

 x1 (t ) 
i.e. y(t) = [0 1]  x 2 (t )  (11.9)
 

Thus we can model the RLC network by the matrix equations (11.8) and (11.9)
which is known as the state space model. Equation (11.8) is known as the state
equation and (11.9) is known as the output equation.

11.3 Conversion nth Order Differential Equation to n First Order Differential

Equations

The main concept in state space modeling is conversion of an nth order

differential equation to n first order differential equations.

Consider the differential equation

d 3 y (t ) d 2 y (t ) dy (t )
3
 a1 2
 a2  a3 y (t )  b0 u (t )
dt dt dt

We can rewrite the above equation in y(t) and u(t) in terms of a set of new variables

x1(t), x2(t) and x3(t).

Set of new variables x1(t), x2(t) and x3(t) are defined as follows:

x1(t) = y(t)
.
x2(t) = y (t ) … (11.10)
..
x3(t) = y (t )

If we take derivative of x1(t), x2(t) and x3(t), we will get the following set of I
order differential equations.

. . .
x1 (t )  y (t )  x 2 (t )  (11.11) (by definition x 2 (t )  y (t ) )
. ..
x 2 (t )  y (t )  x3 (t )  (11.12)
. ... .. .
x3 (t )  y(t )   a1 y (t )  a 2 y (t )  a3 y (t )  b0 u (t )

.. .
Substituting for y (t ) , y (t ) and y(t) from (11.10), we get
.
x3 (t ) = -a1 x3(t) - a2 x2(t) – a3 x1(t) + b0 u(t) … (11.13)
We now rewrite (11.11), (11.12), and (11.13) below
State Space Approach
4
 Imthias Ahamed

.
x1 (t )  x 2 (t )
.
x 2 (t )  x3 (t ) (11.14)
.
x3 (t ) = -a1 x3(t) - a2 x2(t) – a3 x1(t) + b0 u(t)
We write (11.14) in a compact form as given below
 . 
 x1 .(t )   0 1 0   x1 (t )  0
 x (t )  0  0 1   x 2 (t )   0  u (t )  (11.15)
  
 2.  
 x3 (t )   a3  a 2  a1   x3 (t )  b0 
 

11.4 State Space Model Of Linear Time Invariant System

D
.
X +
+ X
u
B + C +
+
y

A
Fig. 11.2: State Space Model of Single Input Single Output LTI System

State space model can be used to model single input single output (SISO) and multi
input multi output (MIMO) system. First let us consider a SISO system.
State space model of a single input single output LTI system is of the form
.
X (t )  AX (t )  BU (t )  (11.16)
Y (t )  CX (t )  DU (t )  (11.17)
Where

Control Systems
5
Ch11_may2008 Imthias Ahamed

 x1 (t ) 
 x 2 (t )
X(t) =   is the state vector
  
 x n (t )

u(t) is the input

y(t) is the output
A is the system matrix of order n x n
B is the input matrix of order nx1
C is the output matrix of order 1xn
D is a 1x1 matrix

Equation (11.16) is called the state equation and (11.17) is called the output
equation.

We can represent the state space model as shown in figure 11.2.

For MIMO system the state space model is similar to the SISO model given above.
The differences are B matrix has as many columns as the number of inputs, and C
Matrix has as many rows as the number of outputs of the system. For example, the
state space model of an nth order system with m-inputs, and p-outputs is of the form
.
X (t )  AX (t )  BU (t )
Y (t )  CX (t )  DU (t )
Where
A is n x n matrix
B is n x m matrix
C is p x n matrix
D is p x m matrix

11.5 State Space Model Of Physical Systems

Let us now learn how to obtain state space model of a physical system. This involves
the following steps.
step 1. Obtain the differential equations describing the system.
step 2. Determine the number of state variables which is equal to the order of the
system.
Step 3. Define the state variables. The state variables are defined in such a way that it
.
is easy to write the first derivative of the state variables (say x ) in terms of the state
variables and input.
. . .
step 4. Obtain x1 , x 2 , ….. x n in terms of x1, x2, ….xn and u.
step 5. Write the state equation in the matrix form
.
X = AX + Bu
step 6. Obtain an equation for the output in terms of the state variable and inputs.
Write the output equation in the matrix form.
Y = CX + Du.

State Space Approach

6
 Imthias Ahamed

Example 11.1
Obtain the state space model of the system given in fig 11.3, where F(t) is
the input and y(t) is the output

y(t)
K

M F(t)
B

Fig 11.3: Example 11.1

Solution:
Given u(t) = F(t)
Step 1: obtain the differential equation of the system:
From Newton’s laws of motions, we get
.. .
F(t) = M y (t) + B y (t) + Ky(t)
Step 2 :
Number of the state variables
= order of the system = 2
Step 3:
Define the state variables:
x1(t) = y(t)
.
x2(t) = y (t)
. .
Step 4: Obtain equations for x1 , x 2 :
. .
x1 (t) = y (t) = x2(t) ….(i)
. ..
K B . 1
x 2 (t) = y (t ) =y (t )  y (t )  F (t )
M M M
Rewriting the above equation in the terms of x1, x2 and u we get
.
K B 1
x 2 (t) = x1 (t )  x 2 (t )  u (t )  (ii )
M M M
Step 5: Write (i) and (ii) in matrix form:

Control Systems
7
Ch11_may2008 Imthias Ahamed

 . 
 x1 .(t )   0 1   x1 (t )  0 
 x (t )    K B    
 2  M   x 2 (t )  1 u (t )  (iii )
M
      
   

Step 6: Obtain an equation for output in terms of state variables:

By definition
y(t) = x1(t)
 x1 (t ) 
i.e. y(t) = [1 0]  x2 (t )
 
Most of the variables which we are considering are functions of time. So if it
is clear from the context, we won’t be explicitly writing the variables as functions of
time.
. .
For example, instead of writing x1 (t), we will write x1 . Instead of writing y(t), we
will write y.
Most of the systems will be represented by more than one higher
order differential equations. Let us now consider a mechanical system represented by
two differential equations.
Example 11.2
Obtain a state space model of the system given in figure 11.4 if input is F(t) and
output is y1(t).

y y
K1 K2

M1 M2 F(t)
B1 B2

Fig 11.4: Example 11.2

State Space Approach

8
 Imthias Ahamed

Solution:

y y

K1 y1 K2(y2-y1)

F(t)
B1 y1 B2(ý2-ý1)
M1 ÿ1 M2 ÿ2

Fig. 11.5: Free body diagram of the system given in example 11.2

Step 1: To obtain the differential equations, we first draw the free

body diagram. (Refer fig. 11.5)
The differential equations representing the system are
.. . .
F  M 2 y 2  B2 ( y 2  y 1 )  k 2 ( y 2  y1 ) ... (i )

. . .. .
k 2 ( y2  y1 )  B2 ( y 2  y 1 )  M 1 y1  B1 y 1  K 1 y1 ... (ii )

Step 2: As there are two second order differential equations, the

order of system = 4. Hence, number of state variables = 4.

Step 3 : Define state variables as

x1 = y1
.
x2 = y 1
x3 = y2
.
x4 = y 2
Step 4:
. .
x1  y1  x1  y1  x 2 ... (iii )
. . ..
x 2  y1  x 2  y1

From (ii), we get

Control Systems
9
Ch11_may2008 Imthias Ahamed

.. k2 B2 . . B1 . K1
y1  ( y 2  y1 )  ( y 2  y1 )  y1  y1
M1 M1 M1 M1

Rewriting the above equations in terms of x1, x2, x3, x4, we get
. k2 B2 B1 K1
x2  ( x3  x1 )  ( x 4  x2 )  x2  x1
M1 M1 M1 M1
Rearranging the above equation, we get
.  K1 k  B  B2 k2 B2
x2  (  2 ) x1  ( 1 ) x2  x3  x4 ... (iv)
M1 M1 M1 M1 M1 M1
. .
x3  y 2  x 3  y 2  x 4 ...(v )
. . ..
x4  y 2  x4  y 2

From (i), we get

.. 1 B2 . . K2
y2  F ( y 2  y1 )  ( y 2  y1 )
M2 M2 M2
Rewriting the above equation in terms of x1, x2, x3, x4, and u, we get
. 1 B K
x4  U  2 ( x 4  x2 )  2 ( x3  x1 )
M2 M2 M2
Rearranging the terms, we get
. K B K B 1
x4  2 x1  2 x2  2 x3  2 x4  U ...(vi )
M2 M2 M2 M2 M2
Step 5: Write equations (iii), (iv), (v) and (vi) in matrix form

.   0 1 0 0  0 
 x1   k  k B  B2 k2 B2   x1   
1 2 1
 .   ( ) ( )  x  0 
x
 2 M1 M1 M1 M1   2  + 0  U
.   
0 0 0 1   x3   
x3   k2 B2  k2  B2     1 
 .  
M 2   x4   M 
 x 4   M 2 M2 M2  2

Step 6 output y = y1 = x1
 x1 
x 
 y  1 0 0 0  
2

 x3 
 
 x4 
All the systems which we have considered are SISO systems. S.S
approach can be used to model and analyse MIMO system also. Let us consider such
a system as our next example

State Space Approach

10
 Imthias Ahamed

Example 11.3

y1 y3

R1 R3

i1 R2 i2

y2
L1 L2

Vc
U1 U2

Fig. 11.6: Example 11.3

Obtain the state space model of the two input, three output system given in fig 11.6.
Input (u1 and u2) and output (y1, y2, and y3) are as shown in figure.
Solution:
Let the loop currents be i1, and i2 as shown in figure 11.6
Applying KVL to loop 1, we get
di
u1  L1 1  R1i1  R2 (i1  i2 )  vc ...(i )
dt
Applying KVL to loop 2, we get
di
u 2  L2 2  R3i2  R2 (i1  i2 )  vc ...(ii)
dt
current through the capacitor
dv
i1  i2  C c ...(iii )
dt
The given RLC network has three storage elements; hence the order of the system is
3. Therefore, number of state variables = 3. Choose inductor currents and capacitor
voltage as state variables. That is,
x1  i1
x 2  i2
x3  vc
. di1
x1  i1  x 1 
dt
From (i), we get
di1 1 R R v
 u1  1 i1  2 (i1  i 2 )  c
dt L1 L1 L1 L1
Rewriting in terms of x1, x2 and x3, we get
Control Systems
11
Ch11_may2008 Imthias Ahamed

. 1 R R x
x1  u1  1 x1  2 ( x1  x 2 )  3
L1 L1 L1 L1
( R1  R2 ) R 1 1
 x1  2 x 2  x3  u1 ...(iv )
L1 L1 L1 L1
. di
x2  i2  x 2  2
dt
From (ii), we get
di2 1 R R 1
 u 2  3 i2  2 (i1  i2 )  vc
dt L2 L2 L2 L2
Rewriting in terms of of x1, x2 and x3, we get
. 1 R R 1
x2  u 2  3 x 2  2 ( x1  x 2 )  x3
L2 L2 L2 L2
R ( R  R3 ) 1 1
  2 x1  2 x2  x3  u 2 ...(v)
L2 L2 L2 L2
. dvc
x3  vc  x 3 
dt
From (iii), we get
dvc 1
= (i1  i 2 )
dt c
Rewriting in terms of of x1, x2 and x3 we get
. 1 1
x 3  x1  x 2 ...(vi )
c c
Now, write equation (iv), (v) and (vi) in matrix form to obtain the state equation.
  ( R1  R2 )  R2  1 1 
.     x  L 0 
x
 u1 
1
   L L L 1
1
1 1 1
   
 .    R2  ( R2  R3 ) 1   x 2   1  
x2    
+ 0
 x3    u2
.   L 2 L 2 L 2 L 2
1 1     0 0 
 x3   0     
   C C   

From fig 11.6, it is clear that

y1  R1i1  y1  R1 x1 ...(vii)
y 2  R2 (i1  i2 )  y 2  R2 x1  R2 x 2 ...(viii)
y3  R3i2  R3 x2 ...(ix)
Write the above three equations in matrix form to obtain the output equation.

 x1 
 y1   R1 0 0 x 
 y   R R2 0  2
 2  2  x3 
 y 3   0 R3 0  
 

State Space Approach

12
 Imthias Ahamed

Example 11.4
Consider a system represented by the following differential equations
y1  2 y 1  4 y1  2u1
y 2  5y 2  3y 2  u1  3u 2
Obtain a state space model.
Solution :
Step 1: Given
y1  2 y 1  4 y1  2u1 (1)
y 2  5y 2  3y 2  u1  3u 2 (2)
We can see that the system is represented by two second order differential
equations. Therefore, the system is a fourth order system.
Step 2:
Number of state variables = 4
Step 3 :
Choose state variables x1, x2 x3 and x4 such that we can find x 1 , x 2 , x 3 and x 4 in
terms of x1, x2, x3, x4, u1 and u2.
Choose
x1  y
x 2  y 1
x 3  y2
x 4  y 2
Step 4 :
x 1  y 1  x 2
x 2  y1  4 y1  2 y 1  2u1  4x1  2x 2  2u1
x 3  y 2  x 4
x 4  y 2  3y 2  5y 2  u1  3u 2
 3x 3  5x 4  u1  3u 2
Step 5 : Write the above equations in a matrix form
 x 1   0 1 0 0   x 1  0 0
x   4  2 0 0   x 2  2 0  u1 
 2   
 x 3   0 0 0 1   x 3  0 0 u 2 
      
x 4   0 0  3  5  x 4  1 3
Control Systems
13
Ch11_may2008 Imthias Ahamed

11.6 Concept of State Variable and State Space

A formal definition of state variable has been deliberately differed, so that
students get an intuitive feel of the terms.
State space model of a system consists of three types of variables,
input variables, output variables and state variables. The input variables are denoted
by u1(t), u2(t), … up(t); the output variables are denoted by y1(t), y2(t), … ym(t) and
state variables are denoted by x1(t), x2(t), …xn(t).
The state variables of a system is a minimal set of variables such that the
knowledge of these variables at t = t0 together with the knowledge of inputs for t  t0
completely determines the behavior of the system for t > t0.
For convineance, state variables may be represented as a vector known as
state vector, X(t);
where
( )
⎡ ⎤
( )
( )= ⎢ ⎥
⎢ ⋮ ⎥
⎣ ( )⎦

X2

X(to)= (X1(to),X2(to)

X1
X1(to)

Fig 11.7 : A two dimensional state space ,

with a specified initial state.

Fig. 11.7 A Two dimensional State Space with a Specified Initial State

At a particular time, say t1, value of state vector, X(t1) denotes the state of the system
at t=t1. For example, X(t0) denotes the initial state of the system. The value of state
vector at t = t0 could be visualized as a point in the n dimensional space with n axis.
For a two dimensional system, the state space is a two dimension space or a state
plane. The initial state
 x1 (t 0 )
X(t0) =   can be marked on the state space as shown in fig 11.7. Similarly state
 x 2 (t 0) 
of the system at any time can be marked on the state space. For example the state of
the system at t = t1 (X(t1)) is marked in fig 11.8. As t increase the state of the system
can change. The trajectory formed by the state points as t increase is known as the
state trajectory. A particular state trajectory is given in fig 11.8.

State Space Approach

14
 Imthias Ahamed

X2

X(t1) X(t2)

X(t3)
X(to)

X1

Fig 11.8 : state trajectory

Fig. 11.8 State Trajectory

To make the concepts of state variable, state space and state trajectory clearer, let us
consider the R.L.C network shown in figure 11.9. The input u(t) of this system is e1(t)
and output y(t) is vc(t), the voltage across the capacitor. The state variables are: x1(t)
= inductor current = iL (t) and x2(t) = capacitor voltage = vc(t).

R L

ei(t) vc(t) C e0(t)

Figure 11.9: RLC Network

Let us now see whether iL(t) and vc(t) would qualify as state variables
according to the definition given. Recall: “The state variables of a system is a
minimal set of variables such that the knowledge of these variables at t = t0 together
with the knowledge of inputs for t  t 0 completely determines the behaviour of the
system for t>t0”.
We know from fundamentals of electric circuit theory that if we know the
initial value of inductor current and the initial value of capacitor voltage; and in
addition if the input voltage is completely specified for t>t0, we can find current
through the circuit iL(t) and the capacitor voltage vc(t) at all instants of time. Hence
iL(t) and vc(t) qualifies as state variables. For this system, state vector
i L (t )  i L (t 0 ) 
X(t) =   and Initial state X(t0) =  
vc (t ) vc (t 0 )
Control Systems
15
Ch11_may2008 Imthias Ahamed

To understand the concept of the state trajectory, consider the

following situation. Let the capacitor be initially charged to 2 V and the current
through the inductor be 0. That is, IL(0) = 0 and vc(0) = 2. In vector notation,
0 
X(0) =  
2
Let u(t) = 0 for t  0. That is, the voltage source is replaced by a short circuit. The
value of initial state and input is shown in fig 11.10(a) and the initial state is marked
on the state space shown in fig 11.10(b). As time progress, the capacitor discharges
through the inductor and resistor. The direction of the current will be from B to A,
and the magnitude of current will slowly increase. In fact, we can solve the
i L (t1 ) 
differential equations describing the system and calculate X(t1) =   , X(t2) =
v c (t1 )
i L (t 2 ) 
v (t ) … etc and plot these points on the state space to get the state trajectory. In
 c 2 
section11.10, we will study how to solve the state equation. Refer ex.11.10 for a
detailed example where the state trajectory of a particular RLC network is plotted.
A B
iL(0)=o
U(t)=ei(t) V(0)=2

(a)

initial state 0
X(t1) 2
X(t2)

state trajectory
of the RLC circuit
given in fig 11.6.4(a)

Fig 11.10 (a) the RLC network shown in fig11.6.3

with e i(t)=0 i L=0 and vc (0)=2

Figure 11.10

11.7 Conversion of Transfer Function to State Space Model

In classical approach, control system analyses were based on transfer function
approach. Hence, in many situations, transfer function may be known and we may
want to find the state space model. In this section, we learn how to obtain state space
model from the transfer function. For a given system, state space model is not unique.
We can obtain different state space models for the same system. We study how to
obtain state space model in two specific forms: Phase variable form and canonical
form.

11.7.1 State space model in phase variable form

Case (i) Systems without numerator dynamics:

State Space Approach

16
 Imthias Ahamed

Y(s) 1
Systems of the form  2 are examples of systems without
U(s) s  as  b
numerator dynamics. In simple terms, these are systems without “s” terms in the
numerator, or in other words these are systems without zeroes.
Steps involved in finding the state space model of an nth order system are
Step 1: Obtain the differential equation from the transfer function
Step 2: Define the output as the first state variable; the derivative of the output as the
second state variable and so on. That is
x1=y
x2  y
d n 1 y
xn 
dt n 1

Step 3 : Obtain x 1 , x 2 .................x n -1 :

They are by definition (step 2) given as
x 1  x 2
x 2  x 3

x n -1  x n
Step 4 : obtain x n from the differential equation obtained in step 1.
Step 5: Write the equations obtained in step 3 and step 4 in the form x  Ax  Bu
Step 6: Obtain the output equation:
By definition,
y = x1
i.e. y = [1 0……….0] X.
We will make the steps involved clear using the following example.
Example 11.5
Obtain a state space model in phase variable form of a system with transfer function
 s  1
 3
 s  s  7s  8s  2
2

Solution :
Step 1: Obtain the differential equation from the transfer function
Cross multiplying, we get
 
s 3  7s 2  8s  2  (s)   ( s)
3 2
s (s)  7s (s)  8s (s)  2 (s)  (s)
Take L-1 on both sides
L-1 (s 3  (s)  7s 2  (s)  8s (s)  2(s))  L-1  (s)

Since L-1 is a linear operator we get

   
L-1 s 3 (s)  L-1 7s 2 (s)  L-1 8s (s)  L-1 2( s)   L-1  (s) ------ (11.7.1)
It may be noted that we denote L(f(t)) as F(s), L(y(t)) as Y(s) and L(u(t)) as  (s).
The usual convention is to use lower case (small) letters for the functions in time
domain and uppercase (capital) letters for functions in “S” domain
According to differentiation theorem
Control Systems
17
Ch11_may2008 Imthias Ahamed

 dy 
L    sΥ s   y(0)
 dt 
By definition of transfer function initial conditions are zero. Therefore,
y0  y 0  y0  0
 dy 
 L    s ( s )
 dt 
 L-1 ss   y (t)
Similarly
 
L-1 s 2 s   y(t)
Substituting for
 
L-1 s  , L-1 s s , L-1 s 2 s  etc, in (11.7.1), we get
y(t)  7 y(t)  8y (t)  2y(t)  u(t) (11.7.2)
Step 2 : Define state variables
x 1  y(t)
x 2  y (t) --------- (11.7.3)
x 3  y(t)
Step 3: obtain x 1 , x 2
x 1  y(t)  x 1  y (t)  x 2 (by defintion)
x 2  y(t)  x 2  y(t)  x 3 ( " )
Step 4: obtain x 3
x 3  y  x 3  y
But
y  - 2y(t) - 8 y (t) - 7y(t)  u(t) (from 11.7.2)
Substituting for y(t), y (t) and y (t) from (11.7.3), we get
x 3  y  -2x1 - 8 x 2 - 7x 3  u
Step 5: Obtain state equation in the matrix form:
From step 3 and step 4, we obtained
x 1  x 2
x 2  x 3
x 3  2x 1 - 8x 2 - 7x 3  u
Writing in the matrix form, we get
 x 1  0 1 0   x 1  0 
 x   0 0 1   x   0  u
   2  2   
 x 3  - 2 - 8 - 7   x 3  1 
Step 6: Obtain the output equation:
By definition
y = x1
x 1 
:. y = [1 0 0] x 
 2
x 3 

State Space Approach

18
 Imthias Ahamed

Example 11.6
Obtain the state space model for the system given in Fig. 11.11

U(S) 2 Y(S)
S(S+1)

Fig 11.11
Figure 11.11
Solution:
The overall transfer function for the system is

2
 (s) s( s  1) 2

 (s) 1  2
s ( s  1) 2
2
s ( s  1) 2
=
( s ( s  1) 2  2) / s ( s  1) 2
2
=
s( s  1) 2  2
 (s) 2
 3
 (s) s  2s 2  s  2
Cross multiplying, we get
 
s 3  2s 2  s  2 (s) = 2  (s)
3 2
s Υ(s)  2s (s)  s (s)  2(s)  2  ( s)
Taking L-1
L-1 (s3 (s)  2s 2 (s)  s(s)  2 (s))  L-1 (2  (s))
y  2y  y  2y  2u ----------------- (11.7.4)
Define state variables
x1  y
x 2  y
x 3  y
Obtain x 1 , x 2 and x 3
x 1  y  x 2
(By definition)
x 2  y  x 3

x 3  y  -2y - y - 2y  2u   2x 1 - x 2 - 2x 3  2u (from eqn. 11.7.4)

Control Systems
19
Ch11_may2008 Imthias Ahamed

Writing the above three equations in matrix form we get

 x 1  0 1 0   x1  0 
 x   0 0 1   x 2   0  u
 2 
 x 3  -2 -1 -2   x 3   2 
By definition y = x1.

x1 
 y  1 0 0  x 2 
 x 3 
Case (ii) Systems with numerator dynamics:
We will illustrate the method for finding state space model for systems with
numerator dynamics using the following example.
Example 11.7
The transfer function of a system is given by
 (s)
7s 2  8s  2
 (s) =
s 3  2 s 2  5s  3

Obtain the state space model in the phase variable form.

Solution:
One approach to solve problem with zeros is to write the transfer function as
 s  X  s  1

X s    s 
 
 7s 2  8s  2 3 2
s  2 s  5s  3
 s 
 7s 2  8s  2 --------------- (11.7.5)
X s 
X s  1
 3 -------------- (11.7.6)
U s  s  2 s  5 s  3
2

Cross multiplying (11.7.6), we get

 
s 3  2s 2  5s  3 X(s)   (s)
Taking L-1, we get
x  2x  5 x  3x  u ------------- (11.7.7)
Define state variables as
x1  x
x 2  x …. (11.7.8)
x 3  x

from 11.7.8, we get

x 1  x  x 1  x  x 2
x 2  x  x 2  x  x 3
x 3  x  x 3  x  - 3x - 5x - 2x  u (from 11.7.7)
= -3x1 - 5x2 – 2x3 +u (from 11.7.8)
Writing the above equations in the matrix form, we get the state equation

State Space Approach

20
 Imthias Ahamed

x 1 0 1 0   x 1  0
x 2  0
   0 1  x 2   0 u
x 3  3 - 5 - 2   x 3  1 
To get the output equation, cross multiply (11.7.5)
 
 (s)  7s 2  8s  2 Xs 
Taking L-1, we get
y = 7x  8x  2x
= 7x3 + 8x2 + 2x1
= 2x1 + 8x2 + 7x3
x 1 
y = [2 8
 
7] x 2 
x 3 
Let us now consider an example, were numerator polynomial and denominator
polynomial are of the same order
Example 11.8
Given,
 s  s 3  7s 2  3s  2

 s  s 3  6s 2  11s  5
Obtain a state space model in the phase variable form
Solution:
Given
 s  s 3  7s 2  3s  2

 s  s 3  6s 2  11s  5
Rewriting the transfer function as
 s  X(s) 1

X s   (s)
 
 s 3  7s 2  3s  2 3 2
s  6 s  11s  5
 s 
X s 

 s 3  7s 2  3s  2  ....(11.7.9)

X s  1
 3 ....(11.7.10)
 s  2
s  6s  11s  5
i.e. (s 3  6s 2  11s  5) X (s)   s  (11.7.16)
-1
Taking L , we get
x  6x  11x  5x  u (11.7.11)

Define
x1  x
x 2  x ….. (11.7.12)
x 3  x

From (11.7.12) and (11.7.11), we get

Control Systems
21
Ch11_may2008 Imthias Ahamed

x 1  x 2
x 2  x 3
x 3  x  -5x - 11x - 6x  u
 - 5x 1 - 11x 2 - 6x 3  u
Therefore,
x 1 0 1 0   x 1  0
x 2  0
   0 1   x 2   0 u
x 3  5 - 11 - 6  x 3  1 
To get the output equation, cross multiply (11.7.9)
 
 (s)  s 3  7s 2  3s  2 X (s)
Take L-1
y(t)  x  7x  3x  2x
Substituting from (11.7.12) we get
y  x 3  7x 3  3x 2  2x1 . But we want an expression for y in terms of x1, x2 and x3.
We know, x 3  - 5x1  11x 2  6x 3  u
 y  - 5x 1 - 11x 2 - 6x 3  u   7x 3  3x 2  2x1
 - 3x 1 - 8x 2  x 3  u
x 1 
 y = [-3 -8 1] x 2  + u
x 3 

11.7.2 State space model in canonical form

If the system matrix of the state space model is diagonal, we say the model is
in the canonical form or diagonal form. Diagonal matrix has many useful properties
which makes the analysis easy.
Finding inverse of a diagonal matrix is very easy. For example,

 λ1 0 0 
If A = 0 λ2 0 

0 0 3 
1 0 0 
 1 
 
A-1 = 0 1 0 
2
 
0 0 1 
 3 
Finding An is also very easy, for example

13 0 0 
3
 3

A = 0 λ2 0 
 3 
 0 0 λ3 
State Space Approach
22
 Imthias Ahamed

Example 11.9
- 1 0 0 
Given A = 0 -4 0 

0 0 - 2
-1 4
Find A , A
Solution :
- 1 0 0 
A -1
= 0 - 1/4 0 

 0 0 - 1/2

(-1) 4 0 0  1 0 0
  
A 4
= 0 (-4) 4
0  = 0 256 0 
0 0 (-2) 4  0 0 16

If the Transfer function is having distinct poles, we can obtain state space
model in the canonical form. In canonical form, the diagonal elements will be poles
of the transfer function. The A-matrix (system matrix) of a system with transfer
function:
p1 0 0
1 
will be 0 p2 0 
(s  p1 )(s  p 2 )(s  p 3 )
0 0 p 3 

If the transfer function is having repeated poles, we cannot obtain the model in the
diagonal form. However, it is possible to obtain a model in almost diagonal form or
Jordan canonical form. For example, the A matrix of a system with transfer function
1
will be
(s - 2) 2 (s - 3)

2 1  0
0 2  0

 
 
0 0  3 
Example 11.10
1
Obtain state space model of a system with transfer function 3 2
in the
s  6s  11s  6
canonical form.
Solution :
 s  1
Given 
 s  3 2
s  6s  11s  6

Step 1: Obtain partial fraction expansion

Factorize s 3  6s 2  11s  6
Control Systems
23
 Ch11_may2008 Imthias Ahamed

s 3  6s 2  11s  6  (s  1) (s  2) (s  3)

 s  1
  G(s) 
 s  ( s  1) (s  2) (s  3)

1 A B C
Let G(s)    
( s  1) (s  2) (s  3) s  1 s  2 s  3

1
A = (s  1) G(s) |s -1 
2
B = (s  2) G(s) |s-2  - 1
1
C = (s  3) G(s) |s -3 
2
1 1
 s  2 1 2
   
 s  s 1 s2 s3

Step 2: Draw a block diagram representing the transfer function in the partial fraction
form. The block diagram is given in fig. 11.12

Step 3 : Define the state variables

Define the output of each block as a state variable as shown in figure 11.12

1/2 X1(S)
S+ 1

-1
U(S) X2(S) + Y(S)
S+2

1/2 X3(S)
S+3

Fig 11.12 Block Diagram Of System With

Fig. 11.12

From figure
1
X 1 s  2

 s  s 1

State Space Approach

24
 Imthias Ahamed

i.e. (s+1) X1(s) = 1  ( s )

2
-1
Taking L , we get
1
x 1  x 1  u
2
1
i.e. x 1  - x 1  u ……. (11.7.13)
2
x 2 s  1
From figure 
u s  s2
i.e. ( s+ 2) x 2 (s) = -u (s)
-1
Taking L , we get x 2  2x 2  - u
x 2   2 x 2 + -u ………(11.7.14)
From Figure
X 3 s  1/2

 s  s3
1
i.e. (s+3) x 3 (s) = u (s)
2
Taking L-1, we get
1
x 3  3x 3  u
2
1
x 3   3x 3  u …….. (11.7.15)
2
Writing equations (11.7.13), (11.7.14) and (11.7.15) in the matrix form, we get

1 
 x 1  - 1 0 0 x1   2
 x   0 -2 0  x   - 1  u
 2   2  
 x 3  0 0 - 3   x 3   1 2 
From fig. 11.12
Y(s) = X1(s) + X2(s) + X3 (s)
Taking L-1, we get
y = x1 + x2 + x3
x1 
i.e y  [1 1 1] x 2 
 x 3 
Example 11.11
The transfer function of a system is given by
2
G(s)  2
s  1 s  2
Obtain a state space model in the Jordan canonical form

Solution :
Note that the system has a repeated pole at s=-1, therefore, it is not possible to obtain
a model in diagonal form. However, it is possible to obtain a model in Jordan -
canonical form
Control Systems
25
Ch11_may2008 Imthias Ahamed

Step 1 : Obtain partial fraction expansion

2 A B C
G(s)    
s  1 s  2 (s  1) (s  1) s  2
2 2

A = (s  1) 2 G(s) |s  -1  2
d d  2 
B =
ds

s  12 G(s)  s  1   
ds  s  2  s 1
 -2 
=  2 
 -2
 (s  2)  s 1
C = (s  2) G(s) |s-2  2

 s  2 2 2
 G(s)    
 s  s  12
s 1 s2
Draw the block diagram of the expanded transfer function (refer fig. 11.13)

From the figure

X1 s  1

X 2 s  s 1
 ( s + 1) x1 s  = x 2 s 
 x 1 = - x 1  x 2

-2

1 1 2 + Y(S)
(S+1) X2(S) (S+1) X4(S)

1 X3(S) 2
(S+2)

Fig 11.13 Block Diagram Of System With

Figure 11.13

From figure 11.13

X 2 s  1

u s  s 1
 ( s +1) x 2 s   u ( s)
 x 2   x 2  u
From the figure

State Space Approach

26
 Imthias Ahamed

X 3 s  1

u s  s2
 (s +2) x 3 s   u ( s )
.
 x 3   2 x3  u
Combining equations for x 1 , x 2 and x 3 we get
x 1 - 1 1 0   x 1  0
x 2  0
   -1 0   x 2   1  u
x 3 0 0 - 2  x 3  1 

From the figure

Y(s) = 2 X1(s) - 2 X2(s) + 2 X3(s)

Taking L-1, we get

y = 2x1 - 2x2 + 2x3

x 1 
 y  [2 - 2 2] x 2 
 x 3 

Example 11.12
The overall transfer function of a closed loop feedback system is
s 3  5s 2  3s  4
G(s) 
s 3  7 s 2  12 s
Obtain the state space model in the canonical form.
Solution :
Note that deriving state space models in canonical form for system with zeroes
is similar to systems without zeroes. However, if the numerator polynomial is of the
same degree as denominator polynomial, we have to divide the numerator polynomial
by denominator polynomial.

s 3  5s 2  3s  4 - 2s 2 - 9s  4
  1   1 + F(s)
s 3  7 s 2  12 s s 3  7 s 2  12 s
Now obtain the partial fraction expansion of second term, F(s).
- 2s 2  9 s  4 - 2s 2  9 s  4
F (S) = 3 
s  7 s 2  12 s s s  3s  4 
- 2s 2  9 s  4 A B C
Let F (s) =   
s s  3s  4  s s  3 s  4
4 1
A = s F (s)  s 0 = 
12 3
B = (s+3) F (s) s = -3

Control Systems
27
Ch11_may2008 Imthias Ahamed

 18  27  4  13
= =
3 3
 32  36  4
C = ( s+ 4) F (s) s = -4 = 2
4
1  13
2
 G (s) = 1 + 3  3 
s s3 s4
U(S)

1/3 X1(S)
S

U(S) -13/3 Y(S)

X2(S) +
S+3

2 X3(S)
S+4

Fig 11.14 State Diagram Example

Figure 11.14

Draw the state diagram and mark the state variables as shown in figure 11.14.
From the figure,
1
x1 s  3

u s  s
1
 s x1 (s) = u ( s)
3
1
 x 1  u
3
13
x2 s 3

u s  s3
 13
 s  3 x 2 s   u s 
3
 13
 x 2   3 x 2  u
3
x 3 s  2

u s  s4
 s  4 x 3 s   2 us 
 x 3   4 x 3  2 u
Writing equations of x 1 , x 2 , x 3 in matrix form, we get

State Space Approach

28
 Imthias Ahamed

1 
 
 x 1  0 0 0  x1   3 
 13 
 x   0 - 3
 2  0   x 2    u
 3 
 x 3  0 0 - 4  x 3   
 2 
 
From the figure
Y s   X 1  s   X 2  s   X 3  s   u  s 
Taking L-1
Y (t) = x1 + x2 + x3+ u
x 1 
 
Y = [ 1 1 1] x 2  + [1] U
x 3 

Example 11.13
Transfer function of a system is given by
Y(s) 1
 2
U(s) s  6s 2  8s
Obtains a state space model with
1
B  1
1

Solution :
Y (s) 1
G (s)   3
U(s) S  6s 2  8s
1
i.e., G(s) 
s(s  2)(s  4)
A B C
  
s s2 s4
1
A  s.G(s) S 0 
8
1
B  (s  2)G(s) S2 
4
1
C  (s  4)G(s) S4 
8
Y(s) 1 8 1/ 4 1 8
   
U(s) s s2 s4
Control Systems
29
Ch11_may2008 Imthias Ahamed

Draw the block diagram as shown in figure 11.15

X1(s)
1 1
s 8

X2(s) Y(s)
U(s) 1 -1
s+2 4
S

1 X3(s) 1
s+4 8

Figure 11.15 : Example 11.13

Since we want B = [1 1 1]T define state variables as shown in figure.

From figure
X1 (s) 1

U(s) s
Cross multiplying and taking
Laplace inverse, we get
x 1  u (i)
X 2 (s) 1
  x 2  2x 2  u (ii)
U(s) s  2
X 3 (s) 1
  x 3  4x 3  u (iii)
U(s) s  4
Again from figure
1 1 1
Y(s)  X1 (s)  X 2 (s)  X 3 (s)
8 4 8
1 1 1
 y  x1  x 2  x3 (iv)
8 4 8
From (i), (ii), (iii) and (iv), we get the required state space model
 x 1  0 0 0   x1  1
x   0  2 0   x   1 u
 2   2   
 x 3  0 0  4  x 3  1

State Space Approach

30
 Imthias Ahamed

 x1 
 1 1 1  
y x2
8 4 8   
 x 3 

11.8 Conversion of State Space Model to Transfer Function

In this section, we study how to obtain transfer function (TF) from
state space model. Let the state space model of an nth order system be
X  AX  BU ( i )
Y  CX  DU ....... (ii )
Y S 
To obtain the transfer function we have to obtain .
U S 
Take Laplace transform of (i) & (ii)
s X s   X (0) = AX (s) + BU (s) ………(iii)
Y (s) = CX (s) + DU (s)………….. (iv)
To obtain Y (s)/U (s), we will first obtain an expression for X(s) in terms of U(s) from
(iii). By definition of transfer function, initial conditions are zero. Therefore X (0)
=0
(iii)  s X (s) - A X (s)  B  (s)
 s I X (s) - A X (s)  B  (s) (Where I is the identity matrix of order
n)
 s I - A  X (s)  B  (s)
Pre-multiplying both sides by s I - A 1
sI - A 1 s I - A  X(s)  sI - A  B  (s)
1

X(s)  sI - A  B  (s)

1

Substituting expression for X(s) in (iv) we get
Y(s)  CsI - A  B  (s)  D  (s)
1


= CsI - A 1 B  D  (s) 
Y(s)
 C sI - A  B  D
1

 (s)
Example 11.14
State space model of a system is
- 2 0 1
x    x   u
0 - 3 1
y  1 - 1X
Obtain the transfer function
Solution:
We know T.F = C sI - A 1 B  D
- 2 0
Given A = 0
 - 3
1 
B = 1 
 
Control Systems
31
Ch11_may2008 Imthias Ahamed

C = 1 -1 
D = 0
Step 1 Find sI - A
1 0 - 2 0 
sI - A = s   - 
0 1 0 - 3
s 0 - 2 0
= 0  - 
 s 0 - 3 
s  2 0 
= 0
 s  3

Step 2 Find [sI – A]-1

Note: inverse of a diagonal matrix can be obtained by substituting reciprocals
of diagonal elements.
 1 
1
s  2 0 
s  2 0 
i.e.    
0 s  3 0 1 
 s  3 
-1
Step 3 Find [sl –A] B
1 0  1
sl - A B 
1  s2   
0 1  1 
 s  3

1 
=  s  2
1 
 s  3
Step 4 find CsI - A 1 B
1 
s  2
CsI - A B  1 - 1 
1

1 
 s  3
1 1
= 
s2 s3
1
=
s  2s  3
Step 5 Find transfer function = CsI - A 1 B  D

1 1
CsI - A  B  D 
1
0
s  2s  3 s  2s  3

State Space Approach

32
 Imthias Ahamed

Example 11.15
Given

X  0 1
X
0
- 1 - 1 1  u
  
Y = [1 0] X
Obtain transfer function.
Solution:
We know
T.F = CsI - A 1 B  D
0 1
Given A = - 1
 - 1
0 
B = 1 
 
C = 1 0
D = 0
s 0 0 1
sI - A  0 
 s  - 1
 - 1
s -1 
= 1
 s  1
For a quick revision on finding inverse refer Appendix III

adjoint of sI - A 
sI - A
1
=
determinant of sI - A 
Adjoint of sI - A  = cofactor of sI - A T
s  1 - 1
Cofactor of sI - A  = 1
 s 
s  1 1 
 Adjoint of sI - A  = - 1
 s 
Determinant of sI - A  =   s2  s 1
s  1 1
- 1 s 
 sI - A 
1
 

(s 1 1 
  
= -1 s 
  
(s 1 1  1 
   0  
sI - A 1 B = -1 1    s 
s   
    

Control Systems
33
Ch11_may2008 Imthias Ahamed

1 

 CsI - A  B  1 0 
1

s 
 
1
=

1
= 2
s  s 1
Example 11.16
State space model of a system is given below
1 
0 0 0  3 
X  0
 -4 0  X  2  u
 
0 0 - 3 - 13 3 
Y = [1 1 1] X + [1] u
Obtain the transfer function
Solution:
0 0 0 
Given A = 0 -4 0 

0 0 - 3
1 
 3 
B = 2 
 
 13 3 
C = 1 1 1; D  [1]
s 0 0 
sI - A = 0 s4 0 

0 0 s  3
1 0 0 
 s 
sI - A -1 = 0 1 0 
 s4 
0 0 1 
 s  3
1 0 0 
 s 
CsI - A -1  [1 1 1] 0 1 0 
 s4 
0 0 1 
 s  3

= 1 s 1
s4
1 
s3

State Space Approach

34
 Imthias Ahamed

1 
 3 
CsI - A 
1
B = 1 s 1
s4
1
s3
 2



 13 3 

1 / 3 2 - 13/3 
=  s  
 s4 s  3 
=
s  4 (s  3)  6s (s  3) - 13s (s  4)
3s (s  4) (s  3)

- 6s 2  27 s  12
=
3 (s 3  7 s 2  12 s)

- 2s 2  9 s  4
CsI - A 
1
B =
s 3  7 s 2  12 s

- 2s 2  9s  4
 CsI - A 1 B  D = 1
s 3  7 s 2  12 s

=
- 2s 2

 9 s  4  (s 3  7 s 2  12s)
s 3  7 s 2  12 s

s 3  5s 2  3s  4
=
s 3  7 s 2  12 s

11.9 Eigen values and Eigen vectors

Eigen values of the system matrix are important parameters of the system
matrix. They are nothing but the poles of the transfer function. You may recall that if
we know the poles of the transfer function, we can say whether the system is stable or
not, we can find the nature of response. Similarly if we know the eigen values, we
can determine the dynamic behavior of the system. For example, if an eigen value of
the system matrix is having negative real part, the system is stable.
We have seen in the last sections, for a given system there can be different
state space models. However, all these state space models will have one characteristic
in common. They all have the same eigen values. Or in other words, eigen values of
state space models are unique.
Let us briefly review the definitions of eigen values and eigen vectors. To
make explanation clear, let us consider a three dimensional state space. (One
advantage of three dimensional state spaces is that we can visualize).
Consider the matrix
0 0 0 
 0 
A = 0 -1
0 0 - 2
If we take the product of A and an arbitrary vector x, the magnitude and
direction of the vector could change.
Control Systems
35
 Ch11_may2008 Imthias Ahamed

5 
For example, consider vector p = 2 given in Fig. 11.15
0 
0 
The resultant vector AP = - 2 is in a different direction
0 
In general, if we multiply a matrix and vector, the magnitude and direction of
vector will change.

With this background of multiplication of a matrix and vector let us define

eigen vector. Eigen vectors of a matrix are those vectors which when multiplied by
the matrix results only in scale change (i.e. the magnitude of the vector will change
but the direction will remain the same).
i.e. if V is an eigen vector of matrix A,
then AV = V
 V – AV = 0
 [ I – A] V = 0
This is a set of simultaneous equations which will have a nontrivial solution if
  I - A = 0

X3

X2

X1

Fig 11.9.1 :multiplication of a matrix

and a vector

Figure 11.16 Multiplication of a Matrix and a Vector

The above equation is known as the characteristic equation of the A matrix,

and solution of the equation is known as the eigen values of A- matrix. If the eigen
values of A matrix are distinct, corresponding to each eigen value, there is an
independent eigen vector. The eigen vector associated with eigen value i can be
obtained by solving the equation.
AVi =  i Vi
State Space Approach
36
 Imthias Ahamed

It may be noted that if Vi is an eigen vector KVi is also an eigen vector. (AVi = i Vi
 A (KVi) = i (KVi)). Hence the above equation has infinite solutions. However
we can get a solution by fixing one of the elements of the eigen vector.

Summary
For a given matrix A, the solution of   I - A = 0 are the eigen values and
solution of AVi = i Vi gives an eigen vector Vi corresponding to the eigen value i.

Example 11.17
Find the eigen values and eigen vectors corresponding to the matrix

0 1 0 
A= 0 0 1 

0 - 12 - 7

Solution :
Find eigen values:

 0 0 0 1 0 
I - A = 0  0 - 0
 0 1 
0 0   0 - 12 - 7
 -1 0 
= 0  - 1 
0 12   7

 I- A =     7   12
=  3 + 72 +12

Characteristic equation is
3 + 72 +12 = 0
Eigen values are solution of the characteristic equation:
3 + 72 +12 =  ( +3) ( + 4) = 0
eigen values are
1 = 0, 2 = -3, 3 = -4
To obtain the eigen vector corresponding to 1 = 0, solve the equation.
. AV1 = 1V1
V11 
Let V1 = V 
 21 
V31 

Control Systems
37
Ch11_may2008 Imthias Ahamed

0 1 0  V11  V11 
 0 0 1  V  = 0 V 
 21   21 
0 - 12 - 7 V31  V31 
 V21 = 0
V31 = 0
0. V11 - 12 V21 – 7 V31 = 0  0. V11 = 0
The third equation implies V11 could be any value, we choose V11 = 1
Therefore an eigen vector corresponding to
1 = 0 is
1 
V1 = 0 
 
0
To find eigen vector corresponding to 2 = -3, solve
AV2 =  2 V2
0 1 0 V12  V12 
 0  0 
1    
V22  = -3 V22 
0 - 12 - 7 V32  V32 
V22 = -3 V12 ………. (i)
V32 = -3 V22 ………. (ii)
-12 V22 - 7 V32 = -3V32
 -12 V22 = 4 V32
 V32 = -3 V22 ………..(iii)
Equation (ii) and (iii) are the same. Thus we have three variables and two equations.
Therefore, the system of equation has many solutions. To get a particular solution fix
any one variable, say V12.
Fix V12 = 1
V22 = -3 V12 = -3
V32 = -3 V22 = 9
1
 
 V2 = - 3
9 
To obtain the eigen vector corresponding to 3= -4 solve the equation.
AV3 = 3V3
0 1 0  V13  V13 
 
0 0 1  V23  = -4 V23 
  
0 - 12 - 7 V33  V33 
 V23 = -4 V13
V33 = -4 V23
-12 V23 – 7 V33 = -4 V33  V33 = -4 V23
Fix V13 = 1
V23 = -4V13 = -4
State Space Approach
38
 Imthias Ahamed

V33 = -4V23 = -4 x -4 = 16
Therefore, the eigen vector corresponding to 3 = -4 is
1 
- 4 
 
16 

Example 11.18
A second order system is represented by the following differential equation
y  5 y  6 y  u
(a) Obtain a state space model with
x1  y; x 2  y
(b) Obtains a state space model with
x 2  2y x 2  y
(c) The model obtained in (a) and (b) are for the same system. What is
common in
the two models.
Solution :
(a) Given
y  5 y  6 y  u (i)
x1  y
x 2  y
 x 1  y  x 2 (ii)
x 2  y  6 y  5y  u from (i)

 x 1   0 1   x1  0
 x    6 5  x   1  u
 2   2  
x 
y  1 0 1 
x 2 
(b) State variable are defined as follows
x1  2 y (iv)
x 2  y (v)
 x 1  2 y  2x 2 (vi)
x 2  y  6y  5y  u (from (i))
Substituting for y and y from (iv) and (v)

Control Systems
39
Ch11_may2008 Imthias Ahamed

x 
 6 1   5x 2  u
2
 3x1  5x 2  u (viii)
From (vi) and (vii), we get
 x 1   0 2   x1  0
 x    3 5  x   1  u
 2   2  
 x1 
y  0.5 0 
x 2 
(c) The eigen values of system matrix of the models obtained in (a) and (b)
are
same. The eigen values are -2 and -3.

11.10 Diagonalisation
We have seen that we can obtain state space model in physical variable
form, phase variable form and canonical form. For mathematical analysis canonical
form or diagonal form has many advantages. Hence, in this section we learn, how to
obtain a state space model in the canonical form, from other forms.
Consider a system with state space model:
X  AX  BU
Y = CX + DU
We can obtain another model by linear transformation.
X = PZ
Where P is a non singular matrix
X  PZ  X   PZ
Substituting for X and X in the original state space model, we get
P Z = APZ+ BU ………………… (i)
Y = CPZ + DU………………… (ii)
Pre multiplying the equation (i) by P-1 we get
~ ~
Z  AZ + B U
~ ~
Y = C Z + DU
~
Where A = P-1AP
~
B = P-1 B
~
C = CP
~
D =D
This alternate model with new set of state variables can be viewed as
dynamics of the system described in another set of axis. For systems having distinct
eigen values, if we choose the transformation matrix as the eigen vector matrix or
model matrix, P-1AP will be diagonal. Eigen vector matrix is the matrix formed by the

State Space Approach

40
 Imthias Ahamed

eigen vectors of the A matrix. That is, if V1, V2 and V3 are the eigen vectors of a
third order matrix, the eigen vector matrix M = [V1 V2 V3]
If A matrix is in the phase variable form, finding Modal matrix is straight
forward. In this case model matrix is known as Vander- Monde matrix. If the eigen
values are 1, 2,………. n, the Vander- Monde matrix is given by.
1 1................. 1 
 2 n 
 1
V = 12 22 2n 
 
   
n -1 n2 -1 nn -1 
 1
Example 11.19
State space model of a system is
0 1 0  0 
  1  X + 0 u

X = 0 0
- 6 - 11 - 6 1 
Y = 1 0 0 x

By suitable linear transformation obtain a state space model in the

diagonal form.

Solution :
0 1 0
Given A = 0 0 1

- 6 - 11 - 6
Step 1: Find eigen values
 0 0 0 1 0
λI - A = 0  0 - 0
 0 1

0 0   - 6 - 11 - 6
 -1 0 
  - 1 
= 0
6 11   6
| λI - A | = λ (  6)  11  1  6
= 3  62  11   6
To find the eigen values, find the roots of the characteristic equation
3  62  11   6  0
The roots of the above equation (eigen values) are
1 = -1, 2 = -2, 3 = -3
Step 2: Find the eigen vector matrix.
We know that, if A matrix is in the phase variable form. Eigen vector matrix M is

Control Systems
41
Ch11_may2008 Imthias Ahamed

1 1 1 
 
1 2 3 
 2 22 32 
 1
 1 1 1
 - 3
M = - 1 -2
 1 4 9 
Step 3 State space model in the diagonal form is given by
~ ~
Z  A Z  B u
~ ~
y C Z  D u
~
Where A  M -1 A M
~
B  M -1 B
~
C C M
~
D D
adj of M
Find M-1 =
det(m)
T
- 6 6 - 2
Adjoint (M) = - 5 8 - 3 

- 1 2 - 1 
- 6 -5 -1 
=  6 8 2

- 2 -3 - 1 
det (M) = 1  -6 -1-6 + 1-2
= -2
 3 5 1 
 2 2
 M-1 =  -3 -4 -1 
 
 1 3 1 
2 2 
1 0 0
Check M-1 M = 0 1 0

 0 0 1
3 5 1  0 1 0
 2 2  1 1 1
~ 
A  M -1 A M =  -3 -4 -1  0 0 1 -1
 -2 -3
 
 1 3 1  - 6 - 11 - 6 1 4 9 
2 2
 1 0 0
=  0 -2 0

 0 0 - 3 

State Space Approach

42
 Imthias Ahamed

3 5 1  0 1 
 2 2     2
~  
B  M -1 B =  -3 -4 -1  0 = - 1 
   
 1 3 1  1   1 2 
2 2
1 1 1
~ 
C  CM = 1 0 0- 1 -2 - 3  [1 1 1]
1 4 9
~
D D 0
 state space model in the canonical forms is given by
1 
- 1 0 0
 2
.

Z  0 -2 
0  Z   1  u
 
0 0 - 3  1 2 
Y = [1 1 1] z

11.11 State Diagram

State diagrams are block diagram representation of transfer functions
which can be used to obtain the state space model. State diagram can also be used to
simulate transfer function using analog computer. It is also possible to represent a
given state space model using state diagram. Thus state diagram can be viewed as a
meeting point of
(i) Transfer function ,
(ii) State space model, and
(iii)Analog computer simulation.
Let us first learn, how state diagrams can be used to obtain state space equation from
transfer function.
Example 11.20
Consider a system with transfer function
Y(s) 1
 2
U(s) s  a1S  a 2
Draw a state diagram and hence obtain a state space model
Solution :
Given
Y(s) 1
 2
U(s) s  a1s  a 2
 (s 2  a 1s  a 2 )Y(s)  U(s)
 s 2 Y(s)  a1sY(s)  a 2 Y(s)  U(s) (i)

It may be noted that if we have s2Y(s) we can obtain sY(s) by multiplying

s2Y(s) be 1/S. In a similar way we can obtain Y(s) from sY(s). Now, if we have
sY(s), Y(s) and U(s) we can obtain s2Y(s) from (i). Using this fact, equation (i) is
represented by the state diagram given in figure 11.17.

Control Systems
43
Ch11_may2008 Imthias Ahamed

U(s) + 2
s Y(s) 1 sY(s) 1 Y(s)
S s s
+ +

-a1 -a2

Figure 11.17: State diagram in s-domain

We know that
d 
L  f (t)   sF(s)  f (0)
 dt 
In the case of transfer function f(0) = 0
d 
 L  f (t)   sF(s)
 dt 
d
 L1 sF(s)  f (t)
dt
Similarly,
F(s)

L ò f (t)dt  s
 F(s) 
 L1    ò f ( t )dt
 s 
Therefore, in the case of transfer function multiplying by “s” in S domain is
1
equivalent to differentiation in time domain, and multiplying by " " is equivalent to
s
integration in time domain.

u y y y
S x2 ò x2 = x1
ò x1

-a2 -a1

Figure 11.18. State diagram in time domain

State Space Approach

44
 Imthias Ahamed

1
Therefore, the block with gain is known as the integrator block. Moreover, we can
s
1
obtain the state diagram in time domain by replacing the block with integrator, and
s

replacing Y(s) and sY(s) by y and y respectively. The state diagram in time domain is
given in figure 11.18. Define the outputs of integrator blocks as state variables as
shown in figure 11.18.
From the diagram, we can see that
x 1  x 2 (i)
x 2  a1x1  a 2 x 2  u (ii)
Y = x1 (iii)

From (i), (ii) and (iii), we get the state space model
 x 1   0 1   x 1  0
x    a  a   x   1 u
 2  1 2  2   
x 
y  [1 0]  1 
x 2 
Example 11.21
Draw the state diagram corresponding to the equation
x 1  x1  u
Solution

u x1 x1
S ò

Figure 11.19 : Example 11.21

Example 11.22
Draw the state diagram corresponding to the state space model.
 x 1   2 0 0   x1  1
 x   0  4 0   x   1 u
 2   2   
 x 3   0 0  1  x 3  1
Solution:
From the state space model,
x 1  2x1  u
x 2  4x 2  u
x 3  x 3  u
y  7 x1  2x 2  3x 3
The state diagram is given in figure 11.20.

Control Systems
45
Ch11_may2008 Imthias Ahamed

x1 x1
+ ò 7

-2

u x2 x2 y
+ ò 2 S
-4

x3 x3
+ ò 3

-1

Figure 11.20 : Example 11.22

1
Note: Since it is very clear that the bock " " in s domain state diagram is
s
1
equivalent to an integrator in time domain, we call the block " " as integrator.
s
Moreover, on the s-domain state diagram, we will mark time domain signals also.
1
For example, if the input to the " " block is sX(s) we will also denote it as x as
s
shown in figure 11.21.

sX 1(s) 1 X 1(s)
x1 s x1

Figure 11.21

Example 11.23
Draw a state diagram and hence obtain a state space model of a system with transfer
function.
Y(s) s2
 2
U(s) s  3s  4
Solution :
X(s) Y(s) 1
 2 (s  2)
U(s) X(s) s  3s  4
X (s) 1
  2 (1)
U (s) s  3s  4

State Space Approach

46
 Imthias Ahamed

Y(s)
s2 (2)
X(s)
From (1)
s2 X(s)  3sX(s)  4X(s)  U(s)
s 2 X(s)  4X(s)  3sX(s)  U(s)
State diagram to obtain X(s) from U(s) is given in figure 11.22

U(s) s2 X(s) sX(s) = X 2(s) X(s) = X 1(s)

1 1
S x2 s s x1
x2 = x1

-3 -4

Figure 11.22 : State diagram representing (1)

Define outputs of the integrators as X1(s) and X2 (s)

From (2), Y(s) = sX(s) + 2X(s).
Now we will add additional paths in figure 11.32 to obtain Y(s). The complete state
diagram is shown in figure 11.23.

U(s) X(s) Y(s)

s2 X(s) 1 sX(s) 1
s s
2 S
U(t) x2 x2 x1

-3 -4

Figure 11.23 : State diagram

From figure 11.23

x 1  x 2
x 2  4 x1  3x 2  u
y  2 x1  x 2
 The state space model is
Control Systems
47
 Ch11_may2008 Imthias Ahamed

 x 1   0 1   x1  0
 x    4  3  x   1 u
 2   2   
x 
y  [2 1]  1 
x 2 

11.11. Solution of state Equation:

We know that state space model of a system is given by the equations
  AX  BU
X
Y  CX  DU
The first equation is known as the state equation and the second equation is the
output equation. It may be noted that the first equation is a set of I-order differential
equations and the second is an algebraic equation in the case of a single output
system. So if we can solve the state equation for a given input and initial condition,
we can find the output easily from the output equation. In this section, we learn how
to solve the state equation.
Solution of state equation depends on the initial condition and the input. First,
we will consider the case when input is not there or the unforced system, such systems
are called autonomous systems; then, we will learn how to solve the state equation in
the presence of input.
11.11.1. Solution of Autonomous System
Let us first clearly understand what is an autonomous system or an unforced
system. It was mentioned earlier that an unforced system is a system without any
input. What is the meaning of a system without input? It means a system with input
zero, or mathematically a system with input function u(t) identically zero.
That is a system with
u(t)  0 t
Why do we want to analyse such systems?
It is a first step to learn how to solve state equation with forcing function. If
input is zero, why does the state of the system change? Initially if the system is not in
equilibrium state, the state will change. For example, consider the RLC network
shown below.

R L

ei(t) vc(t) C e0(t)

Figure11.24: RLC Network

State Space Approach

48
 Imthias Ahamed

The state of the system (say capacitor voltage and inductor current) will
change if capacitor is initially charged to some value even through u(t) = 03. You
may recall capacitor voltage will decay
Consider an autonomous system
X  AX; given X(0)
To solve the above equation, take Laplace transform
L (X )  L[AX]
s X(s) - X(0)  AX(s)
s X(s) - AX(s)  X(0)
s I X(s) - AX(s)  X(0) 4
[s I - A] X(s)  X(0)
Pre – multiplying both sides by
sI - A 1 , we get
X(s) = [sI – A]-1 X(0)

 X(t)  L-1 sI - A  X(0)
1

X(0) is a constant vector so we can take it outside the bracket of L-1.
 
i.e. X(t)  L-1 sI - A 1 X(0)
-1

L sI - A 
1

is known as state transition matrix (S.T.M)
It is denoted by (t). Laplace transform of S.T.M is denoted by (s).

Summary

(s) = [sI –A]-1

S.T.M,  (t)  L-1  (s)


= L-1 sI  A 1 
X(t) = (t) X(0)

Example 11.24:
Solve
  0
X
1
- 2 - 3 X
 
1 
X(0)   
- 2 
Solution:

3
u(t) =0 means the voltage source is short circuited.
4
We want to find X(s). To take X(s) as common factor and to facilitate subtraction we post multiply
X(s) by I, the identity matrix.
Control Systems
49
Ch11_may2008 Imthias Ahamed

0 1 1 
Given A   X(0)   
- 2 - 3 - 2 

Step 1 Find sI - A -1

1 0  0 1 
sI - A  s  
0 1 - 2 - 3
s -1 

2 s  3
s  3 1
Adjoint of [sI-A] =  
- 2 s 
determinant [sI –A],   s 2  3s  2  (s  1) (s  2)

adjoint sI - A 
sI - A 
1

det sI - A 

s  3 1
- 2 s 
= 

s  3 1
  
=  
- 2 s
   

Step 2: Find S.T.M, (t)

s  3 1 
 
 (t)  L-1   
 2 s
   
-1 -1
Note: L of a matrix is L of individual elements of the matrix
 -1  s  3  1
L    L-1   
   
 (t)  
 -1   2   s 
L   L-1  
      
11 12 
=  ---------- (i)
 21  22 
 s  3
11  L-1  
  
 s3 
= L1  
 (s  1)(s  2) 

State Space Approach

50
 Imthias Ahamed

 2 1 
= L1   
s  1 s  2 
= 2e-t + -1*e-2t
= 2e-t - e-2t
 1 
12 = L1  
 s  1s  2  
 1 1 
= L1   
s  1 s  2 
= e-t - e-2t
 2 
 21 = L1  
 s  1s  2  
 -2 2 
= L1   
s  1 s  2 
= -2e-t + 2e-2t
 s 
 22 = L1  
 s  1s  2  
 -1 2 
= L1   
s  1 s  2 
= -e-t + 2e-2t
Substituting for 11, 12, 21 and 22 in (i) we get
 2e -t - e -2t e -t - e -2t 
 (t)   -t - 2t 
- 2e  2e - e - t  2e - 2t 
Step 3 : Find X(t) = (t) X(0)
2e -t - e -2t e -t - e -2t  1 
X(t)   -t - 2t  
- 2e  2e - e - t  2e - 2t   2 
(2e -t - e -2t )  - 2( e -t - e -2t ) 
=  -t - 2t -t - 2t 
(-2e  2e )  - 2(-e  2e ) 
2 t
 x 1 (t)  e 
X(t)      2t 
 x 2 (t)   2e 
i.e.x1(t) = e-2t
x2(t) = -2e-2t
The plot of x1(t) Vs t and x2(t) vs. t is given in figure 11.35.

Control Systems
51
Ch11_may2008 Imthias Ahamed

Figure 11.25

From the figure, it is clear that at t=0 x1(t) = 1 and x2(t)=-2, and as t, x1(t)
and x2(t) approaches zero

Example 11.25
0 1 0 0 

Given X = 0 0 1  X; X(0) 0 

- 6 - 11 - 6 - 1
Find X(t)

Solution:
0 1 0
Given A = 0 0 1
- 6 - 11 - 6
s 0 0 0 1 0 
[sI -A] = 0 s 0 - 0 0 1 
0 0 s  - 6 - 11 - 6

State Space Approach

52
 Imthias Ahamed

s -1 0 
= 0 s - 1 

6 11 s  6
T
s 2  6s  11 -6 - 6s
 
Adj [sI-A] = s  6 s(s  6) - (11s  6)
1 s s2 
 
s 2  6s  11 s6 1
 
=  6 s 2  6s s 
- 6s - (11s  6) s 2 

sI - A   = s 3  6s 2  11s  6
= (s  1)(s  2)(s  3)
adj [sI - A]
sI - A 1 =
sI - A
s 2  6s  11 s6 1
 2 
 6 s  6s s
- 6s - (11s  6) s 2 
= 

2
 s  6s  11 s6 1
 
   
2
 6 s  6s s
=
  
 
 - 6s - 11s  6 s2 
    


Note : For this problem, it is easier to multiply [sI-A]-1 and X(0)and then find L-1.
 s2  6 s  11 s6 1
0 
     
 
-1  -6 s 2  6s s   
[sI-A] X(0) =  0 
     
 
 - 6s - 11s - 6 s2   
- 1
      
-1 
 
 
 -s 
=
 
 2
- s 
  

X(t)  L-1 [sI - A]-1 X(0) 
Control Systems
53
Ch11_may2008 Imthias Ahamed

 
 - 1  L-1  - 1  
      
  
1  - s  -s 
= L  L-1   
      
 2  2 
 - s   -1  - s 
      L 
  
 -1 1
L-1   = L1
 s  1s  2s  3
1 1 1
= 2   2
s 1 s2 s3
= - 1 t
e  e -2t
- 1 e 3t
2 2
-s s
L-1   = L1
 s  1s  2s  3
 1 3 
 2 2 
= L1  2   
 s  1 s  2 s  3

 1 t 3
= e - 2 e - 2t  e 3t
2 2
 1 9 
 s2 1  2 4 2
L1 = L    
s  1s  2s  3  s  1 s2 s  3

1 t - 9
=  e  4 e -2t e 3t
2 2
 - 1 -t - 2t 1 
2 e e - e 3t 
2
 
 - 1 t -2t 3 3t 
 X(t) = e - 2e  e
2 2 
-1 
 e  t  4e 2t - 9 e 3t 
 2 2 

11.11.2 Solution of Non-Homogenous system

Let us now consider the state equation of non-homogenous system
x  Ax  Bu
Taking Laplace transform on both sides we get
sX (s)  X (0)  AX (s)  BU(s)
sX(s)  AX(s)  X(0)  BU(s)
[sI  A]X(s)  X(0)  BU(s)

X(s)  [SI  A]1X(0)  BU(s)

State Space Approach
54
 Imthias Ahamed

X(s)  [SI  A]1X(0)  BU(s) (i)

   
x(t)  L1 [sI  A]1 X(0)  L1 [sI  A]1 BU(s) (ii)

The first term corresponds to the initial condition and the second term is due to
the forcing function. Hence, first term is known as homogenous solution and the
second term is know as the forced solution. To find the complete solution, we may
use (i) also.
Steps for finding solution of state equation :
Step (i) Find [SI-A]
Step (ii) Find Resolvent matrix (s)  [SI  A]1
Step (iii) Find U(s) = L [u(t)]
Step (iv) Find BU(s)
Step (v) Find X(0) + BU(s)
Step (vi) Find X (s )   (s )[ x (0)  BU (s)

Step (vii) Find X(t )  L1[X(s)]

Example 11.26
Obtain the solution of the state equation given below
 1 1 0 1
X   0  1 0  X  1 u
  
 0 0  2 1

1
x (0)  0
1

u (t) = 0 t < 0
=1 t0
Solution:
We know

x( t )  L1 [sI  A]1{X(0)  BU(s)} 
Given
 1 1 0 
A   0 1 0 
 0 0 2 

1 1 
B  1 ; X(0)  0 
1 1 
Control Systems
55
Ch11_may2008 Imthias Ahamed

1
U(s) 
s
Step (i)
Find [sI  A]1

s  1 1 0 
[sI  A]   0 s  1 0 

 0 0 s  2

sI  A    (s  1) 2 (s  2)
T
 (s  1)(s  2) 0 0 

adj ([sI  A])   (s  2) (s  1)(s  2) 0 
 0 0 (s  1) 2 

(s  1)(s  2) (s  2 0 

 0 (s  1)(s  2) 0 
 0 0 (s  1) 2 

 1 1 
 s  1 (s  1) 2 0 
 
1 adj([sI  A])  1 
(sI  A)   0 0 
sI  A (s  1)
 
 1 
 0 0
s  2 

Step (ii) Find X(0) + BU (s)
 s  1
 
 1  1/ s   s 
1 
X(0)  BU(s)   0   1/ s   
 s 
 1  1/ s   
 s  1
 s 
Step (iii) Find [sI  A]1{x(0)  BU(s)}

 1 1 
 s  1 (s  1)2 0   s  1
 
  s 
 1  1 
 0 0 
s 1  s 
  
 0 1   s  1
0
 s  2   s 


State Space Approach

56
 Imthias Ahamed

1 1 
 s  s(s  1) 2 
 
 1 

 s(s  1) 
 s 1 
 
 s(s  2) 

Step (iv) Find X(t)  L1 [sI  A]1[X(0)  BU(s)]

1 1 
 s  s(s  1)2 
 
1  1 
X(t)  L  
 s(s  1) 
 s 1 
 
 s(s  2) 
2 1 1 
 s  (s  1) 2  s  1 
 
1  1 1 
L   
s s 1
 
 1  1 
 2s 2(s  2) 
 
 t t 
2  te  e 
 L1   1  e  t 
 1 1 2 t 
  e 
 2 2 
11.11.3 Response of a system using state space Approach
If the state space model and initial conditions are given, we can find the
response of the system for any input.
Consider a system with state space model,
 (t )  AX( t )  Bu(t )
X
Y ( t )  CY ( t )  Du ( t )
We know that solution of the state equation is


X(t)  L1 [sI  A]1[X(0)  BU(s)] 
We can find the response of the system from output equation.
y ( t )  CX ( t )  Du ( t )
The steps for finding the response is illustrated in example 11.27.

Example 11.27
Control Systems
57
Ch11_may2008 Imthias Ahamed

Obtain the ramp response of a system represented by the following equations.

x 1   x1  x 2  u
x 2   x 2  u
x 3  2x 3  u
y  2 x1  x 3
x1 (0)  x3 (0)  1; x 2 (0)  0
Solution
First let us write the given equation in the standard matrix form
 x 1   1 1 0   x1  1
 x    0 1 0   x   1 u
 2   2  
 x 3   0 0 2  x 3  1

 x1 
y  [2 0 1]   x 2 
 x 3 

1
x (0)  0
1

We have to find ramp response

 u(t )  t
1
 U(s) 
s2
Step 1: Find [sI-A]-1
s  1  1 0 
[sI  A]   0 s  1 0 

 0 0 s  2

(s  1)(s  2) (s  2) 0 

adj([sI  A])   0 (s  1)(s  2) 0 
 0 0 (s  1) 2 

State Space Approach

58
 Imthias Ahamed

det([sI  AI])  (s  1)2 (s  2)

 1 1 
 s  1 (s  1) 2 0 
 
1  1 
 [sI  A]   0 0 
s 1
 
 0 1 
0
 s  2 

Step (ii)
 s2  1
 2 
1  1/ s   s 
2

   1 
X(0)  BU(s)   0   1/ s 2    2 
s
1  1/ s 2   2 
 s  1
 s 2 

Step (iii) 
X(t)  L1 [sI  A]1[X(0)  BU(s)] 
 1 1  2
S 1 0   s  1 
(S  1) 2 2
  s 
1  1  1 
L  0 0  2 
S 1 s
  2 
 0 1   s  1
0
 S  2   s 2 

 s2  1 1 
 2  2 2
 s (s  1) s (s  1) 
1
 L1  2


s (s  1)
 2 
 s 1 
 s 2 (s  2) 

2 3 4 1 
 s 2  s  s  1  (s  1)2 
 
1  1 1 1 
L  2
  
s s s 1
 
 1  1  5 
 2s 2 4s 4(s  2) 
 

Control Systems
59
Ch11_may2008 Imthias Ahamed

 
 2t  3  4e  t  te  t 
 
  1  t  e  t 
 1 1 5 2t 
  t e 
 4 2 4 
Output y = [ 2 0 1] X
25 9 5
  t  8e t  e 2t  2te t
4 2 4
Example 11.28
(a) Obtain a state space model of the system given above
R=10K

+
u(t) _ y(t)=VC(t)
C=0.1mF

Figure 11.26 : Example 11.28

(b) If input u(t) = 0 and VC(0) = 2V, obtain an expression for y(t)

Solution :
Since this is a first order system,
Number of state variables = 1
Choose capacitor voltage as the state variable
x1  VC ( t )
Using KVL,
u (t )  Ri(t )  VC (t )
u ( t )  VC ( t )
i( t ) 
R
dV
C i
dt
u ( t )  VC ( t )

R
1 1
 Cx 1   x1  u
R R
 1   1 
x 1    x1   u
 RC   RC 

State Space Approach

60
 Imthias Ahamed

x 1  103 x1  103 u (i)

From figure,
y  VC (t )  x1
y  [1]x1 (ii)
Equation (i) and (ii) gives the state space model of the system. Note that A matrix is
of order 1x1.
A  [103 ]
Also note that
B  [103 ] and c  [1]
(b) Since u(t) = 0, the given system is an autonomous system.
 x(t)  L-1[sI  A]1 x(0)
 L1[s  103 ]1 x(0)
 1 
 L1  3
2
 s  10 
3
 2e10 t
3
y( t )  c x ( t )  x ( t )  2e10 t
Example 11.29
Solve the following differential equation
x  4x  3x  0
Given x (0)  3, x (0)  0
Solution :
We will represent the differential equation as state equation and solve the state
equation.
Define
x1  x
x 2  x
x 1  x 2 (i)
x 2  x  3x  4x
 3x1  4x 2 (ii)
From (i) and (ii), we get
 x 1   0 1   x1 
x    3  4 x 
 2   2 
Control Systems
61
Ch11_may2008 Imthias Ahamed

We know


X(t)  L1 [sI  A]1 X(0) 
1
 s 1 
1
[sI  A]   
3 s  4 
s  4 1
 
 
3 s
 
  
Where   s 2  4s  3
Given
x(0)  x1 (0)  3  3
  X(0)   

x(0)  x 2 (0)  0  0
 s  4 1 
     3  
 X(t)  L1    0 
  3 s  
     

 1 s  4 
 x1 ( t )  L 3  
x ( t )   1  9 
 2   L 
  
By definition
x ( t )  x1 ( t )
Therefore the solution of given differential equation,
 3(s  4) 
x(t)  L1  2 
 s  4s  3 
 A B 
 L1   
 (s  1) s  3 
9
A
2
3
B
2
 9 / 2 3 / 2 
 x(t)  L-1   
s 1 s  3 
9  t 3 3 t
 e  e
2 2

State Space Approach

62
 Imthias Ahamed

Exercise

11.1 A third order system is represented by the following differential equation.

y  3 y  5 y  7 y  2u
Obtain its state space model.

 x 1   0 1 0   x 1  0 
  
Ans: x 2    0 0 1   x 2   0 u
 x 3   7  5  3  x 3  2

 x1 
y  1 0 0x 2 
 x 3 

11.2 Obtain two different state space models for an armature controlled d.c. motor.
Ans: Model 1:
x1  ia; x 2  ; x 3  

 Ra Kb   1 
 1 
 x 1   La La   x 1   La 
 x   0     
 2  0 1  x 2    0  u
  B   x 3   0 
 x 3   Kt 0   
 J J   
 x1 
y  0 1 0 x 2 
 x 3 

Model 2 :
x1  , x 2   , x 3  

   
 x 1  0 1 0   x1   0 
 x   0 0 1
   
 2   x 2    0 
  BR  K bK t  R B  K
 x 3 0   a
   a     x 3  
t

  JL a L J 
 JL 
    a  a

 x1 
y  1 0 0 x 2 
 x 3 

Control Systems
63
Ch11_may2008 Imthias Ahamed

11.3 Obtain the state space model of the system given below. Given input =ei and
output = e0.
R1 R2

ei
+ e0
_ C R3

(Hint : Choose x1 = voltage across capacitor )

 (R  R 2  R 3 )  1
Ans : x 1   1  x1  u
 CR 1CR 2  R 3  CR 1
R3
y x1
R 2  R3
11.4 Obtain state space model of the system given below

R1 L1 R2 L2 R3

e1 e2
+ +
C1 C2

Given e1 and e2 are inputs and voltage across resistors are the output
(Hint choose x1=i1, x2 = i2, x3 = Vc; x4 = Vc2)
 R1 1 
 L 0 0 
L1
 1  1 0 
 R2 1 1  L
 0  0 
L2 L2 L2   1  u1 
Ans x 1   x 0 0   
 1 1    u 2 
 0 0  0 1
 
 C1 C1   0 C2 R 3 
 1 1 
 0 0 
 C2 C2 R 3 

State Space Approach

64
 Imthias Ahamed

x 
 y1  R 1 0 0 0  1  0 0 
y   0 x u 
 2 1  R2 0 0   2   0 0   1 
x  u
 y3   0 0 0 1  3  0  1  2 
 x4 

11.5 Consider the following system with ei(t) as input, and e0(t) as output. Obtain a
state space model with q and i as state variable.

R L1

e i(t) + C e 0(t)

 0 1  0
Ans : x    1  R X   1 u
 LC L   L 

11.6 Obtain a state space model of the system given in figure below (Hint : let
displacement of M1, be y1 and that of M2 be y2

K1 B1
F
M1 K2

B2

M2

x1  y1; x 2  y 1; x3  y2 ; x 4  y 2

 0 1 0 0   0 
 K2 B B2   0 
 M  2 0 
M2 M2  
Ans : x   2
X   0 u
 0 0 0 1 
B2 K  (B1  B 2 )   1 
 0  1 M 
 M1 M1 M1   1

y  1 0 0 0x
Control Systems
65
Ch11_may2008 Imthias Ahamed

11.7 Obtain a state space model for the system given below.

R1 R2 L3

u(t) c1 c2 c3 y(t)

Ans:
  1 1  1   1 
    0 0  C R 
  R1C1 R 2 C1  C1R 2   1 2
 x 1   1 1 1   x1   
 x    0    0 
 2   R 2 C2 R 2C1 C2   x 2   u

 x 3   1   x3   0 
   0 0 0    
 x 4   C3   x 4   
 1 1   
 0 0   0 
 L3 L3   

 x1 
x 
y  [0 0 1 0] 2 
x3 
 
x 4 

11.7 Consider the system given in figure below. Obtain a state space model with
voltage across the resistors taken as outputs.

R1 L1 R2

c1 c2
u(t)

State Space Approach

66
 Imthias Ahamed

Ans:
 R1 1 
 0  1
L L1   x1   L1 
 x 1   1
 x    1 1 1    
 2  C  x2   0  u
R 2 C1 R 2 C1     
 x 3   1  x 3   0 
 1 1 
 0   
 R 2 C2 R 2 C2 

 x1 
 y1   R1 0 0   
 y    0 1 1  x 2 
 2   x 
 3

11.9 Consider the system given in figure below. Obtain a state space model.

Ans:
 0 1 0 0   0 
 x 1    K1  (B1  B2  f ) B2   x1  
 x   0   0 
 2    M1 M1 M1   x 2  
  u
 x 3   0 0 0 1 x3   0 
   0 B2  B2     1 
 x 4   M2
0 x 
M 2   4   M 2 


 x1 
 
 y1  1 0 0 0  x 2 

 y  0 0 1 0  x 
 2   3
 
x 4 

Control Systems
67
Ch11_may2008 Imthias Ahamed

11.10 Consider the system given below

L1 L2

R1
R2
C1 C2 y(t)
u(t) +

(a) Obtain a state space model with physical variables as state variables
(b) Obtain a state space model in phase variable form taking the output as one of the
state variables

11.11 Transfer function of a system is

Y s  10 s  4 
 3
 s  s  4 s 2  3s
(a) Obtain state space model in the phase variable form.
(b) Obtain state space model in the canonical form.
(c) Determine the eigen values of the system matrix in both cases.

Ans:
0 1 0  0 
.
(a) X  0 0 1  X  0  u
 
0 3 4  1 

y  40 10 0X
 40 
0 0 0   3
.
(b) X   0 1 0  X   15  u
   
 0 0 3  5 / 3 

y  1 1 1X
(c ) Eigen values in both cases are 0, -1, -3

11.12 Obtain state space models in the phase variable form and Jordan canonical
form for the system given below. What are the poles of the given system? What are
the eigen values of system matrices?

State Space Approach

68
 Imthias Ahamed

 s  Y s 
s3  1
2
s s  2 

Ans:
State space model in phase variable form
0 1 0  0 
.
X  0 0 1  X  0  u
 
0 4 4  1 

y  1 4 4 X  [1]U
State space model in Jordan canonical form.
0 0 0  1 
.
X  0 2 1  X  0  u
 
0 0 2 1 

y  1 7 17  X  [1] U
 4 2 4

Eigen values in both cases are 0, -2, -2

11.13 Consider a system with transfer function
Y(s) (s  3)

U(s) (s  4)(s  5)
Obtain a state space model in the diagonal form with B = [1 1]T

11.14 Given
Y(s) s3  5
 3
U(s) s  7s 2  10s
(a) Obtain state space model in canonical form.
(b) Obtain state space model in phase variable form.
(c) Find eigen values of a matrix obtained in (a) and (b)
(d) Find the poles of the transfer function.
Ans :
0 0 0  ½ 
(a) x  0 2 0  X   ½  u; y  [1 1 1]X  [1]u
 
0 0 5  8

Control Systems
69
Ch11_may2008 Imthias Ahamed

0 1 0 0
(b) 
x  0 0 1  X  0 u; y  [5 - 10 - 7] X  [1] u

0  10  7 1

(c) eigen values are 0, -2, -5

(d) poles of the transfer function are 0, -2, -5

11.15 Consider a system with state space model

0 1 0
x    X   u
 2  3 1
y  [1 0]X
(a) Obtain the eigen values of the system matrix.
(b) Obtain the transfer function of the system
(c) Derive a state space model in diagonal form corresponding to the transfer
function in (b)
(d) Obtain the eigen values of the system matrix obtained in (c)
Anss
(a) -1, -2
1
(b) 2
s  3s  2
 1 0  1
(c) x    X    u; y  [1  1]X
 0  2 1
(d) -1, -2

11.16 Consider a system with state space model given below.

0 1 0 0

x  0 0 1  X  0 u

0  10  7 1

y  [5  10  7]X  [1]u
By suitable linear transformation, obtain a state space model in the diagonal
form.

0 0 0 ½ 
Ans : x  0  2 0  X   ½  u; y  [1 1 1]X  [1]u
 
0 0  5  8

State Space Approach

70
 Imthias Ahamed

11.17 State space model of a system is given below.

0 1 0 0

x   0 0 1  X  0 u; y  x 1


 2  5  4 1

Obtain a state space model in Jordan canonical form
 1 1 0 0 
Ans : x   0  1 0  X  1 u; y  [2 - 2 2] X
 
 0 0  2 1

11.18 State space model of a third order system in phase variable form is given
below. Obtain the corresponding transfer function.
0 1 0  0 
 
X  0 0 1   
 X + 0  u
0  12  7  1 

y = 4  9  2 X  1u

Ans:
Y(s) s3  5s 2  3s  4
 3
U(s) s  7s 2  12s

11.19 Consider the system described by the following equations

x 1  5x1  x1  2u
x 2  3x1  x 2  5u
y  x1  2 x 2
x1 (0)  x 2 (0)  0
Y(s)
Obtain
U(s)
12s  59
Ans:
s 2  6s  8

11.20 Consider a system with transfer function

Y (s) 1

U (s) (s  1) 2
Control Systems
71
Ch11_may2008 Imthias Ahamed

Obtain a state space model in Jordan canonical form

 1 1  0
Ans : x    X    u; y  [1 0]X
 0  1 1

0 1  0 
11.21 Given x    X    u; y  [1 0]X
0 0 1 
Y(s)
Obtain
U(s)
1
Ans:
s2

11.22 Given
 2 0 0 
A = 0  1 0 
0 0  3

Find (a) A-1

(b) A4
(c) eigen values of A
(d) eigen vectors of A

Ans :
a)
 1 0 0 
 2 
A 1   0 1 0 
 
 0 0 1 
3
b)
16 0 0 
A   0 1 0 
4

 0 0 81
c)
Eigen values are  2,  1,  3
d)
1  0  0
Eigen vectors are  0 1 
 
0
 
0  0  1 

State Space Approach

72
 Imthias Ahamed

11.23 Given

0 1 0 
A = 0 0 1
0  6  5

Find eigen vectors.

Ans:

1  1 1
Eigen vectors are 0   2
 
 3
 
0   4   9 

2 2 1 
11.24 Given, A  1 3 1
1 2 2

Find eigen values and eigen vectors of A

Ans: eigen values are 1,1,5
 0.9045 0.5774  0.4865
Eigen vectors are  0.3015 0.5774  0.1838
 0.3015 0.5774 20.8541 

11.25 Draw a state diagram and hence obtain a state space model of a system with
10
transfer function
S 2

11.26 Consider a system with transfer function

Y(s) S2  3s  5

U(s) (S  1) 2 (s  2)
Draw a state diagram, and hence obtain state space model in Jordan- canonical
form.

Control Systems
73
Ch11_may2008 Imthias Ahamed

11.27 Obtain a state space model of the system with transfer function
Y(s) s 2  3s  3

U(s) s 2  3s  2
in the diagonal form.
Also draw a state diagram.

11.28 Solve
0 1  1
x    X; X(0)   
  3  2 1

 e t cos( 2t) 

Ans : X(t)   
t t
 2e sin( 2t)  e cos( 2t) 

Y(s) 1
11.29 Given  2
U(s) S  5s  4
(a)Obtain step response using classical approach
(b) Obtain a state space model
(c)Obtain step response using state space approach
Ans :
1 1  t 1 4 t
(a) y( t )   e  e
4 3 12
 4 0   1 / 3
(b) x    X  u; y  [1 1]X
 0  1  1/ 3 
(c)same as (a)
  AX
11.30 Consider a system represented by the state equation X
Given
e t  te  t te  t 
State transition matrix is   t 
 te e  t  te  t 
(a) Find a set of states x1(0) and X2(0) such that x1(2) = 2
(b) Find resolvent matrix
(c) Find the A matrix

11.31 For the system given below, given R=1000 Ω, L=1 H and C = 0.1µF
(a) Obtain the state space model
State Space Approach
74
 Imthias Ahamed

(b) If switch S is closed at t =0 and if Vc (0) = 2 V and iL(0) = 0 A. Obtain

solution of the state equation.
(c) Plot x1 t  vs t and x2 t  vs t
(d) Plot the state trajectory.

R L

S C e0(t)

Ans:

(a) ̇ = −1000 −1
10 0
( . )
−2
.
(b) ( )= ( . )
2 cos(3122.5 ) + 1000
.
(c)

Control Systems
75
Ch11_may2008 Imthias Ahamed

(d)

State Space Approach

76
 Imthias Ahamed

Objective Type Questions

1. The state variable model of a linear autonomous system is
given by
. 0 1 
X  X
1 0 
The roots of the characteristic equation are
(a) 1 and -1 (c) –j1 and –j1
(b) +1 and +1 (d) -j1 and +j1

2. Consider a second order system represented by the equation

.
X  A X  BU
Y  C X  DU
We can obtain an alternate model
.
Z  A1 Z  BU
1

Y  C1 Z  D1U
by following substitution
(a)z1 = 2x1; z2 = 2x2
(b) z1 = 2x1+ 3; z2 = 2x2 + 3
(c) z1 = 2x1+ 3; z2 = 2x2
(d) All the above

0 1
3. A 
 6 5 
- ------- is an eigen vector of A

(a)[1 -2]
(b)[2 -4]
(c)[3 -6]
(d) All the above

0 1
4. A 
 6 5 

---------- is an eigen value of A

(a) 0
(b) -1
(c) -2
(d) All the above

Control Systems
77
Ch11_may2008 Imthias Ahamed

5. A system is represented by the equations

.
X  A X  BU
Y  C X  DU
 2 0 0 
Given A =  0 3 0 
 0 0 4
The poles of the corresponding transfer function are
(a) 2, 3, 4
(b) -2, -3, -4
(c) depends on B and D matrices
(d) depends on C matrix

6. Consider a system is represented by the following equations

.
x1  ax1
.
x 2  bx2  cu
y  x1  x2
The system is stable if
(a) a<0
(b) b<0
(c) c>0
(d) a < 0 & b < 0.

7. Given the homogenous state space equation

.  2 0 
X  X
 0 4 

If X = (0) = [ 1 1]T , the steady state value XSS is

(a) [ 0 0]T
(b) [ 0 1]T
(c) [ 1 0]T
(d) [ 1 1]T

8. Consider a system represented by the following state space model.

. 0 1 0
X   X
 2 3 1

Y = [1 0] x ; x(0) = [ 0 0]T

State Space Approach

78
 Imthias Ahamed

For this system, step response is

(a)oscillatory but stable
(b) oscillatory but unstable
(c) damped
(d) sinusoidal

9. For the system

.
x   2x  u
If x(0) = 0 and
u(t) = 0 t 0
=1 t>0
lim ( ) =
→

(a) 0 (b) 1 (c) 2 (d) ½

10. Consider the RLC circuit given in figure below. Given Vc(0) = 10 V. If input
voltage is zero, ≥ 0 the steady value of the output will be

R L

ei(t) vc(t) C e0(t)

Figure: RLC Network

(a)10 V (b) -10 V (c) 0 V (d) 5V

11. Consider a system represented by the following equations

.
x1   x1  u
.
x2   2 x2  u ;
y  2 x1
x1 (0) = x2(0) = 0.
where x1 and x2 are state variables. If input u, is a unit impulse function, the output y
is
(a)e-t (b) e-2t (c) 2e-t (d) -2e-2t

Control Systems
79
Ch11_may2008 Imthias Ahamed

12. A homogeneous system is represented by

.  2 0 
X  X
 0 1
0
given X (0) = the state of the system at the end of 1 s given by
1
(a) [ 0 e-1 ]T
(b) [ e-1 0]T
( c) [e-2 0]T
(d) [e-2 e-1]T

Short Answer Question

1. What are the advantages of the state space analysis.
2. Define state variable.
3. What do you understand by the term state space.
4. Define eigen values.
5. Show that under linear transformation the eigen values of a linear system are
invariant.
6. Prove that the set of state variables for a system is not unique.
7. Derive a formula to obtain transfer function from state space model.
8. State the properties of state transition matrix.
9. What are the peculiarities of phase variable form.
10. What are the advantages of canonical form.
11. Can we represent any L.T.I system in canonical form.
12. Illustrate any two properties of diagonal matrix.
13. The transfer function of a system is . Obtain a state space model.
14. The transfer function of a system is given by

( ) 5
= .
( ) +2
Determine the differential equation governing the system
15. The transfer function of an integrator is given by
( ) 1
= .
( )
Obtain two different state space model.
16. Can we always transform a state space model to diagonal form
17. What is similarity transformation.
18. How are poles of a system related to its state model.
19. List the properties of state transition matrix
20. Give any two methods for finding STM

State Space Approach

80
 Imthias Ahamed

21. Prove that

∅( − ) ∅( − ) = ∅( − )

22. Solve the state equation of a linear time invariant unforced system.
23. Write an expression for resolvent matrix.

Long Answer Questions

1. Compare and contrast state space approach with classical approach.
2. With the help of illustrations, explain the concept of state, state space and
state trajectory
3. (a) What are the advantages of canonical form
(b)Explain steps involved in obtaining diagonalization matrix

4. Derive, explain and illustrate the properties of state transition matrix.

5. Explain and illustrate how state equation can be solved by Laplace transform
method.
6. (a) Explain the steps involved in transforming a state space model to
canonical form.
(b) Give an example where it is not possible to obtain state space model in
canonical form
(c) Write notes on Jordan canonical form
7. Explain steps involved in transforming a state space model with repeated
eigen values to Jordan – canonical form.

Answers of Objective Questions:

1. (a)
2. (a)
3. (d)
4. ( c)
5. (b)
6. (d)
7. (a)
8. ( c)
9. (d)
10. ( c)
11. ( c)
12. (a)

Control Systems
81