State Diagram of Discrete-Time

System with ZOH

• Before the state diagram for the system can be

constructed, the state diagram for ZOH must be
established.

Let the input and output of ZOH be denoted by e* (t )

and h(t ) , respectively. Then
 h(t )  e(kT ) for kT  t  (k  1)T
Taking the Laplace transform,
e( kT )
H (s)  for kT  t  (k  1)T
s

State diagram of ZOH for kT  t  (k  1)T

Example:

G(s)

State diagram of
G(s)
Using the direct decomposition method, the state
diagram of the controlled process G(s) is obtained.

• The overall state diagram is constructed by

connecting this state diagram to that of ZOH and
using the relationship

e(kT )  r (kT )  c(kT )  r (kT )  x1 (kT )

By letting t 0  kT ,

h(kT )  h(t )  e(kT ) for kT  t  (k  1)T
The complete state diagram is shown below.
Using X 1 ( s ) and X 2 ( s ) as outputs, and x1 (kT ) , x2 (kT ),
and r (kT ) as inputs:

For kT  t  (k  1)T
 1 1 1   1 
  2
 X 1 ( s)  s s ( s  1) s( s  1)   x1 (kT )   s 2 ( s  1) 
 X ( s )       r (kT )
 2    1 1   x2 (kT )  1 
 s ( s  1) s  1   s( s  1) 
• Taking the inverse Laplace transform of both sides,
for kT  t  (k  1)T

 x1 (t )  2  (t  kT )   (t kT ) 1   (t kT )   x1 (kT ) 

 x (t )   ( t kT ) ( t  kT )   
 2   1    2x ( kT ) 
(t  kT )  1   (t kT ) 
 ( t  kT )  r (kT )
 1  
• For responses only at the sampling instants, use
t = (k+1)T.

Thus,

 x1 (k  1)  2  T   T 1   T   x1 (k )  T  1   T 
 x (k  1)   T T     T 
r (kT )
 2   1      x2 ( k )   1   
 Controllability and Observability
The linear multivariable discrete-time system is
described as: x(k  1)  Ax(k )  Bu(k )
c(k )  Dx(k )  Eu(k )
• In control problems, two basic questions need to be
answered in deciding whether or not a control
solution exists. These questions are posed as:
1) Is it possible to transfer any state x(0)  x 0 to any
1
other state x in finite number of sampling periods N,
by proper choice of the input u(k ); k  [0, N  1] ?
0
2) Is it possible to identify any state x at k = 0 by
observing the output c(k ); k  [0, N  1] and input
u(k ); k [0, N  1] over a finite number of sampling
periods N?
• The answers to these questions are conceptualized
by Kalman into what is known as controllability and
observability of a system.
Definition of Controllability:
• This system is said to be completely state controllable
or simply controllable if for any state x(0)  x 0 there
exists a finite positive integer N and input
u(k ); k [0, N  1] that will transfer state x 0 to state
1
x at k = N. Otherwise, the system is said to be
uncontrollable.

Note: There is no constraint imposed on the input or

on the trajectory which the state should follow.
• Further, the system is said to be uncontrollable
although it may be controllable in part, i.e., it may be
possible to transfer certain states to any desired
states or any state vector to certain regions in state
space.
• Controllability is a property of the coupling between
input and state and therefore involves the matrices A
and B. The controllability of such a system is often
referred to as the controllability of the pair {A, B}.

• By complete controllability of a plant, it is meant that

one can make the plant do whatever is desired.
This definition is perhaps too restrictive in the sense
that too much is demanded from the plant.
• But if one shows that the system equations satisfy
this definition, there can certainly be no limitation
on the design of control system for the plant.

• However, if the system turns out to be

uncontrollable, it doesn’t necessarily mean that the
plant can never be operated in a satisfactory
manner.

• Provided that a control system will maintain the

important variables in an acceptable region, the fact
that the plant is not completely controllable is
immaterial.
Definition of Observability:
It is known that a plant is controllable if no variable in
state vector x(k ) is beyond the influence of input u(k ) .
• The fact that the definition of controllability is
satisfied doesn’t provide any information about how
to design an acceptable controller.
• A full-state feedback controller is very often based
on the control law
u(k )  Kx(k )
• In order to implement this control law, the feedback
of all the state variables is needed. Unfortunately, in
practice, not all the state variables are accessible for
measurement. It can be assumed that only the
inputs and outputs are measurable.
• A subsystem is needed to perform the observation of
state variables based on information received from
the measurement of u(k ) and output c(k ) .
• This subsystem is called an observer and its design
is based on observability property of the system.

• The system is said to be completely observable or

simply observable if for any state x 0 at k = 0, there
exists a finite positive integer N such that the
knowledge of input u(k ); k [0, N  1] and the
output c(k ); k  [0, N  1] suffices to determine the
state x 0 , otherwise the system is said to be
unobservable.
• Observability is a property of the coupling between
the state and output, and involves only the matrices A
and D; the observability of the system is often
referred to as the observability of the pair {A,D}.

Controllability Test:
x(k  1)  Ax(k )  Bu(k )
c(k )  Dx(k )  Eu(k )
Theorem: This discrete-time system is controllable if
and only if the rank of the controllability matrix S:
S  [B AB A 2 B  A n1B]
is n; i.e.,  (S)  n
• This theorem gives not only necessary and sufficient
conditions for controllability but also a method to
compute the control.
Consider a single input system described by
x(k  1)  Ax(k )  bu (k )
The system is assumed to be controllable; therefore
 (S)  [b Ab A 2b  A n1b]  n
Suppose that we wish to control the system in such a
way so as to transfer the system from some initial
state x 0 to a desired state x1 .
• The control sequence can be found by successive
substitution method as:
  u ( N  1) 
u ( N  2) 
[b Ab A 2b A N 1b]    x1  A N x 0  x
 
 
 u (0) 
• This control sequence will accomplish the task in N
sampling intervals.
• If N < n, i.e., the number of control periods is less
than the order of the system, the (unique) solution
may not exist for arbitrary initial and desired states.
• When N = n, the solution exists for arbitrary x1 in the
entire state space and is given by
 u (n  1) 
u (n  2)
   [b Ab A 2b  A n1b]1[x1  A n x 0 ]
  
 
 u ( 0) 
• This yields an open-loop deadbeat control law.
Characteristics of Deadbeat Controller:
(i) The system must have zero steady state error at
the sampling instants for the specified reference
input signal.
(ii) The time required to reach the steady state should
be minimum.
(iii) The digital controller must be physically realizable.

Such controllers are called deadbeat controllers

because they beat the system to the dead stop.
Example:
The transfer function of the plant for an antenna
stabilization system is given by
 (s) 1
G( s)  
U ( s ) s ( s  5)
The state model of this system can be obtained as

 x1 (t )  0 1   x1 (t )  0
x (t )          u (t )
 x 2 (t ) 0  5  x2 (t ) 1

where x1   and x2   .
 A digital positioning system
• The input u changes at discrete instants. If these
instants are separated by T = 0.1sec, then the
discrete-time description of the system becomes
x(k  1)  Ax(k )  bu (k )
1 0.0787  0.0043 
A b 
  0.0787 
 0 0.6065 
0.0043 0.0105  ,
Since det[b Ab ]  det    0.0006  0
 0.0787 0.0477 
the system is controllable.
• Suppose it is desired to transfer  (t )  x1 (t ) from 0 rad
to 0.1 rad. It is also desired that when   0.1 , x2  
should be zero. This will ensure that the antenna will
not move from the desired position once it has
reached it.
• It can be easily seen that such a transfer is possible
in 2 sampling periods (0.2 sec). The control sequence
for this control is given by
 u * (1) 

 *   b Ab  x  A x(0)  
1 1 2

 7.6936 

u ( 0)   12 .6936 

and u (k )  0; k  2,3,
*

• The optimal control sequence u * (k ) and the

corresponding responses are shown.
Deadbeat control and responses
Observability Test:

Theorem: The discrete-time system is observable if

and only if the rank of the observability matrix V,
 D 
 DA 
 
V   DA  is n, i.e.,  (V)  n
2

 
  
n 1
DA 

Duality Property:
Comparing the theorems for controllability and
observability tests, the following observations are
made.
• The pair {A, B} is controllable implies that the
pair {A / , B / } is observable.

• The pair {A, D} is observable implies that the

pair {A / , D/ } is controllable.

Thus, the concepts of controllability and observability are

dual concepts.