Systèmes asservis numériques 2019/2020 M1 Instrumentation

Lecture: Stability Analysis of Discrete Time Systems

1 Stability Analysis using Bilinear Transformation and Routh Stability

Criterion
Another frequently used method in stability analysis of discrete time system is the bilinear
transformation coupled with Routh stability criterion. This requires transformation from z-
plane to another plane called w-plane.

The bilinear transformation has the following form.

a + b
z=
c + d

where a, b, c, d are real constants. If we consider a = b = c = 1 and d = −1, then the transformation
takes a form
 +1 z +1
z= or,  =
 −1 z −1
This transformation maps the inside of the unit circle in the z-plane into the left half of the w-
plane. Let the real part of w be α and imaginary part be β ⇒ w = α+jβ. The inside of the unit circle
in z-plane can be represented by:
 + 1  + j + 1
z = = 1
 −1  + j − 1
( + 1) +  2  1   + 1 2 +  2   − 1 2 +  2    0
2

( ) ( )
( − 1) +  2
2

Thus inside of the unit circle in z-plane maps into the left half of w-plane and outside of the unit
circle in z-plane maps into the right half of w-plane. Although w-plane seems to be similar to s-
plane, quantitatively it is not same.
 +1
In the stability analysis using bilinear transformation, we first substitute z = in the
 −1
characteristics equation P ( z ) = 0 and simplify it to get the characteristic equation in  -plane
as Q ( ) = 0 . Once the characteristics equation is transformed as Q ( ) = 0 , Routh stability
criterion is directly used in the same manner as in a continuous time system.

Dr. NAIL Bachir University Centre Tissemsilt 1

Systèmes asservis numériques 2019/2020 M1 Instrumentation

We will now solve the same examples which were used to understand the Jury’s test.

Example 1 The characteristic equation: P ( z ) = z 3 − 1.3z 2 − 0.08 z + 0.24 = 0 Transforming

P ( z ) into  -domain:

  + 1   + 1   + 1
3 2

Q ( ) =  − 1.3 − 0.08   − 1  + 0.24 = 0

  − 1    − 1 

Or, Q ( ) = 0.14 3 − 1.06 2 − 5.1 − 1.98 = 0

We can now construct the outh array as

3 0.14 −5.1

−1.06 2
−1.98
1
−5.36
0
−1.98
There is one sign change in the first column of the Routh array. Thus, the system is unstable with

one pole at right hand side of the  -plane or outside the unit circle of z-plane.

Example 2: The characteristic equation: P(z) = z4 − 1.2z3 + 0.07z2 + 0.3z − 0.08 = 0 Transforming

P(z) into w-domain:

  + 1   + 1   + 1   + 1
4 3 2

Q ( ) =   − 1.2   + 0.07   + 0.3  − 0.08 = 0

  − 1   − 1   − 1   − 1 

or, Q ( ) = 0.09 4 + 1.32 3 + 5.38 2 + 7.32 + 1.89 = 0

We can now construct the Routh array as
4 0.09 5.38 1.89
3 1.32 7.32

2 4.88 1.89

1 6.81

0 1.89
All elements in the first column of Routh array are positive. Thus, the system is stable.

Dr. NAIL Bachir University Centre Tissemsilt 2

Systèmes asservis numériques 2019/2020 M1 Instrumentation

Example 3: Consider the system shown in Figure 1. Find out the range of K for which the system
is stable.
Solution:
(
G ( z) = K 3
0.084 z 2 + 0.17 z + 0.019 )
( z − 1.5z 2 + 0.553z − 0.05)

Figure 1: Example 3

Characteristic equation:

1+ K
( 0.084 z + 0.17 z + 0.019) = 0
2

( z − 1.5z + 0.553z − 0.05)

3 2

or, P ( z ) = z 3 + ( 0.084 K − 15 ) z 2 + ( 0.17 K + 0.553) z + ( 0.019 K − 0.015 ) = 0

Transforming P ( z ) into  -domain:

  + 1   + 1   + 1
3 2

Q ( ) =   + ( 0.084 K − 15 )   + ( 0.17 K + 0.553)  + ( 0.019 K − 0.015 ) = 0

  − 1   − 1   − 1 

or, Q ( ) = ( 0.003 + 0.27 K )  3 + (1.1 − 0.11K )  2 + ( 3.8 − 0.27 K )  + ( 3.1 + 0.07 K ) = 0

We can now construct the Routh array as

3 0.003 + 0.27K 3.8 − 0.27K

2 1.1 − 0.11K 3.1 + 0.07K

0.01K 2 − 1.55K + 4.17

 1

1.1 − 0.11K

0 3.1 + 0.07K

Dr. NAIL Bachir University Centre Tissemsilt 3

Systèmes asservis numériques 2019/2020 M1 Instrumentation

The system will be stable if all the elements in the first column has same sign. Thus, the
conditions for stability, in terms of K, are:

0.003 + 0.27 K  0  K  −0.011

1.1 − 0.11K  0  K  10
0.01K 2 − 1.55K + 4.17  0  K  2.74 or, K  140.98
3.1 + 0.07 K  0  K  −44.3

Combining above four constraints, the stable range of K can be found as

−0.011  K  2.74
1.1 Singular Cases
In Routh array, tabulation may end in occurrence with any of the following conditions.

• The first element in any row is zero.

• All the elements in a single row are zero.

The remedy of the first case is replacing zero by a small number and then proceeding with the
tabulation. Stability can be checked for the limiting case. Second singular case indicates one or
more of the following cases.

• Pairs of real roots with opposite signs.

• Pairs of imaginary roots.

• Pairs of complex conjugate roots which are equidistant from the origin.

When a row of all zeros occurs, an auxiliary equation A ( ) = 0 is formed by using the
coefficients of the row just above the row of all zeros. The roots of the auxiliary equation are
also the roots of the characteristic equation. The tabulation is continued by replacing the row of
dA ( )
zeros by the coefficients of .
d
Looking at the correspondence between w-plane and z -plane, when an all zero row occurs, we
can conclude that following two scenarios are likely to happen.

• Pairs of real roots in the z -plane that are inverse of each other.

• Pairs of roots on the unit circle simultaneously.

Dr. NAIL Bachir University Centre Tissemsilt 4

Systèmes asservis numériques 2019/2020 M1 Instrumentation

Example 4: Consider the characteristic equation

P ( z ) = z 3 − 1.7 z 2 − z + 0.8 = 0

Transforming P ( z ) into  -domain:

  + 1   + 1   + 1
3 2

Q ( ) =   − 1.7  − + 0.8 = 0
  − 1   − 1    − 1 

or, Q ( ) = 0.9 3 + 0.1 2 − 8.1 − 0.9 = 0

The Routh array:
 3 0.9 −8.1

 2 0.1 −0.9

1 0 0
The tabulation ends here. The auxiliary equation is formed by using the coefficients of  2 row,
as:
A ( ) = 0.1 2 − 0.9 = 0
Taking the derivative,
dA ( )
= 0.2
d
Thus, the Routh tabulation is continued as
3 0.9 −8.1

2 0.1 −0.9

1 0.2 0

 0 −0.9
As there is one sign change in the first row, one of the roots lie in  -plane is on the right-hand
side of the w-plane. This implies that one root in z-plane lies outside the unit circle.

To verify our conclusion, the roots of the polynomial z 3 − 1.7 z 2 − z + 0.8 = 0 , are found out to be
z = 0.5, z = −0.8 and z = 2. Thus, one can see that z = 2 lies outside the unit circle and it is inverse
of z = 0.5 which caused the all zero row in  -plane.

References: Digital Control systems Dr. Indrani Kar

Digital Control System Pro. Rolf Isermann Volume 1
Control Engineering, 5th edition, T.J. Katsuhiko, Modern

Dr. NAIL Bachir University Centre Tissemsilt 5

Digital Control Module 3 Lecture 1

Module 3: Stability Analysis of Discrete Time Systems

Lecture Note 1

1 Stability Analysis of closed loop system in z-plane

Stability is the most important issue in control system design. Before discussing the stability
test let us first introduce the following notions of stability for a linear time invariant (LTI)
system.

1. BIBO stability or zero state stbaility

2. Internal stability or zero input stability

Since we have not introduced the concept of state variables yet, as of now, we will limit our
discussion to BIBO stability only.

An initially relaxed (all the initial conditions of the system are zero) LTI system is said to
be BIBO stable if for every bounded input, the output is also bounded.

However, the stability of the following closed loop system

C(z) G(z)
=
R(Z) 1 + GH(z)

can be determined from the location of closed loop poles in z-plane which are the roots of the
characteristic equation

1 + GH(z) = 0

1. For the system to be stable, the closed loop poles or the roots of the characteristic equation
must lie within the unit circle in z-plane. Otherwise the system would be unstable.

2. If a simple pole lies at |z| = 1, the system becomes marginally stable. Similarly if a pair of
complex conjugate poles lie on the |z| = 1 circle, the system is marginally stable. Multiple
poles on unit circle make the system unstable.

Example 1:
Determine the closed loop stability of the system shown in Figure 1 when K = 1.

I. Kar 1
Digital Control Module 3 Lecture 1

G(s)

R(s) + E(s) E ∗ (s) 1 − e−s 1 C(s)

K
T =1 s s(s + 1)
−

Figure 1: Example 1

Solution:
 
1 − e−s 1
G(z) = Z .
s s(s + 1)
 
1
= (1 − z )Z 2
−1
s (s + 1)

Since H(s) = 1,
 
C(z) z−1 z (1 − e−1 )z
∴ = . −
E(z) z (z − 1)2 (z − 1)(z − e−1 )
(z − e−1 ) − (1 − e−1 )(z − 1)
=
(z − 1)(z − e−1 )
z − 0.368 − 0.632z + 0.632
=
(z − 1)(z − 0.368)
0.368z + 0.264
G(z) =
(z − 1)(z − 0.368)

We know that the characteristics equation is ⇒ 1 + G(z) = 0

⇒ (z − 1)(z − 0.368) + 0.368z + 0.264 = 0

⇒ z 2 − z + 0.632 = 0
⇒ z1 = 0.5 + 0.618j
⇒ z2 = 0.5 − 0.618j

Since |z1 | = |z2 | < 1, the system is stable.

Three stability tests can be applied directly to the characteristic equation without solving for
the roots.
→ Schur-Cohn stability test
→ Jury Stability test
→ Routh stability coupled with bi-linear transformation.

Other stability tests like Lyapunov stability analysis are applicable for state space system models

I. Kar 2
Digital Control Module 3 Lecture 1

which will be discussed later. Computation requirements in Jury test is simpler than Schur-
Cohn when the co-efficients are real which is always true for physical systems.

1.1 Jury Stability Test

Assume that the characteristic equation is as follows,

P (z) = a0 z n + a1 z n−1 + ... + an−1 z + an

where a0 > 0.

Jury Table

Row z 0 z 1 z 2 z 3 z 4 ... z n
1 an an−1 an−2 ... ... ... a0
2 a0 a1 a2 ... ... ... ... an
3 bn−1 bn−2 ... ... b0
4 b0 b1 ... ... ... bn−1
5 cn−2 cn−3 ... c0
6 c0 c1 ... ... cn−2
. ............................
. ............................
. ............................
2n − 3 q2 q1 q0

where,

an an−1−k
bk =
a0 ak+1

k = 0, 1, 2, 3, ...., n − 1

bn−1 bn−2−k
ck =
b0 bk+1

k = 0, 1, 2, 3, ...., n − 2

p3 p2−k
qk =
p0 pk+1

This system will be stable if:

I. Kar 3
Digital Control Module 3 Lecture 1

1. |an | < a0
2. P (z)|z=1 >0
3. P (z)|z=−1 > 0 for n even and P (z)|z=−1 < 0 for n odd
4.

|bn−1 | > |b0 |

|cn−2 | > |c0 |
. .............
. ...........
|q2 | > |q0 |

Example 2: The characteristic equation: P (z) = z 4 − 1.2z 3 + 0.07z 2 + 0.3z − 0.08 = 0

Thus, a0 = 1 a1 = −1.2 a2 = 0.07 a3 = 0.3 a4 = −0.08

We will now check the stability conditions.

1. |an | = |a4 | = 0.08 < a0 = 1 ⇒ First condition is satisfied.

2. P (1) = 1 − 1.2 + 0.07 + 0.3 − 0.08 = 0.09 > 0 ⇒ Second condition is satisfied.

3. P (−1) = 1 + 1.2 + 0.07 − 0.3 − 0.08 = 1.89 > 0 ⇒ Third condition is satisfied.

Jury Table
an a0
b3 = = 0.0064 − 1 = −0.9936
a0 an

an a1
b2 = = −0.08 × 0.3 + 1.2 = 1.176
a0 a3
Rest of the elements are also calculated in a similar fashion. The elements are b1 = −0.0756
b0 = −0.204 c2 = 0.946 c1 = −1.184 c0 = 0.315. One can see
|b3 | = 0.9936 > |b0 | = 0.204
|c2 | = 0.946 > |c0 | = 0.315

All criteria are satisfied. Thus the system is stable.

Example 3 The characteristic equation: P (z) = z 3 − 1.3z 2 − 0.08z + 0.24 = 0

Thus a0 = 1 a1 = −1.3 a2 = −0.08 a3 = 0.24.

Stability conditions are:

1. |a3 | = 0.24 < a0 = 1 ⇒ First condition is satisfied.

I. Kar 4
Digital Control Module 3 Lecture 1

2. P (1) = 1 − 1.3 − 0.08 + 0.24 = −0.14 < 0 ⇒ Second condition is not satisfied.

Since one of the criteria is violated, we may stop the test here and conclude that the system
is unstable. P (1) = 0 or P (−1) = 0 indicates the presence of a root on the unit circle and
in that case the system can at the most become marginally stable if rest of the conditions are
satisfied.

The stability range of a parameter can also be found from Jury’s test which we will see in
the next example.

Example 4: Consider the system shown in Figure 1. Find out the range of K for which
the system is stable.

Solution:
K(0.368z + 0.264)
G(z) =
(z − 1)(z − 0.368)
C(z) K(0.368z + 0.264)
The closed loop transfer function: = 2
R(z) z + (0.368K − 1.368)z + 0.368 + 0.264K

Characteristic equation: P (z) = z 2 + (0.368K − 1.368)z + 0.368 + 0.264K = 0

Since it is a second order system only 3 stability conditions will be there.

1. |a2 | < a0

2. P (1) > 0

3. P (−1) > 0 since n = 2=even. This implies:

1. |0.368 + 0.264K| < 1 ⇒ 2.39 > K > −5.18

2. P (1) = 1 + (0.368K − 1.368) + 0.368 + 0.264K = 0.632K > 0 ⇒ K > 0

3. P (−1) = 1 − (0.368K − 1.368) + 0.368 + 0.264K = 2.736 − 0.104K > 0 ⇒ 26.38 > K

Combining all, the range of K is found to be 0 < K < 2.39.

If K = 2.39, system becomes critically stable.

The characteristics equation becomes:

z 2 − 0.49z + 1 = 0 ⇒ z = 0.244 ± j0.97

Sampling period T = 1 sec.

ws 2π 0.97 ∼
wd = ∠z = tan−1 = 1.324 rad/sec
2π 2π 0.244
The above frequency is the frequency of sustained oscillation.

I. Kar 5
Digital Control Module 3 Lecture 1

1.2 Singular Cases

When some or all of the elements of a row in the Jury table are zero, the tabulation ends
prematurely. This situation is referred to as a singular case. It can be avoided by expanding or
contracting unit circle infinitesimally by an amount ǫ which is equivalent to move the roots of
P (z) off the unit circle. The transformation is:
z1 = (1 + ǫ)z
where ǫ is a very small number. When ǫ is positive the unit circle is expanded and when ǫ
is negative the unit circle is contracted. The difference between the number of zeros found
inside or outside the unit circle when the unit circle is expanded or contracted is the number
of zeros on the unit circle. Since (1 + ǫ)n z n ∼
= (1 + nǫ)z n for both positive and negative ǫ, the
transformation requires the coefficient of the z n term be multiplied by (1 + nǫ).

Example 5: The characteristic equation: P (z) = z 3 + 0.25z 2 + z + 0.25 = 0

Thus, a0 = 1 a1 = 0.25 a2 = 1 a3 = 0.25.

We will now check the stability conditions.

1. |an | = |a3 | = 0.25 < a0 = 1 ⇒ First condition is satisfied.

2. P (1) = 1 + 0.25 + 1 + 0.25 = 2.5 > 0 ⇒ Second condition is satisfied.

3. P (−1) = −1 + 0.25 − 1 + 0.25 = −1.5 < 0 ⇒ Third condition is satisfied.

Jury Table
a3 a0
b2 = = 0.0625 − 1 = −0.9375
a0 a3

a3 a1
b1 = = 0.25 − 0.25 = 0
a0 a2
Thus the tabulation ends here and we know that some of the roots lie on the unit circle. If we
replace z by (1 + ǫ)z, the characteristic equation would become:
(1 + 3ǫ)z 3 + 0.25(1 + 2ǫ)z 2 + (1 + ǫ)z + 0.25 = 0
First three stability conditions are satisfied.

Jury Table

Row z0 z1 z2 z3

1 0.25 1+ǫ 0.25(1 + 2ǫ) 1 + 3ǫ

2 1 + 3ǫ 0.25(1 + 2ǫ) 1+ǫ 0.25
3 0.252 − (1 + 3ǫ)2 0.25(1 + ǫ) − 0.25(1 + 2ǫ)(1 + 3ǫ) −(1 + 3ǫ)(1 + ǫ) + 0.252 (1 + 2ǫ)

I. Kar 6
 Digital Control Module 3 Lecture 1

|b2 | = |0.0625 − (1 + 6ǫ + 9ǫ2 )| and |b0 | = |1 + 3.875ǫ + 3ǫ2 − 0.0625|. Since, when ǫ → 0+ ,
1 + 6ǫ + 9ǫ2 > 1 + 3.875ǫ + 3ǫ2 , thus |b2 | > |b0 | which implies that the roots which are not on
the unit circle are actually inside it and the system is marginally stable.

I. Kar 7
Powered by TCPDF (www.tcpdf.org)
Digital Control Module 7 Lecture 1

Module 7: Discrete State Space Models

Lecture Note 1

1 Introduction to State Variable Model

In the preceding lectures, we have learned how to design a sampled data control system or
a digital system using the transfer function of the system to be controlled. Transfer function
approach of system modeling provides final relation between output variable and input variable.
However, a system may have other internal variables of importance. State variable representa-
tion takes into account of all such internal variables. Moreover, controller design using classical
methods, e.g., root locus or frequency domain method are limited to only LTI systems, partic-
ularly SISO (single input single output) systems since for MIMO (multi input multi output)
systems controller design using classical approach becomes more complex. These limitations of
classical approach led to the development of state variable approach of system modeling and
control which formed a basis of modern control theory.

State variable models are basically time domain models where we are interested in the dynamics
of some characterizing variables called state variables which along with the input represent the
state of a system at a given time.

• State: The state of a dynamic system is the smallest set of variables, x ∈ Rn , such that
given x(t0 ) and u(t), t > t0 , x(t), t > t0 can be uniquely determined.
• Usually a system governed by a nth order differential equation or nth order transfer function
is expressed in terms of n state variables: x1 , x2 , . . . , xn .
• The generic structure of a state-space model of a nth order continuous time dynamical
system with m input and p output is given by:
ẋ(t) = Ax(t) + Bu(t) : State Equation (1)
y(t) = Cx(t) + Du(t) : Output Equation
where, x(t) is the n dimensional state vector, u(t) is the m dimensional input vector, y(t)
is the p dimensional output vector and A ∈ Rn×n , B ∈ Rn×m , C ∈ Rp×n , D ∈ Rp×m .

Example

Consider a nth order differential equation

dn y dn−1 y
+ a 1 + . . . + an y = u
dtn dtn−1

I. Kar 1
Digital Control Module 7 Lecture 1

Define following variables,

y = x1
dy
dt
= x2
.. .
. = ..
dn−1 y
dtn−1
= xn
dn y
dtn
= −a1 xn−1 − a2 xn−2 − . . . − an x1 + u
The nth order differential equation may be written in the form of n first order differential
equations as
ẋ1 = x2
ẋ2 = x3
.. .
. = ..
ẋn = −a1 xn−1 − a2 xn−2 − . . . − an x1 + u
or in matrix form as,
ẋ = Ax + Bu
where    
0 1 0 ... 0 0
 0 0 1 ... 0   0 
A= .. .. B= ..
   
 
 . ... .   . 
0 0 0 ... 1 1
The output can be one of states or a combination of many states. Since, y = x1 ,
y = [1 0 0 0 . . . 0]x

1.1 Correlation between state variable and transfer functions models

The transfer function corresponding to state variable model (1), when u and y are scalars, is:
Y (s)
G(s) = = C(sI − A)−1 B + D (2)
U (s)
Q(s)
=
|sI − A|
where |sI − A| is the characteristic polynomial of the system.

1.2 Solution of Continuous Time State Equation

The solution of state equation (1) is given as
Z t
A(t−t0 )
x(t) = e x(t0 ) + eA(t−τ ) Bu(τ )dτ
t0

where eAt = Φ(t) is known as the state transition matrix and x(t0 ) is the initial state of the
system.

I. Kar 2
Digital Control Module 7 Lecture 1

2 State Variable Analysis of Digital Control Systems

The discrete time systems, as discussed earlier, can be classified in two types.

1. Systems that result from sampling the continuous time system output at discrete instants
only, i.e., sampled data systems.

2. Systems which are inherently discrete where the system states are defined only at discrete
time instants and what happens in between is of no concern to us.

2.1 State Equations of Sampled Data Systems

Let us assume that the following continuous time system is subject to sampling process with
an interval of T .

ẋ(t) = Ax(t) + Bu(t) : State Equation (3)

y(t) = Cx(t) + Du(t) : Output Equation

We know that the solution to above state equation is:

Z t
x(t) = Φ(t − t0 )x(t0 ) + Φ(t − τ )Bu(τ )dτ
t0

Since the inputs are constants in between two sampling instants, one can write:

u(τ ) = u(kT ) for, kT ≤ τ ≤ (k + 1)T

which implies that the following expression is valid within the interval kT ≤ τ ≤ (k + 1)T if we
consider t0 = kT :
Z t
x(t) = Φ(t − kT )x(kT ) + Φ(t − τ )Bu(kT )dτ
kT
Z t
Let us denote Φ(t − τ )Bdτ by θ(t − KT ). Then we can write:
kT

x(t) = Φ(t − kT )x(kT ) + θ(t − KT )u(kT )

If t = (k + 1)T ,

x((k + 1)T ) = Φ(T )x(kT ) + θ(T )u(kT ) (4)

Z (k+1)T
where Φ(T ) = eAT and θ(T ) = Φ((k + 1)T − τ )Bdτ . If t′ = τ − kT , we can rewrite
Z T kT

θ(T ) as θ(T ) = Φ(T − t′ )Bdt′ . Equation (4) has a similar form as that of equation (3) if we
0
consider φ(T ) = Ā and θ(T ) = B̄. Similarly by setting t = kT , one can show that the output
equation also has a similar form as that of the continuous time one.

I. Kar 3
Digital Control Module 7 Lecture 1

When T = 1,

x(k + 1) = Φ(1)x(k) + θ(1)u(k)

y(k) = Cx(k) + Du(k)

2.2 State Equations of Inherently Discrete Systems

When a discrete system is composed of all digital signals, the state and output equations can
be described by

x(k + 1) = Ax(k) + Bu(k)

y(k) = Cx(k) + Du(k)

2.3 Discrete Time Approximation of A Continuous Time State Space

Model
Let us consider the dynamical system described by the state space model (3). By approximating
the derivative at t = kT using forward difference, we can write:

1
ẋ(t) t=kT
= [x((k + 1)T ) − x(kT )]
T
1
⇒ [x((k + 1)T ) − x(kT )] = Ax(kT ) + Bu(kT )
T
and, y(kT ) = Cx(kT ) + Du(kT )

Rearranging the above equations,

x((k + 1)T ) = (I + T A)x(kT ) + T Bu(kT )

If, T = 1 ⇒ x(k + 1) = (I + A)x(k) + Bu(k)
and y(k) = Cx(k) + Du(k)

We can thus conclude from the discussions so far that the discrete time state variable model
of a system can be described by

x(k + 1) = Ax(k) + Bu(k)

y(k) = Cx(k) + Du(k)

where A, B are either the descriptions of an all digital system or obtained by sampling the
continuous time process.

I. Kar 4
Digital Control Module 7 Lecture 2

Module 7: Discrete State Space Models

Lecture Note 2

In this lecture we will discuss about the relation between transfer function and state space
model for a discrete time system and various standard or canonical state variable models.

1 State Space Model to Transfer Function

Consider a discrete state variable model

x(k + 1) = Ax(k) + Bu(k)

y(k) = Cx(k) + Du(k) (1)

Taking the Z-transform on both sides of Eqn. (1), we get

zX(z) − zx0 = AX(z) + BU (z)

Y (z) = CX(z) + DU (z)

where x0 is the initial state of the system.

⇒ (zI − A)X(z) = zx0 + BU (z)

or, X(z) = (zI − A)−1 zx0 + (zI − A)−1 BU (z)

To find out the transfer function, we assume that the initial conditions are zero, i.e., x0 = 0,
thus  
Y (z) = C(zI − A)−1 B + D U (z)
Therefore, the transfer function becomes

Y (z)
G(z) = = C(zI − A)−1 B + D (2)
U (z)

which has the same form as that of a continuous time system.

I. Kar 1
Digital Control Module 7 Lecture 2

2 Various Canonical Forms

We have seen that transform domain analysis of a digital control system yields a transfer function
of the following form.

Y (z) β0 z m + β1 z m−1 + . . . + βm
G(z) = = m≤n (3)
U (z) z n + α1 z n−1 + . . . + αn
Various canonical state variable models can be derived from the above transfer function model.

2.1 Controllable canonical form

Consider the transfer function as given in Eqn. (3). Without loss of generality, let us consider
the case when m = n. Let
X̄(z) 1
= n
U (z) z + α1 z n−1 + . . . + αn
In time domain, the above equation may be written as

x̄(k + n) + α1 x̄(k + n − 1) + . . . + αn x̄(k) = u(k)

Now, the output Y (z) may be written in terms of X̄(z) as

Y (z) = β0 z n + β1 z n−1 + . . . + βn X̄(z)

or in time domain as

y(k) = β0 x̄(k + n) + β1 x̄(k + n − 1) + . . . + βn x̄(k)

The block diagram representation of above equations is shown in Figure 1. State variables are
selected as shown in Figure 1.
The state equations are then written as:

x1 (k + 1) = x2 (k)
x2 (k + 1) = x3 (k)
.. .
. = ..
xn (k + 1) = −αn x1 (k) − αn−1 x2 (k) − . . . − α1 xn (k) + u(k)

Output equation can be written as by following the Figure 1.

y(k) = (βn − αn β0 )x1 (k) + (βn−1 − αn−1 β0 )x2 (k) + . . . + (β1 − α1 β0 )xn (k) + β0 u(k)

In state space form, we have

x(k + 1) = Ax(k) + Bu(k)

y(k) = Cx(k) + Du(k) (4)

I. Kar 2
Digital Control Module 7 Lecture 2

y(k)
+ + ... + +

β0 β1 β2 ... βn−1 βn

u(k) xn xn−1 x2 x1 = x̄(k)

+ ...
z −1
z −1
z −1
z −1

α1 α2 ... αn−1 αn

+ + ... +

Figure 1: Block Diagram representation of controllable canonical form

where    
0 1 0 ... 0 0
 0 0  1 ... 0  0
.. .. .. ..
   
A= B =  0
   
 . . . . . 
 . .  .. 
 0 0 0 ... 1  .
−αn −αn−1 −αn−2 . . . −α1 1
C = [βn − αn β0 βn−1 − αn−1 β0 . . . β1 − α1 β0 ] D = β0

2.2 Observable Canonical Form

Equation (3) may be rewritten as

z n + α1 z n−1 + . . . + αn Y (z) = β0 z n + β1 z n−1 + . . . + βn U (z)

or, z n [Y (z) − β0 U (z)] + z n−1 [α1 Y (z) − β1 U (z)] + . . . + [αn Y (z) − βn U (z)] = 0
or, Y (z) = β0 U (z) − z −1 [α1 Y (z) − β1 U (z)] − . . . − z −n [αn Y (z) − βn U (z)]

The corresponding block diagram is shown in Figure 2. Choosing the outputs of the delay blocks

I. Kar 3
Digital Control Module 7 Lecture 2

u(k)

βn βn−1 ... β1 β0

+ x1 + x2 + xn + y(k)
z −1
z −1 ... z −1
- - - -

αn αn−1 ... α1

Figure 2: Block Diagram representation of observable canonical form

as the state variables, we have following state equations

xn (k + 1) = xn−1 (k) − α1 (xn (k) + β0 u(k)) + β1 u(k)
xn−1 (k + 1) = xn−2 (k) − α2 (xn (k) + β0 u(k)) + β2 u(k)
.. .
. = ..
x1 (k + 1) = −αn (xn (k) + β0 u(k)) + βn u(k)
y(k) = xn (k) + β0 u(k)
This can be rewritten in matrix form (4) with
 
0 0 0 . . . 0 −αn  
1 0 0 . . . 0 −αn−1  βn − α β
n 0
  βn−1 − αn−1 β0 
0 1 0 . . . 0 −αn−2 
A=  B= .. C = [0 0 . . . 1] D = β0
 
 .. .. .. ..

. . . . . . .


 . 
β1 − α1 β0
0 0 0 . . . 1 −α1

2.3 Duality
In previous two sections we observed that the system matrix A in observable canonical form is
transpose of the system matrix in controllable canonical form. Similarly, control matrix B in
observable canonical form is transpose of output matrix C in controllable canonical form. So also
output matrix C in observable canonical form is transpose of control matrix B in controllable
canonical form.

I. Kar 4
Digital Control Module 7 Lecture 2

2.4 Jordan Canonical Form

In Jordan canonical form, the system matrix A represents a diagonal matrix for distinct poles
which basically form the diagonal elements of A.

Assume that z = λi , i = 1, 2, . . . , n are the distinct poles of the given transfer function
(3). Then partial fraction expansion of the transfer function yields

Y (z) β̄1 z n−1 + β̄2 z n−2 + . . . + β̄n

= β0 +
U (z) z n + α1 z n−1 + . . . + αn
β̄1 z n−1 + β̄2 z n−2 + . . . + β̄n
= β0 +
(z − λ1 )(z − λ2 ) . . . (z − λn )
r1 r2 rn
= β0 + + + ... + (5)
z − λ1 z − λ2 z − λn
A parallel realization of the transfer function (5) is shown in Figure 3.

β0

+
u(k) + x1 (k) + y(k)
Z −1 r1
+ +
+

λ1

+ x2 (k)
Z −1 r2

λ2

+ xn (k)
Z −1 rn

λn

Figure 3: Block Diagram representation of Jordan canonical form

Considering the outputs of the delay blocks as the state variables, we can construct the state
model in matrix form (4), with

I. Kar 5
Digital Control Module 7 Lecture 2

 
λ1 0 0 ... 0  
 0 λ2 0 1
 ... 0  1 
A =  0 0 λ3 ... 0 B =  ..  C = [r1 r2 . . . rn ] D = β0
  
 .. .. .. .. 

.
. . . ... . 
1
0 0 0 . . . λn
When the matrix A has repeated eigenvalues, it cannot be expressed in a proper diagonal form.
However, it can be expressed in a Jordan canonical form which is nearly a diagonal matrix. Let
us consider that the system has eigenvalues, λ1 , λ1 , λ2 and λ3 . In that case, A matrix in Jordan
canonical form will be  
λ1 1 0 0
 0 λ1 0 0 
A=  0 0 λ2 0 


0 0 0 λ3
1. The diagonal elements of the matrix A are eigenvalues of the same.

2. The elements below the principal diagonal are zero.

3. Some of the elements just above the principal diagonal are one.

4. The matrix can be divided into a number of blocks, called Jordan blocks, along the
diagonal. Each block depends on the multiplicity of the eigenvalue associated with it. For
example Jordan block associated with a eigenvalue z1 of multiplicity 4 can be written as
 
z1 1 0 0
 0 z1 1 0 
A=  0 0 z1 1 


0 0 0 z1

Example: Consider the following discrete transfer function.

0.17z + 0.04
G(z) =
z 2 − 1.1z + 0.24
Find out the state variable model in 3 different canonical forms.

Solution:
The state variable model in controllable canonical form can directly be derived from the
transfer function, where the A, B, C and D matrices are as follows:
   
0 1 0  
A= , B= , C = 0.04 0.17 , D = 0
−0.24 1.1 1
The matrices in state model corresponding to observable canonical form are obtained as,
   
0 −0.24 0.04  
A= , B= , C= 0 1 , D=0
1 1.1 0.17

I. Kar 6
Digital Control Module 7 Lecture 2

To find out the state model in Jordan canonical form, we need to fact expand the transfer
function using partial fraction, as
0.17z + 0.04 0.352 −0.182
G(z) = = +
− 1.1z + 0.24
z2 z − 0.8 z − 0.3
Thus the A, B, C and D matrices will be:
   
0.8 0 1  
A= , B= , C = 0.352 −0.182 , D = 0
0 0.3 1

I. Kar 7
Digital Control Module 7 Lecture 4

Module 7: Discrete State Space Models

Lecture Note 4

In this lecture we would discuss about the solution of discrete state equation, computation
of discrete state transition matrix and state diagram.

1 Solution to Discrete State Equation

Consider the following state model of a discrete time system:

x(k + 1) = Ax(k) + Bu(k)

where the initial conditions are x(0) and u(0). Putting k = 0 in the above equation, we get

x(1) = Ax(0) + Bu(0)

Similarly if we put k = 1, we would get

x(2) = Ax(1) + Bu(1)

Putting the expression of x(1) ⇒ x(2) = A2 x(0) + ABu(0) + Bu(1)

For k = 2,

x(3) = Ax(2) + Bu(2)

= A3 x(0) + A2 Bu(0) + ABu(1) + Bu(2)

and so on. If we combine all these equations, we would get the following expression as a
general solution:

k−1
X
k
x(k) = A x(0) + Ak−1−i Bu(i)
i=0

As seen in the above expression, x(k) has two parts. One is the contribution due to the initial
state x(0) and the other one is the contribution of the external input u(i) for i = 0, 1, 2, · · · , k−1.

I. Kar 1
Digital Control Module 7 Lecture 4

When the input is zero, solution of the homogeneous state equation x(k + 1) = Ax(k) can be
written as

x(k) = Ak x(0)
where Ak = φ(k) is the state transition matrix.

2 Evaluation of φ(k)
Similar to the continuous time systems, the state transition matrix of a discrete state model
can be evaluated using the following different techniques.

1. Using Inverse Z-transform:


φ(k) = Z −1 (zI − A)−1

2. Using Similarity Transformation If Λ is the diagonal representation of the matrix A,

then Λ = P −1 AP . When a matrix is in diagonal form, computation of state transition
matrix is straight forward:  k 
λ1 0 . . . 0
Λk =  0 λk2 . . . 0 
0 0 . . . λkn
Given Λk , we can compute Ak = P Λk P −1

3. Using Caley Hamilton Theorem

Example Compute Ak for the following system using three different techniques and hence find
y(k) for k ≥ 0.
     
0 1 0 k 1
x(k + 1) = x(k) + (−1) ; x(0) =
−0.21 −1 1 0
y(k) = x2 (k)
 
0 1
Solution: A = and eigenvalues of A are −0.3 and −0.7.
−0.21 −1

Method 1

( −1 )
z − 1 −1
Ak = Z −1 (zI − A)−1 = Z −1
1 z−1

I. Kar 2
Digital Control Module 7 Lecture 4

z+1 1
 
z 2 +z+0.21 z 2 +z+0.21
Ak = Z −1  
−0.21 z
z 2 +z+0.21 z 2 +z+0.21
 1.75 0.75 2.5 2.5

z+0.3
− z+0.7 z+0.3
− z+0.7
= Z −1  
−0.525 0.525 −0.75 1.75
z+0.3
+ z+0.7 z+0.3
+ z+0.7
1.75(−0.3)k − 0.75(−0.7) k
2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)
Method 2
(−0.3)k
 
k k k 0
A = P Λ P where Λ =
−1
. Eigen values are −0.3 and −0.7. The corre-
0 (−0.7)k    
1 1
sponding eigenvectors are found, by using equation Avi = λi vi , as and respec-
−0.3 −0.7
tively. The transformation matrix is given by
   
1 1 −1 1.75 2.5
P = ⇒ P =
−0.3 −0.7 −0.75 −2.5
Thus,
Ak = P Λk P −1
(−0.3)k
   
1 1 0 1.75 2.5
=
−0.3 −0.7 0 (−0.7)k −0.75 −2.5
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)
Method 3: Caley Hamilton Theorem
The eigenvalues are −0.3 and −0.7.
(−0.3)k = β0 − 0.3β1
(−0.7)k = β0 − 0.7β1
Solving,
β0 = 1.75(−0.3)k − 0.75(−0.7)k
β1 = 2.5(−0.3)k − 2.5(−0.7)k
Hence,
φ(k) = Ak = β0 I + β1 A
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)

I. Kar 3
Digital Control Module 7 Lecture 4

The solution x(k) is

k−1
X
k
x(k) = A x(0) + Ak−1−i Bu(i)
i=0
k−1 
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k−1−i − 2.5(−0.7)k−1−i
  X 
= + (−1)i
−0.525(−0.3)k + 0.525(−0.7)k −0.75(−0.3)k−1−i + 1.75(−0.7)k−1−i
i=0

Since y(k) = x2 (k), we can write

k−1
X
k k
y(k) = −0.525(−0.3) + 0.525(−0.7) + [−0.75(−0.3)k−1−i + 1.75(−0.7)k−1−i ](−1)i
i=0
k−1
X k−1
X
= −0.525(−0.3)k + 0.525(−0.7)k − 0.75(−0.3)k−1 (1/0.3)i + 1.75(−0.7)k−1 (1/0.7)i
i=0 i=0

Now,
k−1 k−1
X
i
X 1 − (3.33)k
(1/0.3) = (3.33)i = = −0.43[1 − (3.33)k ]
i=0 i=0
1 − 3.33
k−1 k−1
X X 1 − (1.43)k
(1/0.7)i = (1.43)i = = −2.33[1 − (1.43)k ]
i=0 i=0
1 − 1.43

Putting the above expression in y(k)

y(k) = 0.475(−0.3)k − 5.3(−0.7)k + (−0.3)k (3.33)k + 5.825(−0.7)k−1 (1.43)k

3 State Diagram
Conventional signal flow graph method was meant for only algebraic equation, thus these are
generally used for the derivation of input output relation in a transformed domain.

State diagram or state transition signal flow graph is an extension of conventional signal flow
graph which can be applied to represent differential and difference equations as well.

Example 1: Draw the state diagram for the following differential equation.

ÿ(t) + 2ẏ(t) + y(t) = u(t)

Considering the state variables as x1 (t) = y(t) and x2 (t) = ẏ(t), we can write

ẋ1 (t) = x2 (t)

ẋ2 (t) = −x1 (t) − 2x2 (t) + u(t)

I. Kar 4
Digital Control Module 7 Lecture 4

1 1
1 ẋ2 s s x1 1
u(t) y(t)
U (s) ẋ1 = x2 Y (s)

−2

−1

Figure 1: State Diagram of Example 1

The state diagram is shown in Figure 1.

Example 2: Consider a discrete time system described by the following state difference equa-
tions.
x1 (k + 1) = −x1 (k) + x2 (k)
x2 (k + 1) = −x1 (k) + u(k)
y(k) = x1 (k) + x2 (k)
Draw the state diagram.

The state diagram is shown in Figure 2.

1

u(k) 1 x2 (k + 1) z −1 1 x1 (k + 1) z −1 1 y(k)
x2 (k) x1 (k)

−1

−1

Figure 2: State Diagram of Example 2

3.1 State Diagram of Zero Order Hold

State diagram of zero order hold is important for sampled date control systems. Let the input
to and output of a ZOH is e∗ (t) and h(t) respectively. Then, for the inetrval kT ≤ t ≤ (k + 1)T ,
h(t)e(kT )
Or,
e(kT )
H(s) =
s
Therefore, the state diagram, as shown in Figure 3, consists of a single branch with gain s−1 .

I. Kar 5
Digital Control Module 7 Lecture 4

s−1
e(kT ) H(s)

Figure 3: State Diagram of Zero Order Hold

4 System Response between Sampling Instants

State variable method is a convenient way to evaluate the system response between the sampling
instants of a sampled data system. State transition equation is given as:
Z t
x(t) = φ(t − t0 )x(t0 ) + u(t0 ) φ(t − τ )Bdτ
t0

where x(t0 ) is the initial state of the system and u(t) is the external input.
Z t
when t0 = kT, x(t) = φ(t − kT )x(kT ) + u(kT ) φ(t − τ )Bdτ
kT
Since we are interested in response between the sampling instants, let us consider t =
(k + ∆)T where k = 0, 1, 2, · · · and 0 ≤ ∆ ≤ 1. This implies

x((k + ∆)T ) = φ(∆T )x(kT ) + θ(∆T )u(kT )

Z (k+∆)T
where θ(∆T ) = φ((k + ∆)T − τ )Bdτ . By varying the value of ∆ between 0 and 1
kT
all information on x(t) for all t can be obtained.

Example 3: Consider the following state model of a continuous time system.

   
−1 −1 0
ẋ(t) = x(t) + u(t)
0 −2 1
y(t) = x1 (t)

which undergoes through a sampling process with period T . To derive the discrete state
space model, let us first compute the state transition matrix of the continuous time system
using Caley Hamilton Theorem.

λ+1 1
△(λ) = |λI − A| = = (λ + 1)(λ + 2) = 0 ⇒ λ1 = −1, λ2 = −2
0 λ+2

Let f (λ) = eλt

This implies

e−t = β0 − β1 (λ1 = −1)

e−2t = β0 − 2β1 (λ2 = −2)

I. Kar 6
 Digital Control Module 7 Lecture 4

Solving the above equations

β1 = e−t − e−2t β0 = 2e−t − e−2t

Then

eAt = β0 I + β1 A
 −t −2t 
e e − e−t
=
0 e−2t

Thus the discrete state matrix A is given as

 
e−T e−2T − e−T
A = φ(T ) =
0 e−2T

The discrete input matrix B can be computed as

Z T  
′ 0
B = θ(T ) = Φ(T − t ) dt′
0 1
 −T 
e − 0.5e−2T − 0.5
=
0.5 − 0.5e−2T

When t = (k + 1)T , the discrete state equation is described by

 −T −2T   −T 
e e − e−T e − 0.5e−2T − 0.5
x((k + 1)T ) = x(kT ) + u(kT )
0 e−2T 0.5 − 0.5e−2T

When t = (k + ∆)T ,

   −∆T 
e−∆T e−2∆T − e−∆T e − 0.5e−2∆T − 0.5
x(kT + ∆T ) = x(kT ) + u(kT )
0 e−2∆T 0.5 − 0.5e−2∆T

If the sampling period T = 1,

   
0.37 −0.23 −0.2
x(k + 1) = x(k) + u(k)
0 0.14 0.43

At the sampling instants we can find x(k) by putting k = 0, 1, 2 · · · . If ∆ = 0.5, then

between the sampling instants,

   
0.61 −0.24 −0.08
x(k + 0.5) = x(k) + u(k)
0 0.37 0.32

The responses in between the sampling instants, i.e., x(0.5), x(1.5), x(2.5) etc., can be found
by putting k = 0, 1, 2 · · · .

I. Kar 7
Powered by TCPDF (www.tcpdf.org)
Digital Control Module 7 Lecture 4

Module 7: Discrete State Space Models

Lecture Note 4

In this lecture we would discuss about the solution of discrete state equation, computation
of discrete state transition matrix and state diagram.

1 Solution to Discrete State Equation

Consider the following state model of a discrete time system:

x(k + 1) = Ax(k) + Bu(k)

where the initial conditions are x(0) and u(0). Putting k = 0 in the above equation, we get

x(1) = Ax(0) + Bu(0)

Similarly if we put k = 1, we would get

x(2) = Ax(1) + Bu(1)

Putting the expression of x(1) ⇒ x(2) = A2 x(0) + ABu(0) + Bu(1)

For k = 2,

x(3) = Ax(2) + Bu(2)

= A3 x(0) + A2 Bu(0) + ABu(1) + Bu(2)

and so on. If we combine all these equations, we would get the following expression as a
general solution:

k−1
X
k
x(k) = A x(0) + Ak−1−i Bu(i)
i=0

As seen in the above expression, x(k) has two parts. One is the contribution due to the initial
state x(0) and the other one is the contribution of the external input u(i) for i = 0, 1, 2, · · · , k−1.

I. Kar 1
Digital Control Module 7 Lecture 4

When the input is zero, solution of the homogeneous state equation x(k + 1) = Ax(k) can be
written as

x(k) = Ak x(0)
where Ak = φ(k) is the state transition matrix.

2 Evaluation of φ(k)
Similar to the continuous time systems, the state transition matrix of a discrete state model
can be evaluated using the following different techniques.

1. Using Inverse Z-transform:


φ(k) = Z −1 (zI − A)−1

2. Using Similarity Transformation If Λ is the diagonal representation of the matrix A,

then Λ = P −1 AP . When a matrix is in diagonal form, computation of state transition
matrix is straight forward:  k 
λ1 0 . . . 0
Λk =  0 λk2 . . . 0 
0 0 . . . λkn
Given Λk , we can compute Ak = P Λk P −1

3. Using Caley Hamilton Theorem

Example Compute Ak for the following system using three different techniques and hence find
y(k) for k ≥ 0.
     
0 1 0 k 1
x(k + 1) = x(k) + (−1) ; x(0) =
−0.21 −1 1 0
y(k) = x2 (k)
 
0 1
Solution: A = and eigenvalues of A are −0.3 and −0.7.
−0.21 −1

Method 1

( −1 )
z − 1 −1
Ak = Z −1 (zI − A)−1 = Z −1
1 z−1

I. Kar 2
Digital Control Module 7 Lecture 4

z+1 1
 
z 2 +z+0.21 z 2 +z+0.21
Ak = Z −1  
−0.21 z
z 2 +z+0.21 z 2 +z+0.21
 1.75 0.75 2.5 2.5

z+0.3
− z+0.7 z+0.3
− z+0.7
= Z −1  
−0.525 0.525 −0.75 1.75
z+0.3
+ z+0.7 z+0.3
+ z+0.7
1.75(−0.3)k − 0.75(−0.7) k
2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)
Method 2
(−0.3)k
 
k k k 0
A = P Λ P where Λ =
−1
. Eigen values are −0.3 and −0.7. The corre-
0 (−0.7)k    
1 1
sponding eigenvectors are found, by using equation Avi = λi vi , as and respec-
−0.3 −0.7
tively. The transformation matrix is given by
   
1 1 −1 1.75 2.5
P = ⇒ P =
−0.3 −0.7 −0.75 −2.5
Thus,
Ak = P Λk P −1
(−0.3)k
   
1 1 0 1.75 2.5
=
−0.3 −0.7 0 (−0.7)k −0.75 −2.5
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)
Method 3: Caley Hamilton Theorem
The eigenvalues are −0.3 and −0.7.
(−0.3)k = β0 − 0.3β1
(−0.7)k = β0 − 0.7β1
Solving,
β0 = 1.75(−0.3)k − 0.75(−0.7)k
β1 = 2.5(−0.3)k − 2.5(−0.7)k
Hence,
φ(k) = Ak = β0 I + β1 A
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k − 2.5(−0.7)k
 

= 
k k k k
−0.525(−0.3) + 0.525(−0.7) −0.75(−0.3) + 1.75(−0.7)

I. Kar 3
Digital Control Module 7 Lecture 4

The solution x(k) is

k−1
X
k
x(k) = A x(0) + Ak−1−i Bu(i)
i=0
k−1 
1.75(−0.3)k − 0.75(−0.7)k 2.5(−0.3)k−1−i − 2.5(−0.7)k−1−i
  X 
= + (−1)i
−0.525(−0.3)k + 0.525(−0.7)k −0.75(−0.3)k−1−i + 1.75(−0.7)k−1−i
i=0

Since y(k) = x2 (k), we can write

k−1
X
k k
y(k) = −0.525(−0.3) + 0.525(−0.7) + [−0.75(−0.3)k−1−i + 1.75(−0.7)k−1−i ](−1)i
i=0
k−1
X k−1
X
= −0.525(−0.3)k + 0.525(−0.7)k − 0.75(−0.3)k−1 (1/0.3)i + 1.75(−0.7)k−1 (1/0.7)i
i=0 i=0

Now,
k−1 k−1
X
i
X 1 − (3.33)k
(1/0.3) = (3.33)i = = −0.43[1 − (3.33)k ]
i=0 i=0
1 − 3.33
k−1 k−1
X X 1 − (1.43)k
(1/0.7)i = (1.43)i = = −2.33[1 − (1.43)k ]
i=0 i=0
1 − 1.43

Putting the above expression in y(k)

y(k) = 0.475(−0.3)k − 5.3(−0.7)k + (−0.3)k (3.33)k + 5.825(−0.7)k−1 (1.43)k

3 State Diagram
Conventional signal flow graph method was meant for only algebraic equation, thus these are
generally used for the derivation of input output relation in a transformed domain.

State diagram or state transition signal flow graph is an extension of conventional signal flow
graph which can be applied to represent differential and difference equations as well.

Example 1: Draw the state diagram for the following differential equation.

ÿ(t) + 2ẏ(t) + y(t) = u(t)

Considering the state variables as x1 (t) = y(t) and x2 (t) = ẏ(t), we can write

ẋ1 (t) = x2 (t)

ẋ2 (t) = −x1 (t) − 2x2 (t) + u(t)

I. Kar 4
Digital Control Module 7 Lecture 4

1 1
1 ẋ2 s s x1 1
u(t) y(t)
U (s) ẋ1 = x2 Y (s)

−2

−1

Figure 1: State Diagram of Example 1

The state diagram is shown in Figure 1.

Example 2: Consider a discrete time system described by the following state difference equa-
tions.
x1 (k + 1) = −x1 (k) + x2 (k)
x2 (k + 1) = −x1 (k) + u(k)
y(k) = x1 (k) + x2 (k)
Draw the state diagram.

The state diagram is shown in Figure 2.

1

u(k) 1 x2 (k + 1) z −1 1 x1 (k + 1) z −1 1 y(k)
x2 (k) x1 (k)

−1

−1

Figure 2: State Diagram of Example 2

3.1 State Diagram of Zero Order Hold

State diagram of zero order hold is important for sampled date control systems. Let the input
to and output of a ZOH is e∗ (t) and h(t) respectively. Then, for the inetrval kT ≤ t ≤ (k + 1)T ,
h(t)e(kT )
Or,
e(kT )
H(s) =
s
Therefore, the state diagram, as shown in Figure 3, consists of a single branch with gain s−1 .

I. Kar 5
Digital Control Module 7 Lecture 4

s−1
e(kT ) H(s)

Figure 3: State Diagram of Zero Order Hold

4 System Response between Sampling Instants

State variable method is a convenient way to evaluate the system response between the sampling
instants of a sampled data system. State transition equation is given as:
Z t
x(t) = φ(t − t0 )x(t0 ) + u(t0 ) φ(t − τ )Bdτ
t0

where x(t0 ) is the initial state of the system and u(t) is the external input.
Z t
when t0 = kT, x(t) = φ(t − kT )x(kT ) + u(kT ) φ(t − τ )Bdτ
kT
Since we are interested in response between the sampling instants, let us consider t =
(k + ∆)T where k = 0, 1, 2, · · · and 0 ≤ ∆ ≤ 1. This implies

x((k + ∆)T ) = φ(∆T )x(kT ) + θ(∆T )u(kT )

Z (k+∆)T
where θ(∆T ) = φ((k + ∆)T − τ )Bdτ . By varying the value of ∆ between 0 and 1
kT
all information on x(t) for all t can be obtained.

Example 3: Consider the following state model of a continuous time system.

   
−1 −1 0
ẋ(t) = x(t) + u(t)
0 −2 1
y(t) = x1 (t)

which undergoes through a sampling process with period T . To derive the discrete state
space model, let us first compute the state transition matrix of the continuous time system
using Caley Hamilton Theorem.

λ+1 1
△(λ) = |λI − A| = = (λ + 1)(λ + 2) = 0 ⇒ λ1 = −1, λ2 = −2
0 λ+2

Let f (λ) = eλt

This implies

e−t = β0 − β1 (λ1 = −1)

e−2t = β0 − 2β1 (λ2 = −2)

I. Kar 6
Digital Control Module 7 Lecture 4

Solving the above equations

β1 = e−t − e−2t β0 = 2e−t − e−2t

Then

eAt = β0 I + β1 A
 −t −2t 
e e − e−t
=
0 e−2t

Thus the discrete state matrix A is given as

 
e−T e−2T − e−T
A = φ(T ) =
0 e−2T

The discrete input matrix B can be computed as

Z T  
′ 0
B = θ(T ) = Φ(T − t ) dt′
0 1
 −T 
e − 0.5e−2T − 0.5
=
0.5 − 0.5e−2T

When t = (k + 1)T , the discrete state equation is described by

 −T −2T   −T 
e e − e−T e − 0.5e−2T − 0.5
x((k + 1)T ) = x(kT ) + u(kT )
0 e−2T 0.5 − 0.5e−2T

When t = (k + ∆)T ,

   −∆T 
e−∆T e−2∆T − e−∆T e − 0.5e−2∆T − 0.5
x(kT + ∆T ) = x(kT ) + u(kT )
0 e−2∆T 0.5 − 0.5e−2∆T

If the sampling period T = 1,

   
0.37 −0.23 −0.2
x(k + 1) = x(k) + u(k)
0 0.14 0.43

At the sampling instants we can find x(k) by putting k = 0, 1, 2 · · · . If ∆ = 0.5, then

between the sampling instants,

   
0.61 −0.24 −0.08
x(k + 0.5) = x(k) + u(k)
0 0.37 0.32

The responses in between the sampling instants, i.e., x(0.5), x(1.5), x(2.5) etc., can be found
by putting k = 0, 1, 2 · · · .

I. Kar 7