JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

6
Sampled Data Systems
and the z-Transform

A sampled data system operates on discrete-time rather than continuous-time signals. A digital
computer is used as the controller in such a system. A D/A converter is usually connected to
the output of the computer to drive the plant. We will assume that all the signals enter and
leave the computer at the same fixed times, known as the sampling times.
A typical sampled data control system is shown in Figure 6.1. The digital computer performs
the controller or the compensation function within the system. The A/D converter converts the
error signal, which is a continuous signal, into digital form so that it can be processed by the
computer. At the computer output the D/A converter converts the digital output of the computer
into a form which can be used to drive the plant.

6.1 THE SAMPLING PROCESS

A sampler is basically a switch that closes every T seconds, as shown in Figure 6.2. When a
continuous signal r (t) is sampled at regular intervals T , the resulting discrete-time signal is
shown in Figure 6.3, where q represents the amount of time the switch is closed.
In practice the closure time q is much smaller than the sampling time T , and the pulses can
be approximated by flat-topped rectangles as shown in Figure 6.4.
In control applications the switch closure time q is much smaller than the sampling time T
and can be neglected. This leads to the ideal sampler with output as shown in Figure 6.5.
The ideal sampling process can be considered as the multiplication of a pulse train with a
continuous signal, i.e.
r * (t) = P(t)r (t), (6.1)
where P(t) is the delta pulse train as shown in Figure 6.6, expressed as


∞
P(t) = δ(t − nT ); (6.2)
n=−∞

thus,


∞
r * (t) = r (t) δ(t − nT ) (6.3)
n=−∞

Microcontroller Based Applied Digital Control D. Ibrahim

C 2006 John Wiley & Sons, Ltd. ISBN: 0-470-86335-8
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

132 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Input Output
Digital
A/D D/A Plant
computer

sensor

Figure 6.1 Sampled data control system

Sampler
r(t) r* (t)

Continuous Sampled
signal signal

Figure 6.2 A sampler

r(t)

0 T 2T 3T 4T 5T

r*(t)

0 T 2T 3T 4T 5T

Figure 6.3 The signal r (t) after the sampling operation

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE SAMPLING PROCESS 133

r(t)

0 T 2T 3T 4T 5T

r*(t)

0 T 2T 3T 4T 5T

Figure 6.4 Sampled signal with flat-topped pulses

r(t)

0 T 2T 3T 4T 5T

r*(t)

0 T 2T 3T 4T 5T

Figure 6.5 Signal r (t) after ideal sampling

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

134 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

P(t)

t
0 T 2T 3T 4T 5T

Figure 6.6 Delta pulse train

or


∞
r * (t) = r (nT )δ(t − nT ). (6.4)
n=−∞

Now
r (t) = 0, for t < 0, (6.5)
and


∞
r * (t) = r (nT )δ(t − nT ). (6.6)
n=0

Taking the Laplace transform of (6.6) gives

∞
R * (s) = r (nT )e−snT . (6.7)
n=0

Equation (6.7) represents the Laplace transform of a sampled continuous signal r (t).
A D/A converter converts the sampled signal r ∗ (t) into a continuous signal y(t). The D/A
can be approximated by a zero-order hold (ZOH) circuit as shown in Figure 6.7. This circuit
remembers the last information until a new sample is obtained, i.e. the zero-order hold takes the
value r (nT) and holds it constant for nT ≤ t < (n + 1)T , and the value r (nT) is used during
the sampling period.
The impulse response of a zero-order hold is shown in Figure 6.8. The transfer function of
a zero-order hold is given by
G(t) = H (t) − H (t − T ), (6.8)

Sampler
r(t) r*(t)
Zero-order
y(t)
hold
Continuous Sampled
signal signal

Figure 6.7 A sampler and zero-order hold

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE SAMPLING PROCESS 135

g(t)

t
0 T

Figure 6.8 Impulse response of a zero-order hold

where H (t) is the step function, and taking the Laplace transform yields

1 e−T s 1 − e−T s
G(s) = − = . (6.9)
s s s
A sampler and zero-order hold can accurately follow the input signal if the sampling time
T is small compared to the transient changes in the signal. The response of a sampler and a
zero-order hold to a ramp input is shown in Figure 6.9 for two different values of sampling
period.

r(t) and y(t)

r(t)

y(t)

T=1 second

seconds
0 1 2 3 4 5 6

r(t) and y(t)

r(t)

y(t)

T = 0.6 second

seconds
0 1 2 3 4 5 6

Figure 6.9 Response of a sampler and a zero-order hold for a ramp input
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

136 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Zero-order
y(t)
hold
r(t) r*(t)

Figure 6.10 Ideal sampler and zero-order hold for Example 6.1

Zero-order
y(t)
hold
r(t) r*(t)

Figure 6.11 Solution for Example 6.1

Example 6.1
Figure 6.10 shows an ideal sampler followed by a zero-order hold. Assuming the input signal
r (t) is as shown in the figure, show the waveforms after the sampler and also after the zero-order
hold.

Solution
The signals after the ideal sampler and the zero-order hold are shown in Figure 6.11.

6.2 THE z-TRANSFORM

Equation (6.7) defines an infinite series with powers of e−snT . The z-transform is defined so
that
Z = esT ; (6.10)
the z-transform of the function r (t) is Z [r (t)] = R(z) which, from (6.7), is given by


∞
R(z) = r (nT )z −n . (6.11)
n=0

Notice that the z-transform consists of an infinite series in the complex variable z, and
R(z) = r (0) + r (T )z −1 + r (2T )z −2 + r (3T )z −3 + . . . ,
i.e. the r (nT ) are the coefficients of this power series at different sampling instants.
The z-transformation is used in sampled data systems just as the Laplace transformation is
used in continuous-time systems. The response of a sampled data system can be determined
easily by finding the z-transform of the output and then calculating the inverse z-transform,
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 137

r(nT)

t
0 T 2T 3T 4T

Figure 6.12 Unit step function

just like the Laplace transform techniques used in continuous-time systems. We will now look
at how we can find the z-transforms of some commonly used functions.

6.2.1 Unit Step Function

Consider a unit step function as shown in Figure 6.12, defined as


0, n < 0,
r (nT ) =
1, n ≥ 0.
From (6.11),


∞

∞
R(z) = r (nT )z −n = z −n = 1 + z −1 + z −2 + z −3 + z −4 + . . .
n=0 n=0
or
z
R(z) = , for |z| > 1.
z−1

6.2.2 Unit Ramp Function

Consider a unit ramp function as shown in Figure 6.13, defined by


0, n < 0,
r (nT ) =
nT, n ≥ 0.
From (6.11),


∞

∞
R(z) = r (nT )z −n = nTz−n = Tz−1 + 2Tz−2 + 3Tz−3 + 4Tz−4 + . . .
n=0 n=0

r(nT)

4
3
2
1
t
0 T 2T 3T 4T

Figure 6.13 Unit ramp function

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

138 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

r(nT)

4
3
2
1
t
0 T 2T 3T 4T

Figure 6.14 Exponential function

or
Tz
R(z) = , for |z| > 1.
(z − 1)2

6.2.3 Exponential Function

Consider the exponential function shown in Figure 6.14, defined as


0, n < 0,
r (nT ) = −anT
e , n ≥ 0.
From (6.11),


∞

∞
R(z) = r (nT )z −n = e−anT z −n = 1 + e−aT z −1 + e−2aT z −2 + e−3aT z −3 + . . .
n=0 n=0
or
1 z
R(z) = = , for |z| < e−aT . (6.12)
1 − e−aT z −1 z − e−aT

6.2.4 General Exponential Function

Consider the general exponential function


0, n < 0,
r (n) =
p n , n ≥ 0.
From (6.11),


∞

∞
R(z) = r (nT )z −n = p n z −n = 1 + pz −1 + p 2 z −2 + p 3 z −3 + . . .
n=0 n=0
or
z
R(z) = , for |z| < | p|.
z−p
Similarly, we can show that
z
R( p −k ) = .
z − p −1
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 139

6.2.5 Sine Function

Consider the sine function, defined as


0, n < 0,
r (nT ) =
sin nωT. n ≥ 0.
Recall that
e j x − e− j x
sin x = ,
2j
so that
e jnωT − e− jnωT e jnωT e− jnωT
r (nT ) = = − . (6.13)
2j 2j 2j
But we already know from (6.12) that the z-transform of an exponential function is
z
R(e−anT ) = R(z) = .
z − e−aT
Therefore, substituting in (6.13) gives
   
1 z z 1 z(e jωT − e− jωT
R(z) = − =
2 j z − e jωT z − e− jωT 2 j z 2 − z(e jωT + e− jωT ) + 1
or
z sin ωT
R(z) = 2 .
z − 2z cos ωT + 1

6.2.6 Cosine Function

Consider the cosine function, defined as


0, n < 0,
r (nT ) =
cos nωT, n ≥ 0.
Recall that
e j x + e− j x
cos x = ,
2
so that
e jnωT + e− jnωT e jnωT e− jnωT
r (nT ) = = + . (6.14)
2 2 2
But we already know from (6.12) that the z-transform of an exponential function is
z
R(e−anT ) = R(z) = .
z − e−aT
Therefore, substituting in (6.14) gives
 
1 z z
R(z) = +
2 z − e jωT z − e− jωT
or
z(z − cos ωT )
R(z) = 2 .
z − 2z cos ωT + 1
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

140 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

6.2.7 Discrete Impulse Function

Consider the discrete impulse function defined as


1, n = 0,
δ(n) =
0, n = 0.
From (6.11),


∞

∞
R(z) = r (nT )z −n = z −n = 1.
n=0 n=0

6.2.8 Delayed Discrete Impulse Function

The delayed discrete impulse function is defined as


1, n = k > 0,
δ(n − k) =
0, n = k.
From (6.11),


∞

∞
R(z) = r (nT )z −n = z −n = z −n .
n=0 n=0

6.2.9 Tables of z-Transforms

A table of z-transforms for the commonly used functions is given in Table 6.1 (a bigger table is
given in Appendix A). As with the Laplace transforms, we are interested in the output response
y(t) of a system and we must find the inverse z-transform to obtain y(t) from Y (z).

6.2.10 The z-Transform of a Function Expressed as a Laplace

Transform

It is important to realize that although we denote the z-transform equivalent of G(s) by G(z),
G(z) is not obtained by simply substituting z for s in G(s). We can use one of the following
methods to find the z-transform of a function expressed in Laplace transform format:
r Given G(s), calculate the time response g(t) by finding the inverse Laplace transform of G(s).
Then find the z-transform either from the first principles, or by looking at the z-transform
tables.
r Given G(s), find the z-tranform G(z) by looking at the tables which give the Laplace trans-
forms and their equivalent z-transforms (e.g. Table 6.1).
r Given the Laplace transform G(s), express it in the form G(s) = N (s)/D(s) and then use
the following formula to find the z-transform G(z):

p
N (xn ) 1
G(z) = , (6.15)
n=1
D (xn ) 1 − e xn T z −1

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 141

Table 6.1 Some commonly used z-transforms

f (kT ) F(z)
δ(t) 1
z
1
z−1
Tz
kT
(z − 1)2
z
e−a kT
z − e−aT
T ze−aT
kT e−akT
(z − e−aT )2
z
ak
z−a
z(1 − e−aT )
1 − e−a kT
(z − 1)(z − e−aT )
z sin aT
sin a kT
z 2 − 2z cos aT + 1
z(z − cos aT )
cos a kT
z 2 − 2z cos aT + 1

where D  = ∂ D/∂s and the xn , n = 1, 2, . . . , p, are the roots of the equation D(s) = 0. Some
examples are given below.

Example 6.2
Let
1
G(s) = .
s 2 + 5s + 6
Determine G(z) by the methods described above.

Solution
Method 1: By finding the inverse Laplace transform. We can express G(s) as a sum of its
partial fractions:
1 1 1
G(s) = = − . (6.16)
(s + 3)(s + 2) s+2 s+3
The inverse Laplace transform of (6.16) is
g(t) = L −1 [G(s)] = e−2t − e−3t . (6.17)
From the definition of the z-transforms we can write (6.17) as


∞
G(z) = (e−2nT − e−3nT )z −n
n=0

= (1 + e−2T z −1 + e−4T z −2 + . . .) − (1 + e−3T z −1 + e−6T z −2 + . . .)

z z
= −
z − e−2T z − e−3T
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

142 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

or
z(e−2T − e−3T )
G(z) = .
(z − e−2T )(z − e−3T )

Method 2: By using the z-transform transform tables for the partial product. From Table 6.1,
the z-transform of 1/(s + a) is z/(z − e−aT ). Therefore the z-transform of (6.16) is
z z
G(z) = −
z − e−2T z − e−3T
or
z(e−2T − e−3T )
G(z) = .
(z − e−2T )(z − e−3T )

Method 3: By using the z-transform tables for G(s). From Table 6.1, the z-transform of
b−a
G(s) = (6.18)
(s + a)(s + b)
is
z(e−aT − e−bT )
G(z) = . (6.19)
(z − e−aT )(z − e−bT )
Comparing (6.18) with (6.16) we have, a = 2, b = 3. Thus, in (6.19) we get
z(e−2T − e−3T )
G(z) = .
(z − e−2T )(z − e−3T )

Method 4: By using equation (6.15). Comparing our expression

1
G(s) =
s2 + 5s + 6
with (6.15), we have N (s) = 1, D(s) = s 2 + 5s + 6 and D  (s) = 2s + 5, and the roots of
D(s) = 0 are x1 = −2 and x2 = −3. Using (6.15),

2
N (xn ) 1
G(z) =  (x ) 1 − e xn T z −1
n=1
D n

or, when x1 = −2,

1 1
G 1 (z) =
1 1 − e z −1
−2T

and when x1 = −3,

1 1
G 2 (z) = .
−1 1 − e−3T z −1
Thus,
1 1 z z
G(z) = − = −
1− e−2T z −1 1− e−3T z −1 z−e −2T z − e−3T
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 143

or
z(e−2T − e−3T )
G(z) = .
(z − e−2T )(z − e−3T )

6.2.11 Properties of z-Transforms

Most of the properties of the z-transform are analogs of those of the Laplace transforms.
Important z-transform properties are discussed in this section.
1. Linearity property
Suppose that the z-transform of f (nT ) is F(z) and the z-transform of g(nT ) is G(z). Then
Z [ f (nT ) ± g(nT )] = Z [ f (nT )] ± Z [g(nT )] = F(z) ± G(z) (6.20)
and for any scalar a
Z [a f (nT )] = a Z [ f (nT )] = a F(z) (6.21)
2. Left-shifting property
Suppose that the z-transform of f (nT ) is F(z) and let y(nT ) = f (nT + mT ). Then


m−1
Y (z) = z m F(z) − f (i T )z m−i . (6.22)
i=0

If the initial conditions are all zero, i.e. f (i T ) = 0, i = 0,1,2, . . . , m − 1, then,

Z [ f (nT + mT )] = z m F(z). (6.23)
3. Right-shifting property
Suppose that the z-transform of f (nT ) is F(z) and let y(nT ) = f (nT − mT ). Then


m−1
Y (z) = z −m F(z) + f (i T − mT )z −i . (6.24)
i=0

If f (nT ) = 0 for k < 0, then the theorem simplifies to

Z [ f (nT − mT )] = z −m F(z). (6.25)
4. Attenuation property
Suppose that the z-transform of f (nT ) is F(z). Then,
Z [e−anT f (nT )] = F[zeaT ]. (6.26)
This result states that if a function is multiplied by the exponential e−anT then in the
z-transform of this function z is replaced by zeaT .
5. Initial value theorem
Suppose that the z-transform of f (nT ) is F(z). Then the initial value of the time response
is given by
lim f (nT ) = lim F(z). (6.27)
n→0 z→∞
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

144 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

6. Final value theorem

Suppose that the z-transform of f (nT ) is F(z). Then the final value of the time response is
given by
lim f (nT ) = lim (1 − z −1 )F(z). (6.28)
n→∞ z→1

Note that this theorem is valid if the poles of (1 − z −1 )F(z) are inside the unit circle or at
z = 1.

Example 6.3
The z-transform of a unit ramp function r (nT ) is
Tz
R(z) = .
(z − 1)2
Find the z-transform of the function 5r (nT ).

Solution
Using the linearity property of z-transforms,
5T z
Z [5r (nT )] = 5R(z) = .
(z − 1)2
Example 6.4
The z-transform of trigonometric function r (nT ) = sin nwT is
z sin wT
R(z) = 2 .
z − 2z cos wT + 1
find the z-transform of the function y(nT ) = e−2T sin nW T .

Solution
Using property 4 of the z-transforms,
Z [y(nT )] = Z [e−2T r (nT )] = R[ze2T ].
Thus,
ze2T sin wT ze2T sin wT
Z [y(nT )] = =
(ze2T )2 − 2ze2T cos wT + 1 z 2 e4T − 2ze2T cos wT + 1
or, multiplying numerator and denominator by e−4T ,
ze−2T sin wT
Z [y(nT )] = .
z2 − 2ze−2T + e−4T

Example 6.5
Given the function
0.792z
G(z) = ,
(z − 1)(z 2 − 0.416z + 0.208)
find the final value of g(nT ).
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 145

Solution
Using the final value theorem,
0.792z
lim g(nT ) = lim (1 − z −1 )
n→∞ z→1 (z − 1)(z 2 − 0.416z + 0.208)
0.792
= lim
z→1 z 2 − 0.416z + 0.208
0.792
= = 1.
1 − 0.416 + 0.208

6.2.12 Inverse z-Transforms

The inverse z-transform is obtained in a similar way to the inverse Laplace transforms. Gener-
ally, the z-transforms are the ratios of polynomials in the complex variable z, with the numerator
polynomial being of order no higher than the denominator. By finding the inverse z-transform
we find the sequence associated with the given z-transform polynomial. As in the case of
inverse Laplace transforms, we are interested in the output time response of a system. There-
fore, we use an inverse transform to obtain y(t) from Y (z). There are several methods to find
the inverse z-transform of a given function. The following methods will be described here:
r power series (long division);
r expanding Y (z) into partial fractions and using z-transform tables to find the inverse
transforms;
r obtaining the inverse z-transform using an inversion integral.
Given a z-transform function Y (z), we can find the coefficients of the associated sequence
y(nT )at the sampling instants by using the inverse z-transform. The time function y(t) is then
determined as


∞
y(t) = y(nT )δ(t − nT ).
n=0

Method 1: Power series. This method involves dividing the denominator of Y (z) into the
numerator such that a power series of the form
Y (z) = y0 + y1 z −1 + y2 z −2 + y3 z −3 + . . .
is obtained. Notice that the values of y(n) are the coefficients in the power series.

Example 6.6
Find the inverse z-transform for the polynomial
z2 + z
Y (z) = .
z2 − 3z + 4
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

146 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Solution
Dividing the denominator into the numerator gives
1 + 4z −1 + 8z −2 + 8z −3
z − 3z + 4 z + z
2 2

z 2 − 3z + 4
4z − 4
4z − 12 + 16z −1
8 − 16z −1
8 − 24z −1 + 32z −2
8z −1 − 32z −2
8z −1 − 24z −2 + 32z −3
...
and the coefficients of the power series are
y(0) = 1,
y(T ) = 4,
y(2T ) = 8,
y(3T ) = 8,
....
The required sequence is
y(t) = δ(t) + 4δ(t − T ) + 8δ(t − 2T ) + 8δ(t − 3T ) + . . .
Figure 6.15 shows the first few samples of the time sequence y(nT ).

Example 6.7
Find the inverse z-transform for Y (z) given by the polynomial
z
Y (z) = .
z2 − 3z + 2

y(t)

1
time
0 T 2T 3T

Figure 6.15 First few samples ofy(t)

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 147

Solution
Dividing the denominator into the numerator gives

z −1 + 3z −2 + 7z −3 + 15z −4
z 2 − 3z + 2 z
z − 3 + 2z −1
3 − 2z −1
3 − 9z −1 + 6z −2
7z −1 − 6z −2
7z −1 − 21z −2 + 14z −3
15z −2 − 14z −3
15z −2 − 45z −3 + 30z −4
...
and the coefficients of the power series are
y(0) = 0
y(T ) = 1
y(2T ) = 3
y(3T ) = 7
y(4T ) = 15
....
The required sequence is thus
y(t) = δ(t − T ) + 3δ(t − 2T ) + 7δ(t − 3T ) + 15δ(t − 4T ) + . . . .
Figure 6.16 shows the first few samples of the time sequence y(nT ).
The disadvantage of the power series method is that it does not give a closed form of the
resulting sequence. We often need a closed-form result, and other methods should be used
when this is the case.

y(t)

15

3
1
time
0 T 2T 3T 4T

Figure 6.16 First few samples ofy(t)

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

148 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Method 2: Partial fractions. Similar to the inverse Laplace transform techniques, a partial
fraction expansion of the function Y (z) can be found, and then tables of known z-transforms
can be used to determine the inverse z-transform. Looking at the z-transform tables, we see
that there is usually a z term in the numerator. It is therefore more convenient to find the partial
fractions of the function Y (z)/z and then multiply the partial fractions by z to obtain a z term
in the numerator.

Example 6.8
Find the inverse z-transform of the function
z
Y (z) =
(z − 1)(z − 2)

Solution
The above expression can be written as
Y (z) 1 A B
= = + .
z (z − 1)(z − 2) z−1 z−2
The values of A and B can be found by equating like powers in the numerator, i.e.
A(z − 2) + B(z − 1) ≡ 1.
We find A = −1, B = 1, giving
Y (z) −1 1
= +
z z−1 z−2
or
−z z
Y (z) = +
z−1 z−2
From the z-transform tables we find that
y(nT ) = −1 + 2n
and the coefficients of the power series are
y(0) = 0,
y(T ) = 1,
y(2T ) = 3,
y(3T ) = 7,
y(4T ) = 15,
...
so that the required sequence is
y(t) = δ(t − T ) + 3δ(t − 2T ) + 7δ(t − 3T ) + 15δ(t − 4T ) + . . . .

Example 6.9
Find the inverse z-transform of the function
1
Y (z) = .
(z − 1)(z − 2)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 149

Solution
The above expression can be written as
Y (z) 1 A B C
= = + + .
z z(z − 1)(z − 2) z z−1 z−2
The values of A, B and C can be found by equating like powers in the numerator, i.e.
A(z − 1)(z − 2) + Bz(z − 2) + C z(z − 1) ≡ 1
or
A(z 2 − 3z + 2) + Bz 2 − 2Bz + C z 2 − C z ≡ 1,
giving
A + B + C = 0,
−3A − 2B − C = 0,
2A = 1.
The values of the coefficients are found to be A = 0.5, B = −1 and C = 0.5. Thus,
Y (z) 1 1 1
= − +
z 2z z − 1 2(z − 2)
or
1 z z
Y (z) = − + .
2 z − 1 2(z − 2)
Using the inverse z-transform tables, we find
2n
y(nT ) = a − 1 + = a − 1 + 2n−1
2
where

1/2, n = 0,
a=
0, n = 0,
the coefficients of the power series are
y(0) = 0
y(T ) = 0
y(2T ) = 1
y(3T ) = 3
y(4T ) = 7
y(5T ) = 15
...,
and the required sequence is
y(t) = δ(t − 2T ) + 3δ(t − 3T ) + 7δ(t − 4T ) + 15δ(t − 5T ) + . . . .

The process of finding inverse z-transforms is aided by considering what form is taken by
the roots of Y (z). It is useful to distinguish the case of distinct real roots and that of multiple
order roots.
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

150 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Case I: Distinct real roots. When Y (z) has distinct real roots in the form
N (z)
Y (z) = ,
(z − p1 )(z − p2 )(z − p3 ) . . . (z − pn )
then the partial fraction expansion can be written as
A1 A2 A3 An
Y (z) = + + + ... +
z − p1 z − p2 z − p3 z − pn
and the coefficients Ai can easily be found as

Ai = (z − pi ) Y (z)|z= pi for i = 1,2,3, . . . , n.

Example 6.10
Using the partial expansion method described above, find the inverse z-transform of
z
Y (z) = .
(z − 1)(z − 2)

Solution
Rewriting the function as
Y (z) A B
= + ,
z z−1 z−2
we find that

1 
A = (z − 1)  = −1,
(z − 1)(z − 2) z=1

1 
B = (z − 2)  = 1.
(z − 1)(z − 2)  z=2

Thus,
z z
Y (z) = +
z−1 z−2
and the inverse z-transform is obtained from the tables as

y(nT ) = −1 + 2n ,

which is the same answer as in Example 6.7.

Example 6.11
Using the partial expansion method described above, find the inverse z-transform of

z2 + z
Y (z) = .
(z − 0.5)(z − 0.8)(z − 1)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

THE z-TRANSFORM 151

Solution
Rewriting the function as
Y (z) A B C
= + +
z z − 0.5 z − 0.8 z − 1
we find that

z+1 
A = (z − 0.5)  = 10,
(z − 0.5)(z − 0.8)(z − 1) z=0.5

z+1 
B = (z − 0.8)  = −30,
(z − 0.5)(z − 0.8)(z − 1) z=0.8

z+1 
C = (z − 1)  = 20.
(z − 0.5)(z − 0.8)(z − 1) z=1
Thus,
10z 30z 20z
Y (z) = − +
z − 0.5 z − 0.8 z − 1
The inverse transform is found from the tables as

y(nT ) = 10(0.5)n − 30(0.8)n + 20

The coefficients of the power series are

y(0) = 0
y(T ) = 1
y(2T ) = 3.3
y(3T ) = 5.89
...
and the required sequence is

y(t) = δ(t − T ) + 3.3δ(t − 2T ) + 5.89δ(t − 3T ) + . . . .

Case II: Multiple order roots. When Y (z) has multiple order roots of the form
N (z)
Y (z) = ,
(z − p1 )(z − p1 )2 (z− p1 )3 . . . (z − p1 )r
then the partial fraction expansion can be written as
λ1 λ2 λ3 λr
Y (z) = + + + ... +
z − p1 (z − p2 )2 (z − p1 )3 (z − p1 )r
and the coefficients λi can easily be found as

1 dk 
λr −k =  . (6.29)
k! dz [(z − pi ) (X (z)/z)] z= pi
k r
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

152 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Example 6.12
Using (6.29), find the inverse z-transform of
z 2 + 3z − 2
Y (z) = .
(z + 5)(z − 0.8)(z − 2)2

Solution
Rewriting the function as
Y (z) z 2 + 3z − 2 A B C D E
= = + + + +
z z(z + 5)(z − 0.8)(z − 2) 2 z z + 5 z − 0.8 (z − 2) (z − 2)2
we obtain

z 2 + 3z − 2  −2
A =z  = = 0.125,
z(z + 5)(z − 0.8)(z − 2) z=0 2  5 × (−0.8) × 4

z 2 + 3z − 2  8
B = (z + 5)  = = 0.0056,
z(z + 5)(z − 0.8)(z − 2) z=−52  −5 × (−5.8) × 49

z 2 + 3z − 2  1.04
C = (z − 0.8)  = = 0.16,
z(z + 5)(z − 0.8)(z − 2) z=0.82  0.8 × 5.8 × 1.14

z 2 + 3z − 2  8
E = (z − 2) 2  = = 0.48,
z(z + 5)(z − 0.8)(z − 2) z=2 2  2 × 7 × 1.2
 
d z 2 + 3z − 2 
D = 
dz z(z + 5)(z − 0.8) z=2
 
z(z + 5)(z − 0.8(2z + 3) − (z 2 + 3z − 2)(3z 2 + 8.4z − 4) 
=  = −0.29.
(z 3 + 4.2z 2 − 4z)2 
z=2

We can now write Y (z) as

0.0056z 0.016z 0.29z 0.48z
Y (z) = 0.125 + + − +
z+5 z − 0.8 (z − 2) (z − 2)2
The inverse transform is found from the tables as
y(nT ) = 0.125a + 0.0056(−5)n + 0.016(0.8)n − 0.29(2)n + 0.24n(2)n ,
where

1, n = 0,
a=
0, n = 0.

Method 3: Inversion formula method. The inverse z-transform can be obtained using the
inversion integral, defined by
1
y(nT ) = Y (z)z n−1 dz. (6.30)
2π j
r
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 153

Using the theorem of residues, the above integral can be evaluated via the expression

y(nT ) = [residues of Y (z)z n−1 ]. (6.31)
at poles of
[Y (z)z k−1 ]

If the function has a simple pole at z = a, then the residue is evaluated as

[residue]|z=a = [(z − a)Y (z)z n−1 ]. (6.32)

Example 6.13
Using the inversion formula method, find the inverse z-transform of
z
Y (z) = .
(z − 1)(z − 2)

Solution
Using (6.31) and (6.32):
 
z n  z n 
y(nT ) = + = −1 + 2n
z − 2 z=1 z − 1 z=2
which is the same answer as in Example 6.9.

Example 6.14
Using the inversion formula method, find the inverse z-transform of
z
Y (z) = .
(z − 1)(z − 2)(z − 3)

Solution
Using (6.31) and (6.32),
  
zn  zn  zn  1 3n
y(nT ) =  +  +  = − 2 n
+ .
(z − 2)(z − 3) z=1 (z − 1)(z − 3) z=2 (z − 1)(z − 2) z=3 2 2

6.3 PULSE TRANSFER FUNCTION AND MANIPULATION

OF BLOCK DIAGRAMS
The pulse transfer function is the ratio of the z-transform of the sampled output and the input
at the sampling instants.
Suppose we wish to sample a system with output response given by
y(s) = e* (s)G(s). (6.33)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

154 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

e(s) e*(s) y(s) y*(s)

G(s)

Figure 6.17 Sampling a system

as illustrated in Figure 6.17. We sample the output signal to obtain

y * (s) = [e* (s)G(s)]* = e* (s)G * (s) (6.34)
and
y(z) = e(z)G(z). (6.35)
Equations (6.34) and (6.35) tell us that if at least one of the continuous functions has been
sampled, then the z-transform of the product is equal to the product of the z-transforms of
each function (note that [e*(s)]* = [e*(s)], since sampling an already sampled signal has no
further effect). G(z) is the transfer function between the sampled input and the output at the
sampling instants and is called the pulse transfer function. Notice from (6.35) that we have no
information about the output y(z) between the sampling instants.

6.3.1 Open-Loop Systems

Some examples of manipulating open-loop block diagrams are given in this section.

Example 6.15
Figure 6.18 shows an open-loop sampled data system. Derive an expression for the z-transform
of the output of the system.

Solution
For this system we can write
y(s) = e* (s)KG(s)
or
y * (s) = [e* (s)KG(s)]* = e* (s)KG* (s)
and
y(z) = e(z)KG(z).

Example 6.16
Figure 6.19 shows an open-loop sampled data system. Derive an expression for the z-transform
of the output of the system.

e(s) e*(s) y(s) y*(s)

G(s)

Figure 6.18 Open-loop system

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 155

e(s) e*(s) y(s) y*(s)

G1(s) G2(s)

Figure 6.19 Open-loop system

Solution
The following expressions can be written for the system:
y(s) = e* (s)G 1 (s)G 2 (s)
or
y * (s) = [e* (s)G 1 (s)G 2 (s)]* = e* (s)[G 1 G 2 ]* (s)
and
y(z) = e(z)G 1 G 2 (z),
where
G 1 G 2 (z) = Z {G 1 (s)G 2 (s)} = G 1 (z)G 2 (z).
For example, if
1
G 1 (s) =
s
and
a
G 2 (s) = ,
s+a
then from the z-transform tables,

a z(1 − e−aT )
Z {G 1 (s)G 2 (s)} = Z =
s(s + a) (z − 1)(z − e−aT )
and the output is given by
z(1 − e−aT )
y(z) = e(z) .
(z − 1)(z − e−aT )

Example 6.17
Figure 6.20 shows an open-loop sampled data system. Derive an expression for the z-transform
of the output of the system.

e(s) e*(s) x(s) x*(s) y(s) y*(s)

G1(s) G2(s)

Figure 6.20 Open-loop system

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

156 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

Solution
The following expressions can be written for the system:
x(s) = e* (s)G 1 (s)
or
x * (s) = e* (s)G *1 (s), (6.36)
and
y(s) = x * (s)G 2 (s)
or
y * (s) = x * (s)G 2* (s). (6.37)
From (6.37) and (6.38),
y * (s) = e* (s)G 1* (s)G *2 (s),
which gives
y(z) = e(z)G 1 (z)G 2 (z).
For example, if
1 a
G 1 (s) = and G 2 (s) = ,
s s+a
then
z az
Z {G 1 (s)} = and Z {G 2 (s)} = ,
z−1 z − ze−aT
and the output function is given by
z az
y(z) = e(z)
z − 1 z − ze−aT
or
az
y(z) = e(z) .
(z − 1)(1 − e−aT )

6.3.2 Open-Loop Time Response

The open-loop time response of a sampled data system can be obtained by finding the inverse
z-transform of the output function. Some examples are given below.

Example 6.18
A unit step signal is applied to the electrical RC system shown in Figure 6.21. Calculate and
draw the output response of the system, assuming a sampling period of T = 1 s.

Solution
The transfer function of the RC system is
1
G(s) = .
s+1
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 157

u(s) u*(s) y(s)

Figure 6.21 RC system with unit step input

For this system we can write

y(s) = u * (s)G(s)
and
y * (s) = u * (s)G * (s),
and taking z-transforms gives
y(z) = u(z)G(z).
The z-transform of a unit step function is
z
u(z) =
z−1
and the z-transform of G(s) is
z
G(z) = .
z − e−T
Thus, the output z-transform is given by
z z z2
y(z) = u(z)G(z) = = ;
z − 1 z − e−T (z − 1)(z − e−T )
since T = 1 s and e−1 = 0.368, we get
z2
y(z) = .
(z − 1)(z − 0.368)
The output response can be obtained by finding the inverse z-transform of y(z). Using partial
fractions,
y(z) A B
= + .
z z − 1 z − 0.368
Calculating A and B, we find that
y(z) 1.582 0.582
= −
z z−1 z − 0.368
or
1.582z 0.582z
y(z) = − .
z−1 z − 0.368
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

158 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

y(nT)
1.571
1.552
1.503

1.367

1.0

T
0 T 2T 3T 4T

Figure 6.22 RC system output response

From the z-transform tables we find

y(nT ) = 1.582 − 0.582(0.368)n .
The first few output samples are
y(0) = 1,
y(1) = 1.367,
y(2) = 1.503,
y(3) = 1.552,
y(4) = 1.571,
and the output response (shown in Figure 6.22) is given by
y(nT ) = δ(T ) + 1.367δ(t − T ) + 1.503δ(t − 2T ) + 1.552δ(t − 3T ) + 1.571δ(t − 4T ) + . . . .

It is important to notice that the response is only known at the sampling instants. For example,
in Figure 6.22 the capacitor discharges through the resistor between the sampling instants,
and this causes an exponential decay in the response between the sampling intervals. But this
behaviour between the sampling instants cannot be determined by the z-transform method of
analysis.

Example 6.19
Assume that the system in Example 6.17 is used with a zero-order hold (see Figure 6.23). What
will the system output response be if (i) a unit step input is applied, and (ii) if a unit ramp input
is applied.

G1(s) G2(s)
u(s) u*(s) y(s)
Z.O.H 1
s+1

Figure 6.23 RC system with a zero-order hold

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 159

Solution
The transfer function of the zero-order hold is
1 − e−T s
G 1 (s) =
s
and that of the RC system is
1
G(s) = .
s+1
For this system we can write
y(s) = u * (s)G 1 G 2 (s)
and
y * (s) = u * (s)[G 1 G 2 ]* (s)
or, taking z-transforms,
y(z) = u(z)G 1 G 2 (z).
Now, T = 1 s and
1 − e−s 1
G 1 G 2 (s) = ,
s s+1
and by partial fraction expansion we can write
 
1
−s 1
G 1 G 2 (s) = (1 − e ) − .
s s+1
From the z-transform tables we then find that
 
−1 z z 0.63
G 1 G 2 (z) = (1 − z ) − = .
z − 1 z − e−1 z − 0.37

(i) For a unit step input,

z
u(z) =
z−1
and the system output response is given by
0.63z
y(z) = .
(z − 1)(z − 0.37)
Using the partial fractions method, we can write
y(z) A B
= + ,
z z − 1 z − 0.37
where A = 1 and B = −1; thus,
z z
y(z) = − .
z − 1 z − 0.37
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

160 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

y(nT)
0.98
0.95
0.86

0.63

T
0 T 2T 3T 4T

Figure 6.24 Step input time response of Example 6.19

From the inverse z-transform tables we find that the time response is given by
y(nT ) = a − (0.37)n ,
where a is the unit step function; thus
y(nT ) = 0.63δ(t − 1) + 0.86δ(t − 2) + 0.95δ(t − 3) + 0.98δ(t − 4) + . . . .
The time response in this case is shown in Figure 6.24.
(ii) For a unit ramp input,
Tz
u(z) =
(z − 1)2
and the system output response (with T = 1) is given by
0.63z 0.63z
y(z) = = 3 .
(z − 1) (z − 0.37)
2 z − 2.37z + 1.74z − 0.37
2

Using the long division method, we obtain the first few output samples as
y(z) = 0.63z −2 + 1.5z −3 + 2.45z −4 + 3.43z −5 + . . .
and the output response is given as
y(nT ) = 0.63δ(t − 2) + 1.5δ(t − 3) + 2.45δ(t − 4) + 3.43δ(t − 5) + . . . ,
as shown in Figure 6.25.

Example 6.20
The open-loop block diagram of a system with a zero-order hold is shown in Figure 6.26.
Calculate and plot the system response when a step input is applied to the system, assuming
that T = 1 s.
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 161

y(nT)
2.45

1.5

0.63

T
2T 3T 4T

Figure 6.25 Ramp input time response of Example 6.19

G1(s) G2(s)
u(s) u*(s) y(s)
1
Z.O.H
s(s + 1)

Figure 6.26 Open-loop system with zero-order hold

Solution
The transfer function of the zero-order hold is
1 − e−T s
G 1 (s) =
s
and that of the plant is
1
G(s) = .
s(s + 1)
For this system we can write
y(s) = u * (s)G 1 G 2 (s)
and
y * (s) = u * (s)[G 1 G 2 ]*(s)
or, taking z-transforms,
y(z) = u(z)G 1 G 2 (z).
Now, T = 1 s and
1 − e−s
G 1 G 2 (s) =
s 2 (s + 1)
or, by partial fraction expansion,
 
1 1 −s1
G 1 G 2 (s) = (1 − e ) 2 − +
s s s+1
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

162 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

y(nT)
0.9145

0.7675

0.3678

T
0 T 2T 3T

Figure 6.27 Output response

and the z-transform is given by

 
1 1 1
G 1 G 2 (z) = (1 − z −1 )Z − + .
s2 s s+1
From the z-transform tables we obtain
 
−1 z z z ze−1 + 1 − 2e−1
G 1 G 2 (z) = (1 − z ) − + =
(z − 1)2 z − 1 z − e−1 (z − 1)(z − e−1 )
0.3678z + 0.2644
= .
z 2 − 1.3678z + 0.3678
After long division we obtain the time response
y(nT ) = 0.3678δ(t − 1) + 0.7675δ(t − 2) + 0.9145δ(t − 3) + . . . ,
shown in Figure 6.27.

6.3.3 Closed-Loop Systems

Some examples of manipulating the closed-loop system block diagrams are given in this
section.

Example 6.21
The block diagram of a closed-loop sampled data system is shown in Figure 6.28. Derive an
expression for the transfer function of the system.

Solution
For the system in Figure 6.28 we can write
e(s) = r (s) − H (s)y(s) (6.38)
and
y(s) = e* (s)G(s). (6.39)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 163

r(s) e(s) e*(s)

G(s) y(s)
+ −

H(s)

Figure 6.28 Closed-loop sampled data system

Substituting (6.39) into (6.38),

e(s) = r (s) − G(s)H (s)e* (s) (6.40)
or
e* (s) = r * (s) − GH* (s)e* (s)
and, solving for e* (s), we obtain
r * (s)
e* (s) = (6.41)
1 + GH* (s)
and, from (6.39),
r * (s)
y(s) = G(s) . (6.42)
1 + GH* (s)
The sampled output is then
r * (s)G * (s)
y * (s) = (6.43)
1 + GH* (s)
Writing (6.43) in z-transform format,
r (z)G(z)
y(z) = (6.44)
1 + GH(z)
and the transfer function is given by
y(z) G(z)
= . (6.45)
r (z) 1 + GH(z)

Example 6.22
The block diagram of a closed-loop sampled data system is shown in Figure 6.29. Derive an
expression for the output function of the system.

Solution
For the system in Figure 6.29 we can write
y(s) = e(s)G(s) (6.46)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

164 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

r(s) e(s)
G(s) y(s)
+ −

y*(s)
H(s)

Figure 6.29 Closed-loop sampled data system

and
e(s) = r (s) − H (s)y * (s). (6.47)
Substituting (6.47) into (6.46), we obtain
y(s) = G(s)r (s) − G(s)H (s)y * (s) (6.48)
or
y * (s) = Gr * (s) − GH* (s)y * (s). (6.49)
Solving for y*(s), we obtain
Gr* (s)
y * (s) = (6.50)
1 + GH* (s)
and
Gr(z)
y(z) = . (6.51)
1 + GH(z)

Example 6.23
The block diagram of a closed-loop sampled data control system is shown in Figure 6.30.
Derive an expression for the transfer function of the system.

Solution
The A/D converter can be approximated with an ideal sampler. Similarly, the D/A converter
at the output of the digital controller can be approximated with a zero-order hold. Denoting

r(s) Digital y(s)

A/D D/A Gp(s)
controller
Plant

H(s)

Sensor

Figure 6.30 Closed-loop sampled data system

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

PULSE TRANSFER FUNCTION AND MANIPULATION OF BLOCK DIAGRAMS 165

G(s)

r(s) e(s) y(s)

D*(s) 1 − e−Ts Gp(s)
s

H(s)

Figure 6.31 Equivalent diagram for Example 6.23

the digital controller by D(s) and combining the zero-order hold and the plant into G(s), the
block diagram of the system can be drawn as in Figure 6.31. For this system can write
e(s) = r (s) − H (s)y(s) (6.52)
and
y(s) = e* (s)D * (s)G(s). (6.53)
Note that the digital computer is represented as D*(s). Using the above two equations, we can
write
e(s) = r (s) − D * (s)G(s)H (s)e* (s) (6.54)
or
e* (s) = r * (s) − D * (s)GH* (s)e* (s)
and, solving for e* (s), we obtain
r * (s)
e* (s) = (6.55)
1 + D (s)GH* (s)
*
and, from (6.53),
r * (s)
y(s) = D * (s)G(s) . (6.56)
1 + D * (s)GH* (s)
The sampled output is then
r * (s)D * (s)G * (s)
y * (s) = , (6.57)
1 + D * (s)GH* (s)
Writing (6.57) in z-transform format,
r (z)D(z)G(z)
y(z) = (6.58)
1 + D(z)GH(z)
and the transfer function is given by
y(z) D(z)G(z)
= . (6.59)
r (z) 1 + D(z)GH(z)
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

166 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

r (s) e(s) e*(s) 1

y(s)
s(s + 1)
+ −

Figure 6.32 Closed-loop system

6.3.4 Closed-Loop Time Response

The closed-loop time response of a sampled data system can be obtained by finding the inverse
z-transform of the output function. Some examples are given below.

Example 6.24
A unit step signal is applied to the sampled data digital system shown in Figure 6.32. Calculate
and plot the output response of the system. Assume that T = 1 s.

Solution
The output response of this system is given in (6.44) as
r (z)G(z)
y(z) = .
1 + GH(z)
where
z z(1 − e−T )
r (z) = , G(z) = , H (z) = 1;
z−1 (z − 1)(z − e−T )
thus,
z/z − 1 z(1 − e−T )
y(z) = .
1 + (z(1 − e−T )/(z − 1)(z − e )) (z − 1)(z − e−T )
−T

Simplifying,
z 2 (1 − e−T )
y(z) = .
(z 2 − 2ze−T + e−T )(z − 1)
Since T = 1,
0.632z 2
y(z) = .
z 3 − 1.736z 2 + 1.104z − 0.368
After long division we obtain the first few terms
y(z) = 0.632z −1 + 1.096z −2 + 1.25z −3 + . . . .
The first 10 samples of the output response are shown in Figure 6.33.

6.4 EXERCISES
1. A function y(t) = 2 sin 4t is sampled every T = 0.1 s. Find the z-transform of the resultant
number sequence.
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

EXERCISES 167

Figure 6.33 First 10 output samples

2. Find the z-transform of the function y(t) = 3t.

3. Find the inverse z-transform of the function
z
y(z) = .
(z + 1)(z − 1)
4. The output response of a system is described with the z-transform
z
y(z) = .
(z + 0.5)(z − 0.2)

(i) Apply the final value theorem to calculate the final value of the output when a unit
step input is applied to the system.
(ii) Check your results by finding the inverse z-transform of y(z).
5. Find the inverse z-transform of the following functions using both long division and the
method of partial fractions. Compare the two methods.
0.2z 0.1(z + 1)
(i) y(z) = (ii) y(z) =
(z − 1)(z − 0.5) (z − 0.2)(z − 1)
0.2 z(z − 1)
(iii) y(z) = (iv) y(z) =
(z − 3)(z − 1) (z − 2)2
JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

168 SAMPLED DATA SYSTEMS AND THE Z-TRANSFORM

r(s) r*(s) y(s)

G(s)

Figure 6.34 Open-loop system for Exercise 6

G(s)
r(s) r*(s) y(s)
Z.O.H 2
(s + 1)(s + 2)

Figure 6.35 Open-loop system with zero-order hold for Exercise 10

6. Consider the open-loop system given in Figure 6.34. Find the output response when a unit
step is applied, if
0.2
G(s) = .
s(s + 1)

7. Draw the output waveform of Exercise 6.

8. Find the z-transform of the following function, assuming that T = 0.5 s:
s+1
y(s) = .
(s − 1)(s + 3)
9. Find the z-transforms of the following functions, using z-transform tables:
s+1 (ii) y(s) =
s
(i) y(s) = (s + 1)2
s(s + 2)
s 2 0.4
(iii) y(s) = (iv) y(s) =
(s + 1)2 (s + 2) s(s + 1)(s + 2)

10. Figure 6.35 shows an open-loop system with a zero-order hold. Find the output response
when a unit step input is applied. Assume that T = 0.1 s and
2
G(s) = .
(s + 1)(s + 2)
11. Repeat Exercise 10 for the case where the plant transfer function is given by
0.1 2s
(i) G(s) = (ii) G(s) =
s(s + 2) (s + 1)(s + 4)
12. Derive an expression for the transfer function of the closed-loop system whose block
diagram is shown in Figure 6.36.

r(s) e(s) e*(s)

G(s) y(s)
+ −

H(s)

Figure 6.36 Closed-loop system for Exercise 12

JWBK063-06 JWBK063-Ibrahim December 22, 2005 15:18 Char Count= 0

FURTHER READING 169

e(s) e*(s)
r(s)
G(s) y*(s)
+ −

H(s)

Figure 6.37 Closed-loop system for Exercise 13

13. Derive an expression for the output function of the closed-loop system whose block dia-
gram is shown in Figure 6.37.

FURTHER READING

[Astrom and Wittenmark, 1984] Astrom, K.J. and Wittenmark, B. Computer Controlled Systems. Prentice Hall,
Englewood Cliffs, NJ, 1984.
[D’Azzo and Houpis, 1966] D’Azzo, J.J. and Houpis, C.H. Feedback Control System Analysis and Synthesis,
2nd edn., McGraw-Hill, New York, 1966.
[Dorf, 1992] Dorf, R.C. Modern Control Systems, 6th edn., Addison-Wesley, Reading, MA,
1992.
[Evans, 1954] Evans, W.R. Control System Dynamics. McGraw-Hill, New York, 1954.
[Houpis and Lamont, 1962] Houpis, C.H. and Lamont, G.B. Digital Control Systems: Theory, Hardware,
Software, 2nd edn., McGraw-Hill, New York, 1962.
[Hsu and Meyer, 1968] Hsu, J.C. and Meyer, A.U. Modern Control Principles and Applications.
McGraw-Hill, New York, 1968.
[Jury, 1958] Jury, E.I. Sampled-Data Control Systems. John Wiley & Sons, Inc., New York,
1958.
[Katz, 1981] Katz, P. Digital Control Using Microprocessors. Prentice Hall, Englewood Cliffs,
NJ, 1981.
[Kuo, 1963] Kuo, B.C. Analysis and Synthesis of Sampled-Data Control Systems. Englewood
Cliffs, NJ, Prentice Hall, 1963.
[Lindorff, 1965] Lindorff, D.P. Theory of Sampled-Data Control Systems. John Wiley & Sons,
Inc., New York, 1965.
[Ogata, 1990] Ogata, K. Modern Control Engineering, 2nd edn., Prentice Hall, Englewood
Cliffs, NJ, 1990.
[Phillips and Harbor, 1988] Phillips, C.L. and Harbor, R.D. Feedback Control Systems. Englewood Cliffs,
NJ, Prentice Hall, 1988.
[Raven, 1995] Raven, F.H. Automatic Control Engineering, 5th edn., McGraw-Hill, New York,
1995.
[Smith, 1972] Smith, C.L. Digital Computer Process Control. Intext Educational Publishers,
Scranton, PA, 1972.
[Strum and Kirk, 1988] Strum, R.D. and Kirk, D.E. First Principles of Discrete Systems and Digital Signal
Processing. Addison-Wesley, Reading, MA, 1988.
[Tou, 1959] Tou, J. Digital and Sampled-Data Control Systems. McGraw-Hill, New York,
1959.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

8
System Stability

This chapter is concerned with the various techniques available for the analysis of the stability
of discrete-time systems.
Suppose we have a closed-loop system transfer function
Y (z) G(z) N (z)
= = ,
R(z) 1 + GH(z) D(z)
where 1 + GH(z) = 0 is also known as the characteristic equation. The stability of the system
depends on the location of the poles of the closed-loop transfer function, or the roots of the
characteristic equation D(z) = 0. It was shown in Chapter 7 that the left-hand side of the
s-plane, where a continuous system is stable, maps into the interior of the unit circle in the z-
plane. Thus, we can say that a system in the z-plane will be stable if all the roots of the
characteristic equation, D(z) = 0, lie inside the unit circle.
There are several methods available to check for the stability of a discrete-time system:
r Factorize D(z) = 0 and find the positions of its roots, and hence the position of the closed-
loop poles.
r Determine the system stability without finding the poles of the closed-loop system, such as
Jury’s test.
r Transform the problem into the s-plane and analyse the system stability using the well-
established s-plane techniques, such as frequency response analysis or the Routh–Hurwitz
criterion.
r Use the root-locus graphical technique in the z-plane to determine the positions of the system
poles.
The various techniques described in this section will be illustrated with examples.

8.1 FACTORIZING THE CHARACTERISTIC EQUATION

The stability of a system can be determined if the characteristic equation can be factorized. This
method has the disadvantage that it is not usually easy to factorize the characteristic equation.
Also, this type of test can only tell us whether or not a system is stable as it is. It does not
tell us about the margin of stability or how the stability is affected if the gain or some other
parameter is changed in the system.

Microcontroller Based Applied Digital Control D. Ibrahim

C 2006 John Wiley & Sons, Ltd. ISBN: 0-470-86335-8
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

188 System Stability

e(s) e*(s) 1 − e−Ts 4 y(s)

r(s)
s s+2
+ −

Figure 8.1 Closed-loop system

Example 8.1
The block diagram of a closed-loop system is shown in Figure 8.1. Determine whether or not
the system is stable. Assume that T = 1 s.

Solution
The closed-loop system transfer function is
Y (z) G(z)
= , (8.1)
R(z) 1 + G(z)
where

 
 
1 − e−T s 4 4 2z(1 − e−2T )
G(z) = Z = (1 − z −1 )Z = (1 − z −1 )
s s+2 s(s + 2) (z − 1)(z − e−2T )
2(1 − e−2T )
= . (8.2)
z − e−2T
For T = 1 s,
1.729
G(z) = .
z − 0.135
The roots of the characteristic equation are 1 + G(z) = 0, or 1 + 1.729/(z − 0.135) = 0,
the solution of which is z = −1.594 which is outside the unit circle, i.e. the system is not
stable.

Example 8.2
For the system given in Example 8.1, find the value of T for which the system is stable.

Solution
From (8.2),
2(1 − e−2T )
G(z) = .
z − e−2T
The roots of the characteristic equation are 1 + G(z) = 0, or 1 + 2(1 − e−2T )/(z − e−2T ) =
0, giving
z − e−2T + 2(1 − e−2T ) = 0
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

JURY’S STABILITY TEST 189

or
z = 3e−2T − 2.
The system will be stable if the absolute value of the root is inside the unit circle, i.e.
|3e−2T − 2|< 1,
from which we get
1
2T < ln 3
or T < 0.549.
Thus, the system will be stable as long as the sampling time T < 0.549.

8.2 JURY’S STABILITY TEST

Jury’s stability test is similar to the Routh–Hurwitz stability criterion used for continuous-
time systems. Although Jury’s test can be applied to characteristic equations of any order, its
complexity increases for high-order systems.
To describe Jury’s test, express the characteristic equation of a discrete-time system of order
n as
F(z) = an z n + an−1 z n−1 + . . . + a1 z + a0 = 0, (8.3)
where an > 0. We now form the array shown in Table 8.1. The elements of this array are defined
as follows:
r The elements of each of the even-numbered rows are the elements of the preceding row, in
reverse order.
r The elements of the odd-numbered rows are defined as:
     
a an−k   b bn−k−1   c cn−2−k 
bk =  0 , ck =  0 , dk =  0 , ···.
an ak  n n−1 bk  cn−2 ck 

Table 8.1 Array for Jury’s stability tests

z0 z1 z2 ... z n−k ... z n−1 zn

a0 a1 a2 ... an−k ... a n−1 an
an an−1 an−2 ... ak ... a1 a0
b0 b1 b2 ... bn−k ... bn−1
bn−1 bn−2 bn−3 ... bk−1 ... b0
c0 c1 c2 ... cn−k ...
cn−2 cn−3 cn−4 ... ck−2 ...
... ... ... ... ...
... ... ... ... ...
l0 l1 l2 l3
l3 l2 l1 l0
m0 m1 m2
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

190 System Stability

The necessary and sufficient conditions for the characteristic equation (8.3) to have roots
inside the unit circle are given as
F(1) > 0, (−1)n F(−1) > 0, |a0 |< an , (8.4)

|b0 | > bn−1

|c0 | > cn−2
|d0 | > dn−3
(8.5)
...
...
|m 0 | > m 2 .
Jury’s test is then applied as follows:
r Check the three conditions given in (8.4) and stop if any of these conditions is not satisfied.
r Construct the array given in Table 8.1 and check the conditions given in (8.5). Stop if any
condition is not satisfied.
Jury’s test can become complex as the order of the system increases. For systems of or-
der 2 and 3 the test reduces to the following simple rules. Given the second-order system
characteristic equation
F(z) = a2 z 2 + a1 z + a 0 = 0, where a2 > 0,
no roots of the system characteristic equation will be on or outside the unit circle provided that
F(1) > 0, F(−1) > 0, |a0 |< a2 ..
Given the third-order system characteristic equation
F(z) = a3 z 3 + a2 z 2 + a1 z + a 0 = 0, where a3 > 0,
no roots of the system characteristic equation will be on or outside the unit circle provided that
F(1) > 0, F(−1) < 0, |a0 |< a3 ,
     
 a3   a1 
det a0 a
> det 0 .
 a3 a0   a3 a2 
Examples are given below.

Example 8.3
The closed-loop transfer function of a system is given by
G(z)
,
1 + G(z)
where
0.2z + 0.5
G(z) = .
z2 − 1.2z + 0.2
Determine the stability of this system using Jury’s test.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

JURY’S STABILITY TEST 191

Solution
The characteristic equation is
0.2z + 0.5
1 + G(z) = 1 + =0
z 2 − 1.2z + 0.2
or
z 2 − z + 0.7 = 0.
Applying Jury’s test,
F(1) = 0.7 > 0, F(−1) = 2.7 > 0, 0.7 < 1.
All the conditions are satisfied and the system is stable.

Example 8.4
The characteristic equation of a system is given by
K (0.2z + 0.5)
1 + G(z) = 1 + = 0.
z 2 − 1.2z + 0.2
Determine the value of K for which the system is stable.

Solution
The characteristic equation is
z 2 + z(0.2K − 1.2) + 0.5K = 0, where K > 0.
Applying Jurys’s test,
F(1) = 0.7K − 0.2 > 0, F(−1) = 0.3K + 2.2 > 0, 0.5K < 1.
Thus, the system is stable for 0.285 < K < 2.

Example 8.5
The characteristic equation of a system is given by
F(z) = z 3 − 2z 2 + 1.4z − 0.1 = 0.
Determine the stability of the system.

Solution
Applying Jury’s test, a3 = 1, a2 = −2, a1 = 1.4, a0 = −0.1 and
F(1) = 0.3 > 0, F(−1) = −4.5 < 0, 0.1 < 1.
The first conditions are satisfied. Applying the other condition,
   
 −0.1 1   −0.1 1.4 
 = −0.99 and   = −1.2;
 1 −0.1   1 −2 
since |0.99| < | − 1.2|, the system is not stable.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

192 System Stability

8.3 ROUTH–HURWITZ CRITERION

The stability of a sampled data system can be analysed by transforming the system characteristic
equation into the s-plane and then applying the well-known Routh–Hurwitz criterion.
A bilinear transformation is usually used to transform the left-hand s-plane into the interior
of the unit circle in the z-plane. For this transformation, z is replaced by
1+w
z= . (8.6)
1−w
Given the characteristic equation in w,
F(w) = bn w n + bn−1 w n−1 + . . . + b1 w + b0 = 0,
then the Routh–Hurwitz array is formed as follows:

wn  bn bn−2 bn−4 ...
w n−1  bn−1 bn−3 bn−5 ...
w n−2  c1 c2 c3 ...
. . .  . . . . . . . . . ...
w 1  j1
w 0  k1
The first two rows are obtained from the equation directly and the other rows are calculated
as follows:
bn−1 bn−2 − bn bn−3
c1 = ,
bn−1
bn−1 bn−4 − bn bn−5
c2 = ,
bn−1
bn−1 bn−6 − bn bn−7
c3 = ,
bn−1
c1 bn−3 − bn−1 c2
d1 = ,
c1
....
The Routh–Hurwitz criterion states that the number of roots of the characteristic equation in
the right hand s-plane is equal to the number of sign changes of the coefficients in the first
column of the array. Thus, for a stable system all coefficients in the first column must have the
same sign.

Example 8.6
The characteristic equation of a sampled data system is given by
z 2 − z + 0.7 = 0.
Determine the stability of the system using the Routh–Hurwitz criterion.

Solution
Transforming the characteristic equation into the w-plane gives

1+w 2 1+w
− + 0.7 = 0,
1−w 1−w
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

ROUTH–HURWITZ CRITERION 193

r(s) e(s) e*(s) 1 − e−Ts K y(s)

+ −
s s( s + 1)

Figure 8.2 Closed-loop system

or
2.7w 2 + 0.6w + 0.7 = 0.
Forming the Routh–Hurwitz array,

w 2  2.7 0.7
w 1  0.6 0
w 0  0.7
there are no sign changes in the first column and thus the system is stable.

Example 8.7
The block diagram of a sampled data system is shown in Figure 8.2. Use the Routh–Hurwitz
criterion to determine the value of K for which the system is stable. Assume that K > 0 and
T = 1 s.

Solution
The characteristic equation is 1 + G(z) = 0, where
1 − e−T s K
G(s) = .
s s(s + 1)

The z-transform is given by


K
G(z) = (1 − z −1 )Z ,
s 2 (s + 1)
which gives
K (0.368z + 0.264)
G(z) = .
(z − 1)(z − 0.368)
The characteristic equation is
K (0.368z + 0.264)
1+ = 0,
(z − 1)(z − 0.368)
or
z 2 − z(1.368 − 0.368K ) + 0.368 + 0.264K = 0.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

194 System Stability

Transforming into the w-plane gives

 
1+w 2 1+w
− (1.368 − 0.368K ) + 0.368 + 0.264K = 0
1−w 1−w
or

w2 (2.736 − 0.104K ) + w(1.264 − 0.528K ) + 0.632K = 0.

We can now form the Routh–Hurwitz array


w 2  2.736 − 0.104K 0.632K
w 1  1.264 − 0.528K 0
w0  0.632K
The system is stable if there is no sign change in the first column. Thus, for stability,

1.264 − 0.528K > 0

or

K < 2.4.

8.4 ROOT LOCUS

The root locus is one of the most powerful techniques used to analyse the stability of a closed-
loop system. This technique is also used to design controllers with required time response
characteristics. The root locus is a plot of the locus of the roots of the characteristic equation
as the gain of the system is varied. The rules of the root locus for discrete-time systems are
identical to those for continuous systems. This is because the roots of an equation Q(z) = 0 in
the z-plane are the same as the roots of Q(s) = 0 in the s-plane. Even though the rules are the
same, the interpretation of the root locus is quite different in the s-plane and the z-plane. For
example, a continuous system is stable if the roots are in the left-hand s-plane. A discrete-time
system, on the other hand, is stable if the roots are inside the unit circle. The construction and
the rules of the root locus for continuous-time systems are described in many textbooks. In this
section only the important rules for the construction of the discrete-time root locus are given,
with worked examples.
Given the closed-loop system transfer function
G(z)
,
1 + GH(z)
we can write the characteristic equation as 1 + k F(z) = 0, and the root locus can then be
plotted as k is varied. The rules for constructing the root locus can be summarized as follows:

1. The locus starts on the poles of F(z) and terminate on the zeros of F(z).
2. The root locus is symmetrical about the real axis.
3. The root locus includes all points on the real axis to the left of an odd number of poles and
zeros.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

ROOT LOCUS 195

4. If F(z) has zeros at infinity, the root locus will have asymptotes as k → ∞. The number of
asymptotes is equal to the number of poles n p , minus the number of zeros n z . The angles
of the asymptotes are given by
180r
θ= , where r = ±1, ±3, ±5, . . . .
n p − nz
The asymptotes intersect the real axis at σ , where
poles of F(z) − zeros of F(z)
σ = .
n p − nz
5. The breakaway points on the real axis of the root locus are at the roots of
dF(z)
= 0.
dz
6. If a point is on the root locus, the value of k is given by
1
1 + kF(z) = 0 or k=− .
F(z)

Example 8.8
A closed-loop system has the characteristic equation
0.368(z + 0.717)
1 + GH(z) = 1 + K = 0.
(z − 1)(z − 0.368)
Draw the root locus and hence determine the stability of the system.

Solution
Applying the rules:

1. The above equation is in the form 1 + kF(z) = 0, where

0.368(z + 0.717)
F(z) = .
(z − 1)(z − 0.368)
The system has two poles at z = 1 and at z = 0.368. There are two zeros, one at z = −0.717
and the other at minus infinity. The locus will start at the two poles and terminate at the two
zeros.
2. The section on the real axis between z = 0.368 and z = 1 is on the locus. Similarly, the
section on the real axis between z = −∞ and z = −0.717 is on the locus.
3. Since n p − n z = 1, there is one asymptote and the angle of this asymptote is
180r
θ= = ±180◦ for r = ±1.
n p − nz
Note that since the angles of the asymptotes are ±180◦ it is meaningless to find the real
axis intersection point of the asymptotes.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

196 System Stability

4. The breakaway points can be found from

dF(z)
= 0,
dz
or
0.368(z − 1)(z − 0.368) − 0.368(z + 0.717)(2z − 1.368) = 0,
which gives
z 2 + 1.434z − 1.348 = 0
and the roots are at
z = −2.08 and z = 0.648.
5. The value of k at the breakaway points can be calculated from

1 
k=−
F(z) z=−2.08,0.648
which gives k = 15 and k = 0.196.

The root locus of the system is shown in Figure 8.3. The locus is a circle starting from the
poles, breaking away at z = 0.648 on the real axis, and then joining the real axis at z = −2.08.

Figure 8.3 Root locus for Example 8.8

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

ROOT LOCUS 197

Figure 8.4 Root locus with unit circle

At this point one part of the locus moves towards the zero at z = −0.717 and the other moves
towards the zero at −∞.
Figure 8.4 shows the root locus with the unit circle drawn on the same axis. The system will
become marginally stable when the locus is on the unit circle. The value of k at these points
can be found either from Jury’s test or by using the Routh–Hurwitz criterion.
Using Jury’s test, the characteristic equation is
0.368(z + 0.717)
1+K = 0,
(z − 1)(z − 0.368)
or

z 2 − z(1.368 − 0.368K ) + 0.368 + 0.263K = 0.

Applying Jury’s test

F(1) = 0.631 for K > 0 .

Also,

|0.263K + 0.368| < 1

which gives K = 2.39 for marginal stability of the system.

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

198 System Stability

Example 8.9
For Example 8.8, calculate the value of k for which the damping factor is ζ = 0.7.

Solution
In Figure 8.5 the root locus of the system is redrawn with the lines of constant damping factor
and constant natural frequency.
From the figure, the roots when ζ = 0.7 are read as s1,2 = 0.61 ± j0.25 (see Figure 8.6).
The value of k can now be calculated as

1 
k=−
F(z) 
z=0.61± j0.25

which gives k = 0.324.

Example 8.10
A closed-loop system has the characteristic equation
(z − 0.2)
1 + GH(z) = 1 + K = 0.
z2 − 1.5z + 0.5
Draw the root locus and hence determine the stability of the system. What will be the value
of K for a damping factor ζ > 0.6 and a natural frequency of ωn > 0.6 rad/s?

Figure 8.5 Root locus with lines of constant damping factor and natural frequency
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

ROOT LOCUS 199

Figure 8.6 Reading the roots when ζ = 0.7

Solution
The above equation is in the form 1 + kF(z) = 0, where
z − 0.2
F(z) = .
z2 − 1.5z + 0.5

The system has two poles at z = 1 and at z = 0.5. There are two zeros, one at z = −0.2 and
the other at infinity. The locus will start from the two poles and terminate at the two zeros.

1. The section on the real axis between z = 0.5 and z = 1 is on the locus. Similarly, the section
on the real axis between z = −∞ and z = 0.2 is on the locus.
2. Since n p − n z = 1, there is one asymptote and the angle of this asymptote is
180r
θ= = ±180◦ for r = ±1
n p − nz
Note that since the angle of the asymptotes are ±180◦ it meaningless to find the real axis
intersection point of the asymptotes.
3. The breakaway points can be found from
dF(z)
=0
dz
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

200 System Stability

or
(z 2 − 1.5z + 0.5) − (z − 0.2)(2z − 1.5) = 0,
which gives
z 2 − 0.4z − 0.2 = 0
and the roots are at
z = −0.290 and z = 0.689.
4. The value of k at the breakaway points can be calculated from

1 
k=−
F(z) z=−0.290,0.689
which gives k = 0.12 and k = 2.08. The root locus of the system is shown in Figure 8.7.
It is clear from this plot that the system is always stable since all poles are inside the unit
circle for all values of k.
Lines of constant damping factor and constant angular frequency are plotted on the same
axis in Figure 8.8.
Assuming that T = 1 s, ωn > 0.6 if the roots are on the left-hand side of the constant angular
frequency line ωn = 0.2π/T . The damping factor will be greater than 0.6 if the roots are below
the constant damping ratio line ζ = 0.6. A point satisfying these properties has been chosen

Figure 8.7 Root locus for Example 8.10

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

NYQUIST CRITERION 201

Figure 8.8 Root locus with lines of constant damping factor and natural frequency

and shown in Figure 8.9. The roots at this point are given as s1,2 = 0.55 ± j0.32. The value
of k can now be calculated as

1 
k=−
F(z)  z=0.55± j0.32

which gives k = 0.377.

8.5 NYQUIST CRITERION

The Nyquist criterion is one of the widely used stability analysis techniques in the s-plane,
based on the frequency response of the system. To determine the frequency response of a
continuous system transfer function G(s), we replace s by jω and use the transfer function
G( jω). In the s-plane, the Nyquist criterion is based on the plot of the magnitude |GH ( jω)|
against the angle  GH ( jω) as ω is varied.
In a similar manner, the frequency response of a transfer function G(z) in the z-plane can
be obtained by making the substitution z = e jωT . The Nyquist plot in the z-plane can then be
obtained by plotting the magnitude of |GH (z)|z=e jωT against the angle  GH (z)|z=e jωT as ω is
varied. The criterion is then
Z = N + P,
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

202 System Stability

Figure 8.9 Point for ζ > 0.6 and ωn > 0.6

where N is the number of clockwise circles around the point −1, P the number of poles of
GH(z) that are outside the unit circle, and Z the number of zeros of GH(z) that are outside the
unit circle.
For a stable system, Z must be equal to zero, and hence the number of anticlockwise circles
around the point –1 must be equal to the number of poles of GH(z).
If GH(z)has no poles outside the unit circle then the criterion becomes simple and for stability
the Nyquist plot must not encircle the point −1.
An example is given below.

Example 8.11
The transfer function of a closed-loop sampled data system is given by
G(z)
,
1 + GH(z)
where
0.4
GH(z) = .
(z − 0.5)(z − 0.2)
Determine the stability of this system using the Nyquist criterion. Assume that T = 1 s.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

NYQUIST CRITERION 203

Solution
Setting z = e jωT = cos ωT + j sin ωT = cos ω + j sin ω,
0.4
G(z)|z=e jωT =
(cos ω + j sin ω − 0.5)(cos ω + j sin ω − 0.2)
or
0.4
G(z)|z=e jωT = .
(cos2 ω − sin ω − 0.7 cos ω + 0.1) + j(2 sin ω cos ω − 0.7 sin ω)
2

This has magnitude

0.4
|G(z)| =
(cos2 ω − sin ω − 0.7 cos ω + 0.1)2 + (2 sin ω cos ω − 0.7 sin ω)2
2

and phase
2 sin ω cos ω − 0.7 sin ω
 G(z) = tan−1 .
cos2 ω − sin2 ω − 0.7 cos ω + 0.1
Table 8.2 lists the variation of the magnitude of G(z) with the phase angle. The Nyquist plot
for this example is shown in Figure 8.10. Since N = 0 and P = 0, the closed-loop system has
no poles outside the unit circle in the x-plane and the system is stable.

The Nyquist diagram can also be plotted by transforming the system into the w-plane and
then using the standard s-plane Nyquist criterion. An example is given below.

Example 8.12
The open-loop transfer function of a unity feedback sampled data system is given by
z
G(z) = .
(z − 1)(z − 0.4)
Derive expressions for the magnitude and the phase of |G(z)| by transforming the system
into the w-plane.

Solution
The w transformation is defined as
1+w
z=
1−w
which gives
(1 + w)/(1 − w) 1+w
G(w) = = ,
((1 + w/1 − w) − 1)((1 + w/1 − w) − 0.4) 2w(0.6 + 1.4w)
or
1+w
G(w) = .
1.2w + 2.8w 2
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

Table 8.2 Magnitude and phase of G(z)

w |G(z)|  G(z)
0 1.0000E+000 0
1.0472E−001 9.8753E−001 −1.9430E+001
2.0944E−001 9.5248E−001 −3.8460E+001
3.1416E−001 9.0081E−001 −5.6779E+001
4.1888E−001 8.3961E−001 −7.4209E+001
5.2360E−001 7.7510E−001 −9.0690E+001
6.2832E−001 7.1166E−001 −1.0625E+002
7.3304E−001 6.5193E−001 −1.2096E+002
8.3776E−001 5.9719E−001 −1.3492E+002
9.4248E−001 5.4789E−001 −1.4820E+002
1.0472E+000 5.0395E−001 −1.6089E+002
1.1519E+000 4.6505E−001 −1.7308E+002
1.2566E+000 4.3075E−001 1.7518E+002
1.3614E+000 4.0058E−001 1.6384E+002
1.4661E+000 3.7408E−001 1.5283E+002
1.5708E+000 3.5082E−001 1.4213E+002
1.6755E+000 3.3044E−001 1.3168E+002
1.7802E+000 3.1259E−001 1.2147E+002
1.8850E+000 2.9698E−001 1.1146E+002
1.9897E+000 2.8337E−001 1.0162E+002
2.0944E+000 2.7154E−001 9.1945E+001
2.1991E+000 2.6130E−001 8.2401E+001
2.3038E+000 2.5250E−001 7.2974E+001
2.4086E+000 2.4501E−001 6.3646E+001
2.5133E+000 2.3872E−001 5.4404E+001
2.6180E+000 2.3354E−001 4.5232E+001
2.7227E+000 2.2939E−001 3.6118E+001
2.8274E+000 2.2622E−001 2.7050E+001
2.9322E+000 2.2399E−001 1.8015E+001
3.0369E+000 2.2266E−001 9.0018E+000

Figure 8.10 Nyquist plot for Example 8.11

204
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

BODE DIAGRAMS 205

But since the w-plane can be regarded an analogue of the s-plane, i.e. s = σ + jw in the
s-plane and w = σw + jww in the z-plane, for the frequency response we can set
w = jww
which yields
1 + ww
G( jww ) = .
1.2 jww − 2.8w 2 w
The magnitude and the phase are then given by
1 + ww
|G( jww )| =  2
(1.2ww )2 + 2.8w 2 w
and
1.2
 G( jww ) = tan−1 ,
2.8ww
where ww is related to w by the expression

wT
ww = tan .
2

8.6 BODE DIAGRAMS

The Bode diagrams used in the analysis of continuous-time systems are not very practical
when used directly in the z-plane. This is because of the e jωT term present in the sampled
data system transfer functions when the frequency response is to be obtained. However, it is
possible to draw the Bode diagrams of sampled data systems by transforming the system into
the w-plane by making the substitution
1+w
z= , (8.7)
1−w
where the frequency in the w-plane (ww ) is related to the frequency in the s-plane (w) by the
expression

wT
ww = tan . (8.8)
2
It is common in practice to use a similar transformation to the one given above, known as the
w -plane transformation, which gives a closer analogy between the frequency in the s-plane
and the w  -plane. The w  -plane transformation defined as
2 z−1
w = , (8.9)
T z+1
or
1 + (T /2)w 
z= , (8.10)
1 − (T /2)w 
and the frequencies in the two planes are related by the expression
2 wT
w = tan . (8.11)
T 2
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

206 System Stability

Note that for small values of the real frequency (s-plane frequency) such that wT is small,
(8.11) reduces to

2 wT 2 wT
w  = tan ≈ =w (8.12)
T 2 T 2
Thus, the w -plane frequency is approximately equal to the s-plane frequency. This approx-
imation is only valid for small values of wT such that tan(wT /2) ≈ wT , i.e.
wT π
≤ ,
2 10
which can also be written as
2π
w≤
10T
or
ws
w≤ , (8.13)
10
where ws is the sampling frequency in radians per second. The interpretation of this is that the
w -plane and the s-plane frequencies will be approximately equal when the frequency is less
than one-tenth of the sampling frequency.
We can use the transformations given in (8.10) and (8.11) to transform a sampled data system
into the w  -plane and then use the standard continuous system Bode diagram analysis.
Some example Bode plots for sampled data systems are given below.

Example 8.13
Consider the closed-loop sampled data system given in Figure 8.11. Draw the Bode diagram
and determine the stability of this system. Assume that T = 0.1 s.

Solution
From Figure 8.11,


1 − e−sT 5 1 − e−0.5
G(z) = Z = ,
s s+5 z − e−0.5
or
0.393
G(z) = .
z − 0.606

+
r(s) 1 − e−sT 5 y(s)
s+5
− s

Figure 8.11 Closed-loop system

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

BODE DIAGRAMS 207

Transforming the system into the w  -plane gives

0.393 − 0.0196w
G(w  ) = ,
0.08w − 0.393
or
4.9(1 − 0.05w  )
G(w ) = .
w  + 4.9
The magnitude of the frequency response and the phase can now be calculated if we set w  = jv
where v is the analogue of true frequency ω. Thus,
4.9(1 − 0.05 jv)
G( jv) = .
jv + 4.9
The magnitude is
4.9 1 + (0.25v)2
|G( jv)| = √
v 2 + 4.92
and
v 
 G( jv) = − tan−1 (0.05) − tan−1 .
4.9
The Bode diagram of the system is shown in Figure 8.12. The system is stable.

Figure 8.12 Bode diagram of the system

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

208 System Stability

Example 8.14
The loop transfer function of a unity feedback sampled data system is given by
0.368z + 0.264
G(z) = 2 .
z − 1.368z + 0.368
Draw the Bode diagram and analyse the stability of the system. Assume that T = 1 s.

Solution
Using the transformation
1 + (T /2)w 1 + 0.5w
z= = ,
1 − (T /2)w 1 − 0.5w
we get
0.368(1 + 0.5w/1 − 0.5w) + 0.264
G(w) =
(1 + 0.5w/1 − 0.5w)2 − 1.368(1 + 0.5w/1 − 0.5w) + 0.368
or
0.0381(w − 2)(w + 12.14)
G(w) = − .
w(w + 0.924)
To obtain the frequency response, we can replace w with jv, giving
0.0381( jv − 2)( jv + 12.14)
G( jv) = − .
jv( jv + 0.924)

The magnitude is then

√ √
0.0381 v 2 + 22 v 2 + 12.142
|G( jv)| = √
v v 2 + 0.9242
and
v v v
 G( jv) = tan−1
+ tan−1 − 90 − tan−1 .
2 12.14 0.924
The Bode diagram is shown in Figure 8.13. The system is stable with a gain margin of 5 dB
and a phase margin of 26◦ .

8.7 EXERCISES
1. Given below are the characteristic equations of some sampled data systems. Using Jury’s
test, determine if the systems are stable.
(a) z 2 − 1.8z + 0.72 = 0
(b) z 2 − 0.5z + 1.2 = 0
(c) z 3 − 2.1z 2 + 2.0z − 0.5 = 0
(d) z 3 − 2.3z 2 + 1.61z − 0.32 = 0
2. The characteristic equation of a sampled data system is given by
(z − 0.5)(z 2 − 0.5z + 1.2) = 0.
Determine the stability of the system.
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

EXERCISES 209

Figure 8.13 Bode diagram of the system

3. For the system shown in Figure 8.14, determine the range of K for stability using Jury’s
test
4. Repeat Exercise 3 using the Routh–Hurwitz criterion.
5. Repeat Exercise 3 using the root locus.
6. The forward gain of a unity feedback sampled data system is given by

K (z − 0.2)
G(z) = .
(z − 0.8)(z − 0.6)

(a) Write an expression for the closed-loop transfer function of the system.
(b) Draw the root locus of the system and hence determine the stability.

+
r(s) K ( z + 0.6) y(s)
− ( z − 1)( z − 0.8)

Figure 8.14 System for Exercise 3

JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

210 System Stability

7. Check the stability of the transfer function

1
G(z) =
z3 + 2.7z 2 + 1.5z + 0.2
using:
(a) Jury’s test;
(b) the Routh–Hurwitz criterion;
(c) the oot locus
8. Repeat Exercise 7 using the Bode diagram.
9. A process with the transfer function
K
s(1 + as)
is preceded by a zero-order hold and is connected to form a unity feedback sampled data
system.
(a) Assuming the sampling time is T , derive an expression for the closed-loop transfer
function of the system.
(b) Draw the root locus of the system and hence determine the value of K for which the
system becomes marginally stable.
10. The open-loop transfer function of a sampled data system is given by
1
G(z) = .
z 3 − 3.1z 2 + 3.1z − 1.1
The closed-loop system is formed by using a unity gain feedback. Use Jury’s criterion to
determine the stability of the system.
11. Use the Bode diagram to determine the stability of the sampled data system given by
z
G(z) = .
(z − 1)(z − 0.6)
12. Repeat Exercise 11 using the Nyquist criterion.
13. The open-loop transfer function of a sampled data system is given by
K (z − 0.6)
G(z) = .
(z − 0.8)(z − 0.4)
(a) Plot the Bode diagram by calculating the frequency response, assuming K = 1.
(b) From the Bode diagram determine the phase margin and the gain margin.
(c) Find the value of K for marginal stability.
(d) If the system is marginally stable, determine the frequency of oscillation.
14. The block diagram of a closed-loop sampled data system is shown in Figure 8.15. Deter-
mine the range of K for stability by:
(a) finding the roots of the characteristic equation;
(b) using Jury’s test;
(c) using the Routh–Hurwitz criterion;
JWBK063-08 JWBK063-Ibrahim December 22, 2005 15:37 Char Count= 0

FURTHER READING 211

+ K ( z + 0. 4)
r(s) y(s)
( z − 0.6)( z − 0.8)
−

Figure 8.15 Closed-loop system for Exercise 14

(d) using the root locus;

(e) drawing the Bode diagram;
(f) drawing the Nyquist diagram.
Which method would you prefer in this exercise and why?
15. Explain the mapping between the s-plane and the simple w-plane. How are the frequency
points mapped?

FURTHER READING
[Bode, 1945] Bode, H.W. Network Analysis and Feedback Amplifier Design, Van Nostrand,
Princeton, NJ, 1945.
[D’Azzo and Houpis, 1966] D’Azzo, J.J. and Houpis, C.H. Feedback Control System Analysis and Synthesis,
2nd edn., McGraw-Hill. New York, 1966.
[Dorf, 1992] Dorf, R.C. Modern Control Systems, 6th edn., Addison-Wesley, Reading, MA, 1992.
[Evans, 1948] Evans, W.R. Graphical analysis of control systems. Trans. AIEE, 67, 1948, pp. 547–
551.
[Evans, 1954] Evans, W.R. Control System Dynamics. McGraw-Hill. New York, 1954.
[Houpis and Lamont, 1962] Houpis, C.H. and Lamont, G.B. Digital Control Systems: Theory, Hardware, Soft-
ware, 2nd edn., McGraw-Hill, New York, 1962.
[Jury, 1958] Jury, E.I. Sampled-Data Control Systems. John Wiley & Sons, Inc., New York, 1958.
[Katz, 1981] Katz, P. Digital Control Using Microprocessors. Prentice Hall, Englewood Cliffs,
NJ, 1981.
[Kuo, 1962] Kuo, B.C. Automatic Control Systems. Prentice Hall, Englewood Cliffs, NJ, 1962.
[Kuo, 1963] Kuo, B.C. Analysis and Synthesis of Sampled-Data Control Systems. Prentice Hall,
Englewood Cliffs, NJ, 1963.
[Lindorff, 1965] Lindorff, D.P. Theory of Sampled-Data Control Systems. John Wiley & Sons, Inc.,
New York, 1965.
[Phillips and Harbor, 1988] Phillips, C.L. and Harbor, R.D. Feedback Control Systems. Englewood Cliffs, NJ,
Prentice Hall, 1988.
[Soliman and Srinath, 1990] Soliman, S.S. and Srinath, M.D. Continuous and Discrete Signals and Systems.
Prentice Hall, Englewood Cliffs, NJ, 1990.
[Strum and Kirk, 1988] Strum, R.D. and Kirk, D.E. First Principles of Discrete Systems and Digital Signal
Processing. Addison-Wesley, Reading, MA, 1988.