Z Transforms

Chapter 75
75.1 Samplers and Holds
75.2 Equivalence of Pulses and Impulses
75.3 Analysis of Pulse Trains
75.4 Transforms of Variables
75.5 Difference Equations
75.6 Inverse Z Transformation
75.7 Worked Example

Z transforms are used in the analysis and design of 75.1 Samplers and Holds
systems in which the signals are sampled on a regu-
Consider the sampler and hold device of Fig-
lar basis.Referred to as a sampled data signals,such
ure 75.1.
a signal consists of a train of pulses which represent
The sampler is essentially a switch which closes
some continuous variable. Sampled data systems
for a finite time t every T s, the sampling period T
are common in practice. For example, many ana-
normally being significantly shorter than the dom-
lytical instruments are intermittent in nature: they
inant time constants of the system. This has the
sample their inputs and produce pulse outputs.
effect of converting the continuous signal x(t) into
Digital controllers are inherently sampled data in
a train of pulses x*(t) at the sampling instants, as
nature: their algorithms are executed in phase with
depicted in Figure 75.2.
the sampling of the I/O signals.
This chapter introduces some important con- x(t)
cepts in relation to the sampling process and pulse
trains,defines the Z transform,introduces its prop-
erties, and shows how to use the Z transform for t
solving difference equations. These provide the ba-
sis for Chapter 76 on sampled data systems analy-
sis and Chapter 77 on impulse compensator design. x* ( t )
Many texts cover Z transforms and sampled data
theory, the reader is referred in particular to Dut- t
ton (1997) and Ogata (2002).
T
Δt
x h (t)

x(t ) x* ( t ) x h (t) t
G h (s)
T
Fig. 75.2 The effect of a sample/hold device on a continuous
Fig. 75.1 Representation of sampler and hold devices signal
 622 75 Z Transforms

Note that the height of the pulse varies according zero duration, infinite magnitude and unit area. A
to the value of x(t) at the sampling instant, and useful insight into the nature of an impulse is given
that the sampler transmits no information in be- by finding its Laplace transform:
tween the sampling instants. Note also that two
Lim
completely different signals that happen to have L {ı(0)} = f(s)
the same values at the sampling instants would t → 0  
Lim 1
produce the same sampled signal. It is therefore = 1 − e−t.s
obvious that a sampled data signal contains less t → 0 t.s
  
information than its original continuous signal. Lim 1
The hold device converts the sampled signal = 1 − 1 − t.s
t → 0 t.s
x*(t) into a continuous one xh (t) which approxi- 
mately reproduces the continuous signal x(t). The 1 1
simplest, and most common, hold device is the so- + (t.s)2 − (t.s)3 + . . . =1
2 3!
called zero order hold (ZOH) which converts the
sampled signal into one which is constant between A classical assumption in sampled data theory is
consecutive sampling instants. The ZOH is often that a pulse function may be represented by an im-
described as a latch because it latches onto the pulse function, where the magnitude of the pulse
value of the last pulse. Its transfer function is as function is equal to the area under the impulse
follows: function. This is depicted in Figure 75.4.
1 − e−Ts
Gh (s) = (75.1)
s x* ( t ) x* ( t ) Area=x(t)
which is two steps, in opposite directions, the sec-
Δt
≡
ond being delayed by the sampling period T. x(t)
t t

75.2 Equivalence of Pulses and Fig. 75.4 The equivalence of a pulse and an impulse function
Impulses This assumption may be justified by considering
Consider a pulse at time t = 0 of finite duration t a sampled data signal to be the product of a con-
and magnitude 1/t, as depicted in Figure 75.3. tinuous and an impulse signal. Consider the two
signals depicted in Figure 75.5.
1 x(t )
Δt Area=1

t x (T )
0 Δt t
Fig. 75.3 A pulse of unit area 0

δ( T )
The Laplace transform for this is:

1
f(s) = 1 − e−t.s
t.s t
0 T
The impulse ı(t) may be considered to be the lim-
iting case of the pulse as t tends to zero, i.e. it is of Fig. 75.5 The product of a continuous and an impulse signal
 75.3 Analysis of Pulse Trains 623

Suppose that x(t) is a continuous signal and that x∗ (t) = ı(t)x(0) + ı(t − T)x(T)
ı(T) is a single impulse of unit area, referred to as + ı(t − 2T)x(2T) + . . .
the unit impulse, that occurs when t = T. If x∗ (t) n
is the product of the two signals, it can be seen by = ı(t − kT).x(kT)
inspection that: k=0

for t < T and t > T x∗ (t) = 0 Laplace transform, remembering that L {ı(t)} = 1,
and for t = T x∗ (T) = ı(T).x(T) this gives:

where x∗ (T) is the unit impulse weighted by a fac- ∞


tor of x(T), which is consistent with the classical x∗ (s) = x(kT).e−kTs
assumption. k=0

Now define z to be a complex variable according to

z = eTs . It can be seen that z is equivalent to a time
75.3 Analysis of Pulse Trains advance of duration T. Also, since both T and s are
Consider the sampler shown in Figure 75.6. independent of time, so too is z. Whence:

∞

x(t ) x* ( t ) x∗ (s) = x(kT).z−k (75.2)
T k=0

Fig. 75.6 A sampler with sampling period T

This is the definition of the Z transform. By nota-
Suppose that x(t) is sampled at regular intervals of tion:
time, i.e. t = 0, T, 2T, 3T, 4T. . .. ' (
This results in a train of pulses of varying mag- Z {x(t)} = x(z) = x∗ (s) = L x∗ (t) (75.3)
nitude, each of which may be represented by a
weighted impulse. Thus:

x∗ (t) = ıT (t).x(t)
75.4 Transforms of Variables
= ı(0)x(0) + ı(T)x(T) + ı(2T)x(2T) + . . . The formation of Z transforms is best illustrated
by means of a couple of examples.
where ıT (t) is a train of unit impulses spaced a
time T apart, as depicted in Figure 75.7. Example 1
The unit step function x(t) = 0 for t < 0, and
x(t) = 1 for t ≥ 0.

∞

Z {1(t)} = 1(kT).z−k
t k=0
0 T 2T 3T 4T
= + 1 + z−1 + z−2 + z−3 + · · ·
Fig. 75.7 A train of unit impulses z.Z {1(t)} = z + 1 + z−1 + z−2 + z−3 + · · ·

However, each of the unit impulses is identical to (z − 1) .Z {1(t)} = z

the first, i.e. that at t = 0, except for being delayed z
Z {1(t)} =
by some multiple of the sampling period T. Thus: (z − 1)
 624 75 Z Transforms

Example 2 Table 75.1 Table of commonly used Z transforms

The exponential function x(t) = 0 for t < 0, and x(t)t≥0 x(z)
x(t) = e−at for t ≥ 0. (x(t)t<0 = 0)
∞
' (  ı(t) 1
Z e−at = e−akT .z−k
k=0 ı(t − kT) z−k
= +1 + e−aT z−1 + e−2aT z−2 z
1
+ e−3aT z−3 + · · · z−1
' (
eaT .z.Z e−at = e+aT z + 1 + e−aT z−1 ak
z
z−a
+e −2aT −2
z +e −3aT −3
z +···
' ( ak cos(k )
z
eaT z − 1 .Z e−at = eaT z z+a
' ( z
Z e−at = Tz
z − e−aT t
(z − 1)2
T2 z(z + 1)
Example 3 t2
(z − 1)3
The function whose Laplace transform is: z
e−at
1 z − e−aT
x(s) =
s (s + 1) 1 − e−aT z
1 − e−at
(z − 1)(z − e−aT )
Expanding into partial fractions gives:
1 1 Tze−aT
t.e−at
x(s) = − (z − e−aT )2
s s+1
z. sin(!T)
The Z transform is found from the inverse Laplace sin(!t)
z2 − 2z. cos(!T) + 1
transform as follows:
' ( z.(z − cos(!T))
x(z) = Z {x(t)} = Z L−1 {x(s)} cos(!t)
2 2 33 z2 − 2z. cos(!T) + 1
1 1 ' (
= Z L−1 − = Z 1(t) − e−t z.e−aT . sin(!T)
s s+1 e−at sin(!t)
z2 − 2z.e−aT . cos(!T) + e−2aT
' ( z z
= Z {1(t)} − Z e−t = − z2 − z.e−aT . cos(!T)
z − 1 z − e−T e−at cos(!t)
z 1 − e−T z2 − 2z.e−aT . cos(!T) + e−2aT
=
(z − 1) z − e−T Note that the notation has been simplified in the
Note that an important property of Z transforms, tables by writing x(k) instead of x(kT), x(k + 1)
that the transform of the sum of two functions is instead of x((k + 1)T), etc.
equal to the sum of the transforms of the individual
functions, has been used in this calculation.
Fortunately,it is not normally necessary to have 75.5 Difference Equations
to work out the transform of a function from first
principles. Extensive tables of Z transforms exist. Just as Laplace transforms are used for solving dif-
Table 75.1 lists the more commonly used ones in ferential equations, so Z transforms are used for
process control.Table 75.2 lists the more important solving difference equations. Again there is a four
properties of Z transforms. step procedure which exploits the first two proper-
 75.5 Difference Equations 625

Table 75.2 Important properties of Z transforms 4. Find the inverse transforms, i.e. the values of
x(kT) in the left hand side of the table, corre-
x(k)k≥0 x(z)
sponding to the partial fractions. This results
(x(k)k<0 = 0)
in an equation in which the dependant variable
x1 (k) + x2 (k) x1 (z) + x2 (z) x is expressed as a function of time kT.
c.x(k) c.x(z) This procedure is best illustrated by means of a
x(k + 1) z.x(z) − z.x(0) couple of examples.
x(k + 2) z2 .x(z) − z2 .x(0) − z.x(1)
x(k + m) zm .x(z) − zm .x(0) Example 4
−zm−1 .x(1)
Find the response x(k) of the system whose be-
− . . . − z.x(m − 1)
haviour is described by the difference equation:
d
k.x(k) −z. x(z)
dz x(k + 2) + 3x(k + 1) + 2x(k) = 0
e−ak .x(k) x(z.ea )
z given the initial conditions that x(0) = 0 and
ak .x(k) x x(1) = 1.
a Transform both sides using the third and
x(∞) Lim{(z − 1).x(z)}, provided that fourth properties in Table 75.2, the forward dif-
z→1
−1 ference properties:
.x(z) is analytic on and
z
outside the unit circle z2.x(z) − z + 3z.x(z) + 2x(z) = 0
x(0) Lim x(z), provided that the
z→∞ Solve for x(z) and find partial fractions:
limit exists

∞ z z z
x(k) x(1) x(z) = = −
k=0
z2 + 3z + 2 z + 1 z + 2
n
Inverse transform both sides:
x(k).y(n − k) x(z).y(z)
k=0
x(k) = (−1)k − (−2)k for k = 0, 1, 2, 3, . . .
ties of Table 75.2. These enable the transforms, and
Example 5
their inverses, of expressions to be found one term
at a time and independently of their coefficients. Solve the difference equation:
x(k + 2) − 3x(k + 1) + 2x(k) = u(k)
1. Transform both sides of the difference equation
using the transforms of Table 75.1 as appropri- given the initial conditions that x(k) = 0 for k ≤ 0,
ate. This results in an algebraic equation with and that the forcing function u(0) = 1 and u(k) = 0
the complex variable z replacing kT as the in- for k < 0 and k > 0.
dependent variable. The initial condition x(1) will also be required
2. Solve the algebraic equation resulting from Step for the solution. This may be found by putting
1 such that the transformed variable x(z) is ex- k = −1 in the difference equation, which gives
pressed as a function of z. x(1) = 0.
3. Reduce the function x(z) resulting from Step Transform both sides:
2 into partial fractions that occur on the right
z2 .x(z) − 3z.x(z) + 2x(z) = u(z)
hand side of Table 75.1. For simple functions
this step is trivial but for complex ones it can 
∞
However, u(z) = u(kT).z−k = 1
be very tedious. k=0
 626 75 Z Transforms

Solve for x(z) and split into partial fractions to give: By inspection:

1 −1 1 x(0) = 0, x(T) = 1, x(2T) = −3, x(3T) = 7,

x(z) = 2
= +
z − 3z + 2 z − 1 z − 2 etc. which agrees with the analytical solution:
To find the inverse transform from Table 75.1 it is x(k) = (−1)k − (−2)k for k = 0, 1, 2, 3, . . .
necessary to have a z in the numerator of each par-
Note that in general it is difficult to obtain the an-
tial fraction. The forward difference property may
alytical solution x(k) from the set of values x(0),
be exploited:
x(1), x(2), x(3), etc..
Z {x(k + 1)} = z. (x(z) − x(0)) Finding the inverse by long division is often im-
practicable unless only the first few values of the
= z.x(z)
−z z time series are required, perhaps to get a feel for
= + the form of the response.Otherwise the inverse has
z−1 z−2
to be found by splitting x(z) into partial fractions.
Inverse transform both sides gives: Note the non-uniqueness of partial fractions.
Again consider the x(z) of Example 4. Direct ex-
x(k + 1) = −1 + 2k for k = 0, 1, 2, 3 . . . pansion yields the partial fractions:
z 2 1
Alternatively: x(z) = = −
z2 + 3z + 2 z + 2 z + 1
x(k) = −1 + 2k−1 for k = 1, 2, 3, 4 . . . These are of no use for inverse transformation be-
cause there are no zs in the numerators. However,
x(z)/z may be expanded into partial fractions as
follows:
75.6 Inverse Z Transformation x(z) 1 1 1
The simplest method of inverse transformation is = 2 = −
z z + 3z + 2 z + 1 z + 2
by expansion of x(z) into a time series in z −1 , which
This gives:
can be solved by inspection since:
z z
∞ x(z) = −
 z+1 z+2
x(z) = x(kT).z−k
whose partial fractions contain a z in their numer-
k=0
ator, thereby enabling the transforms in Table 75.1
= x(0) + x(T).z−1 + x(2T).z−2 to be used for inverse transformation.
+ x(3T).z−3 + . . .

Consider again the function x(z) of Example 4:

z
75.7 Worked Example
x(z) = 2
z + 3z + 2 Find x(k) for the function:

This can be rewritten as: z 1 − e−T

x(z) =
(z − 1) z − e−T
z−1
x(z) = Expanding x(z) into a convergent time series in z−1
1 + 3z−1 + 2z−2
gives:
which gives by long division:
1 − e−T .z−1
x(z) =
x(z) = z −1
− 3z −2
+ 7z −3
− 15z −4
+ ... 1 − 1 + e−T .z−1 + e−T .z−2
 75.7 Worked Example 627

Long division yields: x(z) = x(0) + x(1).z−1 + x(2).z−2 + x(3).z−3 + . . .

x(z) = 0 + 1 − e−T .z−1 + 1 − e−2T .z−2
which, by inspection gives:
+ 1 − e−3T .z−3 + . . .
This may be compared with Equation 75.2: x(k) = 1 − e−kT