ELE8311: The z-Transform

Dr. Hassan A. Bashir

ELE8311: Digital Control Engineering

2.0 Module 2: The z-Transform

2.1 Difference Equation

Suppose 𝑦𝑦(𝑘𝑘) is a discrete-time signal, where the independent variable (𝑘𝑘), usually time, is
assumed to have a discrete set of possible values, then a difference equation can be used to
represent the system shown in Fig.2.1 as in (2.1).

𝑦𝑦(𝑘𝑘 + 𝑛𝑛) + 𝑎𝑎𝑛𝑛−1 𝑦𝑦(𝑘𝑘 + 𝑛𝑛 − 1) + ⋯ + 𝑎𝑎1 𝑦𝑦(𝑘𝑘 + 1) + 𝑎𝑎0 𝑦𝑦(𝑘𝑘)

= 𝑏𝑏𝑛𝑛 𝑢𝑢(𝑘𝑘 + 𝑛𝑛) + 𝑏𝑏𝑛𝑛−1 𝑢𝑢(𝑘𝑘 + 𝑛𝑛 − 1) + ⋯ + 𝑏𝑏1 𝑢𝑢(𝑘𝑘 + 1) + 𝑢𝑢(𝑘𝑘) (2.1)

Fig. 2.1: Discrete System Block Diagram

where, 𝑢𝑢(𝑘𝑘) = 𝑢𝑢(𝑘𝑘𝑘𝑘), 𝑘𝑘 = 0,1,2, … is an input sequence (also called a forcing function), and
𝑦𝑦(𝑘𝑘) = 𝑦𝑦(𝑘𝑘𝑘𝑘) is an output sequence (also called the response). A difference equation, such
as (2.1), is said to be of order 𝑛𝑛 because the difference between the highest and lowest time
arguments of 𝑦𝑦(. ) and 𝑢𝑢(. ) is 𝑛𝑛 . If the coefficients 𝑎𝑎𝑖𝑖 , 𝑏𝑏𝑖𝑖 | 𝑖𝑖 = 0, 1, 2, …, are constant, the
difference equation is referred to as linear time invariant (LTI). If the forcing function 𝑢𝑢(𝑘𝑘) is
equal to zero, the equation is said to be homogeneous.

1
 ELE8311: The z-Transform

Exercise – Identifying a Difference Equation

Example: Determine the order of the following difference equations. Identify whether the
equation is: (i) linear, (ii) time invariant, or (iii) homogenous.

(a) 𝑦𝑦(𝑘𝑘 + 2) + 0.8𝑦𝑦(𝑘𝑘 + 1) + 0.07𝑦𝑦(𝑘𝑘) = 𝑢𝑢(𝑘𝑘)

(b) 𝑦𝑦(𝑘𝑘 + 4) + sin(0.4𝑘𝑘) 𝑦𝑦(𝑘𝑘 + 1) + 0.3𝑦𝑦(𝑘𝑘) = 0
(c) 𝑦𝑦(𝑘𝑘 + 1) = −0.1𝑦𝑦 2 (𝑘𝑘)

Solution:

(a) The equation is second order. All terms are linear with constant coefficients, thus it is
LTI. There is a forcing function in the equation, thus, it is non-homogeneous.
(b) The difference equation is fourth order. The second coefficient is time dependent, but
all terms are linear. It is therefore a linear time varying, homogeneous equation.
(c) This is a first order difference equation with a nonlinear term at the right hand side. It
has no forcing function. The equation is therefore nonlinear, time invariant, and
homogeneous.

Example:

Consider the LTI difference equation (2.2), find its response due to the initial conditions
𝑦𝑦(−2) = 1, 𝑦𝑦(−1) = 2 under a unit step input sequence 𝑢𝑢(𝑘𝑘) = 1, for = 0,1,2, … .

3𝑦𝑦(𝑘𝑘) + 2𝑦𝑦(𝑘𝑘 − 1) − 𝑦𝑦(𝑘𝑘 − 2) = 2𝑢𝑢(𝑘𝑘 − 1) − 3𝑢𝑢(𝑘𝑘 − 2) (2.2)

Solution:

We first re-write equation (2.2) as follows:

1
𝑦𝑦(𝑘𝑘) = [−2𝑦𝑦(𝑘𝑘 − 1) + 𝑦𝑦(𝑘𝑘 − 2) + 2𝑢𝑢(𝑘𝑘 − 1) − 3𝑢𝑢(𝑘𝑘 − 2)]
3

Then, the response for the given initial conditions and forcing function is obtained recursively
as follows:

1 5
𝑦𝑦(0) = [−2𝑦𝑦(−1) + 𝑦𝑦(−2) + 2𝑢𝑢(−1) − 3𝑢𝑢(−2)] =
3 3
1 10
𝑦𝑦(1) = [−2𝑦𝑦(0) + 𝑦𝑦(−1) + 2𝑢𝑢(0) − 3𝑢𝑢(−1)] = −
3 9
1 26
𝑦𝑦(2) = [−2𝑦𝑦(1) + 𝑦𝑦(0) + 2𝑢𝑢(1) − 3𝑢𝑢(0)] =
3 27
2
 ELE8311: The z-Transform

Note that this course will focus on LTI difference equations and the z-transform is a
convenient approach for solving LTI problems.

2.2 z-Transform

z-transform is a vital tool for analysis and deisgn of discrete-time systems. Similar to the
Laplace transforms in continuous domain, z-transform converts LTI difference equations to
algebraic equations and convolutions to multiplications.

Def. 2.1:

For a causal signal (𝑘𝑘 ≥ 0) defined by the sequence 𝑓𝑓(𝑘𝑘) = {𝑢𝑢0 , 𝑢𝑢1 , 𝑢𝑢2 , … , 𝑢𝑢𝑘𝑘 , … } , the z-
transform is defined as in (2.2):
∞

𝐹𝐹(𝑧𝑧) = 𝑧𝑧[𝑓𝑓(𝑘𝑘)] = 𝑢𝑢0 + 𝑢𝑢1 𝑧𝑧 −1 + 𝑢𝑢2 𝑧𝑧 −2 + ⋯ + 𝑢𝑢𝑘𝑘 𝑧𝑧 −𝑘𝑘 = � 𝑢𝑢𝑘𝑘 𝑧𝑧 −𝑘𝑘

𝑘𝑘=0

= � 𝑓𝑓(𝑘𝑘)𝑧𝑧 −𝑘𝑘 (2.2)

𝑘𝑘=0

To express 𝐹𝐹(𝑧𝑧) explicitly, we write:

𝐹𝐹(𝑧𝑧) = 𝑓𝑓(0) + 𝑓𝑓(1)𝑧𝑧 −1 + 𝑓𝑓(2)𝑧𝑧 −2 + 𝑓𝑓(3)𝑧𝑧 −3 + ⋯

where, 𝑧𝑧 −𝑖𝑖 indicates the 𝑖𝑖th sampling instant. Thus, 𝑧𝑧 0 corresponds to the initial time 𝑘𝑘 = 0;
𝑧𝑧 −1 indicates the time instant at 𝑘𝑘 = 1. Hence, 𝑧𝑧 −1 is called the unit-delay element.

Def. 2.2:

Alternatively, the z-transform of a given function can be obtained from its Laplace transforms
as follows.

Consider the following impulse train representation of a discrete-time signal:

∞
∗ (𝑡𝑡)
𝑢𝑢 = � 𝑢𝑢𝑘𝑘 𝛿𝛿(𝑡𝑡 − 𝑘𝑘𝑘𝑘).
𝑘𝑘=0

In Laplace transform, we have:

3
 ELE8311: The z-Transform

∞
∗ (𝑠𝑠) −𝑠𝑠𝑠𝑠 −2𝑠𝑠𝑠𝑠 −𝑘𝑘𝑘𝑘𝑘𝑘
𝑈𝑈 = 𝑢𝑢0 + 𝑢𝑢1 𝑒𝑒 + 𝑢𝑢2 𝑒𝑒 + ⋯ + 𝑢𝑢𝑘𝑘 𝑒𝑒 + ⋯ = � 𝑢𝑢𝑘𝑘 (𝑒𝑒 −𝑠𝑠𝑠𝑠 )𝑘𝑘 ,
𝑘𝑘=0

setting 𝑧𝑧 = 𝑒𝑒 𝑠𝑠𝑠𝑠 yields similar results as in equation (2.2) above.

2.3 Standard Discrete-time Signals

The commonly used excitation signals include discrete-time impulse, sampled step,
exponential and sinusoidal signals. The following identities are essential.

Useful identities:
𝑛𝑛
1 − 𝑎𝑎𝑛𝑛 +1
� 𝑎𝑎𝑘𝑘 = ; 𝑎𝑎 ≠ 1
1 − 𝑎𝑎
𝑘𝑘=0

∞
1
� 𝑎𝑎𝑘𝑘 = ; |𝑎𝑎| < 1
1 − 𝑎𝑎
𝑘𝑘=0

a. Unit Impulse Signal

Fig. 2.2: Unit Impulse Signal

Suppose the discrete-time signal shown in Fig. 2.2 is defined by:

1, 𝑘𝑘 = 0
𝑢𝑢(𝑘𝑘) = 𝛿𝛿(𝑘𝑘) = �
0, 𝑘𝑘 ≠ 0

The z-transform of this signal as defined in (2.2) is:

𝑧𝑧[𝑢𝑢(𝑘𝑘)] = 𝑈𝑈(𝑧𝑧) = 1.
4
 ELE8311: The z-Transform

b. Unit Step Signal

Let the sequence {𝑢𝑢𝑘𝑘 }∞

𝑘𝑘=0 = {1,1,1, … } be a sampled step signal. The z-transform is defined as:

∞
−1 −2 −3 −𝑘𝑘
𝑈𝑈(𝑧𝑧) = 1 + 𝑧𝑧 + 𝑧𝑧 + 𝑧𝑧 + ⋯ + 𝑧𝑧 + ⋯ = � 𝑧𝑧 −𝑘𝑘 .
𝑘𝑘=0

Fig. 2.3: Unit Step Signal

Recall that the above series have the following closed-form:

1 𝑧𝑧
𝑈𝑈(𝑧𝑧) = −1
=
1 − 𝑧𝑧 𝑧𝑧 − 1

c. Exponential Signal

Let a sampled exponential signal be defined by:

𝑎𝑎𝑘𝑘 , 𝑘𝑘 ≥ 0
𝑢𝑢(𝑘𝑘) = �
0, 𝑘𝑘 < 0

Then,

𝑧𝑧[𝑈𝑈(𝑘𝑘)] = 𝑈𝑈(𝑧𝑧) = 1 + 𝑎𝑎𝑧𝑧 −1 + 𝑎𝑎2 𝑧𝑧 −2 + ⋯ + 𝑎𝑎𝑘𝑘 𝑧𝑧 −𝑘𝑘 + ⋯.

Similarly, this has a closed-form:

1 𝑧𝑧
𝑈𝑈(𝑧𝑧) = = .
1 − 𝑎𝑎𝑧𝑧 −1 𝑧𝑧 − 𝑎𝑎

5
 ELE8311: The z-Transform

Fig. 2.4: Unit Exponential Signal

z-Transforms – summary

Table 2.1: Summary of z-Transform

6
 ELE8311: The z-Transform

Table 2.2: Derived z-Transform

2.4 Properties of the z-transform

a. Convergence of z-transform

The existence of a unique z-transform depends on the availability of its region of absolute
convergence. Recall that for an impulse signal 𝛿𝛿(𝑛𝑛), let:

𝑢𝑢1 (𝑛𝑛) = 𝑎𝑎𝑛𝑛 1(𝑛𝑛)

𝑢𝑢2 (𝑛𝑛) = −𝑎𝑎𝑛𝑛 1(−𝑛𝑛 − 1)

Then, for the first sequence:

∞ ∞
𝑛𝑛 −𝑛𝑛
1 𝑧𝑧
𝑈𝑈1 (𝑧𝑧) = � 𝑎𝑎 𝑧𝑧 = �(𝑎𝑎𝑧𝑧 −1 )𝑛𝑛 = −1
=
1 − 𝑎𝑎𝑧𝑧 𝑧𝑧 − 𝑎𝑎
𝑛𝑛=0 𝑛𝑛 =0

7
 ELE8311: The z-Transform

if the sum exists, i.e., 𝑎𝑎𝑧𝑧 −1 > 1, (Condition 1)

For the second sequence,

∞ −∞

𝑈𝑈2 (𝑧𝑧) = − � 𝑎𝑎𝑛𝑛 1(−𝑛𝑛 − 1)𝑧𝑧 −𝑛𝑛 = − � 𝑎𝑎𝑛𝑛 𝑧𝑧 −𝑛𝑛

𝑛𝑛 =−∞ 𝑛𝑛=−1

Let 𝑚𝑚 = −𝑛𝑛, we have

∞ ∞
−𝑚𝑚 𝑚𝑚
𝑈𝑈2 (𝑧𝑧) = − � 𝑎𝑎 𝑧𝑧 = 1 − � (𝑎𝑎−1 𝑧𝑧)𝑚𝑚 .
𝑚𝑚 =1 𝑚𝑚 =0

Therefore,

1 𝑧𝑧
𝑈𝑈2 (𝑧𝑧) = 1 − =
1 − 𝑎𝑎−1 𝑧𝑧 𝑧𝑧 − 𝑎𝑎

If the sum exists, i.e., 𝑎𝑎−1 𝑧𝑧 < 1, (Condition 2)

Notice that, although the two sequences 𝑢𝑢1 and 𝑢𝑢2 yield the same z-transform, the
conditions upon which their regions of convergence exist are completely different. Since z-
transform are defined in terms of polynomials in z, which can be real or complex, the
complete definition for case 1 and 2 are as follows:
𝑧𝑧
𝑢𝑢1 (𝑛𝑛) ↔ , |𝑎𝑎𝑧𝑧 −1 | < 1 𝑜𝑜𝑜𝑜 |𝑧𝑧| > |𝑎𝑎|
𝑧𝑧 − 𝑎𝑎
𝑧𝑧
𝑢𝑢2 (𝑛𝑛) ↔ , |𝑎𝑎−1 𝑧𝑧| < 1 𝑜𝑜𝑜𝑜 |𝑧𝑧| < |𝑎𝑎|
𝑧𝑧 − 𝑎𝑎

Fig. 2.4: Region of Convergence of 𝑢𝑢1 (𝑛𝑛) (left) and 𝑢𝑢2 (𝑛𝑛) (right).

8
 ELE8311: The z-Transform

Example 2.1:

Suppose a signal 𝑢𝑢(𝑘𝑘) consist of a sum of two real exponentials:

1 𝑘𝑘 1 𝑘𝑘
𝑢𝑢(𝑘𝑘) = � � 1(𝑘𝑘) + � � 1(𝑘𝑘)
2 3

From the definition of z-transform, the z-transform of 𝑢𝑢(𝑘𝑘) is:

∞
1 𝑘𝑘 1 𝑘𝑘
𝑈𝑈(𝑧𝑧) = � �� � 1(𝑘𝑘) + � � 1(𝑘𝑘)� 𝑧𝑧 −𝑘𝑘
2 3
𝑘𝑘=−∞

∞ ∞
1 𝑘𝑘 1 𝑘𝑘
= � � � 1(𝑘𝑘)𝑧𝑧 + � � � 1(𝑘𝑘)𝑧𝑧 −𝑘𝑘
−𝑘𝑘
2 3
𝑘𝑘=−∞ 𝑘𝑘=−∞

∞ ∞
1 𝑘𝑘 −𝑘𝑘 1 𝑘𝑘 −𝑘𝑘
= � � � 𝑧𝑧 + � � � 𝑧𝑧
2 3
𝑘𝑘=0 𝑘𝑘=0

Now, for the first term:

∞
1 𝑘𝑘 1 1
� � � 𝑧𝑧 −𝑘𝑘 = ∶ |𝑧𝑧| > ,
2 1 2
𝑘𝑘=0 1 − 2 𝑧𝑧 −1

For the second term:

∞
1 𝑘𝑘 −𝑘𝑘 1 1
� � � 𝑧𝑧 = ∶ |𝑧𝑧| >
3 1 3
𝑘𝑘=0 1 − 3 𝑧𝑧 −1

Notice that both regions of convergence intersect. Thus, at |𝑧𝑧| > 1/2 both sums converge
(i.e., their z-transform exist).

Example 2.2:

Find the z-transform of {𝑢𝑢(𝑘𝑘)} = {1(𝑘𝑘)} + {0.5𝑘𝑘 1(−𝑘𝑘)}

∞ ∞
−𝑘𝑘
𝑈𝑈(𝑧𝑧) = � 1(𝑘𝑘)𝑧𝑧 + � 0.5𝑘𝑘 1(−𝑘𝑘)𝑧𝑧 −𝑘𝑘
𝑘𝑘=−∞ 𝑘𝑘=−∞

∞ 0

= � 𝑧𝑧 −𝑘𝑘 + � 0.5𝑘𝑘 𝑧𝑧 −𝑘𝑘

𝑘𝑘=−∞ 𝑘𝑘=−∞
9
 ELE8311: The z-Transform

For the first term, the z-transform is 𝑧𝑧/(𝑧𝑧 − 1) for |𝑧𝑧| > 1. For the second term, however, we
have 0.5/(0.5 − 𝑧𝑧) for |𝑧𝑧| < 0.5.

Now, since the regions of convergence for the first and second term (|𝑧𝑧| > 1 and |𝑧𝑧| < 0.5)
do not intersect (are mutually exclusive), the z-transform of this example does not exist (no
convergence).

Note that we are only interested on signals which have their z-transform defined.

b. Linearity

Similar to linearity property of Laplace transforms in the continuous domain, we have:

𝑍𝑍[𝛼𝛼𝑓𝑓1 (𝑘𝑘) + 𝛽𝛽𝑓𝑓2 (𝑘𝑘)] = 𝛼𝛼𝐹𝐹1 (𝑧𝑧) + 𝛽𝛽𝐹𝐹2 (𝑧𝑧)

That is, it obeys both superposition and homogeneity.

Example: given that: 𝑓𝑓(𝑘𝑘) = 2 × 1(𝑘𝑘) + 4𝛿𝛿(𝑘𝑘), 𝑘𝑘 = 0,1,2, …

The z-transform is obtained as follows:

2𝑧𝑧 6𝑧𝑧 − 4
𝑍𝑍[𝑓𝑓(𝑘𝑘)] = 𝐹𝐹(𝑧𝑧) = 2𝑍𝑍[1(𝑘𝑘)] + 4𝑍𝑍[𝛿𝛿(𝑘𝑘)] = +4=
𝑧𝑧 − 1 𝑧𝑧 − 1

c. Time Delay

This also follows from the time delay property of the Laplace transform:

𝑍𝑍[𝑓𝑓(𝑘𝑘 − 𝑛𝑛)] = 𝑧𝑧 −𝑛𝑛 𝐹𝐹(𝑧𝑧)

Example 2.3

Find the z-transform of the causal sequence

4, 𝑘𝑘 = 2,3, …
𝑓𝑓(𝑘𝑘) = �
0, otherwise

This is a sequence that is delayed to start at 𝑘𝑘 = 2 rather than 𝑘𝑘 = 0 . Using the delay
property above:

4𝑧𝑧 4
𝐹𝐹(𝑧𝑧) = 𝑧𝑧{4 × 1(𝑘𝑘 − 2)} = 4𝑧𝑧 −2 𝑧𝑧{1(𝑘𝑘)} = 𝑧𝑧 −2 =
𝑧𝑧 − 1 𝑧𝑧(𝑧𝑧 − 1)

10
 ELE8311: The z-Transform

Example 2.4:

Find the z-transform of

𝑢𝑢(𝑘𝑘) = 2𝛿𝛿(𝑘𝑘) − 3𝛿𝛿(𝑘𝑘 − 2) + 4𝛿𝛿(𝑘𝑘 − 5)

Solution:
∞ ∞ ∞
−𝑘𝑘 −𝑘𝑘
𝑧𝑧[𝑢𝑢(𝑘𝑘)] = 𝑈𝑈(𝑧𝑧) = 2 � 𝛿𝛿(𝑘𝑘)𝑧𝑧 − 3 � 𝛿𝛿(𝑘𝑘 − 2)𝑧𝑧 + 4 � 𝛿𝛿(𝑘𝑘 − 5)𝑧𝑧 −𝑘𝑘
𝑘𝑘=−∞ 𝑘𝑘=−∞ 𝑘𝑘=−∞

2𝑧𝑧 5 − 3𝑧𝑧 3 + 4
= 2 − 3𝑧𝑧 −2 + 4𝑧𝑧 −5 = ; |𝑧𝑧| > 0
𝑧𝑧 5

Example 2.5

Find the z-transform of

𝑢𝑢(𝑘𝑘) = [2 + 4(−3)𝑘𝑘 ]{1(𝑘𝑘)}

∞
2𝑧𝑧 4𝑧𝑧
𝑈𝑈(𝑧𝑧) = �[2 + 4(−3)𝑘𝑘 ]𝑧𝑧 −𝑘𝑘 = + ; |𝑧𝑧| > 1 ∩ |𝑧𝑧| > 3
𝑧𝑧 − 1 𝑧𝑧 + 3
𝑘𝑘=0

Therefore,

6𝑧𝑧 2 + 2𝑧𝑧
𝑈𝑈(𝑧𝑧) = ; |𝑧𝑧| > 3
(𝑧𝑧 − 1)(𝑧𝑧 + 3)

d. Time Advance

𝑧𝑧{𝑓𝑓(𝑘𝑘 + 1)} = 𝑧𝑧𝑧𝑧(𝑧𝑧) − 𝑧𝑧𝑧𝑧(0)

𝑧𝑧{𝑓𝑓(𝑘𝑘 + 𝑛𝑛)} = 𝑧𝑧 𝑛𝑛 𝐹𝐹(𝑧𝑧) − 𝑧𝑧 𝑛𝑛 𝑓𝑓(0) − 𝑧𝑧 𝑛𝑛 −1 𝑓𝑓(1) − ⋯ − 𝑧𝑧𝑧𝑧(𝑛𝑛 − 1)

Example 2.6:

Find the z-transform of the following sequence

{𝑓𝑓(𝑘𝑘)} = {4,8,16, … }

Solution:

The sequence can be rewritten as:

𝑓𝑓(𝑘𝑘) = 2𝑘𝑘+2 = 𝑔𝑔(𝑘𝑘 + 2), 𝑘𝑘 = 0,1,2, …

11
 ELE8311: The z-Transform

Using the advance property, we have:

𝑧𝑧 4𝑧𝑧
𝐹𝐹(𝑧𝑧) = 𝑧𝑧 2 𝐺𝐺(𝑧𝑧) − 𝑧𝑧 2 𝑔𝑔(0) − 𝑧𝑧𝑧𝑧(1) = 𝑧𝑧 2 − 𝑧𝑧 2 − 2𝑧𝑧 =
𝑧𝑧 − 2 𝑧𝑧 − 2

Notice that, the solution can also be obtained by rewriting the sequence as:

{𝑓𝑓(𝑘𝑘)} = 4{1,2,4, … }

and using the linearity property:

4𝑧𝑧
𝑧𝑧 − 2

e. Multiplication by Exponent

𝑧𝑧{𝑎𝑎−𝑘𝑘 𝑓𝑓(𝑘𝑘)} = 𝐹𝐹(𝑎𝑎𝑎𝑎)

Example 2.7:

Find the z-transform of cos 𝑤𝑤𝑤𝑤 1(𝑘𝑘) and sin 𝑤𝑤𝑤𝑤 1(𝑘𝑘).

Solution: Using Euler’s identity, it is convenient to benefit from the simplicity of sequences of
the form 𝑎𝑎𝑘𝑘 .

𝑒𝑒 𝑗𝑗𝑗𝑗𝑗𝑗 = cos 𝑤𝑤𝑤𝑤 + 𝑗𝑗 sin 𝑤𝑤𝑤𝑤

Taking the z-transform of both sides yields:

∞ ∞
𝑗𝑗𝑗𝑗𝑗𝑗 𝑗𝑗𝑗𝑗𝑗𝑗 −𝑘𝑘
𝑧𝑧�𝑒𝑒 1(𝑘𝑘)� = � 𝑒𝑒 𝑧𝑧 1(𝑘𝑘) = � 𝑒𝑒 𝑗𝑗𝑗𝑗𝑗𝑗 𝑧𝑧 −𝑘𝑘
𝑘𝑘=−∞ 𝑘𝑘=0

𝑧𝑧
= ; |𝑧𝑧| > �𝑒𝑒 𝑗𝑗𝑗𝑗 � = 1
𝑧𝑧 − 𝑒𝑒 𝑗𝑗𝑗𝑗

𝑧𝑧�𝑧𝑧 − 𝑒𝑒 −𝑗𝑗𝑗𝑗 �
=
(𝑧𝑧 − 𝑒𝑒 𝑗𝑗𝑗𝑗 )(𝑧𝑧 − 𝑒𝑒 −𝑗𝑗𝑗𝑗 )

12
 ELE8311: The z-Transform

Fig. 2.4: Poles and Zeros of cos 𝑤𝑤𝑤𝑤 1(𝑘𝑘) (left) and sin 𝑤𝑤𝑤𝑤 1(𝑘𝑘) (right).

𝑧𝑧(𝑧𝑧 − cos 𝑤𝑤 + 𝑗𝑗 sin 𝑤𝑤)

=
(𝑧𝑧 − 𝑒𝑒 𝑗𝑗𝑗𝑗 )(𝑧𝑧 − 𝑒𝑒 −𝑗𝑗𝑗𝑗 )

𝑧𝑧(𝑧𝑧 − cos 𝑤𝑤) 𝑧𝑧 sin 𝑤𝑤

= + 𝑗𝑗
(𝑧𝑧 − 𝑒𝑒 )(𝑧𝑧 − 𝑒𝑒
𝑗𝑗𝑗𝑗 −𝑗𝑗𝑗𝑗 ) (𝑧𝑧 − 𝑒𝑒 )(𝑧𝑧 − 𝑒𝑒 −𝑗𝑗𝑗𝑗 )
𝑗𝑗𝑗𝑗

By equating the real and imaginary terms, we have:

𝑧𝑧(𝑧𝑧 − cos 𝑤𝑤)

(cos 𝑤𝑤𝑤𝑤)1(𝑘𝑘) ↔
(𝑧𝑧 − 𝑒𝑒 𝑗𝑗𝑗𝑗 )(𝑧𝑧 − 𝑒𝑒 −𝑗𝑗𝑗𝑗 )

𝑧𝑧 sin 𝑤𝑤
(sin 𝑤𝑤𝑤𝑤)1(𝑘𝑘) ↔
(𝑧𝑧 − 𝑒𝑒 𝑗𝑗𝑗𝑗 )(𝑧𝑧
− 𝑒𝑒 −𝑗𝑗𝑗𝑗 )

f. Effect of Damping

The presence of an exponential in time sequence results in the following expression for its z-
transform:

𝑧𝑧[𝑢𝑢(𝑘𝑘)𝑒𝑒 −𝑎𝑎𝑎𝑎 ] = 𝑈𝑈(𝑧𝑧𝑒𝑒 𝑎𝑎 )

That is, U evaluated at 𝑧𝑧𝑒𝑒 𝑎𝑎 .

From the definition,

∞ ∞
−𝑎𝑎𝑎𝑎 −𝑘𝑘
� 𝑢𝑢(𝑘𝑘)𝑒𝑒 𝑧𝑧 = � 𝑢𝑢(𝑘𝑘)(𝑒𝑒 𝑎𝑎 𝑧𝑧)−𝑘𝑘 = 𝑈𝑈(𝑧𝑧𝑒𝑒 𝑎𝑎 )
𝑘𝑘=−∞ 𝑘𝑘=−∞

13
 ELE8311: The z-Transform

g. Initial and Final Value Theorems for Causal Signals

If the limit exists, the z-transform definition of the initial value of a causal signal is obtained
as:

𝑢𝑢(0) = lim 𝑈𝑈(𝑧𝑧)

z→∞

From the definition of causal signals, we have

∞

𝑈𝑈(𝑧𝑧) = � 𝑢𝑢(𝑘𝑘)𝑧𝑧 −𝑘𝑘 = 𝑢𝑢(0) + 𝑢𝑢(1)𝑧𝑧 −1 + 𝑢𝑢(2)𝑧𝑧 −2 + ⋯

𝑘𝑘=−∞

Taking the limit yields:

lim 𝑈𝑈(𝑧𝑧) = 𝑢𝑢(0).

𝑧𝑧→∞

Similarly, if the limit exists, the final value of causal signals is defined as:

𝑧𝑧 − 1
lim 𝑢𝑢(𝑘𝑘) = lim(1 − 𝑧𝑧 −1 )𝑈𝑈(𝑧𝑧) = lim 𝑈𝑈(𝑧𝑧)
𝑘𝑘→∞ 𝑧𝑧→1 𝑧𝑧→1 𝑧𝑧

The above is true if 𝑈𝑈(𝑧𝑧) converges for all |𝑧𝑧| > 1 and if all poles of 𝑈𝑈(𝑧𝑧)(𝑧𝑧 − 1) are inside
the unit circle. The only possible pole of 𝑈𝑈(𝑧𝑧) not strictly inside the unit circle is a simple pole
at 𝑧𝑧 = 𝑎𝑎, which is removed in (𝑧𝑧 − 1)𝑈𝑈(𝑧𝑧).

Example 2.8:

Using the final value theorem, calculate the steady state value of (0.5𝑘𝑘 − 0.5)1(𝑘𝑘). Verify that
this agrees with the direct result.

Solution

The z-transform pair are:

𝑧𝑧 0.5𝑧𝑧
(0.5𝑘𝑘 − 0.5)1(𝑘𝑘) ↔ − , |𝑧𝑧| > 1.
𝑧𝑧 − 0.5 𝑧𝑧 − 1

From the LHS, the steady state value is:

lim (0.5𝑘𝑘 − 0.5)1(𝑘𝑘) → −0.5

𝑘𝑘→∞

Also, from the RHS, the steady state value is:

𝑧𝑧 0.5𝑧𝑧
lim � − � (𝑧𝑧 − 1)/𝑧𝑧 → −0.5.
𝑧𝑧→1 𝑧𝑧 − 0.5 𝑧𝑧 − 1

14
 ELE8311: The z-Transform

Example 2.9

Does the final value theorem applies to this sequence 2𝑘𝑘 1(𝑘𝑘)?

Solution:
𝑧𝑧
2𝑘𝑘 1(𝑘𝑘) ↔ , |𝑧𝑧| > 2
𝑧𝑧 − 2

Notice that, from the LHS, as 𝑘𝑘 → ∞ we have a growing sequence. Also, from the RHS, we
have:
𝑧𝑧
lim (𝑧𝑧 − 1)/𝑧𝑧
𝑧𝑧→1 𝑧𝑧 − 2

Thus, since the RHS is valid only for |𝑧𝑧| > 2, the theorem cannot even be applied. Hence, the
final value for causal signals cannot be found for growing sequences.

Example 2.10

Find the initial value and final value of a causal sequence whose z-transform is given by:

0.792𝑧𝑧 2
𝑈𝑈(𝑧𝑧) = .
(𝑧𝑧 − 1)(𝑧𝑧 2 − 0.416𝑧𝑧 + 0.208)

Solution

The initial value is

0.792𝑧𝑧 2
lim 𝑈𝑈(𝑧𝑧) = = 0.
𝑧𝑧→∞ 𝑧𝑧 3

The final value is

0.792
lim(𝑧𝑧 − 1)𝑈𝑈(𝑧𝑧) = = 1.
𝑧𝑧→1 1 − 0.416 + 0.208

Exercise: Verify the above initial and final values by expanding the expression for 𝑈𝑈(𝑧𝑧) by
long division.

15
 ELE8311: The z-Transform

h. Convolution

If 𝑈𝑈(𝑧𝑧) and 𝐺𝐺(𝑧𝑧) are z-transforms of {𝑢𝑢(𝑛𝑛)} and {𝑔𝑔(𝑛𝑛)}, respectively, then 𝑈𝑈(𝑧𝑧)𝐺𝐺(𝑧𝑧) is the z-
transform of {𝑢𝑢(𝑛𝑛)} ∗ {𝑔𝑔(𝑛𝑛)}.

In other words, if 𝑢𝑢(𝑛𝑛) ↔ 𝑈𝑈(𝑧𝑧) and 𝑔𝑔(𝑛𝑛) ↔ 𝐺𝐺(𝑧𝑧), then

𝑔𝑔(𝑛𝑛) ∗ 𝑢𝑢(𝑛𝑛) ↔ 𝐺𝐺(𝑧𝑧)𝑈𝑈(𝑧𝑧).

Let 𝑦𝑦(𝑛𝑛) = 𝑔𝑔(𝑛𝑛) ∗ 𝑢𝑢(𝑛𝑛). By convolution, we have:

∞

𝑦𝑦(𝑛𝑛) = � 𝑔𝑔(𝑘𝑘)𝑢𝑢(𝑛𝑛 − 𝑘𝑘)

𝑘𝑘=−∞

Taking the z-transform yields:

∞ ∞ ∞ ∞
−𝑛𝑛 −𝑘𝑘
𝑌𝑌(𝑧𝑧) = � � 𝑔𝑔(𝑘𝑘)𝑢𝑢(𝑛𝑛 − 𝑘𝑘)𝑧𝑧 = � 𝑔𝑔(𝑘𝑘)𝑧𝑧 � 𝑢𝑢(𝑛𝑛 − 𝑘𝑘)𝑧𝑧 −(𝑛𝑛−𝑘𝑘)
𝑛𝑛=−∞ 𝑘𝑘=−∞ 𝑘𝑘=−∞ 𝑛𝑛=−∞

∴ 𝑌𝑌(𝑧𝑧) = 𝐺𝐺(𝑧𝑧)𝑈𝑈(𝑧𝑧).

Similarly, for a causal signal

∞

𝑦𝑦(𝑛𝑛) = � 𝑔𝑔(𝑘𝑘)𝑢𝑢(𝑛𝑛 − 𝑘𝑘)

𝑘𝑘=0

Taking the z-transform,

∞ ∞

𝑌𝑌(𝑧𝑧) = � � 𝑔𝑔(𝑘𝑘) 𝑢𝑢(𝑛𝑛 − 𝑘𝑘)𝑧𝑧 −𝑛𝑛

𝑛𝑛=0 𝑘𝑘=0

∞ ∞
−𝑘𝑘
= � 𝑔𝑔(𝑘𝑘)𝑧𝑧 � 𝑢𝑢(𝑛𝑛 − 𝑘𝑘)𝑧𝑧 −(𝑛𝑛−𝑘𝑘) ,
𝑘𝑘=0 𝑛𝑛=0

with 𝑚𝑚 = 𝑛𝑛 − 𝑘𝑘, we have

∞ ∞

𝑌𝑌(𝑧𝑧) = � 𝑔𝑔(𝑘𝑘)𝑧𝑧 −𝑘𝑘 � 𝑢𝑢(𝑚𝑚)𝑧𝑧 −𝑚𝑚 .

𝑘𝑘=0 𝑚𝑚 =0

16
 ELE8311: The z-Transform

Example 2.11

Determine the step response of a system with impulse response is 𝑔𝑔(𝑛𝑛) = 0.5𝑛𝑛 1(𝑛𝑛).

Solution

Using the z-transform approach, we simply multiply the signal 𝑢𝑢 and the step input 𝑔𝑔 as
follows:
∞ ∞
−𝑗𝑗
1
𝑈𝑈(𝑧𝑧) = � 𝑢𝑢(𝑗𝑗)𝑧𝑧 = � 0.5𝑗𝑗 𝑧𝑧 −𝑗𝑗 = , |𝑧𝑧| > 0.5
1 − 0.5𝑧𝑧 −1
𝑗𝑗 =−∞ 𝑗𝑗 =0

We also have,
∞ ∞
−𝑗𝑗
1
𝐺𝐺(𝑧𝑧) = � 𝑔𝑔(𝑗𝑗)𝑧𝑧 = � 𝑧𝑧 −𝑗𝑗 = , |𝑧𝑧| > 1
1 − 𝑧𝑧 −1
𝑗𝑗 =−∞ 𝑗𝑗 =0

On multiplying, we have:

1 1
𝑌𝑌(𝑧𝑧) = 𝐺𝐺(𝑧𝑧)𝑈𝑈(𝑧𝑧) = , |𝑧𝑧| > 1.
1 − 0.5𝑧𝑧 1 − 𝑧𝑧 −1
−1

i. Complex Differentiation

𝑚𝑚
𝑑𝑑 𝑚𝑚
𝑧𝑧{𝑘𝑘 𝑓𝑓(𝑘𝑘)} = �−𝑧𝑧 � 𝐹𝐹(𝑧𝑧)
𝑑𝑑𝑑𝑑

Example 2.12

Find the z-transform of the sampled ramp sequence

𝑓𝑓(𝑘𝑘) = 𝑘𝑘, 𝑘𝑘 = 0,1,2, …

Solution

Recall that the z-transform of a step signal is simply,

𝑧𝑧
𝐹𝐹(𝑧𝑧) =
𝑧𝑧 − 1

But, 𝑓𝑓(𝑘𝑘) = 𝑘𝑘 × 1, 𝑘𝑘 = 0,1,2, …

Then applying the complex differentiation property yields

17
 ELE8311: The z-Transform

𝑑𝑑 𝑧𝑧 (−𝑧𝑧)�(𝑧𝑧 − 1) − 𝑧𝑧� 𝑧𝑧
𝑧𝑧{𝑘𝑘 × 1} = �−𝑧𝑧 �� �= =
𝑑𝑑𝑑𝑑 𝑧𝑧 − 1 (𝑧𝑧 − 1) 2 (𝑧𝑧 − 1)2

18
ELE8311: The z-Transform

19