The Z-Transform

EE 453 / CE 352
Saad Baig

Slides adapted from Dr. Muhammad Umer Tariq’s offering of DSP in 2021
Z-Transform
• Analysis and design of DSP systems require the capability to analyze
and design Discrete-Time systems.
• We focus on LTI Discrete-Time systems due to their mathematical
tractability.
• Analysis of LTI DT techniques:
− Time-domain techniques (such as convolution)
− Transform-domain techniques (such as Fourier Transform)

EE 453 / CE 352 - Digital Signal Processing 2

Z-Transform
• Why transform-domain analysis?
• Analysis simplification.
• E.g. convolution of two time-domain signals are equivalent to multiplication
of their corresponding transforms.
• Z-transform provides us additional means of characterizing an LTI
system.

EE 453 / CE 352 - Digital Signal Processing 3

Z-Transform
• Direct z-Transform:
∞

𝑋 𝑧 = ෍ 𝑥(𝑛)𝑧 −𝑛
𝑛=−∞
• Notation:
𝑋(𝑧) ≡ Z 𝑥(𝑛)
Z
𝑥(𝑛) ՞ 𝑋(𝑧)

EE 453 / CE 352 - Digital Signal Processing 4

Region of Convergence
∞

𝑋 𝑧 = ෍ 𝑥(𝑛)𝑧 −𝑛
𝑛=−∞

• The ROC of 𝑋(𝑧) is the set of all

values of 𝑧 for which 𝑋(𝑧)
attains a finite value.
• The z-transform is uniquely
characterized by the expression
of 𝑋(𝑧) as well as its ROC.

EE 453 / CE 352 - Digital Signal Processing 5

Region of Convergence
∞

𝑋 𝑧 = ෍ 𝑥(𝑛)𝑧 −𝑛
𝑛=−∞

• The ROC plays an important role

when we’re looking at inverting
z-transforms and understanding
qualities like causality and
stability.

EE 453 / CE 352 - Digital Signal Processing 6

Region of Convergence: Finite-Duration Signals
• Example 1: 𝑥1 𝑛 = 1,2,5,7,0,1
↑

𝑋1 𝑧 = 𝑧 2 + 2𝑧 + 5 + 7𝑧 −1 + 𝑧 −3

ROC: Entire plane except 𝑧 = 0 and

𝑧 = ∞.

EE 453 / CE 352 - Digital Signal Processing 7

Region of Convergence: Finite-Duration Signals
• Example 2: 𝑥2 𝑛 = δ(𝑛 − 2)
• 𝑋2 𝑧 = 𝑧 −2
• ROC: Entire z-plane except 𝑧 = 0.

• Example 3: 𝑥3 𝑛 = δ(𝑛 + 2)
• 𝑋3 𝑧 = 𝑧 2
• ROC: Entire z-plane except 𝑧 = ∞.

EE 453 / CE 352 - Digital Signal Processing 8

Region of Convergence: Infinite-Duration Signal
• Practice: 𝑥 𝑛 = 𝛼 𝑛 𝑢 𝑛

𝑋 𝑧 = σ∞
𝑛=−∞ 𝑥(𝑛)𝑧
−𝑛

𝑋 𝑧 = σ∞ 𝑛 −𝑛
𝑛=0 𝛼 𝑧

𝑋 𝑧 = σ∞
𝑛=0 (𝛼𝑧 −1 )𝑛

1 𝑧
𝑋 𝑧 = −1
=
1 − 𝛼𝑧 𝑧−𝛼

ROC: 𝛼𝑧 −1 < 1 → 𝑧 > 𝛼

EE 453 / CE 352 - Digital Signal Processing 9

Region of Convergence: Infinite-Duration Signal
• Example: 𝑥 𝑛 = −𝛼 𝑛 𝑢 −𝑛 − 1

𝑋 𝑧 = σ−1 𝑛 −𝑛
𝑛=−∞(−𝛼 )𝑧

Let 𝑡 = −𝑛:
𝑋 𝑧 = − σ∞ −1 𝑡
𝑡=1(𝛼 𝑧)

𝛼 −1 𝑧 𝑧
𝑋 𝑧 =− −1
=
1−𝛼 𝑧 𝑧−𝛼

ROC: 𝛼 −1 𝑧 < 1 → 𝑧 < 𝛼

EE 453 / CE 352 - Digital Signal Processing 10

Region of Convergence: Infinite-Duration Signal
1 𝑛 1 𝑛
• Practice: 𝑔 𝑛 = 𝑢𝑛 − 𝑢 −𝑛 − 1
4 2
𝑛
1 Z 𝑧
𝑢𝑛 ՞ , 𝑧 > 1Τ4
4 Τ
𝑧−1 4

𝑛
1 Z 𝑧
− 𝑢 −𝑛 − 1 ՞ , 𝑧 < 1Τ2
2 𝑧 − 1Τ2

𝑧 𝑧
𝐺 𝑧 = +
𝑧 − 1Τ4 𝑧 − 1Τ2

ROC: 1Τ4 < 𝑧 < 1Τ2

EE 453 / CE 352 - Digital Signal Processing 11

Region of Convergence: Infinite-Duration Signal
3 𝑛
• Practice: 𝑥 𝑛 = 𝑢 𝑛 + − 𝑢 −𝑛
4

𝑧 < 3Τ4 , 𝑧 >1

𝑋(𝑧) does not exist!

EE 453 / CE 352 - Digital Signal Processing 12

Region of Convergence: Infinite-Duration Signal
• For a power series: • For a power series:
∞ ∞
𝑎1 𝑎2
𝑓 𝑧 = ෍ 𝑎𝑛 (𝑧 − 𝑐)𝑛 = 𝑎0 + 𝑎1 𝑧 − 𝑐 + 𝑎2 (𝑧 − 𝑐)2 + ⋯ 𝑓 𝑧 = ෍ 𝑎𝑛 (𝑧 − 𝑐)−𝑛 = 𝑎0 + + +⋯
𝑧−𝑐 (𝑧 − 𝑐)2
𝑛=0 𝑛=0

• There exists a number 0 ≤ 𝑟 ≤ ∞ such that • There exists a number 0 ≤ 𝑟 ≤ ∞ such that
the series: the series:
1. Converges for 𝑧 − 𝑐 < 𝑟 1. Converges for 𝑧 − 𝑐 > 𝑟
2. Diverges for 𝑧 − 𝑐 > 𝑟 2. Diverges for 𝑧 − 𝑐 < 𝑟
3. May or may not converge for values on 3. May or may not converge for values on
𝑧−𝑐 =𝑟 𝑧−𝑐 =𝑟

In short, the convergence of a power series will only occur if it is absolutely summable.

EE 453 / CE 352 - Digital Signal Processing 14

Region of Convergence: Infinite-Duration Signal

EE 453 / CE 352 - Digital Signal Processing 15

Region of Convergence: Infinite-Duration Signal

𝒓𝟏 > 𝒓𝟐 𝒓𝟏 < 𝒓𝟐

EE 453 / CE 352 - Digital Signal Processing 16

Region of Convergence: Summary

EE 453 / CE 352 - Digital Signal Processing 17

z-Transform: Properties

EE 453 / CE 352 - Digital Signal Processing 18

Linearity Property

EE 453 / CE 352 - Digital Signal Processing 19

Time Shifting Property

EE 453 / CE 352 - Digital Signal Processing 20

Time Shifting Property: ROC

EE 453 / CE 352 - Digital Signal Processing 21

Scaling in the z-Domain

EE 453 / CE 352 - Digital Signal Processing 22

Convolution Property

EE 453 / CE 352 - Digital Signal Processing 23

Rational z-Transforms
• 𝑋(𝑧) is a rational function if it can be represented as the ratio of two
polynomials in 𝑧 −1 (or 𝑧):
𝑏0 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2 + ⋯ + 𝑏𝑀 𝑧 −𝑀
𝑋 𝑧 =
𝑎0 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ + 𝑎𝑁 𝑧 −𝑁
• For LTI systems represented by LCCDEs, the z-transform of the unit
sample response ℎ(𝑛), denoted by 𝐻(𝑧) = Z{ℎ(𝑛)}, is rational.
𝑌 𝑧 σ𝑀 𝑏
𝑘=0 𝑘 𝑧 −𝑘
𝐻 𝑧 = = 𝑁
𝑋 𝑧 σ𝑘=0 𝑎𝑘 𝑧 −𝑘

EE 453 / CE 352 - Digital Signal Processing 24

Rational z-Transforms
• Many signals of practical interest have a rational z-Transform.
• For LTI systems represented by LCCDEs, the z-transform of the unit
sample response ℎ(𝑛), denoted by 𝐻(𝑧) = Z{ℎ(𝑛)}, is rational.
𝑌 𝑧 σ𝑀 𝑏
𝑘=0 𝑘 𝑧 −𝑘
𝐻 𝑧 = = 𝑁
𝑋 𝑧 σ𝑘=0 𝑎𝑘 𝑧 −𝑘

EE 453 / CE 352 - Digital Signal Processing 25

Rational z-Transforms: Poles and Zeros
• Zeroes of 𝑋(𝑧):
Values of 𝑧 for which 𝑋(𝑧) = 0.

• Poles of 𝑋(𝑧):
Values of 𝑧 for which 𝑋(𝑧) = ∞.

EE 453 / CE 352 - Digital Signal Processing 26

Rational z-Transforms: Poles and Zeros
• Let 𝑎0 , 𝑏0 ≠ 0:
𝑌 𝑧 𝑏0 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2 + ⋯ + 𝑏𝑀 𝑧 −𝑀
𝐻 𝑧 = =
𝑋 𝑧 𝑎0 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ + 𝑎𝑁 𝑧 −𝑁

𝑏0 𝑧 −𝑀 𝑧 𝑀 + 𝑏1 Τ𝑏0 𝑧 𝑀−1 + ⋯ + 𝑏𝑀 Τ𝑏0

𝐻(𝑧) =
𝑎0 𝑧 −𝑁 𝑧 𝑁 + 𝑎1 Τ𝑎0 𝑧 𝑁−1 + ⋯ + 𝑎𝑁 Τ𝑏0

𝑏0 −𝑀+𝑁 𝑧 − 𝑟1 𝑧 − 𝑟2 … 𝑧 − 𝑟𝑀
= 𝑧
𝑎0 𝑧 − 𝑝1 𝑧 − 𝑝2 … 𝑧 − 𝑝𝑁

ς 𝑀
𝑁−𝑀 𝑘=1 𝑧 − 𝑟𝑘
= 𝐺𝑧
ς𝑁
𝑘=1 𝑧 − 𝑝𝑘

EE 453 / CE 352 - Digital Signal Processing 27

Poles and Zeros of Rational z-Transform
ς 𝑀
𝑁−𝑀 𝑘=1 𝑧 − 𝑟𝑘 𝑏𝑜
𝐻 𝑧 = 𝐺𝑧 , where 𝐺 ≡
ς𝑁
𝑘=1 𝑧 − 𝑝𝑘 𝑎𝑜
• 𝑋(𝑧) has 𝑀 finite zeroes at 𝑟 = 𝑟1 , 𝑟2 , … , 𝑟𝑀
• 𝑋(𝑧) has 𝑁 finite poles at 𝑝 = 𝑝1 , 𝑝2 , … , 𝑝𝑀

• Poles and zeroes at 𝒛 = 𝟎: • Poles and zeroes at 𝒛 = ∞:

• If 𝑁 > 𝑀, there are 𝑁 − 𝑀 zeroes at 𝑧 = 0. • A zero exists at 𝑧 = ∞ if 𝑋(∞) = 0.
• If 𝑁 < 𝑀, there are 𝑁 − 𝑀 poles at 𝑧 = 0. • A pole exists at 𝑧 = ∞ if 𝑋(∞) = ∞.

If we count the poles and zeros at zero and infinity:

total number of zeroes = total number of poles

EE 453 / CE 352 - Digital Signal Processing 28

Rational z-Transforms: Poles and Zeros
2𝑧 2 −2𝑧+1
• Practice: 𝑋(𝑧) = 𝑧 3
16𝑧 +6𝑧+5

1 1 1 1
𝑧− +𝑗 𝑧− −𝑗
2 2 2 2
𝑋 𝑧 = 𝑧−0
1 3 1 3 1
𝑧− +𝑗 𝑧− +𝑗 𝑧− −
4 4 4 4 2

1 1
• Zeroes: 0, ±𝑗
2 2
1 3 1
• Poles: ± 𝑗 , −
4 4 2

EE 453 / CE 352 - Digital Signal Processing 29

Pole-Zero Plot and Conjugate Pairs
2𝑧 2 −2𝑧+1
𝑋(𝑧) = 𝑧 3
16𝑧 +6𝑧+5

• For real time-domain signals, the coefficients

of 𝑋(𝑧) are necessarily real.
• Complex poles and zeros must occur in
conjugate pairs.
• Real poles and zeros do not have to be paired up.

EE 453 / CE 352 - Digital Signal Processing 30

Pole-Zero Plot, Causality and Stability

EE 453 / CE 352 - Digital Signal Processing 31

Pole-Zero Plot, Causality and Stability

EE 453 / CE 352 - Digital Signal Processing 32

Poles/Zeroes and Time Behavior of Signals

EE 453 / CE 352 - Digital Signal Processing 33

Practice
• A LTI System is characterized by the system function:
𝑧2
𝐻 𝑧 =
(𝑧 − 0.5)(𝑧 + 2)
• What are the possibilities of its ROC? What are the implications of
these ROC choices on the stability and causality of the LTI system?

EE 453 / CE 352 - Digital Signal Processing 34

Practice
• Given a causal system:
𝑦 𝑛 = 0.9𝑦 𝑛 − 1 + 𝑥(𝑛)
• Determine 𝐻(𝑧) and sketch its pole-zero plot.

EE 453 / CE 352 - Digital Signal Processing 35

Practice
• Given that the following is a causal system:
𝑧+1
𝐻 𝑧 = 2
𝑧 − 0.9𝑧 + 0.81
• Sketch its pole-zero plot and find its difference equation
representation.

EE 453 / CE 352 - Digital Signal Processing 36

The Inverse Z-Transform
EE 453 / CE 352
Saad Baig

Slides adapted from Dr. Muhammad Umer Tariq’s offering of DSP in 2021
Inversion of the z-Transform

EE 453 / CE 352 - Digital Signal Processing 38

Inversion of Rational z-Transform
𝑏0 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2 + ⋯ + 𝑏𝑀 𝑧 −𝑀
𝑋 𝑧 =
𝑎0 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ + 𝑎𝑁 𝑧 −𝑁
In the table lookup method, we attempt to express the function X(z) as a linear
combination:
𝑋 𝑧 = 𝛼1 𝑋1 𝑧 + 𝛼2 𝑋2 𝑧 + ⋯ + 𝛼𝑘 𝑋𝑘 𝑧
Each of these terms have an inverse transform available in a table. Then due to
linearity:
𝑥(𝑛) = 𝛼1 𝑥1 𝑛 + 𝛼2 𝑥2 𝑛 + ⋯ + 𝛼𝑘 𝑥𝑘 𝑛
Example:
1 Partial Fractions 2 1 Table Lookup
𝑋 𝑧 = 𝑋 𝑧 = − 𝑥 𝑛 = 2(1)𝑛 𝑢 𝑛 − (0.5)𝑛 𝑢(𝑛)
1 − 1.5𝑧 −1 + 0.5𝑧 −2 1−𝑧 −1 1 − 0.5𝑧 −1

EE 453 / CE 352 - Digital Signal Processing 39

Rational z-Transforms: Proper vs Improper
𝑏0 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2 + ⋯ + 𝑏𝑀 𝑧 −𝑀
𝑋 𝑧 =
𝑎0 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ + 𝑎𝑁 𝑧 −𝑁

• A rational function of the above form is called proper, if 𝑎𝑁 ≠ 0 and

𝑀 < 𝑁.
• An improper rational function (𝑀 > 𝑁) can always be written as the
sum of a polynomial and a proper rational function.

EE 453 / CE 352 - Digital Signal Processing 40

Proper Rational z-Transforms
• Let 𝑋(𝑧) be a proper rational function with 𝑎𝑁 ≠ 0 and 𝑀 < 𝑁:
𝐵(𝑧) 𝑏0 + 𝑏1 𝑧 −1 + 𝑏2 𝑧 −2 + ⋯ + 𝑏𝑀 𝑧 −𝑀
𝑋 𝑧 = =
𝐴(𝑧) 1 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ + 𝑎𝑁 𝑧 −𝑁
• To simplify, we eliminate negative powers of 𝑧 by multiplying both the
numerator and denominator by 𝑧 𝑁 :
𝑏0 𝑧 𝑁 + 𝑏1 𝑧 𝑁−1 + ⋯ + 𝑏𝑀 𝑧 𝑁−𝑀
𝑋 𝑧 =
𝑧 𝑁 + 𝑎1 𝑧 𝑁−1 + ⋯ + 𝑎𝑁
which contains only positive powers of 𝑧 and is always proper.
EE 453 / CE 352 - Digital Signal Processing 41
Proper Rational z-Transforms
𝑏0 𝑧 𝑁 + 𝑏1 𝑧 𝑁−1 + ⋯ + 𝑏𝑀 𝑧 𝑁−𝑀
𝑋 𝑧 =
𝑧 𝑁 + 𝑎1 𝑧 𝑁−1 + ⋯ + 𝑎𝑁
• Our task in partial fraction expansion is to express this as a sum of simple
fractions.
𝑋(𝑧) 𝐴1 𝐴2 𝐴𝑁
= + + ⋯+
𝑧 𝑧 − 𝑝1 𝑧 − 𝑝2 𝑧 − 𝑝𝑁
• where 𝑝1 , 𝑝2 , … , 𝑝𝑁 are the roots of the denominator polynomial 𝑧 𝑁 + 𝑎1 𝑧 𝑁−1 + ⋯ + 𝑎𝑁 .
• If X(z) has a pole of multiplicity 𝑚, i.e. it contains in its denominator the factor
(𝑧 − 𝑝𝑘 )𝑚 , then the partial fraction expansion must contain the terms:
𝑋(𝑧) 𝐴1𝑘 𝐴2𝑘 𝐴𝑚𝑘
= + 2
+ ⋯+
𝑧 𝑧 − 𝑝𝑘 𝑧 − 𝑝𝑘 𝑧 − 𝑝𝑘 𝑚
EE 453 / CE 352 - Digital Signal Processing 42
Common Transform Pairs

EE 453 / CE 352 - Digital Signal Processing 43

Partial-Fraction Expansion
• Example: Find 𝑥(𝑛), given:
1
𝑋 𝑧 =
1 + 2𝑧 −1 1 − 𝑧 −1 2

𝑋(𝑧) 𝑧2
= 2
𝑧 𝑧+2 𝑧−1
𝑧2 𝐴1 𝐴2 𝐴3
2
= + + 2
𝑧+2 𝑧−1 𝑧+2 𝑧−1 𝑧−1

EE 453 / CE 352 - Digital Signal Processing 44

Partial-Fraction Expansion
𝑧2 𝐴1 𝐴2 𝐴3
2
= + + 2
𝑧+2 𝑧−1 𝑧+2 𝑧−1 𝑧−1
For 𝐴1 :
𝑧2 𝑧 + 2 𝐴1 𝑧 + 2 𝐴2 𝑧 + 2 𝐴3 𝑧 + 2
2
= + +
𝑧+2 𝑧−1 𝑧+2 𝑧−1 𝑧−1 2
𝑧2 𝐴2 𝑧 + 2 𝐴3 𝑧 + 2
2
= 𝐴1 + +
𝑧−1 𝑧−1 𝑧−1 2
4
Setting 𝑧 = −2, we get: 𝐴1 =
9

EE 453 / CE 352 - Digital Signal Processing 45

Partial-Fraction Expansion
𝑧2 𝐴1 𝐴2 𝐴3
2
= + + 2
𝑧+2 𝑧−1 𝑧+2 𝑧−1 𝑧−1
For 𝐴3 :
𝑧2 𝑧 − 1 2 𝐴1 𝑧 − 1 2 𝐴2 𝑧 − 1 2 𝐴3 𝑧 − 1 2
2
= + +
𝑧+2 𝑧−1 𝑧+2 𝑧−1 𝑧−1 2
𝑧2 𝐴1 𝑧 − 1 2
= + 𝐴2 (𝑧 − 1) + 𝐴3
𝑧+2 𝑧+2
1
Setting 𝑧 = 1, we get: 𝐴3 =
3

EE 453 / CE 352 - Digital Signal Processing 46

Partial-Fraction Expansion
For 𝐴2 :
𝑧2 𝐴1 𝑧 − 1 2
= + 𝐴2 (𝑧 − 1) + 𝐴3
𝑧+2 𝑧+2
𝑑 𝑧2 𝑑 𝐴1 𝑧 − 1 2
= + 𝐴2 (𝑧 − 1) + 𝐴3
𝑑𝑧 𝑧 + 2 𝑑𝑧 𝑧+2
5
Setting 𝑧 = 1, we get: 𝐴2 =
9
𝑋(𝑧) 4 1 5 1 1 1
= + + 2
𝑧 9𝑧 + 2 9𝑧 − 1 3 𝑧 − 1
EE 453 / CE 352 - Digital Signal Processing 47
Partial-Fraction Expansion

𝑋(𝑧) 4 1 5 1 1 1
= + + 2
𝑧 9 𝑧+2 9 𝑧−1 3 𝑧−1

4 𝑧 5 𝑧 1 𝑧
𝑋 𝑧 = + + 2
9 𝑧+2 9 𝑧−1 3 𝑧−1

4 1 5 1 1 1
𝑋 𝑧 = −1
+ −1
+
9 1 + 2𝑧 9 1−𝑧 3 1 − 𝑧 −1 2

EE 453 / CE 352 - Digital Signal Processing 48

Using Table Lookup
4 1 5 1 1 1
𝑋 𝑧 = + +
9 1 + 2𝑧 −1 9 1 − 𝑧 −1 3 1 − 𝑧 −1 2

Let’s assume we are dealing with a causal system:

4 𝑛
5 1
𝑋 𝑧 = −2 𝑢 𝑛 + 𝑢 𝑛 + 𝑛𝑢(𝑛)
9 9 3

4 𝑛
5 𝑛
𝑋 𝑧 = −2 + + 𝑢 𝑛
9 9 3

EE 453 / CE 352 - Digital Signal Processing 49

Improper Rational z-Transforms
• Practice: Express the improper rational transform in terms of a
polynomial and a proper function.
1 + 3𝑧 −1 + 11
6 𝑧 −2 + 1𝑧 −3
3
𝑋 𝑧 =
1 + 56𝑧 −1 + 16𝑧 −2

1 −1
𝑧
• Answer: 𝑋 𝑧 = 1 + 2𝑧 −1 + 5
6
1
1+ 𝑧 −1 + 𝑧 −2
6 6

EE 453 / CE 352 - Digital Signal Processing 50

Improper Rational z-Transforms

EE 453 / CE 352 - Digital Signal Processing 51

Examples from MATLAB
11 1
1+3𝑧 −1 + 𝑧 −2 + 𝑧 −3
• Example: 𝑋 𝑧 = 5
6
1
3
1+6𝑧 −1 +6𝑧 −2

1 −1
𝑧
• 𝑋 𝑧 = 1 + 2𝑧 −1 + 5
6
1
1+6𝑧 −1 +6𝑧 −2

• Comparison with MATLAB’s residuez() function:

EE 453 / CE 352 - Digital Signal Processing 52

Examples from MATLAB
• Example: 𝑋 𝑧 = 1
1−0.9 𝑧−1 −0.81𝑧−2 +0.729𝑧−3

•b=1 a = [1 -0.9 -0.81 0.729]

EE 453 / CE 352 - Digital Signal Processing 53

Z-Transform and DTFT
EE 453 / CE 352
Saad Baig

Slides adapted from Dr. Muhammad Umer Tariq’s offering of DSP in 2021
Big Picture Till Now
DT LTI Systems
(Described by LCCDEs)
Analysis
{𝑎𝑘 , 𝑏𝑘 } 𝐻(𝜔)

Design
{𝑎𝑘 , 𝑏𝑘 } 𝐻(𝜔)

DT LTI Systems as Frequency Rational z-Transforms

Selective Filters (Pole-Zero Plots)
(Frequency Response)

EE 453 / CE 352 - Digital Signal Processing 55

Z-Transform and DTFT
∞

𝑋 𝑒 𝑗𝜔 = ෍ 𝑥 𝑛 𝑒 −𝑗𝜔𝑛
𝑛=−∞

𝑋 𝑧 = ෍ 𝑥(𝑛)𝑧 −𝑛 Source: 12.1: Z-Transform - Engineering LibreTexts

𝑛=−∞
• With the DTFT, we have a complex-valued
function of a real-valued variable 𝜔 (and 2𝜋
𝑋 𝑒 𝑗𝜔 = 𝑋 𝑧 for all 𝑧 = 𝑒 𝑗𝜔 periodic).
• The z-transform is a complex-valued function
of a complex valued variable 𝑧.

EE 453 / CE 352 - Digital Signal Processing 56

Z-Transform and DTFT
• The Fourier Transform can be viewed as the z-transform of the
sequence evaluated on the unit circle.
• If 𝑋(𝑧) does not converge in the region |𝑧| = 1, the Fourier transform
𝑋(𝜔) does not exist.
• I.e. if the unit circle is not contained in the ROC of 𝑋(𝑧).

EE 453 / CE 352 - Digital Signal Processing 57

Z-Transform and DTFT

• If H(z) converges on the unit circle,

then we can obtain the frequency
response by letting 𝑧 = 𝑒 𝑗𝜔 :
∞

𝐻 𝑤 =𝐻 𝑧 ቤ 𝑗𝜔𝑛 = ෍ ℎ 𝑛 𝑒 −𝑗𝜔𝑛
𝑧=𝑒
𝑛=−∞
EE 453 / CE 352 - Digital Signal Processing 58
Z-Transform and DTFT
𝑌 𝑧 σ𝑀
𝑘=0 𝑏𝑘 𝑧 −𝑘
𝐻 𝑧 = =
𝑋 𝑧 1 + σ𝑁
𝑘=1 𝑎 𝑘 𝑧 −𝑘

−𝑀 𝑀 𝑏𝑀
𝑏𝑜 𝑧 𝑧 +⋯+
𝑏𝑜
𝐻 𝑧 =
𝑧 −𝑁 𝑧 𝑁 + ⋯ + 𝑎𝑁
𝑁−𝑀
𝑧 − 𝑟1 𝑧 − 𝑟2 … 𝑧 − 𝑟𝑀
𝐻(𝑧) = 𝑏0 𝑧
𝑧 − 𝑝1 𝑧 − 𝑝2 … 𝑧 − 𝑝𝑁
𝑒 𝑗𝜔 − 𝑟1 𝑒 𝑗𝜔 − 𝑟2 … 𝑒 𝑗𝜔 − 𝑟𝑀
𝐻(𝜔) = 𝑏0 𝑒 𝑗(𝑁−𝑀)𝜔
𝑒 𝑗𝜔 − 𝑝1 𝑒 𝑗𝜔 − 𝑝2 … 𝑒 𝑗𝜔 − 𝑝𝑁

EE 453 / CE 352 - Digital Signal Processing 59

Z-Transform and DTFT
𝑁−𝑀
𝑧 − 𝑟1 𝑧 − 𝑟2 … 𝑧 − 𝑟𝑀
𝐻(𝑧) = 𝑏0 𝑧
𝑧 − 𝑝1 𝑧 − 𝑝2 … 𝑧 − 𝑝𝑁
𝑒 𝑗𝜔 − 𝑟1 𝑒 𝑗𝜔 − 𝑟2 … 𝑒 𝑗𝜔 − 𝑟𝑀
𝐻(𝜔) = 𝑏0 𝑒 𝑗(𝑁−𝑀)𝜔
𝑒 𝑗𝜔 − 𝑝1 𝑒 𝑗𝜔 − 𝑝2 … 𝑒 𝑗𝜔 − 𝑝𝑁

• Geometric Interpretation:
• The factor 𝑒 𝑗𝜔 − 𝑧𝑘 can be interpreted as a vector in
the complex z-plane from a zero 𝑧𝑘 to the unit circle at
𝑧 = 𝑒 𝑗𝜔 .
• The factor 𝑒 𝑗𝜔 − 𝑝𝑘 can be interpreted as a vector in
the complex z-plane from a pole 𝑝𝑘 to the unit circle at
𝑧 = 𝑒 𝑗𝜔 .

EE 453 / CE 352 - Digital Signal Processing 60

Z-Transform and DTFT
𝑒 𝑗𝜔 − 𝑟1 𝑒 𝑗𝜔 − 𝑟2 … 𝑒 𝑗𝜔 − 𝑟𝑀
𝐻(𝜔) = 𝑏0 𝑒 𝑗(𝑁−𝑀)𝜔
𝑒 𝑗𝜔 − 𝑝1 𝑒 𝑗𝜔 − 𝑝2 … 𝑒 𝑗𝜔 − 𝑝𝑁

𝑒 𝑗𝜔 − 𝑟1 𝑒 𝑗𝜔 − 𝑟2 … 𝑒 𝑗𝜔 − 𝑟𝑀
|𝐻 𝜔 | = 𝑏0 𝑗𝜔
𝑒 − 𝑝1 𝑒 𝑗𝜔 − 𝑝2 … 𝑒 𝑗𝜔 − 𝑝𝑁
Product of length of vectors from Zeros to the point on Unit Circle for 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
Product of length of vectors from Poles to the point on Unit Circle for 𝑒 𝑗𝜔

EE 453 / CE 352 - Digital Signal Processing 61

System Function vs. Frequency Response

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 62
System Function vs. Frequency Response
Moving the pole locations along the unit circle:

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 63
System Function vs. Frequency Response
Moving the pole locations along the unit circle:

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 64
System Function vs. Frequency Response

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 65
System Function vs. Frequency Response
Moving the pole locations close to (0,0):

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 66
System Function vs. Frequency Response
Moving the pole locations close to (0,0):

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 67
System Function vs. Frequency Response

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 68
System Function vs. Frequency Response
Moving the zero location:

𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝒁𝒆𝒓𝒐𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
|𝐻 𝜔 | = 𝑏0
𝑃𝑟𝑜𝑑𝑢𝑐𝑡 𝑜𝑓 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑣𝑒𝑐𝑡𝑜𝑟𝑠 𝑓𝑟𝑜𝑚 𝑷𝒐𝒍𝒆𝒔 𝑡𝑜 𝑡ℎ𝑒 𝑝𝑜𝑖𝑛𝑡 𝑜𝑛 𝑈𝑛𝑖𝑡 𝐶𝑖𝑐𝑟𝑙𝑒 𝑓𝑜𝑟 𝑒 𝑗𝜔
EE 453 / CE 352 - Digital Signal Processing 69
Z-Transform and DTFT
𝑒 𝑗𝜔 − 𝑟1 𝑒 𝑗𝜔 − 𝑟2 … 𝑒 𝑗𝜔 − 𝑟𝑀
𝐻(𝜔) = 𝑏0 𝑒 𝑗(𝑁−𝑀)𝜔
𝑒 𝑗𝜔 − 𝑝1 𝑒 𝑗𝜔 − 𝑝2 … 𝑒 𝑗𝜔 − 𝑝𝑁
𝑀 𝑁

∠𝐻 𝜔 = 0 or 𝜋 + 𝑁 − 𝑀 𝜔 + ෍ ∠ 𝑒 𝑗𝜔 − 𝑟𝑘 − ෍ ∠ 𝑒 𝑗𝜔 − 𝑝𝑘
1 1

∠𝐻 𝜔 = constant factor + linear phase factor + nonlinear phase factor

• Constant factor: The phase of the gain term 𝑏0 , which is either 0 or 𝜋

depending on whether 𝑏0 is positive or negative.
• Non-linear phase factor:
Sum of angles from “zero vectors” - sum of angles from “pole vectors”

EE 453 / CE 352 - Digital Signal Processing 70