Digital Signal Processing

디지털 신호처리

Chapter 4.

Z-Transform
Introduction
◼ 𝑒 𝒋𝝎𝟎 𝒏 is an eigen-function of an LTI system:

h(n)
e j n
H(ω)
H ( )e j n
H ( ) =  h ( n )e
n =−
− j n

◼ Response to any arbitrary sequence 𝑥(𝑛) can be easily computed

in the frequency domain 𝒀 𝝎 = 𝑿 𝝎 𝑯(𝝎)
◼ However, two shortcomings :
◼ There are many signals for which DTFT does not exist.
◼ Transient response cannot be computed using DTFT.

Not energy signals

◼ Can we extend DTFT to address 2 problems?

h(n)
z n
H(z)
H ( z) z n
H ( z) =  h( n) z
n =−
−n
Y ( z) = X ( z)H ( z)

2
Bilateral z-Transform
◼ z-transform : 𝑿 𝒛 = σ∞
𝒏=−∞ 𝒙(𝒏)𝒛−𝒏
∞ ∞

𝑿 𝒛 = ෍ 𝒙 𝒏 × 𝒓−𝒏 𝒆−𝒋𝝎𝒏 = ෍ 𝒙 𝒏 (𝒓𝒆𝒋𝝎 )−𝒏

𝒏=−∞ 𝒏=−∞
◼ 𝒛 = 𝒓𝒆𝒋𝝎 , 𝒛 = 𝒓, ∠𝒛 = 𝝎
◼ 𝑿 𝒛 is defined in a region (ROC) :: 𝑹𝟏 < 𝒛 < 𝑹𝟐
Im(z) Im(z) Im(z)

R1 R2 R1 R2
Re(z) Re(z) Re(z)

◼ Inverse z-transform ::
𝟏
◼ 𝒙 𝒏 = 𝒁−𝟏 𝑿 𝒛 = ‫ׯ‬ 𝑿(𝒛)𝒛𝒏−𝟏 𝒅𝒛
𝟐𝝅𝒋 𝑪

3
Unilateral z-Transform
◼ z-transform : 𝑿 𝒛 = σ∞
𝒏=𝟎 𝒙(𝒏)𝒛−𝒏
∞ ∞

𝑿 𝒛 = ෍ 𝒙 𝒏 × 𝒓−𝒏 𝒆−𝒋𝝎𝒏 = ෍ 𝒙 𝒏 (𝒓𝒆𝒋𝝎 )−𝒏

𝒏=𝟎 𝒏=𝟎
◼ 𝒛 = 𝒓𝒆𝒋𝝎 , 𝒛 = 𝒓, ∠𝒛 = 𝝎
◼ 𝑿 𝒛 is defined in a region (ROC) :: 𝑹 < 𝒛
Im(z)

R
Re(z)

◼ Inverse z-transform ::
𝟏
◼ 𝒙 𝒏 = 𝒁−𝟏 𝑿 𝒛 = ‫ׯ‬ 𝑿(𝒛)𝒛𝒏−𝟏 𝒅𝒛
𝟐𝝅𝒋 𝑪

4
Concept of z-Transform
◼ Generalization of DTFT
◼ also applicable to signals for which DTFT does not exist
◼ brings new ideas : Stability & Causality

𝑋 𝜔 = 𝑋(𝑧)|𝑧=𝑒 𝑗𝜔 ➔ DTFT = z-Transform on unit circle

(ROC must include the unit circle.)

◼ Discrete-time counter to the Laplace transform

∞
𝑥 𝑡 ↔ 𝑋 𝑠 = න 𝑥(𝑡)𝑒 −𝑠𝑡 𝑑𝑡 , 𝑠 = 𝜎 + 𝑗𝜔
−∞

𝑥 𝑛 ↔ 𝑋 𝑧 = ෍ 𝑥(𝑛)𝑧 −𝑛 , 𝑧 = 𝑟𝑒 𝑗𝜔
𝑛=−∞

5
Laplace Transform & z-Transform
 
X ( s ) =  x(t )e dt
− st
X ( z ) =  x ( n) z − n
0 n =0

Im[s] Im[z]
s-plane z-plane

1
𝜎 Re[s] r Re[z]

ROC: Re[s] > σ ROC: 𝑧 > r

CTFT 𝑋(Ω) is defined DTFT 𝑋(𝜔) is defined

on the 𝑗Ω axis on the |𝑧| = 1 unit circle
6
Plots : 𝑿(𝝎) & 𝑿(𝒛)
◼ 𝒙(𝒏) ↔ 𝑿(𝝎) vs. 𝒙(𝒏) ↔ 𝑿(𝒛)
function of function of
real variable 𝜔 complex variable 𝑧

𝟏 𝟏
◼ 𝑿 𝝎 = vs. 𝑿 𝒛 =
𝟏−𝟎.𝟓𝟓𝟓𝒆−𝒋𝝎 𝟏−𝟎.𝟓𝟓𝟓𝒛−𝟏

Im[z]

Re[z]
𝜔

7
Practice
◼ 𝑿 𝒛 = σ∞
𝒏=−∞ 𝒙(𝒏)𝒛
−𝒏

◼ ROC : set of 𝒛 for which 𝑿(𝒛) converges(exists)

◼ function of |𝑧|, not phase of 𝑧

◼ 𝒙 𝒏 = 𝜹 𝒏 − 𝒏𝒐 ↔ 𝑿 𝒛 = 𝒛−𝒏𝟎 , 𝑧 ≠ 0 for 𝑛0 > 0, 𝑧 ≠ ∞ for 𝑛0 < 0

𝟏
◼ 𝒙 𝒏 = 𝒂𝒏 𝒖 𝒏 ↔ 𝑿 𝒛 = , 𝑧 > |𝑎| (causal sequence)
𝟏−𝒂𝒛−𝟏

𝟏
◼ 𝒙 𝒏 = −𝒂𝒏 𝒖 −𝒏 − 𝟏 ↔ 𝑿 𝒛 = , 𝑧 < |𝒂| (anti-causal sequence)
𝟏−𝒂𝒛−𝟏

◼ 𝒙 𝒏 = {𝒙 𝟎 , … , 𝒙 𝑵 − 𝟏 } ↔ σ𝑵−𝟏 −𝒏
𝒏=𝟎 𝒙(𝒏)𝒛 , entire z plane except 𝑧 = 0

8
X(z) Uniqueness & ROC
◼ 𝒙 𝒏 = 𝒂𝒏 𝒖 𝒏 = {𝟎, 𝟎, … , 𝟎, 𝟏, 𝒂, 𝒂𝟐 , … }
𝟏 𝒛
◼ 𝑿 𝒛 = = , 𝒛 > |𝒂|
𝟏−𝒂𝒛−𝟏 𝒛−𝒂 ×
◼ causal sequence :: ROC➔원 외부 |a|

◼ 𝒙 𝒏 = −𝒂𝒏 𝒖 −𝒏 − 𝟏 = {… , −𝒂−𝟐 , −𝒂−𝟏 , 𝟎, 𝟎, … }

𝟏 𝒛
◼ 𝑿 𝒛 =𝟏− 𝒛 = , 𝒛 < |𝒂|
𝟏− 𝒛−𝒂
𝒂

◼ anti-causal :: ROC➔원 내부 ×
|a|

(*) To uniquely determine x(n) from X(z), ROC is a MUST!

9
Example 4.3
◼ 𝒙 𝒏 = 𝒂𝒏 𝒖 𝒏 − 𝒃𝒏 𝒖(−𝒏 − 𝟏) Im(z)

causal anti-causal
ROC1: 𝑧 > |𝑎| ROC2: 𝑧 < |𝑏| |𝒂| |𝒃|
Re(z)

◼ ROC of 𝑥 𝑛 ∷ ROC1 ∩ ROC2 ROC: 𝑎 < 𝑧 < |𝑏|

◼ 𝑋(𝑧) = σ∞ 𝑛 −𝑛 − σ−1 𝑏 𝑛 𝑧 −𝑛
𝑛=0 𝑎 𝑧 −∞
𝑧 𝑧
= +
𝑧−𝑎 𝑧−𝑏

10
Example
𝟑 𝒏
◼ 𝒙 𝒏 =𝒖 𝒏 + − 𝒖(−𝒏)
𝟒
𝟑 𝒏
◼ 𝑥 𝑛 =𝑢 𝑛 +𝛿 𝑛 + − 𝒖(−𝒏 − 𝟏)
𝟒
1 3
◼ 𝑢 𝑛 ↔ , 𝑧 >1
1−𝑧 −1
4
◼ 𝛿 𝑛 ↔1
3 1 1
◼ −(− )𝑛 𝑢 −𝑛 − 1 ↔ 3 −1
, 𝑧 < 3/4
4 1+ 𝑧
4

Some two-sided sequences do not have z-transform.

11
Poles & Zeros
𝑷(𝒛)
◼ For rational functions :: 𝑿 𝒛 =
𝑸(𝒛)
◼ zeros : 𝑿 𝒛 |𝒛=𝐳𝐞𝐫𝐨 = 𝟎 ➔ 𝑷 𝒛 |𝒛=𝐳𝐞𝐫𝐨 = 𝟎
◼ poles : 𝑿 𝒛 |𝒛=𝐩𝐨𝐥𝐞 = ∞ ➔ 𝑸 𝒛 |𝒛=𝐩𝐨𝐥𝐞 = 𝟎

Im[z] Im[z]

X X

o o
X X
Re[z] Re[z]
o o

causal sequence anti-causal sequence

12
Poles & Zeros Example
𝟏
◼ 𝒙 𝒏 = 𝒂𝒏 𝒖 𝒏 ↔ 𝑿 𝒛 = , 𝒛 > |𝒂|
𝟏−𝒂𝒁−𝟏

𝟏
◼ 𝒙 𝒏 = −𝒂𝒏 𝒖 −𝒏 − 𝟏 ↔ 𝑿 𝒛 = , 𝒛 < |𝒂|
𝟏−𝒂𝒁−𝟏

𝟏 𝒏 𝟏 𝒏
◼ 𝒙 𝒏 = 𝒖(𝒏) + − 𝒖(𝒏)
𝟒 𝟐

𝟏 𝒏 𝟏 𝒏
◼ 𝒙 𝒏 = 𝒖 𝒏 − − 𝒖(−𝒏 − 𝟏)
𝟒 𝟐

13
Poles & Zeros Example

𝒂𝒏 , 𝒏 = 𝟎, 𝟏, … , 𝑵 − 𝟏
◼ 𝒙 𝒏 =ቊ
𝟎, otherwise

𝒛+𝟏 (𝒛+𝟐)(𝒛−𝟏)
◼ 𝑿 𝒛 = , 𝒀 𝒛 =
(𝒛+𝟐)(𝒛−𝟏) 𝒛+𝟏

14
Properties of ROC
◼ Always bounded by a circle
◼ ROC for right-sided sequences is always outside of a circle; so is
the causal sequence
◼ ROC for left-sided sequences is always inside of a circle; so is the
anti-causal sequence
◼ ROC for two-sided is always an open ring
◼ ROC for finite-duration sequences is the entire z-plane (except
possibly 𝒛 = 𝟎 and/or 𝒛 = ∞)
◼ ROC cannot include a pole.
◼ ROC∋ 𝒛 = 𝟏 (unit circle) ↔ DTFT exists.

15
Properties of z-Transform
Properties ROC
Linearity
ROC1 ∩ ROC2
𝒁 𝒂 𝟏 𝒙𝟏 𝒕 + 𝒂 𝟐 𝒙𝟐 𝒕 = 𝒂𝟏 𝑿𝟏 𝒛 + 𝒂𝟐 𝑿𝟐 𝒛
Sample shifting
ROCx
𝒁 𝒙 𝒏 − 𝒏𝟎 = 𝒛−𝒏𝟎 𝑿(𝒛)
Frequency shifting
ROCx scaled by |a|
𝒁 𝒂𝒏 𝒙 𝒏 = 𝑿(𝒛/𝒂)
Folding
Inverted ROCx
𝒁 𝒙 −𝒏 = 𝑿(𝟏/𝒛)
Complex conjugation
ROCx
𝒁 𝒙∗ 𝒏 = 𝑿∗ (𝒛∗ )
Differentiation in the z-domain
𝒅𝑿(𝒛) ROCx
𝒁 𝒏𝒙 𝒏 = −𝒛
𝒅𝒛
Multiplication
𝟏 ROC1 ∩ ~ROC2
𝒁 𝒙𝟏 𝒏 𝒙𝟐 𝒏 = ර 𝑿𝟏 (𝝂)𝑿𝟐 (𝒛/𝝂)𝝂−𝟏 𝒅𝝂
𝟐𝝅𝒋 𝑪
Convolution
ROC1 ∩ ROC2
𝒛 𝒙𝟏 𝒏 ∗ 𝒙𝟐 𝒏 = 𝑿𝟏 (𝒛)𝑿𝟐 (𝒛)
16
Example 4.4
◼ 𝑿𝟏 𝒛 = 𝟐 + 𝟑𝒛−𝟏 + 𝟒𝒛−𝟐 , 𝑿𝟐 𝒛 = 𝟑 + 𝟒𝒛−𝟏 + 𝟓𝒛−𝟐 + 𝟔𝒛−𝟑
◼ 𝑋3 𝑧 = 𝑋1 𝑧 𝑋2 𝑧 ?
(1) direct multiplication ⇒ power series
𝑋3 𝑧 = 𝟐 + 𝟑𝒛−𝟏 + 𝟒𝒛−𝟐 𝟑 + 𝟒𝒛−𝟏 + 𝟓𝒛−𝟐 + 𝟔𝒛−𝟑 = 6 + ⋯ + 24𝑧 −5
(2) using convolution
𝑥3 𝑛 = 𝑥1 𝑛 ∗ 𝑥2 𝑛 = 2,3,4 ∗ {3,4,5,6}

3 4 5 6
2 6 8 10 12
3 9 12 15 18
4 12 16 20 24

X 3 ( z ) = 6 + 17 z −1 + 34 z −2 + 43 z −3 + 38 z −4 + 24 z −5

17
Example 4.5
◼ 𝑿𝟏 𝒛 = 𝒛 + 𝟐 + 𝟑𝒛−𝟏 , 𝑿𝟐 𝒛 = 𝟐𝒛𝟐 + 𝟒𝒛 + 𝟑 + 𝟓𝒛−𝟏
◼ 𝑋3 𝑧 = 𝑋1 𝑧 𝑋2 𝑧
(1) direct multiplication
𝑋3 𝑧 = 𝒛 + 𝟐 + 𝟑𝒛−𝟏 𝟐𝒛𝟐 + 𝟒𝒛 + 𝟑 + 𝟓𝒛−𝟏 = 𝟐𝒛𝟑 + ⋯ + 𝟏𝟓𝒛−𝟐
(2) using convolution
𝑥3 𝑛 = 𝑥1 𝑛 ∗ 𝑥2 𝑛 = 1,2,3 ∗ {2,4,3,5}

2 4 3 5
1 2 4 3 5
2 4 8 6 10
3 6 12 9 15

X 3 ( z ) = 2 z 3 + 8 z 2 + 17 z1 + 23 + 19 z −1 + 15 z −2

18
Some z-Transform Pairs
Sequence Transform ROC
𝜹(𝒏) 𝟏 ∀𝒛
𝟏
𝒖(𝒏) 𝒛 >𝟏
𝟏 − 𝒛−𝟏
𝟏
−𝒖(−𝒏 − 𝟏) 𝒛 <𝟏
𝟏 − 𝒛−𝟏
𝟏
𝒂𝒏 𝒖(𝒏) 𝒛 > |𝒂|
𝟏 − 𝒂𝒛−𝟏
𝟏
−𝒃𝒏 𝒖(−𝒏 − 𝟏) 𝒛 < |𝒃|
𝟏 − 𝒃𝒛−𝟏
(𝒂sin𝝎𝟎 )𝒛−𝟏
𝒂𝒏 sin𝝎𝟎 𝒏 𝒖(𝒏) 𝒛 > |𝒂|
𝟏 − 𝟐𝒂cos𝝎𝟎 𝒛−𝟏 + 𝒂𝟐 𝒛−𝟐
𝒏 𝟏 − (𝒂cos𝝎𝟎 )𝒛−𝟏
𝒂 cos𝝎𝟎 𝒏 𝒖(𝒏) 𝒛 > |𝒂|
𝟏 − 𝟐𝒂cos𝝎𝟎 𝒛−𝟏 + 𝒂𝟐 𝒛−𝟐
𝒏 𝒂𝒛−𝟏 𝒛 > |𝒂|
𝒏𝒂 𝒖(𝒏)
(𝟏 − 𝒂𝒛−𝟏 )𝟐

𝒏 𝒃𝒛−𝟏
−𝒏𝒃 𝒖(−𝒏 − 𝟏) 𝒛 < |𝒃|
(𝟏 − 𝒃𝒛−𝟏 )𝟐

19
Inversion of z-Transform
◼ Formula::
𝟏
𝒙 𝒏 = 𝒁−𝟏 𝑿 𝒛 = ‫ׯ‬ 𝑿(𝒛)𝒛𝒏−𝟏 𝒅𝒛 : impractical
𝟐𝝅𝒋 𝑪

◼ Using power series expansion

◼ Partial fraction expansion
𝟏
𝑿 𝒛 =  𝒂𝒏 𝒖(𝒏) or −𝒂𝒏 𝒖(−𝒏 − 𝟏)
𝟏−𝒂𝒛−𝟏
◼ Long division

◼ Matlab
𝐵(𝑧)
 ( n) 𝐻 𝑧 = x ( n)
𝐴(𝑧)

20
Z-Transform as a Power Series
◼ Z-transform is a power series ::
◼ 𝑋 𝑧 = σ∞
𝑛=−∞ 𝑥 𝑛 𝑧
−𝑛

◼ = ⋯ + 𝑥 −2 𝑧 2 + 𝑥 −1 𝑧1 + 𝑥 0 + 𝑥 1 𝑧 + ⋯

◼ Example: 𝑿 𝒛 = 𝟐𝒛𝟓 + 𝒛𝟑 − 𝒛𝟐 + 𝟏 + 𝟑𝒛−𝟏 − 𝟒𝒛−𝟒

ROC?

21
Power Series Expansion
◼ 𝑿 𝒛 = exp(−𝟎. 𝟓𝒛−𝟏 ) ➔ 𝒙 𝒏 =?
𝑥𝑛
◼ Use Taylor series expansion :: exp 𝑥 = σ∞
𝑛=0 𝑛!
(−0.5𝑧−1 )𝑛 𝑛
(−0.5)
◼ 𝑋 𝑧 = σ∞
𝑛=0 = σ∞
𝑛=0 𝑧 −𝑛
𝑛! 𝑛!
(−0.5)𝑛
◼ 𝑥 𝑛 = 𝑢(𝑛)
𝑛!

22
Example
𝟏−𝒛−𝟏 +𝒛−𝟐
◼ 𝑿 𝒛 = , ROC:𝟏 < 𝒛 < 𝟐
(𝟏−𝟎.𝟓𝒛−𝟏 )(𝟏−𝟐𝒛−𝟏 )(𝟏−𝒛−𝟏 )

pole=0.5 pole=2 pole=1

causal seq. anti-causal seq. causal seq.
1
X X X 2

23
 𝒛
Example 4.7 𝑿 𝒛 = 𝟑𝒛𝟐 −𝟒𝒛+𝟏
⇒𝒙 𝒏 ?

◼ Using partial fraction expansion::

1 1 1 1
◼ 𝑋 𝑧 = − :: poles → 𝑧1 = 1, 𝑧2 = 1/3
2 1−𝑧 −1 2 1−1𝑧 −1
3

𝟏 𝟏
ROC : 𝒛 > 𝟏 ROC : 𝒛 < ROC : < 𝒛 < 𝟏
𝟑 𝟑

24
Homework
𝒛𝟑 −𝟏𝟎𝒛𝟐 −𝟒𝒛+𝟏
◼ 𝑿 𝒛 = , 𝑹𝑶𝑪 ∶ 𝒛 < 𝟏
𝟐𝒛𝟑 −𝟐𝒛𝟐 −𝟒𝒛

25
 𝒛
Example 4.8 𝑿 𝒛 = 𝟑𝒛𝟐 −𝟒𝒛+𝟏
⇒𝒙 𝒏 ?

◼ Using MATLAB::
residuez
◼ b=[0,1]; a=[3,-4,1]; [R,p,C]=residuez(b,a); 부분분수꼴로 해석해 주는 함수
◼ R = [ -0.5, 0.5], p = [1/3, 1], C=[ ]
−0.5 0.5
◼ 𝑋 𝑧 = 1 −1
+ 1−1∙𝑧 −1 + 0
1− 𝑧
3
1
◼ 𝑥 𝑛 = −0.5(3)𝑛 𝑢 𝑛 + 0.5(1)𝑛 𝑢(𝑛)

◼ b=[0,1]; a=[3,-4,1]; [delta,n]=impseq(0,0,7); x=filter(b,a,delta);

0𝑧 2 + 𝑧
delta={1,0,0,0,0…,0} x = {0,1/3,4/9,…}
3𝑧 2 − 4𝑧 + 1

filter
𝐵(𝑧)/𝐴(𝑧)로 표현한 시스템의 출력 구하는 함수

26
Z-Transform Analysis of LTI Systems
◼ LTI systems described by difference equations play an
important role in signal processing.
◼ They are widely used as filters.
◼ They are used as models for distortion.
◼ They are used as models for characterizing signals.
◼ Important questions
◼ What is the stable conditions?
◼ What is the causal conditions?
◼ If 𝐻(𝑧) is a model for a system that introduces distortion, can it be
reversed or inverted?
◼ How does the frequency response relate to the system function?

27
LTI Systems : Difference Equations
◼ LTI systems described by difference equations play an
important role in signal processing.
◼ Filters : FIR & IIR
◼ Filter design
◼ Filter implementation
◼ Models
◼ For a system introducing distortion
◼ For characteristics of a random process

28
System Function in z-Transform
◼ 𝒀 𝒛 = 𝑯 𝒛 𝑿(𝒛)

x ( n) 𝒉(𝒏) y ( n) H ( z ) =  h( n) z − n
−

◼ When LTI system is described by 차분방정식:

𝑁 𝑀
𝑌(𝑧) σ𝑀
𝑙=0 𝑏𝑙 𝑧
−𝑙
𝐵(𝑧)
𝑦(𝑛) + ෍ 𝑎𝑘 𝑦(𝑛 − 𝑘) = ෍ 𝑏𝑙 𝑥(𝑛 − 𝑙) 𝐻(𝑧) = = =
𝑋(𝑧) 1 + σ𝑁𝑘=1 𝑎𝑘 𝑧
−𝑘 𝐴(𝑧)
𝑘=1 𝑙=0

𝑁 𝑀 𝑁−𝑀
ς𝑀
𝑙=1(𝑧 − 𝑧𝑙 ) zeros
𝐻(𝑧) = 𝑏0 𝑧
𝑌(𝑧) + ෍ 𝑎𝑘 𝑧 −𝑘 𝑌(𝑧) = ෍ 𝑏𝑙 𝑧 −𝑙 𝑥(𝑛 − 𝑙) ς𝑁
𝑘=1(𝑧 − 𝑝𝑘 ) poles
𝑘=1 𝑙=0

𝑦(𝑛 − 𝑘) ↔ 𝑌(𝑧)𝑧 −𝑘

29
 Example
(𝒛+𝟏)𝟐
◼ Given 𝑯 𝒛 = , find difference equation
(𝒛−𝟎.𝟓)(𝒛+𝟎.𝟐𝟓)

(1) causal case : ROC > 0.5 stable (2) anti-causal case : ROC < 0.25 unstable

𝑌(𝑧) 𝑧 2 + 2𝑧 + 1 1 + 2𝑧 −1 + 𝑧 −2 𝑌(𝑧) 𝑧 2 + 2𝑧 + 1
𝐻 𝑧 = = = 𝐻 𝑧 = =
𝑋(𝑧) 𝑧 2 − 1 𝑧 − 1 1 − 1 𝑧 −1 − 1 𝑧 −2 𝑋(𝑧) 𝑧 2 − 1 𝑧 − 1
4 8 4 8 4 8
1 1
1 1 𝑌 𝑧 [𝑧 2 − 𝑧 − ] = 𝑋(𝑧)[𝑧 2 + 2𝑧 + 1]
𝑌 𝑧 [1 − 𝑧 −1 − 𝑧 −2 ] = 𝑋(𝑧)[1 + 2𝑧 −1 + 𝑧 −2 ] 4 8
4 8

1 1
𝑦 𝑛+2 − 𝑦 𝑛+1 − 𝑦 𝑛
4 8
1 1
𝑦 𝑛 = 𝑦 𝑛−1 + 𝑦 𝑛−2 +𝑥 𝑛 = 𝑥(𝑛 + 2) + 2𝑥(𝑛 + 1) + 𝑥(𝑛)
4 8
+ 2𝑥(𝑛 − 1) + 𝑥(𝑛 − 2)

𝑦 𝑛 = −2𝑦 𝑛 + 1 + 8𝑦 𝑛 + 2
−8𝑥 𝑛 + 2 − 16𝑥 𝑛 + 1 − 8𝑥(𝑛)

30
Transfer Function Representation
◼ If ROC ∋(𝒛 = 𝒆𝒋𝝎 )
𝑧 = 𝑒 𝑗𝜔 ς𝑀
𝑙=1(𝑧 − 𝑧𝑙 )
ς𝑴 𝒋𝝎
𝟏 (𝒆 −𝒛𝒍 ) 𝐻(𝑧) = 𝑏0 𝑧 𝑁−𝑀
◼ 𝑯 𝝎 = 𝒃𝟎 𝒆𝒋(𝑵−𝑴)𝝎 ς𝑁
ς𝑵 𝒋𝝎
𝟏 (𝒆 −𝒑𝒌 ) 𝑘=1(𝑧 − 𝑝𝑘 )

ς𝑴 𝒋𝝎
𝟏 |𝒆 −𝒛𝒍 |
◼ |𝑯 𝝎 | = |𝒃𝟎 | ς𝑵 𝒋𝝎
𝟏 |𝒆 −𝒑𝒌 |

◼ ∠𝑯 𝝎 = 0 or π + 𝑁 − 𝑀 𝜔 + σ𝑀
1 ∠ 𝑒
𝑗𝜔 − 𝑧 − σ𝑁 ∠ 𝑒 𝑗𝜔 − 𝑝
𝑙 1 𝑘

31
MATLAB Implementation
◼ [H,w] = freqz(b,a,N) freqz : 𝐵(𝑧)/𝐴(𝑧)로 표현된 시스템의 주파수 응답 구하기

◼ Example 4.11 : 𝒚 𝒏 = 𝟎. 𝟗𝒚 𝒏 − 𝟏 + 𝒙(𝒏)

1
◼ 𝐻 𝑧 = zplane : 𝐵(𝑧)/𝐴(𝑧)로 표현된 시스템의 pole/zero 그리기
1−0.9𝑧 −1
◼ b=[1,0]; a=[1,-0.9]; zplane(b,a);
◼ [H,w]=freqz(b,a,100); magH=abs(H); phaH=angle(H);

32
 Example 4.12
𝑅(1) 𝑅(2)
𝐻 𝑧 = 1.2345 + + 𝑧 −1 +𝑧 −2
1 − 𝑝(1)𝑧 −1 1 − 𝑝(2)𝑧 −1 using MATLAB :: 𝐻 𝑧 = 1−0.9𝑧 −1+0.81𝑧 −2
𝒛+𝟏
◼ 𝑯 𝒛 =
𝒛𝟐 −𝟎.𝟗𝒛+𝟎.𝟖𝟏

◼ For causal system, ROC?

◼ 𝐻 𝜔 = 𝐻(𝑧 = 𝑒 𝑗𝜔 )

◼ Difference equation ::

◼ impulse response ℎ(𝑛) ::

ℎ 𝑛 = 𝐶𝛿 𝑛 + 𝑅 1 𝑝(1)𝑛 + 𝑅 2 𝑝(2)𝑛

ℎ 𝑛 = 𝐶𝛿 𝑛 + 𝑅 1 |𝑝(1)|𝑛 𝑒 𝑗θ(1) + 𝑅 2 |𝑝(2)|𝑛 𝑒 𝑗θ(2)

33
Relationships between System & Representations

34
Stability & Causality

◼ Stable : σ∞
𝒏=−∞ 𝒉 𝒏 < ∞
◼ Causal : 𝒉 𝒏 = 𝟎, 𝒏 < 𝟎

◼ Difference equation description

𝑁 𝑀

෍ 𝑎𝑘 𝑦 𝑛 − 𝑘 = ෍ 𝑏𝑘 𝑥(𝑛 − 𝑘)
𝑘=0 𝑘=0

σ𝑀 𝑏
𝑘=0 𝑘 𝑧 −𝑘 𝑏0 ς𝑀
𝑘=1 (1 − 𝑐𝑘 𝑧 −1 )
𝐻 𝑧 = 𝑁 =
σ𝑘=0 𝑎𝑘 𝑧 −𝑘 𝑎0 ς𝑁 𝑘=1(1 − 𝑑 𝑘 𝑧 −1 )

𝑥(𝑛) ℎ(𝑛) 𝑦 𝑛 ROC?

35
Causality
◼ Causality: right-sided inverse transforms for all poles
◼ ROC extends outward from pole with largest radius

◼ 𝑧 > max| 𝑑𝑘 |
𝑘

Im[z]

x
x x Re[z]
x
x

36
Stability
◼ Stability : absolute summability of 𝒉(𝒏)
◼ ∞>σℎ 𝑛 = σ ℎ 𝑛 |𝑧|−𝑛 ||𝑧|=1 = σ |ℎ(𝑛)𝑧 −𝑛 ||𝑧|=1
◼ ≥ | σ ℎ 𝑛 𝑧 −𝑛 | 𝑧 =1 = |𝐻 𝑧 ||𝑧|=1

◼ 𝐻 𝑧 < ∞ on 𝑧 = 1 ➔ ROC includes unit circle

Im[z]

x
x x Re[z]
x
x unit circle
𝑧 =1

37
Example
◼ 𝒚 𝒏 − 𝟏. 𝟓𝒚 𝒏 − 𝟏 − 𝒚 𝒏 − 𝟐 = 𝟐𝒙 𝒏 − 𝒙(𝒏 − 𝟏)
𝟐−𝒛−𝟏 𝟐−𝒛−𝟏 𝟏.𝟐 𝟎.𝟖
◼ 𝑯 𝒛 = = = +
𝟏−𝟏.𝟓𝒛−𝟏 −𝒛−𝟐 (𝟏+𝟎.𝟓𝒛−𝟏 )(𝟏−𝟐𝒛−𝟏 ) 𝟏−𝟐𝒛−𝟏 𝟏+𝟎.𝟓𝒛−𝟏

38
Stability & Causality
◼ Stable & causal ➔ all poles inside 𝒛 = 𝟏

◼ Examples
◼ 𝒚 𝒏 + 𝟎. 𝟕𝟓𝒚 𝒏 − 𝟏 + 𝟎. 𝟏𝟐𝟓𝒚 𝒏 − 𝟐 = 𝒙 𝒏 − 𝒙 𝒏 − 𝟏

𝒛𝟐 +𝟐𝒛+𝟏
◼ 𝑯 𝒛 =
𝒛−𝟎.𝟓

◼ 𝒚 𝒏 = 𝟎. 𝟖𝟏𝒚 𝒏 − 𝟐 + 𝒙 𝒏 − 𝒙(𝒏 − 𝟐)

39
Inverse Systems for LTI Systems

𝒙(𝒏) ℎ(𝑛) 𝒚(𝒏) ℎ𝐼 (𝑛) 𝒙(𝒏)

𝑦 𝑛 =ℎ 𝑛 ∗𝑥 𝑛 𝑥 𝑛 = ℎ𝐼 𝑛 ∗ 𝑦 𝑛 ⇒ ℎ𝐼 𝑛 ∗ ℎ 𝑛 = 𝛿 𝑛
𝟏
𝐻 𝑧 𝐻𝐼 𝑧 = 1 ➔ 𝑯𝑰 𝒛 = 𝑯(𝒛)

◼ For stability, ROC for 𝑯𝑰 𝒛 ∋ 𝒛 = 𝟏

◼ 𝐻 𝑧 > 0 for z = 𝑒 𝑗𝜔
𝐻(𝑒 𝑗𝜔 )

➔ no inverse system

40
Stable/Causal Inverse System
𝒃𝟎 ς(𝟏−𝒄𝒌 𝒛−𝟏 )
◼ 𝑯 𝒛 =
𝒂𝟎 ς(𝟏−𝒅𝒌 𝒛−𝟏 )
◼ For stable/causal system, 𝑑𝑘 < 1
➔poles : inside unit circle

𝒂𝟎 ς(𝟏−𝒅𝒌 𝒛−𝟏 )
◼ 𝑯𝑰 𝒛 =
𝒃𝟎 ς(𝟏−𝒄𝒌 𝒛−𝟏 )
◼ For stable/causal system, 𝑐𝑘 < 1
➔ zeros : inside unit circle

41
Example : Multipath Communication

𝑦 𝑛 = 𝑥 𝑛 + 𝑎𝑥(𝑛 − 1)

𝑥(𝑛)

◼ Inverse system exists?

1
◼ 𝐻 𝑧 = 1 + 𝑎𝑧 −1 & 𝐻𝐼 𝑧 =
1+𝑎𝑧 −1
◼ Requires :: 𝑎 < 1
◼ Inverse system :: 𝑦 𝑛 + 𝑎𝑦 𝑛 − 1 = 𝑥(𝑛)

42
Example : Inverse System
𝟏−𝟐𝒛−𝟏
◼ 𝑯 𝒛 =
𝟏−𝟎.𝟑𝒛−𝟏

𝟏−𝟎.𝟓𝒛−𝟏
◼ 𝑯 𝒛 =
𝟏−𝟎.𝟑𝒛−𝟏

43
MAGNITUDE FREQUENCY RESPONSE
FROM POLE & ZEROS

44
Magnitude Freqeuncy Response
σ 𝒃𝒌 𝒆−𝒋𝝎𝒌
◼ 𝑯 𝝎 = 𝑯 𝒛 |𝒛=𝒆𝒋𝝎 = σ 𝒂𝒌 𝒆−𝒋𝝎𝒌
𝒃𝟎 ς(𝟏−𝒄𝒌 𝒆−𝒋𝝎 )
◼ In pole/zero form : 𝑯 𝝎 =
𝒂𝟎 ς(𝟏−𝒅𝒌 𝒆−𝒋𝝎 )

𝒃𝟎 ς 𝒆𝒋𝝎 − 𝒄𝒌
𝑯 𝜔 =
|𝒂𝟎 | ς 𝒆𝒋𝝎 − 𝒅𝒌

𝑳𝒌,𝒑 = |𝒆𝒋𝝎 − 𝒅𝒌 |
𝒆𝒋𝝎
𝒅𝒌
𝑳𝒌,𝒛 = |𝒆𝒋𝝎 − 𝒄𝒌 |

𝒄𝒌
𝒃𝟎 ς(distance from 𝒆𝒋𝝎 to zero)
𝑯 𝜔 =
|𝒂𝟎 | ς(distance from 𝒆𝒋𝝎 to pole)

45
Magnitude Freqeuncy Response
◼ When 𝑒 𝒋𝜔 is close to a zero, |𝑯(𝑒 𝒋𝜔 )| is small.
◼ When 𝑒 𝒋𝜔 is close to a pole, |𝑯(𝑒 𝒋𝜔 )| is large.

𝟏
◼ Example : 𝐇 𝐳 = → zero=0, pole=0.75
𝟏−𝟎.𝟕𝟓𝒛−𝟏

46
Example
𝟏−𝒛−𝟐
◼ 𝑯 𝒛 =
(𝟏−𝟎.𝟗𝟓𝒆𝒋𝟎.𝟐𝟓𝝅 𝒛−𝟏 )(𝟏−𝟎.𝟗𝟓𝒆−𝒋𝟎.𝟐𝟓𝝅 𝒛−𝟏 )

◼ Poles push magnitude response UP.

◼ Zeros pull magnitude response DOWN.
47
Example

48
Example

49
 SOLUTION OF
DIFFERENCE EQUATION

50
One-Sided z-Transform
◼ 𝒁+ 𝒙 𝒏 = 𝒁 𝒙 𝒏 𝒖 𝒏 = 𝑿+ 𝒛 = σ∞ 𝒏=𝟎 𝒙(𝒏)𝒛−𝒏

◼ It can handle “initial conditions”.

◼ 𝒁+ 𝒙 𝒏 − 𝒌 = σ−𝟏
𝒎=−𝒌 𝒙 𝒎 𝒛− 𝒎+𝒌
+ 𝒛−𝒌 𝑿+ (𝒛)
initial conditions

◼ 𝒚 𝒏 + σ𝑵
𝒌=𝟏 𝒂𝒌 𝒚 𝒏 − 𝒌 = σ 𝑴
𝒎=𝟎 𝒃𝒎 𝒙 𝒏 − 𝒎 , 𝒏 ≥ 𝟎

◼ subject to : {𝑦(𝑖), 𝑖 = −1, … , −𝑁} and {𝑥(𝑖), 𝑖 = −1, … , −𝑀} 초기조건

51
Example 4.14

◼ 𝒚 𝒏 − 𝟏. 𝟓𝒚 𝒏 − 𝟏 + 𝟎. 𝟓𝒚 𝒏 − 𝟐 = 𝒙 𝒏 , 𝒏 ≥ 𝟎
◼ 𝑥 𝑛 = (0.25)𝑛 𝑢(𝑛), 𝑦 −1 = 4, 𝑦 −2 = 10

◼ 𝑌 𝑧 − 1.5{𝑌 𝑧 𝑧 −1 + 𝑦(−1)} +0.5{𝑌(𝑧)𝑧 −2 + 𝑦(−1)𝑧 −1 + 𝑦(−2)} = 𝑋(𝑧)

1
◼ 𝑌 𝑧 − 1.5𝑌 𝑧 𝑧 −1 − 6 + 0.5𝑌 𝑧 𝑧 −2 + 2𝑧 −1 + 5 = 1−0.25𝑧 −1

1
◼ 𝑌 𝑧 1 − 1.5𝑧 −1 + 0.5𝑧 −2 = 1−0.25𝑧 −1 + 1 − 2𝑧 −1

1
◼ 𝑌 𝑧 1 − 0.5𝑧 −1 1 − 𝑧 −1 = 1−0.25𝑧 −1 + 1 − 2𝑧 −1
0.5 1 0.25
system poles input pole 초기값에 의한 부분

𝒚 𝒏 = 𝑨(𝟎. 𝟓)𝒏 𝒖 𝒏 + 𝑩(𝟏)𝒏 𝒖 𝒏 + 𝑪(𝟎. 𝟐𝟓)𝒏 𝒖(𝒏)

52