79

Chap. 8 Z-transform

7.1 Infinite Impulse Response(IIR) system

• Input & output in a LTI system

x[n] → h[n] → y[n] = x[n] ~ h[n]

∑
M ∑
N
• Difference equation : y[n] = bl x[n − l] + ak y[n − k]
l=0 k=1
1
• EX: y[n] = x[n] + y[n − 1], where y[n] = 0 for n < 0.
2
⇒ when x[n] = δ[n], h[n] = δ[n] + 12 h[n − 1]
– n = 0, h[0] = δ[0] + 12 h[−1] = 1
– n = 1, h[1] = δ[1] + 12 h[0] = 1
2 ·1= 1
2

– n = 2, h[2] = δ[2] + 12 h[1] = 1

2 · 1
2 = ( 12 )2
– n = 3, h[3] = δ[3] + 12 h[2] = 1
2 · ( 12 )2 = ( 12 )3
⇒ h[n] = ( 12 )n u[n] ⇒ n → ∞, h[n] → 0 ⇒ stable ystem

• EX: y[n] = x[n] + 2y[n − 1], where y[n] = 0 for n < 0.

⇒ when x[n] = δ[n], h[n] = δ[n] + 2h[n − 1]

– n = 0, h[0] = δ[0] + 2h[−1] = 1
– n = 1, h[1] = δ[1] + 2h[0] = 2 · 1 = 2
– n = 2, h[2] = δ[2] + 2h[1] = 2 · 2 = (2)2
– n = 3, h[3] = δ[3] + 2h[2] = 2 · 22 = (2)3
⇒ h[n] = (2)n u[n] ⇒ n → ∞, h[n] → ∞ ⇒ unstable system

• FIR system vs IIR System

FIR system IIR system

∑M ∑
M ∑N
y[n] = bl x[n − l] y[n] = bl x[n − l] + ak y[n − k]
l=0 l=0 k=1
Non-recursive structure Recursive structure
Finite length impulse response Infinite length impulse response
Always stable Stable / Unstable
Linear phase is possible : ]H(ω̂) = −αω̂ + β Linear phase is impossible
Slower convergence than IIR system Faster convergence than FIR system
 80

7.2 Z-Transform

• Definition of z-transform
∞
∑
X(z) = x[n]z −n where z = rejθ
n=−∞

• Basic properties of z-transform

∞
∑
– Impulse signal : x[n] = δ[n − n0 ] ←→ X(z) = δ[n − n0 ]z −n = z −n0
n=−∞
– Impulse response :
∞
∑
h[n] = h[k]δ[n − k] = · · · h[−1]δ[n + 1] + h[0]δ[n] + h[1]δ[n − 1] · · ·
k=−∞

⇕
∞
∑
H(z) = · · · h[−1]z −1 + h[0]z 0 + h[1]z −1 + h[2]z −2 = h[k]z −k
k=−∞

– Time delay : x[n − n0 ] ←→ z −n0 X(z)

– Convolution :
∞
∑
y[n] = x[n] ~ h[n] = h[k]x[n − k]
k=−∞
⇕
∞
∑ ∞ ( ∑
∑ ∞ )
−n
Y (z) = y[n]z = h[k]x[n − k] z −n
n=−∞ n=−∞ k=−∞
∞
∑ ( ∑
∞ ) ∞
∑ ( ∑
∞ )
= h[k] x[n − k]z −n = (z −k X(z)) = h[k]z −k X(z)
k=−∞ n=−∞ k=−∞ k=−∞

= H(z) · X(z)

– Difference Equation & Transfer function :

∑
L ∑
M ∑
L ∑
M
y[n] = bl x[n − l] + ak y[n − k] ←→ Y (z) = bl z −l X(z) + ak z −k Y (z)
l=0 k=1 l=0 k=1
∑
L
bl z −l
Y (z) l=0
⇒ H(z) = =
X(z) ∑M
1− ak z −k
k=1

⋄ Ex) y[n] = 0.5y[n − 1] + 0.3y[n − 2] − x[n] + 3x[n − 1] − 2x[n − 2]

⇒ Y (z) = (0.5z −1 + 0.3z −2 )Y (z) − (1 − 3z −1 + 2z −2 )X(z)

−1 + 3z −1 − 2z −2
⇒ H(z) =
1 − 0.5z −1 − 0.3z −2

1 − 3z −1
⋄ Ex) H(z) = Y (z)/X(z) =
1 − 0.8z −1 + 0.6z −3 + 0.3z −4
⇒ y[n] = 0.8y[n − 1] − 0.6y[n − 3] − 0.3y[n − 4] + x[n] − 3x[n − 1]
 81

• Region of convergence (ROC)

– Causal signal :

∞
∑ ∞
∑  1
|z| > |a|
x[n] = an u[n] ↔ X(z) = an z −n = (az −1 )n = 1−az −1
 ∞ Otherwise
n=0 n=0

– Noncausal signal :

−1
∑ ∞
∑  b−1 z
|z| < |b|
n −n −1 1−b−1 z
x[n] = b u[−n − 1] ↔ X(z) =
n
b z = (b n
z) =
 ∞ Otherwise
n=−∞ n=1

– x[n] = an u[n] + bn u[−n − 1] ⇒ ROC : {|z| > |a|} ∩ {|z| < |b|}

i) When |a| > |b|,

ii) When |a| < |b|,

 82

• Poles and Zeros

∑
L /( ∑
M ) ∏L
−l −k (z − zl )
H(z) = bl z 1− ak z = A · ∏Ml=0
l=0 k=1 k=0 (z − Pk )

 Zeros : {z } H(z) =0
l
⇒ z=zl
 Poles : {P } H(z) =∞
k z=Pk

• Inverse z -transform (for causal system)

1) Factorize the denominator polynomials in form of (1 − Pk z −1 )
∏
L
(1 − zl z −1 )
l=0
H(z) = A · (L < M )
∏
M
−1
(1 − Pk z )
k=0
2) Construct partial fraction expansions of H(z)
A0 A1 AM
H(z) = −1
+ −1
+ ··· +
1 − P0 z 1 − P1 z 1 − PM z −1
3) Write down the inverse as

h[n] = A0 (P0 )n u[n] + A1 (P1 )n u[n] + · · · + AM (PM )n u[n]

1 − 2.1z −1
– EX: X(z) =
1 − 0.3z −1 − 0.4z −2
1 − 2.1z −1 1 − 2.1z −1 A B
X(z) = −1 −2
= = +
1 − 0.3z − 0.4z (1 + 0.5z −1 )(1 − 0.8z −1 ) 1 + 0.5z −1 1 − 0.8z −1

1 − 2.1z −1 1 + 4.2
A = X(z)(1 + 0.5z −1 ) = = =2
z=−0.5 1 − 0.8z −1 1 + 1.6
z=−0.5

1 − 2.1z −1 1 − 2.1/0.8
B = X(z)(1 − 0.8z −1 ) = = = −1
z=0.8 1 + 0.5z −1 1 + 0.5/0.8
z=0.8

2 1
⇒ X(z) = −1
−
1 + 0.5z 1 − 0.8z −1

⇒ x[n] = 2(−0.5)n u[n] − (0.8)n u[n]

 83

2 − 2.4z −1 − 0.4z −2
– EX: X(z) =
1 − 0.3z −1 − 0.4z −2

2 − 2.4z −1 − 0.4z −2 2 − 2.4z −1 − 0.4z −2

X(z) = =
1 − 0.3z −1 − 0.4z −2 (1 + 0.5z −1 )(1 − 0.8z −1 )
1 − 2.1z −1 + (1 − 0.3z −1 − 0.4z −2 ) 1 − 2.1z −1
= −1 −2
= +1
1 − 0.3z − 0.4z 1 − 0.3z −1 − 0.4z −2
2 1
= − +1
1 + 0.5z −1 1 − 0.8z −1

⇔ x[n] = 2 · (−0.5)n u[n] − (0.8)n u[n] + δ[n]

• Properties of poles and zeros

- N-degree real polynomial : p(x) = a0 + a1 x + · · · + aN xN with real ai .

⇒ There are N roots.

⇒ If x0 is a root, the conjugate of x0 is also a root. That is, if p(x0 ) = 0, then p(x∗0 ) = 0.
A(z)
- The z -transform of a system : H(z) =
B(z)
⇒ Poles or zeros must occur in complex conjugate pairs
⇒ If a zero is z0 = |z0 |ejθ0 , z0∗ = |z0 |e−jθ0 is also a zero.
If a pole is P0 = |P0 |ejϕ0 , P0∗ = |P0 |e−jϕ0 is also a pole.

The pole and zero locations in conjugate pairs

• Pole and zero locations for BIBO system

- Stable BIBO system : If |x[n]| < Mx , then |y[n]| < My

- Sufficient condition for BIBO system.
∞
∑ ∞
∑ ∞
∑
|y[n]| = h[k[x[n − k] ≤ |h[k]||x[n − k]| ≤ Mx · |h[k]| ≤ My
k=0 k=0 k=0
∞
∑
⇒ Sufficient condition for a BIBO stable system : |h[n]| ≤ Mp
n=0
 84

- In view of system response

∞
∑ ∞ ∑
∑ M ∑
M ∞
∑ ∞
∑
|h[n]| = Ak (Pk ) u[n] ≤
n
|Ak | (Pk ) n
≤ |Pk |n ≤ MP
n=0 n=0 k=0 k=0 n=0 n=0
∞
∑
⇒ |Pk |n < ∞ ⇒ |Pk | < 1 for k = 0, 1, · · · , M
n=0

⇒ For a causal BIBO system, all poles {Pk } must be within the unit circle |z| < 1.

⇒ Zero locations do not affect stability.

The pole and zero locations of BIBO system.

• Relation between z -transform and DTFT

∞
∑ ∞
∑
H(ω̂) = h[n]e−j ω̂n , H(z) = h[n]z −n ⇒ H(ω̂) = H(z)
z=ej ω̂
n=−∞ n=∞
 85

• Relations and usages of system representations

• Frequency responses at pole and zero frequencies

1 − z −2 (1 − z −1 )(1 + z −1 )
H(z) = =
1 − 0.9z −1 + 0.81z −2 (1 − 0.9ej 3 z −1 )(1 − 0.9e−j 3 z −1 )
π π

( √ √ )
−1 −2 ∗ 0.9 ± (0.9)2 − 4 × 0.81 0.9 ± 0.9 3j ±j π3
1 − 0.9z + 0.81z = 0 ⇒ z = = = 0.9e
2 2

– Zeros : +1(= ej0π ), 1(= e−jπ ), Poles :0.9ej 3 , 0.9e−j 3

π π

– Frequency response at zero freq. ω̂ = 0, π

( ) ( )
H(ej ω̂ ) = |H(z)|z=ej0π = 0, H(ej ω̂ ) = |H(z)|z=ejπ = 0
ω̂=0 ω̂=π

– Frequency response at pole freq. ω̂ = π3 , − π3

( ) 1 − z −2
H(ej ω̂ ) = |H(z)|z=e±j π3 =
ω̂=±π/3 1 − 0.9z −1 + 0.81z −2 π
z=e±j 3

±2π √
|1 − e−j 3 | |1.5 ± j0.3 3|
= ±π = √ = 10.522
|1 − 0.9e−j + 0.81e−±j 3 | |0.145 ± j0.045 3|
2π
3
 86

∗ Bandwidth

– The width of the peak of freq. response.

– Ex: H(z) = (1 − z −2 )/(1 − 0.9z −1 + 0.81z − 2)

⇒ |H(ω̂ ∗ )| = √1 Hpeak = √1 |H(ω̂

2 2
= ± π3 )| = 0.707 · 10.526 = 7.443
⇒ ω̂1∗ = ∗
0.302π , ω̂2 = 0.369π
⇒ 3dB bandwidth : △ ω = |ω̂1∗ − ω̂2∗ | = 0.067π

* In matlab, ’freqz’ (frequency response)

– y[n] = 0.8y[n − 1] + 2x[n] + 2x[n − 1] ⇒ y[n] − 0.8y[n − 1] = 2x[n] + 2x[n − 1]

>> a=[1 -0.8]

>> b=[2 2]
>> [HH theta]=freqz(b, a, [-pi:pi/100:pi]); %frequency range:[-π π ], freq. resolution : π /100.
>> plot(theta, abs(HH))
>> plot(theta, angle(HH))
 87

• Frequency response via z -domain

∏
L
(z − zl )
l=0
H(z) = A where zl = |zl |ejθl , pk = |pk |ejθk
∏
M
(z − pk )
k=0
- Frequency response
∏
L
(ej ω̂ − zl )
l=0
H(ω̂) = H(z) =A
z=ej ω̂
∏
M
(ej ω̂ − Pk )
k=0

- Magnitude of Frequency response : |H(ω̂)|

∏
L ∏
L
(ej ω̂ − zl ) Dzl (ω̂)
l=0
|H(ω̂)| = A , A l=0 where Dzl (ω̂) = |ej ω̂ − zl |, Dpk (ω̂) = |ej ω̂ − pk |
∏
M ∏M
(ej ω̂ − pk ) Dpk (ω̂)
k=0 k=0

( )
∑
L ∑
M
⇒ log |H(ω̂)| = A Dzl (ω̂) − Dpk (ω̂)
l=0 k=0

⇒ As ej ω̂ approches to a zero, Dzl (ω̂) decreases. ⇒ |H(ω̂)| decreases.

⇒ As ej ω̂ approches to a pole, Dpk (ω̂) decreases. ⇒ |H(ω̂)| increases.

Magnitude of frequency response for ω̂ .

 88

– EX : Highpass filter by a zero

– EX : Lowpass filter by a zero

– EX : Highpass filter by a pole

– EX : Lowpass filter by a pole

 89

Digital Signal Processing HW# 7

1. Problem 8.11 on page 313 of textbook

2. Problem 8.13, 8.14 for S1 , S2
3. Problem 8.16, 8.17
4. Problem 8.19