Chapter Five

Transform Analysis Of LTI-System

Lecture #9

Rediet Million

AAiT, School Of Electrical and Computer Engineering

[email protected]

February, 2020

(Rediet Million) DSP-Lecture #9 February, 2020 1 / 29

5.1 Introduction

Given a linear time-invariant system with a unit sample response h(n),

the input and output are related by a convolution sum
P∞
y (n) = h(n) ∗ x(n) = h(k)x(n − k).
k=−∞

The above relationship implies that Y (e jω ) = H(e jω )X (e jω ) where

H(e jω ), the frequency response of the system, is the discrete-time
Fourier transform of h(n).
This relationship between x(n) and y (n) may also be expressed in the
z-transform domain as Y (z) = H(z)X (z) where H(z), the
z-transform of h(n), is the system function of the LTI system.
In this chapter, we look at the characterization of LTI system in terms
of its system function and we also discuss special types of LTI system
such as all-pass systems, minimum phase and linear phase systems.

(Rediet Million) DSP-Lecture #9 February, 2020 2 / 29

5.2 System function and Frequency response

The transfer function of LTI system is the z-transform of a unit

impulse response.
∞
X
H(z) = h(n)z −n
n=−∞
.
For LTI system that are described by a linear constant coefficient
difference equation can be best developed by using the z-transform .
PN M
P
ak y (n − k) = bk x(n − k)
k=0 k=0
I Using linearity and time shifting properties of the z-transform, we
get
N M
ak z −k Y (z) = bk z −k X (z)
P P
k=0 k=0

(Rediet Million) DSP-Lecture #9 February, 2020 3 / 29

System function and Frequency response
M
bk z −k
P
Y (z)
H(z) = = k=0
N
X (z)
ak z −k
P
k=0
M
bk z −k
P
k=0
When a0 = 1 , H(z) = N
........(1)
z −k
P
1+ ak
k=1
I It also convenient to express H(z) in factored form as
b0 M (1 − ck z −1 )
Q
H(z) = QNk=1
a0 k=1 (1 − dk z −1 )
- Each factor (1 − ck z −1 ) in the numerator contributes a zero at z = ck
and a pole at z = 0
- Each factor (1 − dk z −1 ) in the denominator contributes a pole at z = dk
and a zero at z = 0
(Rediet Million) DSP-Lecture #9 February, 2020 4 / 29
System function and Frequency response

From the general form of system function we obtain two important

special forms.

I The first special form is when ak = 0 for 1 ≤ k ≤ N. Equation (1)

reduces to
M 1 P M
bk z −k = bk z M−k
P
H(z) =
k=0 z M k=0

-This forms of H(z) contains M zeros and an M th - order pole at the origin
z =0.
-Since the system contains M non trivial zeros and all poles at z = 0, it is
an all-zero system and has finite-duration impulse response (FIR).

(Rediet Million) DSP-Lecture #9 February, 2020 5 / 29

System function and Frequency response

IThe second special form is when bk = 0 for 1 ≤ k ≤ M, the system

function reduces to
b0 b0 z N
H(z) = N
= N
, a0 = 1
−k N−k
P P
1+ ak z 1+ ak z
k=1 k=0

I This form of H(z) consists of N poles and an N th - order zero at the

origin z = 0. Since the system contains only non-trivial poles and its called
an all-pole system.
I Due to the presence of poles, the impulse response of such a system is
infinite in duration, and hence it is an infinite impulse response (IIR)
system.

(Rediet Million) DSP-Lecture #9 February, 2020 6 / 29

System function and Frequency response

In general, given one description of a system all other equivalent

descriptions for the system can also be determined.

(Rediet Million) DSP-Lecture #9 February, 2020 7 / 29

System function and Frequency response
Stability and Causality

If we assume the system is stable, the impulse response must be

absolutely summable i.e
P∞
|h(n)| < ∞
n=−∞
since this is identical to the condition
∞ ∞
|h(n)z −n | ≤ |h(n)||z −n | < ∞
P P
|H(z)| ≤
n=−∞ n=−∞

I For |z| = 1, the condition for stability is equivalent to the condition that
the ROC of H(z) include the unit circle.

I Because the unit impulse response of a causal system is right-sided,

h(n) = 0 for n < 0, H(z) has a ROC being outside the outermost pole.

(Rediet Million) DSP-Lecture #9 February, 2020 8 / 29

System function and Frequency response
Inverse systems

The inverse system of LTI system H(z) is defined to be the system

with system function Hi (z) such that when cascaded with H(z) the
overall effective system function is unity.
G (z) = H(z)Hi (z) = 1
1 1
Hi (z) = and Hi (e jω ) =
H(z) H(e jω )
I Consider a class of system with rational system function
b0 M (1 − ck z −1 )
Q
H(z) = QNk=1
a0 k=1 (1 − dk z −1 )
-Poles of H(z) becomes the zeros of Hi (z) and the zeros of H(z) becomes
the poles of Hi (z) i.e
a0 N −1 )
Q
k=1 (1 − dk z
Hi (z) =
b0 M −1 )
Q
k=1 (1 − ck z

(Rediet Million) DSP-Lecture #9 February, 2020 9 / 29

System function and Frequency response
Inverse systems

An inverse system, Hi (z), is stable and causal if and only if both the
poles and zeros of H(z) are inside the unit circle, such system are
referred to as a minimum phase system.
I The ROC that is associated with the inverse system is determined
by the requirement that H(z) and Hi (z) have an overlap ROC.
Example : If
1 − 0.5z −1
H(z) = , |z| > 0.8
1 − 0.8z −1
the inverse system is
1 1 − 0.8z −1
Hi (z) = =
H(z) 1 − 0.5z −1
- Two possible ROC for hi (n) i.e |z| > 0.5 and |z| < 0.5. Because
|z| < 0.5 does not overlap the ROC for H(z). The only possible ROC
is |z| > 0.5
hi (n) = (0.5)n u(n) − 0.8(0.5)n−1 u(n − 1)
(Rediet Million) DSP-Lecture #9 February, 2020 10 / 29
System function and Frequency response
Impulse response of rational system function

The partial fraction method maybe used to find the inverse

z-transform of a rational system function of the form
b0 M (1 − ck z −1 )
Q
H(z) = QNk=1
a0 k=1 (1 − dk z −1 )
I Any rational function of z −1 with distinct poles maybe expressed as
M−N N Ak
Bk z −k +
P P
H(z) = −1
k=0 k=1 1 − dk z

-Term in the first summation exists if M ≥ N. If the system is

assumed to be causal, then
M−N N
Ak (dk )n u(n)
P P
h(n) = Bk δ(n − k) +
k=0 k=1

(Rediet Million) DSP-Lecture #9 February, 2020 11 / 29

System function and Frequency response
Impulse response of rational system function

- If H(z) has no poles except at z = 0 then

M
Bk z −k = B0 + B1 z −1 + B2 z −2 + B3 z −3 + ..BM z −M
P
H(z) =
k=0

and the impulse response becomes

M
P
h(n) = Bk δ(n − k) = B0 δ(n) + B1 δ(n − 1) + B2 δ(n − 2) + ..BM δ(n − M)
k=0

- This system is called Finite Impulse Response (FIR) filter.

I Note that for FIR systems the difference equation is identical to the I/O
convolution sum, i.e
M
X
y (n) = Bk x(n − k)
k=0

(Rediet Million) DSP-Lecture #9 February, 2020 12 / 29

System function and Frequency response
Frequency response for for rational system functions

The frequency response of stable systems is obtained by evaluating

the rational transfer function H(z) on the unit circle.
H(ω) = H(z)|z=e jω = |H(ω)|e jϕ(ω)
M
bk e −jωk
P
b0 M (1 − ck e −jω )
Q
k=0
H(ω) = N = QNk=1
a0 k=1 (1 − dk e −jω )
ak e −jωk
P
k=0
I The magnitude of the response, H(ω), is
QM −jω )|
b0 k=1 |(1 − ck e
|H(ω)| = | | QN
a0 k=1 |(1 − dk e
−jω )|
I The magnitude-squared function is defined as
b0 2 M (1 − ck e −jω )(1 − ck∗ e jω )
Q
∗
|H(ω)| = H(ω)H (ω) = | | QNk=1
2
a0 −jω )(1 − d ∗ e jω )
k=1 (1 − dk e k

(Rediet Million) DSP-Lecture #9 February, 2020 13 / 29

System function and Frequency response
Frequency response for for rational system functions

I The logarithm of the magnitude-squared function, referred to as the log

magnitude or gain of H(ω) and expressed in dB, is
Gain in dB=10 log10 |H(ω)|2 = 20 log10 |H(ω)|

b0 M M
20 log10 |(1 − ck e −jω )| − 20 log10 |(1 − dk e −jω )|
P P
= 20 log10 | |+
a0 k=1 k=1
I Similarly, the phase response for a rational system function will have the
form

I Another important parameter associated with a rational system function

is the group delay, τ , defined as
d
τ (ω) = − ϕ(ω)
dω

(Rediet Million) DSP-Lecture #9 February, 2020 14 / 29

System function and Frequency response

Example-1:
Suppose the system function of a LTI system is given by

(1 + z −1 )2
H(z) =
1 3
(1 − z −1 )(1 + z −1 )
2 4
Find the i) I/O difference equation, ii) pole/zero pattern and frequency
response, iii) impulse response, iv) I/O convolution equation.

i) I/O difference equation

- H(z) may be represented as
Y (z) 1 + 2z −1 + z −2
H(z) = =
X (z) 1 3
1 + z −1 − z −2
4 8
- Cross multiplying, we get
1 3
Y (z)(1 + z −1 − z −2 ) = X (z)(1 + 2z −1 + z −2 )
4 8
(Rediet Million) DSP-Lecture #9 February, 2020 15 / 29
System function and Frequency response

- Inverse transforming both sides using time shift property of the

z-transform, we get the difference equation
1 3
y (n) + y (n − 1) − y (n − 2) = x(n) + 2x(n − 1) + x(n − 2)
4 8
ii) pole/zero pattern and frequency response
- All the poles of H(z) are inside the unit circle. Assuming a causal
system, the frequency response of the system is obtained by replacing z by
e jω into H(z), i.e
(1 + e jω )2
H(e jω ) =
1 3
(1 − e jω )(1 + e jω )
2 4
- Using the identity, for any
p real a,
|1 − ae jω | = 1 − 2a cos(ω) + a2

(Rediet Million) DSP-Lecture #9 February, 2020 16 / 29

System function and Frequency response

- We get the magnitude response as

2 + 2 cos(ω)
|H(e jω )| = r
1 3 9
[1 − cos(ω) + ][1 + cos(ω) + ]
4 4 10
32
- At ω = 0 and z = 1, |H(e jω )|w =0 = |H(z)|z=1 =
7
and
- At ω = π and z = −1, |H(e jω )|w =π = |H(z)|z=−1 = 0

(Rediet Million) DSP-Lecture #9 February, 2020 17 / 29

System function and Frequency response

iii) impulse response

To obtain the impulse response, we use partial fraction expansion of H(z),
i.e
1 + 2z −1 + z −2 A1 A2
H(z) = = A0 + +
1 3 1 3
1 + z −1 − z −2 1 − z −1 1 + z −1
4 8 2 4
where the partial fraction coefficients are obtained as follows :
(z + 1)2 8
A0 = H(z)z=0 = |z=0 = −
1 3 3
(z − )(z + )
2 4
1 18
A1 = (1 − z −1 )H(z)| 1 =
2 z= 5
2
3 1
A2 = (1 + z −1 )H(z)| 3 =
4 z=− 15
4

(Rediet Million) DSP-Lecture #9 February, 2020 18 / 29

System function and Frequency response

i.e
18 1
8 5 15
H(z) = − + +
3 1 −1 3
1− z 1 + z −1
2 4
Thus, the impulse response is, assuming the system is causal,
8 18 1 1 3
h(n) = − δ(n) + ( )n u(n) + (− )n u(n)
3 5 2 15 4
iii) The I/O convolution equation relates input, output, and impulse
response and has the form
∞
P
y (n) = h(k)x(n − k)
k=0

y (n) = h0 xn + h1 xn−1 + h2 xn−2 + h3 xn−3 ...

Thus,
y (n) = xn + 1.75xn−1 + 0.9375xn−2 + ..
(Rediet Million) DSP-Lecture #9 February, 2020 19 / 29
 (#1 ) Class exercises & Assignment
1) When the input to a linear time-invariant system is x(n) and the output
is y (n)
a. x(n) = (1/2)n u(n) + (2)n u(−n − 1)
y (n) = 6(1/2)n u(n) − 6(3/4)n u(n)
b. x(n) = 5u(n)
y (n) = 2(1/2)n u(n) + 3(−3/4)n u(n)
i. Find the system function H(z) of the system.Plot the pole-zero diagram
and indicate ROC.
ii. Find the impulse response h(n) of the system for all value of n.
iii. Write the difference equation that characterize the system.
iv. Is the system stable ? is it causal ?
2) Find the inverse of
1 − 2z −1
H(z) = , |z| > 0.6
1 − 0.6z −1

(Rediet Million) DSP-Lecture #9 February, 2020 20 / 29

5.2 All-pass and Minimum-phase systems
All-pass systems

A system for which the frequency response magnitude is a constant is

called an all-pass system.
Consider the system function H(z) characterized by one pole at z = a
1
and its conjugate reciprocal zero at z = ∗
a
1
z −1 − a∗ −a∗ (1 − ∗ z −1 )
Hap (z) = = a
1 − az −1 1 − az −1
I This condition on the poles and zeros states that if one is inside the
unit circle then the other must be outside the unit circle.

(Rediet Million) DSP-Lecture #9 February, 2020 21 / 29

All-pass and Minimum-phase systems
All-pass systems

I If we put causality and stability requirements, then each pole has to be

inside the unit circle and as a result each conjugate reciprocal zero will be
outside the unit circle.

(Rediet Million) DSP-Lecture #9 February, 2020 22 / 29

All-pass and Minimum-phase systems
All-pass systems

I The frequency response of the all-pass system will be

e −jω − a∗ 1 − a∗ e jω
Hap (e jω ) = −jω
= e −jω ( )
1 − ae 1 − ae −jω
I The Magnitude of H(e jω )
1 − a∗ e jω
|H(e jω )| = |e −jω || |=1
1 − ae −jω
since |e −jω | and |1 − a∗ e jω | = |1 − ae −jω |
- Such a system is called an all- pass system since it has unity magnitude
gain independent of the frequency, ω.
I The general form of the system function of an all-pass system, with Mr
real poles and Mc complex pole pairs, is
Mr M
Y z −1 − dk Yc (z −1 − ek∗ )(z −1 − ek )
H(z) =
1 − dk z−1 (1 − ek z −1 )(1 − ek∗ z −1 )
k=1 k=1
Where the dk0 s are real poles and ek0 s are the complex poles of Hz
(Rediet Million) DSP-Lecture #9 February, 2020 23 / 29
All-pass and Minimum-phase systems
All-pass systems

I The phase function, ϕ, for a first order all-pass system with a pole
z = a = re jθ is expressed as

- and the group delay is

(Rediet Million) DSP-Lecture #9 February, 2020 24 / 29

All-pass and Minimum-phase systems
All-pass systems

Example:
Show that the following causal LIT system described by this difference
equation represent an all-pass filter.
1 1
y (n) − y (n − 2) = x(n − 2) − x(n)
4 4
-Taking the z-transform of both sides and rearranging
Y (z) −1/4 + z −2
H(z) = =
X (z) 1 − 1/4z −2
1
Since two poles, at z = ± , and two zeros, at z = ±2, occurs in
2
conjugate reciprocal pairs the system is an all-pass system.

(Rediet Million) DSP-Lecture #9 February, 2020 25 / 29

All-pass and Minimum-phase systems
Minimum-phase system

A rational system function with all of its poles and zeros inside the
unit circle is said to be minimum phase system.
I A minimum phase system is uniquely defined by the magnitude of its
Fourier transform, |H(e jω |. The procedure to find H(z) from |H(e jω )| is as
follows.
- Given |H(e jω |, we find |H(e jω |2 , which is a function of cos(kω).
1 k
-Then by replacing cos(kω) with (z + z −k ), we have
2
G (z) = H(z)H(z −1 )
- Finally, the minimum phase system is then formed from the poles and
zeros of G (z) that are inside the unit circle.

(Rediet Million) DSP-Lecture #9 February, 2020 26 / 29

All-pass and Minimum-phase systems
Minimum phase and all-pass decomposition

A stable and causal system may be factored in to a product of a

minimum phase and all-phase system
H(z) = Hmin (z).Hap (z)
1
I Suppose H(z) has one zero outside the unit circle at z = , |c| < 1
c∗
H(z) = H1 (z)(z −1 − c ∗ )
(z −1 − c ∗ )
= H1 (z)(1 − cz −1 )
| {z } (1 − cz −1 )
| {z }
Hmin Hap
I The procedure for performing this factorization is as follows
- First all other zeros of H(z) that are outside the unit circle are reflected
inside the unit circle to their conjugate reciprocal
- Then, the all-pass filter is selected so that it reflect the appropriate set of
zeros of Hmin (z) back outside the unit circle.
(Rediet Million) DSP-Lecture #9 February, 2020 27 / 29
 Chapter Five reading assignment
Linear Systems with Generalized Linear Phase

(Rediet Million) DSP-Lecture #9 February, 2020 28 / 29

 (#2 ) Class exercises & Assignment
1) For each of the following system functions, state whether or not it is a
minimum-phase system . Justify your answer
(1 − 2z −1 )(1 + 1/2z −1 ) (1 + 1/4z −1 )(1 − 1/4z −1 )
a. H(z) = b. H(z) =
(1 − 1/3z −1 )(1 + 1/3z −1 ) (1 − 2/3z −1 )(1 + 2/3z −1 )
(1 − 1/3z −1 ) z −1 (1 − 1/3z −1 )
a. H(z) = b. H(z) =
(1 − 1/2z −1 )(1 + 1/2z −1 ) (1 − 1/2z −1 )(1 + 1/2z −1 )
2) Find the minimum phase system that has a magnitude response given
by
5
17 1 − cos(ω)
jω 2
a. |H(e )| = − cos(ω) b. |H(e )| = 4
jω 2
16 2 10 2
− cos(ω)
9 3
3) A discrete time causal LTI system has the system function
(1 + 0.2z −1 )(1 − 9z −2 ) (1 + 3z −1 )
a.H(z) = b.H(z) =
(1 + 0.8z −2 ) (1 + 1/2z −1 )
find the expression for a minimum phase system, Hmin (z), and all-pass
system Hap (z) such that H(z) = Hmin (z).Hap (z)
(Rediet Million) DSP-Lecture #9 February, 2020 29 / 29