CHAPTER 8

IIR Filter
Design

IIR filters have infinite-duration impulse responses, and hence they can
be matched to analog filters, all of which generally have infinitely long im-
pulse responses. Therefore, the basic technique of IIR filter design trans-
forms well-known analog filters into digital filters using complex-valued
mappings. The advantage of this technique lies in the fact that both
analog filter design (AFD) tables and the mappings are available exten-
sively in the literature. This basic technique is called the A/D (analog-
to-digital) filter transformation. However, the AFD tables are available
only for lowpass filters. We also want to design other frequency-selective
filters (highpass, bandpass, bandstop, etc.). To do this, we need to apply
frequency-band transformations to lowpass filters. These transformations
are also complex-valued mappings, and they are also available in the liter-
ature. There are two approaches to this basic technique of IIR filter design:
Approach 1:

Apply Frequency-Band Apply Filter

Design Analog Desired IIR
−→ Transformation −→ Transformation −→
Lowpass Filter Filter
s→s s→z

Approach 2:

Apply Filter Apply Frequency-Band

Design Analog Desired IIR
−→ Transformation −→ Transformation −→
Lowpass Filter Filter
s→z z→z

370
Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 Some Preliminaries 371

The first approach is used in MATLAB to design IIR filters. A

straightforward use of these MATLAB functions does not provide any
insight into the design methodology. Therefore, we will study the second
approach because it involves the frequency-band transformation in the
digital domain. Hence in this IIR filter design technique, we will follow
the following steps:
• Design analog lowpass filters.
• Study and apply filter transformations to obtain digital lowpass filters.
• Study and apply frequency-band transformations to obtain other digi-
tal filters from digital lowpass filters.
The main problem with these approaches is that we have no control
over the phase characteristics of the IIR filter. Hence IIR filter designs
will be treated as magnitude-only designs. More sophisticated techniques,
which can simultaneously approximate both the magnitude and the phase
responses, require advanced optimization tools and hence will not be cov-
ered in this book.
We begin with a discussion on the analog filter specifications and the
properties of the magnitude-squared response used in specifying analog
filters. Next, before we delve into basic techniques for general IIR filters,
we consider the design of special types of digital filters—for example,
resonators, notch filters, comb filters, and so on. This is followed by a
brief description of the characteristics of three widely used analog fil-
ters: Butterworth, Chebyshev, and elliptic filters. Finally, we will study
transformations to convert these prototype analog filters into different
frequency-selective digital filters and conclude this chapter with several
IIR filter designs using MATLAB.

8.1 SOME PRELIMINARIES

We discuss two preliminary issues in this section. First, we consider the

magnitude-squared response specifications, which are more typical of ana-
log (and hence of IIR) filters. These specifications are given on the relative
linear scale. Second, we study the properties of the magnitude-squared
response.

8.1.1 RELATIVE LINEAR SCALE

Let Ha (jΩ) be the frequency response of an analog filter. Then the lowpass
filter specifications on the magnitude-squared response are given by
1
≤ |Ha (jΩ)|2 ≤ 1, |Ω| ≤ Ωp
1 + 2 (8.1)
1
0 ≤ |Ha (jΩ)|2 ≤ 2 , Ωs ≤ |Ω|
A

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 372 Chapter 8 IIR FILTER DESIGN

|Ha( j Ω)|2

1
1+

1
A2
0 Ω
0 Ωp Ωs

FIGURE 8.1 Analog lowpass filter specifications

where is a passband ripple parameter, Ωp is the passband cutoff fre-

quency in rad/sec, A is a stopband attenuation parameter, and Ωs is the
stopband cutoff in rad/sec. These specifications are shown in Figure 8.1,
2
from which we observe that |Ha (jΩ)| must satisfy

1
|Ha (jΩp )|2 = at Ω = Ωp
1 + 2
(8.2)
1
|Ha (jΩs )|2 = at Ω = Ωs
A2
The parameters and A are related to parameters Rp and As , respec-
tively, of the dB scale. These relations are given by

1 7
Rp = −10 log10 2
=⇒ = 10Rp /10 − 1 (8.3)
1+

and
1
As = −10 log10 =⇒ A = 10As /20 (8.4)
A2
The ripples, δ1 and δ2 , of the absolute scale are related to and A by
 √
1 − δ1 1 2 δ1
= =⇒ =
1 + δ1 1 + 2 1 − δ1

and
δ2 1 1 + δ1
= =⇒ A =
1 + δ1 A δ2

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 Some Preliminaries 373

8.1.2 PROPERTIES OF |Ha (jΩ)|2

Analog filter specifications (8.1), which are given in terms of the
magnitude-squared response, contain no phase information. Now to eval-
uate the s-domain system function Ha (s), consider

Ha (jΩ) = Ha (s)|s=jΩ

Then we have

|Ha (jΩ)| = Ha (jΩ)Ha∗ (jΩ) = Ha (jΩ)Ha (−jΩ) = Ha (s)Ha (−s)|s=jΩ

2

or

2
Ha (s)Ha (−s) = |Ha (jΩ)|  (8.5)
Ω=s/j

Therefore, the poles and zeros of the magnitude-squared function are dis-
tributed in a mirror-image symmetry with respect to the jΩ axis. Also for
real filters, poles and zeros occur in complex conjugate pairs (or mirror-
image symmetry with respect to the real axis). A typical pole-zero pat-
tern of Ha (s)Ha (−s) is shown in Figure 8.2. From this pattern, we can
construct Ha (s), which is the system function of our analog filter. We
want Ha (s) to represent a causal and stable filter. Then all poles of Ha (s)
must lie within the left half-plane. Thus we assign all left-half poles of
Ha (s)Ha (−s) to Ha (s). However, zeros of Ha (s) can lie anywhere in the
s-plane. Therefore, they are not uniquely determined unless they all are
on the jΩ axis. We will choose the zeros of Ha (s)Ha (−s) lying left to or
on the jΩ axis as the zeros of Ha(s). The resulting filter is then called a
minimum-phase filter.

jΩ

s-plane

FIGURE 8.2 Typical pole-zero pattern of Ha (s)Ha (−s)

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 374 Chapter 8 IIR FILTER DESIGN

8.2 SOME SPECIAL FILTER TYPES

In this section, we consider the design of several special types of digital

filters and describe their frequency response characteristics. We begin by
describing the design and characteristics of a digital resonator.

8.2.1 DIGITAL RESONATORS

A digital resonator is a special two-pole bandpass filter with a pair of
complex-conjugate poles located very near the unit circle, as shown in
the left plot of Figure 8.3. The magnitude of the frequency response of
the filter is shown in the top right plot of Figure 8.3. The name resonator
refers to the fact that the filter has a large magnitude response in the
vicinity of the pole position. The angle of the pole location determines
the resonant frequency of the filter. Digital resonators are useful in many
applications, including simple bandpass filtering and speech generation.
Let us consider the design of a digital resonator with a resonant peak
at or near ω = ω0 . Hence we select the pole position as

p1,2 = re±jω0 (8.6)

Digital Resonator Responeses

Pole-Zero Plot Magnitude Response

1 1
Magnitude

0.8
0.8
0.6
0.6 0.4
Imaginary Part

0.4 0.2
0
0.2 –1 –1/3 0 1/3 1
0
2
Phase Response
–0.2
0.5
Radians / π

–0.4
–0.6
0
–0.8
–1
–0.5
–1 –0.5 0 0.5 1 –1 –1/3 0 1/3 1
Real Part ω in π Units
FIGURE 8.3 Pole positions and frequency response of a digital resonator with
r = 0.9 and ω0 = π/3

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 Some Special Filter Types 375

The corresponding system function is

b0
H(z) = −1
(1 − jω
re z )(1
0 − re−jω0 z −1 )
b0
= (8.7)
1 − (2r cos ω0 )z −1 + r2 z −2

where b0 is a gain parameter. The frequency response of the resonator is

  b
H ejω =   0  (8.8)
1− re−j(ω−ω0 ) 1 − re−j(ω+ω0 )
  
Since H ejω  has its peak
 at or near ω = ω0 , we select the gain param-

eter b0 so that H e jω0 
= 1. Hence
  jω  b0
H e 0  =
|(1 − r)(1 − re−j2ω0 )|
b0
= √ (8.9)
(1 − r) 1 + r2 − 2r cos 2ω0

Consequently, the desired gain parameter is

7
b0 = (1 − r) 1 + r2 − 2r cos 2ω0 (8.10)

The magnitude of the frequency response H(ω) may be expressed as

  jω  b0
H e  = (8.11)
D1 (ω)D2 (ω)

where D1 (ω) and D2 (ω) are given as

7
D1 (ω) = 1 + r2 − 2r cos(ω − ω0 ) (8.12a)
7
D2 (ω) = 1 + r2 − 2r cos(ω + ω0 ) (8.12b)

For a given value of r, D1 (ω) takes its minimum value (1 − r) at ω = ω0 ,

and the product D1 (ω)D2 (ω) attains a minimum at the frequency
 
−1 1 + r2
ωr = cos cos ω0 (8.13)
2r

which defines precisely the resonant frequency of the filter. Note that
when r is very close to unity, ωr ≈ ω0 , which is the angular position of
the pole. Furthermore, as r approaches unity, the resonant peak becomes
sharper (narrower) because D1 (ω) changes rapidly in the vicinity of ω0 .

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 376 Chapter 8 IIR FILTER DESIGN

A quantitative measure of the width of the peak is the 3 dB bandwidth

of the filter, denoted as ∆(ω). For values of r close to unity,

∆ω ≈ 2(1 − r) (8.14)

Figure 8.3 illustrates the magnitude and phase responses of a digital res-
onator with ω0 = π/3, r = 0.90. Note that the phase response has its
greatest rate of change near the resonant frequency ωr ≈ ω0 = π/3.
This resonator has two zeros at z = 0. Instead of placing zeros at the
origin, an alternative choice is to locate the zeros at z = 1 and z = −1.
This choice completely eliminates the response of the filter at the frequen-
cies ω = 0 and ω = π, which may be desirable in some applications. The
corresponding resonator has the system function

G(1 − z −1 )(1 + z −1 )
H(z) =
(1 − rejω0 z −1 )(1 − re−jω0 z −1 )
1 − z −2
=G (8.15)
1 − (2r cos ω0 )z −1 + r2 z −2

and the frequency response characteristic

  1 − e−j2ω
H ejω = G (8.16)
[1 − rej(ω0 −ω) ][1 − re−j(ω0 +ω) ]
  
where G is a gain parameter that is selected so that H ejω0  = 1.
The introduction of zeros at z = ±1 alters both the magnitude and
phase response of the resonator. The magnitude response may be ex-
pressed as
  jω  N (ω)
H e  = G (8.17)
D1 (ω)D2 (ω)
where N (ω) is defined as
7
N (ω) = 2(1 − cos 2ω) (8.18)

Due to the presence of the zeros at z = ±1, the resonant frequency of the
resonator is altered from the expression given by (8.13). The bandwidth
of the filter is also altered. Although exact values for these two parameters
are rather tedious to derive, we can easily compute the frequency response
when the zeros are at z = ±1 and z = 0 and compare the results.
Figure 8.4 illustrates the magnitude and phase responses for the cases
z = ±1 and z = 0, for pole location at ω = π/3 and r = 0.90. We observe
that the resonator with z = ±1 has a slightly smaller bandwidth than
the resonator with zeros at z = 0. In addition, there appears to be a very
small shift in the resonant frequency between the two cases.

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 Some Special Filter Types 377

Magnitude Response
1

Magnitude
0.8
0.6
0.4
0.2
0
–1 –1/3 0 1/3 1

Phase Response
0.5
Radians/ π

–0.5
–1 –1/3 0 1/3 1
ω in π Units
FIGURE 8.4 Magnitude and phase responses of digital resonator with zeros at
z = ±1 (solid lines) and z = 0 (dotted lines) for r = 0.9 and ω0 = π/3

8.2.2 NOTCH FILTERS

A notch filter is a filter that contains one or more deep notches or, ideally,
perfect nulls in its frequency response. Figure 8.5 illustrates the frequency
response of a notch filter with a null at the frequency ω = ω0 . Notch filters
are useful in many applications where specific frequency components must
be eliminated. For example, instrumentation systems require that the
power line frequency of 60 Hz and its harmonics be eliminated.
To create a null in the frequency response of a filter at a frequency
ω0 , we simply introduce a pair of complex-conjugate zeros on the unit
circle at the angle ω0 . Hence the zeros are selected as

z1,2 = e±jω0 (8.19)

Then the system function for the notch filter is

H(z) = b0 (1 − ejω0 z −1 )(1 − e−jω0 z −1 )

= b0 (1 − (2 cos ω0 )z −1 + z −2 ) (8.20)

where b0 is a gain factor. Figure 8.6 illustrates the magnitude response of

a notch filter having a null at ω = π/4.
The major problem with this notch filter is that the notch has a rela-
tively large bandwidth, which means that other frequency components

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 378 Chapter 8 IIR FILTER DESIGN

Pole-Zero Plot Magnitude Response

10
1 0

Decibels
0.8
0.6
0.4
Imaginary Part

0.2 –50
–1 0 1
0
2
Phase Response
–0.2
1

Radians / π
–0.4
0.5
–0.6
0
–0.8
–0.5
–1
–1
–1 –0.5 0 0.5 1 –1 0 1
Real Part ω in π Units
FIGURE 8.5 Frequency response of a typical notch filter

around the desired null are severely attenuated. To reduce the band-
width of the null, we may resort to the more sophisticated, longer FIR
filter designed according to the optimum equiripple design method de-
scribed in Chapter 7. Alternatively, we could attempt to improve the fre-
quency response of the filter by introducing poles in the system function.

Pole-Zero Plot Magnitude Response

10
1 0
Decibels

0.8
0.6
0.4
Imaginary Part

0.2 –50
–1 –1/4 0 1/4 1
0
2
Phase Response
–0.2
1
–0.4
Radians / π

0.5
–0.6
0
–0.8
–0.5
–1
–1
–1 –0.5 0 0.5 1 –1 –1/4 0 1/4 1
Real Part ω in π Units
FIGURE 8.6 Frequency response of a notch filter with ω0 = π/4

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 Some Special Filter Types 379

Pole-Zero Plot Magnitude Response

10
1 0

Decibels
0.8
0.6
0.4
Imaginary Part

0.2 –50
–1 –1/4 0 1/4 1
0
Phase Response
–0.2
1
–0.4

Radians / π
0.5
–0.6
0
–0.8
–0.5
–1
–1
–1 –0.5 0 0.5 1 –1 –1/4 0 1/4 1
Real Part ω in π Units
FIGURE 8.7 Magnitude and phase responses of notch filter with poles (solid
lines) and without poles (dotted lines) for ω0 = π/4 and r = 0.85

In particular, suppose that we select the poles at

p1,2 = re±jω0 (8.21)

Hence the system function becomes

1 − (2 cos ω0 )z −1 + z 2
H(z) = b0 (8.22)
1 − (2r cos ω0 )z −1 + r2 z −2
  
The magnitude of the frequency response H ejω  of this filter is illus-
trated in Figure 8.7 for ω0 = π/4 and r = 0.85. Also plotted in this figure
is the frequency response without the poles. We observe that the effect of
the pole is to introduce a resonance in the vicinity of the null and, thus, to
reduce the bandwidth of the notch. In addition to reducing the bandwidth
of the notch, the introduction of a pole in the vicinity of the null may re-
sult in a small ripple in the passband of the filter due to the resonance
created by the pole.

8.2.3 COMB FILTERS

In its simplest form, a comb filter may be viewed as a notch filter in which
the nulls occur periodically across the frequency band, hence the analogy
to an ordinary comb that has periodically spaced teeth. Comb filters are
used in many practical systems, including the rejections of power-line
harmonics and the suppression of clutter from fixed objects in moving-
target indicator (MTI) radars.

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 380 Chapter 8 IIR FILTER DESIGN

We can create a comb filter by taking our FIR filter with system
function
M
H(z) = h(k)z −k (8.23)
k=0
L
and replacing z by z , where L is a positive integer. Thus the new FIR
filter has the system function

M
HL (z) = h(k)z −kL (8.24)
k=0
 
If the frequency response of the original FIR filter is H ejω , the frequency
response of the filter given by (8.24) is

  M
 
HL ejω = h(k)e−jkLω = H ejLω (8.25)
k=0
 
Consequently, the
 jωfrequency
 response characteristic HL ejω is an L-order
repetition of H e in the
 range
 0≤ ω ≤ 2π. Figure 8.8 illustrates the
relationship between HL ejω and H ejω for L = 4. The introduction of
a pole at each notch may be used to narrow the bandwidth of each notch,
as just described.

 
FIGURE 8.8 Comb filters with frequency response HL ejω obtained from
 jω 
H e for L = 4

Copyright 201 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.