Chapter 7

IIR Filters

7.1 Introduction
The specification of a digital filter will be desired frequency response, Hd(ejw ). The desired impulse
response, hd(n) of the digital filter can be obtained by taking inverse Fourier transform of Hd(ejw ). Now, the
hd(n) will be an infinite duration discrete time signal defined for all values of n in the range –¥ to +¥ . The
filters designed by considering all the infinite samples of impulse response are called IIR (Infinite Impulse
Response) filters.
In digital domain, the processing of infinite samples of impulse response is practically not possible.
Hence direct design of IIR filter is not possible. Therefore, the IIR filters are designed via analog filters.
In design of IIR filter, the specification of an IIR filter is transformed to specification of an analog filter
and an analog filter with transfer function, H(s) is designed to satisfy the specification. Then the analog filter
is transformed to digital filter with transfer function, H(z).
We know that the analog filter with transfer function H(s) is stable if all its poles lie in the left half of the
s-plane. Consequently, if the conversion technique is to be effective, it should possess the following desirable
properties.
1. The imaginary axis in the s-plane should map into the unit circle in the z-plane. Thus there will
be a direct relationship between the two frequency variables in the two domains.
2. The left-half of the s-plane should map into the interior of the unit circle in the z-plane. Thus a
stable analog filter will be converted to a stable digital filter.
The analog filter is designed by approximating the ideal frequency response using an error function.
A number of solutions to the approximation problem of analog filter design are well developed. The popular
among them are Butterworth and Chebyshev approximation. The popular transformation techniques used for
transforming analog filter transfer function H(s) to digital filter transfer function H(z) are bilinear and impulse
invariant transformation. The digital transfer function H(z) can be realized in a software that runs on a digital
hardware (or it can be implemented in firmware).
The frequency response H(ejw ) of the digital filter can be obtained by letting z = ejw in the transfer
function H(z) of the filter.
Chapter 7 - IIR Filters 7. 2
The designed transfer function of the filter should represent a stable and causal system. For stability
and causality of analog filter, the analog transfer function should satisfy the following requirements.
1. The H(s) should be a rational function of "s" and the coefficients of "s" should be real.
2. The poles should lie on the left half of s-plane.
3. The number of zeros should be less than or equal to number of poles.
For stability and causality of digital filter, the digital transfer function should satisfy the following
requirements.
1. The H(z) should be a rational function of "z" and the coefficients of "z" should be real.
2. The poles should lie inside the unit circle in z-plane.
3. The number of zeros should be less than or equal to number of poles.
Advantages of Digital Filters
1. The values of resistors,capacitors and inductors used in the analog filters changes with
temperature. Since digital filters do not have these components, they have high thermal
stability.
2. In digital filters the precision of the filter depends on the length (or size) of the registers used to
store the filter coefficients. Hence by increasing the register bit-length (in hardware) the
performance characteristics of the filter like accuracy, dynamic range, stability and frequency
response tolerance, can be enhanced.
3. The digital filters are programmable. Hence the filter coefficients can be changed at any time to
implement adaptive features.
4. A single filter can be used to process multiple signals by using the techniques of multiplexing.
Disadvantages of Digital Filters
1. The bandwidth of the discrete signal is limited by the sampling frequency. The bandwidth of
real discrete signal is half the sampling frequency.
2. The performance of the digital filter depends on the hardware (i.e., depends on the bit length of
the registers in the hardware) used to implement the filter.
Important Features of IIR Filters
1. The physically realizable IIR filters do not have linear phase.
2. The IIR filter specifications include the desired characteristics for the magnitude response
only.
Table 7.1 : Comparison of Digital and Analog Filters

Digital Filter Analog Filter

1. Operates on digital samples 1. Operates on analog signals
(or sampled version) of the signal. (or actual signals).
2. It is governed (or defined) by linear 2. It is governed (or defined) by linear
difference equation. differential equation.
7. 3 Digital Signal Processing
Table 7.1 : continued...
Digital Filter Analog Filter
3. It consists of adders, multipliers and 3. It consists of electrical components
delays implemented in digital logic like resistors, capacitors and inductors.
(either in hardware or software or both)
4. In digital filters the filter coefficients 4. In analog filters the approximation
are designed to satisfy the desired problem is solved to satisfy the desired
frequency response. frequency response.

7.2 Frequency Response of Analog and Digital IIR Filters

The filters are frequency selective devices and so they are designed to pass the spectral content of the
input signal in a specified band of frequencies. Hence, based on frequency response the filters are classified
into four basic types. They are lowpass, highpass, bandpass and bandstop filters.
The ideal magnitude response, |Hd(jW )| of the four basic types of analog filters are shown in fig 7.1 (a),
(b), (c) and (d). The ideal magnitude response has sudden transition from passband to stopband which is
practically not realizable. Hence the ideal response is approximated using a filter approximation function.
The approximation problem is solved to meet a specified tolerance in the passband and stopband. The
shaded areas in the fig 7.1 shows the tolerance regions of the ideal frequency response. In the passband the
magnitude is approximated to unity within an error of dp. In the stopband the magnitude is approximated to
zero within an error of ds. Here the dp and ds are the limits of the tolerance in the passband and stopband. The
dp and ds are also called ripples.
The magnitude response of practical or approximated analog filters, |H(jW )| are shown in fig 7.1 (e), (f),
(g) and (h). The frequency repsonse of practical analog filter shows edges for passband and stopband so that
the tolerances are within specified limits. Now, the specification of practical analog filter will be the following.
W p = Passband edge frequency in rad/second.
W s = Stopband edge frequency in rad/second.
Ap = Gain at passband edge frequency
As = Gain at stopband edge frequency
The ideal magnitude response, |Hd(ejw )| of the four basic types of digital IIR filters are shown in fig 7.2
(a), (b), (c) and (d). The ideal magnitude response has sudden transition from passband to stopband which is
practically not possible. The transformation of analog to digital filter will preserve the magnitude response,
and so the magnitude response of digital filter will be similar to analog filter, (but the frequency response of
digital filter is periodic with period 2p). Therefore the practical frequency response of digital IIR filters will be
similar to analog filter as shown in fig 7.2 (e), (f), (g) and (h). The magnitude response of practical digital IIR
filter shows edges for passband and stopband so that the tolerances are within specified limits. Now, the
specifications of practical digital IIR filter will be the following.
w p = Passband edge frequency in rad/sample
w s = Stopband edge frequency in rad/sample
Ap = Gain at passband edge frequency
As = Gain at stopband edge frequency
Chapter 7 - IIR Filters 7. 4
|H (j Ω)|
|H d (j Ω)|
1
δp
1 Ap
A p = 1 − δp
1 A s = δs
= 0.707
2

As
δs
0 Ωp Ωc Ωs Ω
0 Ωc Ω
P as sband T ransition S topband
P as sband S topband band
F ig a : N orm alized m ag n itud e respo n se o f id ea l F ig e : N orm alized m ag n itud e respo n se o f p ra ctica l
a na log lo w pa ss filter. a na log lo w pa ss filter.
|H d (j Ω)|
|H (j Ω)|
1 1
δp
Ap
1
= 0.707
2

δs
As

0 Ωc Ω 0 Ωs Ωc Ωp Ω
S topband P as sband S topband T ransition P as sband
band
F ig b : N orm alized m ag n itud e respo n se o f id eal F ig f : N orm alized m ag n itud e respo n se o f p ra ctical
a na log h ig hp ass filter. a na log h ig hp ass filter.

|H (j Ω)|
1
|H d (j Ω)| δp
Ap
1 1
= 0.707
2

As
δs
0 Ω

0 Ωc2
Ω
Ωc1 P as sband

S topband P as sband S topband S topband

F ig c : N orm alized m ag n itud e respo n se o f id eal F ig g : N orm alized m ag n itud e respo n se o f p ra ctical
a na log b a nd pa ss filter. a na log b a nd pa ss filter.

|H (j Ω)|
1
|H d (j Ω)| δp
Ap
1 1
= 0.707
2

As
δs
0 Ω

0 Ωc2
Ω
Ωc1
S topband

P as sband S topband P as sband P as sband

F ig d : N orm alized m ag n itud e respo n se o f id eal F ig h : N orm alized m ag n itud e respo n se o f p ra ctical
a na log b an d sto p filter. a na log b an d sto p filter.
F ig 7.1 : N o rm a lized freq u en c y resp o nse of id ea l a n d pra ctica l a n alo g filte rs.
7. 5 Digital Signal Processing
jω
jω
|H d (e )| |H (e )|
1 δp
1
Ap
A p = 1 − δp
1 A s = δs
= 0.707
2

As
δs
ωp ωc ωs ω
0 ωc
ω 0

P as sband T ransition S topband

P as sband S topband
band
F ig a : N orm alized m ag n itud e respo n se o f id eal F ig e : N orm alized m ag n itud e respo n se o f p ra ctical
d igital IIR lo w p ass filter. d igital IIR lo w p ass filter.
jω
|H d ( e )| jω
|H ( e )|
1 1
δp
Ap
1
= 0.707
2

δs
As

ω ωs ωc ωp ω
0 ωc 0

S topband P as sband S topband T ransition P as sband

band
F ig b : N orm alized m ag n itud e resp on se o f id eal F ig f : N orm alized m ag n itud e resp on se o f p ra ctica l
d igital IIR high pa ss filter. d igital IIR high pa ss filter.
jω
|H ( e )|
jω
1
|H d ( e )| δp
Ap
1 1
= 0.707
2

As
δs
0
ω

ωc1 ωc2
ω
0 P as sband

S topband P as sband S topband S topband

F ig c : N orm alized m ag n itud e respo n se o f id eal F ig g : N orm alized m ag n itud e respo n se o f p ractica l
d igital IIR ba n dp ass filter. d igital IIR ba n dp ass filter.

jω
|H ( e )|
1
jω
|H d ( e )| δp
Ap
1 1
= 0.707
2

As
δs
0
ω

0 ωc1 ωc2
ω
S topband

P as sband S topband P as sband P as sband

F ig d : N orm alized m ag n itud e respo n se o f id eal F ig h : N orm alized m ag n itud e respo n se o f p ra ctical
d igita l IIR b an dstop filter. d igita l IIR ba n dstop filter.
F ig 7.2 : N o rm a lized freq u en c y resp o n se o f id ea l a n d p ra c tica l d igital IIR filte rs.
Chapter 7 - IIR Filters 7. 6

7.3 Impulse Invariant Transformation

The objective of impulse invariant transformation is to develop an IIR filter transfer function whose
impulse response is the sampled version of the impulse response of the analog filter. The main idea behind
this technique is to preserve the frequency response characteristics of the analog filter. It can be stated that
the frequency response of digital filter will be identical with the frequency response of the corresponding
analog filter if the sampling time period T is selected sufficiently small (or the sampling frequency should be
high) to minimize (or avoid completely) the effects of aliasing.
Let, h(t) = Impulse response of analog filter
The Laplace transform of the analog impulse response h(t) gives the transfer function of analog filter.
\ Transfer function of analog filter, H(s) = L{h(t)}.
When H(s) has N number of distinct poles, it can be expressed as shown in equation (7.1) by partial
fraction expansion.
N
Ai A1 A2 AN .....(7.1)
H ( s) = ∑ s + pi
=
s + p1
+
s + p2
+ ..... +
s + pN
i=1
1
On taking inverse Laplace transform of equation (7.1) we get, o t
L e − at u( t ) =
s+a
N
h( t ) = ∑ A i e − p t u( t )
i = A1 e− p1t u( t ) + A 2 e− p 2 t u( t ) + ..... + A N e − p N t u( t ) ....(7.2)
i=1

where, u(t) = Continuous time unit step function.

Let, T = Sampling period.
h(n) = Impulse response of digital filter.
The impulse response of the digital filter is obtained by uniformly sampling the impulse response of
the analog filter.
∴ h( n) = h( t ) = h( nT)
t = nT

Therefore the impulse response h(n) can be obtained from equation (7.2) by replacing t by nT.
N
∴ h( n) = h( t )
t = nT
= h( nT) = ∑ A i e− p nT u( nT)
i

i=1
.....(7.3)
= A1 e− p1nT u( nT) + A 2 e− p 2 nT u( nT) + ..... + A N e − p N nT u( nT)

On taking Z-transform of equation (7.3) we get, 1

o t
Z e − anT u( nT) =
1 − e − aT z −1
1 1
H( z) = Z{h( n)} = A1 + A2 + .....
1 − e − p1T z −1 1 − e − p 2 T z −1
N
1 1 .....(7.4)
+ AN
1 − e − pNT
z −1
= ∑ A i 1 − e − p T z −1
i
i=1
7. 7 Digital Signal Processing
Comparing the expression of H(s) and H(z) [i.e., equations (7.1) and (7.4)] we can say that,

1 1 .....(7.5)
(
is transformed to )
→ − p i T −1
s + pi 1 − e z

by impulse invariant transformation, where T is the sampling time period.

When a discrete time signal is obtained by sampling analog signal, the frequency spectrum of
discrete signal will be scaled by a factor 1/T (Refer section 4.7 of Chapter 4). Due to this fact, the transfer
function obtained by impulse invariant method is amplified by the factor 1/T for small values of T. If this
amplification is undesirable then the transfer function obtained by impulse invariant tranformation can be
multiplied by T to obtain magnitude normalized transfer function HN(z).
\ HN(z) = T ´ H(z) .....(7.6)
7.3.1 Relation Between Analog and Digital Filter Poles in Impulse Invariant Transformation
The analog poles are given by the roots of the term (s + pi), for i = 1, 2, 3, ....., N. The digital poles are
given by the roots of the term (1 − e− p i T z−1 ), for i = 1, 2, 3, ....., N. From equation (7.5) we can say that the
−p T
analog pole at s = –pi is transformed into a digital pole at z = e i
Consider the digital pole, zi = e− pi T ..... (7.7) jΩ s-p lan e
Put, –pi = si in equation (7.7).
LH P RHP
jΩi si
∴ zi = e − p i T = e si T ..... (7.8)

We know that, "s i" is a point on s-plane. Let the

σi
coordinates of si be si and jW i as shown in fig 7.3.
F ig 7 .3 : s-pla n e.
∴ si = σ i + jΩi ..... (7.9)

Using equation (7.9), the equation (7.8) can be written as,

zi = e(σ i + jΩ i )T
= e σ i T e jΩ i T

We know that "z i" is a complex number. Hence "z i" can be expressed in polar coordinates as, zi = |zi| Ð z i.

∴ |z i | ∠z i = eσ i T e jΩ i T ..... (7.10)

On separating the magnitude and phase of equation (7.10) we get,

|zi | = e σ i T and ∠z i = Ω iT ..... (7.11)

From equation (7.11) the following observations can be made.

1. If si < 0 (i.e., si is negative), then the analog pole "si" lie on Left Half (LHP) of s-plane. In this
case, |z i| < 1, hence the corresponding digital pole "z i" will lie inside the unit circle in z-plane.
2. If si = 0 (i.e., real part is zero), then the analog pole "si" lie on imaginary axis of s-plane. In this
case, |z i| = 1, hence the corresponding digital pole "z i" will lie on the unit circle in z-plane.
3. If si > 0 (i.e., si is positive), then the analog pole "si" lie on the Right Half (RHP) of s-plane. In
this case |z i| >1, hence the corresponding digital pole will lie outside the unit circle in z-plane.
Chapter 7 - IIR Filters 7. 8
The above discussions are applicable for mapping any point on s-plane to z-plane. In general the
impulse invariant transformation maps all points in the s-plane given by,
2 πk ..... (7.12)
si = σ i + jΩi + j , for k = 0, ± 1, ± 2 .....
T
into a single point in the z-plane as For integer k,
FG σi + jΩi + IJ
j2 πk
T ej2pk = 1
zi = eH T K = eσ i T e jΩi T e j2 πk = eσ i T e jΩi T ..... (7.13)
From equations (7.12) and (7.13) we can say that the strip of width 2p/T in the s-plane for values of s
in the range –p/T £ W £ +p/T is mapped into the entire z-plane. Similarly the strip of width 2p/T in the
s-plane for values of s in the range p/T £ W £ 3p/T is also mapped into the entire z-plane. Likewise the strip
of width 2p/T in the s-plane for values of s in the range -3p/T £ W £ -p/T is also mapped into the entire
z-plane.
In general any strip of width 2p/T in the s-plane for values of s in the range, (2k – 1)p/T £ W £ (2k + 1) p/T
(where k is an integer), is mapped into the entire z-plane. The left half portion of each strip in s-plane maps into
the interior of the unit circle in z-plane, right half portion of each strip in s-plane maps into the exterior of the
unit circle in z-plane and the imaginary axis of each strip in s-plane maps into the unit circle in z-plane as
shown in fig 7.4. Therefore we can say that the impulse invariant mapping is many-to-one mapping (and does
not provide one-to-one mapping).

jΩ
3 π/T
jv
LHP RHP U nit circ le
j1

π/T

σ −1 1
u
−π/T

−j1
−3 π/T

F ig 7 .4 a : s-p la n e. F ig 7 .4 b : z-p la n e.
F ig 7 .4 : M a p p in g o f s-p la n e in to z-p la n e in im p u lse in v a ria n t tra n sfo rm a tio n .

The stability of a filter (or system) is related to the location of the poles. For a stable analog filter the
poles should lie on the left half of the s-plane. Since the left half of s-plane maps inside the unit circle in
z-plane we can say that, for a stable digital filter the poles should lie inside the unit circle in z-plane.
7.3.2 Relation Between Analog and Digital Frequency in Impulse Invariant Transformation
Let, W = Analog frequency in rad/second.
w = Digital frequency in rad/sample.
Let, z = rejw be a point on z-plane,
and s = s + jW be the corresponding point in s-plane.
Then by impulse invariant transformation,
z = esT ..... (7.14)
7. 9 Digital Signal Processing
Put, z = r ejw and s = s + jW in equation (7.14).

∴ r e jω = e( σ + jΩ ) T
..... (7.15)
r e jω = eσT e jΩT
On equating the phase on either side of equation (7.15) we get,
ω
Digital frequency, w = W T or Ana log frequency, Ω = ..... (7.16)
T
When impulse invariant transformation is employed the equation (7.16) can be used to compute
the digital frequency for a given analog frequency and vice versa.
The mapping of analog to digital frequency is not one-to-one. Since w is unique over the range
(-p to +p), the mapping w = W T implies that the interval -p/T £ W £ +p/T maps into the corresponding
values of -p £ w £ +p. In general the interval (2k-1) p/T £ W £ (2k + 1) p/T (where k is an integer) maps into
the corresponding values of -p £ w £ +p. Thus the mapping from the analog frequency W to the digital
frequency w is many-to-one. This reflects the effects of aliasing due to sampling.
7.3.3 Useful Impulse Invariant Transformation

The following transformations are given without proof. The equation (7.17) can be used when the
analog real poles has a multiplicity of m. The equations (7.18) and (7.19) can be used when the analog poles
are complex conjugate.

1 ( −1) m − 1 d m − 1 1 .....(7.17)

→ m−1 − p i T −1
(s + pi ) m ( m − 1)! dpi 1 − e z

(s + a) 1 − e − aT (cos bT) z −1 ....(7.18)


→
(s + a ) 2 + b2 1 − 2e − aT (cos bT) z −1 + e −2 aT z−2

b e − aT (sin bT) z−1


→
(s + a ) 2 + b2 1 − 2e − aT (cos bT) z −1 + e −2 aT z−2 ....(7.19)

Example 7.1
2
For the analog transfer function, H(s) = 2
, determine H(z) using impulse invariant
s + 3s + 2
transformation if (a) T = 1 second and (b) T = 0.1 second.

Solution The roots of quadratic,

2 2 s 2 + 3s + 2 = 0 are,
Given that, H(s) = 2
=
s + 3s + 2 (s + 1) (s + 2)
−3 ± 32 − 4 × 2
By partial fraction expansion technique we can write, s=
2
2 A B
H(s) = = + −3 ± 1
(s + 1) (s + 2) s + 1 s + 2 = = −1, − 2
2
2 2
A= × (s + 1) = =2
(s + 1) (s + 2) s = −1 − 1 +2

2 2
B= × (s + 2) = = −2
(s + 1) (s + 2) s = −2 −2 +1
Chapter 7 - IIR Filters 7. 10
2 −2
∴ H(s) = +
s+1 s+2
By impulse invariant transformation we know that,
Ai Ai
  →
s + pi (is transformed to )
1 − e − piT z −1

2 −2
∴ H(z) = + where p1 = 1 and p2 = 2
1 − e − p1T z−1 1 − e− p 2T z−1
2 −2
H(z) = −1
+
1− e –T
z 1 − e–2T z−1
(a) When T = 1 second
2 −2
H(z) = +
1 − e−1 z −1 1 − e−2 z−1
2 −2 2(1 − 0.1353z−1) − 2(1 − 0.3679z−1)
H(z) = −1
+ −1
=
1 − 0.3679z 1 − 0.1353z (1 − 0.3679z−1) (1 − 0.1353z−1)

2 − 0.2706z −1 − 2 + 0.7358z−1 0.4652 z−1

= −1 −1 −2
=
1 − 0.1353 z − 0.3679z + 0.0498z 1 − 0.5032z−1 + 0.0498z −2
Alternatively,
0.4652 z−1 0.4652 z−1
H(z) = −1 −2
= −2 2
1 − 0.5032 z + 0.0498 z z (z − 0.5032 z + 0.0498)
0.4652z
=
z2 − 0.5032 z + 0.0498

(b) When T = 0.1 second

2 −2
H(z) = +
1 − e−0.1 z −1 1 − e−0.2 z−1
2 −2 2(1 − 0.8187z −1) − 2(1 − 0.9048z−1)
= −1
+ −1
=
1 − 0.9048z 1 − 0.8187z (1 − 0.9048z−1) (1 − 0.8187z−1)

2 − 1.6374z−1 − 2 + 18096
. z−1 0.1722 z−1
= −1 −1 −2
=
1 − 0.8187 z − 0.9048z + 0.7408 z .
1 − 17235 z−1 + 0.7408 z−2

Alternatively,
0.1722 z−1 0.1722 z−1
H(z) = −1 −2
= −2 2
1 − 1. 7235 z + 0.7408 z z (z − 1. 7235 z + 0.7408)
0.1722z
=
z2 − 1. 7235 z + 0.7408

Since, T < 1, we can compute magnitude normalized transfer function, HN(z).

0.1722 z −1 0.0172 z −1
HN(z) = T × H(z) = 0.1 × −1 −2
=
1 − 1. 7235 z + 0.7408 z 1 − 1. 7235 z −1 + 0.7408z −2

Alternatively,
0.1722 z 0.0172 z
HN (z) = T × H(z) = 0.1 × =
z2 − 1. 7235 z + 0.7408 z2 − 1. 7235 z + 0.7408
7. 11 Digital Signal Processing
Example 7.2
Convert the analog filter with system transfer function,
(s + 0.1)
H(s) =
(s + 0.1)2 + 9
into a digital IIR filter by means of the impulse invariant method.

Solution
Method - I

s + 0.1 s + 0.1
Given that, H(s) = =
(s + 0.1)2 + 9 (s + 0.1)2 + 32
Using transformation of equation (7.18) we can write,

H(z) =
c
1 − e−0.1T cos 3T z−1 h =
c
1 − e−0.1 cos 3 z−1 h
Put, T = 1
1 − 2e −0.1T
ccos 3Th z −1
+e −2 × 0.1T
z −2
1 − 2e ccos 3h z
−0.1 −1
+ e−0.2 z−2

1 + 0.8958 z −1
=
.
1 + 17916 z−1 + 0.8187 z−2

Alternatively,
1 + 0.8958 z −1 1 + 0.8958 z −1 z2 + 0.8958 z
H(z) = −1 −2
= −2 2
= 2
1 + 1. 7916 z + 0. 8187 z z (z + 1. 7916 z + 0. 8187) z + 1. 7916 z + 0. 8187

Method - II The roots of the quadratic

s2 + 0.2s + 9.01 = 0 are
(s + 0.1) s + 0.1
Given that, H(s) = = −0.2 ± 0.22 − 4 × 9.01
(s + 0.1)2 + 9 s 2 + 2 × 0.1 × s + 0.12 + 9 s=
2
s + 0.1 s + 0.1 −0.2 1
= = = ± −36 = −0.1 ± j3
s 2 + 0.2s + 9.01 (s + 0.1 − j3) (s + 0.1+ j3) 2 2
∴ (s 2 + 0.2s + 9.01)
By partial fraction expansion H(s) can be expressed as,
= (s − (−0.1 + j3)) (s − (− 0.1 − j3))
s + 0.1 A A∗
H(s) = = + = (s + 0.1 − j3)(s + 0.1+ j3)
(s + 0.1 − j3) (s + 0.1+ j3) s + 0.1 − j3 s + 0.1 + j3

s + 0.1 −0.1 + j3 + 0.1 j3

A= × (s + 0.1− j3) = = = 0.5
(s + 0.1− j3) (s + 0.1+ j3) s = − 0.1+ j3
−0.1 + j3 + 0.1 + j3 j6

b g ∗
A∗ = 0.5 = 0.5
0.5 0.5
∴ H(s) = +
s + 0.1 − j3 s + 0.1 + j3
By impulse invariant transformation we know that,
Ai Ai
  → and let, T = 1
s + pi (is transformed to
1 − e−piT z−1
0.5 0.5 0.5 0.5
∴ H(z) = + = +
1 − e−(0.1 − j3)T z−1 1 − e−(0.1 + j3)T z−1 1 − e−0.1 e j3 z−1 1 − e−0.1 e− j3 z−1

0.5 (1 − e−0.1 e− j3 z−1) + 0.5(1 − e−0.1 e j3 z−1)

=
(1 − e−0.1 e j3 z−1) (1 − e−0.1 e− j3 z−1)
Chapter 7 - IIR Filters 7. 12

0.5 − 0.5 e−0.1 e− j3 z−1 + 0.5 − 0.5 e−0.1 e j3 z −1

∴ H(z) =
1 − e −0.1 e− j3 z −1 − e −0.1 e j3 z−1 + e −0.1 e j3 e−0.1 e –j3 z−2
1 − 0.5 e−0.1 z −1(e j3 + e− j3 )
=
1 − e −0.1 z−1 (e j3 + e − j3 ) + e−0.2 z−2

1 − 0.5 × (2 cos3) e−0.1 z−1

=
1 − e −0.1 z−1 (2 cos 3) + e−0.2 z−2 e jθ + e − jθ
cosθ =
2
1 − (cos 3) e−0.1 z−1
=
1 − 2(cos 3) e−0.1 z −1 + e−0.2 z −2
Note : Evalutate cos q by keeping
1+ 0.8958 z−1 calculator in radian mode.
=
1 + 17916
. z−1 + 0.8187 z−2
Alternatively,

1 + 0.8958 z −1 1 + 0.8958 z−1

H(z) = −1 −2
= −2 2
1 + 1. 7916 z + 0. 8187 z z (z + 1. 7916 z + 0. 8187)
z2 + 0.8958 z
= 2
z + 1. 7916 z + 0. 8187

Example 7.3
Using impulse invariant transformation convert the following analog filter transfer function to digital
filter transfer function by taking sampling time, T = 0.5 second.
2.8s 2 + 4.8s + 2.9
H(s) =
(s + 3) (s 2 + s + 0.85)
Solution
Method - I

2.8s 2 + 4.8s + 2.9

Given that, H(s) =
(s + 3) (s 2 + s + 0.85)
By partial fraction expansion technique H(s) can be expressed as,
2.8s 2 + 4.8s + 2.9 A Bs + C
H(s) = = +
(s + 3) (s 2 + s + 0.85) s + 3 s 2 + s + 0.85
On cross multiplying the above equation we get,
2.8s 2 + 4.8s + 2.9 = A(s 2 + s + 0.85) + (Bs + C) (s + 3)

2.8s 2 + 4.8s + 2.9 = As 2 + As + 0.85A + Bs 2 + 3Bs + Cs + 3C

On equating coefficients On equating coefficients of s we get, On equating constants we get,

2
of s we get, A + 3B + C = 4.8 0.85 A + 3C = 2.9
A + B= 2.8 Put, B = 2.8 – A Put, C = 2A – 3.6
\ B = 2.8 – A \ A + 3 (2.8 – A) + C = 4.8 \ 0.85A + 3 (2A – 3.6) = 2.9
A + 8.4 – 3A + C = 4.8 0.85A + 6A – 10.8 = 2.9
C = 4.8 –8.4 + 2A 6.85A = 2.9 + 10.8
\ C = 2A – 3.6 2.9 + 10.8
A= =2
6.85
7. 13 Digital Signal Processing
Here, A = 2
\ B = 2.8 – A = 2.8 – 2 = 0.8
C = 2A – 3.6 = 2 ´ 2 – 3.6 = 0.4
(a + b)2 = a2 + 2ab +b2
0 .4
∴ H(s) =
2 0.8s + 0.4
+ 2 =
2
+ 2
0.8 s +
0.8 e j
s + 3 s + s + 0.85 s + 3 s + (2 × 0.5) s + 0.52 − 0.52 + 0.85

2 0.8 (s + 0.5) 2 0.8(s + 0.5)

= + = +
s + 3 (s + 0.5)2 + 0.6 s + 3 (s + 0.5)2 + 0.6 2
e j
1 (s + 0.5)
= 2× + 0.8 ×
s +3 (s + 0.5)2 + 0. 77462
Using equations (7.17)
Now, using impulse invariant transformation,
and (7.18).
1 1 − e−0.5T (cos 0.7746T ) z−1
H(z) = 2 × + 0.8 ×
1 − e−3T z−1 1 − 2e −0.5T
(cos 0.7746T) z−1 + e −2 × 0.5T z−2

2 0.8 − 0.8 e −0.5 × 0.5 (cos 0.7746 × 0.5) z−1

= + Put, T = 0.5
1 − e−3 × 0.5z−1 1 − 2e−0.5 × 0.5(cos 0.7746 × 0.5) z−1 + e −2 × 0.5 × 0.5 z−2

2 0.8 − 0.5769 z−1

= −1
+
1 − 0.2231z 1 − 1.4422 z −1 + 0.6065 z−2
2 (1 − 1.4422 z−1 + 0.6065 z−2 ) + (0.8 − 0.5769 z−1) (1 − 0.2231z−1)
=
(1 − 0.2231z−1) (1 − 1.4422 z−1 + 0.6065 z−2 )

2 − 2.8844 z−1 + 1.213 z−2 + 0.8 − 0.1785 z−1 − 0.5769 z−1 + 0.1287z−2
=
1 − 1.4422 z−1 + 0.6065 z −2 − 0.2231z−1 + 0.3218 z−2 − 0.1353 z−3
2 .8 − 3.6398 z−1 + 13417
. z−2
=
1 − 1.6653 z + 0.9283 z − 0.1353 z−3
−1 −2

Alternatively,
2 .8 − 3.6398 z −1 + 13417
. z −2 z −3 (2 .8 z3 − 3.6398 z2 + 13417
. z)
H(z) = −1 −2 −3
= −3 3
1 − 1.6653 z + 0.9283 z − 0.1353 z z (z − 1.6653 z 2 + 0.9283 z − 0.1353)

2 .8 z3 − 3.6398 z 2 + 13417
. z
=
z3 − 1.6653 z2 + 0.9283 z − 0.1353

Since, T < 1, we can compute magnitude normalized transfer function, HN(z).

2 .8 − 3.6398 z −1 + 13417
. z −2 14
. − 18199
. z −1 + 0.6709 z −2
HN(z) = T × H(z) = 0.5 × −1 −2 −3
=
1 − 1.6653 z + 0.9283 z − 0.1353 z 1 − 1.6653 z −1 + 0.9283 z −2 − 0.1353 z −3

Alternatively,
2 .8z3 − 3.6398 z 2 + 13417
. z . z3 − 18199
14 . z2 + 0.6709 z
HN (z) = T × H(z) = 0.5 × 3 2
= 3 2
z − 1.6653 z + 0.9283 z − 0.1353 z − 1.6653 z + 0.9283 z − 0.1353

Method - II The roots of the quadratic

2 s2 + s + 0.85 = 0 are,
2.8s + 4.8s + 2 .9
Given that, H(s) =
(s + 3) (s 2 + s + 0.85) s=
−1 ± 12 − 4 × 0.85
2
2.8s 2 + 4.8s + 2 .9 −1 ± j1.5492
= =
(s + 3) (s + 0.5 − j0.7746) (s + 0.5 + j0.7746) 2
= −0.5 ± j 0.7746
Chapter 7 - IIR Filters 7. 14
By partial fraction expansion technique H(s) can be expressed as,
2.8s 2 + 4.8s + 2 .9 A B B∗
H(s) = = + +
(s + 3) (s + 0.5 − j0.7746) (s + 0.5 + j0.7746) s + 3 s + 0.5 − j0.7746 s + 0.5 + j0.7746

2.8s 2 + 4.8s + 2 .9 2.8s 2 + 4.8s + 2 .9

A= × (s + 3) =
(s + 3) (s + 0.5 − j0.7746) (s + 0.5 + j0.7746)
s =−3
s 2 + s + 0.85 s = −3

2.8 × ( −3)2 + 4.8( −3) + 2 .9

= =2
( −3)2 + (−3) + 0.85
2.8s 2 + 4.8s + 2 .9
B= × (s + 0.5 − j0.7746)
(s + 3) (s + 0.5 − j0.7746) (s + 0.5 + j0.7746)
s= −0.5+ j0.7746

2
2 .8(−0.5 + j0.7746) + 4.8(−0.5 + j0.7746) + 2 .9
=
(−0.5 + j0.7746 + 3) (−0.5 + j0.7746 + 0.5 + j0.7746)
2 .8( −0.5 + j0.7746)2 + 0.5 + j3.7181
= = 0.4
(2.5 + j0.7746) ( j1.5492)

B∗ = (0.4)∗ = 0.4
2 0.4 0.4
∴ H(s) = + +
s + 3 s + 0.5 − j0.7746 s + 0.5 + j0.7746
1 1 1
= 2× + 0.4 × + 0.4 ×
s +3 s + (0.5 − j0.7746) s + (0.5 + j0.7746)

Using impulse invariant transformation, H(s) is transformed to H(z) as shown below.

1 1 1 Using
H(z) = 2 × + 0.4 × + 0.4 ×
1− e −3T
z −1
1− e − ( 0.5 − j0.7746 )T
z −1
1− e − ( 0.5 + j0.7746 )T −1
z equation (7.17).
2 0.4 0.4
= + + Put, T = 0.5
1 − e−3 × 0.5 z−1 1 − e−( 0.5 − j0.7746) × 0.5 z −1 1 − e −( 0.5+ j0.7746) × 0.5z −1
2 0.4 0.4
= + +
1 − e−1.5 z−1 1 − e−0.25 e j0.3873 z −1 1 − e−0.25 e − j0.3873z−1

=
2
+
d i d
0.4 1 − e−0.25 e− j0.3873z−1 + 0.4 1 − e−0.25 e j0.3873z−1 i
1 − 0.2231 z−1 d id
1 − e−0.25 e j0.3873 z −1 1 − e −0.25 e − j0.3873 z −1 i
2 0.4 − 0.4 e −0.25 e − j0.3873z −1 + 0.4 − 0.4 e −0.25 e j0.3873z −1
= −1
+
1 − 0.2231 z 1 − e −0.25 e − j0.3873 z −1 − e−0.25 e j0.3873 z −1 + e −0.5 z −2

=
2
+
e
0.8 − 0.4 e−0.25 ej0.3873 + e − j0.3873 z−1 j
−1
1 − 0.2231z e
1 − e−0.25 e j0.3873 + e − j0.3873 z−1 + e−0.5 z−2 j
=
2
+
0.8 − 0.4 e −0.25
2 cos 0.3873 z−1 c h
c
1 − 0.2231z −1 1 − e−0.25 2 cos 0.3873 z−1 + e −0.5 z−2 h
2 0.8 − 0.5769 z−1
= +
1 − 0.2231z −1 1 − 1.4422 z−1 + 0.6065 z−2

=
e j e
2 1 − 1.4422 z−1 + 0.6065 z −2 + 0.8 − 0.5769 z −1 1 − 0.2231z−1 je j
e1 − 0.2231z je1 − 1.4422 z
−1 −1
+ 0.6065 z −2
j
7. 15 Digital Signal Processing
2 − 2.8844 z−1 + 1213
. z−2 + 0.8 − 0.1785 z−1 − 0.5769 z−1 + 0.1287 z−2
∴ H(z) =
1 − 1.4422 z −1 + 0.6065 z−2 − 0.2231z −1 + 0.3218 z−2 − 0.1353 z −3

2.8 − 3.6398 z−1 + 1.3417 z−2

=
1 − 1.6653 z−1 + 0.9283 z−2 − 0.1353 z−3

Alternatively,

2.8 − 3.6398 z−1 + 1.3417 z −2

H(z) =
1 − 1.6653 z−1 + 0.9283 z−2 − 0.1353 z−3

=
e
z−3 2.8 z3 − 3.6398 z2 + 1.3417 z j
z −3
ez
3
− 16653
. 2
z + 0.9283 z − 0.1353 j
2 .8 z3 − 3.6398 z2 + 1.3417 z
=
z − 1.6653 z2 + 0.9283 z − 0.1353
3

7.4 Bilinear Transformation

The bilinear transformation is a conformal mapping that transforms the imaginary axis of s-plane
into the unit circle in the z-plane only once, thus avoiding aliasing of frequency components. In this
mapping all points in the left half of s-plane are mapped inside the unit circle in the z-plane and all points in
the right half of s-plane are mapped outside the unit circle in the z-plane.
The bilinear transformation can be linked to the trapezoidal formula for numerical integration. Any
analog system is governed by a differential equation in time domain. Consider the first order differential
equation of an analog system as shown in equation (7.20).

dy( t ) .....(7.20)
Let , = x( t )
dt
On integrating both sides of equation (7.20) we get,
nT nT The trapezoidal rule when integration is

z dy( t )
dt
dt =
z x( t ) dt approximated by two trapezoids is,
b
( n − 1) T

y( t )
nT
( n − 1) T

=
nT

z x(t) dt
z
a
f(x) dx =
b−a
2
f ( a ) + f ( b)

( n −1) T
( n − 1) T
nT

b
y( nT) − y ( n − 1) T = g z
( n − 1) T
x(t) dt
.....(7.21)

The integral on the right side of equation (7.21) can be approximated by the trapezoidal rule, so that,

.....(7.22)
y( nT) − y[( n − 1) T] = ejT
2
b
x( nT) + x ( n − 1)T g
For discrete time system, the equation (7.22) can be written as,

y( n) − y( n − 1) =
T
x( n) + x( n − 1) .....(7.23)
2
Chapter 7 - IIR Filters 7. 16
On taking Z-transform of equation (7.23) we get,
T
Y ( z ) − z −1 Y ( z ) = 2
X(z) + z−1 X(z)
T
1 − z−1 Y(z) = 2
1 + z −1 X(z)

2(1 − z−1 )
Y(z) = X(z) .....(7.24)
T(1 + z −1 )
On taking Laplace transform of equation (7.20) we get,
s Y(s) = X(s) .....(7.25)
On comparing equations (7.24) and (7.25) we can say that,
2 1 − z −1
s Y( s) ( → Y(z) .....(7.26)
is transformed to) T
1 + z −1
by bilinear transformation, where T is the sampling time period.
2 1− z −1
Hence in the s-domain transfer function, if "s" is substituted by the term the resulting
T
1 + z −1
transfer function will be z-domain transfer function.
7.4.1 Relation Between Analog and Digital Filter Poles in Bilinear Transformation
The mapping of s-domain function to z-domain function by bilinear transformation is a one to one
mapping, that is, for every point in z-plane, there is exactly one corresponding point in s-plane and
vice versa. The transformation is accomplished when,
2 1 − z −1
s= .....(7.27)
T
1 + z −1
The equation (7.27) can be rearranged as shown below to express "z" in terms of "s".
2 1 − z −1 T 1 − z −1 T z −1 ( z − 1)
s= ⇒ s= ⇒ s=
T
1 + z −1 2
1 + z −1 2
z −1 ( z + 1)
Tz −1 .....(7.28)
∴ 2
s=
z +1
On cross multiplying equation (7.28) we get,
T T T T T
2
s(z + 1) = z −1 ⇒ 2
s z + 2 s = z −1 ⇒ 2
s z − z = −1 − 2 s

e T
∴ – z 1– 2s = − 1 + j e T
2
s j
T
1 + 2
s
∴ z= T .....(7.29)
1 − 2
s

In equation (7.29), the variable "s" represent a point on s-plane and "z" is the corresponding point in
z-plane.
Let, si = si + jW i.
On substituting, si = si + jW i in equation (7.29) we get,
T T T
1 + 2 ( σ i + jΩi ) 1 + 2 σ i + j 2 Ωi .....(7.30)
zi = T
= T T
1− 2
(σi + jΩi ) 1− σ
2 i
− j 2 Ωi
7. 17 Digital Signal Processing
The magnitude of equation (7.30) is given by,

1
LM e1 + σ j + e Ω j OP
T
2 i
2 T
2 i
2 2
..... (7.31)
= M
MN e1 − σ j + e Ω j PPQ
zi 2 2
T T
i − i
2 2

From equation (7.31) the following observations can be made,

1. If si < 0 (i.e., si is negative), then the point si = si + jW i, lie on the left half of s-plane. In this
case, |zi| < 1, hence the corresponding point in z-plane will lie inside the unit circle in z-plane.
2. If si = 0 (i.e., real part is zero), then the point si = si + jW i lie on the imaginary axis in the
s-plane. In this case, |zi| = 1, hence the corresponding point in z-plane will lie on the unit circle
in z-plane.
3. If si > 0 (i.e., si is positive), then the point si = si + jW i lie on the right half of s-plane. In this
case, |zi| > 1, hence the corresponding point in z-plane will lie outside the unit circle in z-plane.
The above discussions are applicable for mapping poles and zeros from s-plane to z-plane. The
stability of the filter is associated with location of poles. We know that for a stable analog filter the poles
should lie on the left half of s-plane. In bilinear transformation, the points on left half of s-plane are mapped
as points inside unit circle in z-plane. Hence for stability of digital filter the digital poles should lie inside the
unit circle in z-plane.

jΩ jv
U n it c ircle j1

LHP RHP

σ −1 1 u

−j1

F ig 7 .5a : s-pla ne. F ig 7 .5b : z-pla ne.

F ig 7 .5 : M a p p in g o f s-p la n e in to z-p la n e in b ilin e a r tra nsform a tio n.

7.4.2 Relation Between Analog and Digital Frequency in Bilinear Transformation

Let, s = jW be points on imaginary axis and the corresponding points on the z-plane on unit circle
are given by z = ejw . For bilinear transformation,
2 1 − z −1 .....(7.32)
s=
T 1 + z −1
Put, s = jW and z = e jw in equation (7.32)

Fe jω − jω I
2 1 − e 2 − jω
GH 2 e 2 − e − jω JK
∴ jΩ = = e jθ e− jθ = 1
T 1 + e − jω
T Fe jω − jω I
GH 2 e 2 + e − jω JK
Chapter 7 - IIR Filters 7. 18
− jω Fe jω − jω I e jθ − e − jθ
2
e 2
GH 2 −e 2
JK ω
2 2 j sin 2
sinθ =
2j
jΩ = =
T − jω Fe jω − jω I T 2 cos ω
e 2
GH 2 +e 2
JK 2 cosθ =
e jθ + e − jθ
2
ω
2 sin 2 2 ω
∴Ω = = tan
T cos ω T 2 .....(7.33)
2
2 ω
∴ Analog frequency, Ω = tan
T 2
The equation (7.33) relates the analog frequency, W and digital frequency, w.
From equation (7.33) we get,
ΩT ω ω ΩT
= tan ⇒ = tan −1
2 2 2 2
ΩT .....(7.34)
∴ Digital frequency, ω = 2 tan −1
2
The equation (7.34) can be used to estimate the digital frequency w for a given analog
frequency, W . The equation (7.33) is used to calculate the analog frequency for a given digital frequency.
From the above analysis it is evident that the analog frequency W and digital frequency w has a nonlinear
relationship, because the entire negative imaginary axis in the s-plane (from W = –¥ to 0) is mapped into the
lower half of unit circle in z-plane (from w = –p to 0) and the entire positive imaginary axis in the
s-plane (from W = 0 to +¥ ) is mapped into the upper half of unit circle in z-plane (from w = 0 to +p). This
nonlinear mapping introduces a distortion in the frequency axis, which is called frequency warping.

ω = ΩT (Impulse invariant)

π −1
ω = 2tan ΩT/2

(Bilinear transformation)

2 Warping

0 1 2 3 4 5 6 7 8 9 10
ΩT
F ig 7 .6 : C o rresp on d e nce b etw een a n alo g a n d d igital freq u en c ies
resu ltin g fro m th e b ilin ea r tra nsfo rm a tio n .

The effect of warping on the magnitude response can be explained by considering an analog filter
with a number of passbands as shown in fig 7.7. The corresponding digital filter will have same number of
passbands, but with disproportionate bandwidth, as shown in fig 7.7.
7. 19 Digital Signal Processing
In designing digital filter using bilinear transformation the effect of warping on amplitude response
can be eliminated by prewarping the analog filter. In this method, the specified digital frequencies are
converted to analog equivalent using equation (7.33). This analog frequencies are called prewarp frequencies.
Using the prewarp frequencies, the analog filter transfer function is designed and then it is transformed to
digital filter transfer function.
The effect of warping on the phase response can be explained by considering an analog filter with
linear phase response as shown in fig 7.8. The phase response of corresponding digital filter will be nonlinear.
From the above discussions it can be stated that the bilinear transformation preserves the magnitude
response of an analog filter only if the specification requires piecewise constant magnitude, but the phase
response of the analog filter is not preserved. Therefore the bilinear transformation can be used only to
design digital filters with prescribed magnitude response with piecewise constant values. A linear phase
analog filter cannot be transformed to a linear phase digital filter using bilinear transformation.

jω
|H (e )| Ω ∠H (e jω) Ω

Ω Ω

|H (j Ω)| ∠H (j Ω)
F ig 7.7 : T he w a rp ing effect o n m ag n itu d e resp o n se. F ig 7.8 : T he w a rp ing effect o n p h a se respo n se .

Example 7.4
2
For the analog transfer function, H(s) = , determine H(z) using bilinear transformation if
s 2 + 3s + 2
(a) T = 1 second and (b) T = 0.1 second.

Solution
2
Given that, H(s) =
s 2 + 3s + 2
2 1 − z −1
Put, s = in H(s) to get H(z).
T 1 + z −1
2
∴ H(z) =
F2 1− z −1 2 I F2 I
1 − z−1
GH T 1 + z−1 JK +3 GH T JK
1 + z−1
+2

2
=
−1 2
4 e1 − z j +
6 e1 − z j + 2
−1

T2 −1 2 T e1 + z j
−1
e1 + z j
Chapter 7 - IIR Filters 7. 20
2
∴ H(z) = 2 2
d
4 1 − z −1 i d id
+ 6T 1 − z−1 1 + z −1 + 2T 2 1 + z −1 i d i
2
d
T 2 1 + z−1 i
2

=
d
2T 2 1 + z −1 i (a + b) (a – b) = a2 – b2
2 2
d
4 1 − z −1 i d
+ 6T 1 − z −2 + 2T 2 1 + z−1 i d i
(a) T = 1 second
2

∴ H(z) =
d i 2 1 + z −1
−1 2 −1 2
4d1 − z i + 6d1 − z i + 2 d1 + z i −2

2d1 + 2z + z i −1 −2
(a + b)2 = a 2 + 2ab + b2
=
4 d1 − 2z + z i + 6d1 − z i + 2 d1 + 2z
−1 −2 −2 −1
+z −2
i (a − b)2 = a 2 − 2ab + b2

2 + 4z −1 + 2z−2 2 + 4z−1 + 2z−2

= =
12 − 4z−1
12 1 −
FG
4 −1
z
IJ
12 H K
2 4 −1 2 −2
+ z + z 0.1667 + 0.3333 z −1 + 0.1667 z−2
= 12 12 12 =
4 −1 1 − 0.3333 z−1
1− z
12
Alternatively,

0.1667 + 0.3333 z −1 + 0.1667 z −2 z −2(0.1667 z 2 + 0.3333 z + 0.1667)

H(z) = =
1 − 0.3333 z −1 1 − 0.3333 z −1

0.1667 z2 + 0.3333 z + 0.1667

=
z2 − 0.3333 z

(b) T = 0.1 second

2

H(z) =
2 × 0.12 1 + z−1 d i
2 2
d
4 1 − z −1 i d
+ 6 × 0.1 1 − z −2 + 2 × 0.12 1 + z −1 i d i
=
d
0 .02 1 + 2z−1 + z−2 i
d
4 1 − 2z −1
+z −2
i + 0.6d1 − z i + 0.02 d1 + 2z −2 −1
+ z −2 i
0.02 + 0.04z−1 + 0.02z−2
=
4.62 − 7.96 z−1 + 3.42 z−2
0.02 0.04 −1 0.02 −2
+ z + z 0.0043 + 0.0087 z −1 + 0.0043 z −2
= 4.62 4.62 4.62 =
7.96 −1 3.42 −2 1 − 17229
. z−1 + 0.7403 z−2
1− z + z
4.62 4.62
Alternatively,

0.0043 + 0.0087 z −1 + 0.0043 z −2 z −2(0.0043z2 + 0.0087 z + 0.0043)

H(z) = =
1 − 17229
. z−1 + 0.7403 z−2 z −2(z2 − 1. 7229 z + 0.7403)
0.0043z2 + 0.0087 z + 0.0043
=
z2 − 1. 7229 z + 0.7403
7. 21 Digital Signal Processing
Example 7.5
2s
Obtain H(z) from H(s) when T = 1 second and H(s) =
s 2 + 0.2s + 1
Solution

2s
Given that, H(s) =
s 2 + 0.2s + 1

2 1 − z −1
Put , s = in H(s) to get H(z).
T 1 + z −1
F2 1 − z −1 I 4 (1 − z−1)
2 GH T 1+ z −1 JK 1 + z −1
∴ H(z) = = Put, T = 1
F2 1 − z −1 I 2
F2 1 − z−1 I −1 2
4 (1 − z )
−1 2 +
0.4 (1 − z−1)
+1
GH T 1 + z −1 JK + 0.2GH T 1 + z−1 JK
+1 (1 + z ) 1 + z −1

4(1 − z−1)
(1 + z−1) 4(1 − z−1)(1 + z −1)
= −1 2 −1 −1 −1 2
=
4(1 − z ) + 0.4(1 − z ) (1+ z ) + (1 + z ) 4(1 − z ) + 0.4(1 − z−1) (1+ z −1) + (1 + z−1)2
−1 2

−1 2
(1 + z )
4(1 − z −2 )
= −1
(a + b) (a – b) = a2 – b2
4(1 − 2z + z ) + 0.4(1 − z−2 ) + (1 + 2z−1 + z−2 )
−2
(a + b)2 = a 2 + 2ab + b2
4 4 −1 (a − b)2 = a 2 − 2ab + b2
− z
4 − 4z−2 5.4 5.4
= =
5.4 − 6z−1 + 4.6z−2 6 −1 4.6 −2
1− z + z
5.4 5.4
0 .7407 − 0. 7407 z−1
=
1 − 1.111z−1 + 0.8519 z−2

Alternatively,

0 .7407 − 0. 7407 z −1 0 .7407 − 0. 7407 z −1

H(z) = =
1 − 1.111z −1 + 0.8519 z −2 z (z2 − 1.111z + 0.8519)
−2

0 .7407 z2 − 0. 7407 z
=
z2 − 1.111z + 0.8519

Example 7.6
s3
Obtain H(z) from H(s) when T =1 second, and H(s) =
(s + 1) (s 2 + s + 1)
Solution

s3
Given that, H(s) =
(s + 1) (s 2 + s + 1)
2 1 − z −1
Put, s = in H(s) to get H(z).
T 1 + z −1
Chapter 7 - IIR Filters 7. 22

F2 1 − z−1 I 3

GH T 1+ z−1 JK
∴ H(z) =
F2 1 − z −1 I LF 2
+ 1J MG
1 − z−1 I 2
2 1 − z−1 OP
GH T 1 + z −1 K MNH T 1 + z−1 JK +
T 1 + z−1
+1
PQ
8(1 − z−1)3
(1 + z−1)3
=
LM 2(1 − z −1
)
+1
OP LM 4(1 − z−1)2 2(1 − z−1)
+ +1
OP Put, T = 1
MN 1 + z−1
PQ MN (1 + z−1)2 1 + z −1 PQ
8(1 − z −1)3
(1 + z−1)3
=
LM 2(1 − z ) + (1 + z
−1 −1
) OP LM 4(1 − z−1)2 + 2(1 − z−1) (1 + z −1) + (1 + z−1)2 OP
MN (1 + z ) −1
PQ MN (1 + z−1)2 PQ
8(1 − z −1)3
=
2(1 − z ) + (1 + z ) 4(1 − z −1)2 + 2(1 − z−1) (1+ z −1) + (1 + z−1)2
−1 −1

8(1 − z−1)(1 − 2z −1 + z −2 )
=
−1 −1
2 − 2z + 1+ z 4(1 − 2z −1 + z −2 ) + 2(1 − z −2 ) + (1 + 2z −1 + z −2 )

8(1 − 2z−1 + z −2 − z −1 + 2z −2 − z −3 )
=
3 − z −1 7 − 6z−1 + 3z −2

8 1 − 3z−1 + 3z−2 − z−3

=
21 − 18z−1 + 9z −2 − 7z −1 + 6z−2 − 3z−3
8 − 24z−1 + 24z−2 − 8z −3
=
21 − 25z−1 + 15z −2 − 3z −3
8 24 −1 24 −2 8 −3
− z + z − z
= 21 21 21 21
25 −1 15 −2 3 −3
1− z + z − z
21 21 21
0.381 − 11429
. z −1 + 11429
. z −2 − 0.381z−3
=
.
1 − 11905 z + 0.7143 z − 0.1429 z−3
−1 −2

Alternatively,

0.381 − 11429
. z−1 + 1.1429 z−2 − 0.381z −3
H(z) =
1 − 11905
. z + 0.7143 z−2 − 0.1429 z −3
−1

z −3 0.381z3 − 11429
. z2 + 11429
. z − 0.381
=
z−3 z3 − 11905
. z2 + 0.7143 z − 0.1429

0.381z3 − 11429
. z2 + 11429
. z − 0.381
= 3 2
z − 11905
. z + 0.7143 z − 0.1429
7. 23 Digital Signal Processing
Example 7.7
Convert the analog filter with system function H(s) into digital filter using bilinear transformation.

s + 0.3
H(s) = ; Take T = 0.5
(s + 0.3)2 + 16

Solution

s + 0.3 s + 0.3 s + 0.3

Given that, H(s) = = =
(s + 0.3)2 + 16 s 2 + 0.6s + 0.09 + 16 s 2 + 0.6s + 16.09

2 1 − z −1
Put, s = in H(s) to get H(z).
T 1 + z −1

2 1 − z −1 2(1 − z −1)
−1 +
0.3 + 0.3
T 1+ z T(1+ z −1)
∴ H(z) = =
F2 1 − z−1 I 2
F 2 1 − z−1 I −1
4(1 − z ) 2
+
1.2(1 − z −1)
+ 16.09
GH T 1 + z−1 JK GH
+ 0.6
T 1 + z−1 JK
+ 16.09 2 −
T (1+ z )1 2
T(1+ z−1)

2(1 − z−1) + 0.3T(1 + z −1)

T(1+ z −1)
= −1 2
. T(1 − z −1)(1 + z −1) + 16.09 T 2 (1 + z −1)2
4(1 − z ) + 12
T 2(1+ z−1)2
2(1 − z−1) + 0.3T(1 + z −1) T(1 + z−1)
=
4(1 − z −1)2 + 12
. T(1 − z−2 ) + 16.09 T 2 (1 + z−1)2

2(1 − z−1) + 0.3 × 0.5(1 + z−1) 0.5(1 + z−1) Put, T = 0.5

=
4(1 − z−1)2 + 12
. × 0.5(1 − z−2 ) + 16.09 × 0.52(1 + z−1)2
(1 − z−1)(1 + z−1) + 0.075(1 + z−1)2
=
4(1 − z−1)2 + 0.6 (1 − z−2 ) + 4.0225 (1 + z −1)2

(1 − z−2 ) + 0.075 (1 + 2z−1 + z−2 )

= −1
4(1 − 2z + z−2 ) + 0.6 (1 − z−2 ) + 4.0225 (1 + 2z−1 + z−2 )
1.075 + 0.15z−1 − 0.925 z−2
=
8.6225 + 0.045 z−1 + 7.4225 z−2
1.075 0.15 −1 0.925 −2
+ z − z 0.1247 + 0.0174 z−1 − 0.1073 z−2
= 8.6225 8.6225 8.6225 =
0.045 −1 7.4225 −2 1 + 0.0052 z−1 + 0.8608 z−2
1+ z + z
8.6225 8.6225

Alternatively,

0.1247 + 0.0174 z−1 − 0.1073 z−2

H(z) =
1 + 0.0052 z−1 + 0.8608 z−2
z−2 (0.1247 z2 + 0.0174z − 0.1073)
=
z−2(z2 + 0.0052 z + 0.8608)
0.1247 z2 + 0.0174 z − 0.1073
=
z2 + 0.0052 z + 0.8608
Chapter 7 - IIR Filters 7. 24

7.5 Specifications of Digital IIR Lowpass Filter

Let, H(ejw ) = Frequency response of IIR filter.
|H(ejw )| = Magnitude response of IIR filter.
The magnitude response, |H(ejw )| of IIR filter will have a passband, transition band and stop band. The
specification of the IIR filter can be expressed in any one of the following three different ways.
Case i : Gain at passband and stopband edge frequency
Case ii : Attenuation at passband and stopband edge frequency
Case iii : Ripple at passband and stopband edge frequency
Case i : Gain at passband and stopband edge frequency
The gain can be expressed either in normal values or in decibels (dB).
The maximum value of normalized gain is unity and so the gain at band edge frequencies will be less
than 1. Therefore, the dB-gain will be negative.
Let, w p = Passband edge digital frequency in rad/sample.
w s = Stopband edge digital frequency in rad/sample.
Ap = |H(ejw )|w = w p = Gain (or magnitude) at passband edge frequency.
As = |H(ejw )|w = w s = Gain (or magnitude) at stopband edge frequency.
Ap,dB = 20 log [|H(ejw )|w = w p] = dB-Gain (or dB-magnitude) at passband edge frequency.
As,dB = 20 log [|H(ejw )|w = w s] = dB-Gain (or dB-magnitude) at stopband edge frequency.
The gain in normal values can be converted to dB-gain or vice versa as shown below.
Ap,dB = 20 log Ap Þ Ap = 10(Ap,dB /20)
As,dB = 20 log As Þ As = 10(As,dB /20)
Example
Let, Ap = 0.8
As = 0.2
The gain in normal values can be converted to dB-gain as shown below.
\ Ap,dB = 20 log Ap = 20 log 0.8 = – 1.9382 dB » –2dB
As,dB = 20 log As = 20 log 0.2 = – 13.9794 dB » –14dB
The dB-gain can be converted back to gain in normal values as shown below.
Ap = 10(Ap,dB /20) = 10(–1.9382/20) = 0.8
As = 10(As,dB /20) = 10(–13.9794/20) = 0.2
The magnitude response and log-magnitude response of a digital IIR lowpass filter are shown in fig 7.9.
jω
A = |H (e )| ωp ωs
jω
0 ω
A = |H (e )|
A p, dB = −2dB
1
δp
A p = 0 .8

A s = 0.2
δs
A s , dB = −14dB
0 ωp ωs ω

F ig 7.9a : G a in vs ω. F ig 7.9b .
F ig 7.9 : M a gn itu d e resp o n se o f d ig ita l IIR lo w p a ss filter.
7. 25 Digital Signal Processing
Case ii : Attenuation at passband and stopband edge frequency
Alternatively, the specification of the filter can be attenuation at passband and stopband edge
frequencies. The attenuation in normal value is inverse of the gain in normal value. The attenuation is usually
expressed in decibels (dB). Since the gain at edge frequencies are less than 1, the attenuation in normal values
will be greater than1, and the dB-attenuation is positive.
1 1
Let , α p = = jω
= Attenuation at passband edge frequency
Ap H (e )
ω =ω p

1 1
αs = = = Attenuation at stopband edge frequency
A s H ( e jω )
ω =ω s

LM 1 OP LM 1 OP
α p,dB = 20 log M P = 20 log M PP = dB- Attenuation at passband edge frequency
MN A p PQ M H ( e jω )
N ω =ω p
Q
LM 1 OP LM 1 OP
α s,dB = 20 log M P = 20 log M jω PP = dB- Attenuation at stopband edge frequency
MN A PQ MN H(e )
s
ω =ω s Q
The attenuation in normal values can be conveted to dB-attenuation or vice-versa as shown below.
a p,dB = 20 log a p Þ a p = 10(ap,dB/20)
a s,dB = 20 log a s Þ a s = 10(as,dB/20)
The attenuation can be converted to gain or vice versa using the following equations.
1 1
Ap = αp =
αp Ap
1 1
As = αs =
αs As
A p ,dB = −α p,dB α p ,dB = − A p ,dB
A s,dB = −α s,dB α s,dB = − A s,dB

Example
Let, a p,dB = +1.9382dB » +2dB
a s,dB = +13.9794dB » +14dB
The dB-attenuation can be converted to normal values as shown below.
a p = 10(ap,dB /20) = 10(1.9382/20) = 1.25
a s = 10(as,dB /20) = 10(13.9794/20) = 5
The attenuation can be converted to gain as shown below.
1 1
Ap = = = 0.8
α p 1.25
1 1
As = = = 0. 2
αs 5
The gain in normal values can be converted to dB-gain as shown below.
Ap,dB = 20 log Ap = 20 log 0.8 = –1.9382 dB
As,dB = 20 log As = 20 log 0.2 = –13.9794 dB
Note : The dB-gain and dB-attenuation are numerically same, but dB-gain is negative and dB-attenuation
is positive.
Chapter 7 - IIR Filters 7. 26
The attenuation response and log-attenuation response of a digital IIR lowpass filter are shown in
fig 7.10.
1
α=
H ( e jω )
αs , dB = 14dB
αs = 5

αp = 1.2 5
δp , d B
αp , d B = 2dB

0 ωp ωs ω 0 ωp ωs ω
F ig 7.10a : A tte n u atio n v s ω. F ig 7.10b : d B -atte n ua tio n v s ω.
F ig 7 .1 0 : A tten u a tio n resp o n se o f d ig ita l IIR lo w p a ss filte r.
Case iii : Ripple at passband and stopband edge frequency
Sometimes, the specifications are given in terms of ripple or tolerance in the passband and stopband.
The ripple can be in normal values or in decibels (dB).
Let, dp = Passband ripple.
ds = Stopband ripple.
dp, dB = 20 log dp = Passband ripple in dB.
ds,dB = 20 log ds = Stopband ripple in dB.
The dB-ripples can be converted to normal values as shown below.
dp = 1 – 10(–dp,dB/20)
ds = 10(–ds,dB/20)
The ripples in normal values can be converted to gain or attenuation as shown below.
1 1
Ap = 1 − δp αp = =
Ap 1 − δp
1 1
As = δs αs = =
As δs
The ripples in dB can be converted to dB-gain or dB-attenuation as shown below. Usually, the ripples
are specified as positive dB.
A p ,dB = −δ p ,dB α p ,dB = δ p ,dB
A s,dB = −δ s,dB α s,dB = δ s,dB

The ripples in dB can be converted to gain or attenuation in normal values as shown below.
( − δ p ,dB / 20 ) ( δ p ,dB / 20)
A p = 10 α p = 10
( − δ s,dB / 20) ( δ s,dB / 20 )
A s = 10 α s = 10
7. 27 Digital Signal Processing
Example
Let, dp,dB = +1.9382 dB » +2 dB
ds,dB = +13.9794 dB » +14 dB
The dB-ripples can be converted to ripples in normal values as shown below.
dp = 1–10(-dp,dB /20) = 1–10(–1.9382/20) = 0.2
ds = 10(-ds,dB /20) = 10(–13.9794/20) = 0.2
The dB ripples can be converted to dB-gain and dB-attenuation as shown below.
Ap,dB = –dp,dB = – 1.9382 dB » –2 dB
As,dB = –ds,dB = – 13.9794 dB » –14 dB
a p,dB = dp,dB = + 1.9382 dB » +2 dB
a s,dB = ds,dB = + 13.9794 dB » +14 dB
The dB-ripples can be converted to gain and attenuation in normal values.
Ap = 10(-dp,dB /20) = 10(–1.9382/20) = 0.8
As = 10(-ds,dB /20) = 10(–13.9794/20) = 0.2
a p = 10(dp,dB /20) = 10(1.9382/20) = 1.25

a s = 10(ds,dB /20) = 10(13.9794/20) = 5

7.6 Design of Lowpass Digital Butterworth Filter

The popular methods of designing IIR digital filter involves the design of equivalent analog filter and
then converting the analog filter to digital filter. Hence to design a Butterworth IIR digital filter, first an analog
butterworth filter transfer function is determined using the given specifications. Then the analog filter transfer
function is converted to a digital filter transfer function by using either impulse invariant transformation or
bilinear transformation.
7.6.1 Analog Butterworth Filter
The analog butterworth filter is designed by approximating the ideal analog filter frequency response,
H(jW ) using an error function. The error function is selected such that the magnitude is maximally flat in the
passband and monotonically decreasing in the stopband. (Strictly saying, the magnitude is maximally flat at
the origin i.e., at W = 0, and monotonically decreasing with increasing W ).
The magnitude response of lowpass filter obtained by this approximation is given by,
2 1
H ( jΩ) = 2N ..... (7.35)
1+ e j
Ω
Ωc

Since, |H(jW )|2 = H(jW ) H*(jW ) = H(jW ) H(–jW ), the equation (7.35) can be written as shown below.

1 ..... (7.36)
H ( j Ω ) H ( − jΩ ) = 2N
1+ e j
Ω
Ωc

We know that the frequency response H(jW ) of an analog filter is obtained by letting s = jW in the
analog transfer function H(jW ). Hence substituting W by s/j in equation (7.36) gives the system transfer
function.
Chapter 7 - IIR Filters 7. 28
1 1 ..... (7.37)
∴ H (s) H(– s) = =
F s jI
1+ G J
2N
F s I
1+ G
2
N

HΩ K c H j Ω JK
2 2
c

In equation (7.37), when s/W c is replaced by sn (i.e., letting W c = 1 rad/sec) the transfer function is
called normalized transfer function.

1
∴ H (sn ) H(– sn ) = ..... (7.38)
1 + ( – s2n ) N

The transfer function of equation (7.38) will have 2N poles which are given by the roots of the
denominator polynomial. It can be shown that the poles of the transfer function symmetrically lies on a unit
circle in s-plane with angular spacing of p/N. (Refer example 7.8 to example 7.13).
Properties of Butterworth Filters
1. The Butterworth filters are all pole designs. (i.e., the zeros of the filters exist at infinity).

2. At the cutoff frequency W c the magnitude of normalized Butterworth filter is 1 2

(i.e., |H(jW )| = 1 2 = 0.707). Hence the dB magnitude at the cutoff frequency will be 3 dB
less than the maximum value.
3. The filter order N completely specifies the filter.
4. The magnitude is maximally flat at the origin.
5. The magnitude is a monotonically decreasing function of W .
6. The magnitude response approaches the ideal response as the value of N increases.
7.6.2 Poles of Butterworth Lowpass Filter
Let us equate the denominator polynomial of equation (7.38) to zero and solve the 2N poles of
Butterworth lowpass filter.

∴ 1 + ( − s2n ) N = 0 ⇒ 1 + ( −1) N s2N .....(7.39)

n =0

case i : When N is odd

When N is odd, (–1)N = –1
Hence the equation (7.39) can be written as,
1
1 − s2N
n =0 ⇒ s2N
n = 1 ⇒ sn = 1
2N

For integer k,
Now, sn will have 2N values which are given by 2N roots of unity. These
e j2pk = cos2pk + jsin2pk
2N roots can be evaluated by taking 1 as ej2pk, where k is an iteger.
= 1 + j0 = 1
1 1 πk
j
∴ sn = 12 N = e j2 πk e j 2N =e N

Therefore, when N is odd, the 2N poles of Butterworth filter are given by the equation,
πk
j
sn = e N ; for k = 1, 2, 3, ..... 2N ......(7.40)
7. 29 Digital Signal Processing
case ii : When N is even
When N is even, (–1)N = 1
Hence the equation (7.39) can be written as,
1
1 + s2N
n =0 ⇒ s2N
n = −1 ⇒ sn = ( −1)
2N

Now, sn will have 2N values which are given by 2N roots of –1. These 2N roots can be evaluated by
taking –1 as ej(2k –1)p, where k is an iteger. For integer k,
1 1 ( 2 k −1) π
e j(2k – 1)p = cos(2k – 1)p + jsin(2k – 1)p
j
e
∴ sn = ( −1) 2 N = e j( 2 k − 1) π j 2N =e 2N
= –1 + j0 = –1
Therefore, when N is even the 2N poles of Butterworth filter are given by,
( 2 k −1) π
j
∴ sn = e 2N ; for k = 1, 2, 3, ..... 2N ......(7.41)

jΩ s-plane
Example 7.8
Unit circle
Determine the poles of lowpass Butterworth filter for
N = 1. Sketch the location of poles on s-plane and hence
determine the normalized transfer function of lowpass filter.

Solution p1 p2
X X σ
jπk −1 1
When N = 1, from equation (7.40), sn = e ; for k = 1, 2

When k = 1 ; sn = ejπ ×1 = 1∠π = cos π + j sin π = −1 + j0 = p1

When k = 2 ; sn = e jπ × 2 = 1∠2π = cos 2π + j sin 2π = 1 + j0 = p2

The transfer function is formed using the poles lying on left half
of s-plane. The pole lying on left half of s-plane is p1. F ig 1 : L o c atio n o f p oles o n s-p lan e ,
w h en N = 1.
\ sn = p1 = –1 Þ sn – p1 = 0 Þ sn + 1 = 0
1 1
∴ Normalized transfer function, H(sn ) = =
sn − p1 s + 1

jΩ
Example 7.9 s -plane
U nit circ le

Determine the poles of lowpass Butterworth filter for p2 + j0.7 07 p1

X
X

N = 2. Sketch the location of poles on s-plane and hence

determine the normalized transfer function of lowpass filter.
3 π/4
π/4
−0.707 0.707 1 σ

Solution X
p 1∗
X

p ∗2 −j0.707
When N = 2, from equation (7.41),

j
( 2k − 1) π F ig 1 : L o c atio n o f p oles o n s-p lan e ,
sn = e 4 ; for k = 1, 2, 3, 4 w h en N = 2.
( 2− 1) π π
j
4
j
4
π π
When k = 1 ; sn = e =e = 1∠π / 4 = cos + j sin = 0. 707 + j0. 707 = p1
4 4
Chapter 7 - IIR Filters 7. 30
( 4 −1) π 3π
j
4
j
4
3π 3π
When k = 2 ; sn = e =e = 1∠3π / 4 = cos + j sin = −0. 707 + j0. 707 = p2
4 4
(6 −1) π 5π
j j 5π 5π
When k = 3 ; sn = e 4 =e 4 = 1∠5π / 4 = cos + j sin = −0. 707 − j0. 707 = p∗2
4 4
( 8 −1) π 7π
j j 7π 7π
When k = 4 ; sn = e 4 =e 4 = 1∠7π / 4 = cos + j sin = 0. 707 − j0. 707 = p1∗
4 4
The transfer function is formed using the poles lying on left half of s-plane. The poles lying on left half
of s-plane are p2 and p2*.

\ sn = p2 = –0.707 + j0.707 Þ sn – p2 = 0 Þ (sn + 0.707 –j0.707) = 0

sn = p2* = –0.707 – j0.707 Þ sn – p2* = 0 Þ (sn + 0.707 + j0.707) = 0

1
∴ Normalized transfer function, H(sn ) =
(sn − p 2 ) (sn − p∗2 )
1
∴ H(sn ) =
(sn + 0.707 − j0.707) (sn + 0.707 + j0.707)
1 (a + b) (a – b) = a2 – b2
=
(sn + 0.707)2 − (j0.707)2
1
=
(sn + 0.707)2 + 0.7072
1 1
= =
sn2 + 2 × 0.707 sn + 0.707 2 + 0.7072 sn2 + 1414
. sn + 1

Example 7.10 jΩ s -plane

p2
X p1
Determine the poles of lowpass Butterworth filter for
X
+ j0.8 66 U nit circ le
N = 3. Sketch the location of poles on s-plane and hence
determine the normalized transfer function of lowpass filter.
2 π/3
p3 π/3 p4
X 0.5
X σ
0.5 1

−j0.866 X
Solution
X

p ∗2 p 1∗

When N = 3, from equation (7.40), F ig 1 : L o c atio n o f p oles o n s-p lan e ,

w h en N = 3.
πk
j
sn = e 3 ; for k = 1, 2, 3, 4, 5, 6
π ×1
j
3
π π
When k = 1 ; sn = e = 1∠π / 3 = cos + j sin = 0.5 + j0. 866 = p1
3 3
π×2
j
3
2π 2π
When k = 2 ; sn = e = 1∠2π / 3 = cos + j sin = −0.5 + j0. 866 = p2
3 3
π ×3
j
3
3π 3π
When k = 3 ; sn = e = 1∠3π / 3 = cos + j sin = −1 + j0 = p3
3 3
7. 31 Digital Signal Processing
π×4
j 4π 4π
When k = 4 ; sn = e 3 = 1∠4π / 3 = cos + j sin = −0.5 − j0.866 = p∗2
3 3
π×5
j 5π 5π
When k = 5 ; sn = e 3 = 1∠5π / 3 = cos + j sin = 0.5 − j0.866 = p1∗
3 3
π ×6
j
3
6π 6π
When k = 6 ; sn = e = 1∠6π / 3 = cos + j sin = 1 + j0 = p4
3 3
The transfer function is formed using the poles lying on left half of s-plane. The poles lying on left half
of s-plane are p2, p2* and p3.

\ sn = p2 = –0.5 + j0.866 Þ sn – p 2 = 0 Þ (sn + 0.5 –j0.866) = 0

sn = p2* = –0.5 – j0.866 Þ sn – p2* = 0 Þ (sn + 0.5 + j0.866) = 0

sn = p3 = –1 Þ sn – p 3 = 0 Þ (sn + 1) = 0

1
∴ Normalized transfer function, H(sn ) =
(sn − p3 ) (sn − p2 ) (sn − p∗2 )

1
∴ H(sn ) =
(sn + 1) (sn + 0.5 − j0.866) (sn + 0.5 + j0.866)
1
= (a + b) (a – b) = a2 – b2
(sn + 1) ((sn + 0.5)2 − (j0.866)2 )
1 1
= =
(sn + 1) ((sn + 0.5)2 + 0.8662 ) (sn + 1) (sn2 + 2 × 0.5 sn + 0.52 + 0.8662 )
1
=
(sn + 1) (sn2 + s + 1)

Example 7.11 p3
jΩ s -plane
p2
U nit circ le
X

Determine the poles of lowpass Butterworth filter

X

for N = 4. Sketch the location of poles on s-plane and

hence determine the normalized transfer function of p4
X X p1
lowpass filter. 3 π/8
π/8
σ

p ∗4 X
X
p ∗1

Solution
X
X

p ∗3 p ∗2
When N = 4, from equation (7.41),
F ig 1 : L o c atio n o f p oles o n s-p lan e ,
( 2k − 1) π w h en N = 4.
j
sn = e 8 ; for k = 1, 2, 3, 4, 5, 6, 7, 8
( 2− 1) π π
j
8
j
8
π π
When k = 1 ; sn = e =e = 1∠π / 8 = cos + j sin = 0.924 + j0. 383 = p1
8 8
( 4 − 1) π 3π
j
8
j
8
3π 3π
When k = 2 ; sn = e =e = 1∠3π / 8 = cos + j sin = 0.383 + j0. 924 = p2
8 8
(6 −1) π 5π
j
8
j
8
5π 5π
When k = 3 ; sn = e =e = 1∠5π / 8 = cos + j sin = −0.383 + j0. 924 = p3
8 8
Chapter 7 - IIR Filters 7. 32
( 8 −1) π 7π
j
8
j
8
7π 7π
When k = 4 ; sn = e =e = 1∠7π / 8 = cos + j sin = −0.924 + j0. 383 = p4
8 8
(10 − 1) π 9π
j j 9π 9π
When k = 5 ; sn = e 8 =e 8 = 1∠9π / 8 = cos + j sin = −0.924 − j0. 383 = p∗4
8 8
(12−1)π 11π
j j 11π 11π
When k = 6 ; sn = e 8 =e 8 = 1∠11π / 8 = cos + j sin = −0.383 − j0. 924 = p3∗
8 8
(14 −1) π 13π
j j 13π 13π
When k = 7 ; sn = e 8 =e 8 = 1∠13π / 8 = cos + j sin = 0.383 − j0. 924 = p∗2
8 8
(16 − 1) π 15π
j j 15π 15π
When k = 8 ; sn = e 8 =e 8 = 1∠15π / 8 = cos + j sin = 0.924 − j0. 383 = p1∗
8 8
The transfer function is formed using the poles lying on left half of s-plane. The poles lying on left half
of s-plane are p3, p3*, p4 and p4*.
\ sn = p3 = –0.383 + j0.924 Þ sn – p3 = 0 Þ (sn + 0.383 –j0.924) = 0

sn = p3* = –0.383 – j0.924 Þ sn – p3* = 0 Þ (sn + 0.383 + j0.924) = 0

sn = p4 = –0.924 + j0.383 Þ sn – p4 = 0 Þ (sn + 0.924 – j0.383) = 0

sn = p4* = –0.924 – j0.383 Þ sn – p4* = 0 Þ (sn + 0.924 + j0.383) = 0

1
∴ Normalized transfer function, H(sn ) =
(sn − p3 ) (sn − p3∗ ) (sn − p 4 ) (sn − p∗4 )

1
∴ H(sn ) =
(sn + 0.383 − j0.924) (sn + 0.383 + j0.924) (sn + 0.924 + j0.383) (sn + 0.924 + j0.383)
1
=
d
(sn + 0.383)2 − (j0.924)2 (sn + 0.924)2 − (j0.383)2 id i (a + b) (a – b) = a2 – b2
1
=
d(s n id
+ 0.383)2 + 0.9242 (sn + 0.924)2 + 0.3832 i
1
=
(sn2 + 2 × 0.383 sn + 0.3832 + 0.9242 ) (sn2 + 2 × 0.924 sn + 0.9242 + 0.3832 )
1
=
(sn2 + 0.766 sn + 1) (sn2 + 1848
. sn + 1)

Example 7.12 p3
jΩ s -plane
p2
U nit circ le
X

Determine the poles of lowpass Butterworth filter

X

for N = 5. Sketch the location of poles on s-plane and p4

hence determine the normalized transfer function of
X X p1
lowpass filter. 2 π/5
p5 p6
π/5
X X σ

X
p ∗4 X p 1∗
X
X

p ∗3 p ∗2
F ig 1 : L o c atio n o f p oles o n s-p lan e ,
w h en N = 5.
7. 33 Digital Signal Processing
Solution πk
j
When N = 5, from equation (7.40), sn = e 5 ; for k = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
π ×1
j π π
When k = 1 ; sn = e 5 = 1∠π / 5 = cos + j sin = 0.809 + j0. 588 = p1
5 5
π×2
j
5
2π 2π
When k = 2 ; sn = e = 1∠2π / 5 = cos + j sin = 0.309 + j0. 951 = p2
5 5
π ×3
j
5
3π 3π
When k = 3 ; sn = e = 1∠3π / 5 = cos + j sin = −0.309 + j0. 951 = p3
5 5
π×4
j
5
4π 4π
When k = 4 ; sn = e = 1∠4π / 5 = cos + j sin = −0.809 + j0. 587 = p4
5 5
π×5
j
5
5π 5π
When k = 5 ; sn = e = 1∠5π / 5 = cos + j sin = −1 + j0 = p5
5 5
π ×6
j 6π 6π
When k = 6 ; sn = e 5 = 1∠6π / 5 = cos + j sin = −0.809 − j0. 587 = p∗4
5 5
π ×7
j 7π 7π
When k = 7 ; sn = e 5 = 1∠7π / 5 = cos + j sin = −0.309 − j0. 951 = p∗3
5 5
π×8
j 8π 8π
When k = 8 ; sn = e 5 = 1∠8π / 5 = cos + j sin = 0.309 − j0. 951 = p∗2
5 5
π×9
j 9π 9π
When k = 9 ; sn = e 5 = 1∠9π / 5 = cos + j sin = 0.809 − j0. 588 = p1∗
5 5
π ×10
j
5
10π 10π
When k = 10 ; sn = e = 1∠10π / 5 = cos + j sin = 1 + j0 = p6
5 5
The transfer function is formed using the poles lying on left half of s-plane. The poles lying on left half
of s-plane are p3, p3*, p4, p4* and p5.
\ sn = p3 = –0.309 + j0.951 Þ sn – p3 = 0 Þ (sn + 0.309 –j0.951) = 0

sn = p3* = –0.309 – j0.951 Þ sn – p3* = 0 Þ (sn + 0.309 + j0.951) = 0

sn = p4 = –0.809 + j0.587 Þ s n – p4 = 0 Þ (sn + 0.809 – j0.587) = 0

sn = p4* = –0.809 – j0.587 Þ sn – p4* = 0 Þ (sn + 0.809 + j0.587) = 0

sn = p5 = –1 Þ sn – p5 = 0 Þ (sn + 1) = 0
1
∴ Normalized transfer function, H(sn ) =
(sn − p5 ) (sn − p3 ) (sn − p3∗ ) (sn − p4 ) (sn − p∗4 )
1
∴ H(sn ) =
(sn + 1) (s + 0.309 − j0.951) (sn + 0.309 + j0.951) (sn + 0.809 − j0.587) (sn + 0.809 + j0.587)
1
= (a + b) (a – b) = a2 – b2
(sn + 1) ((sn + 0.309)2 − (j0.951)2 ) ((sn + 0.809)2 − (j0.587)2 )
1
=
(sn + 1) ((sn + 0.309)2 + 0.9512 ) ((sn + 0.809)2 + 0.5872 )
1
=
(sn + 1) (sn2 + 2 × 0.309 sn + 0.3092 + 0.9512 ) (sn2 + 2 × 0.809 sn + 0.8092 + 0.5872 )
1
=
(sn + 1) (sn2 + 0.618 sn + 1) (sn2 + 1.618 sn + 1)
Chapter 7 - IIR Filters 7. 34
jΩ
Example 7.13 s -plane
p4
p3
Determine the poles of lowpass Butterworth filter

X
U nit circ le

X
for N = 6. Sketch the location of poles on s-plane and p5 p2
X

X
hence determine the normalized transfer function of
lowpass filter.
p6 3 π/1 2
X X p1
π/12
σ
X X p 1∗
p ∗6

X
p ∗5 p ∗2
Solution

X
X
When N = 6, from equation (7.41), p ∗4 p ∗3

j
( 2k −1)π F ig 1 : L o c atio n o f p oles o n s-p lan e ,
sn = e 12 ; for k = 1, 2, 3, ...... 12 w h en N = 6.
( 2−1) π π
j
12
j
12
π π
When k = 1 ; sn = e =e = 1∠π / 12 = cos + j sin = 0.966 + j0. 259 = p1
12 12
( 4 −1) π 3π
j
12
j
12
3π 3π
When k = 2 ; sn = e =e = 1∠3π / 12 = cos + j sin = 0.707 + j0. 707 = p2
12 12
(6 −1) π 5π
j
12
j
12
5π 5π
When k = 3 ; sn = e =e = 1∠5π / 12 = cos + j sin = 0.259 + j0. 966 = p3
12 12
(8 −1) π 7π
j
12
j
12
7π 7π
When k = 4 ; sn = e =e = 1∠7π / 12 = cos + j sin = −0.259 + j0. 966 = p4
12 12
(10 −1) π 9π
j
12
j
12
9π 9π
When k = 5 ; sn = e =e = 1∠9π / 12 = cos + j sin = −0.707 + j0. 707 = p5
12 12
(12−1) π 11π
j
12
j
12
11π 11π
When k = 6 ; sn = e =e = 1∠11π / 12 = cos + j sin = −0.966 + j0. 259 = p6
12 12
(14 −1) π 13π
j j 13π 13π
When k = 7 ; sn = e 12 =e 12 = 1∠13π / 12 = cos + j sin = −0.966 − j0. 259 = p∗6
12 12
(16 −1) π 15π
j j 15π 15π
When k = 8 ; sn = e 12 =e 12 = 1∠15π / 12 = cos + j sin = −0.707 − j0. 707 = p∗5
12 12
(18 −1) π 17 π
j j 17π 17π
When k = 9 ; sn = e 12 =e 12 = 1∠17π / 12 = cos + j sin = −0.259 − j0. 966 = p∗4
12 12
( 20 −1)π 19π
j j 19π 19π
When k = 10 ; sn = e 12 =e 12 = 1∠19π / 12 = cos + j sin = 0.259 − j0. 966 = p∗3
12 12
( 22− 1) π 21π
j j 21π 21π
When k = 11 ; sn = e 12 =e 12 = 1∠21π / 12 = cos + j sin = 0.707 − j0. 707 = p∗2
12 12
( 24 − 1) π 23π
j j 23π 23π
When k = 12 ; sn = e 12 =e 12 = 1∠23π / 12 = cos + j sin = 0.966 − j0. 259 = p1∗
12 12
The transfer function is formed using the poles lying on left half of s-plane. The poles lying on left half
of s-plane are p4, p4*, p5, p5*, p6 and p6*.

\ sn = p4 = –0.259 + j0.966 Þ sn – p4 = 0 Þ (sn + 0.259 –j0.966) = 0

sn = p4* = –0.259 – j0.966 Þ sn – p4* = 0 Þ (sn + 0.259 + j0.966) = 0

sn = p5 = –0.707 + j0.707 Þ sn – p5 = 0 Þ (sn + 0.707 – j0.707) = 0

7. 35 Digital Signal Processing
sn = p5* = –0.707 – j0.707 Þ sn – p5* = 0 Þ (sn + 0.707 + j0.707) = 0

sn = p6 = –0.966 + j0.259 Þ s n – p6 = 0 Þ (sn + 0.966 – j0.259) = 0

sn = p6* = –0.966 – j0.259 Þ sn – p6* = 0 Þ (sn + 0.966 + j0.259) = 0

1
∴ Normalized transfer function, H(sn ) =
(sn − p4 ) (sn − p∗4 ) (sn − p5 ) (sn − p∗5 ) (sn − p6 ) (sn − p∗6 )

1
∴ H(sn ) =
(sn + 0.259 − j0.966) (sn + 0.259 + j0.966) (sn + 0.707 − j0.707) (sn + 0. 707 + j0.707)
(sn + 0.966 − j0.259) (sn + 0.966 + j0.259)
1
=
d(s n + 0.259)2 − ( j0.966)2 i d(s
n + 0.707)2 − (j0.707)2 i d(s n + 0.966)2 − (j0.259)2 i
1
=
d i d
(sn + 0.259)2 + 0.9662 + (sn + 0.707)2 + 0.7072 i d(sn + 0.966)2 + 0.2592 i
1
∴ H(sn ) =
(sn2 + 2 × 0.259 sn + 0.2592 + 0.9662 ) (sn2 + 2 × 0.707 sn + 0.7072 + 0.7072 )
(sn2 + 2 × 0.966 sn + 0.9662 + 0.2592 )
1
=
(sn2 + 0.518 sn + 1) (sn2 + 1.414 sn + 1) (sn2 + 1.932 sn + 1)

7.6.3 Transfer function of Analog Butterworth Lowpass Filter

For a stable and causal filter the poles should lie on the left half of s-plane. Hence the desired filter
transfer function is formed by choosing the N-number of left half poles. When N is even, all the poles are
complex and exist as conjugate pair. When N is odd, one of the poles is real and all other poles are complex
and exist as conjugate pair. Therefore the transfer function of Butterworth filters will be a product of second
order factors. (Refer example 7.8 to example 7.13). The analog filter transfer function of normalized and
unnormalized butterworth lowpass filters are given below.
Normalized Butterworth Lowpass Filter Transfer Function
Let, N be the order of the filter.
Let, H(sn) be the normalized Butterworth lowpass filter transfer function.
When N is even,
N
2
1 .....(7.42)
H (sn ) = ∏ s2n + b k sn + 1
k=1

When N is odd,
N−1
2
1 1 .....(7.43)
H(sn ) =
sn + 1 ∏ s2n + b k sn + 1
k=1

where, b k = 2 sin ( 2 k – 1) π .....(7.44)

2N
Chapter 7 - IIR Filters 7. 36
Table 7.2 : Summary of Butterworth Lowpass Filter Normalized Transfer Function

Order, N Normalized tansfer function, H(sn)

1
1 sn + 1

1
2 s2n + 1414
. sn + 1

1
3 (sn + 1) (s2n + sn + 1)

1
4 (s2n + 0.765sn + 1) (s2n + 1848
. sn + 1)

1
5 (sn + 1) ( s2n + 0.618 sn + 1) (s2n + 1618
. sn + 1)

1
6 (s2n + 1932
. sn + 1) (s2n + 1414
. sn + 1) (s2n + 0.518 sn + 1)

Unnormalized Butterworth Lowpass Filter Transfer Function

The unnormalized transfer function is obtained by replacing sn by s/W c, in the normalized transfer
function, where W c is the 3-dB cutoff frequency of the lowpass filter.
Let, N be the order of the filter.
Let, H(s) be the unnormalized Butterworth lowpass filter transfer function.
When N is even, H(s) is obtained by letting sn ® s/W c in equation (7.42).
N
2
1
∴ H (s) = ∏ s2 + b k sn + 1
k=1 n
s
sn =
Ωc
N
2
Ωc2 .....(7.45)
= ∏ s2 + b k Ωcs + Ωc2
k=1
When N is odd, H(s) is obtained by letting sn ® s/W c in equation (7.43).
N−1
2
1 1
∴ H ( s) =
sn + 1 ∏ s2n + b k sn + 1
k=1
s
sn =
Ωc
N−1
Ωc 2
Ωc2
=
s + Ωc ∏ s2 + b k Ωc + Ωc2
.....(7.46)
k=1
 7. 37 Digital Signal Processing
7.6.4 Frequency Response of Analog H (j Ω)
1.0
Lowpass Butterworth Filter N =10

The frequency response of Butterworth filter N =1

N =2
depends on the order N. The magnitude response 1 N =4
= 0 .707
(frequency response) for different values of N are 2
shown in fig 7.11. From fig 7.11 it can be observed
that the approximated magnitude response
Id eal
approaches the ideal response as the value of N respo nse
increases. N=1

N = 10
N=2
N=4

ΩC Ω
³ F ig 7 .11 : M a g n itu d e resp o n se o f b u tterw o rth
lo w p a ss filter fo r va rio u s va lu es o f N .
7.6.5 Order of the Lowpass Butterworth Filter
In Butterworth filters the frequency response of the filter depends on the order, N. Hence the order N
has to be estimated to satisfy the given specifications.
Usually the specifications of the filter are given in terms of gain at a passband and stopband frequency.
Let, Ap = Gain or Magnitude at a passband frequency W p.
As = Gain or Magnitude at a stopband frequency W s.
Calculate a parameter N1 using equation (7.47) and correct it to nearest integer. Choose N such that N ³ N1.

log
LM e1/A 2s − 1
j OP
N1 = 1
MN e
1/A 2p − 1
j PQ ..... (7.47)
2
log
FG IJ
Ωs
H K
Ωp

Sometimes, the specifications of the filter are given in terms of dB-attenuation at a passband and
stopband frequency.
Let, a p, dB = dB-attenuation at a passband frequency W p.
a s, dB = dB-attenuation at a stopband frequency W s.
Calculate a parameter N1 using equation (7.48) and correct it to nearest integer. Choose N such that N ³ N1.
LMF 0.1α s,dB I
1
2
OP
10 −1
log MG J PP
MNH 10 0.1α p ,dB
− 1K
Q ..... (7.48)
N1 = Ωs
log Ωp

7.6.6 Cutoff Frequency of Lowpass Butterworth Filter

The IIR filters are designed to satisfy a prescribed gain or attenuation at a passband and stopband
frequency. But practically the 3-dB cutoff frequency, W c is used to decide the useful frequency range of the
filter. Therefore, in Butterworth filter design the passband and stopband specifications are used to estimate
the order, N of the filter and Nth order normalized Butterworth lowpass filter is designed. Then the normalized
lowpass filter is unnormalized using the cutoff frequency.
The cutoff frequency of lowpass Butterworth filter can be calculated using the following equations.
Chapter 7 - IIR Filters 7. 38
Case i : When the specifications are Ap, As, w p, w s
Ωs
Cutoff frequency, Ωc = 1 ..... (7.49)
FH 1/ A s2 IK −1 2 N
Alternatively,
Ωp ..... (7.50)
Cutoff frequency, Ωc = 1
FH 1/ A 2p IK −1 2N

The equation (7.49) is preferable to equation (7.50), because the cutoff frequency W c calculated using
equation (7.49) ensures smallest amplitude distortion (or ripple) in the passband.
For bilinear transformation,
2 ωp 2 ω
Ωp = tan ; Ωs = tan s
T 2 T 2
For impulse invariant transformation,
ωp ωs
Ωp = ; Ωs =
T T
where T is the sampling time.
Case ii : When the specifications are a p,dB, a s,dB, w p, w s

Ωs
Cutoff frequency, Ωc = 1
0.1α s,dB .....(7.51)
e
10 −1 j 2N

Alternatively ,
Ωp
Cutoff frequency, Ωc = 1
0.1α p ,dB .....(7.52)
e
10 −1 j 2N

The calculation of W p and W s are same as that of case (i).

The equation(7.51) is preferable to equation (7.52), because, the cutoff frequency, W c calculated by
using equation (7.51) ensures smallest magnitude distortion (or ripple) in the passband.
7.6.7 Design Procedure for Lowpass Digital Butterworth IIR Filter
Let, w p = Passband edge digital frequency in rad/sample.
w s = Stopband edge digital frequency in rad/sample.
1
T= = Sampling time in sec.
Fs
where, Fs = sampling frequency in Hz.
Ap = Gain at a passband frequency w p.
As = Gain at a stopband frequency w s.
 7. 39 Digital Signal Processing
Note-1: If passband dB-attenuation, a p,dB and stopband dB-attenuation, a s,dB are specified, then convert
them to Ap and As as shown below.
Ap = 10(–ap,dB / 20) a p,dB and as,dB are positive dB
As = 10(–as,dB / 20)
Bd d n a b s s a p t eshnt i sdeNote-2:
emii ft iem
c oeSp s s t t a - Bd d n a b s s a p f o d a e , n o i t aun e
ap,dB. Remember that ap,dB equal to dp,dB. (refer section 7.5).
Note-3: If T is not specified then take T = 1 second.
1. Choose either bilinear or impulse invariant transformation, and determine the specifications of
equivalent analog filter. The gain or attenuation of analog filter is same as digital filter. The band edge
frequencies are calculated using the following equations.
Let, Wp = Passband edge analog frequency corresponding to w p.
Ws = Stopband edge analog frequency corresponding to w s.
For bilinear transformation,
ωp Note : If either T or Fs is not
2 .....(7.53)
Ωp = tan specified then take T= 1 second.
T 2
1
2 ω .....(7.54) If Fs is specified, then T =
Ωs = tan s Fs
T 2
For impulse invariant transformation,
ωp .....(7.55)
Ωp =
T
ωs .....(7.56)
Ωs =
T
2. Decide the order N of the filter. In order to estimate the order N, calculate a parameter N1 using the
following equation.

log
LM e1/ A 2s − 1
j OP
N1 = 1
MN e
1/ A 2p − 1
j PQ .....(7.57)
2 F I
log G J Ωs
H K Ωp

Choose N such that, N ³ N1. Usually N is chosen as nearest integer just greater than N1.
3. Determine the normalized transfer function, H(sn) of the analog lowpass filter.
When N is even,
N
2
1 .....(7.58)
H ( sn ) = ∏ s2n + b k sn + 1
k=1

When N is odd,
N−1
2
1 1
H(sn ) =
sn + 1 ∏ s2n + b k sn + 1
.....(7.59)
k=1

where, b k = 2 sin ( 2 k −1) π .....(7.60)

2N
Chapter 7 - IIR Filters 7. 40
4. Calculate the analog cutoff frequency, W c.

Ωs .....(7.61)
Cutoff frequency, Ωc = 1
1/ A 2s − 1 2 N
e j
5. Determine the unnormalized analog transfer function H(s) of the lowpass filter.
H ( s) = H (sn ) s s
n=
Ωc

When the order N is even, H(s) is obtained by letting sn ® s/W c in equation (7.58).
N N
2
1 2
Ωc2
∴ H (s) = ∏ s2n + b k sn + 1
= ∏= s2 + b k Ωcs + Ωc2 .....(7.62)
k=1 k 1
s
sn =
Ωc

When the order N is odd, H(s) is obtained by letting sn ® s/W c in equation (7.59).
N−1 N−1
1 2
1 Ωc 2
Ωc2
∴ H(s) =
sn + 1 ∏ s2n + b k sn + 1
=
s + Ωc ∏= s2 + b k Ωcs + Ωc2 .....(7.63)
k=1 k 1
s
sn =
Ωc

6. Determine the transfer function of digital filter, H(z). Using the chosen transformation in step-1,
transform H(s) to H(z). When impulse invariant transformation is employed, if T < 1, then multiply
H(z) by T to normalize the magnitude.
7. Realize the digital filter transfer function H(z) by a suitable structure.
8. Verify the design by sketching the frequency response H(ejw ).

H (e jω ) = H ( z)
z = e jω

Note : The basic filter design is lowpass filter design. The highpass, bandpass or bandstop filters are
obtained from lowpass filter design by frequency transformation.

7.7 Design of Lowpass Digital Chebyshev Filter

For designing a Chebyshev IIR digital filter, first an analog filter is designed using the given
specifications. Then the analog filter transfer function is transformed to digital filter transfer function by
using either impulse invariant transformation or bilinear transformation.
Analog Chebyshev Filter
The analog Chebyshev filter is designed by approximating the ideal frequency response using an
error function. The approximation function is selected such that the error is minimized over a prescribed
band of frequencies. There are two types of Chebyshev approximation. In type-1 approximation, the error
function is selected such that, the magnitude response is equiripple in the passband and monotonic in the
stopband. In type-2 approximation the error function is selected such that, the magnitude response is monotonic
in passband and equiripple in stopband. The type-2 magnitude response is also called inverse Chebyshev
response. The type-1 design is presented in this book.
7. 41 Digital Signal Processing
The magnitude response of Type-1 lowpass filter is given by,
2 1
H ( jΩ) =
1 + ∈2 C2N
FG Ω IJ ..... (7.64)
HΩ K c

where Î is attenuation constant and CN(W /W c) is the Chebyshev polynomial of the first kind of degree N.
1
L1 O 2
The attenuation constant, ∈ = M − 1P ..... (7.65)
MN A 2
p PQ
where, Ap is the gain or magnitude at passband edge frequency W p
For small values of N the Chebyshev polynomial is given by,
R|cos( N cos x)−1
; for |x| ≤ 1
C N ( x) = S|cosh( N cosh x) −1 ..... (7.66)
T ; for |x| > 1

For large values of N the Chebyshev polynomial is given by the recurrence relation,
CN(x) = 2xCN – 1(x) – CN – 2 (x) ..... (7.67)
with initial values C0(x) = 1 and C1(x) = x
The transfer function of the analog system can be obtained from equation (7.64) by substituting
W by s/j.
1
∴ H ( s) H (– s) =
1 + ∈2 C 2N
FG s j IJ ..... (7.68)
HΩ K c

For the normalized transfer function, let us replace s/W c by sn.

1
∴ H ( sn ) H (– sn ) = ..... (7.69)
1 + ∈2 C2N ( − jsn )
For the transfer function of equation (7.69) we can determine 2N poles which are given by the roots of
the denominator polynomial. It can be shown that the poles of the transfer function symmetrically lies on an
ellipse in s-plane.
Properties of Chebyshev Filters (Type-1)

1. The magnitude |H( jW )| oscillates between 1 and 1 1+ ∈2 within the passband and so the filter
is called equiripple in the passband.

2. The normalized magnitude response has a value of 1 1+ ∈2 at cutoff frequency W c.

3. The magnitude is monotonic outside the passband.

4. The Chebyshev Type-1 filters are all pole designs.

5. With large values of N, the transition from passband to stopband becomes more sharp and
approaches ideal characteristics.
Chapter 7 - IIR Filters 7. 42
7.7.1 Transfer Function of Analog Chebyshev Lowpass Filter
For a stable and causal filter the poles should lie on the left half of s-plane. Hence the desired filter
transfer function is obtained by selecting N number of left half poles. When N is even all the poles are complex
and exist as conjugate pair. When N is odd, one of the pole is real and all other poles are complex and exist as
conjugate pair. Therefore the transfer function of Chebyshev filters will be a product of second-order factors.
Normalized Chebyshev Lowpass Filter Transfer Function
Let, N be the order of the filter.
Let, H(sn) be the normalized Chebyshev lowpass filter transfer function.
When N is even,
N
2
Bk
H ( sn ) = ∏ s2n + b k sn + c k
..... (7.70)
k=1

When N is odd,
N−1
2
B0 Bk
H ( sn ) =
s + c0 ∏ s2n + b k sn + c k
......(7.71)
k=1

where, b k = 2 y N sin e ( 2 k − 1) π
2N
j ..... (7.72)

c k = y 2N + cos2 e ( 2 k − 1)π
2N
j ..... (7.73)

c0 = yN ..... (7.74)

R 1 1 U|
1 |LMF OP L OP −
1 N 1 N

S 1 IK 2 1
− MFH
1 IK 2 1
V|
2 |MH
yN = + 1 + + 1 +
∈2 ∈ PQ MN ∈2 ∈ PQ ..... (7.75)
|TN |W
For even values of N the parameter Bk are evaluated using the equation (7.76)
1
H ( sn ) s = 1
n = 0 ..... (7.76)
(1 + ∈2 ) 2
For odd values of N the parameter Bk are evaluated using the equation (7.77)
H ( sn ) s = 1 ..... (7.77)
n = 0

While evaluating Bk using equation (7.76) or (7.77), it is normal practice to take, B0 = B1 = B2 = ..... = Bk.
Unnormalized Chebyshev Lowpass Filter Transfer Function
The unnormalized transfer function is obtained by replacing sn by s/W c in the normalized transfer
function, where W c is the cutoff frequency of the lowpass filter.
Let, N be the order of the filter.
7. 43 Digital Signal Processing
Let, H(s) be the normalized Chebyshev lowpass filter transfer function.
When N is even, H(s) is obtained by letting sn ® s/W c in equation (7.70).
N N
2 2
Bk Bk Ω2c
∴ H (s) = ∏ s2 + b k sn + c k
= ∏ s + b k Ωc s + c k Ω2c
2
k=1 n k=1 ..... (7.78)
s
sn =
Ωc

When N is odd, H(s) is obtained by letting sn ® s/W c in equation (7.71).

N −1 N −1
2 2
B0 Bk B0 Ωc Bk Ω2c
∴ H (s) =
sn + c 0 ∏ s2 + b k sn + c k
=
s + c0 Ω c ∏ 2
s + b k Ωc s + c k Ω2c ..... (7.79)
k=1 n k=1
s
sn =
Ωc

7.7.2 Order of Analog Lowpass Chebyshev Filter

In Chebyshev filters the frequency response of the filter depends on the order, N.
Hence the order has to be estimated to satisfy the given specifications.
Usually the specifications of the filter are given in terms of gain at a passband and stopband frequency.
Let, Ap = Gain or Magnitude at a passband frequency, W p.
As = Gain or Magnitude at a stopband frequency, W s.
Calculate a parameter N1, using equation (7.80) and correct it to nearest integer. Then choose N such
that N ³ N1.

LMF e j I
1
1/ A 2s − 1 2
OP
cosh −1
MMGH e J
1/ A p j − 1 K
2 PP
N1 = N Q .....(7.80)

cosh −1
FG IJ
Ωs
H K
Ωp

Sometimes, the specifications of the filter are given in terms of dB-attenuation at a passband and
stopband frequency.
Let, ap,dB = dB-attenuation at a passband frequency, W p.
Let, a s,dB = dB-attenuation at a stopband frequency, W s.
Calculate a parameter N1 using equation (7.81) and correct it to nearest integer. Choose N such
that N ³ N1.
LMF 0.1α s,dB I
1
2
OP
−1
cosh −1
MMGH 1010 0.1α p ,dB J
− 1K
PP
N1 = N Q
cosh −1
FΩ I .....(7.81)
GH Ω JK
s

p
Chapter 7 - IIR Filters 7. 44
7.7.3 Cutoff Frequency of Analog Lowpass Chebyshev Filter
The IIR filters are designed to satisfy a prescribed gain or attenuation at a passband and stopband
frequency. But practically the cutoff frequency, W c is used to decide the useful frequency range of the filter.
Therefore, in Chebyshev filter design the passband and stopband specifications are used to estimate the
order, N of the filter and Nth order normalized Chebyshev lowpass filter is designed. Then the normalized
lowpass filter is unnormalized using the cutoff frequency.
In Chebyshev filters the passband edge frequency, W p is considered as cutoff frequency, W c and this
cutoff is not equal to 3 dB cutoff frequency, W 3dB.
The 3 dB cutoff frequency of Chebyshev filter is given by, Ω3dB = Ωc cosh e 1
N
cosh −1 ∈1 j ..... (7.82)
7.7.4 Frequency Response of Analog Chebyshev Lowpass Filter
The frequency response of Chebyshev |H (j Ω)|
N =5
filter depends on the order N as shown in fig 1.0
7.12. It can be observed that the approximated
magnitude response approaches the ideal 1

response as the value of N increases. The 1+ε

2
N =2 N =4
magnitude response of Type-1 and Type-2 Ideal N =2
Chebyshev filters are shown in fig 7.13. res ponse

N =4
N =5
Ωc Ω
F ig 7 .1 2 : M a g n itu d e resp o n se o f C h eb y sh e v typ e -1
lo w p a ss filter fo r va rio u s va lu e o f N .
|H (j Ω)| |H (j Ω)|
1 1

1 1
Ap = Ap =
2 2
1+ε 1+ε

As As

Ωp Ωs Ω Ωp Ωs Ω
F ig a : C h eb yshev type-1 , w h en N is odd. F ig b : C h eb yshev type-1 , w h en N is even.

|H (j Ω)| |H (j Ω)|
1 1

1 1
Ap = Ap =
2 2
1+ε 1+ε

As As

Ωp Ωs Ω Ωp Ωs Ω
F ig c : C h eb yshev type-2 , w h en N is odd. F ig d : C h eb yshev type-2 , w h en N is even.
F ig 7 .1 3 : M a g n itu d e respo n se o f a n a lo g C h e b ysh ev filters.
 7. 45 Digital Signal Processing
7.7.5 Design Procedure for Lowpass Digital Chebyshev IIR Filter
Let, w p = Passband edge digital frequency in rad/sample.
w s = Stopband edge digital frequency in rad/sample.
1
T= = Sampling time in seconds.
Fs
where, Fs = sampling frequency in Hz.
Ap = Gain at a passband frequency w p.
As = Gain at a stopband frequency w s.
Note-1: If passband dB-attenuation, a p,dB and stopband dB-attenuation, a s,dB are specified, then convert
them to Ap and As as shown below.
Ap = 10(–ap,dB/20) ap,dB and a s,dB are positive dB
As = 10(–as,dB/20)
Bd d n a b s s a p t eshnt i sdeemii ft ie2:
mc oeSp s s t t a - Bd d n a b s s a p f o d a e , n o i t aun e
ap,dB. Remember that ap,dB equal to dp,dB. (refer section 7.5).
3: If T is not specified then take T = 1 second.
1. Choose either bilinear or impulse invariant transformation, and determine the specifications of
equivalent analog filter. The gain or attenuation of analog filter is same as digital filter. The band edge
frequencies are calculated using the following equations.
Let, Wp = Passband edge analog frequency corresponding to w p.
Ws = Stopband edge analog frequency corresponding to w s.
For bilinear transformation,
Note : If either T or Fs is not
2 ωp .....(7.83)
Ωp = tan specified then take T= 1 sec.
T 2
1
2 ω .....(7.84) If Fs is specified, then T =
Ωs = tan s Fs
T 2
For impulse invariant transformation,
ωp .....(7.85)
Ωp =
T
ω .....(7.86)
Ωs = s
T
2. Decide the order N of the filter. In order to estimate the order N, calculate a parameter N1 using the
following equation. Choose N such that N ³ N1. Usually N is chosen as nearest integer just greater than N1.

LMF e j I
1
1/ A 2s − 1 2
OP
cosh −1
MMGH e J
1/ A p j − 1 K
2 PP
N1 = N Q .....(7.87)
cosh −1
FG IJ
Ωs
H K
Ωp
Chapter 7 - IIR Filters 7. 46
3. Determine the normalized transfer function H(sn), of the filter.
When the order N is even,
N
2
Bk
H ( sn ) = ∏ 2
k = 1 sn + b k sn + c k
.....(7.88)
When the order N is odd,
N−1
2
B0 Bk
H ( s) =
s + c0 ∏ s2 + bk s + ck
.....(7.89)
k=1

where, b k = 2 y N sin e ( 2 k − 1) π
2N
j .....(7.90)

c k = y 2N + cos2 e ( 2 k − 1)π
2N
j .....(7.91)

c0 = yN .....(7.92)
R|L 1 OP
1
N LMF 1 OP −
1
N
U|
yN =
1
2
S|MMFH 1
∈2
+1 IK 2
+
1
∈ PQ –
MNH
1
∈2
+1 IK 2
+
1
∈ PQ V| .....(7.93)
|TN |W
1
∈ = FH 1/ A 2p IK −1 2 .....(7.94)

For even values of N, find Bk such that,

1
H(0) = 1
.....(7.95)
(1 + ∈2 ) 2

For odd values of N, find Bk such that,

H(0) = 1 .....(7.96)
(It is normal practice to take B0 = B1 = B2 ..... = Bk).
4. Determine the unnormalized analog transfer function H(s) of the lowpass filter.

H ( s) = H ( sn ) s s
n=
Ωc

Here, W c = W p = Passband edge frequency.

When the order N is even, H(s) is obtained by letting sn ® s/W c in equation (7.88).
N N
2
Bk 2
Bk Ω2c
∴ H ( s) = ∏= s2n + b k sn + c k
= ∏= s + b k Ωc s + c k Ωc2
2 .....(7.97)
k 1 k 1
s
sn =
Ωc
7. 47 Digital Signal Processing
When the order N is odd, H(s) is obtained by letting sn ® s/W c in equation (7.89).
N−1
B0 2
B B0 Ωc Bk Ω2c
∴ H (s) =
s + c0 ∏= s2n + bk ksn + ck =
s + c0 Ωc s + b k Ωcs + c k Ω2c
2
.....(7.98)
k 1
s
sn =
Ωc

5. Determine the transfer function of digital filter, H(z). Using the chosen transformation,
in step-1transform H(s) to H(z). When impulse invariant transformation is employed, if T < 1, then
multiply H(z) by T to normalize the magnitude.

6. Realize the digital filter transfer function H(z) by a suitable structure.

7. Verify the design by sketching the frequency response H(ejw ).

H ( e jω ) = H ( z ) z = e jω

Note : The highpass, bandpass and bandstop filters are obtained from lowpass filter design by frequency
transformation.

7.8 Frequency Transformation

The four basic types of filters are lowpass, highpass, bandpass and bandstop filters.

The highpass or bandpass or bandstop filters are designed by designing a lowpass filter and then
using frequency transformation, the transfer function of the desired filter is obtained. The frequency
transformation can be carried in s-domain (analog) or in z-domain (digital).

7.8.1 Analog Frequency Transformation

Using analog frequency transformation the following filters can be designed from the normalized
lowpass filter. For normalized lowpass the cutoff frequency, W c = 1 rad/second.

1. Lowpass filter with cutoff frequency, W c.

2. Highpass filter with cutoff frequency, W c.

3. Bandpass filter with center frequency, W 0 and quality factor, Q.

4. Bandstop filter with center frequency, W 0 and quality factor, Q.

Ω0
where, Ω0 = Ωp Ωs and Q =
Ωs − Ωp

To design a filter, first design a normalized lowpass filter from the given specifications, and determine
the analog normalized transfer function (either Butterworth or Chebyshev transfer function) of the lowpass
filter. Then choose the transformation from the table 7.2 and determine the analog transfer function of the
desired filter.
Chapter 7 - IIR Filters 7. 48
Table 7.2 : Summary of Transformation for Analog Filter

Filter Type Transformation

s
Lowpass sn →
Ωc

Ωc
Highpass sn →
s

Q (s2 + Ω02 )
Bandpass sn →
Ω0s

Ω 0s
Bandstop sn →
Q (s2 + Ω02 )

From the analog transfer function H(s) the digital transfer function H(z) is obtained by either bilinear
transformation or impulse invariant transformation.
7.8.2 Digital Frequency Transformation
Table 7.3 : Summary of Transformation for Digital Filter

Filter Type Transformation Design Parameters

sin
FG ω ′ + ω IJ
c c
−1
z − α α =
H 2 K
Lowpass z −1 → F ω ′ − ω IJ
1 − αz −1 sinG c c
H 2 K
cos
FG ω ′ + ω IJ
c c

z −1 →
z −1 + α α =
H 2 K
Highpass F ω ′ − ω IJ
1 + αz −1 cosG c c
H 2 K
cosG
F ω + ω IJ
s p

α =
H 2 K = cos ω
z −2 −
2 αk − 1
k + 1
z +
k − 1
k + 1 cosG
F ω − ω IJ
s p
0

Bandpass
z −1
→ −
k − 1 −2 2 αk − 1 H 2 K
z − z + 1
k + 1 k + 1
k = cosG
F ω − ω I tan ω
s p
H 2 JK 2
c

cosG
F ω + ω IJ
s p

2αk −1 k − 1 α =
H 2 K = cos ω
−1
z −2 −
k + 1
z +
k + 1 cosG
F ω − ω IJ
s p
0

Bandstop
z →
k − 1 −2 2αk −1 H 2 K
z − z + 1
k + 1 k + 1
k = cosG
F ω − ω I tan ω ′
s p
H 2 JK 2
c
7. 49 Digital Signal Processing
Using digital frequency transformation the following filters can be designed from the lowpass digital
filter with cutoff frequency, w c' .
1. Lowpass filter with cutoff frequency, w c.
2. Highpass filter with cutoff frequency w c.
3. Bandpass filter with center frequency w 0 and lower and upper cutoff frequency w 1 and w 2.
4. Bandstop filter with center frequency w 0 and lower and upper cutoff frequency w 1 and w 2.
To design a filter, first design a lowpass digital filter from the given specifications, (either Butterworth
or Chebyshev) and determine H(z). Then choose the transformation from table 7.3 and determine the digital
transfer function of the desired filter.

Example 7.14
The normalized transfer function of an analog filter is given by,
1
H(sn ) =
sn2 + 1.4142sn + 1
Convert the analog filter to a digital filter with a cutoff frequency of 0.4p, using bilinear transformation.

Solution
To preserve the magnitude response the prewarping of analog filter has to be performed. For this
the analog cutoff frequency is determined using bilinear transformation and the analog transfer function is
unnormalized using this analog cutoff frequency. Then the analog transfer function is converted to digital
filter transfer function using bilinear transformation.
Given that, digital cutoff frequency, w c = 0.4p rad/sample. Let T = 1 second.
In Bilinear transformation,

2 0.4π 0.4π
Analog cutoff frequency, Ωc = tan = 2 tan = 1.4531 rad / second
T 2 2
Normalized analog transfer function,
1
H(sn ) =
sn2 + 1.4142sn + 1

The analog transfer function is unnormalized by replacing sn by s/W c

∴ Unnormalized U|
analog filter
|V H(s) = 1
|| FsI 2
F s I +1
transfer function W GH Ω JK
c
+1.4142 GH Ω JK
c

Ω 2c
=
s 2 +14142
. Ω cs + Ω 2c

1.45312
=
s + (1.4142 × 1.4531)s + 1.45312
2

2.1115
= 2
s + 2 .055s + 2 .1115

2 1– z –1
The H(z) is obtained by substituting, s = in H(s)
T 1+ z –1
Chapter 7 - IIR Filters 7. 50

∴ Digital filter |UV H(z) = 2.1115

transfer functionW| LM 2(1 − z ) OP LM 2(1 − z ) OP + 2.1115
−1 2 −1
+ 2.055
MN 1 + z ) PQ −1
MN 1 + z PQ
−1

2.1115
∴ H(z) =
4(1 − z−1)2 + 4.11(1 − z−1) (1+ z −1) + 2.1115(1+ z −1)2
(1 + z−1)2
2.1115(1 + z −1)2
= −1
4(1 − 2z + z ) + 4.11( 1 − z−2 ) + 2.1115(1+ 2z−1 + z−2 )
−2

2.1115(1+ 2z−1 + z−2 )

= −1
4(1 − 2z + z ) + 4.11( 1 − z−2 ) + 2.1115(1+ 2z−1 + z−2 )
−2

2.1115 + 4.223z−1 + 2.1115z−2

=
10.2215 − 3.777z−1 + 2.0015z−2

2.1115 4.223 −1 2.1115 −2

+ z + z
= 10 .2215 10.2215 10.2215
3.777 −1 2.0015 −2
1− z + z
10.2215 10.2215
0.2066 + 0.4131z −1 + 0.2066z−2
=
1 − 0.3695z−1 + 0.1958 z−2

Alternatively,
0.2066 + 0.4131z −1 + 0.2066z −2
H(z) =
1 − 0.3695z −1 + 0.1958 z −2
z −2 (0.2066 z2 + 0.4131z + 0.2066) 0.2066 z2 + 0.4131z + 0.2066
= −2 2
=
z (z − 0.3695z + 0.1958) z 2 − 0.3695z + 0.1958

Example 7.15
Design a Butterworth digital IIR lowpass filter using bilinear transformation by taking T = 0.1second, to
satisfy the following specifications.

0.6 £ |H(ejw )| £ 1.0 ; for 0 £ w £ 0.35p

jw
|H(e )| £ 0.1 ; for 0.7p £ w £ p

Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.

Alternatively,
Passband ripple £ 4.436 dB
Stopband attenuation ³ 20 dB
Passband edge frequency = 0.35p rad/sample
Stopband edge frequency = 0.7p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e −δ p, dB / 20 j = 10b −4.436 / 20g = 0.6

A s = 10
e − α s ,dB / 20j = 10b −20 / 20g = 0.1
7. 51 Digital Signal Processing
Solution
Specifications of digital IIR lowpass filter

Passband edge digital frequency, w p = 0.35p rad/sample

Stopband edge digital frequency, w s = 0.7p rad/sample

Gain in normal value at passband edge, Ap = 0.6

Gain in normal value at stopband edge, As = 0.1

Sampling time, T = 0.1second

Specifications of analog IIR lowpass filter

Gain in normal value at passband edge, Ap = 0.6

Gain is same in analog
Gain in normal value at stopband edge, As = 0.1 and digital filter.
For bilinear transformation,

2
ωp
Passband edge analog frequency, Ωp = T
tan
2
Using equation (7.53).
2 0.35π
= tan = 12.256 rad / second
0.1 2

2ωs
Stopband edge analog frequency, Ω s = T
tan Using equation (7.54).
2
2 0.7π
= tan
0.1 2
= 39.2522 rad / second
Order of the filter

LM FH 1/ A 2 IK − 1O LM e 1/ 0.12 j − 1OP
log
MN FH
s
2 I − 1P
P log
N1 =
1 1/ Ap K Q=1 MN e 1/ 0.6 2 j − 1PQ
Using equation (7.57).
2 Ω
log s 2 log 39.2522
Ωp 12.256

log
99 LM OP
=
1 1. 7778 N
= 17267
. Q
2 log 39.2522
12 .256

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 2.
Normalized transfer function, H(sn) of Butterworth lowpass filter

For even N,
N
2
1
b g ∏s
H sn = 2
n + bk sn + 1
Using equation (7.58).
k =1

where, bk = 2 sin LM b g
2k −1 π OP Using equation (7.60).
N 2N Q
Chapter 7 - IIR Filters 7. 52

N 2
Here, N = 2, ∴ k= 2
=
2
=1

1
b g
∴ H sn =
sn2 + b1 sn + 1
Calculate sin q using
When k = 1 ; bk = b1 = 2 sin LM b g OP = 1.4142
2−1 π
calculator in radian mode.
N Q2× 2

1
b g
∴ H sn =
sn2 + 1.4142 sn + 1

Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 39.2522 Using equation (7.61).

Ωc = 1
= 1
= 12.4439 rad / sec
1/ A s2 − 1 2N
e j c h
1/ 0.12 − 1 4

1
H(s) = H sn b g s
=
sn2 + 1.4142 sn + 1 s s
sn = n = Ω
Ωc c

1 1
∴ H(s) = = 2
s2 s s + 1.4142 Ω cs + Ω 2c
+ 1.4142 +1
Ω 2c Ωc Ω 2c
Ω 2c 12.44392
= = 2
s + 1.4142 Ω cs + Ω c s + 1.4142 × 12.4439 s + 12.44392
2 2

154.8506
=
s 2 + 17.5982 s + 154.8506

Digital IIR lowpass filter transfer function, H(z)

For bilinear transformation,

154.8506
H(z) = H(s) =
s 2 + 17.5982 s + 154.8506
2 1− z −1 2 1− z −1
s = s =
T 1+ z −1 T 1+ z −1
154.8506
=
F 2 1− z I −1 2 F 2 1 − z I + 154.8506
−1
GH T 1 + z JK −1 + 17.5982 GH T 1 + z JK −1

154.8506
=
−1 2
e j
4 1− z
+
e
35.1964 1 − z−1 j + 154.8506
−1 2
T e1 + z j
2
e
T 1+ z j
−1

154.8506
=
−1 2 2
e
4 1− z j e je j
+ 35.1964 T 1 − z−1 1 + z −1 + 154.8506 T 2 1 + z−1 e j
2
e
T 2 1 + z −1 j
7. 53 Digital Signal Processing

154.8506 T 2 (1 + z −1)2 Put, T = 0.1

∴ H(z) =
−1 2 −1 2
e
4 1− z j e je
+ 35.1964 T 1 − z−1 1 + z−1 + 154.8506 T 2 1 + zj e j
2 −1 −2
154.8506 × 0.1 (1 + 2z +z )
=
e
4 1 − 2z −1
+z −2
j + 35.1964 × 0.1e1− z j + 154.8506 × 0.1 e1 + 2z
−2 2 −1
+ z−2 j
1.5485(1 + 2z−1 + z−2 ) (a + b) (a – b) = a2 – b2
=
e
4 1 − 2z −1
+z −2
j + 3.5196 e1 − z j + 15485
. e1 + 2z
−2 −1
+z −2
j (a + b)2 = a 2 + 2ab + b2
(a − b)2 = a 2 − 2ab − b2
1.5485 + 3.097 z−1 + 1.5485 z−2
=
9.0681 − 4.903 z−1 + 2 .0289 z−2
.
15485 3.097 −1 15485
.
+ z + z−2 0.1708 + 0.3415z−1 + 0.1708 z−2
= 9 .0681 9.0681 9 .0681 =
4.903 −1 2.0289 −2 1 − 0.5407 z −1 + 0.2237 z−2
1− z + z
9.0681 9.0681

Alternatively,

0.1708 + 0.3415 z −1 + 0.1708 z−2

H(z) =
1 − 0.5407 z −1 + 0.2237 z−2

=
d
z−2 0.1708 z2 + 0.3415 z + 0.1708 i = 0.1708 z 2
+ 0.3415 z + 0.1708
z −2
dz 2
− 0.5407 z + 0.2237 i 2
z − 0.5407 z + 0.2237

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.1708 + 0.3415 z −1 + 0.1708 z −2

Let , H(z) = =
X(z) 1 − 0.5407 z −1 + 0.2237 z −2

On cross multiplying the above equation we get,

Y(z) – 0.5407z–1Y(z) + 0.2237z–2Y(z) = 0.1708 X(z) + 0.3415z–1X(z) + 0.1708z–2X(z)
\ Y(z) = 0.1708X(z) + 0.3415z–1X(z) + 0.1708z–2X(z) + 0.5407z–1Y(z) – 0.2237z–2Y(z) .....(1)
Using equation (1), the direct form-I structure is drawn as shown in fig 1.

X (z ) 0 .1708 X ( z ) Y (z )
0.17 08 + +
−1 −1
z z

z −1X (z) 0 .3415 z −1 X ( z) 0.5407 z −1 Y (z)

−1
z Y (z)
0.34 15 + + 0.54 07

−1 −1
z z
−2
z −2 X (z) 0.1708 z X ( z) −0 .2237 z −2 Y (z ) −2
z Y (z)
0.17 08 −0.2237

F ig 1 : D irect fo rm -I stru ctu re o f 2 n d o rder dig ita l IIR lo w p a ss filter.

Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.1708 + 0.3415 z −1 + 0.1708 z −2

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 0.5407 z −1 + 0.2237 z −2
Chapter 7 - IIR Filters 7. 54
W(z) 1
where, = .....(2)
X(z) 1 − 0.5407 z−1 + 0.2237 z −2
Y(z)
= 0.1708 + 0.3415 z−1 + 0.1708 z −2 .....(3)
W(z)
On cross multiplying equation (2) we get,
W(z) – 0.5407z–1W(z) + 0.2237z–2 W(z) = X(z)
\ W(z) = X(z) + 0.5407z–1W(z) – 0.2237z–2 W(z) .....(4)
On cross multiplying equation (3) we get,
Y(z) = 0.1708W(z) + 0.3415z–1W(z) + 0.1708z–2W(z) .....(5)
Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.

X (z) W (z) 0.1708 W ( z) Y (z )

+ 0.17 08
+

−1
z
0 .5407 z −1W ( z ) 0.3415 z −1W ( z)
z −1W ( z )
+ 0.54 07 0.34 15 +
−1
z
−0.2237 z −2 W ( z ) 0 .1708 z −2 W ( z )
z −2 W ( z)
−0.2237 0.17 08

F ig 2 : D irec t fo rm -II struc ture o f 2 n d o rd er d ig ita l IIR lo w p ass filter.

Frequency Response, H(ejw )

0.1708 + 0.3415 z−1 + 0.1708 z −2

d i
H e jω = H(z)
z = e jω
=
1 − 0.5407 z −1 + 0.2237 z−2 z = e jω

0.1708 + 0.3415 e − jω + 0.1708 e− j2ω

=
1 − 0.5407 e − jω + 0.2237 e − j2ω

=
c h
0.1708 + 0.3415 cos ω − j sin ω + 0.1708 cos 2ω − j sin 2ω c h e− jθ = cos θ − j sin θ
c h
1 − 0.5407 cos ω − j sin ω + 0.2237 cos 2ω − j sin 2ω c h
c0.1708 + 0.3415 cos ω + 0.1708 cos 2ωh + j c−0.3415 sin ω − 0.1708 sin 2ωh
=
c1 − 0.5407 cos ω + 0.2237 cos 2ωh + j c0.5407 sin ω − 0.2237 sin 2ωh
H de i c0.1708 + 0.3415 cos ω + 0.1708 cos 2ω h + j c −0.3415 sin ω − 0.1708 sin 2ω h
N
jω

Let , Hde i =
jω
=
H de iD
jω
c1 − 0.5407 cos ω + 0.2237 cos 2ωh + j c0.5407 sin ω − 0.2237 sin 2ωh
where, HN(ejw ) = (0.1708 + 0.3415cosw + 0.1708 cos2w) + j(– 0.3415sinw - 0.1708sin2w)
HD(ejw ) = (1 – 0.5407cosw + 0.2237cos2w) + j(0.5407sinw - 0.2237sin2w)
The frequency response H(ejw ) and hence the magnitude response |H(ejw )| are calculated for various
values of w and listed in table 1. Using the values listed in table 1, the magnitude response of lowpass filter is
sketched as shown in fig 3.

Note : Verify the result with MATLAB program 7.1.

7. 55 Digital Signal Processing
Table 1: H(ejww ) and |H(ejww )| for various values of w .
w HN(ejww ) HD(ejww ) H(ejww ) |H(ejww )|
0× π
16
0.6831 + j0 0.683 + j0 1.0000 + j0 1.0000
1×π
16
0.6635 – j0.1320 0.6764 + j0.0199 0.9743 – j0.2238 0.9997
2×π
16
0.6071 – j0.2515 0.6586 + j0.0487 0.8887 – j0.4476 0.9951
3×π
16
0.5201 – j0.3475 0.6360 + j0.0937 0.7216 – j0.6527 0.9730
4× π
16
0.4123 – j0.4123 0.6177 + j0.1586 0.4654 – j0.7870 0.9143
5×π
16
0.2952 – j0.4417 0.6140 + j0.2429 0.1696 – j0.7865 0.8046
6× π
16
0.1807 – j0.4363 0.6349 + j0.3414 –0.0659 – j0.6518 0.6551
7× π
16
0.0796 – j0.4003 0.6878 + j0.4447 –0.1838 – j0.4632 0.4983
8× π
16
0 – j0.3415 0.7763 + j0.5407 –0.2063 – j0.2962 0.3610
9× π
16
–0.0536 – j0.2696 0.8988 + j0.6159 –0.1804 – j0.1763 0.2523
10 × π
16
–0.0807 – j0.1947 1.0487 + j0.6577 –0.1388 – j0.0986 0.1703
11×π
16
–0.0843 – j0.1261 1.2148 + j0.6562 –0.0971 – j0.0531 0.1099
12× π
16
–0.0707 – j0.0707 1.3823 + j0.6060 –0.0617 – j0.0241 0.0662
13× π
16
–0.0478 – j0.0319 1.5352 + j0.5071 –0.0343 – j0.0095 0.0355
14 × π
16
–0.0239 – j0.0099 1.6577 + j0.3651 –0.0150 – j0.0027 0.0152
15× π
16
–0.0063 – j0.0013 1.7370 + j0.1911 –0.0037 – j0.0003 0.0037
16 × π
16
0 + j0 1.7644 + j0 0 + j0 0

|H (e jω)|
1.0

0.9

0.8

0.707
0.7 ωc = 2 tan −1 GH Ω2 T JK
c

0.6
= 2 tan −1
FG 12.4439 × 0.1 IJ
0.5 H 2 K
1.1132
0.4 = 1.1132 = × π
π

0.3 = 0.35 π rad / sam ple

0.2

0.1

ω
0 π 2π 4π
3π 5 π 6π 7π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
ωc = 0.35 π ( π/2 ) ( π)

F ig 3 : F req u en c y resp on se o f 2 n d o rd er d ig ita l B utterw o rth IIR lo w p ass filter.

Chapter 7 - IIR Filters 7. 56
Example 7.16
Design a Butterworth digital IIR highpass filter using bilinear transformation by taking T = 0.1second, to
satisfy the following specifications.
0.6 £ |H(ejw )| £ 1.0 ; for 0.7p £ w £ p
|H(ejw )| £ 0.1 ; 0 £ w £ 0.35p
Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.
Alternatively,
Passband ripple £ 4.436 dB
Stopband attenuation ³ 20 dB
Passband edge frequency = 0.7p rad/sample
Stopband edge frequency = 0.35p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e −δ p,dB / 20 j = 10b −4.436 / 20g = 0.6

As = 10
e − α s ,dB / 20j = 10b −20 / 20g = 0.1

Solution
Specifications of digital IIR highpass filter
Passband edge digital frequency, w p = 0.7p rad/sample
Stopband edge digital frequency, w s = 0.35p rad/sample
Gain in normal value at passband edge, Ap = 0.6
Gain in normal value at stopband edge, As = 0.1
Sampling time, T = 0.1second
The highpass filter is designed via lowpass filter using frequency transformation technique. Hence the
given specifications of IIR highpass filter are converted to corresponding specification of IIR lowpass filter.
Specifications of digital IIR lowpass filter
The specification of lowpass filter is obtained by taking passband edge of highpass as stopband edge of
lowpass and stopband edge of highpass as passband edge of lowpass. The gain of passband and stopband
remain same.
\ Passband edge digital frequency, w p = 0.35p rad/sample
\ Stopband edge digital frequency, w s = 0.7p rad/sample
Gain in normal value at passband edge, Ap = 0.6
Gain in normal value at stopband edge, As = 0.1

Specifications of analog IIR lowpass filter

Gain in normal value at passband edge, Ap = 0.6 Gain is same in analog
Gain in normal value at stopband edge, As = 0.1 and digital filter.
7. 57 Digital Signal Processing
For bilinear transformation,
ωp
Passband edge analog frequency, Ωp = 2
tan Using equation (7.53).
T
2
2 0.35π
= tan = 12.256 rad / second
0.1 2

2 ωs
Stopband edge analog frequency, Ω s = T
tan Using equation (7.54).
2
2 0.7π
= tan
0.1 2
= 39.2522 rad / second
Order of the filter

LM e 1/ A 2 j − 1O LM e 1/ 0.12 j − 1O
IK − 1PP
s
log
MN FH 2
log
MN e 1/ 0.6 2
P
j − 1PQ
N1 =
1 1/ Ap
Q =
1 Using equation (7.57).
2 Ω
log s 2 log 39.2522
Ωp 12.256

log
LM 99 OP
=
1 N1. 7778 Q = 17267
.
2 log 39.2522
12.256

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 2.
Normalized transfer function, H(sn) of Butterworth lowpass filter

For even N,
N
2
1
b g ∏s
H sn = 2
n + bk sn + 1
Using equation (7.58).
k =1

where, bk = 2 sin LM b g
2k −1 π OP Using equation (7.60).
N 2N Q
N 2
Here, N = 2, ∴ k= 2
=
2
=1

1
∴ H sn =b g sn2 + b1 sn + 1

Calculate sin q using

When k = 1 ; bk = b1 = 2 sin LM b g OP = 1.4142
2−1 π
calculator in radian mode.
N Q
2× 2

1
∴ H sn =b g sn2 + 1.4142 sn + 1

Unnormalized transfer function, H(s) of Butterworth highpass filter

The highpass filter with cutoff frequency, W c can be obtained from normalized lowpass filter using the
transformation, sn ® Wc/s.
Chapter 7 - IIR Filters 7. 58

∴ H(s) = H sn b g Ωc
sn =
s

where, W c = Cutoff frequency.

Ωs 39.2522 Using equation (7.61).

Ωc = 1
= 1
= 12.4439 rad / second
e1/ A 2s j − 1 2N FH IK
1/ 0.12 −1 4

1
∴ H(s) = H sn b g Ω
=
sn2 + 1.4142 sn + 1 Ωc
sn = c sn =
s s
1 1 s2
= = = 2
Ω 2c Ωc Ω 2c .
+ 14142 Ω cs + s 2
s + 1.4142 Ω cs + Ω 2c
2 + 1.4142 +1 2
s s s
s2 s2
= 2 2
= 2
s + 1.4142 × 12 .4439 s + 12.4439 s + 17.5982 s + 154.8506
Digital IIR highpass filter transfer function, H(z)
For bilinear transformation,

s2
H(z) = H(s) = 2
s + 17.5982 s + 154.8506
2 1− z −1 2 1− z −1
s = s =
T 1+ z −1 T 1+ z −1

F 2 1− z I −1 2
GH T 1 + z JK −1
=
F 2 1 − z I + 17.5982 2 F 1 − z I + 154.8506
−1 2 −1
GH T 1 + z JK −1
T GH 1 + z JK
2
T −1

−1 2
4 e1 − z j
−1 2
T e1 + z j 2
=
−1 2
4 e1 − z j 35.1964 e1 − z j −1
+ + 154.8506
−1 2 T e1 + z j −1
T e1 + z j
2

−1 2
4 e1 − z j
−1 2
T e1 + z j 2
=
−1 2 −1 2
4 e1 − z j + 35.1964 Te1 − z je1 + z j + 154.8506 T e1 + z j −1 −1 2

−1 2
T e1 + z j 2

−1 2
4 e1 − z j (a + b) (a – b) = a – b 2 2

= 2 2 2
−1 2 (a + b) = a + 2ab + b −1 2
4 e1 − z j + 35.1964 Te1 − z je1 + z j + 154.8506 T e1 + z j −1 −1 2
2 2 2
(a − b) = a − 2ab − b
4 e1 − 2z + z j −1 −2
= Put, T = 0.1
4 e1 − 2z + z j + 35.1964 × 0.1e1 − z j + 154.8506 × 0.1 e1 + 2z + z j
−1 −2 −2 2 −1 −2
7. 59 Digital Signal Processing

=
e
4 1 − 2z−1 + z−2 j
e
4 1 − 2z −1
+z −2
j + 3.5196 e1 − z j + 1.5485e1 + 2z
−2 −1
+ z −2 j
4 − 8z−1 + 4z−2
=
9.0681 − 4.903 z−1 + 2.0289 z−2
4 8 4
− z−1 + z−2
= 9.0681 9.0681 9.0681
4.903 −1 2.0289 −2
1− z + z
9.0681 9.0681
0.4411 − 0.8822 z−1 + 0.4411z−2
=
1 − 0.5407 z−1 + 0.2237 z −2
Alternatively,
0.4411 − 0.8822 z −1 + 0.4411z −2
H(z) =
1 − 0.5407 z −1 + 0.2237 z−2

=
d
z −2 0.4411z2 − 0.8822 z + 0.4411 i
d
z−2 z2 − 0.5407 z + 0.2237 i
0.4411z2 − 0.8822 z + 0.4411
=
z2 − 0.5407 z + 0.2237

Direct form-I structure of digital IIR highpass filter

Y(z) 0.4411 − 0.8822 z −1 + 0.4411z −2

Let , H(z) = =
X(z) 1 − 0.5407 z −1 + 0.2237 z −2

On cross multiplying the above equation we get,

Y(z) – 0.5407z–1Y(z) + 0.2237z–2Y(z) = 0.4411X(z) – 0.8822z–1X(z) + 0.4411z–2X(z)
\ Y(z) = 0.4411X(z) – 0.8822z–1X(z) + 0.4411z–2X(z) + 0.5407z–1X(z) – 0.2237z–2X(z) .....(1)
Using equation (1), the direct form-I structure is drawn as shown in fig 1.

X (z ) 0 . 4411 X (z) Y (z )
0.44 11 + +

−1 −1
z z
−1 −1
z −1X (z) −0. 8822 z X (z ) 0. 5407 z Y (z) −1
z Y (z)
−0.8822
+ + 0.54 07

−1 −1
z z
−2 −2
0. 4411 z X (z ) −0. 2237 z Y (z) −2
z −2 X (z) z Y (z)
0.44 11 −0.2237

F ig 1 : D irec t fo rm -I stru ctu re o f 2 n d o rder dig ita l IIR hig h p ass filter.
Direct form-II structure of digital IIR highpass filter

Y(z) W(z) Y(z) 0.4411 − 0.8822 z −1 + 0.4411z −2

Let , H(z) = = × =
X(z) X(z) W( z ) 1 − 0.5407 z −1 + 0.2237 z −2
Chapter 7 - IIR Filters 7. 60
W(z) 1
where, = .....(2)
X(z) 1 − 0.5407 z−1 + 0.2237 z−2

Y(z)
= 0.4411 − 0.8822 z−1 + 0.4411z −2 .....(3)
W(z)
On cross multiplying equation (2) we get,

W(z) – 0.5407z–1W(z) + 0.2237z–2W(z) = X(z)

\ W(z) = X(z) + 0.5407z–1W(z) – 0.2237z–2W(z) .....(4)

On cross multiplying equation (3) we get,

Y(z) = 0.4411W(z) – 0.8822z–1W(z) + 0.4411z–2W(z) .....(5)

Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.

X (z) W (z) 0 . 4411W (z ) Y (z )

+ 0.44 11
+

−1
z
−1 −1
0 .5407 z W ( z ) −0. 8822 z W (z)
z −1W ( z )
+ 0.54 07 −0.8822
+
−1
z
−2 −2
−0.2237 z W ( z ) 0. 4411 z W (z )
z −2 W ( z)
−0.2237 0.44 11

F ig 2 : D irec t form -II stru cture of 2 n d o rder dig ita l IIR hig h pa ss filter.

Frequency Response, H(ejw )

0.4411 − 0.8822 z−1 + 0.4411z −2

d i
H e jω = H(z)
z = e jω
=
1 − 0.5407 z −1 + 0.2237 z−2 z = e jω
− jω − j2ω e − jθ = cos θ − j sin θ
0.4411 − 0.8822 e + 0.4411e
=
1 − 0.5407 e− jω + 0.2237 e− j2ω

=
c h
0.4411 − 0.8822 cos ω − j sin ω + 0.4411 cos 2ω − j sin 2ω c h
c c h h
1 − 0.5407 cos ω − j sin ω + 0.2237 cos 2ω − j sin 2ω

=
c h c h
0.4411 − 0.8822 cos ω + 0.4411cos 2ω + j 0.8822 sin ω − 0.4411sin 2ω
c1 − 0.5407 cos ω + 0.2237 cos 2ωh + j c0.5407 sin ω − 0.2237 sin 2ωh
H de i c0.4411 − 0.8822 cos ω + 0.4411cos 2ω h + j c0.8822 sin ω − 0.4411sin 2ω h
N
jω

Let , Hde i =
jω
=
H de i
D
jω
c1 − 0.5407 cos ω + 0.2237 cos 2ωh + j c0.5407 sin ω − 0.2237 sin 2ωh
where, HN(ejw ) = (0.4411 – 0.8822cosw + 0.4411cos2w) + j(0.8822sinw - 0.4411sin2w )

HD(ejw ) = (1 – 0.5407cosw + 0.2237cos2w) + j(0.5407sinw - 0.2237sin2w )

The frequency response H(ejw ) and hence the magnitude response |H(ejw )| are calculated for various
values of w and listed in table 1. Using the values listed in table 1, the magnitude response of highpass filter is
sketched as shown in fig 3.

Note : Verify the result with MATLAB program 7.2.

7. 61 Digital Signal Processing
Table 1: H(ejww ) and |H(ejww )| for various values of w
w HN(ejww ) HD(ejww ) H(ejww ) |H(ejww )|
0× π
16
0 + j0 0.683 + j0 0 + j0 0
1×π
16
–0.0166 + j0.0033 0.6764 + j0.0199 –0.0244 + j0.0056 0.0250
2×π
16
–0.0620 + j0.0257 0.6586 + j0.0487 –0.0908 + j0.0451 0.1016
3×π
16
–0.1236 + j0.0826 0.636 + j0.0937 –0.1715 + j0.1551 0.2312
4× π
16
–0.1827 + j0.1827 0.6177 + j0.1586 –0.2062 + j0.3487 0.4051
5×π
16
–0.2178 + j0.326 0.614 + j0.2429 –0.1251 + j0.5804 0.5938
6× π
16
–0.2084 + j0.5031 0.6349 + j0.3414 0.0759 + j0.7516 0.7554
7× π
16
–0.1385 + j0.6964 0.6878 + j0.4447 0.3196 + j0.8058 0.8669
8× π
16
0 + j0.8822 0.7763 + j0.5407 0.533 + j0.7652 0.9325
9× π
16
0.2057 + j1.0341 0.8988 + j0.6159 0.6922 + j0.6762 0.9677
10 × π
16
0.4668 + j1.1270 1.0487 + j0.6577 0.8032 + j0.5709 0.9854
11× π
16
0.7624 + j1.1470 1.2148 + j0.6562 0.8786 + j0.4647 0.9939
12× π
16
1.0649 + j1.0649 1.3823 + j0.6060 0.9295 + j0.3629 0.9978
13× π
16
1.3434 + j0.8976 1.5352 + j0.5071 0.9631 + j0.2665 0.9993
14 × π
16
1.5681 + j0.6495 1.6577 + j0.3651 0.9845 + j0.175 0.9999
15× π
16
1.7139 + j0.3409 1.737 + j0.1911 0.9962 + j0.0867 1.0000
16 × π
16
1.7644 + j0 1.7644 + j0 1.0000 + j0 1.0000

|H (e jω)|
1.0

0.9

0.8

0.707 Ωc T
0.7 ω c = 2 tan −1
2
0.6
= 2 tan −1
FG 12.4439 × 0 .1IJ
0.5 H 2 K
1.1132
0.4 = 1.1132 = × π
π

0.3 = 0.35 π rad / sam ple

0.2

0.1

ω
0 π 2π 4π
3π 5π 6π 7π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
ωc = 0.35 π ( π/2 ) ( π)

F ig 3 : F req u en c y resp on se o f 2 n d o rd er d ig ita l B utterw o rth IIR hig h pa ss filte r.

Chapter 7 - IIR Filters 7. 62
Example 7.17
Design a Butterworth digital IIR lowpass filter using bilinear transformation by taking T = 0.5second, to
satisfy the following specifications.

0.707 £ |H(ejw )| £ 1.0 ; for 0 £ w £ 0.45p

|H(ejw )| £ 0.2 ; for 0.65p £ w £ p

Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.

Alternatively,
Passband ripple £ 3.01 dB
Stopband attenuation ³ 13.97 dB
Passband edge frequency = 0.45p rad/sample
Stopband edge frequency = 0.65p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e −δ p,dB / 20j = 10b −3.01/ 20g = 0. 707

As = 10
e −α s,dB / 20j = 10b −13.97 / 20g = 0.2
Solution
Specifications of digital IIR lowpass filter

Passband edge digital frequency, w p = 0.45p rad/sample

Stopband edge digital frequency, w s = 0.65p rad/sample

Gain in normal value at passband edge, Ap = 0.707

Gain in normal value at stopband edge, As = 0.2

Sampling time, T = 0.5second

Specifications of analog IIR lowpass filter

Gain in normal value at passband edge, Ap = 0.707 Gain is same in analog

and digital filter.
Gain in normal value at stopband edge, As = 0.2

For bilinear transformation,

2
ωp
Passband edge analog frequency, Ωp = T
tan Using equation (7.53).
2
2 0.45π
= tan
0.5 2
= 3.4163 rad / second

2 ωs
Stopband edge analog frequency, Ωs = T
tan Using equation (7.54).
2
2 0.65π
= tan
0.5 2
= 6.5274 rad / second
7. 63 Digital Signal Processing
Order of the filter

LM FH 1/ A 2
s
IK − 1O
P LM FH 1/ 0.22 IK − 1 O
P
log log
1 MN FH 1/ A 2
s
IK − 1P
Q=1 MN FH1/ 0.707 2 IK − 1P
Q Using equation (7.57).
N1 =
2 Ω 2 6.5274
log s log
Ωp 3.4163

log
24 LM OP
=
1 .
10006 N
= 2.4538 Q
2 log 6.5274
3.4163

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 3.
Normalized transfer function, H(sn) of Butterworth lowpass filter

For odd N,
N−1
2
1 1
b g
H sn =
sn + 1 ∏ sn2 + bk sn + 1
Using equation (7.59).
k =1

where, bk = 2 sin LM b g
2k −1 π OP Using equation (7.60).
N 2N Q
N − 1 3 − 1
Here, N = 3, ∴ k= 2
=
2
=1

1
∴ H sn =b g (sn + 1) (sn2 + b1 sn + 1)

Calculate sin q using

When k = 1 ; bk = b1 = 2 sin LM b g OP = 1
2−1 π
calculator in radian mode.
N Q
2× 3

1
∴ H sn =b g (sn + 1) (sn2 + sn + 1)
1 1
= =
sn3 + sn2 + sn + sn2 + sn + 1 sn3 + 2 sn2 + 2 sn + 1

Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 6.5274 Using equation (7.61).

Ωc = 1
= 1
= 3.8433 rad / second
e 1/ A s2 j −1 2N e1/ 0.22 j − 1 2×3

1
∴ H(s) = H sn b g s
=
sn3 + 2 sn2 + 2 sn + 1 s s
sn = n =
Ωc Ωc
Chapter 7 - IIR Filters 7. 64
1 1
∴ H(s) = =
FG s IJ 3
F s IJ
+ 2G
2
+2
s
+1
s3 + 2 Ω c s 2 + 2 Ω 2c s + Ω3c
HΩ K c HΩ K c Ωc Ω3c

Ω3c
=
s + 2 Ωc s + 2 Ω 2c s + Ω3c
3 2

3.84333
=
s + 2 × 3.8433 s + 2 × 3.84332 s + 3.84333
3 2

56.7692
=
s3 + 7.6866 s 2 + 29.5419s + 56 .7692
Digital IIR lowpass filter transfer function, H(z)
For bilinear transformation,

56.7692
H(z) = H(s) =
s3 + 7.6866 s 2 + 29.5419 s + 56.7692
2 1− z −1 2 1− z −1
s = s =
T 1+ z −1 T 1+ z −1
56.7692
=
F 2 1− z I −1 3 F 2 1− z I −1 2 F 2 1 − z I + 56.7692
−1
GH T 1 + z JK −1
+ 7.6866 G
H T 1 + z JK −1
+ 29.5419 GH T 1 + z JK
−1

56.7692
=
8(1 − z−1)3 + 7.6866 × 4T (1 − z−1)2 (1 + z−1) + 29.5419 × 2T 2 (1 − z−1)(1 + z−1)2
Put,
+ 56.7692 × T3 (1 + z−1)3
3 −1 3 T = 0.5
T (1 + z )
56.7692 × 0.53 (1 + z−1)3
=
8 (1 − z−1)3 + 7.6866 × 4 × 0.5 (1 − z−1)2 (1 + z−1) + 29.5419 × 2 × 0.52 (1 − z−1)(1 + z−1)2

+ 56.7692 × 0.53 (1 + z −1)3

7.0962 (1 + z−1)3 (a + b) (a – b) = a2 – b2
=
8 (1 − z−1)3 + 15.3732 (1 − z−1)2 (1 + z−1) (a + b)3 = a3 + 3a 2b + 3ab 2 + b3
+ 14.771(1 − z−1)(1 + z−1)2 + 7.0962 (1 + z−1)3 (a − b)3 = a3 − 3a 2b + 3ab2 − b3

7.0962 (1 + 3 z−1 + 3 z −2 + z−3 )

= −1 −2
8 (1 − 3 z + 3z − z ) + 15.3732 (1 − z−1)(1 − z−2 ) + 14.771(1 − z−2 )(1 + z−1)
−3

+ 7.0962 (1 + 3 z−1 + 3 z−2 + z −3 )

7.0962 + 212886
. z−1 + 21.2886 z−2 + 7.0962z−3
= −1 −2
8 (1 − 3 z + 3z − z ) + 15.3732 (1 − z−1 − z−2 + z−3 ) + 14.771(1 + z−1 − z−2 − z−3 )
−3

+ 7.0962 (1 + 3 z−1 + 3 z−2 + z−3 )

7.0962 + 212886
. z−1 + 21.2886 z−2 + 7.0962z−3
=
45.2404 − 3.3136 z−1 + 15.1444 z−2 − 0.3016 z −3
7.0962 212886
. 212886
. 7.0962 −3
+ z −1 + z −2 + z
= 45 .2404 45. 2404 45. 2404 45.2404
3.3136 −1 15.1444 −2 0.3016 −3
1− z + z − z
45.2404 45.2404 45.2404
0.1569 + 0.4706 z −1 + 0.4706 z−2 + 0.1569 z−3
=
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z−3
7. 65 Digital Signal Processing

Alternatively,

0.1569 + 0.4706 z −1 + 0.4706 z−2 + 0.1569 z −3

H(z) =
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z−3

=
d
z−3 0.1569 z3 + 0.4706 z2 + 0.4706 z + 0.1569 i
z −3
dz 3 2
− 0.0732 z + 0.3348 z − 0.0067 i
3 2
0.1569 z + 0.4706 z + 0.4706 z + 0.1569
=
z3 − 0.0732 z2 + 0.3348 z − 0.0067

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.1569 + 0.4706 z −1 + 0.4706 z −2 + 0.1569 z −3

Let , H(z) = =
X(z) 1 − 0.0732 z −1 + 0.3348 z −2 − 0.0067 z −3

On cross multiplying the above equation we get,

Y(z) – 0.0732z–1Y(z) + 0.3348z–2Y(z) – 0.00677 z–3Y(z) = 0.1569X(z)
+ 0.4706z–1X(z) + 0.4706z–2X(z) + 0.1569z–3X(z)
\ Y(z) = 0.1569X(z) + 0.4706z–1X(z) + 0.4706z–2X(z) + 0.1569z–3X(z) + 0.0732z–1Y(z)
– 0.3348z–2Y(z) + 0.0067z–3Y(z) .....(1)
Using equation (1), the direct form-I structure is drawn as shown in fig 1.

0.1569X (z)
X (z) 0.15 69 + + Y (z)

−1 −1
z z
−1 −1
−1 0.4706z X (z) 0.0732z Y (z) −1
z X (z) z Y (z)
0.4706 + + 0.07 32

−1 −1
z z
−2 0.4706z −2 X (z ) −0.3348z −2 Y (z ) −2
z X (z) z Y (z )
0.47 06 + + −0.3348

−1 −1
z z
−3 0.1569z −3 X (z ) 0.0067z −3 Y (z )
z X (z ) −3
0.1569 0.00 67 z Y (z )

F ig 1 : D irec t fo rm -I stru ctu re o f 3 rd o rd e r d ig ita l IIR lo w pa ss filte r..

Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.1569 + 0.4706 z −1 + 0.4706 z −2 + 0.1569 z −3

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 0.0732 z −1 + 0.3348 z −2 − 0.0067 z −3
W(z) 1
where, = .....(2)
X(z) 1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z−3
Y(z)
= 0.1569 + 0.4706 z−1 + 0.4706 z−2 + 0.1569 z−3 .....(3)
W(z)
On cross multiplying equation (2) we get,
W(z) – 0.0732z–1W(z) + 0.3348z–2 W(z) – 0.0067z–3W(z) = X(z)
\ W(z) = X(z) + 0.0732z–1W(z) – 0.3348z–2 W(z) + 0.0067z–3W(z) .....(4)
Chapter 7 - IIR Filters 7. 66
On cross multiplying equation (3) we get,
Y(z) = 0.1569W(z) + 0.4706z–1W(z) + 0.4706z–2W(z) + 0.1569z–3W(z) .....(5)
Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.

W (z) 0.1569W (z )
X (z ) + 0.15 69
+ Y (z )

−1
z
−1 −1
0.0732z W (z ) z W (z ) 0.4706z −1W (z )
+ 0.07 32 0.4706
+

−1
z
−0.3348 z −2 W (z) −2 0.4706z −2 W (z)
z W (z)
+ −0.3348 0.4706
+

−1
z
0.0067z −3 W (z) −3 0.1569z −3 W (z )
z W (z )
0.00 67 0.1569

F ig 2 : D irec t fo rm -II stru cture o f 3 rd o rd er d ig ita l IIR lo w pa ss filter.

Frequency Response, H(ejw )

0.1569 + 0.4706 z−1 + 0.4706 z−2 + 0.1569 z −3

e j
H e jω = H(z)
z = e jω
=
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z −3 z = e jω

0.1569 + 0.4706 e− jω + 0.4706 e− j2ω + 0.1569 e− j3ω

=
1 − 0.0732 e− jω + 0.3348 e− j2ω − 0.0067 e− j3ω

=
c h c
0.1569 + 0.4706 cos ω − j sin ω + 0.4706 cos 2ω − j sin 2ω + 0.1569 cos 3ω − j sin 3ω h c h
c h c
1 − 0.0732 cos ω − j sin ω + 0.3348 cos 2ω − j sin 2ω − 0.0067 cos 3ω − j sin 3ω h c h
(0.1569 + 0.4706 cos ω + 0.4706 cos 2ω + 0.1569 cos 3ω )
+ j ( −0.4706 sin ω − 0.4706 sin 2ω − 0.1569 sin 3ω )
=
c
1 − 0.0732 cos ω + 0.3348 cos 2ω − 0.0067 cos 3ω h
c
+ j 0.0732 sin ω − 0.3348 sin 2ω + 0.0067 sin 3ω h
(0.1569 + 0.4706 cos ω + 0.4706 cos 2ω + 0.1569 cos 3ω )

Let , Hee j =
jω e j=
HN e jω + j ( −0.4706 sin ω − 0.4706 sin 2ω − 0.1569 sin 3ω )
H ee j
D
jω (1 − 0.0732 cos ω + 0.3348 cos 2ω − 0.0067 cos 3ω )
+ j (0.0732 sin ω − 0.3348 sin 2ω + 0.0067 sin 3ω )

where, HN(ejw ) = (0.1569 + 0.4706cosw + 0.4706 cos2w + 0.1569 cos3w)

+ j(– 0.4706sinw - 0.4706sin2w - 0.1569sin3w)
jw
HD(e ) = (1 – 0.0732cosw + 0.3348cos2w – 0.0067cos3w)
+ j(0.0732sinw - 0.3348sin2w + 0.0067sin3w)
The frequency response H(ejw ) and hence the magnitude response |H(ejw )| are calculated for various
values of w and listed in table 1. Using the values listed in table 1, the magnitude response of lowpass filter is
sketched as shown in fig 3.

Note : Verify the result with MATLAB program 7.3.

7. 67 Digital Signal Processing
Table 1 : H(ejww ) and |H(ejww )| for various values of w
w HN(ejww ) HD(ejww ) H(ejww ) |H(ejww )|
0× π
16
1.255 + j0 1.2549 + j0 1 + j0 1.0000
1×π
16
1.1837 – j0.3591 1.232 – j0.1101 0.979 – j0.204 1.0000
2×π
16
0.9845 – j0.6578 1.1665 – j0.2025 0.9143 – j0.4052 1.0000
3×π
16
0.6977 – j0.8501 1.0686 – j0.2621 0.7999 – j0.5993 0.9995
4× π
16
0.3787 – j0.9143 0.953 – j0.2783 0.6243 – j0.7771 0.9968
5×π
16
0.0844 – j0.8567 0.8378 – j0.2471 0.3701 – j0.9134 0.9855
6× π
16
–0.1407 – j0.7075 0.7414 – j0.1717 0.0296 – j0.9474 0.9479
7× π
16
–0.2732 – j0.5112 0.6801 – j0.0619 –0.3306 – j0.7817 0.8488
8× π
16
–0.3137 – j0.3137 0.6652 + j0.0665 –0.5136 – j0.4202 0.6636
9× π
16
– 0.2825 – j0.1510 0.7012 + j0.1943 –0.4296 – j0.0963 0.4402
10 × π
16
– 0.211 – j0.042 0.7851 + j0.3018 –0.2521 + j0.0434 0.2558
11×π
16
– 0.1308 + j0.0129 0.906 + j0.3715 –0.1186 + j0.0629 0.1342
12× π
16
– 0.0649 + j0.0269 1.047 + j0.3913 –0.046 + j0.0429 0.0629
13× π
16
– 0.0237 + j0.0194 1.1877 + j0.3566 –0.0138 + j0.0205 0.0247
14 × π
16
– 0.0052 – j0.0077 1.3069 + j0.2709 –0.0026 + j0.0064 0.0070
15× π
16
– 0.0003 + j0.0011 1.3867 + j0.1461 –0.0001 + j0.0008 0.0007
16 × π
16
0 1.4147 + j0 0 + j0 0
jω
|H (e )|
1.0

0.9

0.8

0.707
0.7
Ωc T
0.6 ω c = 2 tan −1
2

0.5
= 2 tan −1
FG 3.8433 × 0.5 IJ
H 2 K
0.4
1.5306
= 1.5306 = × π
π
0.3
= 0.49 π rad / sam ple
0.2

0.1

ω
0 π 2π 4π
3π 5π 6π 7 π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
ωc= 0.49 π ( π/2 ) ( π)

F ig 3 : F req u ency resp o nse of 3 rd ord er d ig ita l B u tterw o rth IIR h ig hp a ss filter.
Chapter 7 - IIR Filters 7. 68
Example 7.18
Design a Butterworth digital IIR highpass filter using bilinear transformation by taking T = 0.5second, to
satisfy the following specifications.
0.707 £ |H(ejw )| £ 1.0 ; for 0.65p £ w £ p
jw
|H(e )| £ 0.2 ; for 0 £ w £ 0.45p
Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.
Alternatively,
Passband ripple £ 3.01dB
Stopband attenuation ³ 13.97dB
Passband edge frequency = 0.65p rad/sample
Stopband edge frequency = 0.45p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e −δ p, dB / 20 j = 10b −3.01/ 20 g = 0.707

A s = 10
e − α s ,dB / 20j = 10b −13.97 / 20 g = 0.2

Solution
Specifications of digital IIR highpass filter
Passband edge digital frequency, w p = 0.65p rad/sample
Stopband edge digital frequency, w s = 0.45p rad/sample
Gain in normal value at passband edge, Ap = 0.707
Gain in normal value at stopband edge, As = 0.2
Sampling time, T = 0.5 second.
The highpass filter is designed via lowpass filter using frequency transformation technique. Hence the
given specifications of IIR highpass filter are converted to corresponding specification of IIR lowpass filter.
Specifications of digital IIR lowpass filter
The specification of lowpass filter is obtained by taking passband edge of highpass as stopband edge of
lowpass and stopband edge of highpass as passband edge of lowpass. The gain of passband and stopband
remain same.
\ Passband edge digital frequency, w p = 0.45p rad/sample
Stopband edge digital frequency, w s = 0.65p rad/sample
Gain in normal value at passband edge, Ap = 0.707
Gain in normal value at stopband edge, As = 0.2
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.707 Gain is same in analog
and digital filter.
Gain in normal value at stopband edge, As = 0.2
For bilinear transformation,
2 ωp Using equation (7.53).
Passband edge analog frequency, Ωp = tan
T 2
2 0.45π
= tan = 3.4163 rad / second
0.5 2
7. 69 Digital Signal Processing
2 ωs
Stopband edge analog frequency, Ω s = T
tan Using equation (7.54).
2
2 0.65π
= tan = 6.5274 rad / second
0.5 2
Order of the filter

LM c h − 1OP
1/ A 2
s Lc 1/ 0.22 h − 1 OP
1
log
MN d i − 1PQ 1 log MMN c
2
1/ Ap 1/ 0.707 2 h − 1PQ Using equation (7.57).
N1 = =
2 Ωs 2 6.5274
log log
Ωp 3.4163

log
LM 24 OP
=
1 N1. 0006 Q = 2.4538
2 6.5274
log
3.4163

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 3.
Normalized transfer function, H(sn) of Butterworth lowpass filter

For odd N,
N− 1
2
1 1
b g
H sn =
sn + 1 ∏ sn2 + bk sn + 1
Using equation (7.59).
k =1

where, bk = 2 sin LM b g
2k −1 π OP Using equation (7.60).
N 2N Q
N − 1 3 − 1
Here, N = 3, ∴ k= 2
=
2
=1

b g bs + 1g ds 1+ b s + 1i
∴ H sn = 2
n n 1 n

When k = 1 ; b = b = 2 sin LM b g OP = 1
2−1 π Calculate sin q using
k
N Q 1 2× 3 calculator in radian mode.

1 1 1
∴ Hbs g = n = =
b g e + s + 1j s
s + 1 s n
2
n n
3
n + sn2 + sn + sn2 + sn + 1 sn3 + 2 sn2 + 2 sn + 1

Unnormalized transfer function, H(s) of Butterworth highpass filter

The highpass filter with cutoff frequency, W c can be obtained from normalized lowpass filter using the
transformation, sn ® Wc/s.

∴ H(s) = H sn b g Ωc
sn =
s

where, W c = Cutoff frequency.

Ωs 6.5274 Using equation (7.61).

Ωc = 1
= 1
= 3.8433 rad / second
e 1/ A 2j
s −1
2N c
1/ 0.22 h− 12 × 3
Chapter 7 - IIR Filters 7. 70

1
∴ H(s) = H sn b g Ω
=
sn3 + 2sn2 + 2sn + 1 Ωc
sn = c sn =
s s
1 1
= =
FG Ω IJ
c
3
+2
FG Ω IJ c
2
+2
Ωc
+1
Ω3c + 2 Ω 2cs + 2 Ω c s 2 + s3
HsK HsK s s3
s3 s3
= = 3
s + 2 Ω cs + 2 Ωc s + Ω c s + 2 × 3.8433 s + 2 × 3.84332 s + 3.84333
3 2 2 3 2

s3
= 3 2
s + 7.6866s + 29.5419s + 56.7692
Digital IIR highpass filter transfer function, H(z)
For bilinear transformation,

s3
H(z) = H(s) = 3 2
s + 7.6866s + 29.5419s + 56.7692
2 1− z −1 2 1− z −1
s = s =
T 1+ z −1 T 1+ z −1

F 2 1− z I −1 3
GH T 1 + z JK −1
=
F 2 1− z −1 3 I + 7.6866 F 2 1 − z I + 29.5419 F 2 1 − z I + 56.7692
−1 2 −1
Put, T = 0.5
GH T 1 + z −1 JK GH T 1 + z JK GH T 1 + z JK
−1 −1

8 (1 − z−1)3
T3(1 + z−1)3 (a + b) (a – b) = a2 – b2
=
8 (1 − z −1)3 + 7.6866 × 4T(1 − z−1)2 (1+ z−1) (a + b)3 = a3 + 3a 2b + 3ab2 + b3
2 −1 −1 2 3 −1 3
+29.5419 × 2T (1 − z )(1 + z ) + 56.7692 × T (1 + z ) (a − b)3 = a3 − 3a 2b + 3ab2 − b3
T3(1 + z−1)3
8 (1 − z−1)3
=
8 (1 − z ) + 7.6866 × 4 × 0.5 (1 − z−1)2 (1+ z −1)
−1 3

+ 29.5419 × 2 × 0.52(1 − z−1)(1 + z−1)2 + 56.7692 × 0.53(1 + z−1)3

8 (1 − z−1)3
=
8 (1 − z−1)3 + 15.3732 (1 − z−1)2 (1+ z−1) + 14.771(1 − z−1) (1+ z −1)2 + 7.0962 (1 + z−1)3
8 (1 − 3z −1 + 3z−2 − z −3 )
= −1 −2
8(1 − 3z + 3z − z ) + 15.3732 (1 − z−1)(1 − z−2 ) + 14.771(1 − z−2 )(1 + z−1)
−3

+ 7.0962 (1 + 3z−1 + 3z−2 + z−3 )

8 − 24z−1 + 24z−2 − 8z−3
=
e
8 1− 3z −1
+ 3z −2
−z −3
j + 15.3732(1 − z −1
− z−2 + z−3 ) + 14.771(1 + z−1 − z−2 − z−3 )
+ 7.0962 (1 + 3z−1 + 3z−2 + z−3 )
8 − 24z + 24z −2 − 8z −3 −1
=
45.2404 − 3.3136 z−1 + 15.1444 z−2 − 0.3016 z −3
8 24 24 8
− z −1 + z −2 − z−3
= 45.2404 45.2404 45.2404 45.2404
3.3136 −1 15.1444 −2 0.3016 −3
1− z + z − z
45.2404 45.2404 45.2404
0.1768 − 0.5305 z−1 + 0.5305 z −2 − 0.1768 z−3
=
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z−3
7. 71 Digital Signal Processing
Alternatively,

0.1768 − 0.5305 z−1 + 0.5305 z−2 − 0.1768 z−3

H(z) =
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z −3

=
e
z−3 0.1768 z3 − 0.5305 z2 + 0.5305 z − 0.1768 j
z −3
ez
3 2
− 0.0732 z + 0.3348 z − 0.0067 j
3 2
0.1768 z − 0.5305 z + 0.5305 z − 0.1768
=
z3 − 0.0732 z2 + 0.3348 z − 0.0067

Direct form-I structure of digital IIR highpass filter

Y(z) 0.1768 − 0.5305 z −1 + 0.5305 z −2 − 0.1768 z −3

Let , H(z) = =
X(z) 1 − 0.0732 z −1 + 0.3348 z −2 − 0.0067 z −3
On cross multiplying the above equation we get,
Y(z) – 0.0732z–1Y(z) + 0.3348z–2Y(z) – 0.0067z–3Y(z) = 0.1768X(z) – 0.5305z–1X(z)
+ 0.5305z–2X(z) – 0.1768z–3X(z)
\ Y(z) = 0.1768X(z) – 0.5305z X(z) + 0.5305z X(z) – 0.1768z–3X(z) + 0.0732z–1Y(z)
–1 –2

– 0.3348z–2Y(z) + 0.0067z–3Y(z) .....(1)

Using equation (1), the direct form-I structure is drawn as shown in fig 1.
0.1768X (z)
X (z) 0.17 68 + + Y (z)

−1 −1
z z
−1 −1
−1 −0.5305z X (z) 0.0732z Y (z ) −1
z X (z) z Y (z)
−0.5305 + + 0.07 32

−1 −1
z z
−2 −2
−2 0.5305z X (z ) −0.3348z Y (z ) −2
z X (z) z Y (z )
0.53 05 + + −0.3348

−1 −1
z −3 z
−3
−3 −0.1768z X (z) 0.0067z Y (z ) −3
z X (z ) −0.1768 0.00 67 z Y (z )

F ig 1 : D irec t fo rm -I stru ctu re o f 3 rd o rd e r d ig ita l IIR h ig h pa ss filte r..

Direct form-II structure of digital IIR highpass filter

Y(z) W(z) Y(z) 0.1768 − 0.5305 z −1 + 0.5305 z −2 − 0.1768 z −3

Let , H(z) = = × =
X(z) X(z) W( z ) 1 − 0.0732 z −1 + 0.3348 z −2 − 0.0067 z −3

W(z) 1
where, = .....(2)
X(z) 1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z−3

Y(z)
= 0.1768 − 0.5305 z−1 + 0.5305 z−2 − 0.1768 z−3 .....(3)
W(z)
On cross multiplying equation (2) we get,
W(z) – 0.0732z–1W(z) + 0.3348z–2W(z) – 0.0067z–3W(z) = X(z)
\ W(z) = X(z) + 0.0732z–1W(z) – 0.3348z–2W(z) + 0.0067z–3W(z) .....(4)
Chapter 7 - IIR Filters 7. 72
On cross multiplying equation (3) we get,
Y(z) = 0.1768W(z) – 0.5305z–1W(z) + 0.5305z–2W(z) – 0.1768z–3W(z) .....(5)
Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.
W (z) 0.1768W (z )
X (z ) + 0.17 68
+ Y (z )

−1
z
−1 −1 −1
0.0732z W (z ) z W (z ) −0.5305z W (z)
+ 0.07 32 −0.5305
+

−1
z
−2 −2
−0.3348z W (z ) −2 0.5305z W (z )
z W (z)
+ −0.3348 0.53 05 +

−1
z
−3
0.0067z W (z ) −3 −3
−0.1768z W (z)
z W (z )
0.00 67 −0.1768

F ig 2 : D irec t fo rm -II stru cture o f 3 rd o rder dig ita l IIR h ig h pa ss filter.

Frequency Response, H(ejw )

0.1768 − 0.5305 z−1 + 0.5305 z−2 − 0.1768 z−3

e j
H e jω = H(z)
z = e jω
=
1 − 0.0732 z−1 + 0.3348 z−2 − 0.0067 z −3 z = e jω

− jω − j2ω − j3ω
0.1768 − 0.5305 e + 0.5305 e − 0.1768 e
=
1 − 0.0732 e− jω + 0.3348 e− j2ω − 0.0067 e− j3ω

=
c h c
0.1768 − 0.5305 cos ω − j sin ω + 0.5305 cos 2ω − j sin 2ω − 0.1768 cos 3ω − j sin 3ω h c h
c h c
1 − 0.0732 cos ω − j sin ω + 0.3348 cos 2ω − j sin 2ω − 0.0067 cos 3ω − j sin 3ω h c h
c0.1768 − 0.5305 cos ω + 0.5305 cos 2ω − 0.1768 cos 3ωh
+ j c0.5305 sin ω − 0.5305 sin 2ω + 0.1768 sin 3ω h
=
c1 − 0.0732 cos ω + 0.3348 cos 2ω − 0.0067 cos 3ω h
+ j c0.0732 sin ω − 0.3348 sin 2ω + 0.0067 sin 3ω h

c0.1768 − 0.5305 cos ω + 0.5305 cos 2ω − 0.1768 cos 3ωh

H ee j N
jω
+ j c0.5305 sin ω − 0.5305 sin 2ω + 0.1768 sin 3ω h
Let , Hee j =
jω
==
H ee j D
jω
c1 − 0.0732 cos ω + 0.3348 cos 2ω − 0.0067 cos 3ωh
+ j c0.0732 sin ω − 0.3348 sin 2ω + 0.0067 sin 3ω h

where, HN(ejw ) = (0.1768 – 0.5305cosw + 0.5305cos2w - 0.1768cos3w)

+ j(0.5305sinw - 0.5305sin2w + 0.1768sin3w )

jw
HD(e ) = (1 – 0.0732cosw + 0.3348cos2w - 0.0067cos3w)
+ j(0.0732sinw - 0.3348sin2w + 0.0067sin3w)
The frequency response H(e ) and hence the magnitude response |H(ejw )| are calculated for various
jw

values of w and listed in table 1. Using the values listed in table 1, the magnitude response of highpass filter is
sketched as shown in fig 3.
Note : Verify the result with MATLAB program 7.4.
7. 73 Digital Signal Processing
Table 1: H(ejww ) and |H(ejww )| for various values of w
w HN(ejww ) HD(ejww ) H(ejww ) |H(ejww )|
0× π
16
0 + j0 1.2549 + j0 0 + j0 0
1×π
16
–0.0004 – j0.0013 1.232 – j0.1101 –0.0002 – j0.0011 0.0011
2×π
16
–0.0059 – j0.0088 1.1665 – j0.2025 –0.0036 – j0.0082 0.0089
3×π
16
–0.0268 – j0.022 1.0686 – j0.2621 –0.0189 – j0.0252 0.0315
4× π
16
–0.0733 – j0.0304 0.953 – j0.2783 –0.0623 – j0.0501 0.0799
5×π
16
–0.1475 – j0.0145 0.8378 – j0.2471 –0.1573 – j0.0637 0.1697
6× π
16
–0.238 + j0.0473 0.7414 + j0.1717 –0.3187 + j0.100 0.3189
7× π
16
–0.3186 + j0.1703 0.6801 + j0.0619 –0.4872 + j0.2061 0.5290
8× π
16
–0.3537 + j0.3537 0.6652 + j0.0665 –0.4738 + j0.5791 0.7482
9× π
16
–0.3080 + j0.5763 0.7012 + j0.1943 –0.1964 + j0.8763 0.8981
10 × π
16
–0.1586 + j0.7976 0.7851 + j0.3018 0.1642 + j0.9528 0.9668
11× π
16
0.0951 + j0.9657 0.906 + j0.3715 0.4640 + j0.8756 0.9910
12× π
16
0.4269 + j1.0306 1.047 + j0.3913 0.6806 + j0.7300 0.9980
13× π
16
0.7864 + j1.9583 1.1877 + j0.3566 0.8296 + j0.5578 0.9997
14 × π
16
1.1097 + j0.7415 1.3069 + j0.2709 0.9269 + j0.3752 1.0000
15× π
16
1.3342 + j0.4047 1.3867 + j0.1461 0.9820 + j0.1884 0.9999
16 × π
16
1.4146 + j0 1.4147 + j0 1.9999 + j0 0.9999

|H (e jω)|
1.0

0.9

0.8

0.707
0.7
Ωc T
0.6 ω c = 2 tan −1
2

0.5
= 2 tan −1
FG 3.8433 × 0.5 IJ
H 2 K
0.4
1.5306
= 1.5306 = × π
π
0.3
= 0.49 π rad / sam ple
0.2

0.1

ω
0 π 2π 4π
3π 5π 6π 7 π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
ωc= 0.49 π ( π/2 ) ( π)

F ig 3 : F req u ency resp o nse of 3 rd ord er d ig ita l B u tterw o rth IIR h ig hp a ss filter.
Chapter 7 - IIR Filters 7. 74
Example 7.19
Design a Butterworth digital IIR lowpass filter using impulse invariant transformation by taking
T = 1second, to satisfy the following specifications.
0.707 £ |H(ejw )| £ 1.0 ; for 0 £ w £ 0.3p
jw
|H(e )| £ 0.2 ; for 0.75p £ w £ p
Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.
Alternatively,
Passband ripple £ 3.01dB
Stopband attenuation ³ 13.97dB
Passband edge frequency = 0.3p rad/sample
Stopband edge frequency = 0.75p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e − δp1,dB / 20j = 10b −3.01/ 20 g = 0.707

A s = 10
e −α s , dB / 20j = 10b −13.97 / 20 g = 0.2

Solution
Specifications of digital IIR lowpass filter
Passband edge digital frequency, w p = 0.3p rad/sample
Stopband edge digital frequency, w s = 0.75p rad/sample
Gain in normal value at passband edge, Ap = 0.707
Gain in normal value at stopband edge, As = 0.2
Sampling time, T = 1second
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.707 Gain is same in analog
and digital filter.
Gain in normal value at stopband edge, As = 0.2
For impulse invariant transformation,
ωp 0.3π Using equation (7.55).
Passband edge analog frequency, Ωp = = = 0.9425π rad / second
T 1
ωs 0.75π
Stopband edge analog frequency, Ω s = = = 2.3562 rad / second Using equation (7.56).
T 1
Order of the filter

LM FH1/ A 2
s
IK − 1O
P LM FH 1/ 0.22 IK − 1 O
P LM 24 OP
log log
1 MN FH 2
1/ Ap IK − 1P
Q=1 MN FH
1/ 0.707 2 IK − 1P
Q 1
log
N1. 0006 Q = 17339
N1 = = . Using equation (7.57).
2 Ω 2 0.9425 2 log 0.9425
log s log
Ωp 2.3562 2.3562

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 2.
7. 75 Digital Signal Processing
Normalized transfer function, H(sn) of Butterworth lowpass filter

For even N,
N
2
1
b g ∏s
H sn = 2
n + bk sn + 1
Using equation (7.58).
k =1

where, bk = 2 sin LM b 2k −1 π g OP Using equation (7.60).

N 2N Q
N 2
Here, N = 2, ∴ k= 2
=
2
=1
1
b g
∴ H sn =
sn2 + b1 sn + 1
Calculate sin q using
When k = 1 ; bk = b1 = 2 sin LM b g OP = 1.4142
2−1 π calculator in radian mode.
N Q2× 2

1
b g
∴ H sn =
sn2 + 1.4142 sn + 1

Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 2.3562 Using equation (7.61).

Ωc = 1
= 1
.
= 10645 rad / second
e1/ As2 j − 1 2N c h
1/ 0.22 − 1 4

1
∴ H(s) = H sn b g s
=
sn2 + 1.4142 sn + 1 s s
sn = n = Ω
Ωc c
1 1 Ω2c
= 2
= 2 2
= 2
s s s + 1.4142 Ω cs + Ω c s + 1.4142 Ω cs + Ω2c
+ 1.4142 +1
Ω 2c Ωc Ωc2

2
.
10645 1.1332
= 2 2
= 2
s + 1.4142 × 10645
. s + 10645
. s + 1.5054 s + 1.1332

To convert the analog transfer function to digital transfer function, the above equation can be modified
as follows.
1.1332
∴ H(s) =
s 2 + 0.7527 × 2s + 0.75272 − 0.75272 + 1.1332
1.1332 (s + a)2 = s 2 + 2as + a 2
=
bs + 0.7527g 2
+ 0.5666
2a = 15054
. ⇒ a=
1.5054
= 0.7527
1.1332 0.7527 2
= ×
0.7527 2
b
s + 0.7527 + 0.75272 g
0.7527
= 1.5055 ×
bs + 0.7527g 2
+ 0.75272
Chapter 7 - IIR Filters 7. 76
Digital IIR lowpass filter transfer function, H(z)

In impulse invariant transformation,

b
  →
c
e − aT sin bT z−1 h Using equation (7.19).

bs + a g 2
+b 2 is transformed to
1− 2e − aT
ccos bTh z −1
+e −2aT
z −2

Using the above transformation, the H(s) can be transformed to H(z) as shown below.

∴ H(z) = 1.5055 ×
e−0.7527 ×1
csin 0.7527 × 1h z −1
Put, T = 1
1 − 2 e−0.7527 ×1
(cos 0.7527 × 1) z−1 + e−2 × 0.7527 × 1 −2
z
0.3220 z −1 0.4848 z−1
= 1.5055 × −1 −2
=
1 − 0.6877 z + 0.2219 z 1 − 0.6877 z−1 + 0.2219 z −2
Alternatively,

0.4848 z −1 0.4848 z −1
H(z) = −1 −2
= −2 2
1 − 0.6877 z + 0.2219 z z (z − 0.6877 z + 0.2219)
0.4848 z
=
z2 − 0.6877 z + 0.2219

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.4848 z −1
Let , H(z) = =
X(z) 1 − 0.6877 z −1 + 0.2219 z −2

On cross multiplying the above equation we get,

Y(z) – 0.6877z–1Y(z) + 0.2219z–2Y(z) = 0.4848z–1X(z)
\ Y(z) = 0.4848z–1X(z) + 0.6877z–1Y(z) – 0.2219z–2Y(z) .....(1)
Using equation (1), the direct form-I structure is drawn as shown in fig 1.

X (z) + Y (z)

−1 −1
z z
−1 −1 −1
z X (z ) 0.4848 z X (z) 0.6877z Y (z ) −1
z Y (z )
0.48 48
+ 0.68 77

−1
z
−2
−0.2219z Y (z ) z −2 Y (z )
−0.2219

F ig 1 : D irect fo rm -I stru ctu re o f 2 n d o rder dig ita l IIR lo w p ass filter.

Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.4848 z −1

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 0.6877 z −1 + 0.2219 z −2

W(z) 1
where, = .....(2)
X(z) 1 − 0.6877 z + 0.2219 z−2
−1

Y(z)
= 0.4848 z−1 .....(3)
W(z)
7. 77 Digital Signal Processing
On cross multiplying equation (2) we get,

W(z) – 0.6877z–1 + 0.2219z–2 W(z) = X(z)

\ W(z) = X(z) + 0.6877z–1W(z) – 0.2219z–2 W(z) .....(4)

On cross multiplying equation (3) we get,

Y(z) = 0.4848z–1W(z) .....(5)

Using equation (4) and (5), the direct form-II structure is drawn as shown in fig 2.

W (z)
X (z ) + Y (z )

−1
z
−1 −1
0.6877z W (z ) −1 0.4848z W (z)
z W (z)
+ 0.68 77 0.48 48

−1
z
−2
−0.2219z W (z) −2
z W (z )
−0.2219

F ig 2 : D irect fo rm -II structure of 2 n d order digita l IIR lo w p ass filter.

Frequency Response, H(ejw )

0.4848 z−1
e j
H e jω = H(z)
z = e jω
=
1 − 0.6877 z−1 + 0.2219 z−2 z = e jω

0.4848 e − jω
=
1 − 0.6877 e− jω + 0.2219 e− j2ω

=
c
0.4848 cos ω − j sin ω h
c h
1 − 0.6877 cos ω − j sin ω + 0.2219 cos 2ω − j sin 2ω c h
0.4848 cosω − j0.4848 sin ω
=
c1 − 0.6877 cos ω + 0.2219 cos 2ωh + j c0.6877 sin ω − 0.2219 sin 2ωh

e j
Let , H e jω =
e j=
HN ejω 0.4848 cosω − j0.4848 sin ω
e j
HD ejω c1 − 0.6877 cos ω + 0.2219 cos 2ω h + j c0.6877 sin ω − 0.2219 sin 2ω h

where, HN(ejw ) = 0.4848cosw – j0.4848sinw

HD(ejw ) = (1 – 0.6877cosw + 0.2219cos2w) + j(0.6877sinw - 0.2219sin2w)

The frequency response H(ejw ) and hence the magnitude response |H(ejw )| are calculated for various
values of w and listed in table 1. Using the values listed in table 1, the magnitude response of lowpass filter is
sketched as shown in fig 3.

Note : Verify the result with MATLAB program 7.5.

Chapter 7 - IIR Filters 7. 78
Table 1: H(ejww ) and |H(ejww )| for various values of w
w HN(ejww ) HD(ejww ) H(ejww ) H(ejww )|
0× π
16
0.4848 + j0 0.5342 + j0 0.9075 + j0 0.9075
1×π
16
0.4755 – j0.0946 0.5305 + j0.0492 0.8723 – j0.2592 0.9099
2×π
16
0.4479 – j0.1855 0.5215 + j0.1063 0.7550 – j0.5096 0.9101
3×π
16
0.4031 – j0.2693 0.5131 + j0.1770 0.5403 – j0.7112 0.8932
4× π
16
0.3428 – j0.3428 0.5137 + j0.2644 0.2560 – j0.7990 0.8390
5×π
16
0.2693 – j0.4031 0.5330 + j0.3668 –0.0103 – j0.7492 0.7493
6× π
16
0.1855 – j0.4479 0.5799 + j0.4784 –0.1888 – j0.6166 0.6449
7× π
16
0.0946 – j0.4755 0.6608 + j0.5896 –0.2778 – j0.4718 0.5475
8× π
16
0 – j0.4848 0.7781 + j0.6877 –0.3092 – j0.3498 0.4669
9× π
16
–0.0946 – j0.4755 0.9292 + j0.7594 –0.3118 – j0.2569 0.4040
10 × π
16
–0.1855 – j0.4479 1.1062 + j0.7923 –0.3025 – j0.1882 0.3563
11× π
16
–0.2693 – j0.4031 1.2971 + j0.7768 –0.2898 – j0.1372 0.3206
12× π
16
–0.3428 – j0.3428 1.4863 + j0.7082 –0.2775 – j0.0984 0.2944
13× π
16
–0.4031 – j0.2693 1.6567 + j0.5871 –0.2673 – j0.0678 0.2758
14 × π
16
–0.4479 – j0.1855 1.7923 + j0.4201 –0.2599 – j0.0426 0.2634
15× π
16
–0.4755 – j0.0946 1.8795 + j0.2191 –0.2554 – j0.0205 0.2562
16 × π
16
–0.4848 + j0 1.9096 + j0 –0.2539 + j0 0.2539
jω
|H (e )|
1.0

0.9

0.8
0.707
0.7
ωc = Ωc T = 1.0645 × 1
0.6
1.0645
= ×π
π
0.5
= 0.34 π rad / sam ple
0.4

0.3

0.2

0.1

ω
0 π 2π 4π 5π
3π 6π 7π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
Ωc = 0.3 4 π ( π/2 ) ( π)
F ig 3 : F req u en c y resp o n se o f 2 n d ord e r d igital B u tterw o rth IIR low p a ss filte r.
7. 79 Digital Signal Processing
Example 7.20
Design a Butterworth digital IIR lowpass filter using impulse invariant transformation by taking T = 1second,
to satisfy the following specifications.

0.9 £ |H(ejw )| £ 1.0 ; 0 £ w £ 0.35p

|H(ejw )| £ 0.275 ; 0.7p £ w £ p

Draw direct form-I and II structure of the filter. Verify the design by sketching the frequency response.

Alternatively,
Passband ripple £ 0.9151dB
Stopband attenuation ³ 11.2133dB
Passband edge frequency = 0.35p rad/sample
Stopband edge frequency = 0.7p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e −δ p,dB / 20j = 10b −0.9151/ 20g = 0.9

As = 10
e −α s ,dB / 20j = 10b −11.2133/ 20g = 0.275

Solution
Specifications of digital IIR lowpass filter

Passband edge digital frequency, w p = 0.35p rad/sample

Stopband edge digital frequency, w s = 0.7p rad/sample
Gain in normal value at passband edge, Ap = 0.9
Gain in normal value at stopband edge, As = 0.275
Sampling time, T = 1second
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.9 Gain is same in analog
and digital filter.
Gain in normal value at stopband edge, As = 0.275
For impulse invariant transformation,
ωp 0.35π
Passband edge analog frequency, Ωp = = = 10996
. rad / second Using equation (7.53).
T 1
ωs 0.7π
Stopband edge analog frequency, Ω s = = = 2.1991 rad / second Using equation (7.54).
T 1
Order of the filter

LM FH1/ A s2 IK − 1O
P LM FH
1/ 0.2752 IK − 1O
P
1
log
MN FH IK P
2 −1
1/ A p
Q=1
log
MN FH 1/ 0.92 IK −1 PQ 1 log LMN120.2346
.2231O
PQ = 2.8518 Using equation (7.57).
N= =
2 Ωs 2 2.1991 2 2.1991
log log log
Ωp .
10996 10996
.

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 3.
Chapter 7 - IIR Filters 7. 80
Normalized transfer function, H(sn) of Butterworth lowpass filter

For odd N,
N−1
2
1 1
b g
H sn =
sn + 1
∏ sn2 + bk sn + 1
Using equation (7.59).
k =1

where, bk = 2 sin LM b g
2k −1 π OP Using equation (7.60).
N 2N Q
N − 1 3 − 1
Here, N = 3, ∴ k= 2
=
2
=1

1 1
b g
∴ H sn =
sn + 1 sn2 + b1 sn + 1

When k = 1; bk = b1 = 2 sin LM b g OP = 1
2−1 π Calculate sin q using
N Q2× 3 calculator in radian mode.
1 1
b g
∴ H sn = =
(sn + 1) (sn2 + sn + 1) sn3 + 2sn2 + 2sn + 1

Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 2.1991
Ωc = 1
= 1
= 14489
. rad / sec
e1/ A 2
s j−1 2N F 1 I 6
H 0.2752
−1
K
1 Using equation (7.61).
∴ H(s) = H sn b g s
=
(sn + 1) (sn2 + sn + 1) s
sn = sn =
Ωc Ωc

1 1
=
F s + 1I F s 2
s I = Fs+Ω I Fs 2
+ s Ω c + Ω 2c I
GH Ω JK GH Ω
c
2
c
+
Ωc
+1 JK GH Ω JK GH c
c
Ω c2 JK
Ω3c 1.44893
= 2
=
( s + Ω c ) (s + s Ω c + Ω 2c ) 2
(s + 1.4489) (s + s × 14489
. + 14489
. 2
)
3.0417
= .....(1)
(s + 14489
. ) (s 2 + 1.4489 s + 2.0993)
3.0417
=
s3 + 2.8978s 2 + 4.1986s + 3.0417
To convert the analog transfer function to digital transfer function using impulse invariant transformation,
the equation (1) can be simplified as follows.
3.0417
H(s) =
(s + 14489
. ) (s2 + 1.4489 s + 2.0993)
By partial fraction expansion H(s) can be expressed as

3.0417 A Bs + C
= = + .....(2)
(s + 14489
. ) (s2 + 1.4489 s + 2.0993) s + 1.4489 s 2 + 14489
. s + 2.0993
7. 81 Digital Signal Processing
On cross multiplying the equation (2) we get
3.0417 = A(s2 + 1.4489s + 2.0993) + (Bs + C) (s + 1.4489)
3.0417 = As2 + 1.4489As + 2.0993A + Bs2 + 1.4489Bs + Cs + 1.4489C .....(3)

On equating the coefficients On equating the coefficients of s in On equating constants we get,

of s2 in equation (3) we get, equation (3) we get,
2.0993 A + 1.4489 C = 3.0417
A+B=0 1.4489 A + 1.4489 B + C = 0
Put, C = 0,
\ B= –A Put, B = – A
\ 2.0993 A = 3.0417
\ 1.4489 A + 1.4489(– A) + C = 0
3.0417
\C=0 ∴ A=
2.0993
= 1.4489
\ B = –A = –1.4489

A Bs + C
∴ H(s) = +
s + 1.4489 s 2 + 1.4489s + 2.0993 (s + a)2 = s 2 + 2as + a 2
1.4489 −1.4489s 14489
.
= 2a = 14489
. ⇒ a= = 0.7245
s + 14489
. s 2 + 14489
. s + 2.0993 2
1.4489 1.4489s
= −
s + 14489
. 2
(s + 2 × 0.7245s + 0.72452 ) + FH 2.0993 − 0.72452 IK 2

1.4489 s + 0.7245 − 0.7245

= − 14489
.
s + 14489
. (s + 0.7245)2 + 12548
. 2

1.4489 s + 0.7245 0.7245

= − 14489
. + 14489
.
s + 14489
. (s + 0.7245)2 + 12548
. 2
(s + 0.7245)2 + 1.25482
1.4489 s + 0.7245 14489
. × 0.7245 12548
.
= − 14489
. 2 2
+
s + 14489
. (s + 0.7245) + 12548
. 12548
. (s + 0.7245)2 + 12548
. 2

1.4489 s + 0.7245 .
12548
= .
− 14489 + 0.8366 ......(4)
.
s + 14489 (s + 0.7245)2 + 12548
. 2
(s + 0.7245)2 + 12548
. 2

Digital IIR lowpass filter transfer function, H(z)

In impulse invariant transformation,

Using equations (7.17),
1 Ai
 
is transformed to
→ −1
(7.18) and (7.19).
s + pi 1 − e−piT z
(s + a ) 1 − e− aT (cos bT)z−1
  →
(s + a)2 + b2 is transformed to − aT
1 − 2e (cos bT )z−1 + e−2aT z−2
b e− aT (sin bT )z−1
  →
(s + a)2 + b2 is transformed to
1 − 2e (cos bT )z−1 + e−2aT z−2
− aT

Using the above transformation, the H(s) of equation (4) can be transformed to H(z) as shown below.

14489
. 1 − e −0.7245 (cos 1.2548)z −1
∴ H(z) = −1.4489 −1
− 1.4489 −0.7245
Put, T = 1.
1− e z 1 − 2e (cos 1.2548)z −1 + e −2× 0.7245z −2

e −0.7245 (sin 1.2548)z −1

+ 0.8366 −0.7245
1 − 2e (cos 1.2548)z −1 + e −2× 0.7245z −2
Chapter 7 - IIR Filters 7. 82

1.4489 −1.4489 + 0.2182 z−1 0.3853 z−1

∴ H(z) = −1
+ −1 −2
+
1 − 0.2348 z 1 − 0.3012 z + 0.2348 z 1 − 0.3012 z−1 + 0.2348 z−2

1.4489 −1.4489 + 0.6035 z−1

= −1
+
1 − 0.2348 z 1 − 0.3012 z −1 + 0.2348 z−2

1.4489 (1 − 0.3012 z−1 + 0.2348 z −2 ) + (−1.4489 + 0.6035z −1) (1 − 0.2348 z−1)

=
(1 − 0.2348 z−1) (1 − 0.3012 z−1 + 0.2348 z−2 )

1.4489 − 0.4364 z−1 + 0.3402 z −2 − 1.4489 + 0.3402z−1 + 0.6035z−1 − 0.1417 z−2

=
1 − 0.3012 z−1 + 0.2348 z−2 − 0.2348 z−1 + 0.0707 z −2 − 0.0551z−3

0.5073 z−1 + 0.1985 z−2

=
1 − 0.536 z−1 + 0.3055 z −2 − 0.0551z−3
Alternatively,
0.5073 z−1 + 0.1985 z−2 z−3 (0.5073 z−2 + 0.1985 z)
H(z) = =
1 − 0.536 z−1 + 0.3055 z−2 − 0.0551z−3 z−3(1 − 0.536 z2 + 0.3055 z − 0.0551)
0.5073 z−2 + 0.1985 z
=
1 − 0.536 z2 + 0.3055 z − 0.0551

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.5073 z −1 + 0.1985 z −2

Let , H(z) = =
X(z) 1 − 0.536 z −1 + 0.3055 z −2 − 0.0551z −3
On cross multiplying the above equation we get,
Y(z) – 0.536z–1Y(z) + 0.3055z–2Y(z) – 0.0551z–3Y(z) = 0.5073z–1X(z) + 0.1985z–2X(z)
\ Y(z) = 0.5073z–1X(z) + 0.1985z–2X(z) + 0.536z–1X(z) – 0.3055z–2Y(z) + 0.0551z–3Y(z) .....(5)
Using equation (5), the direct form-I structure is drawn as shown in fig 1.

X (z) + Y (z)

−1 −1
z z
−1 −1
−1
z X (z) 0.5073z X (z) 0.536z Y (z) −1
0.50 73 + + 0 .5 3 6 z Y (z)

−1 −1
z z
−2 −2
−2 0.1985z X (z ) −0 .3055z Y (z) −2
z X (z) z Y (z )
0.19 85
+ −0.3055

−1
−3
z
0.0051z Y (z ) −3
0.00 51 z Y (z )

F ig 1 : D irec t fo rm -I stru ctu re o f 3 rd o rd e r d ig ita l IIR lo w pa ss filte r.

Direct form-II structure of digital IIR lowpass filter
Y(z) W(z) Y(z) 0.5073 z −1 + 0.1985z −2
Let , H(z) = = × =
X(z) X(z) W(z) 1 − 0.536 z −1 + 0.3055 z −2 − 0.0551z −3
W(z) 1
where, = .....(6)
X(z) 1 − 0.536 z + 0.3055 z−2 − 0.0551z−3
−1

Y(z)
= 0.5073 z−1 + 0.1985 z−2 .....(7)
W(z)
7. 83 Digital Signal Processing
On cross multiplying equation (6) we get,
W(z) – 0.536z–1W(z) + 0.3055z–2 W(z) –0.0551 z–3W(z) = X(z)
W(z) = X(z) + 0.536z–1W(z) – 0.3055z–2 W(z) + 0.0551z–3W(z) .....(8)
On cross multiplying equation (7) we get,
Y(z) = 0.5073z–1W(z) + 0.1985z–2W(z) .....(9)
Using equations (8) and (9), the direct form-II structure is drawn as shown in fig 2.

X (z) W (z) Y (z )
+

−1
z
0.536 z −1W ( z) 0.5073 z −1W ( z )
z −1W ( z)
+ 0 .5 3 6 0.50 73 +
−1
z
−0 .3055 z −2 W ( z )
z −2 W ( z ) 0 .1985 z −2 W ( z )
+ −0.3055 0.19 85

−1
z
−3
0 .0551 z W ( z)
z −3 W ( z)
0.05 51

F ig 2 : D irec t fo rm -II stru cture o f 3 rd o rd er d ig ita l IIR lo w pa ss filter.

Frequency Response, H(e jw )

0.5073 z −1 + 0.1985 z−2

e j
H e jω = H(z)
z = e jω
=
1 − 0.536 z−1 + 0.3055 z−2 − 0.0551z−3 z = e jω

0.5073e− jω + 0.1985 e− j2ω

=
1 − 0.536 e− jω + 0.3055 e− j2ω − 0.0551e− j3ω
e − jθ = cos θ − j sin θ
0.5073 (cos ω − j sin ω ) + 0.1985 (cos 2ω − j sin 2ω )
=
1 − 0.536 (cos ω − j sin ω ) + 0.3055 (cos 2ω − j sin 2ω ) − 0.0551(cos 3ω − j sin 3ω )
(0.5073 cos ω + 0.1985 cos ω ) + j( −0.5073 sin ω − 0.1985 sin 2ω )
=
(1 − 0.536 cos ω + 0.3055 cos 2ω − 0.0551cos 3ω ) + j(0.536 sin ω − 0.3055 sin 2ω + 0.0551sin 3ω )

e j
Let , H e jω =
e j = (0.5073 cos + 0.1985 cos ω) + j(−0.5073 sin ω − 0.1985 sin 2ω)
HN ejω
H ee j
D
jω (1 − 0.536 cos ω + 0.3055 cos 2ω − 0.0551cos 3ω )
+ j(0.536 sin ω − 0.3055 sin 2ω + 0.0551sin 3ω )

where, HN(ejw ) = (0.5073 cosw + 0.1985 cos2w) + j(– 0.5073sinw - 0.1985sin2w)

HD(ejw ) = (1 – 0.536cosw + 0.30554cos2w -0.0551 cos3w)
+ j(0.536sinw - 0.3055sin2w + 0.0551sin3w )
The frequency response H(e ) and hence the magnitude response |H(ejw )| are calculated for various
jw

values of w and listed in table 1. Using the values listed in table 1, the magnitude response of lowpass filter is
sketched as shown in fig 3.

Note : Verify the result with MATLAB program 7.6.

Chapter 7 - IIR Filters 7. 84
Table 1: H(ejww ) and |H(ejww )| for various values of w
w HN(ejww ) HD(ejww ) H(ejww ) |H(ejww )|
0× π
16
0.7058 + j0 0.7144 + j0 0.9880 + j0 0.9880
1×π
16
0.6809 – j0.1749 0.7107 + j0.0183 0.9511 – j0.2706 0.9888
2×π
16
0.6090 – j0.3345 0.6997 + j0.0400 0.8403 – j0.5261 0.9914
3×π
16
0.4978 – j0.4652 0.6820 + j0.0696 0.6535 – j0.7488 0.9939
4× π
16
0.3587 – j0.5572 0.6600 + j0.1125 0.3883 – j0.9104 0.9898
5×π
16
0.2059 – j0.6052 0.6393 + j0.1742 0.0597 – j0.9629 0.9648
6× π
16
0.0538 – j0.6090 0.6298 + j0.2581 –0.2662– j0.8579 0.8982
7× π
16
–0.0844 – j0.5735 0.6438 + j0.3630 –0.4806 – j0.6198 0.7843
8× π
16
–0.1985 – j0.5073 0.6945 + j0.4809 –0.5351 – j0.3500 0.6449
9× π
16
– 0.2824 – j0.4216 0.7917 + j0.5968 –0.4834 – j0.1681 0.5118
10 × π
16
– 0.3345 – j0.3283 0.9382 + j0.6901 –0.3984 – j0.0569 0.4024
11×π
16
– 0.3578 – j0.2384 1.1268 + j0.7387 –0.3191 – j0.0023 0.3191
12× π
16
– 0.3587 – j0.1602 1.3400 + j0.7235 –0.2572 – j0.0193 0.2580
13× π
16
– 0.3458 – j0.0985 1.5519 + j0.6341 –0.2132 – j0.0236 0.2145
14 × π
16
– 0.3283 – j0.0538 1.7323 + j0.4720 –0.1843 + j0.0192 0.1853
15× π
16
– 0.3142 – j0.0230 1.8538 + j0.2521 –0.1681 + j0.0104 0.1684
16 × π
16
– 0.3088 + j0 1.8966 + j0 –0.1628 + j0 0.1628

|H (e jω)|
1.0

0.9

0.8

0.707
0.7

0.6 ωc = ΩcT = 1.4489 × 1

0.5 1.4489
= × π = 0.46 π
π
0.4

0.3

0.2

0.1

ω
0 π 2π 4π
3π 5π 6π 7π 8π 9π 10 π 11π 12 π 13 π 14 π 15 π 16 π
16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16
Ωc = 0.46 π ( π/2 ) ( π)
F ig 3 : F req u en c y resp o nse of 3 rd ord er d ig ita l B u tterw o rth IIR lo w p a ss filte r.
7. 85 Digital Signal Processing

Example 7.21
Design a Butterworth digital IIR lowpass filter using impulse invariant transformation by taking T = 1second,
to satisfy the following specifications.

0.8 £ |H(ejw )| £ 1.0 ; for 0 £ w £ 0.2p

|H(ejw )| £ 0.2 ; for 0.32p £ w £ p

Draw direct form-I and II structure of the filter.

Alternatively,
Passband ripple £ 1.9 dB
Stopband attenuation ³ 13.97dB
Passband edge frequency = 0.2p rad/sample
Stopband edge frequency = 0.32p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e − δp,dB / 20j = 10b −1.9 / 20 g = 0.8

As = 10
e − α s ,dB / 20j = 10b −13.97 / 20g = 0.2

Solution
Specifications of digital IIR lowpass filter

Passband edge digital frequency, w p = 0.2p rad/sample

Stopband edge digital frequency, w s = 0.32p rad/sample
Gain in normal value at passband edge, Ap = 0.8
Gain in normal value at stopband edge, As = 0.2
Sampling time, T = 1second
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.8 Gain is same in analog
and digital filter.
Gain in normal value at stopband edge, As = 0.2
For impulse invariant transformation,
0.2π ωp
Passband edge analog frequency, Ωp = = = 0.6283 rad / second Using equation (7.55).
T 1
ω 0.32π
Stopband edge analog frequency, Ωs = s = = 1.0053 rad / second Using equation (7.56).
T 1
Order of the filter

LM FH1/ A 2 IK − 1O
P LM FH 1/ 0.22 IK − 1O
P
log
MN FH
s

1/ A p IK − 1 P
2
log
MN FH
1/ 0.8 2 IK − 1P log
LM 24 OP
N=
1 Q=1 Q= 1 N
0.5625 Q
= 3.9928 Using equation (7.57).
2 Ωs 2 .
10053 .
2 log 10053
log log
Ωp 0.6283 0.6283

Choose order N, such that N ³ N1 and N is an integer.

Let, order, N = 4.
Chapter 7 - IIR Filters 7. 86
Normalized transfer function, H(sn) of Butterworth lowpass filter

For even N,
N
2
1 Using equation (7.58).
b g ∏
H sn =
sn2 + bk sn + 1
k =1

where, bk = 2 sin LM b 2k −1 π g OP
N 2N Q Using equation (7.60).

Here, N = 4, ∴ k = 1, 2

1 1
∴ H sn = b g ×
sn2 + b1 sn + 1 sn2 + b2 sn + 1

When k = 1 ; bk = b1 = 2 sin LM b g OP = 0.7654

2−1 π Calculate sinq using
N Q 2× 4 calculator in radian mode.

= 2 sin LM b g O = 1.8478
2× 2−1 π
When k = 2 ; bk = b2
N QP
2 ×4

1
b g
H sn =
(sn2 + 0.7654 sn + 1) (sn2 + 18478
. sn + 1)
1
=
sn4 + 2.6132 sn2 + 3.4143 sn2 + 2.6132 sn + 1

Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 1.0053
Ωc = 1
= 1
= 0.6757 rad / second
e1/ A 2
s −1 j
2N c1/ 0.22 h −18

1
∴ H(s) = H snb g =
(sn2 + 0.7654 sn + 1)(sn2 + 18478
. sn + 1)
Using equation (7.61).
sn = s sn = s
Ωc Ωc
1 1
=
Fs 2
s IF s 2
s I = Fs 2
+ 0.7654 Ω c s + Ω c2 IFs 2
.
+ 18478 Ωc s + Ω 2c I
GH Ω 2
c
+ 0.7654
Ωc
+1 JK GH Ω 2
c
+ 1.8478
Ωc
+1 JK GH Ω 2c JK GH Ωc2 JK
Ω 4c 0.67574
= 2
=
(s + 0.7654 Ω cs + Ω c )(s 2
2
.
+ 18478 Ωc s + Ω 2c ) (s + 0.7654 × 0.6757 s + 0.67572 )
2

(s 2 + 18478
. × 0.6757 s + 0.67572 )

0.2085
= ....(1)
(s + 0.5172 s + 0.4566)(s2 + 12486
2
. s + 0.4566)
0.2085
= 4
s + 12486
. s3 + 0.4566 s 2 + 0.5172 s3 + 0.6458 s 2 + 0.2362 s + 0.4566 s 2 + 0.5701s + 0.2085
0.2085
=
s 4 + 1.7658 s3 + 1559
. s 2 + 0.8063 s + 0.2085
7. 87 Digital Signal Processing
To convert the analog transfer function to digital transfer function using impulse invariant transformation,
the equation (1) is simplified as follows.

The roots of the quadratic The roots of the quadratic

s 2 + 0.5172 s + 0.4566 = 0 are, s 2 + 1.2486 s + 0.4566 = 0 are,

−b ± b2 − 4ac −b ± b2 − 4ac
= =
2a 2a
−0.5172 ± 0.51722 − 4 × 1 × 0.4566 −1.2486 ± 12486
. 2
− 4 × 1 × 0.4566
= =
2 2
−0.5172 ± j1.2486 −1.2486 ± j0.5170
= = −0.2586 ± j0.6243 = = −0.6243 ± j0.2586
2 2
= (s − ( −0.2586 + j0.6243)) (s − ( −0.2586 − j0.6243)) = (s − ( −0.6243 + j0.2586)) (s − ( −0.6243 − j0.2586))
= (s + 0.2586 − j0.6243)(s + 0.2586 + j0.6243) = (s + 0.6243 − j0.2586)(s + 0.6243 + j0.2586)

0.2085
H(s) =
(s 2 + 0.5172 s + 0.4566)(s 2 + 1.2486 s + 0.4566)
0.2085
=
(s + 0.2586 − j0.6243) (s + 0.2586 + j0.6243)
(s + 0.6243 − j0.2586) (s + 0.6243 + j0.2586)

By partial fraction expansion H(s) can be expressed as,

A1 A1∗
H(s) = +
(s + 0.2586 − j0.6243) (s + 0.2586 + j0.6243)

A2 A∗2
+ +
(s + 0.6243 − j0.2586) (s + 0.6243 + j0.2586)
where, A1, A1*, A2, A2* are residues
0.2085 × (s + 0.2586 − j0.6243)
A1 =
(s + 0.2586 − j0.6243)(s + 0.2586 + j0.6243) (s + 0.6243 − j0.2586) (s + 0.6243 + j0.2586) s = − 0.2586 + j0.6243

0.2085
=
(−0.2586 + j0.6243 + 0.2586 + j0.6243) ( −0.2586 + j0.6243 + 0.6243 − j0.2586)
( −0.2586 + j0.6243 + 0.6243 + j0.2586)

0.2085
= = −0.3121 + j0.1293
j12486
. (0.3657 + j0.3657) (0.3657 + j0.8829)

A1∗ = conjugate of A1 = −0.3121 − j0.1293

0.2085 × (s + 0.6243 − j0.2586)

A2 =
bs + 0.2586 − j0.6243g bs + 0.2586 + j0.6243g (s + 0.6243 − j0.2586) bs + 0.6243 + j0.2586)g s = − 0.6243+ j0.2586

0.2085
=
(−0.6243 + j0.2586 + 0.2586 − j0.6243) ( −0.6243 + j0.2586 + 0.2586 + j0.6243)
(−0.6243 + j0.2586 + 0.6243 + j0.2586)

0.2085
= = 0.3121 − j0.7536
(−0.3657 − j0.3657) ( −0.3657 + j0.8829) j0.5172

A∗2 = conjugate of A 2 = 0.3121 + j0.7536

Chapter 7 - IIR Filters 7. 88
−0.3121 + j0.1293 −0.3121 − j0.1293
∴ H(s) = +
s + 0.2586 − j0.6243 s + 0.2586 + j0.6243
0.3121 − j0.7536 0.3121 + j0.7536
+ +
s + 0.6243 − j0.2586 s + 0.6243 + j0.2586

Digital IIR lowpass filter transfer function, H(z)

For impulse invariant transformation,

Ai Ai
→ Using equation (7.17).
s + pi 1 − e − p i T z −1

Using the above transformation, the H(s) can be transformed to H(z) as shown below,

−0.3121 + j0.1293 −0.3121 − j0.1293

H(z) = +
1 − e−(0.2586 − j0.6243) z−1 1 − e −(0.2586 + j0.6243) z −1
0.3121 − j0.7536 0.3121 + j0.7536 Put, T = 1
+ − ( 0.6243 − j0.2586 ) −1
+
1− e z 1 − e−(0.6243+ j0.2586 )z−1
( −0.3121 + j0.1293)(1 − e−0.2586 − j0.6243 z−1) + (−0.3121 − j0.1293)(1 − e−0.2586 + j0.6243 z−1)
=
(1 − e−0.2586 + j0.6243 z−1) (1 − e −0.2586 − j0.6243 z−1)

(0.3121 − j0.7536) (1 − e−0.6243 − j0.2586 z−1) + (0.3121 + j0.7536)(1 − e−0.6243+ j0.2586z−1)

+
(1 − e−0.6243+ j0.2586 z −1) (1 − e−0.6243 − j0.2586z −1)

−0.3121+ 0.3121e−0.2586 − j0.6243 z −1 + j0.1293 − j0.1293 e−0.2586 − j0.6243 z−1

− 0.3121+ 0.3121e −0.2586 + j0.6243 z−1 − j0.1293 + j0.1293 e−0.2586 + j0.6243 z−1
=
1 − e−0.2586 − j0.6243 z−1 − e −0.2586 + j0.6243 z −1 + e−2 × 0.2586 z−2

0.3121 − 0.3121e−0.6243 − j0.2586 z−1 − j0.7536 + j0.7536 e−0.6243 − j0.2586 z−1

0.3121 − 0.3121e−0.6243 + j0.2586 z −1 + j0.7536 − j0.7536 e−0.6243 + j0.2586 z−1
+
1 − e−0.6243 − j0.2586 z −1 − e−0.6243 + j0.2586 z−1 + e−2 × 0.6243 z −2

=
e j
−0.6242 + 0.3121e−0.2586 e j0.6243 + e− j0.6243 z−1 + j0.1293 e−0.2586 e j0.6243 − e− j0.6243 z−1 e j
1 − e−0.2586 (e j0.6243 + e− j0.6243 )z−1 + e−0.5172 z−2

+
e j
0.6242 − 0.3121e−0.6243 ej0.2586 + e− j0.2586 z−1 − j0.7536 e−0.6243 e j0.2586 − e− j0.2586 z−1 e j
−0.6243 j0.2586 − j0.2586 −1 −1.2486 −2
1− e (e +e )z +e z

=
− 0.6242 + 0.3121e b2 cos 0.6243g z + j0.1293 e c2 j sin 0.6243h z
−0.2586 −1 −0.2586 −1

1− e b2 cos 0.6243gz + e z
−0.2586 −1 −0.5172 −2

+
0.6242 − 0.3121e b2 cos 0.2586g z − j0.7536 e c2 j sin 0.2586h z
−0.6243 −1 −0.6243 −1

1− e b2 cos 0.2586gz + e z
−0.6243 −1 −1.2486 −2

− 0.6242 + 0.3911z −1 − 0.1167 z−1 0.6242 − 0.3232 z−1 + 0.2065 z−1

= +
.
1 − 12530 z−1 + 0.5962 z−2 1 − 1.0357 z−1 + 0.2869 z−2

− 0.6242 + 0.2744 z −1 0.6242 − 0.1167 z −1

= −1 −2
+
1 − 1.2530 z + 0.5962 z 1 − 1.0357 z −1 + 0.2869 z −2
7. 89 Digital Signal Processing

e
− 0.6242 + 0.2744 z−1 1 − 1.0357 z−1 + 0.2869 z −2 + 0.6242 − 0.1167 z−1 j e j
e1− 1.2530 z −1
+ 0.5962 z−2 j
∴ H(z) =
e1 − 1.2530 z −1
+ 0.5962 z −2
j e1 − 1.0357 z −1
+ 0.2869 z −2
j
−0.6242 + 0.6465 z−1 − 0.1791z−2 + 0.2744 z−1 − 0.2842 z−2 + 0.0787 z−3
+ 0.6242 − 0.7821z−1 + 0.3721z−2 − 0.1167 z−1 + 0.1462 z −2 − 0.0696 z−3
=
1 − 1.0357 z−1 + 0.2869 z−2 − 12530
. z−1 + 1.2977 z−2 − 0.3595 z−3
+ 0.5962z−2 − 0.6175 z−3 + 0.171z−4

0.0221z −1 + 0.055 z−2 + 0.0091z −3

=
1 − 2.2887 z−1 + 2.1808 z−2 − 0.977 z −3 + 0.171z−4
Alternatively,
0.0221z−1 + 0.055 z−2 + 0.0091z−3
H(z) =
1 − 2.2887 z−1 + 2.1808 z−2 − 0.977 z−3 + 0.171z−4
z−4 (0.0221z3 + 0.055 z2 + 0.0091z)
=
z (z − 2.2887 z3 + 2.1808 z2 − 0.977z + 0.171)
−4 4

0.0221z3 + 0.055 z2 + 0.0091

=
z − 2 .2887 z3 + 2.1808 z2 − 0.977z + 0.171
4

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.0221z −1 + 0.055 z −2 + 0.0091z −3

Let , H(z) = =
X(z) 1 − 2 .2887 z −1 + 2.1808 z −2 − 0.977 z −3 + 0.171z −4

On cross multiplying the above equation we get,

Y(z) –2.2887z–1Y(z) + 2.1808z–2Y(z) –0.977z–3Y(z) + 0.171z–4Y(z)
= 0.0221z–1X(z) + 0.055z–2X(z) + 0.0091z–3X(z)
\ Y(z) = 0.0221z–1X(z) + 0.055z–2X(z) + 0.0091z–3X(z)
+ 2.2887z–1Y(z) – 2.1808z–2Y(z) + 0.977z–3Y(z) – 0.171z–4Y(z) .....(2)
Using equation (2), the direct form-I structure is drawn as shown in fig 1.
X (z) + Y (z)

−1 −1
z z
−1
−1 0.0221z X (z ) 2.2887z −1Y (z)
z X (z) −1
0.02 21 + + 2.28 87 z Y (z)

−1 −1
z z
−2 −2
−2 0.055z X (z ) −2.1808z Y (z ) −2
z X (z) z Y (z )
0.05 5 + + −2.1808

−1
z −1
−3 −3
z
−3 0.0091z X (z) 0.977z Y (z )
z X (z) −3
0.00 91
+ 0 .9 7 7 z Y (z )

−1
−4 z
−0.171z Y (z ) −4
−0.171 z Y (z)

F ig 1 : D irec t fo rm -I stru ctu re o f 4 th o rd er d ig ita l IIR lo w p a ss filter.

Chapter 7 - IIR Filters 7. 90
Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.0221z −1 + 0.055 z −2 + 0.0091z −3

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 2 .2887 z −1 + 2.1808 z −2 − 0.977 z −3 + 0.171z −4

W(z) 1
where, = .....(3)
X(z) 1 − 2 .2887 z−1 + 2.1808 z−2 − 0.977 z−3 + 0.171z−4
Y(z)
= 0.0221z−1 + 0.055 z−2 + 0.0091z−3 .....(4)
W(z)
On cross multiplying equation (3) we get,
W(z) –2.2887z–1W(z) + 2.1808z–2W(z) –0.977z–3W(z) + 0.171z–4W(z) = X(z)
\ W(z) = X(z) + 2.2887z–1W(z) – 2.1808z–2W(z) + 0.977z–3W(z) – 0.171z–4W(z) .....(5)
On cross multiplying equation (4) we get,
Y(z) = 0.0221z–1W(z) + 0.055z–2W(z) + 0.0091z–3W(z) .....(6)
Using equations (5) and (6), the direct form-II structure is drawn as shown in fig 2.
X (z) W (z) Y (z )
+

−1
z
−1
2 .2887 z −1W ( z) 0.0221 z W (z )
z −1W ( z)
+ 2.28 87 0.02 21 +
−1
z
−2 .1808 z −2 W ( z) −2
0.055 z W (z )
z −2 W ( z )
+ −2.1808 0.05 5
+
−1
z
−3 −3
0 .977 z W (z ) 0.0091 z W (z )
z −3 W ( z)
+ 0 .9 7 7 0.00 91

−4 −1
−0.171 z W (z) z
−0.171 z −4 W ( z )

F ig 2 : D irec t fo rm -II stru cture o f 4 rd o rd er d ig ita l IIR lo w pa ss filter.

Note : Verify the result with MATLAB program 7.7.
Alternate method for transforming H(s) to H(z)
By partial fraction expansion H(s) can be expressed as,
0.2085
H(s) =
(s 2 + 0.5172 s + 0.4566) (s 2 + 12486
. s + 0.4566)
As + B Cs + D
= + .....(7)
(s 2 + 0.5172 s + 0.4566) (s 2 + 12486
. s + 0.4566)
On cross multiplying equation (7) we get,
0.2085 = (As + B) (s2 +1.2486s + 0.4566) + (Cs + D) (s2 + 0.5172s + 0.4566)
= As3 + 1.2486 As2 + 0.4566 As + Bs2 +1.2486 Bs + 0.4566B
+ Cs3 + 0.5172 Cs2 + 0.4566 Cs + Ds2 + 0.5172 Ds + 0.4566 D .....(8)
7. 91 Digital Signal Processing

On equating the coefficients of On equating the coefficients of s2 On equating coefficients of s in

s3 in equation (8) we get, in equation (8) we get, equation (8) we get,

A+C=0 1.2486 A + B + 0.5172 C + D = 0 0.4566 A + 1.2486B + 0.4566C

+0.5172 D = 0
\C= –A Put, C = – A
Put, C = – A
\ 1.2486 A +B – 0.5172A + D = 0
and, D = –0.7314 A –B
\ 0.7314 A + B + D = 0
0.4566 A + 1.2486 B – 0.4566 A
\ D = –0.7314 A – B
+ 0.5172 (–0.7314 A –B) = 0
–0.3783 A + 0.7314 B = 0
0.3783
∴ B= A = 0.5172A
0.7314

On equating constants of equation (8) we get,

0.4566 B + 0.4566 D = 0.2085
Put, D = –0.7314 A – B
\ 0.4566 B + 0.4566 (–0.7314 A – B) = 0.2085 Þ – 0.334A = 0.085
0.2085
∴ A=− = −0.6243
0.334
Here, A = –0.6243,
\ B = 0.5172A = 0.5172 ´ (–0.6243) = –0.3229
C = –A = 0.6243
D = –0.7314 A – B = –0.7314 (–0.6243) – (–0.3229) = 0.7795
Now, H(s) can be written as,
−0.6243s – 0.3229 0.6243s + 0.7795
H(s) = +
(s2 + 0.5172s + 0.4566) (s2 + 1.2486 s + 0.4566)
(s + a1)2 = s 2 + 2a1s + a12
F 0.3229IJ
–0.6243 G s + 0.5172
=
H 0.6243 K 2a1 = 0.5172 ⇒ a1 =
2
(s + 2 × 0.2586s + 0.2586 ) + F 0.4566 − 0.2586 I
2
2 2 2
H K = 0.2586

F 0.7795IJ
0.6243 G s + (s + a 2 )2 = s 2 + 2a 2s + a 22
+
H 0.6243K 12486
.
(s + 2 × 0.6243s + 0.6243 ) + F 0.4566 − 0.6243 I
2 2 2
2 2a 2 = 12486
. ⇒ a2 =
2
H K = 0.6243
−0.6243(s + 0.5172) 0.6243(s + 1.2486)
= +
(s + 0.2586)2 + 0.62432 (s + 0.6243)2 + 0.2586 2
−0.6243(s + 0.2586 + 0.2586) 0.6243(s + 0.6243 + 0.6243)
= +
(s + 0.2586)2 + 0.62432 (s + 0.6243)2 + 0.2586 2
(s + 0.2586) 0.6243
= – 0.6243 – 0.2586
(s + 0.2586)2 + 0.62432 (s + 0.2586)2 + 0.62432
(s + 0.6243) 0.6243 × 0.6243 0.2586
+ 0.6243 +
(s + 0.6243)2 + 0.25862 0.2586 (s + 0.6243)2 + 0.25862
Chapter 7 - IIR Filters 7. 92
(s + 0.2586) 0.6243
∴ H(s) = – 0.6243 – 0.2586
(s + 0.2586)2 + 0.62432 (s + 0.2586)2 + 0.62432
(s + 0.6243) 0.2586
+ 0.6243 + 15072
.
(s + 0.6243)2 + 0.2586 2 (s + 0.6243)2 + 0.2586 2
Digital IIR lowpass filter transfer function, H(z)
In impulse invariant transformation,

(s + a ) 1– e –aT (cos bT) z –1 Using equations (7.18)

2 2
  →
(s + a ) + b is transformed to
1 – 2 e –aT (cos bT ) z –1 + e –2aT z –2 and (7.19).
b e–aT (sin bT) z–1
2 2
  →
(s + a ) + b is transformed to
1 – 2 e (cos bT ) z –1 + e –2aT z –2
–aT

Using the above transformation, the H(s) can be transformed to H(z) as shown below.

1 − e−0.2586 (cos 0.6243) z−1 Put, T = 1

∴ H(z) = –0.6243 −0.2586
1− 2 e (cos 0.6243) z −1 + e−2 × 0.2586 z −2

e−0.2586 (sin 0.6243) z −1

– 0.2586 −0.2586
1− 2 e (cos 0.6243) z −1 + e−2 × 0.2586
z −2

1 − e−0.6243(cos 0.2586) z −1
+ 0.6243 −0.6243
1− 2 e (cos 0.2586) z−1 + e−2 × 0.6243 z−2

e−0.6243(sin 0.2586) z−1

+ 15072
. −0.6243
1 − 2e (cos 0.2586) z−1 + e−2 × 0.6243
z −2

−0.6243 + 0.3911z−1 −0.1167z−1

= −1 −2
+
1 − 12530
. z + 0.5962z 1 − 12530
. z−1 + 0.5962z−2
0.6243 − 0.3233z −1 0.2065z−1
+ −1 −2
+
1 − 10357
. z + 0.2869z 1 − 1.0357z−1 + 0.2869z−2

−0.6243 + 0.2744z −1 0.6243 − 0.1168z −1

= −1 −2
+
1 − 12530
. z + 0.5962z 1 − 10357
. z−1 + 0.2869z−2

( −0.6243 + 0.2744 z−1)(1 − 10357

. z−1 + 0.2869z −2 )

=
e
+ 0.6243 − 0.1168z−1 1 − 12530
. je
z−1 + 0.5962z−2 j
e .
1 − 12530z−1 + 0.5962z −2 j e1 − 10357
. z −1
+ 0.2869z−2 j
−0.6243 + 0.6466z−1 − 0.1791z−2 + 0.2744z −1 − 0.2842z−2 + 0.0787z−3
+ 0.6243 − 0.7822z−1 + 0.3722z−2 − 0.1168z−1 + 0.1464z−2 − 0.0696z−3
=
1 − 10357
. z−1 + 0.2869z−2 − 12530
. z−1 + 12977
. z−2 − 0.3595z−3
+ 0.5962z −2 − 0.6175z−3 + 0.171z−4

0.022z−1 + 0.0554z2 + 0.0091z−3

H(z) =
1 − 2.2887z−1 + 2.1808z−2 − 0.977z−3 + 0.171z−4

Note : The H(z) obtained by both the methods are same. The small difference in the coefficients are due to
the corrections (or rounding) made in calculations.
7. 93 Digital Signal Processing
Example 7.22
Design a Butterworth digital IIR lowpass filter using bilinear transformation by taking T = 1second, to
satisfy the following specifications.

0.707 £ |H(ejw )| £ 1.0 ; 0 £ w £ 0.2p

|H(ejw )| £ 0.08 ; 0.4p £ w £ p

Draw direct form-I and II structure of the filter.

Alternatively,

Passband ripple £ 3.0116dB

Stopband attenuation ³ 21.9382dB

Passband edge frequency = 0.2p rad/sample

Stopband edge frequency = 0.4p rad/sample

The above specifications can be converted to Ap and As as shown below.

Ap = 10
e− δp,dB / 20j = 10b−3.0116 / 20g = 0. 707

As = 10
e −α s ,dB / 20j = 10b−21.9382/ 20g = 0.08

Solution
Specifications of digital IIR lowpass filter

Passband edge digital frequency, w p = 0.2p rad/sample

Stopband edge digital frequency, w s = 0.4p rad/sample

Gain in normal value at passband edge, Ap = 0.707

Gain in normal value at stopband edge, As = 0.08

Sampling time, T = 1second

Specifications of analog IIR lowpass filter

Gain in normal value at passband edge, Ap = 0.707 Gain is same in analog

and digital filter.
Gain in normal value at stopband edge, As = 0.08

For bilinear transformation,

2 ωp 2 0.2π Using equation (7.53).

Passband edge analog frequency, Ω p = tan = tan
T 2 1 2
= 0.6498 rad / second

2 ω 2 0.4π Using equation (7.54).

Stopband edge analog frequency, Ω s = tan s = tan
T 2 1 2
= 1.4531 rad / second
Chapter 7 - IIR Filters 7. 94
Order of the filter

LM FH 1/ A 2 IK − 1O LM FH IK − 1 O
1/ 0.08 2
log
s
P log P
MN FH s K − 1P MN FH1/ 0.707 2 IK − 1P
1/ A 2 I
N1 =
1 Q=1 Q Using equation (7.57).
2 Ω 2 .
14531
log s log
Ωp 0.6498

log
155.25 LM OP
=
1 .
10006 N
= 3.1341 Q
.
2 log 14531
0.6498
Choose order N1 such that N ³ N1 and N is an integer.
Let, order, N = 4.
Normalized transfer function, H(sn) of Butterworth lowpass filter

For even N,
N
2
1
b g ∏s
H sn = 2
n + bk s n + 1
Using equation (7.58).
k =1

where, bk = 2 sin LM c 2k − 1 πh OP
N 2N
Q Using equation (7.60).
Here, N = 2, ∴ k = 1, 2

When k = 1, bk = b1 = 2 sin
LM b2 × 1 − 1gπ OP = 0.7654
MN 2 × 4 PQ Calculate sin q using

When k = 2, bk = b2 = 2 sin M
L b2 × 2 − 1gπ OP = 18478 calculator in radian mode.
MN 2 × 4 PQ .
1
b g
∴ H sn =
(sn2 + 0.7654 sn + 1) (sn2 + 18478
. sn + 1)
1
=
sn4 + 2.6132 sn3 + 3.4143 sn2 + 2.6132 sn + 1
Unnormalized transfer function, H(s) of Butterworth lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Ωs 14531
. Using equation (7.61).
Ωc = 1
= 1
= 0.7734 rad / second
e1/ A 2s j − 1 2N F 1 I 2× 4
H 0.08 2
−1
K
1
∴ H(s) = H sn b g =
sn4 + 2.6132 sn3 + 3.4143 sn2 + 2.6132 sn + 1 s
sn = s n =
s
Ωc Ωc
7. 95 Digital Signal Processing
1
∴ H(s) =
FG s IJ 4
F s IJ
+ 2.6132 G
3
+ 3.4143
FG s IJ 2
+ 2.6132
FG s IJ + 1
HΩ K c HΩ K c HΩ K
c HΩ K
c

1
=
s4 s3 s2 s
4 + 2.6132 3 + 3.4143 2 + 2.6132 +1
Ωc Ωc Ωc Ωc

Ω 4c
=
s + 2.6132 Ω c s + 3.4143 Ω 2c s 2 + 2.6132 Ω3c s + Ω 4c
4 3

0.77344
=
s + 2.6132 × 0.7734 s + 3.4143 × 0.77342 s 2 + 2.6132 × 0.77343 s + 0.77344
4 3

0.3578
=
s 4 + 2.021s3 + 2.0423 s 2 + 12089
. s + 0.3578
Digital IIR lowpass filter transfer function, H(z)
For bilinear transformation,

0.3578
H(z) = H(s) =
s 4 + 2.021s3 + 2.0423 s 2 + 12089
. s + 0.3578
1− z −1 1− z −1
s = 2 s = 2
T 1+ z −1 T 1+ z −1

0.3578
=
F 2 1− z I−1 4 F 2 1− z I −1 3 F 2 1− z I −1 2 F 2 1 − z I + 0.3578
−1
GH T 1 + z JK
−1
+ 2.021 GH T 1+ z JK −1
+ 2.0423 GH T 1 + z JK −1
+ 1.2089 GH T 1 + z JK
−1

0.3578
=
16(1 − z−1)4 + 2.021 × 8T (1 − z−1)3 (1 + z−1) + 2.0423 × 4T 2 (1 − z−1)2 (1 + z −1)2

+ 1.2089 × 2T3 (1 − z−1)(1 + z−1)3 + 0.3578 × T 4 (1 + z−1)4

T 4 (1 + z−1)4
Put, T = 1
0.3578 (1 + z−1)4
=
16 (1 − z−1)2(1 − z−1)2 + 16.168 (1 − 3 z−1 + 3z−2 − z−3 )(1 + z−1) + 8.1692 (1 − z−1)2(1 + z−1)2
+ 2 .4178 (1 − z−1)(1 + 3 z−1 + 3z−2 + z−3 ) + 0.3578 (1 + z−1)2(1 + z−1)2

0.3578 (1 + 2 z−1 + z−2 )(1 + 2 z−1 + z−2 )

= −1 −2 −1
16 (1 − 2 z + z )(1 − 2 z + z−2 ) + 16.168 (1 − 3 z−1 + 3z−2 − z−3 + z−1 − 3z−2 + 3z−3 − z−4 )
+8.1692 (1 − 2 z−1 + z −2 )(1 + 2 z−1 + z−2 ) + 2.4178 (1 + 3 z−1 + 3z−2 + z−3 − z−1 − 3z −2 − 3z −3 − z−4 )
+ 0.3578 (1 + 2 z−1 + z−2 )(1 + 2 z−1 + z−2 )

0.3578 (1 + 4 z−1 + 6z−2 + 4 z−3 + z−4 )

= −1 −2 −3
16 (1 − 4 z + 6z − 4z + z−4 ) + 16.168 (1 − 2z−1 + 2 z−3 − z−4 ) + 8.1692 (1 − 2z−2 + z−4 )
+ 2.4178 (1 + 2z −1 − 2 z−3 − z−4 ) + 0.3578 (1 + 4 z−1 + 6z −2 + 4 z−3 + z−4 )
0.3578 + 14312
. z−1 + 2.1468 z−2 + 14312
. z−3 + 0.3578 z −4
=
43.1128 − 90.0388 z−1 + 818084
. z−2 − 35.0684 z−3 + 5.7412z−4
Chapter 7 - IIR Filters 7. 96
0.3578 14312
. 2.1468 −2 14312
. 0.3578 −4
+ z−1 + z + z−3 + z
∴ H(z) = 43.1128 43 .1128 43 .1128 43 .1128 43 .1128
90.0388 −1 818084 . 35.0684 −3 5.7412 −4
1− z + z−2 − z + z
43.1128 43.1128 43.1128 43.1128

0.0083 + 0.0332 z−1 + 0.0498 z−2 + 0.0332 z−3 + 0.0083 z−4

=
1 − 2.0892 z−1 + 18975
. z−2 − 0.8133 z−3 + 0.1378 z−4

Alternatively,

0.0083 + 0.0332 z−1 + 0.0498 z−2 + 0.0332 z−3 + 0.0083 z −4

H(z) =
1 − 2.0892 z−1 + 18975
. z−2 − 0.8133 z −3 + 0.1378 z−4

z−4 (0.0083z4 + 0.0332 z3 + 0.0498 z2 + 0.0332 z + 0.0083)

=
z−4 (z4 − 2.0892 z3 + 18975
. z2 − 0.8133 z + 0.1378)

0.0083z4 + 0.0332 z3 + 0.0498 z2 + 0.0332 z + 0.0083

=
z4 − 2.0892 z3 + 18975
. z2 − 0.8133 z + 0.1378

Note: Verify the result with MATLAB program 7.8.

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.0083 + 0.0332 z −1 + 0.0498 z −2 + 0.0332 z −3 + 0.0083 z −4

Let , H(z) = =
X(z) 1 − 2.0892 z −1 + 18975
. z −2 − 0.8133 z −3 + 0.1378 z −4

On cross multiplying the above equation we get,

Y(z) – 2.0892z–1Y(z) + 1.8975z–2Y(z) – 0.8133 z–3Y(z) + 0.1378 z–4Y(z)
= 0.0083X(z) + 0.0332z–1X(z) + 0.0498z–2X(z) + 0.0332z–3X(z) + 0.0083z–4X(z)
\ Y(z) = 0.0083X(z) + 0.0332z–1X(z) + 0.0498z–2X(z) + 0.0332z–3X(z) + 0.0083z–4X(z)
+ 2.0892z–1Y(z) – 1.8975z–2Y(z) + 0.8133 z–3Y(z) – 0.1378 z–4Y(z) .....(1)
Using equation (1), the direct form-I structure is drawn as shown in fig 1.
0.0083X (z)
X (z) 0.00 83 + + Y (z)

−1 −1
z z
−1 −1
−1 0.0332z X (z) 2.0892 z Y (z ) −1
z X (z) 0.0332 + + 2.08 92 z Y (z)

−1 −1
z z
−2 −2
−2 0.0498z X (z) −1.8975z Y (z ) −2
z X (z) z Y (z )
0.0498
+ + −1.8975

−1 −1
z z
−3 0.0332z −3 X (z ) −3
0.8133z Y (z )
z X (z ) −3
0.03 32 + + 0.81 33 z Y (z )

−1 −1
z −4 z
−4 0.0083z X (z ) −0.1378z −4 Y (z )
−4
z X (z ) 0.00 83 −0.1378 z Y (z )

F ig 1 : D irec t fo rm -I stru ctu re o f 4 th o rd er d ig ita l IIR lo w p a ss filter..

7. 97 Digital Signal Processing
Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.0083 + 0.0332 z −1 + 0.0498 z −2 + 0.0332 z −3 + 0.0083 z −4

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 2.0892 z −1 + 18975
. z −2 − 0.8133 z −3 + 0.1378 z −4

W(z) 1
where, = .....(2)
X(z) 1 − 2.0892 z−1 + 18975
. z−2 − 0.8133 z−3 + 0.1378 z−4

Y(z)
= 0.0083 + 0.0332 z−1 + 0.0498 z−2 + 0.0332 z−3 + 0.0083 z −4 .....(3)
W(z)

On cross multiplying equation (2) we get,

W(z) – 2.0892z–1W(z) + 1.8975z–2W(z) – 0.8133 z–3W(z) + 0.1378 z–4W(z) = X(z)

\ W(z) = X(z) + 2.0892z–1W(z) – 1.8975z–2W(z) + 0.8133 z–3W(z) – 0.1378 z–4W(z) .....(4)

On cross multiplying equation (3) we get,

Y(z) = 0.0083W(z) + 0.0332z–1W(z) + 0.0498z–2W(z) + 0.0332z–3W(z) + 0.0083z–4W(z) .....(5)

Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.

W (z) 0.0083W (z )
X (z ) + 0.008 3
+ Y (z )

−1
z
−1 −1 −1
2.0892 z W (z) z W (z ) 0.0332z W (z )
+ 2.089 2 0.0332
+

−1
z
−2 −2
−1.8975z W (z ) −2 0.0498z W (z )
z W (z)
+ −1.8975 0.0498
+

−1
z
−3
0.8133z W (z ) −3 0.0332z −3 W (z )
z W (z )
+ 0.813 3 0.033 2 +
−1
z
−0.1378z −4 W (z ) −4 −4
0.0083z W (z)
z W (z)
−0.1378 0.008 3

F ig 2 : D irect fo rm -II structu re o f 4 th o rd er d ig ita l IIR lo w pass filter.

Example 7.23
Design a Chebyshev digital IIR lowpass filter using impulse invariant transformation by taking T = 1second,
to satisfy the following specifications.

0.9 £ |H(ejw )| £ 1.0 ; 0 £ w £ 0.25p

jw
|H(e )| £ 0.24 ; 0.5p £ w £ p

Draw direct form-I and II structure of the filter.

Chapter 7 - IIR Filters 7. 98
Alternatively,
Passband ripple £ 0.9151dB
Stopband attenuation ³ 12.3958dB
Passband edge frequency = 0.25p rad/sample
Stopband edge frequency = 0.5p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e− δp,dB / 20j = 10b−0.9151/ 20g = 0.9

As = 10
e −α s ,dB / 20j = 10b−12.3958 / 20g = 0.24

Solution
Specifications of digital IIR lowpass filter
Passband edge digital frequency, w p = 0.25p rad/sample
Stopband edge digital frequency, w s = 0.5p rad/sample
Gain in normal value at passband edge, Ap = 0.9
Gain in normal value at stopband edge, As = 0.24
Sampling time, T = 1second
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.9
Gain is same in analog
Gain in normal value at stopband edge, As = 0.24 and digital filter.
For impulse invariant transformation,

ωp 0.25π
Passband edge analog frequency, Ωp = = = 0.7854 rad / second Using equation (7.85).
T 1
ω s 0.5π
Stopband edge analog frequency, Ωs = = = 1.5708 rad / second Using equation (7.86).
T 1
Order of the filter
1 1
LM FH1/ A s2 IK − 1O 2 LM FH1/ 0.242 IK − 1O 2
cosh−1
MN FH 2I − 1P
P cosh−1
MN FH 1/ 0.92 IK − 1 P
P Using equation (7.87).
N1 =
1/ A p K Q = Q
Ωs .
15708
cosh−1 cosh−1
Ωp 0.7854
1

cosh −1 LM16.3611OP 2

= N 0.2346 Q = 2.1077
.
15708
cosh−1
0.7854
Choose order N1 such that N ³ N1 and N is an integer.
Let, order, N = 3.
7. 99 Digital Signal Processing
Normalized transfer function, H(sn) of Chebyshev lowpass filter

For odd N,
N − 1
2
B0 Bk Using equation (7.89).
H(sn ) =
sn + c 0 ∏ sn2 + bk sn + ck
k = 1

N−1 3−1
Here, N = 3, ∴ k= 2
=
2
=1

B0 B1
∴ H(sn ) = × 2
sn + c0 sn + b1 sn + c1
1

e j
∈ = 1/ Ap2 − 1 2 Using equation (7.94).

e j
= 1 / 0.92 − 1 2 = 0.4843

R|L 1 1 U|
I 1O LF 1 I O
1 1 −
N N
1 |MF 1
G + 1J + P − MG
2 2
+ P
1 |V Using equation (7.93).
yN = S
2 |MH ∈
M|N K ∈PP
2 MMH ∈ + 1JK 2
∈P
PQ ||
T Q N W
1 1
L 1
1 MF I 1
OP 1
2
3 LMF 1 I
1
2 1
OP −
3

= G
2 MH 0.4843
+ 1J +
K 0.4843 PP −2 2 MMGH 0.4843 + 1JK 2
+
0.4843 P
2
MN Q N PQ
1
= [1.6335 − 0.6122] = 0.5107
2
c0 = yN = 0.5107 Using equation (7.94).

bk = 2 yN sin LM c 2k − 1 π h OP Using equation (7.90).

N 2N Q
When k = 1, bk = b1 = 2 × 0.5107 sin
LM (2 × 1 − 1)π OP = 0.5107
N 2×3 Q
2
ck = yN + cos 2 LM c 2k − 1 πh OP Using equation (7.91).
N 2N Q
When k = 1, ck = c1 = 0.51072 + cos 2
LM (2 × 1 − 1)π OP
N 2×3 Q
F πI
LM 1 + cos FGH IJK OP 2π
1 + cos 2θ
= 0.5107 2
+ cos G J = 0.5107 +
2 2 6 cos2 θ =
H 6K MM 2 PP 2
N Q
= 0.2608 + 0.75 = 1.0108

B0 B1 B0 B1
∴ H(sn ) = × 2 = ×
sn + c 0 sn + b1 sn + c1 sn + 0.5107 sn2 + 0.5107 sn + 10108
.
Chapter 7 - IIR Filters 7. 100

To evaluate B0 and B1 , let, H(sn ) = 1.

sn = 0

B0 B1
When sn = 0, H(sn ) = = 19372
. B0 B1
(0.5107) (1.0108)

1
∴ 1.9372 B0 B1 = 1 ⇒ B0 B1 = = 0. 5162
1.9372
Let, B0 = B1 ; ∴ B20 = 0.5162 ⇒ B0 = 0.5162 = 0.7185

∴ B1 = B0 = 0.7185

B0 B1 0.7185 0.7185
∴ H(sn ) = × = ×
sn + 0.5107 (sn2 + 0.5107 sn + 1.0108) (sn + 0.5107) (sn2 + 0.5107 sn + 1.0108)

0.5162
=
(sn + 0.5107) (sn2 + 0.5107 sn + 1.0108)
0.5162
=
sn3 + 10214
. sn2 + 12716
. sn + 0.5162
Unnormalized transfer function, H(s) of Chebyshev lowpass filter

H(s) = H snb g s
sn =
Ωc

where, W c = Cutoff frequency.

Here, W c = W p = 0.7854 rad/sec.

0.5162
∴ H(s) = H sn b g =
(sn + 0.5107)(sn2 + 0.5107 sn + 10108
. )
sn = s sn = s
Ωc Ωc

0.5162
=
FG s IF s
+ 0.5107J G
2
s I
HΩ c KH Ω 2
c
+ 0.5107
Ωc
.
+ 10108 JK
0.5162
=
F s + 0.5107 Ω I F s c
2
+ 0.5107 Ω c + 10108
. Ω 2c I
GH Ω c
JK GH Ωc2 JK
0.5162 Ω3c
= 2
(s + 0.5107 Ω c )(s + 0.5107 Ωc s + 10108
. Ω 2c )

0.5162 × 0.78543
= 2
(s + 0.5107 × 0.7854)(s + 0.5107 × 0.7854s + 10108
. × 0.78542 )

0.2501 .....(1)
=
(s + 0.4011) (s 2 + 0.4011s + 0.6235)
0.2501
=
s3 + 0.8022 s 2 + 0.7844 s + 0.2501)
7. 101 Digital Signal Processing
To convert the analog transfer function to digital transfer function using impulse invariant transformation,
the equation (1) is simplified as shown below.

By partial fraction expansion, the H(s) can be expressed as,

0.2501 A Bs + C
H(s) = 2
= + 2 .....(2)
(s + 0.4011)(s + 0.4011s + 0.6235) s + 0.4011 s + 0.4011s + 0.6235

On cross multiplying the equation (1) we get

0.2501 = A(s2 + 0.4011s + 0.6235) + (Bs + C) (s + 0.4011)

0.2501 = As2 + 0.4011 As + 0.6235 A + Bs2 + 0.4011 Bs+ Cs + 0.4011 C ..... (3)

On equating the coefficients of On equating the coefficients of s in On equating constants of

s2 in equation (3) we get, equation (3) we get, equation (3) we get,

A+B=0 0.4011 A + 0.4011 B + C = 0 0.6235 A + 0.4011 C = 0.2501

Put, C = 0.
\ B = –A Put, B = – A
\ 0.6235 A = 0.2501
\ 0.4011 A – 0.4011 A + C = 0
0.2501
\C=0 ∴ A= = 0.4011
0.6235
B = − A = −0.4011

A Bs + C
∴ H(s) = +
(s + 0.4011) (s 2 + 0.4011s + 0.6235)
(s + a)2 = s 2 + 2as + a 2
0.4011 0.4011s
= − 0.4011
(s + 0.4011) (s 2 + 0.4011s + 0.6235) 2a = 0.4011 ⇒ a = = 0.2006
2
0.4011 0. 4011s
= −
(s + 0.4011)
( s 2 + 2 × 0.2006s + 0.2006 2 )+ FH 0.6235 − 0.20062IK 2

0.4011 s + 0.2006 − 0.2006

= − 0.4011
(s + 0. 4011) ( s + 0.2006)2 + 0.76372

0.4011 s + 0.2006 0.2006

= − 0.4011 + 0.4011
(s + 0. 4011) ( s + 0.2006)2 + 0.76372 ( s + 0.2006)2 + 0.76372

0.4011 s + 0.2006
= − 0.4011
(s + 0.4011) ( s + 0.2006)2 + 0.76372

0.4011 × 0.2006 0.7637

+
0.7637 (s + 0.2006)2 + 0.76372

0.4011 s + 0.2006 07637

= − 0.4011 + 0.1054
(s + 0.4011) ( s + 0.2006)2 + 0.76372 (s + 0.2006)2 + 0.76372
Chapter 7 - IIR Filters 7. 102
Digital IIR lowpass filter transfer function, H(z)
In impulse invariant transformation,
Ai Ai Using equations (7.17),
  → (7.18) and (7.19).
s + pi is transformed to − piT z −1
1− e
(s + a ) 1 − e− aT (cos bT)z−1
  →
(s + a)2 + b2 is transformed to
1 − 2e (cos bT )z−1 + e−2aT z−2
− aT

b e− aT (sin bT )z−1
  →
(s + a)2 + b2 is transformed to − aT
1 − 2e (cos bT )z−1 + e−2aT z−2
Using the above transformation, the H(s) can be transformed to H(z) as shown below.
Put, T = 1
0.4011 1 − e−0.2006 (cos 0.7637)z−1
∴ H(z) = −0.4011 −1
− 0.4011 −0.2006
1− e z 1 − 2e (cos 0.7637) z−1 + e −2× 0.2006 z−2

e−0.2006 (sin 0.7637)z−1

+ 0.1054 −0.2006
1− 2 e (cos 0.7637)z−1 + e−2 × 0.2006 −2
z

0.4011 −0.4011 + 0.2371z−1 0.0596z−1

= −1
+ −1 −2
+
1 − 0.6696 z 1 − 1.1820z + 0.6696 z 1 − 1.1820z −1 + 0.6696 z−2

0.4011 −0.4011 + 0.2371z−1 + 0.0596z−1

= −1
+
1 − 0.6696 z 1 − 1.1820z−1 + 0.6696 z−2

0.4011 −0.4011 + 0.2967z−1

= −1
+
1 − 0.6696 z 1 − 1.1820z−1 + 0.6696 z−2

0.4011 (1 − 1.1820z−1 + 0.6696 z−2 ) + (1 − 0.6696 z −1)( −0.4011 + 0.2967z−1)

=
(1 − 0.6696 z−1) (1 − 1.1820z −1 + 0.6696 z−2 )

0.4011 − 0.4741z−1 + 0.2686z−2 − 0.4011+ 0.2961z−1 + 0.2686z−2 − 0.1988z−2 )

=
1 − 1.1820z−1 + 0.6696 z −2 − 0.6696 z−1 + 0.7915z−2 − 0.4484z−3

0.0906z−1 + 0.0698z −2
=
1 − 18516
. z−1 + 14611
. z−2 − 0.4484z−3
Alternatively,

0.0906z −1 + 0.0698z−2 z−3(0.0906z2 + 0.0698z)

H(z) = −1 −2 −2
= −3 3
− .
1 18516 z + .
14611z − 0.4484z z (z − 18516
. z2 + 14611
. z − 0.4484)
0.0906z2 + 0.0698z
=
z − 1.8516z2 + 1.4611z − 0.4484
3

Note: Verify the result with MATLAB program 7.9.

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.0906z−1 + 0.0698z−2

Let , H(z) = =
X(z) 1 − 18516
. z−1 + 14611
. z−2 − 0.4484z−2

On cross multiplying the above equation we get,

Y(z) – 1.8516z–1Y(z) + 1.4611z–2Y(z) – 0.4484z–3 Y(z) = 0.0906z–1 X(z) + 0.0698z–1X(z)
\ Y(z) = 0.0906z–1X(z) + 0.0698z–2X(z) + 1.8516z–1Y(z) – 1.4611z–2Y(z)
+ 0.4484z–3Y(z) .....(4)
7. 103 Digital Signal Processing
Using equation (4), the direct form -I structure is drawn as shown in fig 1.

X (z) + Y (z)

−1 −1
z z
−1
−1
0.0906z X (z ) 1.8516z −1Y (z) −1
z X (z)
0.09 06 + + z Y (z)

−1 −1
z z
−2 0.0698z −2 X (z ) −1.4611z −2 Y (z ) −2
z X (z) z Y (z )
0.06 98
+ −1.4611

−1
z
0.4484z −3 Y (z ) −3
0.44 84 z Y (z )

F ig 1 : D irec t fo rm -I stru ctu re o f 3 rd o rd e r d ig ita l IIR lo w pa ss filte r.

Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.0906z−1 + 0.0698z−2

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 18516
. z−1 + 14611
. z−2 − 0.4484z−2

W(z) 1
where, = .....(5)
X(z) .
1 − 18516z −1 + 14611
. z−2 − 0.4484z−2

Y(z)
= 0.0906z −1 + 0.0698z−2 .....(6)
W(z)
On cross multiplying equation (5) we get,
W(z)– 1.8516z–1W(z) + 1.4611z–2W(z) – 0.4484z–3 W(z) = X(z)
\ W(z) = X(z) + 1.8516z–1W(z) – 1.4611z–2W(z) + 0.4484z–3 W(z) .....(7)
On cross multiplying equation (6) we get,
Y(z) = 0.0906z–1X(z) + 0.0698z–2W(z) .....(8)
Using equations (7) and (8), the direct form-II structure is drawn as shown in fig 2.

X (z) W (z) Y (z )
+

−1
z
1.8516z −1W (z ) 0.0906z −1W (z)
z −1W ( z)
+ 0.09 06
+
−1
z
−1.4611z −2 W (z)
z −2 W ( z ) 0.0698z −2 W (z )
+ −1.4611 0.06 98

−1
z
0.4484z −3 W (z )
z −3 W ( z)
0.44 84

F ig 2 : D irec t fo rm -II stru cture o f 3 rd o rd er d ig ita l IIR lo w pa ss filter.

Chapter 7 - IIR Filters 7. 104
Example 7.24
Design a Chebyshev digital IIR lowpass filter using bilinear transformation by taking T = 1second, to
satisfy the following specifications.
0.8 £ |H(ejw )| £ 1.0 ; for 0 £ w £ 0.2p
jw
|H(e )| £ 0.2 ; for 0.32p £ w £ p
Draw direct form-I and II structure of the filter.
Alternatively,
Passband ripple £ 1.9dB
Stopband attenuation ³ 13.97dB
Passband edge frequency = 0.2p rad/sample
Stopband edge frequency = 0.32p rad/sample
The above specifications can be converted to Ap and As as shown below.

Ap = 10
e− δp,dB / 20j = 10b−1.9 / 20g = 0.8

As = 10
e −α s ,dB / 20j = 10b−13.97 / 20g = 0.2

Solution
Specifications of digital IIR lowpass filter
Passband edge digital frequency, w p = 0.2p rad/sample
Stopband edge digital frequency, w s = 0.32p rad/sample
Gain in normal value at passband edge, Ap = 0.8
Gain in normal value at stopband edge, As = 0.2
sampling time, T = 1second
Specifications of analog IIR lowpass filter
Gain in normal value at passband edge, Ap = 0.8 Gain is same in analog
and digital filter.
Gain in normal value at stopband edge, As = 0.2
For bilinear transformation,
2 ωp
Passband edge analog frequency, Ωp = tan Using equation (7.83).
T 2
2 0.2π
= tan = 0.6498 rad / second
1 2
2 ω
Stopband edge analog frequency, Ωs = tan s
T 2 Using equation (7.84).
2 0.32π
= tan = 1.0995rad / second
1 2
Order of the filter
1 1

−1
LM FH1/ A 2
s
IK − 1O 2
P −1
LM e1/ 0.22 j − 1OP 2 Using equation (7.87).
cosh cosh
MN FH 2I − 1P
1/ Ap K Q MN e
1/ 0.8 j − 1P
2
Q cosh−1 6.5319
N1 = = = = 2.2944
−1 Ωs .
10995 cosh−1 16921
.
cosh cosh−1
Ωp 0.6498
Choose order N1 such that N ³ N1 and N is an integer.
Let, order, N = 3.
7. 105 Digital Signal Processing
Normalized transfer function, H(sn) of Chebyshev lowpass filter

For odd N,
N − 1
2 Using equation (7.89).
B0 Bk
H(sn ) =
sn + c 0 ∏ sn2 + bk sn + ck
k = 1

N− 1 3−1
Here, N = 3, ∴ k= 2
=
2
=1

B0 B1
∴ H(sn ) = × 2
sn + c 0 sn + b1 sn + c1

e
∈ = 1/ Ap2 − 1 2j Using equation (7.94).

e j
= 1 / 0.82 − 1 2 = 0.75

R|L 1 1 U|
|SMFG 1 + 1IJ + 1OP − LMF 1 I OP
1 1 −
N N
1 2 2 1 |V
yN =
2 ||MMNH ∈ K ∈PPQ
2 MMGH ∈ + 1JK
2
+
∈ PP ||
Using equation (7.93).

T N Q W
1 1
L
1 MF 1 I 1 OP
1
2
3 LMF 1 I 1
2 1
OP −
3

= G
2 MH 0.75
+ 1J +
K 0.75 PP −
MMGH 0.75 + 1JK +
PPQ
2 2
0.75
NM Q N
1
= [1.4422 − 0.6934] = 0.3744
2
c0 = yN = 0.3744 Using equation (7.94).

bk = 2 yN sin LM c 2k − 1 πh OP Using equation (7.90).

N 2N Q
When k = 1, bk = b1 = 2 × 0.3744 sin
LM (2 × 1 − 1)π OP = 0.3744
N 2×3 Q
ck = yN2 + cos 2 LM c h
2k − 1 π OP Using equation (7.91).
N 2N Q
When k = 1, ck = c1 = yN2 + cos 2
LM (2 × 1 − 1)π OP
MN 2 × 3 PQ
L1 + cos FGH 2π IJ O 1 + cos 2θ
= 0.37442 + cos 2
π
= 0.3744 + M2 6 KP cos2 θ =
6 MMN 2 PPQ 2

= 0.1402 + 0.75 = 0.8902

B0 B1 B0 B1
∴ H(sn ) = × 2 = ×
sn + c 0 sn + b1 sn + c1 sn + 0.3744 sn2 + 0.3744 sn + 0.8902
Chapter 7 - IIR Filters 7. 106

To evaluate B0 and B1, let, H sn b g sn = 0

=1

B0 B1
When sn = 0, H(sn ) = = 3B0 B1
(0.3744) (0.8902)
1
∴ 3 B0 B1 = 1 ⇒ B0 B1 = = 0.3333
3
2
Let, B0 = B1 ; ∴ B0 = 0.3333 ⇒ B0 = 0.3333 = 0.5774
∴ B1 = B0 = 0.5774
B0 B1 0.5774 0.5774
H(sn ) = × 2 = × 2
sn + 0.3744 sn + 0.3744 sn + 0.8902 sn + 0.3744 sn + 0.3744 sn + 0.8902
0.3333 0.3333
= =
(sn + 0.3744) (sn2 + 0.3744 sn + 0.8902) sn3 + 0.7488 sn2 + 1.0304 sn + 0.3333
Unnormalized transfer function, H(s) of Chebyshev lowpass filter

H(s) = H sn b g s
sn =
Ωc

where, W c = Cutoff frequency.

Here, Wc = Wp = 0.6498 rad/sec.

0.3333
∴ H(s) = H sn b g s
=
(sn3 + 0.7488 sn2 + 1.0304 sn + 0.3333) s
sn = sn =
Ωc Ωc

0.3333
=
Fs 3
s2 s I
GH Ω 3
c
+ 0.7488 2 + 1.0304
Ωc Ωc
+ 0.3333 JK
0.3333 × Ω3c
=
s + 0.7488 Ω c s 2 + 1.0304 Ω c2 s + 0.3333 Ω3c
3

0.3333 × 0.64983
=
s3 + 0.7488 × 0.6498 s 2 + 1.0304 × 0.64982 s + 0.3333 × 0.64983
0.0914
=
s3 + 0.4865 s 2 + 0.4351s + 0.0914
Digital IIR lowpass filter transfer function, H(z)
For bilinear transformation,

0.0914
H(z) = H s bg s=
2 1− z −1
T 1+ z −1
=
s3 + 0.4865 s 2 + 0.4351s + 0.0914 s = 2 1− z −1
T 1+ z −1

0.0914
=
F 2 1− z I −1 3 F 2 1− z I −1 2 F 2 1− z I + 0.0914
−1

GH T 1 + z JK −1 + 0.4865 GH T 1 + z JK
−1 + 0.4351GH T 1 + z JK
−1

0.0914
=
8(1− z −1)3 1.946(1− z−1)2 0.8702(1− z−1)
+ + + 0.0914
T3 (1 + z−1)3 T 2(1 + z−1)2 T (1 + z−1)
7. 107 Digital Signal Processing
0.0914
∴ H(z) =
8(1− z−1)3 + 1.946 T(1− z−1)2 (1 + z−1) + 0.8702 T 2(1− z−1)(1 + z−1)2 + 0.0914 T3(1 + z−1)3
T3 (1 + z−1)
0.0914 T3 (1 + z−1)3
=
8(1− z ) + 1.946 T(1− z ) (1 + z ) + 0.8702 T 2(1− z−1)(1 + z−1)2 + 0.0914 T3 (1 + z−1)3
−1 3 −1 2 −1

Put, T = 1
0.0914 (1 + 3z−1 + 3z−2 + z−3 )
=
8(1− 3z−1 + 3z −2 − z−3 ) + 1.946 (1− 2z−1 + z −2 )(1 + z−1) + 0.8702(1− z−1)(1 + 2z−1 + z−2 )
+ 0.0914 (1 + 3z−1 + 3z−2 + z−3 )

0.0914 (1 + 3z−1 + 3z−2 + z−3 )

= −1 −2
8(1 − 3 z + 3z − z ) + 1.946(1 − z −1 − z−2 + z−3 ) + 0.8702(1 + z−1 − z −2 − z−3 )
−3

+ 0.0914 (1 + 3z−1 + 3z −2 + z −3 )

0.0914 + 0.2742 z−1 + 0.2742 z−2 + 0.0914z−3

=
10.9076 − 24.893z −1 + 213666
. z−2 − 6.8328z−3 (a + b)2 = a 2 + 2ab + b2
(a − b)2 = a 2 − 2ab + b 2
0.0914 0.2742 z−1 0.2742 z−2 0.0914 −3
+ + + z (a + b)3 = a3 + 3a 2b + 3ab2 + b3
= 10 .9076 10.9076 10.9076 10 .9076
24.893z −1
213666
. z −2
6.8328z −3 (a − b)3 = a3 − 3a 2b + 3ab 2 − b3
1− + −
10.9076 10.9076 10.9076
0.0083 + 0.0251z−1 + 0.0251z−2 + 0.0083z−3
=
1 − 2.2821z−1 + 19589
. z−2 − 0.6264z−3
Alternatively,

0.0083 + 0.0251z−1 + 0.0251z −2 + 0.0083 z−3

H(z) =
1 − 2.2821z −1 + 19589
. z−2 − 0.6264 z−3

z−3(0.0083 z3 + 0.0251z2 + 0.0251z + 0.0083

=
z−3(z3 − 2.2821z2 + 19589
. z − 0.6264)

0.0083 z3 + 0.0251z2 + 0.0251z + 0.0083

=
z3 − 2.2821z2 + 19589
. z − 0.6264

Note: Verify the result with MATLAB program 7.10.

Direct form-I structure of digital IIR lowpass filter

Y(z) 0.0083 + 0.0251z −1 + 0.0251z −2 + 0.0083 z −3

Let , H(z) = =
X(z) 1 − 2.2821z −1 + 19589
. z −2 − 0.6264 z −3

On cross multiplying the above equation we get,

Y(z) – 2.2821z–1Y(z) + 1.9589z–2Y(z) – 0.6264z–3 Y(z) = 0.0083 X(z)

+ 0.0251z–1X(z) + 0.0251z–2X(z) + 0.0083z–3X(z)

\ Y(z) = 0.0083X(z) + 0.0251z–1X(z) + 0.0251z–2X(z) + 0.0083z–3X(z)

+ 2.2821z–1Y(z) – 1.9589z–2Y(z) + 0.6264z–3Y(z) .....(1)

Chapter 7 - IIR Filters 7. 108
Using equation (1), the direct form -I structure is drawn as shown in fig 1.

0.0083X (z)
X (z) 0.00 83 + + Y (z)

−1 −1
z z
−1 −1
−1 0.0251z X(z ) 2.2821z Y (z ) −1
z X (z) z Y (z)
0.0251
+ + 2.28 21

−1 −1
z z
−2 −2
−2 0.0251z X(z ) −1.9589z Y (z ) −2
z X (z) z Y (z )
0.0251
+ + −1.9589

−1 −1
z −3 z
−3
−3
0.0083z X(z ) 0.6264z Y (z ) −3
z X (z) 0.00 83 0.62 64 z Y (z )

F ig 1 : D irec t form -I stru ctu re o f 3 rd o rd e r d ig ita l IIR lo w pa ss filte r..

Direct form-II structure of digital IIR lowpass filter

Y(z) W(z) Y(z) 0.0083 + 0.0251z −1 + 0.0251z −2 + 0.0083 z −3

Let , H(z) = = × =
X(z) X(z) W(z) 1 − 2.2821z −1 + 1.9589 z −2 − 0.6264 z −3

W(z) 1
where, = .....(2)
X(z) 1 − 2.2821z−1 + 19589
. z−2 − 0.6264 z−3

Y(z)
= 0.0083 + 0.0251z−1 + 0.0251z−2 + 0.0083 z−3 .....(3)
W(z)
On cross multiplying equation (2) we get,
W(z) – 2.2821z–1W(z) + 1.9589z–2 W(z) –0.6264 z–3W(z) = X(z)
W(z) = X(z) + 2.2821z–1W(z) – 1.9589z–2 W(z) + 0.6264z–3W(z) .....(4)
On cross multiplying equation (3) we get,
Y(z) = 0.0083W(z) + 0.0168z–1W(z) + 0.0168z–2W(z) + 0.0083z–3W(z) .....(5)
Using equations (4) and (5), the direct form-II structure is drawn as shown in fig 2.

W (z) 0.0083W (z )
X (z ) + 0.00 83
+ Y (z )

−1
z
−1 −1
2.2821z W (z) z W (z ) 0.0251z W (z)
−1

+ 2.28 21 0.0251
+

−1
z
−2 −2
−1.9589z W (z ) −2 0.0251z W (z)
z W (z)
+ −1.9589 0.0251
+

−1
z
−3 −3
0.6264z W (z) −3
0.0083z W (z )
z W (z )
0.62 64 0.00 83

F ig 2 : D irec t fo rm -II stru cture o f 3 rd o rd er d ig ita l IIR lo w pa ss filter.