DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 1
Introduction
Many electronic and communication devices contain digital systems in the form of digital
filters. Filtering is the process of removing or blocking unwanted signal from the desired information
signal that can be realized as lowpass, highpass, bandpass or bandstop filters. The digital filter
performs mathematical operations on a sampled, digitized signal to reduce or enhance certain features
of the processed signal. The two most commonly used filters are Finite Impulse Response (FIR) and
Infinite Impulse Response (IIR) filters. FIR filters are simple to design, implement, and are
guaranteed to be stable systems compared to the IIR filters. They can also be designed to have linear
phase characteristics that is desirable in music and video processing. On the other hand, IIR filters are
useful for high-speed designs because they typically require a lower filter order compared to FIR
filters for the same design specifications, however they do not have linear phase and they can be
unstable if not designed properly.

The digital filter design problem involves the determination of a set of filter coefficients to
meet a set of design specifications: the passband and stopband widths and edge frequencies, the
passband ripple factor and the stopband attenuations. IIR and FIR filters are widely used in digital
signal processing (DSP). The choice of one of these methods depends on the design requirements.
Because of their less complexity, IIR filters are the most efficient type of filter to implement in DSP.
Higher order IIR filters can also be implemented in cascade form using second order sections called
biquads that makes the amount of implementation processing relatively small compared to the direct
form implementations. In most of the cases, IIR filters are designed from their analog counterpart
using bilinear transformations that is based on mappings from the s-plane to the z-plane.

E&C 2023-24 1
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 2
Literature Survey
Paper Authors Highlights Drawbacks

Design and comparison Vijay Kumar, Elliptical filter finding out Cannot adjust to
between IIR Mansi Arya, Amit sharpest roll-off in both automatically
Butterworth and Kumar, passbands and stopbands but change in input
Chebyshev digital Dharmendra with ripple in stopbands, parameters
filters using MATLAB Kumar Jhariya non-linear phase response.
-2024 Chebyshev filter has sharper
roll-off than the Butterworth
filter but with ripples in
either passband or stopband
and.non-linear phase
response.
Butterworth filter has flat
passband response but has a
gradual roll off in the
stopband.

In this work, the steps of IIR filter design using bilinear transformation with MATLAB simulations
will be shown by providing some design examples based on the filter design specifications. In the
end, the cost and complexity of implementation of realizing the filter structures is briefly discussed.

E&C 2023-24 2
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 3
Approach on bilinear transformation
The bilinear transformation is the most commonly used technique for digital filter design as it
preserves the system function frequency characteristics when there is transformation from analog to
digital domain. When doing from s-plane to z-plane mapping following procedure to be followed

• Express z in terms of s (T = sampling time)

and

Apply Taylor series and approximating the equation

Rearrange

Use where and then

Substituting
Then

one can find that the entire left half s-plane (sigma <0) maps into the inside of the unit circle (z=1)
that makes a stable system in the analog or s-domain maps into a stable system in the digital or z-
domain. The right half s-plane (sigma >0) maps into the region outside the unit circle in the z-plane,
which corresponds to unstable systems. The imaginary axis maps into the unit circle in a one-to-one
fashion. Hence there is no aliasing in the frequency domain like that of the impulse invariant method

E&C 2023-24 3
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

as shown in the s-plane to z-plane mapping in Fig. 1.

E&C 2023-24 4
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 4
Design of IIR filter
The steps involved in the IIR filter design using the bilinear transformation technique are :
a) Set the digital filter specifications ωp, ωs, Rp, As.

 ωp: digital passband edge frequency in rad

 ωs: digital stopband edge frequency in rad

 Rp: the passband ripple factor in dB

 As: the stopband attenuation in dB.

b) Transform the digital filter specifications given in part (a) to analog specifications (ωp to omega p
and ωs to omega s) and design the analog filter using Butterworth, Chebyshev, Elliptic or other types
as desired. In the case of bilinear transformation, it is necessary to prewarp the cutoff frequencies ωp
and ωs to calculate Ωp and Ωs during transforming from analog to digital specifications.

 Ωp: the analog passband cutoff frequency in rad/sec

 Ωs: the analog stopband cutoff frequency in rad/sec

c) Apply the bilinear transformation to obtain the digital filter from the corresponding analog filter. It
means the desired analog filter represented by Ha(s) in Laplace domain is transformed to the
equivalent digital filter H(z) in the z- domain by mapping s to z.
Design examples

1) BUTTERWORTH LOWPASS FILTER

The Butterworth filter is characterized by the property that its magnitude response is flat in both
passband and stopband. A stable and causal filter with transfer function, Ha(s), in Laplace domain
will be derived based on the anlaog specifications omega s, omega p Rp, As. Out of the 2N poles that
are symmetrically located with respect to the jΩ axis, N of the poles on the left half of the s-plane
(LHP) will be selected to construct the stable causal system Ha(s). The essence of the design in the
case of Butterworth filter is to obtain the order N and the cutoff frequency Ωc, given these
specifications. These parameters depend on the values of the specifications omega s, omega p Rp, As.
The detail of the equations for N and Ωc can be found . The algorithm for the design of the analog
filter for normalized cutoff frequency (i.e., Ωc = 1) is implemented using butterap (N) function in

E&C 2023-24 5
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

MATLAB signal processing toolbox. Once the analog filter Ha(s) is designed, the corresponding
digital filter represented by H(z) can be derived from the analog filter.
To demonstrate the digital filter design procedures, we would like to design a digital lowpass filter
based on the following digital domain specifications

Fig. 4. shows the MATLAB simulation results based on the specifications as given above using the
bilinear transformation design steps for Butterworth lowpass filter. It shows the magnitude response
in linear scale, in dB, the phase response and the pole-zero plot of the digital filter. As shown in the
Figure, the Butterworth filter achieves a maximally flat response both in the passband and stop band.
The order of the filter is N=24 and the number of poles (‘×’- marked) and zeros (‘o’-marked) is equal
to N and the poles reside inside the unit circle (z=1) in the z-plane confirming the stability of the
system.

2) CHEBYSHEV-I LOWPASS FILTER

Chebyshev filters provide lower order than Butterworth filters for the same specifications due to their
tolerance for ripples.

E&C 2023-24 6
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

and
A stable and causal analog filter represented by Ha(s) is derived from (11) similar to the Butterworth
case. The parameters Ωc, and N are derived given the specifications for Ωp, Ωs, Rp, and As. The
cuttoff frequency is Ωc = Ωp for the Chebyshev-I filter. Similar procedures used for the design of
digital Butterworth filter are used to design digital Chebyshev-I filter. The digital filter specifications
will be transformed to the equivalent analog specifications and then the digital filter will be designed
from the analog filter using the bilinear transformations. Fig. 5. shows the MATLAB simulation
results based on the specifications as given in using the bilinear transformation design steps for
Chebyshev-I lowpass filter. It shows the magnitude response in linear scale, in dB, the phase
response and the pole-zero plot of the digital filter. As expected, there is a ripple in the passband and
no ripple in the stop band. The order of the filter N=10 less than the Butterworth filter since the filter
is designed with a tolerance of a ripple in the passband. The number of poles and zeros is equal to N
and the poles are located inside the unit circle in the z-plane.

3) ELLIPTIC LOWPASS FILTER

E&C 2023-24 7
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

These filters exhibit equi ripple behavior in the passband as well as in the stopband. Elliptic filters are
optimum filters in that they achieve the minimum order N for the given specifications (or alternately,
achieve the sharpest transition band for the given order N). The minimum order N is achieved due to
their tolerance of having ripples in both passband and stopband. The magnitude-squared response of
elliptic filters is given by

where N is the order, omega p is the passband ripple (which is related to Rp), and UN(・) is the Nth-
order Jacobian elliptic function. Fig. 7. shows the MATLAB simulation results based on the
specifications as given above using the bilinear transformation design steps for elliptic lowpass filter.
There is equi ripple in both the passband and stopband. The order of the filter is N=6 and it is the
smallest compared to the Chebyshev and Butterworth filters as expected. The number of poles and
zeros is also equal to N.

4) IMPLEMENTATION OF DIGITAL FILTERS

Discrete time systems described by a transfer function or linear constant coefficient difference
equation are often implemented in a digital hardware by an algorithm or structures that consists of an
interconnection of adders, multipliers and delay or storage registers [6]. An implementation is

E&C 2023-24 8
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

considered cost effective if it minimizes the number of required adders, multipliers and delay
elements used in realizing the structure. For example, consider a second order system relating a
discrete time input x[n] and a discrete time output y[n] described by the system function:

An alternative representation of this system is the difference equation relating the input and the
output and given as:

Fig.. shows the Direct form II or canonical block diagram realization of the system. As seen from the
Figure, the system requires four adders, five multipliers and two delay elements.

E&C 2023-24 9
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 5
Results
Based on this concept, Table I shows the number of adders, multipliers and delay elements for the
lowpass and bandpass filters studied in the previous sections. Since the number of poles and number
zeros are equal for the different digital filters, the table is completed by considering M=N as per the
simulation results studied in the previous sections. As expected, the Butterworth filters require the
highest cost of implementations and the Elliptic filters need optimum number of elements and hence
it has the lowest complexity and cost compared to the others but at the expense of the ripples in the
passband and stopband.

E&C 2023-24 10
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 6
Conclusion
Most modern electronic devices or digital systems have filters implemented in the form of DSP
hardware and/or software. In this digital IIR filter design using bilinear transformation is
demonstrated using MATLAB simulations. An analog prototype filter is the basis for designing the
corresponding digital filter based on the filter specifications in the digital domain. Elliptic filter is the
optimum filter among the others in terms of implementation cost and simplicity. This is achieved in
tradeoff of having ripples in both the passband and stopband that may not be tolerated in some
applications. Frequency selective digital filters such as bandpass filters can be constructed from the
analog lowpass filters by firstly transforming the analog filter to the selected band in the analog
domain and then transforming to digital domain or by designing the digital lowpass and transforming
to the selected band in the digital domain.

E&C 2023-24 11
 DESIGN OF DIGITAL IIR FILTER USING BILINEAR TRANSFORMATION IN MATLAB

Chapter 7
References
[1] John G. Proakis and Dimitris K Manolakis, Digital Signal Processing: Principles, Algorithms
and Applications, 4th Edition, Pearson, 2014.
[2] Vinay K. Ingle, John G. Proakis, Digital Signal Processing Using MATLAB, 3rd Edition,
Cengage Learning, 2012.
[3] L. Litwin, "FIR and IIR digital filters," IEEE Potentials, vol. 19, no. 4, pp. 28-31, 2000.
[4] Design and Comparison Between IIR Butterworth and Chebyshev Digital Filters using MATLAB
by Vijay Kumar, Mansi Arya, Amit Kumar, Dharmendra Kumar Jhariya , 2024 Fourth International
Conference on Advances in Electrical, Computing, Communication and Sustainable Technologies
(ICAECT) | 979-8-3503-4367-0/24/$31.00 ©2024 IEEE | DOI:1109/ICAECT60202.2024.10469026.

E&C 2023-24 12