Department of Electrical Engineering

Faculty Member:____________________ Dated: ____________________

Course/Section:____________________ Semester: __________________

EE-330 Digital Signal Processing

Lab 7 IIR Filter Design using Analog Prototypes

PLO4-CLO4 PLO5- PLO8- PLO9-

CLO5 CLO6 CLO7

Name Reg. No Viva / Analysis Modern Ethics Individual

Quiz / Lab of data Tool and and Team
Performa in Lab Usage Safety Work
nce Report

5 Marks 5 Marks 5 Marks 5 Marks 5 Marks

EE-330 Digital Signal Processing Page 1

 Lab7: IIR Filter Design using Analog Prototypes

Objectives
The objective is to design digital filters with an infinite impulse response (IIR). IIR filters can be designed
either directly, by means of a clever placement of poles and zeros in the z-plane, or “indirectly” by first
designing a corresponding analog filter, which is then transformed to the digital domain. In this lab we will
only consider the later design procedure. Note that the described principle of using analog prototypes for IIR
filter design originally stems from the idea to use digital filters as a substitute for analog systems. This allows
us to use the well-constructed theory of analog filter design and simply transform it to the digital domain.

 Familiarization with IIR filter design procedure.

 Familiarization with frequency transformation.
 Bilinear frequency transformation
 Impulse invariance frequency transformation

Lab Instructions
 This lab activity comprises of three parts: Pre-lab, Lab Exercises, and Post-Lab Viva session.
 The lab report shall be uploaded on LMS.
 Only those tasks that completed during the allocated lab time will be credited to the students.
Students are however encouraged to practice on their own in spare time for enhancing their
skills.
Lab Report Instructions
All questions should be answered precisely to get maximum credit. Lab report must ensure following
items:
 Lab objectives
 MATLAB/C codes
 Results (graphs/tables) duly commented and discussed
 Conclusion

EE-330 Digital Signal Processing Page 2

1 IIR Filter Design using Analog Prototypes
1.1 Introduction
IIR filters can be designed either directly, by means of a clever placement of poles and zeros in the z-
plane, or “indirectly” by first designing a corresponding analog filter, which is then transformed to the
digital domain. Within this practical we will only consider the latter design procedure. Note that the
described principle of using analog prototypes for IIR filter design originally stems from the idea to use
digital filters as a substitute for analog systems.This allows us to use the well constructed theory of analog
filter design and simply transfer it to the digital domain.

1.2 Pre lab:

1.2.1 IIR filter design problem
Ideally, we are interested in a digital filter H(ejω) that simulates a continuous time system Hc(j), i.e. the
following identity is fullfilled

Here ω denotes the dimensionless discrete-time frequency in radians per sample, and the analog
frequency in radians per second. Td is a design parameter for the analog to digital conversion, and does
not need to be the same as T = 1/fs, where fs is the sampling frequency.

1.2.2 Design procedure

Given a digital filter specifications, the design of digital IIR filters can be summarized by a three-step
procedure:

1. Translate the discrete-time filter specifications to the analog design domain,so that after the continuous
to discrete-time transform is applied, they will be proper.

2. Select an analog prototype (Butterworth, Chebyshev, etc.) and design your analog filter Hc(s) using the
translated specifications. Analog filter design it treated in next section .

3. Apply the continuous to discrete-time transform, and convert the analog transfer function Hc(s) into a
discrete-time transfer function H(z). Details on the continuous to discrete-time conversion are given in
Section 3. So before designing an appropriate analog filter, we have to translate the discrete-time filter
specifications, which may be given either in real frequency f in Hz or normalized frequency ω = 2πf/fs in
radians per sample, to analog design frequency . The translation depends on which continuous to discrete
time conversion scheme is used. An overview for the frequency translation is depicted in Figure 1. The
digital IIR filter design steps, described above can be done automatically in the MATLAB design
programs. However, to understand the procedure we will go through the individual steps in one of the
experiments.

EE-330 Digital Signal Processing Page 3

1.2.3 Analog filter design
We briefly describe methods for analog lowpass filter design and then show how to easily obtain
highpass, bandpass or bandstop from analog lowpass prototypes by using frequency transformations.
(1) Bilinear transformation
(2) Impulse invariance method.

Figure 1: Translation of filter specifications (f or ω) to analog frequency Ω.

1.2.3.1 Analog lowpass filter design

There are four established techniques for the design of analog filters, including Butterworth, Chebyshev
Type I and II, and Elliptic analog prototypes. They represent four different combinations of two error
approximation measures. One is based on Taylor series, and results in a maximally flat frequency
response. The other approximation minimizes the maximum difference between the desired and actual
response, and results in an equiripple frequency response. The four lowpass prototypes are:

1. The analog Butterworth filter is based on a Taylor series with expansions at Ω = 0 and Ω= ∞, with
desired values 1 and 0 respectively, and first up to (N-1)-th derivative equal to zero. This results in a
smooth transition between at Ω = 0 and Ω= ∞. The squared magnitude of the normalized frequency
response is given by

where N is the filter order and Ωc is 3 dB cutoff frequency. It can be seen that |Hc(jΩ)|2 = 0.5 at Ω= Ωc.

2. The analog Chebyshev Type I filter has a rippled passband with equally sized ripples and a smooth
behavior in the stopband, as can be seen in Figure 2. The squared magnitude of the normalized frequency
response

EE-330 Digital Signal Processing Page 4

Figure 2: Magnitude specification for a Chebyshev Type I lowpass filter of a Chebyshev Type I filter is
given by

where parameter ε determines the passband ripple amplitude, given by , and VN (x) is the
Chebyshev polynomial of order N, defined recursively as

3. The analog Chebyshev Type II is the reverse counterpart of the Type I, i.e. it has a smooth passband
and a rippled stopband. Here the design parameter specifies the stopband ripple amplitude.

4. The analog Elliptic filter uses a Chebyshev approximation in both the passband and the stopband. Two
design parameters are used to specify the passband and stopband ripple amplitude. Corresponding
MATLAB functions for analog prototype lowpass filter design are buttap, cheb1ap, cheb2ap and ellipap.
They provide a zero pole representation, which can be easily converted to differential coefficient
parameters by zp2tf.

1.2.3.2 Frequency transformation

Sometimes it is simpler to calculate filter coefficients of an analog lowpass prototype only once, or to use
tables. This prototype is then converted by means of frequency transforms, according to the desired
characteristics. In the following, let Hc(s) be an analog lowpass prototype of order N with normalized
band edge Ωc = 1.
• For instance, a lowpass with an arbitrary band edge c can be obtained from the analog prototype Hc(s)
by simply replacing
• Using this principle, we can also convert Hc(s) to a highpass filter with band edge Ωc,by changing
this essentially replaces Ω = 0 and Ω= ∞.
• A bandpass filter with lower and upper band edge Ω1 and Ω2, can be obtained from the lowpass
prototype Hc(s) by changing

Where BP is the geometric mean of the two band edges. This mapping doubles the order of the
filter and is a one-to-one mapping of the frequency domain.In MATLAB you can use the functions lp2lp,
lp2hp, lp2bp and lp2bs for the analog frequency transformations. For more detail we refer to the
corresponding help files.

1.2.4 Continuous to discrete-time conversion

Two methods for discrete-time conversion are considered: the impulse invariance and the bilinear
transformation.

1.2.4.1 Impulse invariance method

The impulse invariance method is based on an intuitive concept. We directly transfer the discrete
specifications into the analog domain, using Ω= ω/Td, and design a corresponding continuous-time filter
Hc(j). We then determine the impulse response of Hc(jΩ), denoted hc(t) and obtain the impulse response
of our discrete-time filter by sampling hc(t), i.e.

EE-330 Digital Signal Processing Page 5

Here Td is the design sampling interval. Due to the fact that the FT of hc(nTd) is a periodic extension of
Hc(jΩ), which is not perfectly bandlimited in practice, aliasing may occur so that (1) does not hold
exactly. The continuous to discrete-time conversion by means of the impulse invariance can be done in
MATLAB using the function impinvar.

1.2.4.2 The bilinear transformation

The bilinear transformation is a one-to-one mapping of the entire analog frequency domain −1 < < 1 onto
the discrete-time frequency interval −π < ω < π. Using complex Laplace and Z-transform variables s and
z, the mapping is given by

In the discrete-time domain we are interested in z variables on the unit circle whereas in
the continuous-time domain, s variables on the imaginary axis s = jΩ are of importance.
By evaluating (3) at and s = jΩ, we obtain the relation between analog and digital
frequency =

We notice that ω = 0 implies Ω= 0 and for increasing ω the continuous frequency also increases and we
obtain . Hence, positive continuous-time frequencies are mapped to the upper half of the
unit circle and negative frequencies to the lower half. Two important properties are:

• The bilinear transform preserves stability. A stable continuous-time system Hc(s) with poles on the left
half of the s-plane, is mapped to a stable discrete-time system H(z) with poles inside the unit circle.
• Identity (1) holds exactly at Ω = 2/Td·tan(ω/2) when using the bilinear transform. Finally, we can
perform the bilinear transformation and convert Hc(s) to H(z) using the relation

In MATLAB, the corresponding function bilinear does this job.

1.3 Lab Tasks

1.3.1 Bilinear transformation and impulse invariance
The N-th order analog Butterworth lowpass filter with cutoff frequency Ωc is given by

1. Design a Butterworth filter with N = 4, ωc = 0.3π and Td = 2 using the impulse invariance method.
Proceed as follows:
(a) Determine Ωc and use (5) to calculate the analog filtering coefficients
(b) In the analog domain, plot hc(t) and |Hc(jΩ)|. For the analog domain plots in MATLAB use the
commands tf, impulse and freqs.
(c) Transform your filter coefficients from the analog to the discrete time domain.
(d) In the discrete-time domain, plot h(n) and |H(ejω)| . Compare your results with (b).
2. Repeat 1. but with the bilinear transformation.
3. Vary the design parameter Td. What is its influence? Further, compare both methods: what are their
strengths and weaknesses?

EE-330 Digital Signal Processing Page 6

1.3.2 Lab Task 2:

1.3.2.1 Characteristics of IIR filters

For simplicity, we now use MATLAB functions which automatically perform all digital IIR filter design
steps, mentioned in the introduction. The commands are e.g. butter, cheby1 or ellip. Use the help
command for more details. Say the filtering specifications for a particular job are:

Sampling frequency 200 Hz

Passband ripple _ 1 dB
Passband edge 32 Hz
Stopband edge 38 Hz
Stopband attenuation > 25 dB

1. Determine the order of a Butterworth, Chebyshev Type I and Elliptic filters that meet these
specifications? Use the buttord, cheb1ord and ellipord commands.
2. Use butter, cheby1 and ellip to design the specified filters. For each, plot the magnitude of the
frequency response, the pole-zero location diagram and the significant part of the impulse response.

EE-330 Digital Signal Processing Page 7