Infinite Impulse Response

Infinite Impulse Response (IIR) Filters
3.1. Introduction
3.2. IIR filter design
3.3. Reference analog prototype filter
3.4. Analog prototype filter to analog filter transformation
3.5. Bilinear transformation
3.6. Examples
3.1 Introduction
IIR filters are digital filters with infinite impulse response. Unlike FIR filters, they have the
feedback (a recursive part of a filter) and are known as recursive digital filters therefore.
Figure 3-1-1. Block diagrams of FIR and IIR filters

For this reason IIR filters have much better frequency response than FIR filters of the same
order. Unlike FIR filters, their phase characteristic is not linear which can cause a problem to the
systems which need phase linearity. For this reason, it is not preferable to use IIR filters in digital
signal processing when the phase is of the essence.
Otherwise, when the linear phase characteristic is not important, the use of IIR filters is an
excellent solution.
There is one problem known as a potential instability that is typical of IIR filters only. FIR filters
do not have such a problem as they do not have the feedback. For this reason, it is always
necessary to check after the design process whether the resulting IIR filter is stable or not.
IIR filters can be designed using different methods. One of the most commonly used is via the
reference analog prototype filter. This method is the best for designing all standard types of
filters such as low-pass, high-pass, band-pass and band-stop filters.
This book describes the design method using reference analog prototype filter. Figure 3-1-2
illustrates the block diagram of this method.
Figure 3-1-2. Block diagram of design method using reference analog prototype filter
FIR filters can have linear phase characteristic, which is not typical of IIR filters. When it is
necessary to have linear phase characteristic, FIR filters are the only available solution. In other
cases when linear phase characteristic is not necessary, such as speech signal processing, FIR
filters are not good solution. IIR filters should be used instead. The resulting filter order is
considerably lower for the same frequency response.
The filter order determines the number of filter delay lines, i.e. number of input and output
samples that should be saved in order that the next output sample can be computed. For instance,
if the filter order is 10, it means that it is necessary to save 10 input samples plus 10 output
samples preceeding the current sample. All these 21 samples will affect the next output sample.
The IIR filter transfer function is a ratio of two polynomials of complex variable z-1. The
numerator defines location of zeros, whereas the denominator defines location of poles of the
resulting IIR filter transfer function.
Figure 3-1-3. illustrates input and output signals of the system with non-linear phase
characteristic.
Figure 3-1-3. The effect of non-linear phase characteristic

The system introduces phase shift of 0 radians at frequency of , and radians at three times
higher frequency. Input signal consists of nature frequency and harmonics with the same
amplitude at three times higher frequency. Figure on the left illustrates an input signal, whereas
Figure on the right illustrates an output signal. As seen, these two signals have different
waveforms. Neither the power of the signal nor amplitudes of particular harmonics have been
changed, but the phase of the second harmonic.
Assume that an input represents a speech signal where the phase is not important. In this case
such phase distorsion would be negligable as the system satisfies the stated requirements.
Otherwise, if the phase is important, such a huge distorsion mustnt be allowed.
3.2 Infinite impulse response (IIR) filter design

The most commonly used IIR filter design method uses reference analog prototype filter. It is the
best method to use when designing standard filters such as low-pass, high-pass, bandpass and
band-stop filters.
The filter design process starts with specifications and requirements of the desirable IIR filter. A
type of reference analog prototype filter to be used is specified according to the specifications
and after that everything is ready for analog prototype filter design.
The next step in the design process is scaling of the frequency range of analog prototype filter
into desirable frequency range. This is how an analog prototype filter is converted into an analog
filter.
After the analog filter is designed, it is time to go through the last step in the digital IIR filter
design process. It is conversion from analog to digital filter. The most popular and most
commonly used converting method is bilinear transformation method. The resulting filter,
obtained in this way, is always stable. However, instability of the resulting filter, when bilinear
transformation is used, may be caused only by the finite word-length side-effect.
3.2.1 Basic concepts and IIR filter specification
First of all, it is necessay to learn the basic concepts that will be used further in this book. You
should be aware that without being familiar with these concepts, it is not possible to understand
analyses and synthesis of digital filters.
Figure 3-2-1 illustrates a low-pass digital filter specification. The word specification refers to the
frequency response specification.
Figure 3-2-1a. Low-pass digital filter specification
Figure 3-2-1b. Low-pass digital filter specification
p normalized passband cut-off frequency;

s normalized stopband cut-off frequency;
1 maximum passband ripples;
2 minimum stopband attentuation;
passband attenuation parameter;
A stopband attenuation parameter;

ap maximum passband ripples [dB]; and
as minimum stopband attenuation [dB].
Frequency normalization can be expressed as follows:
where:
fs is the sampling frequency;

f is the frequency to normalize; and
is the normalized frequency.
Specifications for high-pass, band-pass and band-stop filters are defined almost the same way as
those for low-pass filters. Figure 3-2-2 illustrates a high-pass filter specification, whereas Figure
3-2-3 illustrates a band-pass filter specification.
Figure 3-2-2a. High-pass digital filter specification
Figure 3-2-2b. High-pass digital filter specification

Comparing these two Figures 3-2-1 and 3-2-2, it is obvious that low-pass and high-pass filters
have similar specifications. The same parameters are defined in both cases with the difference
that in the later case the passband is substituded by the stopband and vice versa.
Figure 3-2-3 illustrates a band-pass specification.
Filter 3-2-3a. Band-pass digital filter specification
Figure 3-2-3b. Band-pass digital filter specification

Figure 3-2-4 illustrates band-stop digital filter specification
Figure 3-2-4a. Band-stop digital filter specification
Figure 3-2-4b. Band-stop digital filter specification

3.2.2 Z-transform
The Z-transform is performed upon discrete-time signals. It converts a discrete timedomain
signal into a complex frequency-domin representation. It is very suitable for analyzing discrete
time-domain signals and systems as well. The z-transform is derived from the Fourier discrete
time-domain transformation and is considered the basic operation in digital filter design process.
The Z-transform is defined as:
where z is the complex number.

Example:
Assume that samples of a discrete-time signal x(n) are known. It is necessary to transform this
signal with the z-transform and Fourier fransform.
x(n)={1,2,3,4,5,4,3,2,1} ; 0 n 8
z-transform is defined via expression:
It becomes:
The last expression is the z-transform of the given signal.

The Fourier transformation can be found by rewriting the previous expression in terms of z as
z=e^j. It becomes:
Figure 3-2-5 illustrates the (frequency) spectrum of the given signal.
Figure 3-2-5. Frequency spectrum of the given signal

Comparing Z and Fourier transforms, it is easy to notice some similarities between them:
In polar coordinates, the complex number z may be expressed as follows:
The two last expressions lead us to the conclusion that Fourier transform is just a special form of
the z-transform for r=1.
In the z plane, the Fourier transform is represented as a unit circle, which can be seen in Figure
3-2-6 below.
Figure 3-2-6. Fourier transform in the z plane

The z-transform of the transfer function is of great importance for IIR filters. The location of
poles in the z plane is used for testing stability of designed IIR filter. The poles of the IIR filter
transfer function must be located within the unit circle in order that filter is stable.
Figure 3-2-7a illustrates zeros and poles of the transfer function of a stable IIR filter in the z
plane.
Figure 3-2-7a Stable IIR filter

Transfer function zeros are denoted by small circles, whereas its poles are denoted by small
crosses.
Re Real axis
Im Imaginary axis
As seen in Figure 3-2-7, one transfer function zero is located outside the unit circle. It doesnt
cause any problem as the location of poles is the only thing that matters. All four poles of
transfer function are located within the unit circle, which guarantees tha stability of IIR filters.
According to the location of poles in the z plane, it is easy to determine whether it refers to FIR
or IIR filter. The poles of the FIR filter transfer function are located at the origin. It is obviously
not the case in Figure 3-2-7, which means that it refers to IIR, not FIR filter.
Also, Figure 3-2-7a clearly indicates interrelation between zeros and poles in the z plane. If a
zero or a pole is located on the real axe in the z plane, i.e. the imaginary part is zero, then it is
single. If either of them is not located on the real axis in the z plane, then it has the corresponding
pair having the same real value and the same imaginary value with the opposite sign. In the z
plane, it is illustrated as a pair of zeros or poles which are symmetric around the real axis. Such a
pair is also called a complex-conjugated pair of zeros or poles.
Figure 3-2-7b illustrates the zeros and poles of the transfer function of an instable IIR filter in the
z plane.
Figure 3-2-7b. Instable IIR filter

As seen from Figure 3-2-7b, two poles located outside the unit circle make this IIR filter instable.
If bilinear transformation is used in the filter design, the resulting filter is stable before the
coefficient quantization starts. This quantization changes the location of zeros and poles of the
resulting IIR filter, which can cause one pole or one pair of poles to be located outside the unit
circle. The result of such a quantization is a filter that is not stable.
3.2.3 Transfer function of discrete-time systems
The Z-transform is primarily used for finding the transfer function of linear discrete-time
systems. When the transfer function is found, it is necessary to consider the zeros and poles of
the transfer function in the z plane. The transfer function of discrete-time systems is defined to
be:
where:
bi are the feedforward filter coefficients (non-recursive part);

aj are the feedback filter coefficients (recursive part);
H0 is a constant;
qi are the zeros of the transfer function;
pj are the poles of the transfer function;
B(z) is the transfer function of non-recursive part of the system; and
A(z) the transfer function of recursive part of the system (feedback).
The recursive part of the transfer function is actually a discrete-time system feedback. Unlike the
FIR filters, the IIR filters have feedback which enable them to have greater selectivity as well as
nonlinearity of phase characteristic than FIR filters.
Figure 3-2-8. illustrates block diagram of discrete-time system with feedback.
Figure 3-2-8 Discrete-time system with feedback

In the time domain, the discrete-time system shown in Figure 3-2-8 can be expressed as follows:
OR
The later expression is more convenient for software IIR filters realization.
In the frequency domain, the discrete-time system shown in Figure 3-2-8 can be expressed as the
multiplication of Z-transform input signal X(z) and the transform function H(z):
The first way of representing discrete-time systems is suitable for both software and hardware
IIR filter implementation, whereas the representation in the z domain is suitable for analyzes of
designed filters and synthesis itself (design process).
Example:
The transfer function of a 3th order IIR filter, designed using Chebyshev function is:
The following expression describes the filtering process:
This process is also known as convolution. Another expression for convolusion that is more
useful in practical applications is:
After making substitutions of impulse response coefficients, it becomes:
Using expression:
it is possible to find function for particular normalized frequency.

For example, when = 0.2:
The numerator is computed first:
Then denominator:
Figure 2-2-8 illustrates a hardware realization of this IIR filter.
Figure 3-2-9. Realization of IIR filter in this example

The software realization would require two buffers each of minimum length 3. One buffer would
be used for input samples and another one for output samples. These are usually circular buffers
whose length can be expressed as 2^n, which in this case means that the circular buffer is 4 = 2^2
in length.
By complexity, the given IIR filter corresponds to a 6th order FIR filter. Selectivity and
attenuation of this filter are much higher than those of any 6th order FIR filter. The result of the
feedback, which provides so high selectivity and attenuation, is a non-linear phase characteristic.
3.2.4 Effects of the poles and zeros of the transfer function
The location of poles and zeros of the transfer function is very important for discrete-time system
analyses and synthesis. According to their location it is possible to test stability of a discrete-time
system, detect round-off errors made due to software implementation of a filter as well as
coefficient errors encountered during hardware implementation of a filter.
In order that a discrete-time system is stable, all poles of the discrete-time system transfer
function must be located within the unit circle, as shown in Figure 3-2-6. If this requirement is
not satisfied, the system becomes unstable, which is very dangerous. The location of zeroes
doesnt affect the stabilty of discrete-time systems. Recalling that FIR flters do not have a
feedback, which makes them stable. However, this doesnt apply on IIR filters. Therefore, it is
preferable to use bilinear transformation because it always makes filter stable. In this case, filter
stability is questioned only due to coefficient quantization which is performed at the end of the
design process.
It always happens due to software and hardware implementation that an error in coefficients
representation is produced. In software implementation, an error is triggered by the finite wordlength effect, whereas in hardware implementation, it ocurrs due to impossibility of representing
the coefficients with apsolute accuracy. The result in both cases is that the actual value of
coefficients differs from their value obtained in design process. A direct result of such errors is
deviation of the frequency of designed discrete-time system.
Deviation of frequency depends on the spacing between the zeros and poles of the FIR filter
transfer function and the origin in the z plane. The FIR filter coefficient error affects more the
frequency response as the spacing between the zero and pole of the transfer function and the
origin narrows. This property is particularly typical of high-order filters because their zeros are
very close each other. Besides, the pole quantization, by rule, affects more frequency
characteristic. Slight errors in coefficient representation may cause large frequency deviations.
Figure 3-2-9 illustrates the required and obtained frequency characteristic of an IIR filter. The
finite word-length effect on the transform function of an IIR filter is clearly marked in this
figure.
Figure 3-2-10. Deviation from required frequency characteristic

The frequency deviation shown in Figure 3-2-10 is basically slight deviation, even though it is
very large at certain frequencies. The minimum attenuation and the width of transition region of
the resulting IIR filter remain unchanged, so that such deviation is acceptable.
3.2.5 IIR filters design using bilinear transformation

The IIR filter design using bilinear transformation can be split into several steps:
1. Defining filter specification;
2. Specifying analog prototype filter;
3. Computing the filter order required for a given set of specifications and specified analog
prototype filter;
4. Computing the transfer function of reference analog prototype filter;
5. Conversion into analog filter via scaling;
6. Conversion into digital filter via bilinear transformation; and
7. If the obtained filter doesnt satisfy the given specifications or if it is possible to decrease
the filter order, then it is necessary to do it. The filter order can be increased or decreased
according to needs and after that steps 4, 5 and 6 are repeated as many times as needed.
The final objective of defining IIR filter specifications is to find the desirable normalized cutoff
frequencies (c, c1, c2), transition width, maximum passband attenuation and minimum
stopband attenuation. The type of analog prototype filter as well as the filter order will be
specified according to these parameters.
Now, it is time to specify the type of reference analog prototype filter. Be aware that every type
has its good and bad sides. It is only important that its characteristics can satisfy the given
specifications. However, it is preferable to specify such a type of analog prototype filter that can
produce the lowest order IIR filter.
After this step, that is, when the type of analog proptotype filter is known, it is necessary to
specify or compute the filter order required for a given set of specifications. The initial value of
the filter order is roughly estimated and is changed after that depending on the obtained
characteristics and requirements.
When both type and order of analog prototype filter are known, it is possible to find its transfer
function.
The transfer function of analog prototype filter depends on frequencies which are not scaled into
the desirable range. For this reason, it is necessary to perform scaling of the transfer function so
that cut-off frequencies go into the desirable range. This operation is actually conversion of
reference analog prototype filter into analog filter with desirable characteristic.
Finally, the transfer function of the specified type of reference analog prototype filter is obtained
by converting analog filter into digital one. This book represents the most commonly used
conversion known as bilinear transformation.
If the resulting filter doesnt satisfy the given specifications, or if the filter order can be less than
actual one, it should be changed. IIR filters have much greater selectivity and attenuation than
FIR filters of the same order. For this reason, it is preferable to increase or decrease the filter
order by 1. After changing the filter order, the entire IIR filter design process, i.e. computing of
the transfer function of reference analog prototype filter, scaling and obtaining analogue filters
and conversion into digital filter, is repeated.
3.2.6 IIR filter realization
FIR filter transfer function can be expressed as:
where:
N is the filter order;

bk the coefficient of non-recursive part of IIR filter; and
ak the coefficient of recursive part (feedback) of IIR filter.
The coefficients bk and ak are of interest for IIR filter realization (both hardware and software).
Figure 3-2-11 illustrates the block diagram of IIR filter.
Figure 3-2-11. Block diagram of IIR filter

There are several types of IIR filter realization. This chapter covers direct, direct transpose,
direct canonic, direct transpose canonic and cascade realizations. All of them are very convenient
and most commonly used for both hardware and software IIR filter realization. Each of them will
be described in detail along with their advantages and disadvantages.
3.2.6.1 Direct realization
Direct realization of IIR filters starts with this expression:
The first part of the expression refers to non-recursive part and the other refers to recursive part
of IIR filter. In IIR filter direct realization, these two parts are separately considered and realized.
The realization of non-recursive part of IIR filter is identical to the direct realization of FIR filter.
Figure 3-2-12. illustrates the block diagram of direct realization of non-recursive part of IIR
filter.
Figure 3-2-12. Direct realization of non-recursive part of IIR filter

As seen from Figure 3-2-12 above, multiplication coefficients are identical to those of the
transfer function.
Realization of non-recursive part of IIR filter is similar to that of recursive part. Figure 3-2-13.
illustrates the direct realization of the filter recursive part.
Figure 3-2-13. Direct realization of non-recursive part of IIR filter

As non-recursive and recursive part of IIR filter are separately realized, it doesnt matter which
of them will be used first in filtering process. Figures 3-2-14a and 3-2-14b illustrate block
diagrams of IIR filter realization when non-recursive part is used before and after recursive part
of IIR filter, respectively.
Figure 3-2-14a. IIR filter direct realization, non-recursive part is used first
Figure 3-2-14b. IIR filter direct realization, recursive part is used first
This structure is also known as a direct form I structure. As seen from Figures 3-2-14a and 3-214b, direct realization requires in total of 2N delay lines, (2N+1) multiplications and 2N
additions.
Direct realization is very convenient for software implementation and this is where it is most
commonly used. Some of disadvantages of this realization are the greatest sensitivity to accuracy
of realized coefficients (i.e. the largest finite word-length effect), and the greatest complexity due
to implementation (i.e. needs most resources).
On IIR filter software implementation with direct structure, it is necessary to have two buffers
with at least N+1 samples, where N is the IIR filter order. For their simplicity and effectiveness,
most commonly used are the so called circular buffers the length of which can be expressed as
2^k. The value of constant k is defined as a minimum value for which N 2^k is valid.
Accordingly:
where the operator
represents rounding down to a less value.
Figure 3-2-15. Circular buffer of length 16 = 2^4

Since the buffer length is 16, location addressing in the circular buffer is performed via module
16 operations:
Example:
A 6th order FIR filter is used in this example. It is necessary to design this filter using direct
structure with circular buffer. The length of the buffer needs to be 2^k.
The length of circular buffer is obtained from the following expression:
It means that the minimum length of circular buffer is 2^3 = 8.

The contents of the buffer after receiving the first 10 samples is shown in the table 3-2-1. Input
samples are denoted by x[n] and each shaded cell represents changed location in buffer.
Step
0
1
2
3
4
5
6
7
8
9
10
addr. addr. addr. addr. addr. addr. addr. addr.

7
6
5
4
3
2
1
0
x[7]
x[7]
x[7]
x[6]
x[6]
x[6]
x[6]
x[5]
x[5]
x[5]
x[5]
x[5]
x[4]
x[4]
x[4]
x[4]
x[4]
x[4]
x[3]
x[3]
x[3]
x[3]
x[3]
x[3]
x[3]
x[2]
x[2]
x[2]
x[2]
x[2]
x[2]
x[2]
x[2]
x[1]
x[1]
x[1]
x[1]
x[1]
x[1]
x[1]
x[1]
x[9]
x[0]
x[0]
x[0]
x[0]
x[0]
x[0]
x[0]
x[0]
x[8]
x[8]
Table 2-2-2. Input circular buffer after receiving 10 samples

3.2.6.2 Direct transpose realization
Direct transpose realization is similar to direct realization. The only difference is in the position
of delay lines, i.e. buffer if it is about software implementation. Here, it is also necessary to have
two buffers of minimum length N+1, where N is the filter order.
Figures 3-2-16 and 3-2-17 illustrate the block diagram describing IIR filter direct transpose
realization structure of IIR filter.
Figure 3-2-16. IIR filter direct transpose realization, non-recursive part is used first
Figure 3-2-17. IIR filter direct transpose realization, recursive part is used first
There are no significant differences between direct and direct transpose realizations. Both
structures have the same multiplication coefficients. The only difference is in the position of
delay lines. Similar to direct realization structure, the direct transpose realization structure uses
2N delay lines, (2N+1) multiplications and 2N additions.
3.2.6.3 Direct Canonical Realization
Direct canonical realization structure has reduced number of delay lines to the minimum, that is,
N delay lines. This way, one of the main disadvantages of direct and direct transpose realization
structures is eliminated. Recursive and non-recursive parts of IIR filter are not considered
separately, which causes implementation to be more complex than for direct realization structure.
A good thing is that the coefficients are the same as for direct realization.
Figure 3-2-18 illustrates the block diagram describing direct canonic realization structure of IIR
filter.
Figure 3-2-18. Direct canonic realization structure block diagram
Similarities between direct canonic structure block diagram and direct realization structure
shown in Figure 3-2-14b are obvious. The difference between realization structures shown in
Figures 3-2-14b and 3-2-18 is that non-recursive and recursive part for direct canonic realization
structure cannot be treated separately, although it is easy to differentiate between them.
Direct canonic structure uses N delay elements, (2N+1) multipilications and 2N additions.
Sensitivity to the accuracy of coefficients is the same as for all previously described structures,
which is the main disadvantage of this realization structure.
3.2.6.4 Direct transpose canonical realization
Direct transpose canonical realization structure has reduced number of delay lines to the
minimum of N delay lines as well as reduced number of adders to N+1. Recursive and
nonrecursive parts of IIR filter are not considered separately, which causes implementation to be
more complex than for direct realization structure, but similar to direct canonical structure. A
good thing is that the coefficients are the same as for direct realization.
Figure 3-2-19 illustrates the block diagram describing direct transpose canonical realization
structure of IIR filter.
Figure 3-2-19. Direct transpose canonic realization structure block diagram

Similarities between direct transpose canonical structure block diagram and direct transpose
realization structure shown in Figure 3-2-16 are obvious. The difference between realization
structures shown in Figures 3-2-16 and 3-2-19 is that non-recursive and recursive part for direct
transpose canonical realization structure cannot be treated separately, although it is easy to
differentiate between them.
Direct transpose canonic structure uses N delay elements, (2N+1) multipilication elements and
N+1 adders. Sensitivity to the accuracy of coefficients is the same as for all previously described
structures, which is the main disadvantage of this realization structure.
3.2.6.5 Cascade Realization
Cascade realization structure is the most difficult to obtain from the transfer function (comparing
to other realization structures given in this book). It is very convenient for its modular structure
and less sensitivity to the accuracy of non-recursive and recursive coefficients realization. On
cascade IIR filter realization, a filter is divided into several, mutually independent sections of the
first or second order.
Individual sections are mostly realized in direct canonical or direct transpose canonical structure.
Since the sections are mutually independent after design process, the finite word-length effect on
the accuracy of coefficients, modulation of frequency response and IIR filter stability are
separately examined for each section. The analyse is simplified this way.
The IIR filter transfer function is expressed as:
where:
bi are the coefficients of transfer function numerator (non-recursive part);

aj are the coefficients of transfer function denominator (recursive part);
H0 is a constant;
qi are the zeros of the transfer function;
pj are the poles of the transfer function;
B(z) is the transfer function of non-recursive part;
A(z) is the transfer function of recursive part (feedback); and
M is the number of sections in cascade realization structure.
Cascade realization requires the given expression to be factorized so that the transfer function is
expressed as follows:
where:
a[i, k] are the coefficients of recursive part of the i-th IIR filter section;
b[i, k] are the coefficients of non-recursive part of the i-th IIR filter section.
Individual sections are of the first or second order. Direct transpose canonical structure is most
frequently used in realization. Figure 3-2-20 illustrates a first-order section.
Figure 3-2-20. First-order section

Figure 3-2-21 illustrates a second-order section.
Figure 3-2-21. Second-order section

The use of direct transpose realization structure reduces necessary number of delay lines and
adders as well. Filter dividing in independent sections reduces the sensitivity to the accuracy of
quantization coefficients and simplifies analysing the stability of the resulting filter. Besides, the
possibility that IIR filter becomes instable after quantization is drastically reduced as the
coefficients quantization is performed after dividing filter in sections, so the changes of poles
locations are smaller, therefore.
Software realization requires M buffer of length 2 or 1. Each section must have its own buffer for
saving samples of intermediate signals. Such complexity and needed factorization are two main
disadvantages of this realization structure.
Figure 3-2-21 illustrates the block diagram describing cascade IIR filter structure.
Figure 3-2-22. Cascade IIR filter structure
3.3 Reference Analog Prototype Filter

IIR filter design process starts with reference analog prototype filter. This book explains
Butterworth, Chebyshev (Chebyshev I) and inverse Chebyshev filter (Chebyshev II).
The transform function of analog filter Hsa(s) is expressed as:
where:
s is the complex frequency (s = + j); and
M N.
In order that a system described via expression above is stable, it is necessary that all poles (the
square roots of polinomial Aa(s)) are located in the left half of S plane. Figure 3-3-1 illustrates S
plane.
Figure 3-3-1. S plane and region of stability

A low-pass filter is used for analog filter design. The conversion into the appropriate type of
filter (high-pass, band-pass or band-stop) is performed by converting into analog filter, i.e.
frequency axis scaling.
3.3.1 Butterworth analog filter
Low-pass Butterworth analog filters are filters whose frequency response is a monotonious
descending function. They are also known as maximally flat magnitude filters at the frequency
of = 0, as the first 2N - 1 derivatives of the transfer function when = 0 are equal to zero.
Butterworth filter is characterized by 3dB attenuation at the frequency of =1, no matter the
filter order is. Figure 3-3-2 illustrates frequency responses for a few various parameters N (filter
order).
Figure 3-3-2. Frequency response of Butterworth filter

Butterworth filter is defined via expression:
where:
is the frequency; and
N is the filter order.
Figure 3-3-3. IIR filter specification
Figure 3-3-3 illustrates IIR filter specification with parameters of most interest for Butterworth
filter.
To design Butterworth reference analog prototype filter, it is necessary to know the filter order.
All poles of the resulting filter must be located in the left half of the S plane, i.e. to the left of the
imaginary axis.
Note:
In the z plane, filter is stable if all poles are located within the unit circle. In the s plane, filter is
stable if all poles are located in the left half of the s plane. Z-transform is used for digital filters,
whereas Laplase transform (s plane) is used for analog filters. Even though these two types of
transformations are similar to some extent, they should not be mixed up concerning the filter
stability analyse.
When the filter order is known, it is easy to find its poles using expression:
Butterworth poles are equally allocated (equidistantly) on the unit circle within the left half of
the s plane. The location of poles for N=5 and N=6 is shown in Figure 3-3-4.
Figure 3-3-4. Position of Butterworth filter poles for N=5 and N=6
The transfer function of the Butterworth reference analog prototype filter is expressed as follows:
where:
Sk is the k-th pole of the Butterworth filter transfer function
For N=5, the transfer function is:
3.3.2 Chebyshev Analog Filter

Chebyshev analog low-pass filter of the first kind is a type of analog filter that has the least
oscillation in frequency response in the entire passband. Therefore it is characterized by equal
ripple in the passband and the stopband frequency response is monotoniously descending
function.
Figure 3-3-5 illustrates frequency response for a 4th order band-stop Chebyshev reference analog
filter.
Figure 3-3-5. Frequency response of Chebyshev analog filter

To design Chebyshev reference analog prototype filter, it is necessary to know the filter order.
Chebyshev analog filter is defined via expression:
where:
is the frequency;
is a parameter used to define maximum oscillations in the passband frequency response; and
TN is the Chebyshev polynomial.
The Chebyshev polynomial TN() can be obtained via recursive relations:
If the filter order is known in advance, neither recursive relations nor expression for the square of
frequency response are necessary. The design process starts from the values of poles of a 1st
order Chebyshev reference analog filter.
The values of poles are expressed as:
where:
si is the i-th transfer function pole of analog prototype filter (complex value);
i is the pole; and
i is the imaginary pole.
where:
N is the filter order; and
i=1, 2, ..., N.
The value of parameter is obtained via expression:
Transfer function is expressed as:
The value of A0 is found via expression:
Figure 3-3-6. Location of the poles of Chebyshev filter for N=5

3.3.3 Inverse Chebyshev Analog Filter
Inverse Chebyshev analog filter is also known as Chebyshev analog filter of the second kind.
The frequency response of this filter monotoniously falls in the passband and transition region.
Similar to Butterworth filter, the frequency response is extremely flat function at the frequency
of = 0, as the first 2N - 1 derivatives of the transfer function for = 0 are equal to zero. In the
stopband, inverse Chebyshev filter has the least oscillation in the frequency response.
Figure 3-3-7 illustrates the frequency response for an inverse Chebyshev reference analog bandstop filter of the 4th order.
Figure 3-3-7. Frequency characteristic of inverse Chebyshev analog filter

To design inverse Chebyshev reference analog pototype filter, it is necessary to know the filter
order.
Inverse Chebyshev analog filter is defined via expression:
where: is the frequency;

is the parameter of maximum oscillation in the passband frequency response; and
TN is the Chebyshev polynomial.
The Chebyshev polynomial TN() can be obtained from recursive relation:
If the filter order is familiar in advance, neither these recursive relations nor expression for the
square of frequency response are necessary. The design process starts from the values of poles of
a 1st order inverse Chebyshev reference analog filter.
The poles of the transfer function of inverse Chebyshev analog filter are considered reciprocal
poles of the transfer function of a 1st order Chebyshev analog filter.
The poles of a 1st order Chebyshev analog filter are expressed as:
where:
si is the i-th pole of the transfer function of analog prototype filter (complex value);
i is the real pole; and
i is the imaginary pole.
where:
N is the filter order; and
i=1, 2, ..., N.
The poles of the transfer function of inverse Chebyshev analog filter are found via expression:
where:
si is the pole of the transfer function of a 1st order Chebyshev analog filter; and
s2i is the pole of the transfer function of inverse Chebyshev analog filter.
Transfer function is expressed as:
The coefficient k in numerator can be only an odd number. Table 3-3-1 provides a few examples
of values of k.
N min max values
5 1 5
1, 3, 5
6 1 5
1, 3, 5
7 1 7
1, 3, 5, 7
8 1 7
1, 3, 5, 7
Table 3-3-1. coefficient k in the transfer function numerator
The values k are found via expression:
The value H0 is found via expression:
Figure 3-3-8. Location of poles and zeros of inverse Chebyshev filter for N=5
As seen from Figure 3-3-8 and expression for k, the zeros of the transfer function are always
complex-conjugated values, which is not the case with the poles of the transfer function.
3.4 Analog prototype filter to analog filter transformation

All analog prototype filters, regardless of their type, have scaled frequency range so that the
passband cut-off frequency amounts to = 1. For this reason, it is necessary to scale filter during
the design process, so that passband and stopband cut-off frequencies have the appropriate,
predetermined values.
Reference analog prototype filter is also a low-pass filter so it requires to be transformed into the
appropriate type of filter, i.e. high-pass, band-pass or band-stop filter, if needed.
Transformation from an analog prototype filter to appropriate analog filter is performed before
transforming it in digital filter.
Frequency scaling depends on the type of analog filter being designed. Scaling is explained for
low-pass, high-pass, band-pass and band-stop filters.
All the results obtained in this chapter are tested in the Filter designer tool program.
3.4.1 Low-pass filter
The transformation from a reference analog prototype filter to a low-pass analog filter is the
simplest type of transformation. Analog prototype filter is a low-pass filter with the cut-off
frequency of p=1. In this case, the transformation comes to a simple frequency scaling. In the
transform function, s\\c is used instead of s, where c is a desirable cut-off frequency in the
passband.
Generally, the transform function of the reference analog prototype filter can be expressed as
follows:
where
H0 is a costant;
zk is the k-th zero of the transfer function of the reference analog prototype filter;
M is a number of zeros of the transfer function of the reference analog prototype filter;
pk is the k-th pole of the transfer function of the reference analog prototype filter; and
N is a number of poles of the transfer function of the reference analog prototype filter and the
filter order as well.
By performing the transformation:
each expression within brackets in the transfer function numerator is transformed into:
and the whole numerator is transformed into:
Each bracket in denominator is transformed similarly:
and the entire denominator is transformed into:
By replacing numerator and denominator by their transformed expressions, the transform

function of the analog filter is obtained:
Example:
The transform function of the Butterworth reference analog prototype filter of the 3rd order is
The transformation in a band-pass analog filter with the cut-off frequency c = 0.2929 in the
passband is obtained via expression:
Example:
The transform function of inverse Chebyshev reference analog prototype filter of the 3rd order is
The transformation to a band-pass analog filter with the cut-off frequency c=0.3719 in the
passband is obtained via expression:
3.4.2 High-pass filter

Analog prototype filter is a low-pass filter with the cut-off frequency p = 1. Its transformation
to a high-pass analog filter can be split into two steps. The first step refers to the transformation
to a high-pass analog filter, whereas the second one refers to frequency scaling. The final
objective is that passband cut-off frequency of the resulting high-pass analog filter amounts to
c.
The transformation to a high-pass analog filter:
Scaling of frequency axis is performed by transformation:
These two transformations can be represented by one transformation:
Generally, the transform function of the reference analog prototype filter can be expressed as
follows:
where:
H0 is a constant;
By performing the following transformation:
each bracket in the numerator of the transfer function is transformed into
and the entire numerator is transformed into:
Substituting these transformed expressions for numerator and denominator, the transform
function of the analog filter is obtained:
Example:
The transfer function of the Chebyshev reference analog prototype filter of the 3rd order is
passband is expressed as:
The result is the transfer function of analog filter:
Example:
The transfer function of inverse Chebyshev reference analog prototype filter of the 2nd order is
The transformation in a band-pass analog filter with the cut-off frequency c=0.1434 in the
The result is the transform function of analog filter:
3.4.3 Band-pass filter
The transformation in a band-pass analog filter is more complex than the transformation in a
low-pass and high-pass analog filters. The filter order is doubled by this transformation. This is
why it is not possible to design an odd order band-pass filter.
When designing, the required filter order is divided by two. The resulting filter order is used to
design a low-pass reference analog prototype filter. By transforming it into a band-pass analog
filter the filter order is doubled. The required filter order is obtained this way.
Example:
Assume that it is necessary to design an 8th order band-pass digital filter.
A low-pass reference analog filter of the 4th order is designed first.
Reference analog filter is further transformed in a bandpass analog filter. This transformation
doubles the filter order. The order of the resulting filter is 8.
The transformation into a band-pass analog filter is expressed as:
The value of the constant 0 can be found via expression:
where:
p1 is a lower cut-off frequency in the passband (refer to figure 3-4-1); and
p2 is a higher cut-off frequency in the passband (refer to figure 3-4-1).
Figure 3-4-1. Band-pass filter specification

The value of c, which is necessary for normalization, is found via expression:
The transfer function of an analog prototype filter is transformed first in a band-pass analog
filter, and normalized after that with c.
Generally, the transfer function of the reference analog prototype filter can be expressed as:
where:
H0 is a constant;
By performing the transformation:
each bracket in numerator of the transfer function is transformed into:
and the entire numerator is transformed into:
The scaling of frequency axis is performed after transformation:
The scaling of frequency axis is performed after transformation:
Substituting the transformed expressions for numerator and denominator, the transform function
of analog filter is obtained:
Example:
The transfer function of the Chebyshev reference analog prototype filter of the 2nd order is
expressed as:
3.4.4 Band-stop filter

The transformation in a band-stop analog filter is similar to the transformation in a bandpass
analog filter. Similarly, the filter order is doubled and the order of a band-stop digital filter
cannot be an odd number.
When designing, the required filter order is divided by two. The resulting filter order is used for
designing a low-pass reference analog prototype filter. The transformation into a bandstop
analog filter causes the filter order to double. This is how the required filter order is obtained.
Example:
It is necessary to design a band-stop digital filter of the 10th order.
A low-pass reference analog filter of the 5th order is designed first.
The reference analog filter is transformed in a band-stop analog filter. This transformation causes
the filter order to double. The result is a 10th order filter.
The transformation in a band-stop analog filter:
The value of the constant 0 can be found via expression:
where:
s1 is a lower cut-off frequency in the stopband (refer to figure 3-4-2); and
s2 is a higher cut-off frequency in the stopband (refer to figure 3-4-2).
Figure 3-4-2. Band-stop filter specification

The value of c, which is necessary for normalization, is found via expression:
The transfer function of the reference analog prototype filter is transformed first into a band-stop
analog filter, and normalized after that with c.
Generally, the transfer function of the reference analog prototype filter can be expressed as
follows:
where:
H0 is a constant;
The first step refers to normalization with the frequency c:
The transformation is performed in the second step:
and each bracket in the transfer function numerator is transformed in:
and the entire numerator is transformed in:
and the entire denominator is transformed in:
Substituting the transformed expressions for numerator and denominator, the transform function
of analog filter is obtained:
Example:
The transfer function of inverse Chebyshev reference analog prototype filter of the 2nd order is
expressed as:
The transformation into a band-stop analog filter with the cut-off frequency c=0.252.5727 in
the passband is expressed as:
3.5 Bilinear transformation

Digital IIR filters are designed using analog filters. After the frequency scaling and
transformation into a desirable type of filter have been performed, it is necessary to transform the
resulting analog filter into a digital one. It is done by transforming the analog filter transfer
function into a digital one.
The transformation is supposed to:
1. faithfully approximate the frequency response of analog filter; and
2. provide that the resulting digital filter is guaranteed to be stable.
This transformation also transforms s plane into z plane.
Analog filter is stable if the poles of the transfer function are located in the left half of s plane,
whereas digital filter is stable if the poles are located within the unit circle. For this reason, the
transformation must provide that the left half of s plane coincides with the area within the unit
circle of z plane, as shown in Figure 3-5-1.
Figure 3-5-1. Transformation of s plane into z plane

One of most commonly used method of transforming analog filters into appropriate IIR filters is
known as bilinear transformation. It is defined via expression:
Using the previous expression, the transformation of the analog filter transform function into a
digital one can be expressed as:
As seen, the transformation is performed by a simple change of variable s in the expression for
the transfer function of the resulting analog filter.
The analog filter transfer function can be expressed as:
where:
H0 is a constant. If s=0 then H(s)=H0 ;
zk is the zero of the analog filter transfer function; and
pk is the pole of the analog filter transfer function.
After transformation, the analog filter transfer function is further transformed into:
where:
Hoz is a constant of the digital IIR filter transfer function
Example:
The transfer function of a second-order high-pass analog filter (inverse Chebyshev, fc=2KHz,
fs=44100Hz, 60dB) is expressed as:
It is necessary to transform the given analog filter into the appropriate digital filter by bilinear
transformation.
Using expression for linear transformation:
we obtain:
where:
N=2
M=2
z1=j0.1014
z2=-j0.1014
p1=-2.267+j2.2692
p2=-2.267-j2.2692
The result is the transfer function of a digital high-pass IIR filter. Realization structure is
illustrated in Figure 3-5-2 below.
Figure 3-5-2. IIR filter realization
Digital filters designed via bilinear trasnformation are guaranteed to be stable. However, the
accurate values of coefficients are obtained immediately after the implementation of bilinear
transformation. On filter realization, it is impossible to represent coefficients without an error. In
software digital filter realization (implementation), the resulting coefficients are quantized,
which also generates a certain error. Any error made during the quantization of coefficients
affects more or less the frequency response, which may further cause the stopband attenuation to
decrease.
The quantization effect on digital filter stability is much more dangerous. Special care is required
when quantizing feedback coefficients as it causes the location of the digital IIR filter transfer
function poles to change their location. It is very important to prevent the poles from being
located outside the unite circle. However, if it happens, the resulting IIR filter is not stable and is
useless therefore.
A disadvantage of the bilinear transformation is a non-linear transformation of the analog filter
frequency axis into a digital one. When designing, the cut-off frequencies are defined on the
basis of the given specifications and type of a filter. When transforming, these frequencies have
appropriate locations, which is not the case with the rest of the frequency axis. Such a non-linear
transformation of analog filter frequencies causes the phase characteristic distorsion, so that it is
not linear.
3.6 Examples
This chapter discusses various IIR filter design methods. The four standard types of filters are
used here:
low-pass filter;
high-pass filter;
band-pass filter; and
band-stop filter.
The design method used here is known as bilinear transformation.

The IIR filter design process can be split into several steps described in Chapter 3.2.5 Designing
IIR filters by bilinear transformation. These are:
1. Defining filter specifications;
2. Specifying the type of analog prototype filter;
3. Computing the filter order according to the filter specifications and specified analog
prototype filter;
4. Computing the transfer function of reference analog prototype filter;
5. Transformation into analog filter by range scaling;
6. Transformation into digital filter by bilinear transformation; and
7. If the resulting filter doesnt satisfy the given specifications or if it is possible to decrease
the filter order, then it is necessary to do it. The specified filter order is increased or
decreased according to needs, and steps 4, 5 and 6 are repeated after that as many times
as needed.
Some steps are skipped in some cases. If the filter order is known, step 3 is skipped. If the type
of reference analog prototype filter is predetermined, step 2 is skipped.
In every given example, the IIR filter design process will be described through these steps in
order to make it easier for you to observe similarities and differencies between various design
methodes, analog prototype filters and design of various types of filters as well.
Figure 3-6-1 illustrates the design steps along with input and output data for each of them.
Figure 3-6-1. Steps in designing digital IIR filter

The first block refers to design of reference analog prototype filter of appropriate order. The
output data is a reference analog prototype filter transfer function Ha(s). Regardless of the type
of reference analog prototype filter in use, the transfer function is given by:
where:
H0 is a constant;
zk is the k-th zero of the reference analog prototype filter transfer function;
M is a number of zeros of the reference analog prototype filter transfer function;
pk is the k-th pole of the reference analog prototype filter transfer function; and
N is a number of poles of the reference analog prototype filter transfer function and filter order
as well.
Reference analog prototype filter is always a low-pass filter. The next step is the transformation
into an analog filter of appropriate type. The expression used to transform the reference analog
prototype filter transfer function depends on the type of filter that needs to be obtained. The final
result is the transfer function H(s) given by:
where:
H0 is a constant;
zk is the k-th zero of the reference analog prototype filter transfer function;
M is a number of zeros of the reference analog prototype filter transfer function;
pk is the k-th pole of the reference analog prototype filter transfer function; and
N is a number of poles of the reference analog prototype filter transfer function and filter order
as well.
As seen, the transfer functions of reference analog prototype filter and analog filter are very
similar. They differ only in the value of constant H0, the values of the transfer function poles and
zeros zk and pk as well as in the number of transfer function zeros M. The filter order is the same
if the analog filter is a low-pass or high-pass filter, whereas it is different if the analog filter is a
band-pass or band-stop filter. In the later case, the analog filter order (N) is twice that of the
reference analog prototype filter.
The next step is the transformation into appropriate digital IIR filter using bilinear transformation
given by expression:
Filter Designer Tool is used for testing and analysing the resulting IIR filters in this chapter. All
data are calculated with the accuracy of 4 decimal digits, which is sufficient for most examples.
3.6.1 Filter design using Butterworth filter
3.6.1.1 Example 1
Step 1:
Type of filter low-pass filter
Filter specifications:
Filter order N=2;

Sampling frequency fs=20KHz;
Passband cut-off frequency fc=2.5KHz; and
Minimum stopband attenuation ap=40dB.
Step 2:
Method- filter design using Butterworth reference analog prototype filter.
Step 3:
Filter order is predetermined, N=2.
Step 4:
The Butterworth reference prototype filter transfer function has no zeros, only poles. These can
be computed via expression:
As N = 2, the values of poles are:
The reference analog prototype filter transfer function is:
Step 5:
First it is necessary to compute the analog prototype filter cut-off frequency c.
The analog filter transfer function is obtained from the reference analog prototype filter transfer
function using expression:
As the Butterworth reference prototype filter has no zeros, the expression for transfer function is
simpler:
Step 6:
The transformation into a digital filter through bilinear transformation:
Generally, substituting the complex variable s into the expression for analog filter transfer
function, the following is obtained:
where:
zk are the zeros of analog filter transfer function; and
pk are the poles of analog filter transfer function.
This general expression can be written in a simpler way in this example:
A more condensed form of the previous expression is:
The result is the IIR filter transfer function.

Step 7:
The filter order is predetermined.
There is no need to additionally change it.
Filter realization:
Figure 3-6-2 illustrates the direct realization of designed IIR filter, whereas Figure 3-6-3
illustrates the frequency response of the filter obtained using Filter Designer Tool.
Figure 3-6-2. Digital IIR filter direct realization in this example
Figure 3-6-3. Digital IIR filter frequency characteristic in this example

3.6.1.2 Example 2
Step 1:
Type of filter high-pass filter
Filter order N=3;

Passband cut-off frequency fc=5KHz; and
Minimum stopband attenuation ap=40dB.
Step 2:
Method- filter design using Butterworth reference analog prototype filter
Step 3:
Filter order is predetermined, N = 3.
Step 4:
Step 5:
simpler:
Step 6:
where:

Step 7:
Filter realization:
Figure 3-6-4 illustrates the direct realization of designed IIR filter.

3.6.1.3 Example 3
Step 1:
Type of filter band-pass filter
Filter order N=4;

Sampling frequency fs=20KHz; and
Passband cut-off frequency fc1=4KHz, fc2=6KHz.
Step 2:
Method - filter design using Butterworth reference analog prototype filter
Step 3:
Step 4:
When designing an IIR band-pass filter, the reference prototype filter order is half that of the
required IIR filter order. In this example for N=4, the order of reference prototype filter is 2, so
the values of poles are:
Step 5:
simpler:
Step 6:
Generally, by substituting the complex variable s into the expression for analog filter transfer
where:
In this example, the general expression can be written in a simpler way:

Step 7:
Filter order is predetermined.
Filter realization:

3.6.1.4 Example 4
Step 1:
Type of filter band-stop filter
Filter order N=4;

Sampling frequency fs=20KHz; and
Passband cut-off frequency fc1=3000Hz, fc2=3300Hz.
Step 2:
Method - filter design using Butterworth reference analog prototype filter.
Step 3:
Step 4:
When designing an IIR band-stop filter, the reference prototype filter order is half that of the
required IIR filter order. In this example for N = 4, the order of reference prototype filter is 2, so
Step 5:
simpler:
Step 6:
where:

Step 7:
Filter realization:

3.6.2 Filter design using Chebyshev filter
3.6.2.1 Example 1
Step 1:
Sampling frequency fs=44100Hz;

Passband cut-off frequency fc1=15KHz;
Stopband cut-off frequency fc2=18KHz;
Maximum passband attenuation ap=1dB; and
Minimum stopband attenuation as=40dB.
Step 2:
Method - filter design using Chebyshev reference analog prototype filter
Step 3:
Filter order is not pre-determined, so it is necessary to choose an initial solution from which
iterative method starts. The solution is redefined progressively until some pre-determined
requirements are satisfied. Lets assume that the initial filter order is 4.
Step 4:
The Chebyshev reference prototype filter transfer function has no zeros, only poles. These can be
computed via expression:
The Chebyshev filter transfer function is expressed as:
In this case, the value of constant A0 is:
A0 = 0.2457
so that the Chebyshev reference analog prototype filter transfer function is:
Step 5:
As the Chebyshev reference prototype filter has no zeros, the expression for transfer function is
simpler:
Step 6:
where:

Step 7:
By analyzing the resulting filter using Filter Designer Tool, it is obvious that the attenuation
amounting to 31.2dB approximately at the frequency of 18KHZ is not sufficient. The frequency
characteristic of the resulting digital filter is illustrated in Figure 3-6-10 below.
Figure 3-6-10. Frequency characteristic of the resulting IIR filter

It is necessary to additionally redefine the filter order until the predefined requirements are
satisfied. The filter order is incremented by 1 and is 5 therefore. All steps starting with step 3 are
iterated.
Step3:
The filter order is incremented in the second iteration. A new filter order is 5.
Step 4:
The Chebyhsev filter transfer function is expressed as:
A0 = - 0.1228
so the Chebyshev reference analog prototype filter transfer function is:
Step 5:
simpler:
Step 6:
where:

Step 7:
Figure 3-6-11. Frequency characteristic of the resulting IIR filter

It is not necessary to further increase the filter order as this one is appropriate. Sometimes more
iterations are needed to determine the filter order. The whole procedure is the same, only it takes
more time.
Filter realization:
illustrates the frequency characteristic of the filter obtained using Filter Designer Tool.
Figure 3-6-12. Digital IIR filter direct realization
Figure 3-6-13. Frequency characteristic of digital IIR filter

3.6.1.2 Example 2
Step 1:
Filter order N = 3;
Sampling frequency fs = 20KHz;
Passband cut-off frequency fc = 5KHz; and
Maximum passband attenuation ap = 1dB.
Step 2:
Method - filter design using Chebyshev reference analog prototype filter
Step 3:
Step 4:
A0 = - 0.4913
Step 5:
simpler:
After substitution of poles and c into expression:
Step 6:
where:

Step 7:
Filter realization:
Figure 3-6-14. Direct realization of digital IIR filter in this example
Figure 3-6-15. Frequency characteristic of digital IIR filter in this example

3.6.1.3 Example 3
Step 1:
Filter order N = 4;
Passband cut-off frequencies fc1 = 4KHz, fc2 = 6KHz; and
Step 2:
Method filter design using Chebyshev reference analog prototype filter.
Step 3:
Step 4:
In this example, the value of constant A0 is:
A0 = 0.9826
Step 5:
simpler:
Step 6:
where:
In this expression, the general expression can be written in a simpler way:

Step 7:
Filter realization:
Figure 3-6-16 illustrates direct realization of designed IIR filter.
Filter 3-6-16. Direct realization of digital IIR filter in this example

3.6.1.4 Example 4
Step 1:

Filter order N = 4;
Passband cut-off frequencies fc1 = 3000Hz, fc2 = 3300Hz; and
Step 2:
Method - filter design using Chebyshev reference analog prototype filter.
Step 3:
Step 4:
In this example, the value of constant A0 is:
A0 = 0.9826
Step 5:
simpler:
Step 6:
where:

Step 7:
Filter realization

3.6.3 Filter design using inverse Chebyshev filter
3.6.3.1 Example 1
Step 1:
Sampling frequency fs = 44100Hz;

Passband cut-off frequency fc1 = 15KHz;
Stopband cut-off frequency fc2 = 18KHz;
Maximum passband attenuation ap = 1dB; and
Minimum stopband attenuation as = 40dB.
Step 2:
Method - filter design using inverse Chebyshev reference analog prototype filter.
Step 3:
Filter order is not pre-determined, so it is necessary to choose an initial solution from which
iterative method starts. The solution is redefined progressively until some pre-determined
requirements are satisfied. Lets assume that the initial filter order is 4.
Step 4:
The inverse Chebyhsev filter transfer function is expressed as:
In this example, the value of constant H0 is:
H0 = 0.01
so that the Chebyshev reference analog prototype filter transfer function is:
Step 5:
In this example, the analog filter transfer function becomes:
Step 6:
where:
Step 7:
Figure 3-6-20. Frequency charactersitic of designed IIR filter

The attenuation at 15KHz is higher than 1dB and the resulting filter doesnt satisfy the
specifications therefore. It is necessary to additionally redefine the filter order until the
predefined requirements are satisfied. The filter order is incremented by 1, and becomes 5
therefore. All steps starting with step 3 are iterated.
Step 3:
The filter order is incremented in the second iteration. A new filter order is 5.
Step 4:
The inverse Chebyshev reference prototype filter transfer function has no zeros, only poles.
These can be computed via expression:
The Chebyhsev filter transfer function is expressed as:
In this case, the value of constant H0 is:
H0 = - 0.05
so the inverse Chebyshev reference analog prototype filter transfer function is:
Step 5:
In this example, the analog filter transfer function becomes:
Step 6:
where:

Step 7:
By analysing the resulting filter using Filter Designer Tool, it is obvious that the attenuation at
18KHZ is less than 1dB (approximately 0.8dB). The frequency characteristic of the resulting
digital filter is illustrated in Figure 3-6-21 below.
Figure 3-6-21. Frequency characteristic of IIR filter

It is not necessary to further increase the filter order as this one is approriate. Sometimes more
iterations are needed to determine the filter order. The whole procedure is the same, only it takes
more time.
Filter realization:
illustrates the frequency characteristic of the filter obtained using Filter Designer Tool.

3.6.1.2 Example 2
Step 1:

Filter order N = 3;
Passband cut-off frequency fc = 5KHz; and
Minimum stopband attenuation 40dB.
Step 2:
Step 3:
Filter order is predetermined, N=3.
Step 4:
The inverse Chebyshev filter transfer function is expressed as:
In this case, the value of constant H0 is:
H0 = -0.03
Step 5:
In this example, the transfer function is:
After substituting the poles and c into expression:
Step 6:
where:

Step 7:
Filter realization:

3.6.1.3 Example 3
Step 1:
Filter specification:
Filter order N = 4;
Passband cut-off frequencies fc1 = 4KHz, fc2 = 6KHz; and
Step 2:
Step 3:
Step 4:
required IIR filter order. In this example for N=4, the order of reference prototype filter is 2, so
H0 = 0.01
Step 5:
For this example, the analog filter transfer function is:
Step 6:
where:
Step 7:
Filter realization:
Figure 3-6-26 illustrates direct realization of designed IIR filter.

3.6.1.4 Example 4
Step 1:
Filter specification:
Filter order N=4;

Stopband cut-off frequencies fc1=3000Hz, fc2=3300Hz; and
Step 2:
Step 3:
Step 4:
The inverse Chebyshev reference prototype filter transfer function has no zeros, only poles.
These can be computed via expression:
When designing an IIR band-stop filter, the reference prototype filter order is half that of the
The inverse Chebyshev filter transfer function is expressed as:
H0 = 0.01
Step 5:
In this example, the analog filter transfer function is:
Step 6:
where:

Step 7:
Filter realization:

Infinite Impulse Response

Uploaded by

Copyright:

Available Formats

Infinite Impulse Response

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Infinite Impulse Response

Uploaded by

Copyright:

Available Formats

Infinite Impulse Response (IIR) Filters

Figure 3-1-1. Block diagrams of FIR and IIR filters

Figure 3-1-3. The effect of non-linear phase characteristic

3.2 Infinite impulse response (IIR) filter design

Figure 3-2-1a. Low-pass digital filter specification

Figure 3-2-1b. Low-pass digital filter specification

p normalized passband cut-off frequency;

A stopband attenuation parameter;

Frequency normalization can be expressed as follows:

fs is the sampling frequency;

Figure 3-2-2a. High-pass digital filter specification

Figure 3-2-2b. High-pass digital filter specification

Filter 3-2-3a. Band-pass digital filter specification

Figure 3-2-3b. Band-pass digital filter specification

Figure 3-2-4a. Band-stop digital filter specification

Figure 3-2-4b. Band-stop digital filter specification

where z is the complex number.

The last expression is the z-transform of the given signal.

Figure 3-2-5 illustrates the (frequency) spectrum of the given signal.

Figure 3-2-5. Frequency spectrum of the given signal

In polar coordinates, the complex number z may be expressed as follows:

Figure 3-2-6. Fourier transform in the z plane

Figure 3-2-7a Stable IIR filter

Figure 3-2-7b. Instable IIR filter

bi are the feedforward filter coefficients (non-recursive part);

Figure 3-2-8 Discrete-time system with feedback

The following expression describes the filtering process:

After making substitutions of impulse response coefficients, it becomes:

it is possible to find function for particular normalized frequency.

The numerator is computed first:

Figure 2-2-8 illustrates a hardware realization of this IIR filter.

Figure 3-2-9. Realization of IIR filter in this example

Figure 3-2-10. Deviation from required frequency characteristic

3.2.5 IIR filters design using bilinear transformation

N is the filter order;

Figure 3-2-11. Block diagram of IIR filter

Figure 3-2-12. Direct realization of non-recursive part of IIR filter

Figure 3-2-13. Direct realization of non-recursive part of IIR filter

where the operator

represents rounding down to a less value.

Figure 3-2-15. Circular buffer of length 16 = 2^4

It means that the minimum length of circular buffer is 2^3 = 8.

addr. addr. addr. addr. addr. addr. addr. addr.

Table 2-2-2. Input circular buffer after receiving 10 samples

Figure 3-2-18. Direct canonic realization structure block diagram

Figure 3-2-19. Direct transpose canonic realization structure block diagram

bi are the coefficients of transfer function numerator (non-recursive part);

Figure 3-2-20. First-order section

Figure 3-2-21. Second-order section

Figure 3-2-22. Cascade IIR filter structure

3.3 Reference Analog Prototype Filter

Figure 3-3-1. S plane and region of stability

Figure 3-3-2. Frequency response of Butterworth filter

Figure 3-3-3. IIR filter specification

3.3.2 Chebyshev Analog Filter

Figure 3-3-5. Frequency response of Chebyshev analog filter

Chebyshev analog filter is defined via expression:

Transfer function is expressed as:

The value of A0 is found via expression:

For N=5, the transfer function is: