Design of Lowpass Butterworth Filter: Analog and Digital Implementations

(Material presented in this document is primarily taken from Chapter 6 of “Analog Filter Design” by Van
Valkenburg)

Analog Butterworth filter, published in 1930, has the characteristic magnitude-squared frequency
response given by

( ) ( ) ( ) ( ) | ( )|
( )

where is the order of the analog Butterworth filter.

Butterworth filters enable a very good approximation to ideal brick-wall (or piece-wise constant)
lowpass filter magnitude-frequency response.

*(note that all

Butterworth filters exhibit maximally flat magnitude-frequency response, as seen below
figures, except the last, included in this document are taken from chapter 6 of Van Valkenburg’s book which uses the symbol ω

to denote angular frequency (radians/s) while Oppenheim uses Ω for this purpose. Note also that Valkenburg uses a normalized

cut-off frequency, i.e. Ω = 1 rad/s. This is in accord with the usual practice of analog filter design, in which a normalized cut-off
c
frequency based lowpass filter is first designed which can then be scaled using Frequency Scaling and then transformed to high-

pass- or band-pass-filter etc. using Frequency Transformation techniques. )

.

Usman Zabit: EE-330-Spring 2019 Version-2 Page 1 of 7

A typical 2nd-order system/filter does not very well approximate the ideal brick-wall frequency response,
as seen below.

However, a higher-order Butterworth filter is able to approach the brick-wall magnitude-frequency

response, as seen below for =10.

That being said, a higher order Butterworth filter, if imagined as a cascade of 2nd-order systems, is still
able to achieve the characteristic maximally-flat magnitude-frequency response.

Usman Zabit: EE-330-Spring 2019 Version-2 Page 2 of 7

This can be understood if we consider that each 2nd-order system (making up the composite
Butterworth filter) is having a different quality factor (denoted by ) value, while all 2nd-order systems
have the same cut-off frequency. See the diagram below to visualize how 3 different 2nd-order systems,
each having a different value, combine to achieve the maximally flat magnitude frequency response
when these are put in cascade. Consequently, a Butterworth filter is said to be stagger-tuned.

Let us now return to the magnitude-squared frequency response of Butterworth filter and calculate
where the poles are located in the plane.

( ) ( ) | ( )|
( )

Subsequent calculations will show that all poles of the above-mentioned magnitude-squared
frequency response of Butterworth filter are located on a circle of radius in the plane. However,
the angle at which each of these poles is located in the plane is given by equation 6.27 of
Valkenburg:

( )

Let us derive it from magnitude-squared response of Butterworth filter.

Usman Zabit: EE-330-Spring 2019 Version-2 Page 3 of 7

It is easy to see that poles of the Butterworth filter are found when

( )

Thus,

( )

( )

As the complex exponential is a periodic function with periodicity of , so

( )
( )

Now, roots of the above-mentioned equation can then be obtained by

( )
( )

As , so

( )
( )
( )
( )

( )

( )
( )

Thus, it is readily seen now that all poles of a Butterworth filter are located in the plane on a circle,
centered at origin, with a radius of while each of poles is located at an angle ( ).

Pole locations of 1st, 2nd, and 3rd order Butterworth filters are shown below.

Usman Zabit: EE-330-Spring 2019 Version-2 Page 4 of 7

Note that for the sake of stability, left half plane (LHP) poles are chosen for ( ). Also,

 A pole occurs on the real axis whenever is even, but not when is odd.
 A pole never occurs on the imaginary axis.
 Pole separation is always .
 All complex poles occur in complex-conjugate pairs.

As a result, an order Butterworth filter can then be seen as a cascade of number of 2nd-order
systems (when is even).

Thus, a 6th order Butterworth filter can be imagined to be composed of three ( ) different 2nd-
order systems, as seen below.

Let us now focus on describing such a 2nd-order system, containing a pair of complex-conjugate poles.

In such a case, if a pole occurs at o n , then another

pole also occurs at .

So, o n .

Likewise, o n .

And, o n .

Usman Zabit: EE-330-Spring 2019 Version-2 Page 5 of 7

Hence, a typical Butterworth transfer function, when is even, would include terms of the form

( )
( )( )

( )
( )

Or,

( ) ∏
( )

In case of a Butterworth filter with odd, there will simply be added to such 2nd-order terms, an
additional term corresponding to a pole at i.e. ( ) term appears in
the denominator.

( ) ∏
( )

Further simplification of System Function ( ) of Butterworth Filter

The above-mentioned system function of Butterworth filter can be further simplified by realizing
that irrespective of .

Furthermore, as we are interested in the LHP poles, so defining the angle with respect to negative
real-axis (i.e. ) enables to further simplify the computation of which can now be
obtained using

o o

So,
( )
( ) ( o )

Comparing the above expression with the general expression of a 2nd-order system, it may be noted
that quality-factor of each 2nd-order component of Butterworth filter is given by

Thus, the simplified expression of a typical Butterworth transfer function, when is even, becomes

( ) ∏
( o )

Usman Zabit: EE-330-Spring 2019 Version-2 Page 6 of 7

In case of a Butterworth filter with odd, it will be

( ) ∏
( o )

Design of Discrete-Time Butterworth Filter

So, when an analog Butterworth filter is completely designed then it can be transformed into
corresponding discrete-time Butterworth filter.

For this purpose, two techniques prevail, namely

1. Impulse Invariance Method, and

2. Bilinear Transformation Method.

See Oppenhe m’ art le 7. . and 7. . for deta l of each of these methods respectively.

Usman Zabit: EE-330-Spring 2019 Version-2 Page 7 of 7