Prepared by Professor Eliathamby Ambikairajah, Head of

School
Electrical Engineering and Telecommunications
University of New South Wales (UNSW), Australia

Modified and Delivered by Dr. T. Thiruvaran

Analogue Filter Design ........................................................................

Design of Lowpass Butterworth Filters ....................................
Lowpass to High pass transformation ......................................
Low-pass to Band-pass Transformations .................................
Summary ...................................................................................
Analogue Filter Design

The ideal, brick-wall, low pass filter prototype is one, which has
a unit amplitude frequency response from dc to 1 rad/sec with
response dropping to zero there after.

Ideal low pass filter

This may be defined in term of the following squared magnitude

transfer function:

(0.1)

Where (0.2)
Since is a polynomial in , must
satisfy the lower condition in equation (0.2). It is not practically

Therefore approximations to are made. The

approximation for due to Butterworth is given by
. This yields the following squared magnitude
transfer functions:

(0.3)

(0.4)

where n is a positive integer.

Design of Lowpass Butterworth Filters

A Butterwo n

(0.5) (see eq (0.3))

- Cut-off frequency of the filter.

The following properties are easily determined:

1. for all n.

2. for all n; this implies

and

3. is monotonically decreasing function of (i.e.

throughout pass band and stop band.)
4. As n gets larger approaches an ideal low-pass
frequency response.
5. is called maximally flat at the origin since all order
derivatives exist with respect to and are zero at the origin.

Note:

Power series expansion

|H(j )|2

Figure 1: The magnitude frequency response of a Butterworth filter

It is convenient in many cases to look at the frequency response
in decibels, that is, plot versus .

Figure 2 is a straight line approximation of the frequency

response in decibels (dBs) for the Butterworth filters.

dB =20log|H (j )|

0.1 c c 100 c
0
-3dB
n=1 Slope= -20dB/decade
Actual values for n=3
-20 n=2 Slope= -40dB/decade
-40
n=3 Slope= -60dB/decade
-60dB

Figure 2: Filter gain plot for analogue Butterworth filters

of various orders of n.

Starting with the magnitude square frequency response we

would like to find the system function H(s) that gives the
Butterworth magnitude squared response.
Let

(0.6)

For n even (0.7)

For n odd (0.8)

From equation (0.7) the roots of the denominator, i.e. poles are
given by

=>

where

For n even, the first pole is at [i.e. k=0] and the remaining

poles equally spaced around the unit circle by

j
H(s) H(-s)
1 s-plane

2n 2n
2n 2n

-1
Unit circle
n - even

For instance, when n = 2, k = 0, 1,2,3

1 s-plane

H(s) 2
4

H(-s)
-1
Unit circle

The system H(s)H(-s) is a combination of two cascaded systems

H(s) and H(-s) where the poles of one system is obtained by
reflecting the poles of the other system about the origin; i.e., if s
is a pole of H(s), then s is a pole of H(-s).
From equation (0.8), the roots of the denominator i.e. poles are
given by

=>

provides 2n-2 unique poles

For k=0,

For k=n,

we have 2n poles
For n odd, the poles are equally spaced on the unit circle in
angle

j
H(s) H(-s)
1 s-plane

-1
Unit circle
n - odd

To obtain a stable and causal filter, the poles of H(s) are selected
to be those in the left half plane.

A plot of the poles the Butterworth function for n = 3 (odd) is

shown in Figure 3.
j

1 s-plane

H(-s)
H(s)
0
- 60 60 0

- 60 0 600

-1

Figure 3: The s-plane of a butterworth filter with n = 3.

poles of H(s) H(-s)

A plot of the poles of the Butterworth function for n = 4 (even
Figure 4.

j
s-plane
H(s) 1
/4
H(-s)
/8

- /8

-1
- /4

Figure 4: The s-plane of a butterworth filter with n = 4

n=4 2n = 8 poles of H(s) H(-s)

Example: Find the transfer function H(s) for the normalised
Butterworth filter of order 2 (n = 2)

rad /sec

Since n=2; with k=0,1,2,3

s1 /2 s3
H(-s)
H(s) /4 /4
-1 /4 1
s2 s4

Normalised implies c= 1
Example:
Determine the transfer function of a Butterworth filter of the
low-pass type with order n = 3. Assume that the 3dB cut-off
frequency rad/sec.

For n = 3, the 2n = 6 poles of H(s)H(-s) are located on a circle

of unit radius with angular spacing as shown in Figure 3.

Hence allocating the lest-half plane poles to H(s), we define

The transfer function of a Butterworth filter of order 3 is

therefore

n Butterworth Polynomials
1 s+1
2

5
Example: Design of low-pass Butterworth filters

Design an analogue Butterworth filter that has -2dB or better

cut-off frequency of 20 rad/sec and at least 10dB of attenuation
at 30 rad/sec

|H( )| dB =20log|H(j )| or 10log|H(j )|2

0
k1= -2dB

k2= -10dB

x s

20 rad/sec

filter order (n)

- see equation (0.3)

equal to k1 at

equal to k2 at
(0.9)

(0.10)

(0.11)

similarly (0.12)

From equations (0.11) and (0.12)

2n
x k1

c 10 10 1
2n k2
s 10 10
1
c
(0.13)

If n is an integer we use that value, otherwise we use the next

larger integer.

4th order filter

From equation (0.9), we have

The normalised low-pass Butterworth filter for n = 4

can be found from the table given previously

De-normalised transfer function

c= 21.3868

De-normalised transfer function

Example: Design an analogue Butterworth filter (the filter is
monotonic in the pass and stop bands (i.e. no ripples)) which
meets the following specifications:
Low-pass filter: 0 to 10 kHz (pass band)
Transition band: 10 to 20 kHz
Stop band attenuation: -10dB (starts at 20 kHz)
dB

-10dB

10 kHz 20 kHz

at
Thus we choose n = 2 (even)

Normalised

where
Lowpass to High pass transformation
To transform analogue low-pass filter H(s) with unity cut-off
frequency to low-pass filter H(s) with cut-off frequency , we
substitute

Example: First order Butterworth prototype filter is given by

Normalised

To transform to new cut-off frequency = 5, we replace s with

Note: To transform an analogue low-pass filter H(s) with unit

cut-off frequency to high-pass filter H(s) with cut-off frequency
, we substitute
Example: First order Butterworth prototype filter is given by

Normalised

To transfer to a high-pass filter with cut-off frequency

We replace ;

Note: High-pass filter contains zeros as well as poles

Example:

3rd order Butterworth filter.

Determine the transfer function of the corresponding high-pass

filter with cut off frequency (normalised)

i.e.
normalised
Low-pass to Band-pass Transformations
By definition, a band pass filter rejects both low and high
frequency components and passes a certain band of frequencies
some where between them. Thus the frequency response ,
of a band-pass filter has the following properties.

1. | at both &

2. for a frequency band centered on , where

is the mid frequency of the filter

dB = 20log|H(j )|
1

k1
k2

L 0 u

Lower cut-off frequency Upper cut-off frequency

Bandwidth of the band-pass filter

To convert unity cut off low-pass filter H(s) into a Band-Pass

filter H(s) with lower cut-off frequency and the upper cut-off
frequency , we replace
Similarly to convert unity cut-off lowpass filter H(s) into a
bandstop filter H(s) with lower cutoff frequency L and upper
cut-off frequency u, we replace:

dB

k1 Band stop filter

k2

L 1 2 u

Note (second method): A lowpass to bandpass transformation

can be performed by

B - Bandwidth of the band-pass filter

- centre frequency.
Summary: Analogue to Analogue Transformation

Butterworth Prototype response Transformed filter response

dB = 20log|H(j )| dB

0 0
k1 k1

k2 k2

1 p c
p. c

unity cut-off

dB

k1

k2

c/ p c

0- L
dB dB
0- L u- 0
k1 k1
u- 0

k2 k2

L 1 0 2 u 1 L 0 u 2
Example: Design an analogue band-pass filter with the
following characteristics

(a) -3.0103 dB upper and lower cut-off frequency of 50Hz and

20kHz
(b) a stop-band attenuation of at least 20dB at 20Hz and 45
kHz
(c) a monotonic frequency response.

Solution:

dB =20log|H(j )|

0
-3.01

At least
20dB

1 L u 2

20Hz 50Hz 20 kHz 45 kHz

The monotonic requirement can be satisfied with a Butterworth
filter.

Transformation is

For band-pass filter to satisfy the stop-band attenuation

requirement at we must have equality within the
transformation. i.e.

r is the critical radian

frequency of the lowpass
filter.

Substitute values for

Similarly solving , to satisfy stop-band attenuation

requirement at
The critical frequency r obtained are not equal.

We choose , the most restrictive

value obtained.

First design a low-pass filter H(s) with and then apply

transformation (LP to BP) using to obtain BP filter.

The low-pass filter order n can be calculated as before

n = 2.829 and we choose n=3

s2 L u s2 3.94784 10 7
LP BP s
s u L s(1.25349 10 5 )