Analog Filter Realization Procedure

E.L. Dove
March 1, 1999

Butterworth Filters
Butterworth analog filters can be characterized by
Avo
H ( s) = (1)
Bn (s)
Several Butterworth polynomials are tabulated below (from Millman and Halkias [1]) for various
values of the parameter n. The design of a Butterworth filter involves choosing the mid-band
gain Av0, and n, the order of the filter necessary to yield the desired response. These design
parameters are chosen based on the frequency characteristics of the signal being studied.

If s is replaced by jΩ and the magnitude-squared is considered, then

2
H ( jΩ ) =
(A ) vo
2

(2)
2n
Ω 
1+  
Ω 0 
Equation (2) describes the power that is transmitted by the transfer function H(jΩ ). The first
derivative of Eq. (2) is zero at Ω = 0 and Ω = ∞ , thus indicating that the response of the transfer
function is flat at zero Hz, and as frequency increases towards ∞ . The assumption used to
generate the normalized Butterworth polynomials in the table above is that Ω 0 = 1 rad/s.

The design problem then is to first determine the order of the filter by solving Eq. 3.

-1-
 (
 10 − 0.1K1 − 1  )
n=  10 (
log10  − 0.1K 2
−1

) (3)
Ω 1 
2 log10 Ω  
 2
where K1, K2, Ω 1, and Ω 2 are the fiducial gains and frequencies of the desired frequency
response of the filter being designed. The resulting value of n should be rounded up to the
nearest whole number. This is the order of the filter to be designed and built. Once n has been
determined, then the value of Ω 0, the cutoff or corner frequency, can be determined by
substituting into Eq. (2) with the value of K1 or K2 yielding either

Ω1
Ω0 = 1
(4)
 Κ
− 1  2n
10 10
− 1
 
 
or
Ω2
Ω0 = 1
(5)
 Κ
− 2 
2n
10 10
− 1
 
 

The choice depends on which of the two sets of fiducial points is the more important set, i.e.,
which of the two sets should be exactly achieved.

The design is made considerably more useful by noting that all of the Butterworth polynomials
in Table 13-2 are made of first- and/or second-order polynomials, and these transfer functions
are given by [2]:
Av 0 Av 0
H 1 (s ) = or H 2 ( s ) = 2
(6)
s  s   s 
+ 1   
Ω0 Ω  + 2ςΩ + 1
 0  0

To calculate the parameters for a high-pass filter, simply interchange the resistors and capacitors.
To design a band-reject filter, add appropriately designed high-pass and low-pass filters. To
design a band-pass filter, cascade an appropriately designed low-pass with an appropriately
designed high-pass.

The filters can be realized with the following circuits (from [1]):

-2-
 R1′ 1 1
Av0 = 1 + Ω0 = Ω0 =
R1 RC RC
Figure 1: First- and second-order filter realizations with op amps, and passive
resistor and capacitor networks.
Millman and Halkias [1] show that the circuits in Fig. 1 have the transfer functions given in Eq. 6
if the cutoff frequencies and gain are given by
1
Ω0 = for both orders, and Av 0 = 3 − 2ς for a second - order system. (7)
RC

The value of Av0 for the first-order system is not important to the
design, and can be any value.

Synopsis:
The procedure for designing analog Butterworth filters is:
1. Define the desired filter characteristics (fiducial points K1, K2, Ω 1, and Ω 2).
2. Calculate the required order of the filter n.
3. Calculate the required 3-dB point (cutoff frequency Ω 0).
4. Calculate the required mid-band gain Av0 based on the normalized Butterworth
polynomials (Eq. 7).
5. Calculate required values of R and C in Figure 1.
6. If a hi-pass filter is desired, then switch the R and C components. If a band-
pass filter is desired, then cascade a low-pass with a hi-pass filter. If a band-
reject is desired, then add high-pass and low-pass filters.

References:
[1] Millman, J. and Halkias, C. Electronic Fundamentals for Engineers and
Scientists, McGraw-Hill, New York, pp. 332-336, 1976.
[2] Ludeman, L. Fundamentals of Digital Signal Processing, Harper & Row, New
York, pp. 120-134, 1986.

-3-