Introduction

The Butterworth implementation ensures flat response ('maximally flat') in the pass
band and an adequate roll-off. This type of filter is a good all rounder, simple to
understand and is good for applications such as audio processing.
The Chebyshev implementation gives a much steeper roll-off, but has ripple in the pass
band, so is no use in audio systems. It is however, far superior in applications where there is
only one frequency present in the pass band but several other frequencies that need
removing (e.g. deriving a sinewave from a square wave by filtering out the harmonics).
1.1 Objective:
Basically, in this project we are going to introduce the difference between Butterworth
and Chebyshev and as we know there are many books have been written on active filter
design and they normally include countless pages of equations that frighten most small dogs
and some children. This article sets out to unravel the mystery of filter design and to allow
the reader to design continuous time, analogue filters based on op amps in the minimum of
time and with the minimum of mathematics.
1.2 Scope:
In this project we will compare between Butterworth and Chebyshev filters using the
theory, calculations and we will use the LTSPICE to find the wave form.
The Theory of Analogue Electronics
Analog electronics has two distinct sides: the theory taught by academic institutions
(equations of stability, phase-shift calculations, etc.), and the practical side familiar to most
engineers (avoid oscillation by tweaking the gain with a capacitor, etc.). Unfortunately, filter
design is based firmly on long-established equations and tables of theoretical results. Filter
design from theoretical equations can prove arduous. Consequently, this discussion employs a
minimum of matheither in translating the theoretical tables into practical component
values, or in deriving the response of a general-purpose filter.
The Fundamentals
A simple RC Low Pass Filter has the transfer function
TF=

1
1+ cSR ..eq 3

Cascading filters similar to the one above will give rise to quadratic equations in the
denominator of the transfer function and hence further complicate the response of the filter.
In fact, any second order Low Pass filter has a transfer function with a denominator equal to

as + bs+c

..eq 3.1

and substituting different values of a, b and c determine the response of the filter over
frequency. Now, anyone who remembers high school maths will notice that the above
statement can have a value of zero for certain values of s, given by the equation
S=

b b 24 ac
eq 3.2
2a

Hence our transfer function can present infinite gain.

It is the points at which the quadratic equation in the denominator equals zero that the transfer
function has theoretically infinite gain. These points are known as the poles of the quadratic
equation and it is the values of these poles that establish the performance of each type of filter
over frequency. These poles normally come in pairs and take the form of a complex number
(a+ jb) and its complex conjugate (a jb) , although sometimes jb is zero.
In practice it is the real part, a, of the pole that indicates how the filter will respond
to transients and the imaginary part, job, that indicates how the filter will respond
over frequency. We only ever apply frequencies to a filter, hence need not worry about the
system going unstable.
This text will describe how to transfer the tables of poles found in many text books into
component values suitable for circuit design.
Literature review
The two most common filter characteristics, and the ones discussed in this text, are
Butterworth, Chebyshev, each giving a different response. There are many others, but 90%
of all applications can be solved with one of the above implementations.
Butterworth Filter

This filter is also called as maximally flat or flat flat filter. This class of filters approximates
the ideal filter well in the pass band. Frequency response curves of different types of filters.
The Butterworth filter has an essentially flat amplitude-frequency response up to the cutoff
frequency. Although Butterworth filters achieve the sharpest attenuation, their phase-shift as
a function of frequency is non-linear. It has a monotonic drop in gain with frequency in the
cut-off region and a maximally flat response below cut-off frequency. The Butterworth filter
has characteristic similar as to that of Chebyshev filter. It has a moderate roll-off of the skirt
and a slightly nonlinear phase responses.
Chebyshev Filter

It is also called a equal ripple filter. It gives a sharper cut-off than Butterworth filter in the
passband. Both Butterworth and Chebyshev filters exhibit large phase shifts near the cut-off
frequency. A drawback of the Chebyshev filter is the appearance of gain maxima and
minima below the cut-off frequency. This gain ripple, expressed in db, is an adjustable
parameter in filter design.
The faster the roll-off, the greater the peak-to-peak ripples in the passband. The phase
response is highly non-linear in the skirt region. Such unequal delays of data frequency in
the passband causes severe pulse distortion and thus increased errors at modern
demodulators. This can be overcome somewhat by increasing the BW of the filter so that the
phase region is extended.A Chebyshev filter is used where very sharp roll-off is required.
However, this is achieved at the expense of a gain ripple in the lower frequency passband.

Figure 1: Chebyshev vs. Butterworth High-Pass Response.

Types of Filter
The three most common filter characteristics, and the ones discussed in this text, are
Butterworth, Chebyshev and Bessel, each giving a different response. There are many others,
but 90% of all applications can be solved with one of the above implementations.

The Butterworth implementation ensures flat response ('maximally flat') in the pass
band and an adequate roll-off. This type of filter is a good all rounder, simple to understand
and is good for applications such as audio processing.
The Standard Filter Blocks
Figure 1a shows the generic filter structure and by substituting capacitors or resistor
in place of components G1-G4, either a High Pass or Low Pass implementation can be
realized. Considering the effects of these components on the op amp feedback network, it is
quite easy to picture that a Low Pass filter can be derived from Fig 1a by making components
G2/G4 into capacitors and G1/G3 resistors and vice versa for a High Pass implementation.

FIG 5
The transfer function of the Low Pass Filter shown in Fig 1b is
TF=

GIG 3
s ( C 2 C 4 ) +sC 4 ( G1+G 3 ) +G1 G 3 .eq 5
2

(to understand how this transfer characteristic was determined, please see the article on
Nodal Analysis of Op Amps on this website).
Here the equation is simpler if conductance are used, where the capacitors have a
conductance of sC and the resistors have a conductance of G.
4

If this looks too complicated, life is made much easier if the above equation is 'normalized'.
This involves selecting either the resistors or the capacitors to be equal to 1 or 1F
respectively and changing the surrounding components to fit the response. Thus, making all
resistor values equal to 1 , the Low Pass transfer function is
C 2 C 40+ s ( 2 C 4 ) +1
s2
1
TF=

Equation 5.1

This is the transfer function describing the response of our generic second order Low Pass
Filter. We are now faced with taking the theoretical tables of poles describing the three main
filter responses and translating these into real component values.
The Design Process
The designer should determine the type of filter required by selecting the pass band
performance needed from the descriptions above. The simplest way of determining the
desired order of the filter is to design a second order filter, then cascade multiple stages of
this filter. Check to see if the implementation gives the desired stop band rejection, then
proceed to design with the correct pole locations, shown in the tables in the appendix. Once
the pole locations are established, the component values can soon be calculated.
Firstly, each pole location needs to be transformed into a quadratic expression similar to that
in the denominator of our generic second order filter. If a quadratic equation has poles
of (a jb) , then it has roots of

( sa jb ) and (sa+ jb) .eq 6

When these are multiplied together, the resulting quadratic equation is
(s 22 as+ a2 +b2 ) eq 6.1
In the pole tables, a is always negative, so it is sometimes more convenient to use
(s 2 +2 as+ a2+ b2 )

.. eq 6.2

and use the magnitude of a regardless of its sign.

To put this into practice, consider a fourth order Butterworth filter. Below are the poles of a
fourth order Butterworth filter with the associated quadratic expression of each pole location.
The poles can be found from the file at the end of this document.
Table 6

Quadratic

Poles (a jb)
0.9239 jo .3827
o .3827 jo.9239

S +1.8478 s+ 1
S 2 +o .7654 s+1

From this, a fourth order Butterworth Low Pass Filter can be designed. Simply substitute the
values from the above two quadratic expressions into the denominator of Equation 5.1.
Thus, in the first filter
C2C4 = 1; 2C4 = 1.8478

implying C4 = 0.9239F; C2 = 1.08F

and for the second filter

C2C4 = 1; 2C4 = 0.7654

implying C4 = 0.3827F; C2 = 2.61F

And all the resistors in both filters equal 1.

Cascading both second order filters will yield a fourth order Butterworth response with a roll
off frequency of 1 rads-1 but with components impossible to find. For those who do not find
the above frequency or component values suitable, read on.
It just so happens that as long as the ratio of the reactance to the resistors is maintained, the
circuit response remains unchanged. Therefore one may elect to have 1k resistors. Then
divide the capacitor values by 1000 to ensure that the reactances increase in the same
proportion as the resistances.
Unfortunately, we still have the perfect Butterworth response, but with a roll off frequency of
1rads-1. To change the frequency response of the circuit, we must ensure the ratio of the
reactance to the resistors is maintained, but simply at a different frequency. Keeping the
resistors at 1k, for a roll off at 1 kHz (rather than 1rads -1), the capacitor value has to be
further reduced by a factor of 2 x 1000. Therefore the capacitors reactance does not reach
our original (normalized) value until a higher frequency.
In conclusion, a fourth order Butterworth Low Pass Filter with a 1 kHz roll off takes the form
of FIG2

FIG 6
FIG 2b shows the LTspice plot of the output. Perfectly flat response in the passband and slow
roll off after the break frequency.

FIG 6.1
Cascading any number of second order filters will obtain any even-order filter response using
the above technique. At this point it is important to note that a true fourth order Butterworth
filter is not simply obtained by calculating the components for a second order filter, then
cascading two of these stages. Two second order filters have to be designed, each with
different pole locations.
If the filter we are trying to design has an odd order, we can simply cascade second order
filters, then add an RC network in the circuit to gain the extra pole.
For example, a 5th order, 1dB ripple Chebyshev filter has the following poles
Table 6.1
Poles
0.2265 jo .5918

Quadratic
2
2
S +o .453 s +o .402 2.488 s + 1.127 s+ 1
7

o .08652 jo .9575

s 2 +0.173 s +0.924 1.08 s2 +0.187 s+1

0.2800

See text

The first two quadratics have been multiplied by a constant to ensure the last term equals
unity, thus conforming to the generic filter described by Equation 5.1.
Thus, in the first filter
C2C4 = 2.488; 2C4 = 1.127

implying C4 = 0.5635F; C2 = 4.41F

and for the second filter

C2C4 = 1.08; 2C4 = 0.187

implying C4 = 0.0935F; C2 = 11.55F

Earlier in the text, it was shown that an RC circuit has a pole when
1+sCR=o ..eq 6.3
i.e.

s=

1
CR

eq 6.4

Therefore to obtain the final pole at s = -0.28, a capacitor of C = 3.57F is needed if R = 1.

Normalizing for a roll off frequency of 1 kHz, using 1k resistors gives the circuit shown in
FIG 3

FIG 6.2

Thus the designer can now boldly go and design many Low Pass filters of any order at any
frequency.
Now, all of the above theory can also be applied to the design of a High Pass Filter.
8

It has been shown that a simple RC Low Pass filter has the transfer function.
TF=

1
1+ sCR .eq 6.5

Similarly, a simple RC High Pass filter has the transfer function

R

TF=

R+

1 .eq 6.6
sC

Normalising these to correspond with the normalised pole tables gives:

TF=

1
1+ s

for a Low Pass .eq 6.7

and
TF=

1
1+

1
S

for a High Pass. Eq 6.8

It can be seen that the High Pass pole positions, s, can be obtained by inverting the Low Pass
pole positions. Inserting the values into the High Pass Filter Block will ensure that the correct
frequency response is obtained. To obtain the transfer function of the High Pass filter block,
we need to go back to the transfer function of the Low Pass filter block. Thus from
G1+G 30+G 1G 3
s2 ( C 2 C 4 ) +sC 4
eq 6.9
G 1G 3
TF=
interchanging capacitors and resistors it can easily be seen that the transfer function of the
equivalent High Pass Filter Block is
2

TF=

s C1C3
2
s ( C 1 C 3 ) + sG 4 ( C 1+C 3 )+G 2 G 4

eq 6.10

Again, life is made much simpler if the capacitors are normalized instead of the resistors
thus,
9

TF=

S
2
S + S ( 2 G 4 ) +G 2G 4

S2
G 2G 4
TF=
2
S
2
+S
+1
G 2G 4
G2
and putting
G2=

1
R2

G 4=

1
R4

and

gives
TF=

s2 R 2 R 4
s 2 R 2 R 4+ s ( 2 R 2 ) +1

.. Equation 6.11

This is the transfer function of the High Pass filter block and this time we calculate
the resistor values instead of capacitor values.
Once the general High Pass filter response has been obtained, the High Pass pole positions
can be derived by inverting the Low Pass pole positions and continuing as before.
Now, inverting a complex pole location is easier said than done. Take the example of the
5th order, 1dB ripple Chebyshev filter shown earlier. It has two pole positions at
(-0.2265 j0.5918)
The easiest way of inverting a complex number is to multiply then divide by the complex
conjugate, to obtain a real number in the numerator. Finding the reciprocal is then a matter of
inverting the fraction.
Thus

10

O .2265 JO .5618
(0.2265 JO .5918)
(0.2265 JO .5918)

Gives
TF=

0.4015
(0.2265 JO.5918)

Inverting gives
(0.564 J 1.474)
The newly derived pole positions can then be converted into the corresponding quadratic and
the values calculated as before.
We end up with
Table 6.3
Poles
-0.564 j1.474

Quadratic
s + 1.128s + 2.490 0.401s2 + 0.453s + 1
2

Thus from Equation 6.11 we can calculate the first filter component values to be
R2R4 = 0.401; 2R2 = 0.453

implying R2 = 0.227 ; R4 = 1.77

This process can then be repeated for the other pole locations.
A simpler way to achieve the above is to design for a Low Pass filter using the suitable Low
Pass poles, then treat every pole, s, in the filter as a single CR circuit since it has been shown
that .
s=

1
CR

Inverting each Low Pass pole to obtain the corresponding High Pass pole simply involves
inverting the value of CR. Once the High Pass pole locations have been obtained, we can
interpose the capacitors and resistors to ensure the correct frequency response.
For clarification consider that so far, for the Low Pass implementation a normalised capacitor
value was calculated assuming that R = 1. Hence the value of CR is equal to the value of C.
Finding the reciprocal of the value of C will find the High Pass pole and treating this as the
new value of R will yield the appropriate High Pass component value.
11

Considering again the fate of the 5th order 1dB ripple Chebyshev Low Pass filter, the
capacitor values calculated were:
C4 = 0.5635F; C2 = 4.41F.
To obtain the equivalent High Pass resistor values, invert the values of C (to get the High
Pass pole location) and treat these as the new normalised resistor values. i.e. R4 = 1.77 and
R2 = 0.227.
It can be seen that the same results are obtained as the more formal method mentioned earlier.
Thus the circuit in FIG 3 can now be converted into a High Pass Filter with a roll off at 1 kHz
by inverting the normalised capacitor values, interposing resistors and capacitors and scaling
the values accordingly. Previously, the normalised Low Pass values were scaled by dividing
by (2fR). In this case, the scaling factor used is (2fC), where C is the capacitor value and f
is the frequency in Hertz.
The resulting circuit is thus

Fig 6.3

Below we can see the LT spice schematic and transfer characteristic.

12

FIG 6.4

Conclusions
Using the aforementioned methods, you can design lowpass and highpass filters with
response at any frequency. Band pass and band stop filters can also be implemented (with
single op amps) using techniques similar to those shown, but those applications are beyond
the scope of this article. You can, however, implement band pass and band stop filters by
cascading lowpass and highpass filters. Information on Maxim op amps can be found at the
Amplifiers and Comparators. Therefore, we found that the Butterworth filter is the best
compromise between attenuation and phase response. It has no ripple in the pass band or the
stop band, and because of this is sometimes called a maximally flat filter. The Butterworth
filter achieves its flatness at the expense of a relatively wide transition region from pass band
to stop band, with average transient characteristics.
13

Reference

14

Paul, C. R. (2000). Fundamentals of Circuit Analysis. Wiley.

Alexander, C., Alexander, C. K., & Sadiku, M. (2006). Fundamentals of electric circuits. Urban Media
Comics.

Johnson, D. E., Johnson, J. R., Hilburn, J. L., & Scott, P. D. (1997). Electric circuit
analysis (Vol. 3). Prentice Hall.

15