Active Filters

INTRODUCTION

A filter is a circuit that is designed to pass a specific band of frequencies while

attenuating all signals outside this band. Filter network may be either active or
passive. Passive filter networks contain only resistors, inductors, and capacitors.
Active filters, which are only type covered in this text, employ transistors or OP-
AMPs plus resistors, inductors, and capacitors. Inductors are not often used in
active filters, because they are bulky and costly and may have large internal
resistive components.
There are four types of filters: low-pass, high-pass, bandpass, and band-
elimination (also referred as band-reject or notch) filters. A low-pass filter is a
circuit that has a constant output voltage from DC up to a cutoff frequency, fc. As
the frequency increases above fc , the output voltage is attenuated. Figure 6.1(a) is
a plot of the magnitude of output voltage of a low-pass filter versus frequency.
The solid line is a plot for the ideal low-pass filter, while the dashed line indicates
the curve for practical low-pass filter. The range of frequencies that are
transmitted is known as the passband. The range of frequencies that are
attenuated is known as stop band. The cutoff frequency, fc, is also called the
0.707 frequency, the −3 dB frequency, the corner frequency, or the break
frequency.
High-pass filters attenuate the output voltage for all frequencies below the
cutoff frequency. Above fc, the magnitude of the output voltage is constant. Figure
6.1(b) is the plot for ideal and practical high-pass filters. The solid line is the ideal
curve, while the dashed curves show how practical high-pass filters deviate from
the ideal.
Bandpass filters pass only a band of frequencies while attenuating all
frequencies outside the band. Band-elimination filters perform in an exactly
opposite way; that is, band-elimination filters reject a specified band of
frequencies while passing all frequencies outside the band. Typical frequency-
response plots for passband and band-elimination filters are shown in Figure
6.1(c) and (d). Once again, the solid lines represent the ideal plot, while dashed
lines show the practical curves.
 Active filters have been intensively studied, and many useful circuits have
been invented. Ideally, an active filter circuit should
1. Contain few components.
2. Have a transfer function that is insensitive to component tolerances.
3. Place modest demands on the OP-AMP’s gain-bandwidth product, output
impedance, slew rate, and other specifications.
4. Be easily adjusted.
5. Require a small spread of component values.
6. Allow a wide range of useful transfer functions to be realized.

V0 V0

Passband Stop band Stop band Passband

fc Frequency, f fc Frequency, f
(a) Low-pass filter (b) High-pass filter

V0 V0

Pass- Stop
Stop band Stop band Passband band Passband
band

fcl fc fch Frequency, f fcl fc fch Frequency, f

(c) Bandpass filter (d) Band-elimination filter

Figure 6.1 Frequency response for four categories of filters.

BASIC LOW-PASS FILTER

Introduction

The circuit of Figure 6.2(a) is a commonly used low-pass active filter. The
filtering is done by the RC network, and the OP-AMP is used as a unity-gain
amplifier. The resistor Rf is equal to R and is included for DC offset. The IC301 is
an special-purpose OP-AMP. Assuming virtual ground concept, the voltage across
capacitor C equals output voltage V0, because this circuit is a voltage follower. Vs
divides between R and C. The capacitor voltage equals V0 and is given by
1
j C
V0  Vs (6.1)
R 1
j C
Where  is the frequency of Vs in radians per second ( = 2f) and j is equal to
 1. Rearranging Equation (6.1) to obtain closed-loop gain ACL, we have
V 1
ACL  0  (6.2)
Vs 1  jRC
At very low frequencies, that is, as  approaches zero, ACL = 1, and at very high
frequencies, as  approaches infinity, ACL = 0. Therefore, circuit of Figure 6.2(a)
is a low-pass filter.
Rf V0
ACL  20 Log 10 ACL (dB)
Vs
+V
 1.0 0
301 0.707 −3
R
+
+ +
V0 0.1 −20
C V0 V 
+
 Vs

−40
0.01 
0.1 c c 10 c 100 c
(a) Low-pass filter for a roll-off of −20 dB/decade. (b) Frequency response plot for the circuit of
part (a).
Figure 6.2 Low-pass filter and frequency response plot for a filter with a −20 dB/decade
roll-off.

Figure 6.2(b) is a plot of ACL versus  and shows that for frequencies greater
than the cutoff frequency, c, ACL decreases at a rate of 20 dB/decade. The
cutoff frequency c is defined as that frequency of Vs where ACL is reduced to
1
0.707 times its low-frequency value. Let us assume that   . From Equation
RC
(6.2), we obtain ACL  1 2  0.707. Thus, the cutoff frequency of this filter
becomes
 1
c   2f c . (6.3)
RC
Equation (6.3) may be rearranged to solve for R:
1 1
R  . (6.4)
c C 2f c C
The solid curve in Figure 6.2(b) shows how the magnitude of the actual
frequency response deviates from the straight dashed-line approximation in the
vicinity of c. At 0.1c, ACL  1 (0 dB), and at 10c, ACL  0.1 (-20 dB). Table
6.1 gives both the magnitude and the phase angle for different values of 
between 0.1c and 10c.
Many applications require steeper roll-offs after the cutoff frequency. One
common filter configuration that gives steeper roll-offs is the Butterworth filter.

Table 6.1. Magnitude and phase angle for the low-pass

filter of Figure 6.2(a)
__________________________________________
 ACL Phase angle (degree)
__________________________________________
0.1 c 1.0 6
0.25c 0.97 14
0.5 c 0.89 27
c 0.707 45
2 c 0.445 63
4 c 0.25 76
10c 0.1 84
_________________________________________

Design Procedures

The design of a low-pass filter similar to Figure 6.2(a) is accomplished in three

steps:
1. Choose the cutoff frequency − either c or fc.
2. Choose the capacitance C, usually between 0.001 and 0.1 F.
3. Calculate R from Equation (6.4).

INTRODUCTION TO BUTTERWORTH FILTER

In many low-pass filter applications, it is necessary for the closed-loop gain to be

as close to 1 as possible within the passband. The Butterworth filter is best suited
for this type of application. The Butterworth filter is also called a maximally flat
or flat-flat filter, and all filters in this chapter will be off the Butterworth type.
Figure 6.3 shows the ideal (solid line) and the practical (dashed line) frequency
response for three types of Butterworth filters. As the roll-offs become steeper,
they approach the ideal filter more closely.
V0
ACL 
Vs

1.0 20 dB/decade

0.707
40 dB/decade
0.1 60 dB/decade

≈
 
0.1 c c 10 c

Figure 5.3 Frequency response plot for three types of low-pass Butterworth filters.
The Butterworth transfer function is useful in many low-pass filter
applications. Its magnitude is
H0
H( f )  (6.5)
1  ( f f c ) 2n
In which the integer n is the order of the filter and fc is the 3-dB cutoff frequency.
Substituting f = 0 yields H (0)  H 0 ; thus, H 0 is the DC gain magnitude. Plots of
this transfer function are illustrated in Figure 5.4. Notice that as the order of the
filter increases, the transfer function approaches that of an ideal low-pass filter.
H( f )
H0

1.0
Ideal low-pass filter
0.8 n=1

0.6 n=1
Figure 6.4 n=2
Transfer function 0.4
n=3
magnitude versus 0.2 n=4
frequency for low-pass f
0
Butterworth filters. 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 fc
 Two active filters similar to Figure 6.2(a) could be coupled together to give a
roll-off of −40 dB/decade. This would not be the most economical design, because
it would require two OP-AMPs. In Section 6.4, it is shown how one OP-AMP can
be used to build a Butterworth filter to give a −40 dB/decade roll-off. Then in
Section 6.5, a −40 dB/decade filter will be cascaded with a −20 dB/decade filter to
produce a −60 dB/decade filter.
Butterworth filters are not designed to keep a constant phase angle at the
cutoff frequency. A basic low-pass filter of −20 dB/decade has a phase angle of
−450 at c. A −40 dB/decade Butterworth filter has a phase angle of −900 at c
and a −60 dB/decade Butterworth filter has a phase angle of −1350 at c.
Therefore, for each increase of −20 dB/decade, the phase angle will increase by
−450 at c. We now proceed to a Butterworth filter that has a roll-off steeper than
−20 dB/decade.

40 DB/DECADE LOW-PASS BUTTERWORTH FILTER

The circuit of Figure 6.5(a) is one of the most commonly used low-pass filters. It
produces a roll-off of 40 dB/decade; that is, after the cutoff frequency, the
magnitude of ACL decreases by 40 dB as  increases to 10c. The solid line in
Figure 6.5(b) shows the actual frequency-response plot. Since the OP-AMP circuit
is basically a voltage follower (unity-gain amplifier), the voltage across C1 equals
output voltage, V0.
C2

Rf = 2R V0 20 Log 10 ACL (dB)

ACL 
Vs

+V
 1.0 0
301 0.707 −3
R1 = R R2 = R
+
+
+ V0 0.1 −20
C1 V
+
Vs V0 
 
−40
0.01 
0.1 c c 10 c
(a) Low-pass filter for a roll-off of −40 dB/decade. (b) Frequency response plot for the low-pass
filter of part (a).
Figure 6.5 Circuit and frequency response plot for a low-pass filter of −40 dB/decade.
Cutoff Frequency of 40 dB/Decade Low-Pass Butterworth Filter

The equivalent circuit of Figure 6.5(a) is shown in Figure 6.6. In the derivation of
cutoff frequency we will assume, R1 = R2 = R
Using voltage division for V0,
V0 1 jC1 1 jC1 1
  
Vx R2  1 jC1 R  1 jC1 1  jRC1

 Vx  (1  jRC1 )V0 V0 (6.6)

C2
Applying KCL at Vx , R1 = R R2 = R

Vx +
V  Vs Vx  V0 Vx  V0 C1 V0
 x   0 +
Vs
R1 1 j C 2 R2  
V  Vs Vx  V0 Vx  V0
 x   0
R 1 j C 2 R Figure 6.6
Equivalent circuit of Figure 6.5 (a).
Vx Vx V V V0 V
   x  s   0 0
R 1 j  C 2 R R 1 j C 2 R

 Vx  jRC 2Vx  Vx  Vs  jRC 2V0  V0  0

 ( 2  jRC 2 )Vx  Vs  jRC 2V0  V0  0 (6.7)

Putting the value of Vx from Equation (6.6) into Equation (6.7) we have,
 ( 2  jRC 2 )(1  jRC1 )V0  Vs  jRC 2V0  V0  0
 ( 2  jRC 2  2 jRC1   2 R 2C1C 2 )V0  Vs  jRC 2V0  V0  0
 ( 2  jRC 2  2 jRC1   2 R 2C1C 2  jRC 2  1)V0  Vs
 (1   2 R 2 C1C 2  2 jRC1 )V0  Vs
V 1
 0  ACL  2 2
Vs (1   R C1C 2 )  2 jRC1
1
 ACL  (6.8)
(1   2 R 2C1C 2 ) 2  (2RC1 ) 2
At  = 0, |ACL| = 1 = A0
At  = , |ACL| = 0
Therefore, the circuit of Figure 6.5(a) is a low-pass filter.
A 1
We know, at cutoff   c , ACL  0  . Now from Equation (6.8), we
2 2
obtain
1 1
 
(1   c2 R 2 C1C 2 ) 2  ( 2 c RC1 ) 2 2
 (1   c2 R 2 C1C 2 ) 2  ( 2c RC1 ) 2  2
 1  2 c2 R 2 C1C 2   c4 R 4C12 C 22  4 c2 R 2 C12  2
  c4 R 4 C12C 22  2 c2 R 2 C1 ( 2C1  C 2 )  1 (6.9)

To make the c2 term in Equation (6.9) to zero, we choose

2C1  C 2  0
 C2 = 2C1

Then Equation (6.9) becomes,

4c4 R 4C14  1
1
  c4 R 4C14 
4
1
 c2 R 2C12  (product of all square terms can not be negative)
2
1
  c RC1  (c, R and C1 all are positive and their product can not be
2
negative)
1 0.707
 c  
2 RC1 RC1
0.707
 fc  (6.10)
2RC1
Therefore, the cutoff frequency of a 40 dB/decade low-pass Butterworth filter is
given by Equation (6.10).

Design Procedures

The design of a low-pass filter similar to Figure 6.5(a) is accomplished in five

steps:
 1. Choose the cutoff frequency − either c or fc.
2. Pick capacitance C1, usually between 100 pF and 0.1 F.
3. Make C2= 2C1.
4. Calculate R from Equation (6.10).
5. Choose Rf = 2R.
40 dB/decade 20 dB/decade
C2
Rf2 = R

Rf1 = 2R
+V
+V 
 R3 = R 301
+
R1 = R R2 = R 301 + +
+
+ C3 V0 V V0
+ V01
C1 V  
+ V01 
 Vs 

(a) Low-pass filter for a roll-off of −60 dB/decade.

V0 20 Log 10 ACL (dB)

ACL 
Vs

1.0 0
0.707 −3
0.1 −20

0.01 −40

0.001
−60
0.1 c c 10 c 
(b) Plot of frequency response plot for the circuit of part (a).
Figure 6.7 Circuit and frequency response plot for a low-pass filter of −60 dB/decade.

60 DB/DECADE LOW-PASS BUTTERWORTH FILTER

The low-pass filter of circuit of Figure 6.7(a) is built using one low-pass filter of
40 dB/decade cascaded with another of 20 dB/decade to give an overall roll-off
of 60 dB/decade. The overall closed-loop gain ACL is the gain of the first filter
times the gain of the second filter, or
V V V
ACL  0  01  0 (6.11)
Vs Vs V01

The solid line in Figure 6.7(b) shows the actual frequency-response plot. Since the
OP-AMP circuit is basically a voltage follower (unity-gain amplifier), the voltage
across C1 equals V01 and the voltage across C3 equals output voltage, V0.

Cutoff Frequency of 60 dB/Decade Low-Pass Butterworth Filter

In the derivation of cutoff frequency we will assume, R1 = R2 = R3 = R. The

equivalent circuit of 40 dB/decade section of Figure 6.7(a) is shown in Figure
6.8. Using voltage division for V01,
V01 1 jC1 1 jC1 1
  
Vx R2  1 jC1 R  1 jC1 1  jRC1

 Vx  (1  jRC1 )V01 (6.12)

V01

Applying KCL at Vx , C2

Vx  Vs Vx  V01 Vx  V01 R1 = R R2 = R
   0
R1 1 j C 2 R2 Vx +
+ C1 V01
V  Vs Vx  V01 Vx  V01  Vs 
 x   0
R 1 j C 2 R
Figure 6.8
V Vx V V V01 V Equivalent circuit of
 x   x  s  01  0
R 1 j  C 2 R R 1 j C 2 R 40 dB/decade section.

 Vx  jRC 2Vx  Vx  Vs  jRC 2V01  V01  0

 ( 2  jRC 2 )Vx  Vs  jRC 2V01  V01  0 (6.13)

Putting the value of Vx from Equation (6.12) into Equation (6.13) we have,
 ( 2  jRC 2 )(1  jRC1 )V01  Vs  jRC 2V01  V01  0
  ( 2  jRC 2  2 jRC1   2 R 2 C1C 2 )V01  Vs  jRC 2V01  V01  0
 ( 2  jRC 2  2 jRC1   2 R 2 C1C 2  jRC 2  1)V01  Vs
 (1   2 R 2 C1C 2  2 jRC1 )V01  Vs
V 1
 01  2 2 R3 = R
(6.14)
Vs (1   R C1C 2 )  2 jRC1
+ +
V01 C3 V0
From Figure 6.9, we can write  

V0 1 / j C 3 1 / j C 3
 
V01 R3  1 / jC3 R  1 / jC3 Figure 6.9
Equivalent circuit of
V 1
 0  20 dB/decade section. (6.15)
V01 1  jRC3

Using Equations (6.14) and (6.15) we obtain the closed-loop gain from Equation
(6.11) as,
V V V
ACL  0  01  0
Vs Vs V01
1 1
 2 2

(1   R C1C 2 )  2 jRC1 (1  jRC3 )
1 1
Therefore, ACL   (6.16)
2 2 2 2
(1   R C1C 2 )  ( 2RC1 ) 1  (RC3 ) 2

At  = 0, |ACL| = 1 = A0
A0 1
We know, at cutoff   c , ACL   . Now from Equation (6.16), we
2 2
obtain
1 1 1
  
2 2 2 2 2
1  (c RC3 ) (1   R C1C 2 )  ( 2 c RC1 )
c
2

 
 1  (c RC3 ) 2 (1   c2 R 2 C1C 2 ) 2  ( 2 c RC1 ) 2  2
 (1   c2 R 2 C32 )(1  2 c2 R 2 C1C 2  c4 R 4C12 C 22  4 c2 R 2 C12 )  2
 1  2 c2 R 2 C1C 2  c4 R 4C12 C 22  4 c2 R 2 C12

  c2 R 2 C32  2 c4 R 4 C1C 2 C32   c6 R 6 C12 C 22C32  4 c4 R 4 C12C32  2

 c6 R 6C12C 22 C32   c4 R 4 C1 (C1C 22  2C 2C32  4C1C32 )

 (6.17)
  c2 R 2 ( 4C12  C32  2C1C 2 )  1

To make the c2 and c4 terms in Equation (6.17) to zero, we choose,
C1C22  2C 2C32  4C1C32  0 (6.18)
4C12  C32  2C1C2  0 (6.19)

From Equations (6.18) and (6.19) we have,

C1 = C3/2 and C2 = 2C3

Then Equation (6.17) becomes,

  c6 R 6C36  1
  c2 R 2 C32  1
 c RC3  1 (c, R and C3 all are positive and their product can not be
negative)
1
 c 
RC3
1
 fc  (6.20)
2RC3
Therefore, the cutoff frequency of a 60 dB/decade low-pass Butterworth filter is
given by Equation (6.20).

Design Procedures

The design of a low-pass filter similar to Figure 6.7(a) is accomplished in six

steps:
1. Choose the cutoff frequency − either c or fc.
2. Pick capacitance C3, usually between 0.001 F and 0.1 F.
3. Make C1= C3/2 and C2= 2C3.
4. Calculate R from Equation (6.20).
5. R1 = R2 = R3 = R.
 6. Choose Rf1 = 2R and Rf2 = R. For best results the value of R should be
between 10 k and 100 k. If the value of R is outside this range, you
should go back and pick a new value of C3.

Table 6.2 compares the magnitudes of ACL for the three low-pass filters
presented in this chapter. Note that the ACL for Figure 6.7(a) remains quite close
to 1 (0 dB) until the cutoff frequency, c; then the steep roll-off occurs.

Table 6.2. ACL for the low-pass filters of Figures 6.2(a), 6.5(a), and 6.7(a).
____________________________________________________________
 20 dB/decade; 40 dB/decade; 60 dB/decade;
Figure 6.2(a) Figure 6.5(a) Figure 6.7(a)
____________________________________________________________
0.1 c 1.0 1.0 1.0
0.25 c 0.97 0.998 0.999
0.5c 0.89 0.97 0.992
c 0.707 0.707 0.707
2 c 0.445 0.24 0.124
4 c 0.25 0.053 0.022
10 c 0.1 0.01 0.001
___________________________________________________________
The phase angles for the low-pass filter of Figure 6.7(a) range from 00 at  = 0 (0
dB condition) to 2700 as  approaches . Table 6.3 compares the phase angles
for the three low-pass filters.

Table 6.3. Phase angles for the low-pass filters of Figures 6.2(a), 6.5(a), and 6.7(a).
_______________________________________________________________
 20 dB/decade; 40 dB/decade; 60 dB/decade;
Figure 6.2(a) Figure 6.5(a) Figure 6.7(a)
_______________________________________________________________
0.1 c 60 80 120
0.25 c 140 210 290
0.5 c 270 430 600
c 450 900 1350
2 c 630 1370 2100
4 c 760 1430 2260
10 c 840 1720 2560
______________________________________________________________
BASIC HIGH-PASS FILTER

Introduction

The circuit of Figure 6.10(a) is a commonly used high-pass active filter. Compare
with the low-pass filter of Figure 6.2(a), note that C and R are interchanged in the
basic high-pass filter. The feedback resistor Rf is included to minimize DC offset.
Since the OP-AMP is connected as a unity-gain follower in Figure 6.10(a), the
output voltage V0 equals the voltage across R and is expressed by
1
V0  Vs (6.21)
1  j 1 RC 

Rf

+V

C 301
+
+ + V0
V0 V 
+ R
 Vs


(a) High-pass filter for a roll-off of +20 dB/decade.

V0
ACL  20 Log 10 ACL (dB)
Vs

1.0  0
0.707 −3
Slope =
+20 dB/decade Passband
0.1 −20

−40
0.01 
0.01 c 0.1 c c 10 c 
(b) Frequency response plot for the circuit of part (a).

Figure 6.10 High-pass filter and frequency response plot for a filter with a
+20 dB/decade roll-off.

Rearranging Equation (6.21) to obtain closed-loop gain ACL as given by

 V0 1
ACL   (6.22)
Vs 1  j 1 RC 
At very low frequencies, that is, as  approaches zero, ACL = 0, and at very high
frequencies, as  approaches infinity, ACL = 1. Therefore, circuit of Figure
6.10(a) is a high-pass filter.
Figure 6.10(b) is a plot of ACL versus  and shows that for frequencies greater
than the cutoff frequency, c, ACL increases at a rate of 20 dB/decade. The cutoff
frequency c is defined as that frequency of Vs where ACL is reduced to 0.707
1
times its high-frequency value. Let us assume that   . From Equation
RC
(6.22), we obtain ACL  1 2  0.707. Thus, the cutoff frequency of this filter
becomes
1
c   2f c . (6.23)
RC
Equation (6.23) may be rearranged to solve for R:
1 1
R  . (6.24)
c C 2f c C
The solid curve in Figure 6.10(b) shows how the magnitude of the actual
frequency response deviates from the straight dashed-line approximation in the
vicinity of c. At 0.01c, ACL  0.01 (-40 dB), and at 0.1c, ACL  0.1 (-20 dB).

Design Procedures for +20 dB/decade High-Pass Filter

The design of a high-pass filter similar to Figure 6.10(a) is accomplished in four

steps:
1. Choose the cutoff frequency − either c or fc.
2. Choose the capacitance C, usually between 0.001 and 0.1 F.
3. Calculate R from Equation (6.24).
4. Choose Rf = R.

A low-pass transfer function can be transformed into a high-pass function by

replacing f by  f c  f , where fc is the cutoff frequency. Making this substitution
2

in the low-pass Butterworth transfer function given in Equation (6.5) results in

 H0
H hp ( f )  (6.25)
1  ( fc f )2n
Clearly, this is a high-pass transfer function. Notice that as f approaches infinity,
H hp ( f ) approaches H0. On the other hand, as f approaches zero, H hp ( f )
approaches zero. At f = fc, H hp ( f )  H 0 2 . Therefore, fc is half-power, or 3-dB
cutoff frequency. This transfer function rolls off (asymptotically) at a rate of
n  20 dB/decade for frequencies below fc. Substituting f = 0 yields H hp (0)  H 0 ;
thus, H 0 is the DC gain magnitude. Normalized plots of H hp ( f ) are displayed in
Figure 6.11 for n = 1, 2, 3, and 4.
H hp ( f )
H0 Ideal high-pass filter

1.0

0.8 n=3
n=1 n=2
0.6

Figure 6.11 0.4

Normalized high-pass 0.2 n=4
Butterworth transfer f
0
function. 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 fc

+40 DB/DECADE HIGH-PASS BUTTERWORTH FILTER

The circuit of Figure 6.12(a) is to be designed as a high-pass Butterworth filter

with a roll-off of +40 dB/decade below the cutoff frequency c. To satisfy the
Butterworth criteria, the frequency response must be 0.707 at c and be 0 dB in
the pass band. These conditions will be met if the following design procedure is
followed:

Design Procedure for +40 dB/decade High-Pass Filter

1. Choose the cutoff frequency − either c or fc.

2. Let C1 = C2 = C and choose a convenient value (between 100 pF and 0.1
F).
3. Calculate R1 from
 1.414
R1  (6.26)
cC
R
4. Select R2  1 (6.27)
2
5. To minimize DC offset, let Rf  R1.

R2

Rf = R1

+V

C1 = C C2 = C 301
+
+
V V0
+
Vs R1 


(a) High-pass filter for a roll-off of +40 dB/decade.

V0
ACL  20 Log 10 ACL (dB)
Vs

1.0  0
0.707 −3
Slope =
+40 dB/decade Passband
0.1 −20

−40
0.01 
0.1 c c 10 c 
(b) Frequency response plot for the high-pass filter of part (a).

Figure 6.12 Circuit and frequency response for a +40 dB/decade high-pass
Butterworth filter.

EXERCISE 6.1
For the circuit of Figure 6.12(a), prove that the cutoff frequency is given by the
expression f c  1.414 2R1C .

+60 DB/DECADE HIGH-PASS BUTTERWORTH FILTER

As with the low-pass filter of Figure 6.7(a), a high-pass Butterworth filter of +60
dB/decade can be constructed by cascading a +40 dB/decade filter with a +20
dB/decade filter. This circuit (like the other high- and low-pass filters) is designed
as a Butterworth filter to have the frequency response in Figure 6.13(b). The
design steps for Figure 6.13(a) are as follows:

+40 dB/decade +20 dB/decade

R2
Rf2 = R3
Rf1 = R1
+V
+V 
C3 = C 301
 +
C1 = C C2 = C 301 +
+ V0
+ R3
V 
V V01
+ C1 
 Vs R1

(a) High-pass filter for a roll-off of +60 dB/decade.

V0
ACL  20 Log 10 ACL (dB)
Vs

1.0  0
0.707 −3
Slope =
0.1 +60 dB/decade −20

0.01 −40
−60
0.001 
0.1 c c 10 c 
(b) Frequency response plot for circuit of part (a).

Figure 6.13 Circuit and frequency response for a +60 dB/decade high-pass
Butterworth filter.
Design Procedure for +60 dB/decade High-Pass Filter

1. Choose the cutoff frequency − either c or fc.

2. Let C1 = C2 = C3 = C and choose a convenient value (between 100 pF and
0.1 F).
3. Calculate R3 from
1
R3  (6.28)
c C
4. Select R1  2R3 (6.29)
R
5. Select R2  3 (6.30)
2
6. To minimize DC offset, let Rf1  R1 and Rf2  R3 .

EXERCISE 6.2

For the circuit of Figure 6.13(a), prove that the cutoff frequency is given by
1
fc  .
2R3C

INTRODUCTION TO BANDPASS FILTERS

Frequency Response of Bandpass Filter

A bandpass filter is a frequency selector. It allows one to select or pass only one
particular band of frequencies from all other frequencies that may be present in a
circuit. Its normalized frequency response is shown in Figure 6.14. This type of
filter has a maximum gain at a resonant frequency fr. In this chapter all bandpass
filters will have a gain of 1 or 0 dB at fr. There is one frequency below fr where the
gain falls to 0.707. It is the lower cutoff frequency, fcL. At higher cutoff frequency,
fcH, the gain also equals 0.707, as in Figure 6.14.

Bandwidth

The range of frequencies between fcL and fcH is called bandwidth B, or

B  f cH  f cL (6.31)
The bandwidth is not exactly centered on the resonant frequency. (It is for this
reason that we use the historical name “resonant frequency” rather than “center
frequency” to describe fr.)
If you know the values for fcL and fcH, the resonant frequency can be found
from
f r  f cH f cL (6.32)
If you know the resonant frequency, fr, and bandwidth, B, cutoff frequencies can
be found from
B2 2 B
f cL   fr  (6.33)
4 2
f cH  f cL  B (6.34)

V0
AcL 
Vs
Resonant frequency fr

1.0 Bandwidth
B = fcHfcL
0.707

Frequency
fcL fr fcH

Figure 6.14 A bandpass filter has a maximum gain at resonant frequency fr.
The band of frequencies transmitted lies between fcL and fcH.

Quality Factor

The quality factor Q is defined as the ratio of resonant frequency to bandwidth, or

f
Q r (6.35)
B
Q is a measure of the bandpass filter’s selectivity. A high Q indicates that a filter
selects a smaller band of frequencies (more selective).
Narrow and Wideband Filters

A wideband filter has a bandwidth that is two or more times the resonant
frequency. That is, Q  0.5 for wideband filters. In general, wideband filters are
made by cascading a low-pass filter circuit with a high-pass filter circuit. This
topic is covered in the next section. A narrowband filter (Q > 0.5) can usually be
made with a single stage. This type of filter is presented in Section 6.12.

+V
7.5 k
0.01 F +V CA
CA 3140 +
3140 + V0
+ V 
15 k
V 0.05 F
Vs + 7.5 k
0.005 F

V0
AcL 
Vs
3000 Hz low-pass 300 Hz high-pass

1.0
0.707
Bandwidth
2700 Hz
V0 vs. Vs V0 vs. Vs
0.1

fr = 948 Hz
0.01
10 100 fcL 1000 fcH 104 105
Frequency f

Figure 6.15 A 3000-Hz second-order low-pass filter is connected with a

300-Hz high-pass filter to form a 300- to 3000-Hz bandpass voice filter.
NARROWBAND PASSBAND FILTERS

Narrowband filters exhibit the typical frequency response shown in Figure

6.16(a). The analysis and construction of narrowband filters is considerably
simplified if we stipulate that the narrowband filter will have a maximum gain of
1 or 0 dB at the resonant frequency fr. Equations (6.31) to (6.35) and bandpass
terms were presented in Section 6.10. They gave an introduction to (cascaded
pair) wideband filters. These equations and terms also apply to the narrowband
filters that follow.
V0
AcL 
Vs

Bandwidth
1.0 f
B r
Q
0.707

0.1 Frequency
0.1fr fcL fr fcH 10fr

(a) Typical frequency response curve of a narrowband filter

C = 0.015 F

2R = 42.42 k

C=
R = 21.21 k 0.015 F +15V
CA
Rr = 3140 +
Vs + + 15V
3.03 k V0


(b) Narrowband filter

Figure 6.16 Narrow bandpass filter circuit and its frequency response
for the component values shown; fr = 100 Hz, B = 500 Hz, Q = 2, fcL =
780 Hz, and fcH = 1280 Hz.
Narrowband Filter Circuit

A narrowband filter circuit uses only one OP-AMP, as shown in Figure 6.16(b).
The filter’s input resistance is established approximately by resistor R. If the
feedback resistor (2R) is made two times the input resistor R, the filter’s
maximum gain will be 1 or 0 dB at resonant frequency fr. By adjusting Rr one can
change (or exactly trim) the resonant frequency without changing the bandwidth
or gain.

Performance

The performance of the unity-gain narrowband filter in Figure 6.16(b) is

determined by only a few simple equations. The bandwidth B is hertz is
determined by resistor R and the two (matched) capacitors by
0.1591
B (6.36)
RC
f
where B r (6.37)
Q
Gain is a maximum of 1 at fr provided that feedback resistor 2R is twice the value
of input resistor R.
The resonant frequency fr is determined by resistor Rr according to
R
Rr  (6.38)
2Q 2  1
If component values are given for the circuit, its resonant frequency can be
calculated from
0.1125 R
fr  1 (6.39)
RC Rr

NOTCH FILTERS

Introduction

The notch or band-reject filter is named for the characteristic shape of its
frequency response curve in Figure 6.17. Unwanted frequencies are attenuated in
the stopband B. The desired frequencies are transmitted in the passband that lies
on either side of the notch.
 Notch filters usually have a passband gain of unity or 0 dB. The equations for
Q, B, fcL, fcH, and fr are identical to those of its associated bandpass filter. The
reasons for this last statement are presented next.
V0
AcL  AcL (dB)
Vs

Pass Pass
1.0 Stopband B 0

0.707 3

Figure 6.17 A notch filter

transmits frequencies in the
passband and rejects undesired
frequencies in the stopband.
Frequency
fcL fr fcH

Notch Filter Theory

As shown in Figure 6.18, a notch filter is made by subtracting the output of a
bandpass filter from the original signal. For frequencies in the notch filter’s
passband, the output of the bandpass filter section approaches zero. Therefore,
input Vs is transmitted via adder input resistor R1 to drive V0 to a value equal to
Vs. Thus V0 = Vs in both lower and upper passbands of the notch filter.
Suppose that the frequency of Vs is adjusted to resonant frequency fr of the
narrow bandpass filter component. (Note: fr of the bandpass sets the notch
frequency.) Vs will exit from the bandpass as Vs and then is inverted by R1 and R
to drive V0 to +Vs. However, Vs is transmitted via R2 to drive V0 to Vs. Thus V0
responds to both inputs of the adder and becomes V0 = Vs Vs = 0 V at fr.
R
Figure 5.16 Vs at fr

Narrow R1 = R +V
bandpass filter 
fr, Q, ACL = 1 Adder
Vs + R2 = R +
+ V0
V 

Figure 6.18 A notch filter is made by a circuit that subtracts

the output of a bandpass filter from the original signal.
 In practice, V0 approaches zero only at fr. The depth of the notch depends on
how closely the resistors and capacitors are matched in the bandpass filter and
judicious fine adjustment of resistor R1 at the inverting adder’s input. This
procedure is explained in Section 6.14.

120-Hz NOTCH FILTER

Need for a Notch Filter

In applications where low-level signals must be amplified, there may be present

one or more of an assortment of unwanted noise signals. Examples are 50-, 60-, or
400-Hz frequencies from power lines, 120-Hz ripple from full-wave rectifiers, or
even higher frequencies from regulated switching-type power supplies or clock
oscillators. If both signals and a signal-frequency noise component are passed
through a notch filter, only the desired signals will exit from the filter. The noise
frequency is “notched out”. As an example, let us make a notch filter to eliminate
120-Hz hum.

Statement of the Problem

The problem is to make a notch filter with a notch (resonant) frequency of fr = 120
Hz. Let us select a stopband of B = 12 Hz. Gain of the notch filter in the passband
will be unity (0 dB), so that the desired signals will be transmitted without
attenuations. We use Equation (6.37) to determine a value for Q that is required
by the notch filter:
f 120
Q r   10
B 12
This high value of Q means that (i) the notch and component bandpass filter will
have narrow bands with very sharp frequency-response curves, and (ii) the
bandwidth is essentially centered on the resonant frequency. Accordingly, this
filter will transmit all frequencies from 0 to (120  6) = 114 Hz, and also all
frequencies above (120 + 6) = 126 Hz. The notch filter will stop all frequencies
between 114 Hz and 126 Hz.

Procedure to Make a Notch Filter

The procedure to make a notch filter is performed in two steps:

1. Make a bandpass filter that has the same resonant frequency, bandwidth,
and consequently Q as the notch filter.
2. Connect the inverting adder of Figure 6.19 by selecting equal resistors for
R. Usually, R = 10 k.