ANALOG

FILTERS

By

C. Sauriol

Algonquin College
Ottawa, Ontario

Rev. March 24, 2003 ©

 Analog Filters
TABLE OF CONTENTS

SECTION 1 PASSIVE FILTERS

1. First-order low-pass filter-RC 2-4

2. First-order high-pass filter-RC 4-6
3. Second-order low-pass filter-RCRC 6-11
4. Second-order band-pass filter-RCRC 12-15
5. Second-order low-pass filter-RLC 16-18
6. Second-order high-pass filter-RLC 19-21

SECTION 2 FREQUENCY RESPONSE OF BASIC AMPLIFIERS

1. Open-loop response of op amp 22-23

2. Closed-loop response of non-inverting amplifier 24-25
3. Op amp circuits with resistive feedback networks 26-27
4. Transient response of amplifiers to a square wave 27

SECTION 3 ACTIVE FILTERS

1. First-order filter circuits 28-29
2. Second-order filter circuits 30-35
3. Transient resnponse of first and second-order filters 36-37
th
4. N Order Filters 38-41

Design example 1 Low-pass Butterworth filter 42-45

Design example 2 High-pass Butterworth filter 46-50
Design example 3 Low-pass Chebychev filter 51-54
Design example 4 High-pass Chebychev filter 55-58
Design example 5 Low-pass Bessel filter 59-63

Table of normalized low-pass filter poles 64

Pole conversion equations 65
Calculation of minimum order of filters 66
Effect of component variation on frequency response 67-68
Table of second-order transfer functions 69-70

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 1

 Analog Filters
SECTION 1 PASSIVE FILTERS .

1. First-order low-pass filter

R
Transfer Function
1
1 59 15 Vo ( S ) 1
F (S ) = = SC =
Vin(S)
C
Vo (S) Vin ( S ) 1 1 + (RC )S
10 nF R+
SC

1
V ( jω ) jω C 1 Imag
F ( jω ) = o = =
Vin ( jω ) 1 1 + jωCR
R+
jω C
jωCR
MAG (num) 1 1
F ( jω ) = = = ω)
MAG (den) 1 + jωCR 1 + (ωCR )
2
F(j
/ F ( jω ) = / num − / den = 0 − ATAN (ωCR ) θ
Real
1
Frequency response

o
F Gain (V/V) Gain (dB) Phase( )
Exact Exact Bode Approximation Exact Bode Approximation
−1
For F 〉 FC For 10 FC 〉 F 〉 0.1FC
 1 + (ωCR )2  20 × LOG F ( jω ) A − 20 LOG (F F ) − ATAN (ωCR )
  LF C − 45o × LOG (F 0.1FC )
10 1.0000 0.00 0 -0.573 0
20 0.9998 0.00 0 -1.146 0
40 0.9992 -0.01 0 -2.291 0
80 0.9968 -0.03 0 -4.574 0
100 0.9950 -0.04 0 -5.711 0
200 0.9806 -0.17 0 -11.310 -13.55
400 0.9285 -0.64 0 -21.801 -27.09
800 0.7809 -2.15 0 -38.660 -40.64
1000 0.7071 -3.01 0 -45.000 -45.00
2000 0.4472 -6.99 -6.02 -63.435 -58.55
4000 0.2425 -12.30 -12.04 -75.964 -72.09
8000 0.1240 -18.13 -18.06 -82.875 -85.64
10000 0.0995 -20.04 -20.00 -84.289 -90
20000 0.0499 -26.0314 -26.02 -87.138 -90
40000 0.0250 -32.0439 -32.04 -88.568 -90
80000 0.0125 -38.0625 -38.06 -89.284 -90
100000 0.0100 -40.0004 -40.00 -89.427 -90

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 2

 Analog Filters

Gain Response

10
Fc cutoff frequency
LF asymptote
0
HF asymptote
Slope
-20 dB/dec
Gain (dB)

-10
3 dB
1 dB
down
-20 down
1 dB
down
-30

-40
10 100 1000 10000 100000
Frequency (Hz)
Actual Gain Bode approximation

Phase Response

0.000 0.1 Fc
Bode approximation
-20.000 Fc o
-45 /decade
o
5.7 down
Phase (o)

-40.000 o
- 45 o
5.7 up
-60.000

-80.000
10 Fc
-100.000
10 100 1000 10000 100000
Frequency (Hz)

Actual Phase Bode approximation

Cutoff frequency: defined as the frequency where the attenuation is 3 dB with respect to the pass-band gain.
Here the pass-band is 0 to 1000 Hz and the pass-band gain is 0 dB, therefore when the gain drops to –3 dB
we reach the cutoff frequency. To calculate the cutoff frequency for a first-order denominator of the TF,
equate the real part to the imaginary part and solve for ω.

Vo ( jω ) 1 1 1 1
F ( jω ) = = ⇒ ωCR = 1 ⇒ ω = ω c = or Fc = = = 1000 Hz
Vin ( jω ) 1 + jωCR RC 2π RC 2π 15915 × 10n

ω1 first order
C. Sauriol Analog Filters Rev. 3/24/2003 © Page 3
 Analog Filters
Poles and Zeros

The roots of the numerator of the TF are called the zeros of the TF and the roots of the denominator are called
the poles because they make the TF equal to inifinity. The poles of the TF determine the cutoff frequency of
the filter.

For the above filter, the numerator has no roots, or no zeros and the denominator has one pole (root), that is
1 + RCS = 0 SP = -1/RC . The magnitude of SP is the cutoff frequency in r/s.

−1 1 1
Pole S P = ωC = SP = FC =
RC RC 2π RC
Now if we plot the gain and the phase against frequency for a sinewave input, we obtain the the frequency
response of the circuit. Replacing S with jω in the TF we obtain:

2. First-order high-pass filter

C
Transfer Function

V in ( S )
10 nF
R
F (S ) =
Vo ( S )
=
R
=
(RC )S =
S
1 + (RC )S
Vo (S )
1 59 15 Vin ( S ) 1 1
R+ S+
SC RC

Vo ( jω ) jω
F ( jω ) = = Imag
Vin ( jω )1
jω +
RC
MAG (num) jω ω jω
F ( jω ) = = =
MAG (den) 1  1 
2
jω + den
num

RC ω +
2

 RC 
−1
θ
 Real
 1  
2

F ( jω ) = 1 +  
1
  ωRC   RC
 
/ F ( jω ) = / num − / den = 90 o − ATAN (ωCR )

Cutoff frequency: defined as the frequency where the attenuation is 3 dB with respect to the pass-band gain.
Here the pass-band is 1000 Hz to inifinity and the pass-band gain is 0 dB, therefore when the gain drops to –3
dB we reach the cutoff frequency. To calculate the cutoff frequency for a first-order denominator of the TF,
equate the real part to the imaginary part and solve for ω.

Vo ( jω ) jω 1 1 1
F ( jω ) = = ⇒ ωCR = 1 ⇒ ω = ω c = or Fc = = = 1000 Hz
Vin ( jω ) 1 RC 2π RC 2π 15915 × 10n
jω +
CR
ω1 first order
C. Sauriol Analog Filters Rev. 3/24/2003 © Page 4
 Analog Filters

o
F Gain (V/V) Gain (dB) Phase( )
(Hz) Exact Exact Bode Approximation Exact Bode Approximation
−1
For 10 FC 〉 F 〉 0.1FC
 2  F 〈 FC
 1 +  1   For
90 o  F 
  20 × LOG F ( jω ) AHF − 20 LOG (FC F) 90 o − 45 o LOG 
  ωRC   − ATAN (ωCR )  0.1FC 
10 0.0100 -40.00 89.427 -40 90
20 0.0200 -33.98 88.854 -33.9794 90
40 0.0400 -27.97 87.709 -27.9588 90
80 0.0797 -21.97 85.426 -21.9382 90
100 0.0995 -20.04 84.289 -20 90
200 0.1961 -14.15 78.690 -13.9794 76.45
400 0.3714 -8.60 68.199 -7.9588 62.91
800 0.6247 -4.09 51.340 -1.9382 49.36
1000 0.7071 -3.01 45.000 0 45.00
2000 0.8944 -0.97 26.565 0.00 31.45
4000 0.9701 -0.26 14.036 0.00 17.91
8000 0.9923 -0.07 7.125 0.00 4.36
10000 0.9950 -0.04 5.711 0.00 0
20000 0.9988 -0.01 2.862 0.00 0
40000 0.9997 0.00 1.432 0.00 0
80000 0.9999 0.00 0.716 0.00 0
100000 1.0000 0.00 0.573 0.00 0

Gain Response

10
Fc cutoff frequency
0
LF asymptote
HF asymptote
Slope
+20 dB/dec
Gain (dB)

-10
3 dB
1 dB
-20 down
down
1 dB
down
-30

-40
10 100 1000 10000 100000
Frequency (Hz)
Actual Gain Bode approximation

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 5

 Analog Filters
Phase Response

90.000 0.1 Fc
Bode approximation
70.000 Fc o
-45 /decade
o
5.7 down
Phase (o)

50.000 o
- 45
o
5.7 up
30.000

10.000
10 Fc
-10.000
10 100 1000 10000 100000
Frequency (Hz)

Actual Phase Bode approximation

Poles and Zeros

Vo ( S ) S
F (S ) = = For the transfer function of the high-pass filter, the numerator has one root, or
Vin ( S ) 1 one zero and the denominator has one pole, or one root.
S+
RC
Zeros: numerator, Sz = 0
1
Poles: denominator S+ =0 SP = -1/RC . The magnitude of SP is the cutoff frequency in r/s.
RC
−1 1 1
Pole S P = ωC = SP = FC =
RC RC 2π RC

3. Second-order low-pass filter

T h even in 1
SC1 Vin ( S )
R1 R2
VTH = Vin ( S ) × =
X 1 1 + SC1 R1
R1 +
SC1
V in ( S) C1 C2 Vo (S) VTH is the open circuit voltage

X
ZTH is derived by replacing the source
with a short circuit – an ideal voltage
source has zero internal resistance.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 6

 Analog Filters
1 1
R1 × R1 ×
1 SC1 SC1 R1
Z TH = R1 = = =
SC1 1 SC1 R1 + 1 SC1 R1 + 1
R1 +
SC1 SC1
Z TH
R1 R2

V TH (S) =
V in(S) C2 Vo(S)
C1
1+SC 1R 1

1 1
SC 2 Vin ( S ) SC 2
Vo ( S ) = VTH ( S ) × = ×
1 R1 1 + SC1 R1 1 + SC1 R1 + R2 (SC 2 × (1 + SC1 R1 )) + SC 2 R1
+ R2 +
SC 2 1 + SC1 R1 SC 2 (1 + SC1 R1 )
V (S ) 1 SC 2 (1 + SC1 R1 )
F (S ) = o = ×
Vin ( S ) 1 + SC1 R1 SC 2 (1 + SC1 R1 + R2 (SC 2 × (1 + SC1 R1 )) + SC 2 R1 )
1
V (S ) R1C1 R2 C 2
F (S ) = o =
Vin ( S )  1 1 1  1
S 2 + S  + +  +
 R1C1 R2 C 2 R2 C1  R1C1 R2 C 2
To find the frequency response of the above TF, it is easier if one finds the roots of the denominator (the poles
of the TF) and expresses the TF as follows:

Vo ( S ) − p1 × − p 2 1
F (S ) = = =
Vin ( S ) (S − p1 )× (S − p 2 )  S   S 
1 −  × 1 − 
 p1   p 2 
Numerical example
R1 R2 Let R1 = R2 = R and C1 = C2 = C
The TF simplifies to the form shown below
1 591 5 1 591 5 where the poles yield the two break
frequencies of each first-order term of the
Vin (S ) C1 10 n F C2 10 n F Vo (S ) denominator.

1 1
V (S ) R C2 2 2 2
R C 0.382 2.618
F (S ) = o = = ⇒ p1 = − p2 = −
Vin ( S )  3  1  0.382   2.618  RC RC
S + S
2
 + 2 2 S + ×S + 
 RC  R C  RC   RC 

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 7

 Analog Filters
0.382 0.382 2.618 2.618
ω c1 = p1 = = = 2400.25 r / s ω c 2 = p 2 = = = 16450 r / s
RC 15915 × 10n RC 15915 × 10n
In Hz we have : FC1 = 382 Hz and FC2 = 2618.1 Hz where FC1 × FC 2 = 1000 Hz
1
V (S ) R C2 2 39.48 × 10 6
F (S ) = o = = 2 ⇒ p1 = −2400 p 2 = −16450
Vin ( S )  3  1 S + 18850 S + 39.4809 × 10 6
S + S
2
+ 2 2
 RC  R C
Vo ( S ) 39.48 × 10 6 39.48 × 10 6 1
F (S ) = = = =
Vin ( S ) (S + 2400 )× (S + 16450 )
2400 × 16450 ×
(S + 2400) × (S + 16450)  S   S 
1 +  × 1 + 
2400 16450  2400   16450 
V o ( jω ) 1 1
F ( jω ) = = ×
Vin ( jω )  jω   jω 
1 +  1 + 
 2400   16450 
0 dB -2 0 0 dB -2 0
dB dB
/d e /d e
c c

382 Hz 2618.1 Hz

o o
0 0
-45 o 3820 Hz -45 o 26181 Hz
/d ec /d ec

38.2 Hz o 261.8 Hz o
-90 -90
The individual Bode plots for each term is shown above. The overall Bode plot approximation is simply the
addition of the individual Bode plots.

1 1 1
20 × LOG F ( jω ) = 20 × LOG = 20 × LOG × 20 × LOG
 jω  jω   jω   jω 
1 + 1 +  1 +  1 + 
 2400  16450   2400   16450 

st
1 term in dB 2nd term in dB

     
     
/ F ( jω ) = ANG  1
×
1  = ANG  1  + ANG  1 
 jω   jω    jω    jω  
 1 +  1 +   1 +   1 + 
  2400   16450     2400     16450  

st nd
Angle of 1 term angle of 2 term
-ATAN(ω/2400) -ATAN(ω/16450)
C. Sauriol Analog Filters Rev. 3/24/2003 © Page 8
 Analog Filters
Construction of overall Bode plot – gain response

382 Hz 2618 Hz To calculate the number decades,

use:
0 dB
-20 N dec = LOG(F2)-LOG(F1)
dB
/de N dec = LOG(F2 / F1)
c
N dec = LOG(2618 / 382)
N dec = 0.836

0 dB ∆dB = slope x N dec

-20 ∆dB = -20 dB/dec x 0.836 dec
dB ∆dB = - 16.72 dB
/ de
c
The LF gain is obviously the sum
of the two individual LF gains.

0 dB To construct the overall Bode

-20 plot, add the slopes in each
16.72 dB

dB
/de region of the frequency
Overall gain response c
response.
Bode approximation
0.836 dec
63.28 dB

-4
0
dB
/d
ec

-80 dB
1.582 dec
100 kHz
Note: the actual gain is 3 dB down from Bode plot at each break frequency if they are at least one
decade apart. If the change in slope is 40 dB/dec ( from –20 dB/dec to –60 dB/dec), then the
actual gain is 6 dB down at the break frequency.
The actual gain is 1 dB down from Bode plot at 0.5 Fc and 2 Fc if Fc1 and Fc2 are at least one
decade apart. If the change in slope is 40 dB/dec ( from –20 dB/dec to –60 dB/dec), then the
actual gain is 2 dB down at 0.5 Fc and 2 Fc.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 9

 Analog Filters
Construction of overall Bode plot – phase response

38.2 Hz 261.8 Hz 3820 Hz 26181 Hz

0o

-45 o
/ de c

0o
-45 o
/ de c -90o
-45 o
/ de c

0o
-45 o -90o
/ de c
Overall
phase response -37.62o
Bode -90
o
approximation /d
ec

-142.4o -45 o
/ de c

0.836 1.164 0.836 -180o

decade decade decade

To construct the overall phase response simply add the slopes in each region of the frequency response
And add the individual phase angles to compute the LF and the HF plateaus.
o
The actual or exact phase response always cuts inside the breakpoints and is 5.7 off if the breakpoints are at
o o
least one decade apart and if the change in slope is 45 /dec. If the slope change is 90 /dec, say from –
o o o
45 /dec to –135 /dec, then the actual phase is 11.4 off inside the breakpoint.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 10

 Analog Filters
Actual frequency response and Bode approximation

Gain Response

0
Gain (dB)

-20

-40
10 100 1000 10000
Frequency (Hz)
` Bode approximation Exact response

Phase Response

-45
Phase (o)

-90

-135

-180
10 100 1000 10000 100000
Frequency (Hz)

Bode approximation Exact response

As you can see, the phase response has a Bode approximation that is very close to the actual response
0
except at the first and last breakpoints where it is off by 5.7

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 11

 Analog Filters
4. Second-order band-pass filter
T h e ve ni n
C1 R1 Vin ( S ) × SC1 R1
R2 VTH = Vin ( S ) × =
X 1 1 + SC1 R1
R1 +
R1
SC1
Vi n(S ) C2 Vo (S )
VTH is the open circuit voltage
X
ZTH is derived by replacing the source with a
1 1 short circuit – an ideal voltage source has zero
R1 × R1 × internal resistance.
1 SC1 SC1 R1
Z TH = R1 = = =
SC1 1 SC1 R1 + 1 SC1 R1 + 1
R1 +
SC1 SC1

R1 R2

VTH(S) =
Vin ( S ) × SC1 R1
C2 V o (S )
1 + SC1 R1 C1

1 1
SC 2 V ( S ) × SC1 R1 SC 2
Vo ( S ) = VTH ( S ) × = in ×
1 R1 1 + SC1 R1 1 + SC1 R1 + R2 (SC 2 × (1 + SC1 R1 )) + SC 2 R1
+ R2 +
SC 2 1 + SC1 R1 SC 2 (1 + SC1 R1 )

Vo ( S ) SC1 R1 SC 2 (1 + SC1 R1 )
F (S ) = = ×
Vin ( S ) 1 + SC1 R1 SC 2 (1 + SC1 R1 + R2 (SC 2 × (1 + SC1 R1 )) + SC 2 R1 )

S Amax max gain at Fcen

V (S ) R2 C 2 A ∆ω S
F (S ) = o = = 2 max
Vin ( S )  1 1 1  1 S + ∆ω S + ω cen
2
ω cen = ω C1 × ω C 2
S 2 + S  + +  +
 R1C1 R2 C 2 R2 C1  R1C1 R2 C 2 ∆ω = ω C 2 − ω C1
∆ω is the 3 dB BW
From the above TF we can derive the following design equations:
2
C2 1  Fcen  Amax ∆ω 1
= − 1 −   〉 0 C1 R1 = C 2 R2 =
C1 Amax  Amax ∆F  ω cen
2
Amax ∆ω
To do the Bode plot of the above TF, it is
easier if one finds the roots of the
S
denominator and expresses the TF as shown V (S ) R2 C 2 SR1C1
beside. F (S ) = o = =
Vin ( S ) (S − p1 )× (S − p 2 )  S   S 
1 −  × 1 − 
NOTE: the 3 dB cutoff frequencies are equal
to the pole frequencies only if they are at
 p1   p 2 
least 2 decades apart.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 12

 Analog Filters
Band-pass filter – Equal R’s and C’s

S S 0.382
V (S ) A ∆ω S p1 = −
F (S ) = o = RC = 2 max = RC RC
Vin ( S )  3  1 S + ∆ω S + ω cen 
2
0.382   2.618  p2 = −
2.618
S 2 + S + 2 2 S + ×S + 
 RC  R C  RC   RC  RC
Z1 = 0

0
20 LOG(0.382)
Fp1 Fp2
-10 0.382/(2πRC) 0.382/(2πRC)
Gain (dB)

20 LOG(1/3)
at Fcen=1/(2πRC)
-20

3 dB bandwidth
-30 3/(2πRC)

+20 dB/dec -20 dB/dec

-40
100 1000 10000 100000 1000000
Frequency (Hz)

90

60
PHASE (o)

30 o
0 phase shift
at Fcen
0

-30

-60

-90
100 1000 10000 100000 1000000
Frequency (Hz)

Notice that the 3 dB attenuation frequencies are not equal to the pole frequencies, this occurs when the pole
frequencies are not far enough from each other – when the pole frequencies are at least 1.5 to 2 decades
apart, for unequal R’s and C’s, then they correspond to the 3 dB attenuation frequencies which are normally
0
used to define the bandwidth. Also notice that the exact gain is 1/3 V/V and the phaseshift is exactly 0 at the
center frequency which is given by Fcen =1/(2πRC). To find the 3 dB attenuation frequencies one must solve
the following equations:
1 3
ω C1 × ω C 2 = ω CEN
2
= and ∆ω = ω C 2 − ω C1 =
⇒ ω 22 − ∆ω × ω 2 − ω CEN
2
=0
RC RC
Solve for the roots of the quadratic equation and obtain ωC2 and ωC1 – use absolute value if root is –ve.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 13

 Analog Filters
Band-pass filter design example

C1 R2
Design a second-order band-pass filter with cutoff frequencies
of 1 kHz and 9 kHz and a maximum gain of 0.2V/V. Once
Vin(S ) R1 C2 V o(S )
designed, find the transfer function and then calculate the poles
and graph the actual gain response and the Bode
approximation. Use the following design equations.

2
C2 1  Fcen  Amax ∆ω 1
= − 1 −   〉 0 C1 R1 = C 2 R2 =
C1 Amax  Amax ∆F  ω cen
2
Amax ∆ω

∆F = FC2 – FC1 = 9k – 1k = 8k
1/2 1/2
Fcen = (FC2 FC1) = (9k*1k) = 3 kHz

0.2 × 2π × 8000
2
C2 1  3k  1
= −1−   = 0.4843 C1 R1 = = 28.29 µ C 2 R2 = = 99.47 µ
C1 0.2  0.2 × 8k  (2π × 3000 )2 0.2 × 2π × 8000

C1 std C2 theo C2 std R1 theo R1 std R2 theo R2 std

0.4843*C1 28.28µµ/C1 28.28µµ/C2
1.00E-08 4.84E-09 4.70E-09 2829 2700 21164 22000
1.10E-08 5.33E-09 5.10E-09 2572 2700 19504 20000
1.20E-08 5.81E-09 5.60E-09 2358 2400 17763 18000
1.30E-08 6.30E-09 6.20E-09 2176 2200 16044 16000
1.50E-08 7.27E-09 7.50E-09 1886 1800 13263 13000
1.60E-08 7.75E-09 7.50E-09 1768 1800 13263 13000

Final selection: R1 = 2200 C1 = 13 nF R2 = 16k C2 = 6.2 nF

S
R2 C 2 10081× S A ∆ω S
F (S ) = = = 2 max
 1 1 1  1 S + 49853 × S + 352.47 × 10
2 6
S + ∆ω S + ω cen
2

S 2 + S  + +  +
 R1C1 R2 C 2 R2C1  R1C1 R2 C 2
From the TF coefficients, we have the following results:
1 1 49853 10081
Fcen = × = 2988 Hz ∆F = = 7934 Hz Amax = = 0.2022 V / V
2π 2200 *13n *16k * 6.2n 2π 49853

Actual cutoff frequencies (at 3 dB attenuation points): solve for roots of F 2 − ∆F × F − FCEN
2
=0
F 2 − 7634 × F − 2988 2 = 0 ⇒ FC1 = 997.4 Hz and FC 2 = 8951.4 Hz

Poles and zeros: z1 = 0 p1 = -8529.54 p2 = -41323.46

Pole frequencies: F1 = 1357.5 Hz F2 = 6576.8 Hz

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 14

 Analog Filters
We can re-write the TF in terms of the poles found above:

10081 × S 10081× S
F (S ) = =
(S + 8529.54)× (S + 41323.46) 
8529.54 × 41323.46 × 
S  
+ 1 × 
S 
+ 1
 8529.54   41323.46 
(
F ( S ) = 28.6 × 10 −6 × S ) ×
1
×
1
 S   S 
 + 1  + 1
 8529.54   41323.46 
÷øöçèæ+

1 1357.5 Hz 6576.8 Hz
S 0 dB
1 -20
8529 .54 dB
/de
c

1
S 0 dB
1
41323.46 - 20
dB
/de
-12.27 dB c

-12.27 dB
(28.6 10 −6 S ) dec -20
d B/ Resulting d B/d
+ 20 ec
Bode plot

-10
1357.5 Hz -12.27 dB 6578.8 Hz
-13.88 dB
-15
G
A
-20 2988 Hz
I
N -25 997.4 Hz 8951.4 Hz

-30
(dB)
+20 dB/dec -20 dB/dec
-35

-40
100 1000 10000 100000
Frequency (Hz)
Bode approx Exact Gain

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 15

 Analog Filters
5. Second-order low-pass filter – RLC circuit

R L Transfer Function
1.1 5 1 mH
1 1
14.47
V (S ) SC LC
C V o (S ) F (S ) = o = =
V in (S )
Vin ( S )
R + SL +
1  
R 1
0 .2 2 uF S2 +  S +
SC L LC
3.94914 × 10 9
F (S ) = 2
S + 12572 S + 3.94914 × 10 9
Aoω n2 1 1
F (S ) = 2 where ζ = and ω n = = 62842 r / s and Ao = 1 → LF gain
S + 2ζω n S + ω n2 2Q LC

1 1 R 14.47 1
Fn = = = 10 kHz ζ = = = 0.1 Q= =5
2π LC 2π 1.151m × 0.22 µ 2ω n L 2 × 62842 × 1.151m 2ζ

Poles and zeros

The above TF has two complex poles (second-order denominator) which will result in frequency response
nd
quite different from the 2 order TF seen before with the double RC network where the poles were real. Using
the equation to find the roots of the TF denominator, we have:

− 12572 ± 12572 2 − 4 × 1 × 3.94914 × 10 9

p1 and p 2 = = −6285.84 ± j 62526.98 = 62842 / ± 95.74066 o
2 ×1

If your calculator does not handle complex square roots, calculate ζ and ωn from the TF denomnator
coefficients and then use the following formulas:

− b ± b 2 − 4ac
If ζ < 1, the poles are complex p1, 2 = = ω n / 180 ± a cos(ζ )
2a

− b ± b 2 − 4ac
If ζ > 1, the poles are real p1, 2 = = −ζω n ± ω n ζ 2 − 1
2a
Definition of ζ and ωn

ωn is the natural undamped frequency of oscillation – no damping occurs when ζ = 0 or R = 0 - that is the
frequency at which the circuit would naturally oscillate if there were no loss of energy in the circuit.

ζ is called the damping factor and is an indication of the energy loss in the circuit and as we will see later, it
determines how fast oscillations are damped or diminished.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 16

 Analog Filters
Frequency response
Aoω n2 Aoω n2 Aoω n2
F (S ) = F ( jω ) = =
S 2 + 2ζω n S + ω n2 ( jω )2 + 2ζω n jω + ω n2 ( )
ω n2 − ω 2 + j (2ζω nω )
Aoω n2  2ζω ω 
F ( jω ) = / F ( jω ) = − ATAN  2 n 2 

(ω 2
n −ω2 ) + (2ζω ω )
2
n
2
ωn − ω 
nd
Using the above formulas, we can plot the gain and phase response of a 2 order low-pass filter.

Gain Response
20
Resonant peaks
Gain (dB)

0 HF asymptote
-40 dB/dec

-20 dB/dec
asymptote
-20
for ζ > 1
only
Fn = 1000 Hz

-40
1000 10000 100000
Frequency (Hz)

ζ = 0.1 ζ = 0.2 ζ = 0.5 ζ = 0.707 ζ=1 ζ=2

Phase Response
0
No Bode asymptote
except for ζ > 1.
-30 Transition gets steeper as
ζ gets smaller.
PHASE (o)

-60

-90

-120 Fn = 1000 Hz
o
θ = -90
-150

-180
1000 10000 100000
Frequency (Hz)

ζ = 0.1 ζ = 0.2 ζ = 0.5 ζ = 0.707 ζ=1 ζ=2

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 17

 Analog Filters
Bode Plot

Only the gain response has a Bode plot approximation which has a break frequency equal to Fn , that is at the
resonant frequency of the RLC circuit. After Fn, the Bode plot rolls down directly at –40 dB/dec without going
through an intermediary slope of –20 dB/dec when ζ < 1. If ζ >1, then the poles of the TF are real and
determine the two break frequencies of the Bode plot and we have two slopes, -20 dB/dec and –40 dB/dec.

The phase response will have a Bode plot approximation only if ζ > 1, otherwise the actual phase response
has to be computed and plotted with a computer or a graphing calculator.

Resonant Peak

The gain response exhibits a peak only when ζ < 0.707. To find the peak of F ( jω ) , we can use F ( jω ) 2

and differentiate it wrt ω and equate the derivative to zero and solve for ω.
Ao2ω n4
d
d F ( jω )
2

=
(ω 2
n −ω 2 ) + (2ζω ω )
2
n
2

= 0 → solve for ω → ω peak = ω n 1 − 2ζ 2

dω dω

F ( jω peak ) =
2ζ 1 − ζ
Ao
2
(
→ Peak (dB) = −20 × LOG 2ζ 1 − ζ 2 )
The above equation gives the magnitude of the peak above the LF gain (Ao).

If we measure the actual peak we can then correlate it to a ζ value by solving the peak (dB) equation for ζ and

1 − 1 − 10 −0.1× peak ( dB )
we obtain: ζ meas =
2

For small ζ values – that is ζ < 0.3 – the peak(dB) equation can be approximated with:
Peak (dB) ≈ −20 × LOG (2ζ ) = 20 × LOG (2ζ ) = 20 × LOG (Q )
−1

Measurement of Fn
o
We can measure Fn by measuring the frequency at which the phase shift between Vin and Vo is –90 .
st
We cannot use the 3 dB attenuation method used for 1 order filters to measure the cutoff frequency because
Ao
the actual gain at Fn depends on the damping factor and is given by F ( jω peak ) = . If we vary the
2ζ
frequency until the gain reaches Ao/(2ζ) that frequency will be Fn.

Aoω n2 Aoω n2 Aoω n2

F ( jω ) = → = F ( jω n ) = =
( jω )2 + 2ζω n jω + ω n2 ( jω n )2 + 2ζω n jω n + ω n2 − ω n2 + j 2ζω nω n + ω n2
Aoω n2 A − jAo Ao
F ( jω n ) = = o = = / − 90 0 = QAo / − 90 0
j 2ζω nω n j 2ζ 2ζ 2ζ

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 18

 Analog Filters
6. Second-order high-pass filter – RLC circuit
R C
Transfer Function
14.47 Vo ( S ) SL S2
0 .2 2 uF L F (S ) = = =
1 .1 51 V o (S ) Vin ( S )
R + SL +
1 R 1
V in(S )
mH S2 +  S +
SC L LC
S2
F (S ) =
S 2 + 12572 S + 3.94914 × 10 9
Ao S 2 1 1
F (S ) = where ζ = and ω n = = 62842 r / s and Ao = 1 → HF gain
S 2 + 2ζω n S + ω n2 2Q LC

1 1 R 14.47 1
Fn = = = 10 kHz ζ = = = 0.1 Q= =5
2π LC 2π 1.151m × 0.22 µ 2ω n L 2 × 62842 × 1.151m 2ζ

Poles and zeros

The above TF has two zeros and two poles, that is two roots in the numerator and two roots in the
2
denominator. The zeros are: S = 0, Sz1 = Sz2 = 0 or Z1 = Z2 = 0
nd
The two complex poles will result in frequency response quite different from the 2 order TF seen before with
the double RC network where the poles were real. Using the equation to find the roots of the TF denominator,
we have:

− 12572 ± 12572 2 − 4 × 1 × 3.94914 × 10 9

p1 and p 2 = = −6285.84 ± j 62526.98 = 62842 / ± 95.74066 o
2 ×1

If your calculator does not handle complex square roots, calculate ζ and ωn from the TF denominator
coefficients and then use the following formulas:

If ζ < 1, the poles are complex Pole-zero diagram

− b ± b 2 − 4ac
p1, 2 = = ω n / 180 ± a cos(ζ ) Imag
2a
If ζ > 1, the poles are real p1
+j62527
− b ± b 2 − 4ac
p1, 2 = = −ζω n ± ω n ζ 2 − 1
2a
A graphical representation of the poles and -6286 z1 z2
zeros in the complex plane is called a pole-zero
Real
diagram.

-j62527
p2

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 19

 Analog Filters
Frequency response
Gain Response
Using the Ao S 2 Ao ( jω ) − Aoω 2
2

F (S ) = F ( jω ) = =
formulas
beside, we S 2 + 2ζω n S + ω n2 ( jω )2 + 2ζω n jω + ω n2 ( )
ω n2 − ω 2 + j (2ζω nω )
can plot the
gain and Aoω 2  2ζω ω 
F ( jω ) = / F ( jω ) = 180 o − ATAN  2 n 2 
phase
responses (ω 2
n −ω 2 ) + (2ζω ω )
2
n
2
 ωn −ω 

20
HF asymptote Resonant peaks
+40 dB/dec
Gain (dB)

+20 dB/dec
-20 asymptote
for ζ > 1
Fn = 1000 Hz only

-40
1000 10000 100000
Frequency (Hz)

ζ = 0.1 ζ = 0.2 ζ = 0.5 ζ = 0.707 ζ=1 ζ=2

Phase Response
180
No Bode asymptote
150 except for ζ > 1.
Transition gets steeper as
ζ gets smaller.
PHASE (o)

120

90

60
Fn = 1000 Hz
o
θ = +90
30

0
1000 10000 100000
Frequency (Hz)

ζ = 0.1 ζ = 0.2 ζ = 0.5 ζ = 0.707 ζ=1 ζ=2

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 20

 Analog Filters
Bode Plot

Only the gain response has a Bode plot approximation which has a break frequency equal to Fn , that is at the
resonant frequency of the RLC circuit. Before Fn, the Bode plot rolls down directly at +40 dB/dec without going
through an intermediary slope of +20 dB/dec when ζ < 1. If ζ >1, then the poles of the TF are real and
determine the two break frequencies of the Bode plot and we have two slopes, +20 dB/dec and +40 dB/dec.

The phase response will have a Bode plot approximation only if ζ > 1, otherwise the actual phase response
has to be computed and plotted with a computer or a graphing calculator.

Resonant Peak

The gain response exhibits a peak only when ζ < 0.707. To find the peak of F ( jω ) , we can use F ( jω ) 2

and differentiate it wrt ω and equate the derivative to zero and solve for ω.
Ao2ω 4
d
d F ( jω )
2

=
(ω 2
n −ω 2 ) + (2ζω ω )
2
n
2

= 0 → solve for ω → ω peak =

ωn
dω dω 1 − 2ζ 2

F ( jω peak ) =
2ζ 1 − ζ
Ao
2
(
→ Peak (dB) = −20 × LOG 2ζ 1 − ζ 2 )
The above equation gives the magnitude of the peak above the HF gain (Ao).

If we measure the actual peak we can then correlate it to a ζ value by solving the peak (dB) equation for ζ and

1 − 1 − 10 −0.1× peak ( dB )
we obtain: ζ meas =
2

For small ζ values – that is ζ < 0.3 – the peak(dB) equation can be approximated with:
Peak (dB) ≈ −20 × LOG (2ζ ) = 20 × LOG (2ζ ) = 20 × LOG (Q )
−1

Measurement of Fn
o
We can measure Fn by measuring the frequency at which the phase shift between Vin and Vo is +90 .
st
We cannot use the 3 dB attenuation method used for 1 order filters to measure the cutoff frequency because
Ao
the actual gain at Fn depends on the damping factor and is given by F ( jω peak ) = . If we vary the
2ζ
frequency until the gain reaches Ao/(2ζ) that frequency will be Fn.

− Aoω 2 − Aoω n2 − Aoω n2

F ( jω ) = → = F ( jω n ) = =
( jω )2 + 2ζω n jω + ω n2 ( jω n )2 + 2ζω n jω n + ω n2 − ω n2 + j 2ζω nω n + ω n2
− Aoω n2 − Ao + jAo Ao
F ( jω n ) = = = = / + 90 0 = QAo / + 90 0
j 2ζω nω n j 2ζ 2ζ 2ζ

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 21

 Analog Filters

SECTION 2 FREQUENCY RESPONSE

OF BASIC AM PLIFIERS

1. OPEN-LOOP RESPONSE OF OP AMP

F C: 3 dB attenuation cutoff G
A
frequency I 20 Log(A o)

Ao : low frequency or DC N 3 dB

differential gain (dB) Actual gain -2

0
dB
/d
ec
Ao
AV ( S ) =
 S  0 dB
1+  Fc GBW F (rad/s)

 ωC 

Most op amps exhibit the above frequency response because they are internally stabilised or compensated to
prevent self-oscillations when negative feedback is used. This stability measure is generally called frequency
compensation. Stability of the negative feedback loop is achieved in most op amp circuits by rolling down the
gain response at -20 dB/dec right through the 0 dB level. In some circuits, this compensation method is not
sufficient to stabilise against self-oscillations.

Bode plot

Ao Ao
AV ( jω ) = ⇒ AV ( jF ) =
 jω   jF 
1 +  1+ 
 ωC   FC 

LF asymptote: F << FC , AV ≈ Ao
20 log (Ao)

HF asymptote: F >> FC , AV ≈ Ao FC/jF -20 dB/dec

The two asymptotes intersect at the frequency where the gains given by the above equations are equal, that
is: AV = Ao = Ao FC / F ⇒ F = FC NOTE: magnitude of jF is simply F.

Gain-bandwidth product

The HF asymptote is AV = Ao FC / F ⇒ AV × F = Ao FC = cons tan t

The above result shows that the gain x frequency product is a constant when a point lies on the -20 dB/dec
part of the response. This result only occurs when the slope of the response is exactly -20 dB/dec, not +20
dB/dec, not -40 dB/dec, etc. The constant gain-frequency product is usually called the gain-bandwidth product
or GBW in short.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 22

 Analog Filters
Example: 741 op amp

Op amp data: GBW = 1 MHz typical, LF gain = 100 000 V/V typical.

A) Draw the open loop response of the op amp.

First calculate all relevant points of the response.

LF gain: 20 log (105) = 100 dB

Cutoff frequency: FC x 105 = GBW = 1M, therefore FC = 1M /105 = 10 Hz.
0 dB crossing: 1 x F0dB = GBW = 1M, therefore F0dB = 1M/1 = 1 MHz.

100

G 80 -2
0
A dB
I 60 /d
ec
N
40

(dB) 20

0
10

100

1K

10K

100K

1M
F (Hz)

B) Find the op amp gain at 1 kHz, 10 kHz and 35 kHz using the GBW product.
Gain at 1 kHz : AV = GBW/F = 1M/1k = 1000 or 60 dB
Gain at 10 kHz : AV = GBW/F = 1M/10k = 100 or 40 dB
Gain at 35 kHz : AV = GBW/F = 1M/35k = 28,57 or 29,12 dB

NOTE: For most op amps the gain response has the above shape but the numbers will be different.
The decade: the decade is a logarythmic unit of frequency span defined as follows:

#dec = log F2 - lof F1 = log (F2/F1)

If F1=100 Hz and F2 = 100 kHz, #dec = log(100k/100) = log(1000) = 3 decades

If F1=33 Hz and F2 = 455 Hz, #dec = log(455/33) = log(13,79) = 1,14 decades

Example: If we have two points lying on a

slope of -40 dB/dec and one of the points lies +24 dB
at (120 Hz, 24 dB), determine the gain of the
second point lying at 1800 Hz.
47,04 dB -40 dB/dec
#dec = log(1800/120) = 1,176

slope = ∆dB/∆dec, therefore -23,04 dB

∆dB=slope x ∆dec=-40 dB/dec x 1,176dec
∆dB=-47,04 dB
AV at 1800 Hz = 24 - 47,04 = -23,04 dB
120 1,176 1800
Hz decade Hz

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 23

 Analog Filters
2. CLOSED-LOOP RESPONSE OF NON-INVERTING AMPLIFIER

RX 0A
Vi n
Vo G AV op amp
A -2
0
IF dB
RE IF RF I /d
ec AVF
N
- V F +
20 LOG(AVFideal)
R 1 (dB)
AVF ideal = 1 + F =
RE β V
V RE BW F (Hz)
βV = F =
Vo RE + RF

As will be shown right after, the closed-loop (with feedback) gain response of the non-inverting amplifier is as
shown above where the bandwidth of the amplifier is given by the following expression:

BW = βV x GBW

The above formula also applies to all op amp circuits with resistive feeback networks as will be shown later.

Derivation of bandwidth formula

In the "Intro to Feedback" section we derived the actual closed-loop gain of the non-inv amp to be:

AV 1
AVF = = where AV is the open-loop gain of the op amp. The
1 + β V AV  1 
+βV asymptotes of AVF are determined as follows:
 AV 

LF asymptote: AV is considered very large such as

20 log (1 β V )
to have βV >> 1/AV, therefore AVF ≈ 1/βV
Ideal gain

HF asymptote: AV is considered very small such as -20 dB/dec

to have βV << 1/AV, therefore AVF ≈ AV
AV
Open-loop gain

The intersection point of the two asymptotes will determine the BW of the circuit. At this point we have AV =
1/βV = AVF ideal , and since this point lies on the -20 dB/dec slope of AV, we can use the GBW of the op amp
to find the bandwidth frequency.

F = BW = GBW / AV = GBW / (1/βV) = βV GBW Done!

From this basic result we can now explain the origin of the name "gain-bandwidth product".

GBW = BW / βV = BW x AVF ideal = BW x (1 + RF/RE )

From the above expression we can see that the product of the ideal gain times the bandwidth of the circuit is a
constant duly named "gain-bandwidth product".

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 24

 Analog Filters

Numerical Example: Non-inverting amplifier response

An LF 347 op amp is used in the circuit shown

beside. Draw the typical open-loop and closed loop Vi n
gain responses for feedback resistor values of 10K LF347 Vo
and 100K.

Op amp data: GBW = 4 MHz typical 2K RF

AV LF = 110 dB typical

A) Open-loop response

110 dB ⇒ 10(
110 20 )
= 316228 V / V , FC = GBW/AV = 4M / 316228 = 12,65 Hz

0 dB crossing F0dB = GBW/ (1 Hz) = GBW = 4 MHz

B) Closed-loop response

2K 1 10K 1
βV = = ⇒ AVF ideal = 1 + = 6 V / V or 15,56 dB ⇒ BW = × 4M = 666,6 kHz
2K + 10K 6 2K 6
2K 1 100K 1
βV = = ⇒ AVF ideal = 1 + = 51 V / V or 34,15 dB ⇒ BW = × 4M = 78,43 kHz
2K + 100K 51 2K 51

G 110 dB AV op amp
A -2
0
I dB
/d
ec AVF
N
RF=100K 34,15 dB
(dB) RF=10K 15,56 dB

F (Hz)
12,6

78,4K

667K

4M

NOTE: The actual gain at the bandwidth frequency is 3 dB down from the ideal gain.

Conclusion: BW = βV x GBW = GBW/ AVF ideal This formula and the above example show that if we want
more gain from the amplifier this is done at the cost of a reduced bandwidth. This result is generally true for
all types of amplifiers.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 25

 Analog Filters
3. OP AMP CIRCUITS WITH RESISTIVE FEEDBACK NETWORKS

It can be shown that the actual gain of any op amp circuit with a resistive feedback network is given by the
following expression whose Bode plot is shown as well.

 AV 
AVF actual = ( AVF ideal × β V ) ×  
 1 + β V AV 

The last term in the above equation is the same as that of the non-inverting op amp and it is multiplied by a
(
constant term AVF ideal × β V )
, therefore the bandwidth of any circuit with resistive feedback is given by the
same expression obtained with the non-inverting amplifier, that is:

BW = βV x GBW

where βV must be calculated by determining the ratio VF/Vo while replacing all sources by their internal
resistances and also by opening the feedback loop.

NOTE: Ideal voltage source Rint = 0 , ideal current source Rint = ∞.

Example: Summing amplifier gain response

Determine the open-loop and the closed-loop gain 2K 100K

responses of the circuit shown beside assuming that V1
the op amp has a GBW = 5 MHz and a DC gain of
20K
100 dB.
V2
Vo
1, 8K

R R 
Ideal output: Vo = −  F V1 + F V2  = −50V1 − 5V2 AVF1 = -50 or 34 dB, AVF2 = -5 or 14 dB
 R1 R2 

Calculation of βV 2K 100K

cut
20K 0A X l oop
2K 20K V
VF = Vo × = o = 0,017857 Vo
+
2K 20K 100K 56 - VF +
Vo

β V = VF Vo = 1 / 56 = 0,017857 1, 8K

Bandwidth BW = βV x GBW = (1/56) 5M = 89,3 kHz

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 26

 Analog Filters
Gain response

G 100 dB AV op amp
A -2
0
I dB
/d
ec AVF
N

34 dB A VF1 =V o/V 1
34,96 dB
(dB) A VF2 =V o/V 2
14 dB

0 dB F (Hz)
50 BW

5M
89,4K

Notice that even if AVF1 and AVF2 are different, the bandwidths for both gains are the same which differs from
what we previously obtained for the non-inverting amplifier whose βV changed for different gains while it is the
same here for both gains. Also notice that the close-loop responses do not merge with the open-loop
response of the op amp - there is a 0,96 dB gap between AV and AVF1. This can be readily verified by
calculating AV of the op amp at 89,4 kHz using its GBW figure.

4. TRANSIENT RESPONSE OF AMPLIFIERS TO A SQUARE WAVE

When a square wave is applied to the input, V o(t) +VP

the op amp will output a square wave with 90%
exponential rising and falling edges if the
output peak amplitude does not exceed
t
SR/(4πBW). The 10%-90% rise and fall times
0

are then given by

10%
-V P

tF = tR = 0,35/BW
tr tf
Slopes
V o(t) equal to
+VP slew rate
If VP >> SR/(4πBW), then the edges of the of op
output square wave will be entirely linear and amp
the 0%-100% rise and fall times will be given 0 t
by tR = tF = VPP/SR - if the negative and
positive slew rates are different then tR ≠ tF.
-V P

tr tf

NOTE: When positive feedback is used or if the O/P is driven hard into saturation, the op amp will always
switch at the slew rate and will produce linear edges at the output. When no feedback is present, if the input
differential voltage is large enough (about 100 mV) the output will also switch at the slew rate of the op amp.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 27

 Analog Filters
SECTION 3 ACTIVE FILTERS
1. First-Order Filter Circuits
First-order non-inverting low-pass filter First-order inverting low-pass filter
RE RF CF

RE RF

R Vin
Vo
Vi n

C Rx
Vo

Transfer function Transfer function

1 + (R F R E ) F (S ) =
− RF RE
F (S ) = 1 + RF C F S
1+ R C S
Filter parameters
Filter parameters
RF
ω n = (RF C F )
−1
Ao = ALF
R
= 1+ F ω n = (RC )
−1 Ao = ALF = −
RE RE
Design procedure
Design procedure
1. Select standard C value
1. Select standard CF value

R = (ω n C )
−1
R F = (ω n C F )
2. Calculate −1
2. Calculate
3. To obtain desired gain calculate ratio
R F R E = (Ao − 1) and to minimize DC O/P offset 3. Calculate R E = R F Ao to obtain the
1 1 1 desired gain.
voltage we must have + = . To meet
RE RF R 4. To minimize DC O/P offset voltage we must
both of the above conditions we have the following: have R X = R E R F .

1 1 1 Select closest standard value.

+ = ⇒ solve for R E (theo.)
R E ( Ao − 1) R E R

Try several standard values of RE around RE(theo) to

achieve the desired ratio R F R E = (Ao − 1) as
accurately as possible with standard values of RE NOTE: the inverting filter allows for gains less than
and RF. 1 v/v which cannot be achieve with the non-inverting
NOTE: If the op amp has FET I/P's, it is not configuration.
necessary to balance the inputs for minimum O/P
offset voltage, therefore step 3 takes into account NOTE: If the op amp has FET I/P's, it is not
only the ratio R F R E = (Ao − 1)
necessary to balance the inputs for minimum O/P
offset voltage, therefore RX is not necessary.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 28

 Analog Filters

First-order non-inverting high-pass filter First-order inverting high-pass filter

RE RF
CE RF
RE
Vin Vo

C
Vo
Vin
RB
R

Transfer function
Transfer function
(− R F R E )(R E C E S )
(1 + R F R E )(R C S ) F (S ) =
F (S ) = (1 + R E C E S )
(1 + R C S )
Filter parameters Filter parameters

RF
ω n = (RE C E )
RF
ω n = (RC )
−1 −1
Ao = AHF = 1 + Ao = AHF = −
RE RE
Design procedure Design procedure

1. Select standard C value 1. Select standard CE value

R = (ω n C ) R E = (ω n C E )
−1 −1
2. Calculate 2. Calculate

3. To obtain desired gain calculate ratio R F = Ao × R E

3. Calculate to obtain the
RF R E = (Ao − 1) and to minimize DC O/P offset
desired gain.
1 1 1
voltage we must have + = . To meet 4. To minimize DC O/P offset voltage we must
RE RF R
both of the above conditions we have the following: have R B = R E RF .

1 1 1 Select closest standard value.

+ = ⇒ solve for R E (theo.)
R E ( Ao − 1) R E R

Try several standard values of RE around RE(theo) to

achieve the desired ratio R F R E = (Ao − 1) as
NOTE: the inverting filter allows for gains less than
accurately as possible with standard values of RE 1 v/v which cannot be achieve with the non-inverting
and RF. configuration.

NOTE: If the op amp has FET I/P's, it is not NOTE: If the op amp has FET I/P's, it is not
necessary to balance the inputs for minimum O/P necessary to balance the inputs for minimum O/P
offset voltage, therefore step 3 takes into account offset voltage, therefore RB is not necessary.
only the ratio R F R E = (Ao − 1)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 29

 Analog Filters
2. SECOND-ORDER FILTER CIRCUITS

2A GENERAL LOW-PASS SALLEN-KEY FILTER

Transfer function

G1G 2
Ao
Vo C 3C 4
=
Vin G G + G2  G1G 2
S 2 +  2 (1 − Ao ) + 1 S +
 C4 C3  C3C 4

Parameter equations
 1 1 
ζ = 0.5ab(1 − Ao ) +  a + 
RF 1 R1 C3
Ao = 1 + ωn = where a= and b =
RE R1 R2 C3C 4  b a  R2 C4

The above equations can be used with any of the following specific low-pass Sallen-Key configurations where
constants a, b and Ao differ from one configuration to another.
To desensitize the filter to the op amp GBW, on should, as a rule of thumb, make GBW 〉 100 × Ao Q Fn . The
larger the GBW of the op amp is, the better, for all types of filters - low-pass, high-pass, etc. To this effect,
manucfacturers of active filter IC's often specify a maximum Q X Fn product that should not be exceeded in
any given application.

2B GENERAL HIGH-PASS SALLEN-KEY FILTER

Transfer function

Vo Ao S 2
=
Vin G  1 1  GG
S 2 +  3 (1 − Ao ) + G 4  +  S + 3 4
 C1  C1 C 2  C1C 2

Parameter equations

 1 1 
ζ = 0.5ab(1 − Ao ) +  a + 
RF 1 C2 R4
Ao = 1 + ωn = where a = and b =
RE C1C 2 R3 R4  b a  C1 R3
The above equations can be used with any of the following specific high-pass Sallen-Key configurations where
constants a, b and Ao differ from one configuration to another.
To desensitize the filter to the op amp GBW, on should, as a rule of thumb, make GBW 〉 100 × Ao Q Fn . The
larger the GBW of the op amp is, the better, for all types of filters - low-pass, high-pass, etc. To this effect,
manucfacturers of active filter IC's often specify a maximum Q X Fn product that should not be exceeded in
any given application.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 30

 Analog Filters
2C LOW-PASS SALLEN-KEY WITH MATCHED C's AND UNMATCHED R's

Transfer function

G1G2
Ao
Vo C2
=
Vin  G (2 − Ao ) + G1  GG
S2 +  2  S + 1 22
 C  C

where C3=C4=C

Parameter equations

 1 C3
ζ = 0.5a(2 − Ao ) + 
RF 1 R1
Ao = 1 + ωn = where a = and b = =1
RE C R1 R2  a R2 C4

Design procedure

Given Ao, ωn and ζ,

1. Select standard value for C

2. Calculate R2 from R2 =
1
ω nC
(
ζ + ζ 2 + ( Ao − 2) )
1
Calculate R1 from R1 =
3.
(
ω n C R2
2 2
)
4. To obtain desired gain calculate ratio R F R E = (Ao − 1) and to minimize DC O/P offset voltage we
1 1 1
must have + = . To meet both of the above conditions we have the following:
RE RF R1 + R2
1 1 1
+ = ⇒ solve for RE (theo.)
RE ( Ao − 1) RE R1 + R2

and calculate RF = (Ao − 1)RE in order to achieve the desired ratio RF RE = (Ao − 1) as accurately as
possible.

NOTE: If the op amp has FET I/P's, it is not necessary to balance the inputs for minimum O/P offset
(
voltage, therefore step 4 takes into account only the ratio R F R E = Ao − 1 )

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 31

 Analog Filters
2D HIGH-PASS SALLEN-KEY WITH MATCHED C's AND UNMATCHED R's

Transfer function

Vo Ao S 2
=
Vin  G (1 − Ao ) + 2G4  GG
S2 +  3  S+ 3 4
 C  C

where C1=C2=C

Parameter equations

RF 1  b(1 − Ao ) 1  C2 R4
Ao = 1 + ωn = ζ = +  where a = and b=
RE C R3 R4  2 b C1 R3

Design procedure

Given Ao, ωn and ζ,

1. Select standard value for C

2. Calculate R3 from
1
R3 =
2ω n C
(
ζ + ζ 2 + 2( Ao − 1) )
1
Calculate R1 from R4 =
3.
(
ω n C R3
2 2
)
4. To obtain desired gain calculate ratio R F R E = (Ao − 1) and to minimize DC O/P offset voltage we
1 1 1
must have + = . To meet both of the above conditions we have the following:
RE RF R1 + R2
1 1 1
+ = ⇒ solve for RE (theo.)
RE ( Ao − 1) RE R1 + R2

Calculate RF = (Ao − 1)RE in order to achieve the desired ratio RF RE = (Ao − 1) as accurately as
possible.

NOTE: If the op amp has FET I/P's, it is not necessary to balance the inputs for minimum O/P offset
(
voltage, therefore step 4 takes into account only the ratio R F R E = Ao − 1 )

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 32

 Analog Filters
2E SALLEN-KEY WITH MATCHED R's AND C's

Low-pass filter High-pass filter

Transfer function
Transfer function
G1G2
Ao Vo Ao S 2
Vo C2 =
=  G (1 − Ao ) + 2G4 
 G (2 − Ao ) + G1 
Vin G3 G 4
Vin GG S2 +  3  S +
S2 +  2  S + 1 22  C  C
 C  C

Parameter equations

Referring to the general equations we have a=1 and

Ao = 1 +
RF
ωn =
1
ζ =
(3 − Ao )
b=1 for both circuits, therefore the following RE RC 2
equations apply to both circuits.

Design procedure
Given ωn and ζ,

1. Select standard value for C

2. Calculate R = 1 /(ω n C )

3. To obtain desired gain calculate ratio R F RE = (Ao − 1) = 2(1 − ζ ) and to minimize DC O/P offset
1 1 1
voltage we must have + = . To meet both of the above conditions we have the following:
RE RF 2 R
1 1 1
+ = ⇒ solve for RE (theo.) and calculate RF = (Ao − 1)RE in order to achieve the
RE ( Ao − 1) RE 2 R
desired ratio R F R E = (Ao − 1) as accurately as possible.

NOTE: If the op amp has FET I/P's, it is not necessary to balance the inputs for minimum O/P offset
(
voltage, therefore step 3 takes into account only the ratio R F R E = Ao − 1 . )

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 33

 Analog Filters
2F UNITY-GAIN SALLEN-KEY

Low-pass filter High-pass filter

RF RF

R1 =R2 =R C1 =C2 =C

R C C
R
Vo Vo
Vi n Vi n

C3 C4 R3 R4

Transfer function Transfer function

1 Vo S2
2 =
Vo R C3C 4 Vin  2  1
= S 2 +  S + 2
Vin  2  1  R4 C  C R3 R4
S 2 +  S + 2
 RC 3  R C3C 4
Parameter equations
Parameter equations

1 R3
Ao = 1 ω n =
1
ζ =
C4 Ao = 1 ω n = ζ =
C3 C R3 R4 R4
R C 3C 4
Design procedure
Design procedure
Given Ao=1, ωn and ζ,
Given Ao=1, ωn and ζ,
1. Select standard value for C.
1. Select standard values for C3 and C4
in order to achieve accurate ratio C4 C3 = ζ 2 . R4 = (ζω n C )
−1
2. Calculate

2. Calculate R = 1 /(ω n C3 C 4 ) 3. Calculate R3 = ζ 2 R4

3. RF=2R for minimum O/P offset voltage - 4. RF=R4 for minimum O/P offset voltage -
select closest standard value. select closest standard value.
If op amp has FET I/P's RF is not needed. If op amp has FET I/P's RF is not needed.

NOTE: The unity-gain Sallen-Key circuit is the configuration with ζ being the least sensitive to
component variations of all possible Sallen-Key configurations and therefore should be used when
greater accuracy is needed for ζ - even if overall filter gain is not 0 dB, all that is needed for the extra
gain is a plain non-inverting or inverting amplifier.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 34

 Analog Filters
2G MULTIPLE FEEDBACK FILTER

General transfer function

− Y1Y4 Where Yk are admittances of the components.

F (s ) =
Y3Y4 + Y5 (Y1 + Y2 + Y3 + Y4 ) Y = G = 1/R for a resistor Y = SC for a capacitor

Type Of Filter Y1 Y2 Y3 Y4 Y5

Low-pass G1 SC2 G3 G4 SC5

High-pass SC1 G2 SC3 SC4 G5
Band-pass G1 G2 SC SC G5

Filter parameters

Type Of Filter ωn ζ Ao

G3 G 4 G1 + G3 + G4 R3
Low-pass −
C 2 C5 2ω n C 2 R1
G2 G5 G5 (C1 + C 3 + C 4 ) C1
High-pass −
C3 C 4 2ω n C 3C 4 C3

G5 (G1 + G2 ) G5 R5
Band-pass −
C G1 + G2 2R1

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 35

 Analog Filters
3. TRANSIENT RESNPONSE OF FIRST AND SECOND-ORDER FILTERS
1 st
1 order low-pass filter
0.8
If F<<FC squarewave gets through
0.6 unattenuated and edges are
0.4
exponential
1 0.35
0.2 τ= tr = t f =
2πFC FC
0
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045 0.005

Input output
0.1
If F >> FC squarewave is heavily
attenuated and distorted. Actually,
0.05 the filter now intregrates the
square wave and it becomes
triangular. The waveform shown
0 beside is form a 1 VPP I/P
squarewave and a filter gain of
1V/V
-0.05

-0.1
8.0E-04 8.5E-04 9.0E-04 9.5E-04 1.0E-03
O/P waveforms for two different frequencies

1
st
1 order high-pass filter
0.5
If F >> FC , the square wave gets
through unattenuated but may
0 have some tilt at the peaks if F is
not high enough.

( )
-0.5
% TILT = e πFC F
− 1 × 100
-1
0.008 0.0085 0.009 0.0095 0.01

O/P waveforms for two different frequencies

1.0 If F << FC , the edges of the I/P

squarewave go through but top
0.5 and bottom of squarewave I/P are
blocked by HPF. Exponential
spikes have following parameters:
0.0
1 0.35
-0.5 τ= tr = t f =
2πFC FC
-1.0
0 0.001 0.002 0.003 0.004 0.005
Input output
C. Sauriol Analog Filters Rev. 3/24/2003 © Page 36
 Analog Filters

nd
1 2 order low-pass filter

0.5 If F << Fn , squarewave gets

through unattenuated and ringing
0 occurs if ζ<1. Amplitude of
oscillations (ringing) decays
-0.5 exponentially.

∆V = V P e −t τ τ = 1 (ζω n )
-1
0 0.002 0.004 0.006 0.008 0.01
settling t1% ≈ τ × Ln(100)
t 0.1% ≈ τ × Ln(1000 )
O/P waveforms for two different damping factors
time
0.1
If F >> Fn , I/P squarewave is
0.05
heavily attenuated and turns into
a “parabolic” wave that
0 resembles a sinewave. The
higher F is, the smaller the O/P
-0.05 amplitude

-0.1
0.0015 0.0016 0.0017 0.0018 0.0019 0.002

O/P waveforms for two different frequencies

nd
1 2 order high-pass filter

0.5
If F >> Fn , the square wave
gets through unattenuated but
0 may have some tilt at the peaks
if F is not high enough.
-0.5

-1
0.008 0.0085 0.009 0.0095 0.01
O/P waveforms for two different frequencies

1.0 If F << Fn , edges of I/P

squarewave get through
unattenuated but top and bottom
0.5 of I/P squarewave are blocked
by HPF and ringing occurs if ζ <
0.0 1. Amplitude of oscillations
(ringing) decays exponentially.
-0.5
∆V = V P e −t τ τ = 1 (ζω n )
-1.0 settling time
0 0.002 0.004 0.006 0.008 0.01 t1% ≈ τ × Ln(100)
O/P waveforms for two different damping factors
C. Sauriol Analog Filters Rev. 3/24/2003 © Page 37
 Analog Filters

3. Nth Order Filters

One needs to use a high-order filter in order to obtain more attenuation past the cutoff frequency of the filter.
Generally, the rolloff slope for a low-pass filter of order N will be –N x 20 dB/decade when frequency is at least
half a decade away from FC.
Flat
response Gain response
is the flattest of
all filters.

Transient
response
exhibits some
ringing.

Butterworth Gain Response Butterworth Unit-Step Response

Ripple in Gain response

Pass-band exhibits ripple.
Filter can be
designed to
specified ripple
value. The more
ripple, the more
ringing in
transient
response

Chebychev Gain Response Chebychev Unit-Step Response

Bessel has
the best
transient
response
due to a
flat delay
response-
See next
page

Bessel Gain Response Bessel Unit-Step Response

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 38

 Analog Filters

Bessel filters
have by far the
flattest delay
response which
results in the
least distortion
of signals – see
previous page
for transient
response to a
step input.

th th
Phase response of 4 order filters Delay response of 4 order filters

Chebychev filters
provide more
attenuation than
Bessel or
Butterworth but
also distort the
signals the most.

We can see that

th
a 10 order
Chebychev filter
made up of five
nd
2 order stages
each having their
own resonant
th nd peak will result in
Gain response of 4 order filters Gain response of each 2 order stage
five peaks in the
of a 10th order Chebychev filter
overall response.
Transfer Functions
th
N order low-pass filter
Vo (s) ALF bo Denominator has polynomial form
F (s) = =
( n −1
Vin (s) bn S + bn−1 S +! + b2 S 2 + b1 S + bo
n
)
Vo (s) K
F (s) = = Denominator has product form
Vin (s) (S − S p ( n ) )(S − S p ( n−1) )...(S − S p 2 )(S − S p1 ) where Spk are the poles of the filter

th
N order high-pass filter
Vo (s) AHF bn S n
F (s) = =
(
Vin (s) bn S n + bn −1 S n −1 +! + b2 S 2 + b1 S + bo ) Denominator has polynomial form

Vo (s) AHF S n
F ( s) = = Denominator has product form
Vin (s) (S − S p ( n ) )(S − S p ( n−1) )...(S − S p 2 )(S − S p1 ) where Spk are the poles of the filter

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 39

 Analog Filters

Fourth-order low-pass
Chebychev Vin First 2nd order Second 2nd order Vo
stage stage

Vo ( s ) A1ω n21 A2ω n22

F (s) = = 2 ×
Vin ( s ) S + 2ζ 1ω n1 S + ω n21 S 2 + 2ζ 2ω n 2 S + ω n22

15
R Fn 1 = 0.96368 × FC
E 10
L G
ζ1 = 0.10883
A 5
A
T I 0
I
N
V -5
E 2 dB ripple
-10
(dB) STAGE-1
Fn 2 = 0. 47071 × FC
-15 STAGE-2
OVERALL ζ2 = 0. 53795
-20
0 0.2 0.4 0.6 0.8 1 1.2
RELATIVE FREQUENCY (F / Fc)

Chebychev filters are stagger-tuned which means that each stage is tuned at different Fn value. By selecting
appropriate ζ values, one obtains an equiripple gain response which means that the two humps in the
response are equal (2 dB in the above example).
20

R
0
E
G
L -40
-40
A -20
A dB/dec
I
T
N -40
I
V
-60 -80
E
(dB) STAGE-1 -80
dB/dec
-80 STAGE-2
OVERALL
-100
0 2 4 6 8 10
RELATIVE FREQUENCY (F / Fc)

The HF attenuation rate is -N x 20 dB/dec, but is higher near the cutoff frequency FC. The more pass-band
ripple, the more the attenuation rate (or slope) near FC. More attenuation is obtained at the expense or more
distortion of the signal.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 40

 Analog Filters

Fourth-order Butterworth low-pass filter

4
R Fn 2 = FC
E 2
L G ζ2 = 0.38268
A A 0
T I -3 dB
I
N -2
V
E
-4
(dB) STAGE-1 Fn1 = FC
-6 STAGE-2
ζ1 = 0.92388
OVERALL
-8
0 0.2 0.4 0.6 0.8 1 1.2
RELATIVE FREQUENCY (F / Fc)

Butterworth low-pass and high-pass filters are synchronous which means that each stage is tuned at the same
frequency Fn=FC . In Butterworth filters the choice of ζ values produces a very flat gain response and a 3 dB
attenuation at the cutoff frequency.

Fourth-order Butterworth low-pass filter

10
0
R
G -10
E -40 dB/dec
L A -20
A I
-30
T N
-40
I
V -50
(dB) -60 STAGE-1 -80 dB/dec
E
STAGE-2
-70
OVERALL
-80
0 2 4 6 8 10
RELATIVE FREQUENCY (F / Fc)

The HF attenuation rate is -N x 20 dB/dec, The attenuation rate near FC is less than N x 20 dB/dec and is
therefore not as great as for a Chebychev filter but is higher than a Bessel filter.

NOTE: Frequency axis is not logarythmic, this means that the graphs do not show slopes in dB/dec.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 41

 Analog Filters
DESIGN EXAMPLE – 1 LOW-PASS BUTTERWORTH FILTER .

Amax
G 26
Design a low-pass A
Butterworth filter I 23
20 dB min
that meets the gain N Amin
response curve attenuatio
shown beside.
(dB)
6
max

500 1000 F (Hz)

1. Minimum order required

 10 Amin 10 − 1   10 20 10 − 1  Therefore we will use n=4, Fc=500Hz and ALP=26

log Amax 10  log 3 10  dB. To minimize the number of stages we must
 10 −1  10 − 1  select the minimum reasonnable order required, that
n≥ = = 3.318 is the next higher integer above 3.318, which is 4. If
 Fs   1000  it would have been closer, say 3.5 to 3.99, it would
2 log  2 log 
 F c  500  have been wiser to use n=5 because of component
tolerance and variations with temperature.

2. Circuit diagram

The Sallen-key circuit with unmatched R's is chosen because it allows for a specific gain.

RE RF RE RF

R1 R2 R1 R2 Vo

C C C C

Vi n

3. Filter parameters and transfer function

From the normalized low-pass filter tabe we obtain the following filter parameters.

ORDER( 4 ) SLP /SLP S LP = ω c × SLP ω n = SLP ζ = −COS /SLP

first stage 1 ±112.5 3141.59 /±112.5 3141.59 0.382683

second stage 1 ±157.5 3141.59 /±157.5 3141.59 0.923880

A1ω n21 A2ω n22 4.4087 × 10 7 4.4087 × 10 7

F (S ) = × = ×
S 2 + 2ζ 1ω n1 S + ω n21 S 2 + 2ζ 2ω n 2 S + ω n22 S 2 + 2404.5S + 4.4087 × 10 7 S 2 + 5804.9 S + 4.4087 × 10 7

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 42

 Analog Filters

4. Calculation of filter components

First stage Ao=4.467, ζ=0.382683, ωn=3141.59

A) C = 16 nF standard

B) R2 =
1
ωnC
(
ζ + ζ2 + (Ao − 2) ) = 39775Ω
1
C) R1 = = 9950.6Ω
(ω C 2R2 )
2
n

RF
D) = A o − 1 = 3.467 for desired gain and R E R F = R1 + R 2 = 9950.6 + 39775 = 49725.6Ω to
RE
minimize DC O/P offset voltage.
1 1 1 1 1
To meet both conditions we must have + = + = ⇒ R E = 64068Ω
R E RF R E 3.467R E 49725.6

RF=3.467 RE = 3.467 X 64068 = 222124Ω

Selected components: C=16 nF, R1=9950.6Ω, R2=39775Ω, RE=64068Ω and RF=222124Ω

Second stage Ao=4.467, ζ=0.923880, ωn=3141.59

A) C = 36 nF standard

B) R2 =
1
ωnC
(
ζ + ζ2 + (Ao − 2) ) = 24281Ω
1
C) R1 = = 3220
(ω C 2R2 )
2
n

RF
D) = A o − 1 = 3.467 for desired gain and R E R F = R1 + R 2 = 24281 + 3220 = 27501Ω to
RE
minimize DC O/P offset voltage.

1 1 1 1 1
To meet both condotions we must have + = + = ⇒ R E = 35433Ω
R E RF R E 3.467R E 27501

RF=3.467 RE = 3.467 X 35433 = 122847Ω

Selected components: C=36 nF, R1=3220Ω, R2=24281Ω, RE=35433Ω and RF=122847Ω

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 43

 Analog Filters

FINAL CIRCUIT

64068 222124 35433 122847

9951 39775 3220 24281 Vo

1 6 nF 1 6 nF 3 6 nF 3 6 nF

Vi n

First stage component tolerance 0.35% Second stage component tolerance 0.88%
for ∆ζ < 5% and ∆Fn < 2 % for ∆ζ < 5% and ∆Fn < 2 %

TYPICAL LOW FREQUENCY GAIN RESPONSE

30.00

G
A 24.00
I
N
18.00
(dB)

12.00

6.00

0.00
100 1K

FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 44

 Analog Filters
TYPICAL WIDE RANGE GAIN RESPONSE

50.00

10.00
G
A
I
N -30.00

(dB)
-70.00

-110.00

-150.00
100 1K 10K 100K 1M 10M

FREQUENCY (Hz)

TYPICAL LOW FREQUENCY GROUP DELAY RESPONSE

1.50m

D
E 1.20m
L
A
Y 0.90m

(sec)
0.60m

0.30m

0.00m
50 100 1K 5K
FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 45

 Analog Filters
DESIGN EXAMPLE – 2 HIGH-PASS BUTTERWORTH FILTER .

Design a high-pass Butterworth filter that meets the gain response shown below.

ACTUAL RESPONSE
IDEAL RESPONSE
G 26
A
I 23 24 dB min
N Amin attenuation

(dB)
2
max
400 800 BW1 BW2 F (Hz)
As we can see from the above gain response, a true high-pass filter is impossible to be implemented
with active filters because of the op amps' open loop responses which cut into the ideal response thus
creating high frequency break points in the actual response. We can determine those break points by
assuming all capacitive reactances of the circuit to be null and then analizing for βv (HF) of each stage after
which we can calculate BW 1=GBW* βv 1 (HF), BW 2=GBW* βv 2 (HF), etc.

1. Minimum order required

 10 Amin 10 We will use n=5, Fc=800 Hz and ALP=26 dB. To

−1   10 24 10 − 1 
log Amax 10  log 3 10  minimize the number of stages we must select the
 10 − 1   10 − 1  minimum reasonnable order required, N=4 would be
n≥ = = 3.987 too close, let N=5 which allows some leeway for
F   800  component tolerance and variations with
2 log C  2 log 
 FS   400  temperature.

2. Circuit diagram

The Sallen-key circuit with matched R's and C's is chosen because it provides an easy design
procedure with a locked gain and the first order stage gain is a free parameter and will thus provide the extra
gain required to achieve the overall gain.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 46

 Analog Filters
3. Filter parameters and transfer function

From the normalized low-pass filter table we first obtain the normalized low-pass poles and then
convert them to normalized high-pass poles and finally we de-normalize high-pass poles.

ORDER SLP 1 SHP = ω c × SHP ω n = SHP ζ = −COS / S HP

(5)
/ S LP SHP =
SLP
first stage 1 5026.55
1 180.000 /-180.000 5026.55 N/A
/-180.000
second 1 5026.55
stage 1 ±108.000 /±108.000 /±108.000 5026.55 0.309017
third stage 1 5026.55
1 ±144.000 /±144.000 /±144.000 5026.55 0.809017

A1 S 2 A2 S 2 A3 S
F (S ) = × ×
S + 2ζ 1ω n1 S + ω n1 S + 2ζ 2ω n 2 S + ω n 2 S + ω n 3
2 2 2 2

2.382 × S 2 1.382 × S 2 6.0613 × S

F (S ) = × ×
S + 3106.6 S + 2.5266 × 10
2 7
S + 8133.1 S + 2.5266 × 10
2 7
S + 5026.55

4. Calculation of filter components

Second stage ζ=2.3820, ζ=0.309017, ωn=5026.55

Ao=3-2ζ

A) C = 39 nF standard
B) R = 1 ω n C = 1 (5026.55 × 39n ) = 5101Ω
C) R F R E = A o − 1 = 1.382 for desired gain and R E R F = 2R = 10.2K to minimize DC O/P offset
1 1 1 1 1
voltage. To meet both conditions we must have + = + = ⇒ RE = 17.58K
RE RF RE 1.382R E 10.2 K

R F = 1.382 × RE = 1.382 × 17.58 = 24296Ω

Selected components: C=39 nF, R=5101, RE=17580 and RF=24296

Third stage ζ=1.3820, ζ=0.809017, ωn=5026.55

Ao=3-2ζ

A,B) R and C will be the same for all stages because ωn is the same for all three stages.

C) R F R E = A o − 1 = 0.382 for desired gain and R E R F = 2 R = 10.2 K to minimize DC O/P offset

voltage. To meet both conditions we must have
1 1 1 1 1
+ = + = ⇒ RE = 36.9 K
RE RF RE 0.382 RE 10.2 K

RF = 0.382 × RE = 0.382 × 36900 = 14096Ω

Selected components: C=39 nF, R=5101, RE=36900 and RF=14096

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 47

 Analog Filters
First stage Ao=6.05, ωn=5026.55

A,B) R and C will be the same for all stages because ωn is the same for all three stages.

C) The first stage gain makes up what is missing to achieve an overall gain of 20 dB or 10 V/V.
Note that a first order stage does not have any damping (ζ) specified which means that its gain is a free
parameter here determined by the desired overall gain.

Av tot 19.953
Av1 = = = 6.061 RF RE = Ao − 1 = 5.061 for
Av 2 Av 3 (1 + 14096 36900 )(1 + 24296 17580 )
desired gain and R E R F = R = 5101 to minimize DC O/P offset voltage. To meet both conditions we must
1 1 1 1 1
have + = + = ⇒ RE = 6109Ω
RE RF RE 5.061R E 5101

We now try standard values of RE around 6109_ and such that R F R E = 5.061 because the ratio R F R E
is more important (it affects the overall gain, but is not critical) than minimizing DC O/P offset.

RE standard 4300 4700 5100 5600 6200 6800

RF=5.061 RE 21762 23787 25811 28342 31378 34415
RF standard 22000 24000 27000 27000 30000 33000

Selected components: C=39 nF, R=5101, RE=4.7K and RF=24K

FINAL CIRCUIT
4.7 K 24 K 17580 2429 6 36900 14 09 6

3 9 nF 3 9 nF 39 n F 39 nF 3 9 nF
V in
Vo

5 10 1 5 10 1 5101 5101 5101

All components should have a tolerance of 0.5% or better.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 48

 Analog Filters
TYPICAL LOW FREQUENCY GAIN RESPONSE

TYPICAL WIDE RANGE GAIN RESPONSE

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 49

 Analog Filters
TYPICAL LOW FREQUENCY GROUP DELAY RESPONSE

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 50

 Analog Filters
DESIGN EXAMPLE - 3 LOW-PASS CHEBYCHEV FILTER .

Design a low-pass +2 dB
0
Chebychev filter
that meets the gain G
response curve A
shown beside. I
N

(dB) -40

200
200Hz 800 Hz

1. Minimum order required

Let us be cautious in
 10 Amin 10 − 1  10 40 10 − 1 
COSH −1  A max 10 COSH −1  2 10 choosing the order and use a fourth
 10 − 1  10 − 1  order filter again to allow for
n≥ = = 2.698 component variations.

−1  Fs  −1  800 
COSH  COSH
 Fc   200 

2. Circuit diagram

The Sallen-key circuit with matched R's and unity gain is chosen for this application.

NOTE: The unity-gain Sallen-Key configuration is the most stable with respect to ζ
variations and should be chosen when better performance is required.

RF RF

R R R R Vo

C3 C4 C3 C4

Vi n

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 51

 Analog Filters
3. Filter parameters and transfer function

From the normalized low-pass filter table we obtain the following filter parameters.

ORDER SLP /SLP S LP = ω c × SLP ω n = SLP ζ = −COS /SLP

(4)
1210.996
first stage 0.96368 ±96.248 /±96.248 1210.996 0.1088322

591.512
second stage 0.47071 ±122.544 /±122.544 591.512 0.5379471

A1ω n21 A2ω n22 1.4665 × 10 6 3.499 × 10 5

F (S ) = × = ×
S 2 + 2ζ 1ω n1 S + ω n21 S 2 + 2ζ 2ω n 2 S + ω n22 S 2 + 263.59S + 1.4665 × 10 6 S 2 + 636.4S + 3.499 × 10 5

4. Calculation of filter components

First stage Ao=1, ζ=0.1088322, ωn=1210.996

A) C3 standard 1.0E-7 2.2E-7 2.4E-7 2.7E-7 3.0E-7 4.3E-7

C 4 = ζ C3 2.60E-9 2.84E-9 3.19E-9 3.55E-9 5.09E-9
2
B) 1.18E-9
1.20E-9 2.70E-9 2.70E-9 3.30E-9 3.60E-9 5.10E-9
C4 standard
1
C) R=
(ω n C 3C 4 ) 75382 33882 32439 27664 25127 17633

D) To minimize DC O/P offset voltage RF=2R=2*75382=150.76K, 150K standard.

Selected components: R=75382Ω, C3=100 nF, C4=1.2 nF and RF=150K

Second stage Ao=1.0, ζ=0.5379471, ωn=591.512

4.7E-8 5.1E-8 5.6E-8 6.2E-8 6.8E-8 7.5E-8

A) C3 standard
C 4 = ζ C3
2
B) 1.360E-8 1.476E-8 1.621E-8 1.794E-8 1.968E-8 2.170E-8
1.300E-8 1.500E-8 1.600E-8 1.800E-8 2.000E-8 2.200E-8
C4 standard
1
R=
(ω C 3C 4 )
C) 68394 61123 56478 50606 45842 41619
n

D) To minimize DC O/P offset voltage RF=2R=2*50606=101.2K, 100K standard.

Selected components: R=50606Ω, C3=62 nF, C4=18 nF and RF=100K

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 52

 Analog Filters

FINAL CIRCUIT

150K 100K

7 53 82 7 53 82 5 0606 50606 Vo

1 00 n F 1.2 nF 62 nF 18 nF

Vin

All components should have a tolerance of 1% or better.

TYPICAL LOW FREQUENCY GAIN RESPONSE

5.00

0.00

G
A
I -5.00
N

(dB) -10.00

-15.00

-20.00
3 10 100 300

FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 53

 Analog Filters
TYPICAL WIDE RANGE GAIN RESPONSE
10.00

G
A -40.00
I
N
-90.00

(dB)
-140.00

-190.00

-240.00
10 100 1K 10K 100K 1M 10M

FREQUENCY (Hz)

TYPICAL LOW FREQUENCY GROUP DELAY RESPONSE

10.00m

D 8.00m
E
L
A 6.00m
Y

4.00m

(sec)

2.00m

0.00m
10 100 400

FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 54

 Analog Filters
DESIGN EXAMPLE - 4 HIGH-PASS CHEBYCHEV FILTER .

Design a high-pass Butterworth filter that meets the gain response shown below.

ACTUAL RESPONSE
0,5 dB ripple IDEAL RESPONSE
G 20
A
I
N 20 dB min Amin
attenuation
(dB)
0
max
1.25K 5K BW1 BW2 F (Hz)

As we can see from the above gain response, a true high-pass filter is impossible to be implemented
with active filters because of the op amps' open loop responses which cut into the ideal response thus
creating high frequency break points in the actual response. We can determine those break points by
assuming all capacitive reactances of the circuit to be null and then analizing for βv(HF) of each stage after
which we can calculate BW 1=GBW* βv 1 (HF), BW 2=GBW* βv 2 (HF), etc.

1. Minimum order required

We will use n=3, FC=5 kHz and AHP=20

dB. To minimize the number of stages
10 Amin 10
−1 10 20 10
−1
COSH −1 COSH −1 we must select the minimum
10 Amax 10
−1 10 0.5 10
−1 reasonnable order required, n=2 is too
n≥ = = 1.959 close, n=3 allows some leeway for
−1  FC  −1  5k 
COSH   COSH   component tolerance and variations .
 FS   1.25k 

2. Circuit diagram

The Unity-gain Sallen-key circuit is chosen for the second order stage and a simple non-inverting first
order stage whose gain is a free parameter that will be the overall gain of 20 dB.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 55

 Analog Filters
3. Filter parameters

From the normalized low-pass filter table we first obtain the normalized low-pass poles for 0.5 dB of
ripple and then convert them to normalized high-pass poles and finally we de-normalize high-pass poles.

0.5 dB High-Pass Chebychev

SLP /SLP 1 SHP = ω c × SHP ωn = SHP ζ = − COS /SHP
SHP =
SLP
first 1.596271 50148.34
stage 0.62646 180.000 /-180.000 /-180.000 50148.34 N/A
second 0.935585 29392.27
stage 1.06885 ±107.041 /±107.041 /±107.041 29392.27 0.293056

A1 S 2 A2 S 1× S 2 10 × S
F (S ) = × = ×
S + 2ζ 1ω n1 S + ω n1
2 2
S + ω n2 S + 17227 S + 8.6391 × 10
2 8
S + 50148.34

4. Calculation of filter components

Second stage Ao=1, ζ=0.293056, ωn=29392.27

A) C = 10 nF standard
1 1
B) R4 = = = 11609Ω
ζω n C 0.293056 × 29392.97 × 10n
C) R3 = ζ 2 R4 = 0.293056 2 × 11609 = 997Ω
D) To minimize DC O/P offset voltage, RF = R4 = 12K

Selected components: C=10 nF, R3= 997, R4=11609, and RF=12K

First stage Ao=10, ωn=50148.34

A) C = 3.9 nF standard
B) R = 1 (ω n C ) = 1 (50148.34 × 3.9n ) = 5113Ω
C) The first stage gain makes up what is missing to achieve an overall gain of 20 dB or 10 v/v.
Note that a first order stage does not have any damping (ζ) specified which means that its gain is a free
parameter here determined by the desired overall gain.
Av tot 10
Av1 = = = 10 RF RE = Ao − 1 = 9 for desired gain and R E RF = R = 5113 to minimize
Av 2 1 DC O/P offset
1 1 1 1 1 voltage.
To meet both conditions we must have + = + = ⇒ RE = 5.68 K
RE RF RE 9 RE 5113
We now try standard values of RE around 5.68K and such that R F R E = 9 because the ratio R F R E is
more important (it affects overall gain) than minimizing DC O/P offset.

RE standard 5600 6200 6800 7500 8200 9100

R F = 9 RE 50400 55800 61200 67500 73800 81900
RF standard 51000 56000 62000 68000 75000 82000

Selected components: C=3.9 nF, R=5113, RE=6.2K and RF=56K

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 56

 Analog Filters

FINAL CIRCUIT

6.2K 56K 12K

3.9 n F 1 0 nF 10 nF
Vin
Vo

51 13 9 97 1 16 09

All components should have a tolerance of 1% or better.

TYPICAL LOW FREQUENCY GAIN RESPONSE

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 57

 Analog Filters
TYPICAL WIDE RANGE GAIN RESPONSE

TYPICAL LOW FREQUENCY GROUP DELAY RESPONSE

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 58

 Analog Filters
DESIGN EXAMPLE – 5 LOW-PASS BESSEL FILTER .

Design a low-pass 20
G
Bessel filter that
meets the gain A 17
response curve I
shown beside. N
30 dB min.
(dB)

-10
max

400 1600 F (Hz)

1. Minimum order required

Since we don't have an

N 0 equation to calcultate N
O minimum, we will proceed
graphically by using the
R -10 normalized curves shown
M.
-20
ORDER beside. At
2nd F/Fc=1600/400=4, the
G -30 3rd X attenuation must be
A greater than 30 dB which
4th corresponds to a
I -40
5th normalized gain below -30
N dB. It appears that N=4
-50 6th
is sufficient but to be safe
(dB) 7th let's use N=5 to allow for
-60 component tolerance and
8th
variations with
-70 temperature.
0 1 2 3 4 5
NORMALIZED FREQUENCY ( F/Fc )

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 59

 Analog Filters
2. Circuit diagram

The Sallen-key circuit with matched R's and C's is chosen for second order stages because of easy
design procedure and high ζ values which can be achieved with reasonnable accuracy - if superior accuracy
of ζ is desired, then the Sallen-Key with 0 dB is the best choice because it is the least sensitive. Here the gain
(Ao=3-2ζ) of the second order stages is determined by the ζ values but Ao of the first order stage is a free
parameter which will be determined by the desired overall gain. For the first order stage, a simple non-
inverting filter can be used for gains > 0 dB and a first order inverting filter can be used instead if the gain < 0
dB which cannot be achieved with the non-inverting filter.

RE RF RE RF RE RF

Vi n R R R R R Vo

C C C C C

3. Filter parameters and transfer function

From the normalized low-pass filter table we obtain the following filter parameters.

ORDER SLP /SLP S LP = ω c × SLP ω n = SLP ζ = −COS /SLP A o = 3 − 2ζ

(4)
first stage Av tot
1.50470 180.000 3781.72 3781.72 ---
A v2 Av3
/±180.0
second
1.55876 ±152.530 3917.59 3917.59 0.887252 1.2255
stage
/±152.530
third
stage 1.75812 ±123.062 4418.64 4418.64 0.545546 1.9089
/±123.062

A1ω n21 A2ω n22 Aω

F (S ) = × × 3 n3
S + 2ζ 1ω n1 S + ω n1 S + 2ζ 2ω n 2 S + ω n 2 S + ω n3
2 2 2 2

1.8808 × 10 7 3.727 × 10 7 16166

F (S ) = × ×
S + 6951.8 S + 1.5348 × 10
2 7
S + 4821.1 S + 1.9524 × 10
2 7
S + 3781.7

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 60

 Analog Filters
4. Calculation of filter components

Second stage Ao=1.2255, ζ=0.887252, ωn=3917.59

A) C = 16 nF standard

B) R = 1 ω n C = 15954Ω

RF
C) = A o − 1 = 0.2255 for desired gain and R E R F = 2R = 31908Ω to minimize DC O/P offset
RE
voltage. To meet both conditions we must have
1 1 1 1 1
+ = + = ⇒ R E = 173407Ω
R E RF R E 0.2255 RE 31908

R F = 0.2255 × R E = 0.2255 × 173407 = 39103Ω

Selected components: C=16 nF, R=15954Ω, RE=173407Ω and RF=39103Ω

Third stage Ao=1.9089, ζ=0.545546, ωn=4418.64

A) C = 16 nF standard

B) R = 1 ω n C = 14145Ω

RF
C) = A o − 1 = 0.9089 for desired gain and RE RF = 2R = 28290Ω to minimize DC O/P
RE
offset voltage. To meet both conditions we must have

1 1 1 1 1
+ = + = ⇒ R E = 59415Ω
R E RF R E 0.9089 R E 28290

RF = 0.9089 X RE = 0.9089 X 59415 = 54002Ω

Selected components: C=16 nF, R=14145Ω, RE=59415Ω and RF=54002Ω

First stage Ao=4.2747, ωn=3781.72

A) C = 16 nF standard

B) R = 1 ω n C = 16527Ω
C) The first stage gain makes up what is missing to achieve an overall gain of 20 dB or 10 V/V.
Note that a first order stage does not have any damping (ζ) specified which means that its gain is a free
parameter here determined by the desired overall gain.

Av tot 10 R RF
A v1 = = = 4.2747 = 1+ F ⇒ = 3.2747 for desired gain
A v2 Av3 1.2255 × 1.9089 RE RE

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 61

 Analog Filters
and R E RF = R = 16527Ω to minimize DC O/P offset voltage. To meet both conditions we must have
1 1 1 1 1
+ = + = ⇒ R E = 21574Ω
R E RF R E 3.2747 R E 16527

Since Ao is not critical for a first order stage, we now try standard values of RE around 21574Ω and such that
R F RE = 3.2747 to obtain the correct gain.

RE standard 16K 18K 20K 22K 30K

RF=3.2747 RE 52395 58945 65494 72043 98241
RF standard 51K 56K 68K 75K 100K

Selected components: C=16 nF, R=16527Ω, RE=30K and RF=100K

FINAL CIRCUIT

30K 100K 173407 39103 59415 54002

Vi n 16527 15954 15954 14145 14145 Vo

1 6 nF 1 6 nF 1 6 nF 1 6 nF 1 6 nF

All components should have a tolerance of 1% or better.

TYPICAL LOW FREQUENCY GAIN RESPONSE

25.00

G
A 15.00
I
N

5.00

(dB)

-5.00

-15.00

-25.00
10 100 1K 2K

FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 62

 Analog Filters
TYPICAL WIDE RANGE GAIN RESPONSE

50.00

G
A -10.00
I
N

-70.00

(dB)

-130.00

-190.00

-250.00
10 100 1K 10K 100K 1M 10M

FREQUENCY (Hz)

TYPICAL LOW FREQUENCY GROUP DELAY RESPONSE

1.00m

0.80m
D
E
L
0.60m
A
Y

0.40m

(sec)

0.20m

0.00m
10 100 1K 10K
FREQUENCY (Hz)

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 63

 Analog Filters
TABLE 1 NORMALIZED LOW-PASS FILTER POLES

BUTTERWORTH BESSEL CHEBYCHEV

0.1 dB RIPPLE 0.5 dB RIPPLE 1 dB RIPPLE 2 dB RIPPLE
ORDE
R SLP /SLP SLP /SLP SLP /SLP SLP /SLP SLP /SLP SLP /SLP
(N)
2 1 ±135.000 1.27420 ±150.000 1.82045 ±130.661 1.23134 ±125.373 1.05000 ±121.515 0.90723 ±116.296
3 1 ±120.000 1.32475 180.000 1.29990 ±111.893 1.06885 ±107.041 0.99710 ±104.348 0.94133 ±101.300
3 1 180.000 1.44993 ±136.348 0.96941 180.000 0.62646 180.000 0.49417 180.000 0.36891 180.000
4 1 ±112.500 1.43241 ±163.332 1.15327 ±103.241 1.03127 ±99.790 0.99323 ±98.076 0.96368 ±96.248
4 1 ±157.500 1.60594 ±128.367 0.78926 ±143.902 0.59700 ±135.162 0.52858 ±129.591 0.47071 ±122.544
5 1 ±108.000 1.50470 180.000 1.09313 ±98.763 1.01773 ±96.316 0.99414 ±95.163 0.97579 ±93.964
5 1 ±144.000 1.55876 ±152.530 0.79745 ±123.143 0.69048 ±115.120 0.65521 ±110.944 0.62702 ±106.360
5 1 180.000 1.75812 ±123.062 0.53891 180.000 0.36232 180.000 0.28949 180.000 0.21831 180.000
6 1 ±105.000 1.60653 ±168.458 1.06273 ±96.196 1.01145 ±94.403 0.99536 ±93.582 0.98283 ±92.739
6 1 ±135.000 1.69186 ±144.891 0.83449 ±112.055 0.76812 ±106.033 0.74681 ±103.149 0.73003 ±100.125
6 1 ±165.000 1.90782 ±119.250 0.51319 ±146.521 0.39623 ±137.002 0.35314 ±131.082 0.31611 ±123.681
7 1 ±102.857 1.68713 180.000 1.04520 ±94.601 1.00802 ±93.242 0.99633 ±92.629 0.98723 ±92.007
7 1 ±128.571 1.71911 ±159.923 0.86788 ±105.705 0.82273 ±101.194 0.80837 ±99.116 0.79711 ±96.979
7 1 ±154.286 1.82539 ±139.167 0.57464 ±126.209 0.50386 ±117.262 0.48005 ±112.676 0.46085 ±107.679
7 1 180.000 2.05279 ±116.355 0.37678 180.000 0.25617 180.000 0.20541 180.000 0.15534 180.000
8 1 ±101.250 1.78143 ±171.175 1.03416 ±93.547 1.00595 ±92.485 0.99707 ±92.012 0.99014 ±91.533
8 1 ±123.750 1.83514 ±153.314 0.89381 ±101.762 0.86101 ±98.295 0.85061 ±96.731 0.84249 ±95.138
8 1 ±146.250 1.95645 ±134.699 0.64514 ±115.003 0.59887 ±108.085 0.58383 ±104.807 0.57193 ±101.386
8 1 ±168.750 2.19237 ±114.075 0.38159 ±147.449 0.29674 ±137.648 0.26507 ±131.604 0.23770 ±124.078

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 64

 Analog Filters

TABLE 2 POLE CONVERSION EQUATIONS

TYPE OF NORMALIZED POLES DE-NORMALIZED POLES NATURAL DAMPING

FILTER FREQUENCY FACTOR

Find /SLP from

Find SLP from table 1 or from S LP = ω c × SLP ω n = SLP table 1 or from
LOW-PASS pole equation given below table pole equation
ζ = −COS /SLP

1 S HP = ω c × SHP ω n = SHP ζ = −COS /SHP

HIGH-PASS SHP =
SLP

SLP SLP 2 S BP = (ω c 2 − ω c1 ) × SBP ω n = SBP ζ = −COS /SBP

BAND-PASS 2
SBP = j ω
2 ± cen −
4

1 2 1 S BR = (ω c2 − ω c1 ) × SBR ω n = SBR ζ = −COS /SBR

BAND- SBR = ± j ω −
REJECT 2 SLP cen
4 SLP 2
OR NOTCH

ωc : bandwidth of overall filter for low-pass or high-pass cases.

ωc1 - ωc2 : bandwidth of overall filter for band-pass or band-reject cases.

ω cen
ω cen = : normalized geometric center frequency for bpf or brf filters where ω cen = ω c2 ω c1 .
(ω c 2 − ω c1 )
SHP = SLP = 1 : for Butterworth

ζHP = ζLP : for Butterworth, Bessel and Chebychev filters.

Butterworth low-pass poles: S k = cos(φ k ) + jsin(φk )

Chebychev low-pass poles: S k = cos(φ k )sinh(a) + jsin(φ k )cosh(a)

  2k + 1  1  1 
, k = 0, 1, 2, 3, ..., N − 1 , a = sinh 
−1
where φ k (° ) = 90°+ 90°
  n  N  10 0.1AMAX − 1 
and N is the order.

C. Sauriol Analog Filters Rev. 3/24/2003 © Page 65

Electronics II Theory Analog Filters
CALCULATION OF MINIMUM ORDER OF FILTERS

Butterworth filters
(0.1 AMIN )
10 −1 10 (0.1 AMIN ) − 1
LOG (0.1 AMAX ) LOG (0.1 AMAX )
10 −1 10 −1
N LP 〉 N HP 〉
F FC
2 × LOG S 2 × LOG
FC FS
10 (0.1 AMIN ) − 1 10 (0.1 AMIN ) − 1
LOG LOG
10 (0.1 AMAX ) − 1 10 (0.1 AMAX ) − 1
N BP 〉 2 × N LP (min ) ⇒ N LP 〉 N BR 〉 2 × N LP (min ) ⇒ N LP 〉
F − FS 1 FC 2 − FC1
2 × LOG S 2 2 × LOG
FC 2 − FC1 FS 2 − FS 1

Amin : min. attenuation at FS , Amax : max. gain variation inside pass-band, 3 dB for Butterworth.

Chebychev filters
(0.1 AMIN )
10 −1 10 (0.1 AMIN ) − 1
COSH −1 (0.1 AMAX ) COSH −1
10 −1 10 (0.1 AMAX ) − 1
N LP 〉 N HP 〉
FS FC
COSH −1 COSH −1
FC FS
10 (0.1 AMIN ) − 1 10 (0.1 AMIN ) − 1
COSH −1 COSH −1
10 (0.1 AMAX ) − 1 10 (0.1 AMAX ) − 1
N BP 〉 2 × N LP (min ) ⇒ N LP 〉 N BR 〉 2 × N LP (min ) ⇒ N LP 〉
−1 FS 2 − FS 1 FC 2 − FC1
COSH COSH −1
FC 2 − FC1 FS 2 − FS 1
Amin : min. attenuation at FS , Amax : max. gain variation inside pass-band, ripple(dB) for Chebychev

Bessel filters
N 0
Using min. attenuation at FS /FC, find O
R -10
min. low-pass order from Bessel curves
M.
shown or translate into HP, BP or BR -20 ORDER
minimum order. 2nd
G -30
F 
3rd
A
N LP (min) for ATTmin (dB)at S  I -40
4th

 FC  LP N 5th
-50 6th
(dB) 7th
N HP 〉 N LP (min) -60
8th
F  F  -70
for ATTmin (dB)at  S  =  C  0 1 2 3 4 5
 FC  LP  FS  HP NORMALIZED FREQUENCY ( Fs/Fc )

F   F − FS 1  F   F − FC 1 
N BP 〉 2 N LP (min) for ATT min ( dB )at  S  =  S 2  N BR 〉 2 N LP (min) for ATT min ( dB ) at  S  =  C 2 
 FC  LP  FC 2 − FC 1  BP  FC  LP  FS 2 − FS 1  BR

C. Sauriol Rev. 3/24/2003 © Analog Filters Page 66

Electronics II Theory Analog Filters
EFFECT OF COMPONENT VARIATION ON FREQUENCY RESPONSE - SENSITIVITY

150K 100K

7 53 82 7 53 82 5 0606 50606 Vo

1 00 n F 1.2 nF 62 nF 18 nF

Vin

Unity-gain Sallen-Key: 2% components

2.5
2
1.5
Gain (dB)

1
0.5
0
-0.5
0 50 100 150 200 250
Frequency (Hz)

C. Sauriol Rev. 3/24/2003 © Analog Filters Page 67

Electronics II Theory Analog Filters

1 0K 1 7. 82K 1 0K 9. 24K

8 2576 8 25 76 5 12 31 5 1231 Vo

10 nF 10 nF 33 nF 33 nF

Vin

Sallen-Key with gain:2% components

10

5
Gain (dB)

-5

-10
0 50 100 150 200 250
Frequency (Hz)

C. Sauriol Rev. 3/24/2003 © Analog Filters Page 68

Electronics II Theory Analog Filters

TRANSFER
FUNCTION ζ > 1 ζ < 1
Imag
Low-pass filter +j PEAK (dB)
ω n = Sp1 × Sp2
20 LOG Ao Sp1 G
-2 ω A
2 0
Sp2 dB n I
Ao ωn2 Imag G de
c 20 LOG Ao -4
F(s) = 2
= ζ + ζ2 − 1 A N 0
S p1 dB
( )

-4
S + 2ζωn S + ω2n Sp1 +j I -
/d

0
+ ec
N Real

dB
K S p2

/d
ω

ec
F(s) = ωmax ωn
(S − Sp1 )(S − Sp2 ) S p1 ωn S p2 ω ωn
Real - X X +
Sp2 PEAK(dB) = −20 LOG 2ζ 1− ζ2
Sp2 Sp1 -j ( )
-j

ω n = Sp ζ = −COS Sp ω max = ω n 1 − 2ζ2

Imag
+j PEAK (dB)
High-pass filter ω n = Sp1 × Sp2
Sp1
20 LOG Ao ω G
2 c A
Sp2 de n
Imag dB double zero I
= ζ + ζ2 − 1 0 G ec 20 LOG
AoS2 +2 Ao
N
S p1 /d
F(s) = 2 Sp1 ( ) A
- Sz1= Sz2 dB
0

ec
S + 2ζωn S + ω 2n +j I Real + +4

/d
N
2 S p2

dB
0
KS double zero ω

+4
F(s) = ωn ωmax
Sz1= Sz2
p1 p2 ω S p1 S p2 ωn
(S − S )(S − S ) - ωn
J0
Real X X Sp2
-j PEAK(dB) = −20 LOG 2ζ 1− ζ2
Sp2 Sp1 ( )
-j
2
ω n = Sp ζ = −COS Sp ω max = ω n 1 − 2ζ

Imag
+j + 20 LOG Ao
Band-pass filter ω n = Sp1 × Sp2 Sp1 + 20 LOG Ao - 3 dB
20 LOG Ao ω ∆ω = ωn /Q
2 n G
Sp2 A +20 LOG Ao /Q
= ζ + ζ2 − 1 Imag G c -2 I
de 0 S p1
A
A 2ζω n S Sp1 ( ) 0
dB dB Sz1
+ N dec -20
F(s) = 2 o 2 +j I de Real - dB/
+2 c +20
dB/
S + 2ζω n S + ω n N S p2
dec

KS Sz1 ωn ω
F(s) = S p1 S p2 ω ωn
(S − Sp1)(S − Sp2) - + ωn
Real X X Sp2
-j
Sp2 Sp1 PEAK(dB) = 20 LOG( Q)
-j

ω n = Sp ζ = −COS Sp Q = 1 (2ζ )

C. Sauriol Rev. 3/24/2003 © Analog Filters Page 69

Electronics II Theory Analog Filters

TRANSFER
POLE-ZERO DIAGRAM FREQUENCY RESPONSE
FUNCTION

Band-reject filter Imag

+j
2 2 Sp1
Ao (S + ω n ) Sz1 = +j ωn Ao Ao
F(s) = 2 ω n = SP = Sz ζ = −COS S p G
A
ωn
S + 2ζω n S + ω 2n I
0.707 A o
S p1
N
K (S − Sz1 ) (S − Sz 2 ) Real - + (V/V)
∆ω = ωn /Q
F(s) = S p2
(S − Sp1) (S − Sp2) 0
ωn ω
n
ω
Sp2 -j
Sz2 = ωn
-j

Amplitude equalizer Imag ω n = SP = Sz A o ζ1/ζ2

+ j ζ >ζ
1 2
Sp1
Ao
G
Ao (S2 + 2ζ1ω n S + ω 2n ) A
ωn
Sz1 ζ <ζ
1 2
F(s) = ζ 1 = − COS S z
ωn I
S p1 A o ζ1/ζ2
(S2 + 2ζ2ω nS + ω 2n ) N
Real - + (V/V)
K(S − S z1 ) (S − Sz 2 ) S p2
0
ωn ζ 2 = − COS S p
F(s) = Sz2 n ωn ω
(S − Sp1)(S − Sp2 ) ω
Sp2
- j

Imag F(jω) = oA for all frequencies

All-pass filter
+ j
ω n = SP = Sz
Delay equalizer Sp1 Sz1
G D
2 2 n
R E
ωn
Ao (S − 2ζω n S + ω ) n ω Q =10
tg max = 4Q/ωn
F(s) = ζ = COS S z = −COS S p O L
U A
(S2 + 2ζω nS + ω 2n ) Real - + P Y Q = 5

(sec)
K(S − S z1 ) (S − Sz 2 ) n ωn S p1 = − Sz 2 Sp2 = − Sz1 0
F(s) = ω ωn ω
p1
(S − S )(S − S ) p2
Sp2 Sz2
- j

C. Sauriol Rev. 3/24/2003 © Analog Filters Page 70