LABORATORY MANUAL
DEPARTMENT OF
ELECTRICAL & COMPUTER ENGINEERING
UNIVERSITY OF CENTRAL FLORIDA
EEL 4140
ANALOG FILTERS DESIGN
Revised
September 2005
TABLE OF CONTENTS
SAFETY RULES AND OPERATING PROCEDURES
LABORATORY SAFETY INFORMATION
EXPERIMENT # 1
Study Guide A
Study Guide B
Effect of Op Amp Frequency Dependence on Finite Gain
Amplifiers and Bandwidth Extension Techniques using
Composite Op Amps (CNOA)
Composite Operational Amplifiers: Generation and FiniteGain
Applications
Inverting Integrator and Active Filter Applications of Composite
Operational Amplifiers
EXPERIMENT #2
Real Zero and Pole Synthesis
EXPERIMENT # 3
SallenKey Filters
EXPERIMENT # 4
StateVariable Biquads
EXPERIMENT # 5
Single Op Amp BandPass Filters
EXPERIMENT # 6
Two Op Amps Current Generalized Immittance
Structure (CGIC) Based Biquad
Biquads II: The current Generalized Immittance (CGIC)
Structure
Study Guide C
EXPERIMENT # 7
Study Guide C
HighOrder LowPass Filter Design
Biquads II: The current Generalized Immittance (CGIC)
Structure
EXPERIMENT # 8
Butterworth Filter Approximation
APPENDIX
LIST OF COMPONENTS
RESISTOR COLOR CODE TUTORIAL
ii
Safety Rules and Operating Procedures
1. Note the location of the Emergency Disconnect (red button near the door) to shut off power
in an emergency. Note the location of the nearest telephone (map on bulletin board).
2. Students are allowed in the laboratory only when the instructor is present.
3. Open drinks and food are not allowed near the lab benches.
4. Report any broken equipment or defective parts to the lab instructor.
Do not open, remove the cover, or attempt to repair any equipment.
5. When the lab exercise is over, all instruments, except computers, must be turned off.
Return substitution boxes to the designated location. Your lab grade will be affected if your
laboratory station is not tidy when you leave.
6. University property must not be taken from the laboratory.
7. Do not move instruments from one lab station to another lab station.
8. Do not tamper with or remove security straps, locks, or other security devices.
Do not disable or attempt to defeat the security camera.
9. ANYONE VIOLATING ANY RULES OR REGULATIONS MAY BE DENIED
ACCESS TO THESE FACILITIES.
I have read and understand these rules and procedures. I agree to abide by these rules and
procedures at all times while using these facilities. I understand that failure to follow these rules
and procedures will result in my immediate dismissal from the laboratory and additional
disciplinary action may be taken.
________________________________________
Signature
Date
________________
Lab #
iii
Laboratory Safety Information
Introduction
The danger of injury or death from electrical shock, fire, or explosion is present while conducting
experiments in this laboratory. To work safely, it is important that you understand the prudent
practices necessary to minimize the risks and what to do if there is an accident.
Electrical Shock
Avoid contact with conductors in energized electrical circuits. Electrocution has been reported at
dc voltages as low as 42 volts. 100ma of current passing through the chest is usually fatal. Muscle
contractions can prevent the person from moving away while being electrocuted.
Do not touch someone who is being shocked while still in contact with the electrical conductor or
you may also be electrocuted. Instead, press the Emergency Disconnect (red button located near
the door to the laboratory). This shuts off all power, except the lights.
Make sure your hands are dry. The resistance of dry, unbroken skin is relatively high and thus
reduces the risk of shock. Skin that is broken, wet, or damp with sweat has a low resistance.
When working with an energized circuit, work with only your right hand, keeping your left hand
away from all conductive material. This reduces the likelihood of an accident that results in current
passing through your heart.
Be cautious of rings, watches, and necklaces. Skin beneath a ring or watch is damp, lowering the
skin resistance. Shoes covering the feet are much safer than sandals.
If the victim isnt breathing, find someone certified in CPR. Be quick! Some of the staff in the
Department Office are certified in CPR. If the victim is unconscious or needs an ambulance,
contact the Department Office for help or call 911. If able, the victim should go to the Student
Health Services for examination and treatment.
Fire
Transistors and other components can become extremely hot and cause severe burns if touched. If
resistors or other components on your protoboard catch fire, turn off the power supply and notify
the instructor. If electronic instruments catch fire, press the Emergency Disconnect (red button).
These small electrical fires extinguish quickly after the power is shut off. Avoid using fire
extinguishers on electronic instruments.
Explosions
When using electrolytic capacitors, be careful to observe proper polarity and do not exceed the
voltage rating. Electrolytic capacitors can explode and cause injury. A first aid kit is located on the
wall near the door. Proceed to Student Health Services, if needed.
iv
EEL 4140
ANALOG FILTERS
LABORATORY 1
Effect of Op Amp Frequency Dependence on Finite Gain Amplifiers and Bandwidth
Extension Techniques using Composite Op Amps (CNOA)
I.
Objective
To understand the effect of finite gain bandwidth product of practical Op Amps in finite gain
applications, and to study the concept of Composite Op Amps (CNOA).
II. Introduction
The simplified model of a practical Op Amp is shown in Fig.1.
In this model, RI and RO
are the input and output resistance, respectively. The openloop gain AOL ( j ) can be written
as:
AOL ( j ) =
A0
(1)
1+ j
0
Where A0 is the DC openloop gain, and 0 is the dominantpole frequency.
Va
+
+
RI
Vb
RO
Vo
o
AOL ( j )[Va Vb ]
Fig.1 The simplified model of a practical Opamp
The relationship between
AOL ( j )
and the frequency is shown in Fig.2.
In this figure,
the gainbandwidth product G , or the unitygain bandwidth, is defined as:
G A0 0
(2)
AOL ( j ) (dB)
A0 (dB)
0 (dB)
log(G )
log( 0 )
log( )
Fig.2 The magnitude response of the openloop gain
The positive finite gain amplifier, as shown in Fig. 3, is used to illustrate the bandwidth
shrinkage by the voltage gain K . In this amplifier, the input and output resistances of the Op
Amp are assumed to be infinite and zero, respectively. The transfer function of this amplifier
can be derived as:
T ( j ) =
Vo ( j )
Vi ( j )
A0
A0
1+ j
1+
K
(3)
0 1 +
A0
Normally, it is true that
A0 >> K
Therefore, Equation (3) can be simplified as:
(4)
T ( j ) K
1+ j
(5)
A0
K
and the cutoff frequency, or 3dB frequency, of this amplifier, is given by:
A0
cutoff = 0
(6)
From Equation (6), it is clear that the cutoff frequency is inversely proportional to the gain
of the amplifier, K . Therefore, the bandwidth of a single Op Amp amplifier realization
shrinks by a factor of
1
K
Vi ( j )
Vo ( j )
R1 = R
R2 = (K 1)R
Fig 3 The positive finite gain amplifier
One method to increase the bandwidth is to use the Nstage amplifier, each stage having the
gain of N K to realize an overall gain K .
by
1
N
Consequently, the bandwidth of each stage shrinks
In addition, cascading N stages introduces another shrink factor of
In total, the bandwidth of the Nstage amplifier shrinks by a factor of
1
N
2
1 [1].
1
2 N 1
.
N K
Composite Op Amp (CONA)
The CNOA is constructed using N Op Amps and 2(N1) resistors, resulting in (N1) resistor
ratios. These resistor ratios can be used advantageously to reduce the deviation of the overall
active realization and to guarantee the system stability. The CNOA is versatile since it has three
external terminals that correspond to those of the regular Op Amps. CNOA has other important
properties, such as stability with one and twopole Op Amps model, low sensitivity to
component and Op Amps mismatch, and wide dynamic range.
As shown in References [2, 3] attached, the bandwidth of the amplifier constructed by
CNOA shrinks only by a factor of N
1
K
Compared with the shrinkage factor of
1
K
for the
single Op Amp amplifier, and
2 N 1
N
for the Nstage amplifier, the CNOA considerably
extends the useful bandwidth.
In this experiment, we only discuss C2OA, composed of two Op Amps and 2 resistors.
the general case of CNOA, see References [2, 3].
For
Four different C2OA structures, referred as C2OA1, C2OA2, C2OA3, and C2OA4, are
found to meet the good performance criteria, and are shown in Fig. 4. Here, is the resistor
ratio. The openloop gain of the Op Amps used in the modeling of the C2OAs (assuming a
singlepole model) is
Ai ( j ) =
Ao,i 0,i
j + 0,i
i = 1 or 2
(7)
where Ao, i , and 0, i are the DC openloop gain and the cutoff frequency of the ith Op Amp,
respectively.
The output voltage of the C2OAs is given by:
Vom = Va Aam ( j ) Vb Abm ( j ),
m = 1,2,3,4
(8)
where for C2OA1
Vo1 = Va
A2 ( j )(1 + A1 ( j ) )(1 + )
A ( j ) A2 ( j )(1 + )
Vb 1
A1 ( j ) + (1 + )
A1 ( j ) + (1 + )
(9)
for C2OA2
Vo 2 = Va
A1 ( j ) A2 ( j )(1 + )
A ( j ) A2 ( j )(1 + )
Vb 1
A2 ( j ) + (1 + )
A2 ( j ) + (1 + )
(10)
for C2OA3
Vo 3 = Va
A1 ( j ) A2 ( j )
A ( j )(1 + A1 ( j ) )
Vb 2
(1 + )
(1 + )
(11)
and for C2OA4
Vo 4 = Va
A2 ( j )( A1 ( j ) + )
A ( j )[A1 ( j ) + (1 + )]
Vb 2
(1 + )
(1 + )
C2OA1
b
C2OA2
b
(12)
A1
+
A2
A2

A1
(a)
(b)
C2OA3
a
C2OA4
a
R

A1
A2
o
R
A1
A2
(c)
(d)
Fig 4. The Composite Operational Amplifiers (C2OAs).
(a) C2OA1. (b) C2OA2. (c) C2OA3. (d) C2OA4.
From Equations (9), (10), (11), and (12), the transfer functions of the circuits using C2OAs
can be derived. The applications of the four proposed C2OAs in the positive and negative
finitegain amplifiers are summarized in Table 1. From this table, it is clear that the 3dB
1
.
K
bandwidth of the finite amplifier shrinks by a factor of
Design Procedure for Negative FiniteGain Amplifiers Employing C2OA1
1. Design a negative finitegain amplifier employing C2OA1.
10, and the quality factor Qp of C2OA is 1.
2.
Assume the two Op Amps in the C2OA1 are identical.
The gain of this amplifier k is
That is
1 = 2
(13)
A1 = A2
(14)
and
The quality factor can be simplified as:
QP =
3.
1+
1+ k
Calculate the resistor ratio as:
= QP 1 + k 1
= 2.32
4.
5.
(15)
(15)
Choose the resistor value R in the C2OA1 as:
R = 10k
(16)
R = 23.2k
(17)
Calculate the resistor value as:
III. Design
1.
Positive Finite Gain Amplifiers
a. Design a singlestage positive finite gain amplifier with an overall gain of 100 using LM471
Op Amps. Repeat for an overall gain of 25.
b. Design a twostage positive finite gain amplifier.
overall gain of 100. Use LM471 Op Amps.
Each stage has a gain of 10 to realize the
c. Design a positive finite gain amplifier using C2OA1 ( Q p = 0.707 ) with closeloop gain of
100.
Use LM471 Op Amps.
2. Negative Finite Gain Amplifiers
a. Design a singlestage negative finite gain amplifier with an overall gain of 100 using LM471
Op Amps. Repeat for an overall gain of 25.
b. Design a twostage negative finite gain amplifier.
overall gain of 100. Use LM471 Op Amps.
Each stage has a gain of 10 to realize the
c. Design a negative finite gain amplifier by using C2OA1 ( Q p = 0.707 ) with an overall gain of
100.
Use LM471 Op Amps.
IV. Computer Simulations
1. Positive Finite Gain Amplifiers
a. Simulate the singlestage positive finite gain amplifier with overall gain of 100, and plot the
magnitude responses of the amplifier. Repeat for an overall gain of 25. Record and
compare these two cutoff frequencies. Relate the bandwidth shrinkage to the amplifier
closeloop gain.
b. Simulate the twostage positive finite gain amplifier with closeloop gain of 100.
magnitude response. Determine the cutoff frequency.
Plot the
c. Simulate the positive finite gain amplifier with finite gain of 100 using C2OA1.
magnitude response. Determine the cutoff frequency.
Plot the
Compare these cutoff frequencies of the three realizations of finite overall gain of 100, and
show the useful bandwidth improvement.
2. Negative Finite Gain Amplifiers
a. Simulate the singlestage negative finite gain amplifier with closedloop gain of 100, and plot
the magnitude response of the amplifiers. Repeat for an overall gain of 25. Record and
compare these two cutoff frequencies. From the formulas and the graphs compare the
bandwidth shrinkage to the amplifier closedloop gain.
b. Simulate the twostage negative finite gain amplifier with overall gain of 100. Plot the
magnitude response. Determine the cutoff frequency.
c. Simulate the negative finite gain amplifier with closeloop gain of 100 using C2OA1.
the magnitude response. Determine the cutoff frequency.
Plot
Compare these cutoff frequencies of the three realizations of finite overall gain of 100, and
comment on the useful bandwidth in each realization.
V. Experiments
This lab is a computer simulation lab. No actual experiment.
References
[1]. J. Millman, and C. Halkias, Integrated Electronics; Analog and Digital Circuits and
Systems, McgrawHill, Inc. 1972, pp.386
[2]. Wasfy B. Mikhael, and Sherif Mickael, Composite Operational Amplifiers: Generation and
FiniteGain Applications, IEEE Trans. on Circuits and Systems, Vol. 34, No. 5, pp. 449460,
May 1987.
[3]. Sherif Mickael, and Wasfy B. Mikhael, Inverting Integrator and Active Filter Applications
of Composite Operational Amplifiers, IEEE Trans. on Circuits and Systems, Vol. 34, No. 5, pp.
461470, May 1987.
Table 1 Negative and Positive Finite Gains Vo V Using the C2OAs
i
C2OAi
Negative Finite Gain Positive Finite Gain The
3dB The quality factor
Qp
Transfer Function
Transfer Function
bandwidth
p
C2OA1
Ti
1+
1+
2
s
s
+ 2
pQ p p
C2OA2
Ti
1+
C2OA3
Ti
1+
C2OA4
Ti
1+
1+
(1 + )s
Ti
s
s
+
p Q p w 2p
C2OA
s
s
+ 2
pQ p p
s
s
+ 2
pQ p p
1+
Ti
s
s2
+ 2
pQ p p
1+
1
1+
s
s
+ 2
pQ p p
1+
Vo
1 2
1+
1+ k
1+ k
1 2
1+
1+ k
2
1
s
s2
1+
+
p Q p 2p
Ti
1+
Vi
Ti
Vi
s
1
1+ k
1
2
1 2
(1 + k )(1 + )1
1 2
(1 + k )1
(1 + ) 2
(1 + k )(1 + )
(1 + k )(1 + )
2
s
s
+ 2
pQp p
C2OA
Vo
kR
Vo
= k = Ti
Vi
kR
Vo
= (1 + k ) = Ti
Vi
is the ideal
transfer function
Ti
Study Guide A
EEL 4140
ANALOG FILTERS DESIGN
Composite Operational Amplifiers: Generation and FiniteGain Applications
IEEE TRANSACTIONS
ON CIRCUITS
AND
SYSTEMS,
VOL.
CAS34, NO. 5,
MAY
449
1987
Composite Operational Amplifiers: Generation
and FiniteGain Applications
WASFY B. MIKHAEL,
FELLOW, IEEE, AND
AbstractA
practical and effective general approach is presented for
extending the useful operating frequencies and improving the performance
of linear active networks realized using operational amplifiers (OAs). This
is achieved by replacing each OA in the active network by a composite
operational amplifier (CNOA) constructed using N OAs. The technique
of generating the CNOAs for any given N is proposed. The realizations
employing the CNOA are examined according to a stringent performance
criterion satisfying such important properties as extended bandwidth,
stability with one and twopole OA models, low sensitivity to the components and OA mismatch, and wide dynamic range. Several families of
CNOAs, for N = 2, 3, and 4, are shown to satisfy ,the suggested performance criterion. In this contribution, the CNOAs applications in inverting, noninverting, and differential finitegain amplifiers are given and
shown theoretically and experimentally to compare favorably with the
stateoftheart realizations using the same number of OAs. Applications
of the CNOA in inverting integrator and active filter realizations are
presented in a companion contribution (321.
I.
INEAR
ACTIVE
INTRODUCTION
circuits,
namely,
positive,
negative,
and differential finitegain amplifiers, integrators, and
active filters are usually realized with operational amplifiers (OAs) as the active elements.These active elements
have frequencydependentgains which restrict the operating frequencies of the linear active circuits. The operating
frequencies are defined to be those frequencies for which
the deviation of the actually obtained transfer function
T(s) of an active realization from its ideal value q(s)
(due to the OAs finite gain and frequency dependence)
falls within a predetermined acceptable range. The
frequency limitations due to the passive components are
not addressed here. In practice, the passive components
restrict operation in a higher frequency range relative to
that of the OA.
For practical reasons, extending the useful bandwidth
(BW) of the most commonly used linear active circuits has
received the attention of many researchersin this field.
This has resulted in many contributions, each dealing with
the solution of this frequency limitation in specific applications [l][14]. Generally, three approacheswere considered
to minimize the dependenceof the realization on the active
Manuscript received October 29,1985; revised August 26, 1986.
W. B. hfikhael is with the Electrical Engineering Department, West
Virginia University, Morgantown, WV 26506.
S. Michael is with the Department of Electrical and Computer
Engineering, Naval Postgraduate School, Monterey, CA 93943.
IEEE Log Number 861?469.
SHERIF MICHAEL,
:@p;++
,
(4
Fig. 1.
MEMBER, IEEE
~qfjj&!+!F~
1)Nll.fOl
(b)
Atm
(a) An operational amplifier (VCVS) and (b).its nullor representation.
device parameters and consequently its variations [15],
[16]. In the first approach, for a given fixed number of
OAs, the passive configuration in which the OAs are
embedded (called the companion network) is carefully
designed. In the secondapproach, an increasednumber of
OAs are used to realize a given 7).(s). This results in
increased degreesof freedom in choosing the companion
network: In the third approach, each OA in a given
configuration is simply replaced by an OA that has improved characteristics, such as wider gainbandwidth
product (GBWP).
Recently, the authors suggestedan approach using socalled composite operational amplifiers (CNOAs) that
achieved a considerable performance improvement and
bandwidth extension of almost all linear active networks
used in signal processingand amplification for audio and
video communications as well as instrumentation [17][20].
This has been verified by other researchers[21], [22]. In
addition, the CNOA concept has proved to be useful in
nonlinear and highspeed,highaccuracyapplications, such
as fast A/D and D/A conversion,digital communications,
and switched capacitor filtering [23][26], [34]. The objective of this paper is to present a comprehensivetreatment
of the CNOA method and its application in linear active
signal processing. In this technique, each OA is replaced
by a composite operational amplifier (CNOA) [17][20]
without modifying the topology of the companion network.
In Section II, the procedure for generating the CNOAs
is presented. First, a general technique for the generation
of a number of C20As (N = 2) using imllatornorator
pairing [27][30] is described. Four of the C20As are
found to meet useful performance criteria; these are retained for use in design.To further illustrate the generality
of the proposed technique, the generation of the CNOAs
00984094/87/05000449$01.00 01987 IEEE
450
IEEE TRANSACTIONS
ON CIRCUITS
0)
(4
7h
AND
SYSTEMS,
VOL.
CAS34, NO. 5, MAY 1987
(4
(HI)R
Fig. 2. (a)(d) Four different networks fbr generating the composite operational amplifiers using two single OAs (C20As).
(e) H and (f) +H finitegain amplifier realizations used in Fig. 4(a)(d). (g) The composite operational amplifier
(C20Ai) symbol.
for N > 2 is described. Sample results of C30As and
C40As, which meet the above performance criteria, are
presented.
Use of the proposed CNOArs in inverting and noninverting finitegain applications is given in Section III. It is
shown theoretically and experimentally that appreciable
performance improvements are realized over the present
stateoftheart designs which utilize the same number of
OAS.
II.
GENERATION OF COMPOSITEOPERATIONAL
AMPLIFIERS (CNOAs) USING N SINGLE OAs
A. Generationof the C2OAs TN = 2)
An operational amplifier, shown in Fig. l(a) is a
voltagecontrolled voltage source (VCVS). In the ideal
case, the input impedance Z, + cc. This corresponds to
the model shown in Fig. l(b), which uses nullator and
norator singular elements [27][30]. The ideal OA is replaced by a nullor which is describedby
[:]=[: :I*[q.
(1)
The matrix in (1) is called the nullor chain transmission
matrix of an ideal OA. In any physical circuit that contains N OAs, replacing each OA by a nullor results in a
nullor equivalent network. The nullors can further be split
into nullators and norators to yield a nullatornorator
equivalent network.
Similarly, a nullatornorator equivalent network containing N nullators and N norators yields N! nullor
equivalent networks, since nullators and norators can be
paired into nullors in an arbitrary manner.
Although the nullator (or norator) alone is not an admissible element for modeling a physical network, the
nullor, like the infinitegain controlled source, can be used
for this purpose. The equivalence established is valid
whether A + cc or A +  cc, and so in practice a nullor
can be replaced by a highgain differential controlled
source in two ways. In general,a nullor equivalent network
containing N nullors correspondsto 2 physical networks.
Each of these N! nullor networks yields a physical realization which has a different dependence on the nonideal
active elements. This subject is well documented in the
literature [27][30].
The procedure to generate.the C20As is as follows. In
the first step, a redundant amplifier of finite gain + H
MIKHAEL
AND
MICHAEL:
COMPOSITE
OPERATIONAL
I//
.Fig. 3.
451
AMPLIFIERS
(4
i
/
I/
(4
The composite operational amplifiers (C20As). (a) C20A1. (b) C20A2. (c) C20A3. (d) C20A4
(Fig. 2(e) and (f)) is combined with a single OA such that
the chain matrix of the resulting twoamplifier network
(assuming ideal amplifiers) correspondsto that of a nullor,
as given in (1). That is, although each network contains
two VCVSs, the overall twoport network realizes one
VCVS. Six possible topologies can be obtained for each of
the four networks shown in Fig. 2(a)(d). Two topologies
are obtained, one for + H and the other for  H, for each
position of the threeway switch, leading to six topologies
per network. It is easy to show that 17 of the 24 topologies
realize true nullors; i.e., no special stipulation on network
elements or signals is required. Eight possible OA realizations can be obtained from each of these 17 topologies
(nullor networks). This results in 136 composite operational amplifiers (C20As), each constructed using two
single OAs. The resulting C20As, shown in Fig. 2(g), are
examined according to the following performance criteria.
i) Let A,(s) and Ah(s) denote the noninverting and
inverting openloop gains of each of the 136 C20As
examined. The denominator polynomial coefficients of
A,(s) and Ah(s) should show no change in sign. This
satisfies the necessary(but not sufficient) conditions for
stability. Also, none of the numerator or denominator
coefficients of A,(s) and Ah(s) should be realized through
differences. This eliminates the need for single OAs with
matched GBWPs and results in low sensitivity of the
C20A with respect to its components.
ii) The external threeterminal performance of the C20A
should resemble as closely as possible that of the single
OA.
iii) No righthalf splane (RHS) zeros due to the single
OA pole should be allowed in the closedloop gains of the
C20As (for minimum phase shifts).
iv) The resulting inputoutput relationship T,(s) in the
applications considered should have extended frequency
operation with minimum gain and phase deviation from
the ideal q(s). The improvement should be sufficient to
justify the increased number of OAs.
Four C20As referred to as C20A1, C20A2, C20A3
and C20A4, of the 136 examined are found to meet these
performance criteria, and are shown in Fig. 3.
It is interesting to note that a special case of C20A3
can be derived from the transistor Darlington pair [31],
where the norators are both at ac ground in the Darlington
pair enabling an OA realization.
The openloop gain of the single OAs used in the
modeling of the C20As (assuminga singlepole model) is
Ai===
AoitiLi
wi
WLi+S
i=lor2
s + WLi
(4
where Aoi, wLi and wi are the dc openloop gain, the 3dB
bandwidth, and the GBWP of the ith single OA, respectively.
It can be easily shown that the openloop inputoutput
relationships for the C20A1 to C20A4 are given by
(i=l..4)
V,,= K,Aai(s)GAbi
where for C20A1
v
=
01
Ad+
a
Ad(l+
A,+(l+a)
a)
A,A,(l+
bAI+(l+ci)
a)
(3)
for C20A2
+a>_v AAl +4
Yd=v,AAl
A,+(l+a)
bA2+(1+a)
(4
452
IEEE TRANSACTIONS
ON CIRCUITS
a
I.
I
I
cl&
I
I
Fig. 4.
VOL.
CAS34,
NO.
5,
MAY
1987
C30Ab
a
\
\
\
.
\
From (2) and (7a), the composite amplifier has a singlepole
rolloff from q/A, to w,/(l+ or), where the second pole
occurs. As (Y increases, the dc gain increases while the
frequency of the secondpole decreases.
Also, from (5) and (6), both the C20A3 and the C20A4
have an openloop dc gain given by
and for C20A4
4(4+d
(l+a)
v A,[A,+(l+a)l
b
(1+4
(6)
(7b)
where QIis a resistor ratio, as illustrated in Fig. 3.
Assuming identical OAs, i.e.,
A,l=A,z=A,
and 01=w2=wi
it is interesting to examine the openloop gains given by
(3)(6) in the singleendedinverting application, i.e., when
v, = 0.
For C20A1 and C20A2, the openloop dc gain Aocl
is given by
A
SYSTEMS.
The composite operational amplifiers (C30As). (a) C3OA1. (b) C30A2. (c) C30A3. (d) C30A4. (e) C30A5. (f)
C30A6.
for C20A3
T/od=K
AND
A&+a> = A,(1 + a)
Ocl= 1+ (1 Ia)/A,
for (l+ a) <<A,.
(74
From (2) and (7b), AoC2has double poles (12dB/octave)
at wi/A,, and as (Yincreasesthe dc gain decreaseswithout
affecting. the location of the secondpole.
Only the C20A2 has identical expressionsfor the positive and negative openloop gains A, and A,. Thus, commonmode rejection ratio (CMRR) problems should not
be encountered using C20A2, even for relatively large
commonmode signal applications. From (3), (53, and (6),
the CMRR of the C20A1 and C20A3 is (A,, + l/2),
while that of the C20A4 is (A,, + (Y+ l/2). For singleended applications (small commonmode signal), no prob
MIKHAEL
MICHAEL:
COMPOSITE
OPERATIONAL
453
AMPLIFIERS
Fig. 5. The composite operational amplifiers (C40As). (a) C40A1. (b) C40A2. (c) C40A3.
lems are anticipated in using the C20A1, C20A3, or
C20A4, as verified experimentally later.
It is easy to show that the voltage swing at the first
OA(A,) output, which is an internal node in each of
C20A1 to C20A4, is always less than the output voltage
V,. Hence, the dynamic range is determined by the voltage
swing of the output voltage V,. Consequently, no
dynamicrange reduction of V, or harmonic distortion
problems should arise.
B. Generationof CNOAs (N > 2)
Following an analogous approach, CNOAs for N > 2
can be generated for extending the operating frequencies
at the expense of additional amplifiers. The CNOAs can
be obtained in two different ways. The first approach
starts from the basic single OA with additional redundant
amplifiers. Then, nullatornorator pairing is used as described in Section IIA.
In the second approach, which is used here, the C20As
are used as single OA replacementsin the C20A structure.
(4
C40A4
(e) C40A5.
Although this second approach is not exhaustive,it will be
shown to yield excellent results. Hence, C30As are obtained by starting with one of the proposed C20As and
replacing one of its single OAs by any ,of the C20As in
Fig. 3. Thirtytwo possible combinations of C30As can be
obtained using the four proposed C20As.
Similarly, C40As are generatedhere by replacing each
of the single OAs in a C20A with any of the C20As or
by replacing one of the single OAs in C30As with a
C20A. This results in many possible combinations of
C40As. The process can be continued (by using C20As,
C30As, and C40As) to obtain CNOAs for any number
N. N should be limited in practice; the increasedcomplexity is expected to give rise to practical problems in spite of
the advantagesof an extended operating range.
Samplesof C30A and C40A novel designs,which meet
the performance criteria described in Section IIA, are
shown in Figs. 4 and 5. The openloop expressionsof these
C30As and C40As, as well as others, can be found
elsewhere[33].
IEEE TRANSACTIONS
454
ON CIRCUITS
AND
SYSTEMS,
VOL.
CAS34,
NO. 5, MAY
1987
TABLE
NEGATIVE
CZOAi
Negative
Finite
Gain Trans.
(1.3)
AND
POSITIVE
Function
I
GAINS
FINITE
Positive
Finite
1
C20A1
Ti
+ (S%;)
Function
wP
QP
WlW2
1 + (S/upQ;)
+ (S%;)
I l+k
1 + WwpQp) + (S'/$)
F l+k
T,
1 + (S/opQp)
Ti
+ (S*/$)
Ti
+ (S*/w;)
w2
i WI
(14
fK
l wy
WI
yyaz
(l+Ww1)
Ti
1 + (S/wPQP) + (S%;)
ji$$
1 + (VwPQP) + (S'/$)
vow
igz
*va
kR
kR
!!.c=
vi
fx&
(W
fl+k
1 + (S/wpQp) + (S%;)
Cl+(l+~wwl
Ti
w1w.T
(l+S/u)
Ti
1 + (ShPQP)
C20A4
Gain Trans.
(Ta)
THE C20As
(1Ww)
CZOA3
v] USING
T,
1 + (ShPQP)
C20A2
V,/
F=
1
 k=Ti
(ltk)
= Ti
Ti(idea1
Transfer
Function)
*aR1 e kR (for maximum up).
III.
REALIZATION
OF POSITIVE, NEGATIVE,
DIFFERENTIAL
FINITEGAIN
LR
AND
AMPLIFIERS
V2
A. Finite Gain Amplifiers Using the ProposedC2OAs
The application of the four proposed C20As in positive
and negative finitegain amplification is given in Table I.
For the differential finitegain realization, often referred to
as the instrumentation amplifier shown in Fig 6, C20A2
is used. The network in Fig. 6, when each single OA is
modeled by (2), can be shown to have the inputoutput
relationship given by
1
where
q1 = x(1 + k)/(l+
q2=k
Qp=
020+
(8)
x)
k)
For the differential gain application given above and the
finitegain applications in Table III, the actual inputoutput relationship T, has the form
T,=q.;
Application
N=l+a~=l+~
(1+ Lx).
w1
J
1 + s/oPQP + s ,a, v,
T2
Fig. 6.
of the C20A2 as a differential
fier.
finitegain ampli
where q = the transfer function realized assuming ideal
OAs and
1
1 + s/wPQP + s /o, I v,
+
VI
(9)
(a is zero ( w, + cc) in somecases)
*z
D=1+b,s+b2s2=1+(s/WpQp)+(s2/~;).
Thus, N/D indicates the amplitude and phasedeviation of
T, from q. Also, b, and b, determine the stability of T,,
while a, b,, and b,, and consequently oZ, wP,and QP, are
functions of the circuit parameters or, w2, and (Y.None of
the a and b coefficients is realized through differences,
which guaranteesthe low sensitivity of T,, o,, o,,, and Q.P
to the circuit parameters.On the other hand, the b coefficients are always positive (assuming a singlepole ,OA
model), which is necessaryfor the stability of the transfer
function. From Table I, a mismatch of k 5 percent in w1
and w2 results in a +Spercent change in wP and a
*2.5percent change in QP. Hence, single OAs with mismatched gainbandwidth products within practical ranges
can be used without appreciably affecting the stability or
the sensitivity of the finitegain realizations.
455
MIKHAEL AND MICHAEL: COMPOSITE OPERATIONAL AMPLIFIERS
TABLE II
VALUESOF~
FORMAXIMALLYFLATANDFOR
Q,, =l
C20As ANDTHE
FINITEGAINREALIZATIO~S~SING
CORRESPONDINGBANDWIDTHAND
STABILITYCONDITIONS
dBb
1
CZOAi
1%
c2oA1
PP
Stability
Condition
for (I used
OF!
Satisfied
2
fm
Satisfied
(ind$ment
C20AI
32 
1 C20A3
1 OPmin = m(.
&
I
(l+k)
Unsatisfied
30
30
for op= ,707
a:4
,
,
r
,
,
,
, f kh.
40 50 60 70 80 90 100 110 120 130 140 I50
(4
I
Unsatisfied
m
C20A4
10
1
2(l+k)
zI
IO 20
Proposed design
Unsatisfied
20
&(ltk)
I
30
I
40
50
1) Effect of the Single OAs SecondPole on the Stability
60
of the C2OAs FiniteGain Realizations: In the following,
70
the stability properties of the positive and negative finite80
gain amplifier realizations using a twopole, openloop
model of the single OAs is studied. If the dc gains of the
(b)
(b)
first and second OAs (A, and A,, respectively) are asFi 7. Theoretical responses of the negative finitegain amplifiers using
sumed to be equal, this greatly simplifies the analysis 8 20A  1 and the existing twoOA realizations (assuming OA GBWP = 1
MHz). (a) Frequency response of negative finitegain amplifiers. (b)
without affeCting the reliability.of the conclusions. This is
Phase response of negative finitegain amplifiers.
due to the absenceof gain difference terms in all the gain
expressions obtained, as seenfrom (3)(6) (8), and Table
I
I. Let
comparing the ups in Table II. As oP increasesfor a fixed
A=A,=A,
Q,, both amplitude and phasedeviations of T, from 7; at
a given frequency o (w < wP) decrease.It can be easily
where 114 is given by
shown that the differential finitegain amplifier in Fig. 6
.f= (1+;)(
&+k),
(10) has similar excellent bandwith and stability properties as
those obtained for C2OA2 inpbsitive and negative finite1
gain
applications. In conclusi?n,it is clear that the C2OAI
and wh >> wL.
and
C2OA2
are the most attractive configurationsin finiteBy invoking the RouthHurwitz stability criterion [16],
gain
applications
from SW aid stability considerations.It
the necessaryand sufficient condition for stability using
should
be
noted
that some special cases of finitegain
the C20Ai .or C20A2 [33] is found to be
amplifiers using C20As have been reported in the litera0+9<(1+W
(11) ture [l][3], [5], [6], [8] and Qted for their improved performance.
while for the C20A3 the condition [33] is found to be
2) Comparisons of the Proposed C2OAs FiniteGain
(l&)>/m.
(14 Reizlizationswith Others: The SW of a finitegain amplifier
realized using a single OA shrinks approximately by a
Finally, for the C20A4 the condition [33] is given by
multiplying factor lik relative,to its unity gain 3dB BW
(1+X) >4(1$k).
03) (wi). Also, the optimum n@paily
flat 3dB BW using a
Upon examining (ll), (12), and (13), one finds that e cascade of two (singleOA realization) finitegain ampliimposed hy the stability conditions (not necessarily BW fiers is obtained when eacham+fier has a gain fi to
conditions): is physically realizable for all k. In practice, (Y realize an overall gain k. .The resulting BW shrinks by
should be chosen in the stable range for a given .k that m/G
= 0.66/G relativeto wl[31]. The C20A1 and
results in the best realizable value of QP and o,.Table II C20A2 circuit BWs can be designedto shrink by only a
gives the values of (Y required to yield QP= l/a
and factor of =l/fi
for QP= 0.707 (maximally flat) and
Q, = 1 for the realizations in Table I. The usefui BWs of greater than l/G
for QP= 1 (k> 1); see Table II. In
the different finitegain amplifiers can be obtained by addition, the C20As require only two accurate gain
1987
IEEE TRANSACTIONS ON CIRCUITS AND SY~TEMS,VOL. CAS34, NO. &MAY
456
20 LOG
Vo/Vi
k
I
+2
dB
0
= 0.707
5 %
5 %
_
6
6 I
.f
20
a=
10
20
40
80
60
100
30
40
50
120
140
160
180
200
220
c ,f
LOG
70
80
90
100
II0
120
130
kHz
IhO
kHr
(4
20
60
(b)
Vo/Vin
k
I
dB
20LOG
I
k
b/Vi
k
Qp = 0.707
d0
100
k
a
= 100
=6
4
10
20
30
40
50
60
70
80
so
100
(4
(4
Fig. 8. Experimental results using C2OA1 in negative gain applications. (a) (Qp = 0.707) Maximally flat closedloop
gain =  25,  50,  100 (LM747 OP AMPS). (b) Effect of compensation resistorratio variation by T 5 percent (LM747 OP
AMPS). (c) Effect of active compensation on extending the bandwidth (LM747 OP AMPS). (d) Effect of power supply
variation from T 9 V to T 15 V on the closedloop gain for k = 100 (LM 747 OP AMPS).
determining components, compared with four in the
cascaderealization.
To further illustrate the usefulnessof the C20As, one
of the common applications considered in this paper is
chosen, namely, negative finitegain amplification. The
performance of the C20A1 in this application is illustrated and compared in Fig. 7 with some of the most
recently published negative finitegain realizations which
utilize a similar number of OAs. The results shown in Fig.
7 are for nominal gains > 1 for practical reasonssince an
increase in k decreasesthe useful bandwidth, so that
extending the range of operating frequenciesbecomesmore
important. The proposed realizations are seen to be far
superior in both amplitude and phaseresponsesrelative to
those reported in [7] and [9]. Upon examining those of [2]
and [8], one may erroneously conclude that, in spite of
their inferior amplitude characteristics, they have better
phase response. In fact, the realizations of [2] and [8], in
contrast with the proposed ones, can be easily shown in
theory to be unstable for all useful values of closedloop
gains, due to the second Ok pole. This has been verified
experimentally as well. Indeed, the results in Fig. 7 show
clearly the excellent gain and phase performance of the
proposed realizations.
3) Experimental Results Using C2OAs in FiniteGain
Applications: Experimental results of negative finitegain
amplifier realizations are given in Fig. 8(a) and (b) using
the C20A1 of Fig. 3(a). LM747s with a GBWP ranging
from 1 to 1.5 MHz were used to implement the C2OAs in
this section as well as in the experiments throughout this
work. The stability and low sensitivity to the power supply
and to the active compensation resistor variations are
examined as shown in Fig. 8(c) and (d). Comparing the
results in Fig. 8 with those obtainable using singleamplifier realizations illustrates the considerableimprovement
in the useful BW without sacrificing any of the singleOA
desirable features, namely, its low sensitivity to circuit
elements and power supply variations, stability, and versatility.
B. FiniteGain Applications Using C3OAs and C4OAs
Wideband positive, negative, and differential composite
amplifiers can be designed using the CNOA structures
proposed in Section IIB and shown in Figs. 4 and 5.
Positive and negative finitegain expressionsfor C30A1
through C30A6 are given in Table III, while those for
C40A1 through C40A4 can be found elsewhere[33]. In
the finitegain expressionsof these C30As and C40As,
MIKHAEL
AND
MICHAEL:
COMPOSITE
OPERATIONAL
457
AMPLIFIERS
TABLE III
NEGATIVE
C30Ai
Finite
Gain Transfer
C30A1
1 t (1%)
POSITWE
FINITE
GAINS
V,/ y
USING
THE C30As
Function
k(l+
kc
"i
AND
i3;)
$+
Negative Finite
Gain Trans. Func.
(E)
&+
(l+k)
s+,
52 L
&
(l+k)>a(lw)
""i
V
.o _
q C30A2
!!L
"i
(ltk)(l+
l+(l+#
'
T+a
TlFi w,w:,
++
($1
St
(l+k)
?
4 w2w3
Neqative Finite
Gain Trans. Func.
 K (W./w,)
,+ CL+
w*
l+k
ol(l+B)
1 s+ f&g+
(l+kJ$g
(l'tk)>
1+ [+
+ s
1 S + #$
&
+ (I+@&
9
Negative Finite
Gain Trans. Func.
K
l+($)$+(&&+(l+U&
C30A3
(l+k)>(l+a)(l+B)
L
V
o=
"i
(l+K)(
l+(s)+
1 + &
g+
+ ( $
&
L
)$q+
Positive
Finite
/ Gain Trans. Func.
J
W+&
K
C30A4
l+k
+(K)w
a(l+B)
Positive
Finite
Gain Trans. Func.
(l+K)(l+S/w2)
AZ=
"i
Positive
Finite
Gain Trans. Func.
'1 Neaative
Finite1
Gain Trans. Func
5
SL + (l+k+19
w1w2w3
1 (l+k)>(l+a)(l+6)
Positive
Finite
Gain Trans. Func.
I,
Negative Finite
Gain Configuration
no terms containing differences are encountered; thus, low
coefficient sensitivities are obtained and reasonable OA
mismatch is tolerated. Also, all the denominator coefficients are positive, which is necessaryfor stability. Applying the same technique used in Sections II and IIIA, it
can be shown that the resistor ratios (Y,p, and i can be
chosen to extend the BW and to satisfy the necessary
stability conditions assumingsinglepoleOA models.
1) Comparisons of the Proposed C3OAs and C4OAs
FiniteGain Realizations with Others: The optimum maximally flat 3dB BW using three (four) singleCiAfinitegain
building blocks is obtained by cascadingthree (four) identical blocks, each with gain k113(k114)to realize an overall
gain k. The overall BW shrinks by a multiplying factor
0.51/k/3(0.435/k4)
relative to wi [31]. The BW of the
new proposed C30A (C40A) circuits are found to shrink
by only a factor 1/k/3(1/k4)
(k B 1).
Maximally flat response (Butterworth) as well as
Chebyshev characteristics, using CNOAs, can be achieved
by controlling the resistor ratios OL,/3, and y while still
satisfying the stability conditions. Computer plots of the
C30A1 and C40A1 transfer functions in the positive
and negative finitegain configurations are given in Figs. 9
and 10 for gains of 100. From Figs. 9 and 10, it is seenthat
the 3dB BW available from these C30A (C40A) finitegain amplifiers implemented using lMHz single OAs
corresponds to the BW attainable from a single OA with
zerodB BW in excessof 25 (35) MHz! The performance of
the C30A1 is compared with the performance of recently
published threeOA realizations, called the zero second
derivative (ZSD) amplifiers, proposed in [lo]. A positive
finite gain of 38.7 is chosen to permit direct comparison
with the theoretical results previously published in [lo].
For practical reasons, both ZSD amplifiers are designed
such that the stability condition, using Rouths test on the
thirdorder denominator coefficients, is exceeded by a
margin of 10 percent. The best theoretical results using the
ZSD are obtainable with the minimum stability margin to
allow for maximum bandwidth, i.e., (rl  1) = 1.1 pk. Fig.
11(a) and (b) shows the theoretical magnitude and phase
characteristics of the ZSD amplifiers, as well as the C30A1
and C40A1 amplifiers (which satisfy the stability constraints), with positive finite gain of 38.7, for different
compensation values. The figure depicts the extended
frequency range of operation attainable with these C30A
and C40A designs over the ZSD realization [lo], or a
realization which uses three (four) cascaded singleOA
finitegain stages. It is interesting to note that the condition for stability of the amplifier shown in Table III using
the C30A1 is
l+k
a< 1+p
where k = 38.7.
458
IEEETRANSACTIONS
ON CIRCUITS
AhSYSTEMS,
VOL.
CAS34, NO. 5, MAY 1987
 14i.6
40.09
aI0
dB
3i.81
a=
1.8
p=
I2
k=
100
 71.8
0.0
i3.54
(I i
 71.8
i0.26
26.99
:l43.6
23.71
215.0
32.73
30.29
27
0
60
160
240
( MAXIMALLY
FLAT
320
400
P = 2
v = 3.9
k = 100
.85
kHz
200
f kHr
300
MAXIMALLY
FLAT
(4
(a)
143.6
, dB
37.20
1
36.11
(3 = Il.8
k
34.27
9
71.8
= 2.6
31.35
34.16
32.21
i = 3.9
k = 100
71.8
143.0
28.43
?I5 .o
25.50
0
60
160
240
320
(CHEBYCAEV)
@I
:.
Fig. 9. Computer plots (magnitude and phase) of the C30A1 transfer
function for gain 100. (a) Computed frequency response (amplitude and
phase) of the positive finitegain amplifier (k =lOO) using C30A1
(single OA GBWP = 1 MHz). (b) Computed frequency response (amphtude and phase) of the positive finitegain amplifier (k =lOO) using
C30A1
a2
p=3
= 100
(single
OA
GBWP
(CHEBYCHEV)
(b)
Fig. 10. Computer plots (magnitude and phase) of the;C40A1 transfer
function for, gain 100. (a) Maximally flat computed frequency response
(amplitude and phase) of the negative finitegain .aihplifier (k = 100)
using .C40A1 (single C)A,GBWP = 1 MHz). (b) Chebyshev computed
frequency response (amphtude and phase) of the negative finitegain
amplifier (k = 100) using C40A1 (single OA GBWP = 1 MHz).
= 1 MHz).
This is satisfied by a wide ,margin in. the ,.C30A1 responses in Fig. 11(a) and (b),for both the maximally flat
response and the Chebyshev,response.,All of: these finitegain designs have the same attractive dynamicrange, stability, and lowsensitivity properties as the C20A designs
in Section IIIA. Also, it is interesting to note that some of
the finitegain designs presented here (C30A2, C30A4,
C40A2, C40A3, and C40A5) have identical N/D multiplying factors in the positive and negative gain applications, which makes them suitable in differential gain applications.
2) Experimental Results Using C3OAs and ,C4OAs in
FiniteGain Applications: Only sample experimental results
using the C30A1 and C40A1 are given to illustrate the
performance. Exhaustive test results are documented in
[33]. Fig. 12 gives the experimental results using the
C30A1 in positive finitegain applications of 38.7. The
computer frequency responseplots of the C40A1 negative finitegain realization in Fig. 10 closely agreewith the
experimental results of Fig. 13.
The stability and low sensitivity to power supply as well
as to the active compensation resistor variations were
verified [33].
IV. CONCLUSIONS
A new approach is presented for extend& the useful
operating frequency range in a wide variety *oflinear active
networks which utilize OAs. The extended SW is.achieved
by replacing each of the single OAs in theactive realization by a composite OA (CNOA). The application of the
CNOAs in finitegain amplifiers is also given.
A systematic procedureis given for the generation of the
CNOAs. Each CNOA is constructed using. Nsingle OAs
and 2(N  1) active compensating lowspread and lowaccuracy resistors, resulting in (N  1) resistor ratios. The
CNOA is versatile since it has three external terminals that
correspond to those of a single OA. The suggestedgeneration method gives rise to a large number of QJOAs for a
given N. For N = 2, 3, and 4, CNOAs%aregeneratedand
examined according to a stringent perfor.mancecriterion
that considers stability, sensitivity, dynar$ range,:CMRR,
BW, and the GBWP mismatch effect of single OAs.
Several of the CNOAs, namely the C2OA:J, to C20A4,
C30A1 to C30A6, and C40A1 to C40A5, .meet the
performance criterion and have been found to be very
useful in practice. In these CNOAs, simple resistor ratios
can be used advantaeeouslvto reduce the deviatidh of the
MIKHAEL
AND
MICHAEL:
COMPOSITE
OPERATIONAL
459
AMPLIFIERS
vo/
20 Log
f kHz
200
300
50
60
70
00
90
f KHz
m
300
400
(b)
Fig. 11. Theoretical results for C30A5, C40A5, and [lo] for positive
finitegain applications (gain k = 38.7). 0:
[lo] r =l.l,
/3 =10m8
(stable with min. margin); 0 : [lo] r = 2, B = 0.2 (unstable): @ :
C30A5 (Y=1.9, p = 5.4; @ : C30A5 OL=i, p = 6.4; @ : C40A5
OL= 0.6, p = 2.4, y = 12.5; @ : cascade of three singleOA finitegain
stages, each of gain (38.7)1/3; 0 : c ascade of four singleOA finitegain
stages, each of gain (38.7)i4. (a) Theoretical amplitude responses of
C30A5, C40A5, and [lo] for positive finitegain applications. (b)
Theoretrcal phase responses of C30A5, C40A5, and [lo] in positive
finitegain applications.
33.7.
27.7.
100
e
200
4
6
3 dB
__..
f kHz
300
Fig. 12. Experimental results of C30A1 in positive finitegain application, for a gain of 38.7, and the effect of variation of the compensating
resistor ratios a and j3 (LM747 OP AMPS).
overall active realizations response from the ideal while
guaranteeing stability.
Finitegain applications utilizing the proposed CNOAs
are shown both theoretically and experimentally to be
stable and to exhibit wide dynamic range and low sensitivity. Comparisons with the the stateoftheart realizations
using similar numbers of OAs in these applications show
the appreciable improvement of the realizations obtained
_
IO,
Fig. 13.
30
40
d0
2
(experimental)
50
IO
20
200
4

400
(4
100
d0
+2
vi
k
100
150
200
250
300
350
400
450
f kHz
e
500
The effect of active compensation on extending the bandwidth
of the C40A1 (using LM747 OP AMPS).
using the proposed CNOAs with respect to stability and
useful BW.
Although the examplesgiven which use CNOAs are for
highgain applications, it is easyto show that the deviation
in amplitude and phasefrom the ideal is much lower than
other existing realizations, even for closedloop gains as
low as unity. Also, another attractive feature of the proposed technique is that, in an integrated implementation,
the chip area and the power consumption of a CNOA are
much less than N times those of a single OA. This is
because in a CNOA there is one output OA only that
drives an external load and that may be required to have
power handling capability.
In addition, it is worthwhile to mention that the method
used to generate the CNOAs is actually a composite
dependent source generation technique,i.e., it is applicable
to any of the four types of dependentsources:voltage (and
current) controlled voltage (or current) sources. Novel
composite dependent sources with considerable performance improvements are expectedto result when the procedure described is applied to the other dependentsources.
Moreover, employing elements other than resistors for
active compensation and using different types of OAs in
the same CNOA (e.g., a highaccuracy OA for the input
OA and a highspeed one for the output OA) are promising and challenging topics for further research and are
presently under investigation.
It is to be noted that the CNOAs, when implemented
using ideal OAs, represent ideal nullors independent of
the absolute values of the compensating resistors. In certain configurations using CNOAs implemented using
frequencydependent OAs, the ratios of one or more of
the compensating resistors to other resistors outside the
CNOA appear in the highorder parasitic terms of T, of
the overall active realization. Although such terms are
small, they can be made negligible by the proper choice of
the impedance level of the appropriate compensatingresistors. This is the approach taken here for simplicity while
still yielding excellent results. Other researchersmay find
the impedance level as an added degree of freedom that
may be used advantageously.
460
IEEE TRANSACTIONS
REFERENCES
PI A. M. Soliman, Instrumentation amplifiers with improved bandwidth, IEEE Circuits Syst. Magazine, vol. 3, no. 1, pp. 79, 1981.
M. A. Reddy, R. Ravishankar, B. Ramamurthy, and K. R. Rao, A
PI
highquality
doubleintegrator
buildingblock
for activeladder
filters, IEEE Trans. Circuits Sysr., vol. CAS28, pp. 11741177,
Dec. 1981.
[31 A. Budak, G. Wullink, and R. L. Geiger, Active filters with zero
transfer function sensitivity with respect to the time constant of
operational amplifiers, IEEE Trans. Circuits Syst., vol. CAS27,
pp. 849854, Oct. 1980.
141 B. B. Bhattacharyya, W. B. Mikhael, and A. Antoniou, Design of
RCactive networks by using generalized immittance converters, in
Proc. IEEE Int. Symp. Circuit Theory, Apr. 1973, pp. 290294.
[51 K. R. Rao, M. A. Reddy, S. Ravichandran, B. Ramamurthy, and
R. R. Sankar, An activecompensated doubleintegrator filter
without matched operational amplifiers, IEEE Proc., vol. 68, pp.
534538, Apr. 1980.
[61 A. M. Soliman, A eneralized active compensated noninverting
VCVS with reduced pfl ase error and wide bandwidth, IEEE Proc.,
vol. 67, pp. 963965, June 1979.
171 S. Natarajan and B. B. Bhattacharyya, Design and some applications of extended bandwidth finite gain amplifiers, J. Frank/in
Inrt., vol. 305, no. 6, pp. 320341, June 1978.
PI R. L. Geiger and A. Budak, Active filters with zero amplifier
sensitivity, IEEE Trans. Circuits Syst., vol. CAS26, pp. 277288,
Apr. 1979.
191 A. M. Soliman and M. Ismail, Active compensation of opamps,
IEEE Trans. Circuits Syst., vol. CAS26, pp. 112117, Feb. 1979.
UOI R. L. Geiger and A. Budak, Design of active filters independent of
first and secondorder operational amplifier time constant effects,
IEEE Trans. Circuits Syst,, vol. CAS28, pp. 749757, Aug. 1981.
WI A. M. Soliman, Classification and generation of active compensated noninverting VCVS building blocks, Znt. J. Circuit
Theory and Ap lications, vol. 8, pp. 395405, 1980.
WI G. Wilson, 4 ompensation of some operationalamplifier based
RCactive networks, IEEE Trans. Circuits Syst., vol. CAS23, pp.
443446, July 1976.
voltage amplifier,
1131 A. Nedungadi, A simple invertingnoninverting
IEEE Proc., vol. 68, pp. 414415, Mar. 1980.
u41 R. Nandi and A. K. Bandyopadhyay, A highinput impedance
inverting/noninverting
active gain block, IEEE Proc., vol. 67, pp.
690691, Apr. 1979.
P51 A. Sedra and P. Brackett, Filter Theory and Design: Active and
Passive. Beaverton, OR: Matrix, 1978.
WI M. Ghausi and K. Laker, Modern Filter Design ActiveRC and
Switched Ca acitor. Englewood Cliffs, NJ: PrenticeHall, 1981.
u71 W. B. Mi ki ael and S. Michael, Active filter design for high
frequency operation, in Midwest Symp. Circuits Syst. (Albuquerque, NM), June 1981, pp. 573576.
WI W. B. Mikhael and S. Michael, A systematic general approach for
the generation of composite OAs with some useful applications in
linear active networks, in Proc. 25th Midwest Symp. Circuits Syst.,
(Houghton, MI), Aug. 1982, pp. 454463.
P91 W. B. Mikhael and S. Michael, Generation of actively compensated operational amplifiers and their use in extending the
operating frequencies of linear active networks, in IEEE Znt.
Symp. Circuits Syst. (Newport Beach, CA), May 1983, pp.
12901293.
WI S. Michael and W. B. Mikhael, High frequency filtering and
inductance simulation using new composite generalized immittance
converters, in IEEE Inf. Symp. Circuits Syst. (Kyoto, Japan), June
1985, pp. 299300.
1211 R. Schaumann, Twoamplifier activeRC biquads with minimized
dependence on opamp parameters, IEEE Trans. Circuits Syst.,
vol. CAS30, pp. 797803, Nov. 1983.
P21 W. F. Stephenson, Composite amplifier structures for use in active
RC biquads, IEEE Trans. Circuits Syst., vol. CAS31, pp. 420423,
Apr. 1984.
1231 T. Fleming, Monolithic sample/hold combines speed and precision, Harris Application Note #538, Jan. 1983.
1241 M. Ismail, S. R. Zarabadi, and G. Myers, Application of composite opamps in nonlinear circuits, in 27th Midwest Symp.
Circuits Syst. (Morgantown, WV), June 1984, pp. 4447.
1251 S. Michael and W. B. Mikhael, Highspeed highaccuracy integrated operational amplifiers, in 27th Midwesf Symp. Circuits
Syst. (Morgantown, WV), June 1984, pp. 792795.
WI CLC103, Fast settling wideband operational amplifiers, Comlinear Corp., Loveland,, CO, Nov. 1984.
1271 A. Antomou, Realization of gyrators using operational amplifiers,
and their use in RCactivenetwork synthesis, IEEE Proc., vol.
116, pp. 18381850, Nov. 1969.
P81 A. C. Davies, The significance of nullators, norators and nullors
in activenetwork theory, Radio Electron. Eng., vol. 34, pp.
259267. 1967.
ON CIRCUITS
AND
SYSTEMS,VOL.
CAS34,N0.
5,MAY
1987
1291 J. Braun, Equivalent N.I.C. networks with nullators and norators,
IEEE Trans. Circuit Theory, vol. CT12, pp. 411412, 1965.
[301 B. D. Tellegen, On nullators and norators, IEEE Trans. Circuit
Theory, vol. CT13, pp. 466469, 1966.
1311 M. S. Ghausi, Electronic Devices and Circuits: Discrete and Integrated.
New York: Holt, Rinehart and Winston, 1985.
~321 S. Michael and W. B. Mikhael, Inverting integrators and active
filter applications of composite operational amplifiers, pp.
461470, this issue.
[331 S. Michael, Composite operational amplifiers and their applications in active networks, Ph.D. dissertation, West Virginia University, Morgantown, July 1983.
1341 T. J. Groom, Precision high speed op. amp. parallel transconductance implementation HA2.548, Harris Semiconductor, Melbourne,
FL.
rIc
Wasfy B. Mikhael (S70h473SM83F86)
was
born in Manfalout, Egypt, on November 3,1944.
He received the BSc. degree (honors) in electronics and communications from Assiut University,
Assiut, Egypt, the M.Sc.E.E. degree from the
University of Calgary, Calgary, Alberta, Canada,
and the D.Eng. degree from Sir George Williams
University, Montreal, Quebec, Canada, in 1965,
1970, and 1973, respectively.
From 1965 to 1968, he was an Engineer with
the Telecommunications Organization, Cairo,
Egypt. From 1970 to 1973, he taught in the Computer Science Department at Sir George Williams University, and in 1973 he taught in the
Mathematics Department, Dawson College, Montreal, Quebec, Canada.
Since May of 1973, he has been a Member of the Scientific Staff at
BellNorthern Research, Ottawa, Ontario, Canada, as well as an adjunct
Associate Professor in Electrical Engineering at Sir George Williams
University (now known as Concordia University). In August 1978, he
joined the faculty of West Virginia University, where he is now a
Professor. of Electrical Engineering. His research interests are active
networks, switchedcapacitor circuits, and adaptive signal processing. He
has several patents and publications in the area of communication
networks and active filters.
Dr. Mikhael was the recipient of the Bell Northern Research Outstanding Contribution
Patent Award in 1978, the Outstanding Researcher
Award from the College of Engineering, West Virginia University, in
1982 and 1983, and the Halliburton Best Researcher Award in 1984. He
served as Chairman of the 1984 Midwest Symposium on Circuits and
Systems. Dr. Mikhael is presently Associate Editor of the Letters Section
of the IEEE TRANSACTIONSON
CIRCUITSAND
SYSTEMS.
Sherif Michael (S78M83)
was born in
Alexandria, Egypt. He received the B.Sc. degree
in electrical engineering (electronics and communications) from Cairo University, Cairo,
Egypt, in 1974. He received the MSc. degree in
industrial engineering and the Ph.D. degree in
electrical engineering from West Virginia University, Morgantown, WV, in 1980and 1983,
respectively.
He received technical training at Philips
Industries, Eindhoven, Holland. He served as a
First Lieutenant in an Engineering Corps specializing in water well
drilling, where he conducted, as a Field Engineer, electronic measurements and supervised drilling operations in cooperation with Schlumberger Co. Dr. Michael worked as a research and teaching fellow at West
Virginia University. As a Research Engineer with the National Transportation Research Center, Morgantown, WV, he worked on designing and
implementing a new digital communication system for the Morgantown
Personal Transit System (MPRT). Since 1983, he has been an Assistant
Professor with the Department of Electrical and Computer Engineering
at the Naval Postgraduate School, Monterey, CA. His present research
interests are in the area of analog integrated circuits and active networks
design, radiation hardening, solar cells and space power applications.
Dr. Michael is a member of Eta Kappa Nu, Alpha Pi Mu, and Tau
Beta Pi and is a registered Professional Engineer.
Study Guide B
EEL 4140
ANALOG FILTERS DESIGN
Inverting Integrator and Active Filter Applications of Composite Operational Amplifiers
461
IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,
VOL. CAS34, NO. 5, MAY 1987
Inverting Integrator and Active Filter
Applications of Composite
Operational Amplifiers
SHERIF MICHAEL,
MEMBER, IEEE, AND
Abstract A new approach for extending the useful operating frequency
range of linear active networks realized using operational amplifiers (OAs)
has been reported [l]. The extension in bandwidth (BW) is achieved by
replacing each of the single OAs in the active realization by a suitable
composite OA (CNOA) that has been constructed using N OAs.
The use of the CNOAs to realiie inverting integrators and active filters
is presented here. The considerable performance improvement of these
realizations is demonstrated both theoretically and experimentally. Their
comparison with stateoftheart designs is also given.
I. INTRODUCTION
N RECENT YEARS, a great deal of attention has been
directed toward designing highperformance integrators
and active RC filters. Severalcontributions using a variety
of techniques have been reported to extend the useful.
operating frequency range of thesenetworks and to reduce
their sensitivities with respect to the active elements,
namely, the operational amplifiers (OAs) [3][24]. It is
well known that the poles and zeros actually realized are
displaced from their nominal positions because of the
frequencydependent characteristics of the operational
amplifier gains. This results in both phase and magnitude
errors in the response,especially if the active networks are
designed to operate at high frequencies and/or with high
Q s.
In [l], the authors propose a technique for extending the
operating frequency range of linear active networks by
using composite OAs. A general procedure is described
where N OAs are combined to form a new active device
that resembles externally an OA and is referred to as a
composite operational amplifier CNOA. The technique
generatesa very large number of CNOAs for a given N.
The effect of using suitable CNOAs as onetoonereplacements of the single OAs in finitegain realizations is found
to result in an extended operating frequency range relative
to that of existing realizations that use a similar number of
OAS.
Manuscript received October 29, 1985; revised August 26, 1986.
S. Michael is with the Department of Electrical and Computer
Engineering, Naval Postgraduate School, Monterey, CA.
W. B. Mikhael is with the Department of Electrical Engineering, West
Virginia University, Morgantown, WV 26508.
IEEE Log Number 8613470.
WASFY B. MIKHAEL,
FELLOW, IEEE
In this paper, the use of the CNOA families in the
realization of inverting integrators and active filters is
investigated. In Section II, it is shown, both theoretically
and by computer simulations, that the members of these
new C20A and C30A families extend the operating
frequency range of inverting integrator realizations considerably beyond the presently available stateoftheart designs. In addition, the proposed technique results in practically useful designs,as is demonstratedby experiment later
in Section III.
Appreciable improvements in the performance of different types of modern active filters using the CNOAs are
shown in Section III. Here, the active filters are considered
to belong to two categories.The first category consists of
filters which are realized with functional building blocks
(finitegain amplifiers and integrators), while the second
category consists of filters in which the OAs are embedded in the passivenetwork, so that functional building
blocks cannot be identified from the filter structure. Active
filters in the first category are shown, theoretically in
Section IIIA and experimentally in Section IIIB, to provide significant improvement at high operating frequencies
when the improved inverting integrators proposed here
and finitegain amplifiers proposed in [l] are used. An
application is given in Section IIIB by employing the
C20As in the twoiniegratorloop filters as an example.
The extension in BW is achieved while maintaining low
sensitivity to passive and active elements, wide dynamic
range, and stable operation.
In Section IIIC, the theory for extending the active
filters operating frequencies in the second category is
presented. The application of the theory, supported by
experiments, is given in Section IIID, where a suitable
C30A is used in one of the wellknown multiplefeedback
(MFB) structures. The resulting filter maintains the practical useful features such as stability, low sensitivity to the
compensating elements,and signal handling capacity. Also,
comparison with one of the stateoftheart designsis given
which shows considerableimprovements in bandwidth and
stability of the proposed technique. It is worthwhile to
note that Schaumann [l, ref. [21]] successfully demon
009%4094/87/05000461$01.00 01987 IEEE
IEEE TRANSACTIONSON CIRCUITSAND SYST!2MS,VOL.CAS34,N0. 5,MAY1987
462
TABLE I
INVERTINGINTEGRATORSV,/ I: USINGTHE C20As
Actual
CZOAi
CZOA1
Negative
Integrator
1
T. . c
1
l+S/wpQp+wwpZ
&
Transfer
I*
Function
(T,)
'W,(l+a)
l+T,UJ,(l+al
C20A2
CZOA3
Ti
* [
CZOA4
Ti
. '1
+ S/wpQp
Function
Ti
1 + (,+a)
I
s/w,
YW2
lx
+ SLl~p~l
v
Ideal
(Where
Transfer
Tt
is
the
integrator
= F=
i
time
= &,
constant
= RC
= 11 wt
)
"i
I
i
i/ ,. _
Fig. 1.
The composite operational amplifiers C20As [l].
463
MICHAEL AND MIKHAEL: APPLlCATIONSOFCOMPOSlTEOPERATIONALAMPLIFIERS
strated the drastic active sensitivity improvement of the
Deliyannis filter, which belongs to category 2, by using
C20A4.
TABLE11
VALUESOF~EC~OA'~RESISTORCOMPENSA~ONRATIO &FOR
SELECTEDVALUESOFQ ,ANDT~CORRESPONDINGSTABXLITY
CONDITIONSFORTHE PNVERTINGINTEGRATOFSUSINGTHE
CZOAi
II.
APPLICATIONOFTHEPROPOSEDCNOA'SIN
INVERTING INTEGRATORS
A. Inverthg
Integrators Using C.JOAs
The amplifiers C20A1 through C2OA4 were previously proposed by, the authors and are repeated in Fig. 1
for convenience.The actually realized transfer functions T,
using the C2OA1 to C20A4 in inverting integrator applications, assuming a singlepole OA model, are given in
Table I. The Tas have the samegeneralform found in [l]:
ity
for
(I used
0
CZOA1
Unrealizable
0
CZOA2
Unrealizable
z?
CZOA3
where T. is the transfer function realized assuming ideal
OAs and
N=l+a.s=l+(s/o,)
( ciis zero (0, + w) in somecases)
CZOA4
Thus, N/D determines the amplitude and phase deviation of T, from T.. Also, b, and b, determine the stability
of T,. Here, a, b,, and b, and consequently wZ, wp, and
Qp (as defined in Table I) are functions of the circuit
parameters which are the singleOA GBWPs ot, 02, and
the C20A compensation ratio (Y(as well as TV,the integrator time constant). None of the a and b coefficients are
realized through differences, thus guaranteeing the low
sensitivity of T,, wZ, (jp, and Qp with respect to the circuit
parameters. On the other hand, the b coefficients are
always positive (assuming a singlepole OA model), which
x guaranteesthe stability of the transfer function.
Now let us assumea twopole openloop gain Ai of the
ith single OA (i = 1,2). l/Ai is given by
where
oL, 3dB bandwidth 2 wi/A,,,
oi gainbandwidth product (GBWP) of the singlepole
model of the ith OA,
Aoi dc gain of the ith OA,
w,,, second pole frequency (w,, Z+ wL,).
It is easyto show [l, ref. [33]] that, for identical singleOA
models (wh, = o,, = oh and oi= w2= wi), the necessary
and sufficient stability conditions (the parasitic poles are
in the left half of the s plane) for the C20A1 and
C20A2 integrator realizations are given by
1
A
Unrealizable
= 1
A
3w.
= wj
= qfi
Wh '
Wh >
3"'
I 0.666
wi is the GBWP of each single OA and. w,, is tile singleOA second
pole frequency.
For the C20A3, the condition is given by
(4)
Also, for the C20A4, the stability condition is given by
It is clear from (3), (4), and (5) that, for a given (Y,there
is a minimum value of wh/oi which satisfies the stability
condition. In C20A1 and C20A2 integrators, the condition a = 0 results in the minimum value for wh, since wh
increases as (Yincreases.The C20A2 integrator requires
wh/wi to be greater than 3/2 for stable operation when (Y
equals zero and wh/wi to be greater than 2 when a tends
toward infinity. The C20A4 integrator has an advantage,
since for wh = oi, a value of OLexists (a = l/3) for which
the integrator is stable with excellent frequency response.
In general, as wh increases,the stability improves. From
physical considerations, if o,, + cc, all the integrators become stable, as can be seen from Table I, since the
twopole OA model reducesto a singlepole model (Ai =
w/(si + oL )). The stability conditions for particular values
of (Y are summarized in Table II. From graphical data
sheets of internally compensatedOAs such as741s and
747s, w,, is seen to slightly exceed wi [2]. This results in
stable realizations down to 0dB closedloop gain. For
464
IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,
VOL. CAS34, NO. 5, MAY 1987
externally compensated OAs such as the 702 and 709, (Y
can be chosen in the range that guaranteesstability while
yielding the most desirable T,(s).
A theoretical comparison of the C20A4 negative integrator with stateoftheart negative integrator realizations [3][5] is given in Fig. 2 for ot/oi = 0.05 (w, =l/rt,
where rr ,is the integrator time constant). The theoretical
results in Fig. 2 shows that the percentage deviation in
magnitude and phase from the ideal for the proposed
C20A4 negative integrator, with (Y= l/3, very nearly
matches the excellent performance of the integrator in [3].
(The C20A4 integrator with (Y= 0 is identical to the
integrator in [3].) In practice, the integrator in [3] suffers
from stability problems while the proposed one does not,
as verified experimentally by the authors in active filters
(twointegratorloop biquadSection IIIB) when.internally
compensated OAs with a phasemargin of less than 60 at
0dB closedloop gain (wh/wi < 3/2) are used. Thus, the
importance of the controlling parameter (Y is apparent
since it guarantees stable operation using commercial internally compensated OAs without sacrificing performance.
% deviation
from ideal
IHI
6%
CBOA4
7%
6%
5%
4 %
3 %
2 %
I %
% deviofion
from ideal
B. Inverting Integrators Using C3OAs
In this section, three highfrequency integrators which
use actively compensated multiple OAs (CNOAs) for
N = 3 are introduced. The different structures of these
C30As, together with their actually realized openloop
transfer functions, are given in [l] and [l, ref. [33]]. C30A2,
C30A5, and C30A6, which are employed here, are given
in Fig. 3(a) for convenience.
Using the C30A2, C30A5, and C30A6 in the inverting
integrator
circuit
transfer functions.
Using the C30A2,
T,=q.
yields
the following
integrator
04
Fig. 2. Comparison of the C20A4 negative integrator for (Q, =
0.707, Qp = 0.835) with existing negative integrators proposed in [3][5].
(a) Percentaee deviation from ideal of the transfer function magnitude
versus no&lized
frequency
for the proposed C20A4
i&g&r
and
several other integrators (l/r,w, = 0.05). (b) Percentage deviation from
ideal of the transfer function phase versus normalized frequency for the
proposed C20A4 integrator and several other integrators (l/r,wi = 0.05).
(6)
Using the C30A5,
T,=T.
(7)
Using the C30A5,
T, = q.
(8)
465
MICHAEL AND MIKHAEL: APPLICATIONSOF COMPOSITEOPERATIONALAMPLIFIERS
where wi is the GBWP of the OAs used. T. the ideal
integrator transfer function is equal to  w,/s (T, ) 7J as
the OA Ais + co). In (6)(8), it can be shown that no
difference terms appear in any of the numerator and
denominator coefficients when OAs with different gains
are used. Thus, the low coefficient sensitivities are achieved
and the necessarycondition for stability is satisfied without
the need for matched amplifiers. Assuming a singlepole
OA model, the necessary and sufficient conditions for
stability, i.e., for the roots of D to be strictly in the left
half of the s plane, are as follows.
For the C30A2 integrator,
b
+[;][1+,(1;.)]>l. tsa)
For the C30A5 integrator,
[ l+;f+)][(~+@+~+;(l+P)]>l.
from
ideal
(9b)
For the C30A6 integrator,
(1+0)(1+~)
51.
I
It can be easily shown that these new integrators can be
designed to satisfy the above stability conditions, while
allowing a wide range of (Yand p variations. For illustration, Fig. 3(b) and (c) shows sample theoretical results
using the C30A6 and those obtained using the integrator
in [3] that employs two OAs. The performance improvement in both magnitude and phaseis obvious.
4%
3 %
2%
I %
f ktiz
20
40
60
80
100 120 140 160 180 200 220 240 260 280
III.
(b)
EXTENDING THE ACTIVE FILTER OPERATING
FREQUENCY RANGE USING CNOAs
Active filters have been designedusing a wide variety of
approaches [6][lo]. In this contribution, active filters are
considered to belong to one of two categories. The first
category includes active filters that are realized using functional building blocks, namely, inverting integrators and
finitegain amplifiers. Examples are the positivegain
Sallen and Key filter, the twointegratorloop filter, and
the SFG filter [6]. The second category contains those
filters whose OAs are embeddedin the passive network,
and functional building blocks cannot be isolated in the
filter structure. Examples are the multiplefeedback (MFB)
filter and the generalizedimmittance converter (GIC) filters
WI, [91
5 %
4 % .
20
40
60
60 100 120 I40
I60 I60 200
220 240 260 260
f kHz
(4
Fig. 3. Comparison of C30A6 negative integrator and the one proposed in [3] (0,/w, = 0.1). (a) The composite operational amplifiers
C30A2, C30A5, and C30A6 [l]. (b) Percentage deviation from ideal
of the transfer function magnitude versus normalized frequency for the
proposed C30A6 negative integrator and the one proposed in [3]
(o,/w, = 0.1). (c) Percentage deviation from ideal of the transfer function phase versus normalized frequency for the proposed C30A6
negative integrator and the one proposed in [3] (at/w, = 0.1).
A. Improving the Performance of Active Filters in the
First Category
It is easy to show that for filters in the first category, as
.
the behavior of the active functional building blocks approaches the ideal over a wider frequency range, extended
operating frequencies are obtained. This can be explained
as follows.
IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,
VOL. CAS34, NO. 5, MAY 1987
466
Let us assume an active filter that realizes a transfer
function T, using functional active building blocks
whose individual transfer functions are G,, G,, . . . , G,.
Here, T, is the actually realized transfer function using
nonideal (frequencydependent) active building blocks.
GpGz,*, G,, are the actually realized (nonideal) transfer
functions of the building blocks using frequencydependent active elements(OAs). In addition, let q, Gii,. . . , Gni
be the corresponding ideal transfer functions when the
active elements (OAs) used are frequency independent;
i.e., the gainbandwidth product (GBWP,) of the OAs is
cc (7i = 0, where 7i =l/GBWP =l/bi). T, can be written
as
Ta=f(G1,GZ;,G,).
(10)
Similarly, the jth gain Gj realized using the jth CNOA,
whose OA ris are r,j to rNj, can be expressedas
Gj=g(r1j;r2j,...,r11[1)
01)
where j=l,2;..,
n. Here, f (.) and g( .) denote functional dependence. For simplicity, let n = 1 and N = 2.
The following argument can be shown to be valid for any
integer values of n and N. Also, let G, = G and Gii = Gi.
The Taylors seriesexpansion of T, about Gi is given by
(G Gi)
Tl=%=a,+ dG G=C,
1 aT,
.
(GGJ*+
(12)
Equation (12) can be rewritten as
AT=T(AG)+
;
i
= TAG
T(AG)2+
...
for AG very small
(13)
where
AT=T,T.
and AG=GGi.
Similarly, from a Maclaurin series expansion of (11) in
ri and r2 about their ideal values of ri = r2 = 0, it can be
shown that
of the active filter, but independent of the nonidealnessof
G. Thus, from (13) and (15), to minimize AT, both the
phase and gain deviations in G of the functional building
block, corresponding to AGs real and imaginary parts,
must be appropriately minimized using the CNOAs compensating resistor ratios in the expansion of G given by
(15). The minimization can be carried out in many different ways. A simple but less exact approach is to minimize
each ]AG] individually. This is found useful in practice and
is applied in Section IIIB. Another more exact but more
involved approach is to substitute for AG from (15) in AT
given by (13). Then, the compensating resistor ratios are
chosen to minimize AT as desired. In both approaches,AT
may be minimized at a critical frequency or over a band of
frequencies.
At this point, two pitfalls are pointed out. The first may
occur if one attempts to choosea structure for realizing G
which results in zero lower order derivatives G;, G;, G;, . . .
of G with respect to its parameters. Great care should be
exercised in attempting this, since the higher order derivatives of G may be increased in a manner that offsets the
reduction in AG obtained by nulling the lower order
derivative terms, particularly as the frequency increases.
Note that, for a given number of OAs, a realization that is
characterized by zero loworder sensitivities may not be
optimum; there may exist a structure that fails to satisfy
the zero loworder sensitivity property but results in a
smaller AT. This can be easily seen by comparing the
performance of the C20A4 with the C20A1 in finitegain
applications [l].
The second pitfall may result if one controls the phase
angle of G by moving some of the parasitic poles of G to
the right half of the s plane to cancel the phase shift due
to the other poles and zeros.This is intolerable in finitegain
applications and very undesirable even when the active
building block is embedded in filter structures, since it
leads to local instabilities. This has been examined by
researchers and referred to as stability during activation
[25][27].
To conclude this subsection, it follows from (13) that if
the active building blocks in a filter are replaced by the
proposed building blocks which have a smaller AG, the
filters performance can be improved.
B. Improving the Performance of an Active Filter in the
First Category (A MultipleAmplifier Biquad) Using C2OAs
Equation (14) can be rewritten as
AG = Girl + G$r2+ i [G ;rf + G;rz + 2G$r1r2] + +. . .
(15)
It is to be noted that, evenif Gi is real, AG is complex in
general. Also, T , T , . . . are dependent on the topology
In this subsection, it is shown how the operating frequencies of the first category of filters can be extended
through the use of the functional building blocks that are
presented in [l] and in Section II. These blocks allow the
minimization of AG and the necessary tradeoff between
gain and phase deviations to achievestable highfrequency
operation with low sensitivity to the active compensation
elements. As mentioned above, several techniques are possible for error minimization and bandwidth extension in
this category of filters. A less global and more straightforward approach is employed in this subsection for demonstrating where the bandwidth of each block is individually
extended.
467
MICHAEL AND MIKHAEL: APPLICATIONSOF COMPOSITEOPERATIONALAMPLIFIERS
Cl
CP
350.
300
CZOA2)
250
C20Adrl
200.
150.
(4
I oo50
%
100
')Pf
/
Lop;
SINGLE OA
30
SlNGLE OA
30
2s
20
15
I oo
IO
MEASURED SING C2OAl
50.
OQf
20
40
60
80
(4
100
120
(4
Fig. 4. Experimental results of the twointegratorloop BP filter using the proposed C20A2 and C20A4 and the theoretical
using CZOA2 and C20A4. (b) Percentage
results obtainable with regular OAs. (a) Bandpass filter (twointegratorloop)
variation of Qp, (filter) as a function of Q for bandpass filter (wpf = 50 krad/s). (LM747 OP AMP). (c) Percentage
variation of Qp/ as a function of Qp, for bandpass filter (o , = 30.8 krad/s) (LM747 OP AMP). (d) Percentage variation of
Qp, as a function of fp, for bandpass filter (LM747 OP ARIP).
A biquadratic active filter, which uses the functional where QPr is the complex polepair selectivity factor, and
building blocks, is designed and tested. The filter that is
chosen is the wellknown statevariable filter [lo], and is
(18)
shown in Fig. 4(a). It usestwo inverting integrators and a
differential finitegain amplifier, which are constructed
using the C20A4 integrators proposed in Section II and From [l], for maximally flat response(Q, = l/a)
of the
the C20A2 differential amplifier [l]. The biquads trans differential amplifier, 1y= 9
 1 = 0, independent of
fer function 7;(s) (at the bandpassoutput) is given by
X and QP,. Also, from Sect& II and Table II, a! is set
l+R,/R,_. _
s
equal to unity (maximally flat) in the C20A4s integra.tors. In Fig. 4(b)(d), the experimental results for the filter
1+&/R,
W,
using
the C20As are compared to those utilizing the
WhP
=
RJR,
' (16)
7  s l+R,/R,
r&I
I
single OAs. In addition to the appreciableimprovement in
R,C, l+R,/R,
' R,R,C,C,!
the useful operating frequency range demonstratedin Fig.
4(b)(d), excellent theoretical sensitivity, stability, and dyThe elements are chosen as
namic range are also verified experimentally. Total
harmonic distortion (THD) of much less than 1 percent at
C,=C,=C
R,=R,=R,=R,=R,=R
any of the filter outputs is measuredover a wide range of
and aPr (the complex polepair resonant frequency) = frequencies and signal levels; e.g., the THD was found to
1/RC.
be less than 50 dB below the fundamental at fPf, for an
Referring to [lo], thesedesign values correspond to
output voltage swing of 12 V peak to peak, where
vpower supply = 12 V, r,f = 16 kHz, and QP, =IO. It is very
interesting
to note that when (Yis set to zero (which results
X=2Q*,1+4
(17) in a previously
reported integrator [3], [ll], [12]), the filter
3
&=l.
.J
IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,
VOL. CAS34, NO. 5, MAY 1987
468
From (24) and (26), to minimize AT, t, which is complex
in general, has to be minimized appropriately using the
CNOAs compensating resistor ratios (OLS,Rs, etc). As
stated before, in Section IIIA, error minimization can be
done in many different ways depending on the cost function and the optimization criterion. In the following subsection, a way of improving the performance of a filter in
the second category by employing CNOAs is given.
0 "0
D. Improving the Performance of an Active Filter in the
Second Category (A Singletimplifier Biquad Using C3OA)
For a given filter in this category, it is feasible to derive
closedform expressions,by combining (24) and (26), to
Fig. 5. MFB active BP filter using C30A1.
minimize the error appropriately. In this subsection, a
simpler but equally useful approach is adopted to demonbecame unstable in practice, as predicted from the stability strate the usefulness of the proposed technique. In the
following results, the CNOAs compensatingresistor ratios
analysis in Section II.
(as and Rs) are chosen using the search method in a
straightforward computer program that minimizes the
C. Improving the Performance of Active Filters in the
filters response deviation from the ideal and satisfies
Second Category
In the second category, T, can be expresseddirectly as a stability.
A multiplefeedback (MFB) biquadratic bandpass acfunction of the t's of the CNOAs as follows:
tive filter is considered here (Fig. 5), as an example to
T,=f(t,,t,,,t,)
09) illustrate the improved filter performance using one of the
where t,, t,, . . e,t, are reciprocals of the openloop gains proposed C30As [l], namely, the C30A1. For ideal
of the first, second,. . . , n th composite amplifier, respec OAs, i.e., Ai + cc (i = 1,2,3), the filter transfer function Ti
tively. For a composite amplifier constructed using N is given by
OAs, tj can be expressedas
 R,CsH
$L
(27)
V.I 1+2RCs+R
tj=g(71j,T2j,"',7NI)
(20)
1
1R 3C2s2
where tj is the reciprocal of the openloop gain of the jth where H = R2/(Ri + R2), R, = Ri//R2,
uPr =
CNOA and 7ij is the reciprocal of the GBWP of the i th l/(Cm),
and Qp, = (1/2)/m.
Assuming, for
OA used in constructing
the jth
n =1 and N = 2. Also, let
CNOA.
t,
For simplicity,
let
simplicity,
identical
frequencydependent
OAs and R X
~ii = 7i, and 721= r2. R,, the actually realized transfer function T, is given by
t,
Hence,
T,=f(t)
(21)
T,s
{l+s[zCR,+~
( sCR3H(1+;}/
and
t = &I,
72).
(22)
Thus, the Maclaurin series expansion of (21) about its
ideal value of t = 0 is given by
+wi(l+P)
T,= T,ltzO+ aT,
at (t=o+)+(
f)Jgq=o*(,+
*..
(23)
t=O *t+Yi
=Tt+
;
i
Also,
at,2
zy(t)+
.(t)2+
...
(24)
+S4
to
**.
(25)
is given by
t=
at
aT1
T1+ 71  7* = 0
at
Jr2
(2CR, + CR,)
72 +
7, = T2= 0
(26)
CR,R,
2 CR1
<+ +
A
o;.(l+P)
0,
C2R,R3
w?(1+ a)
1
f&l + a)
C2R,R2 + (2CR, + CR,)
0,(1+/3) +
1 a2T,
at
+s2
C2R,R,
Equation (23) can be rewritten as
AT23
1[
wi
w;(l+ a)
+
w; I
+
(28)
Again, it can be shown that the dynamic range of this filter
is identical to that obtained if a single OA replaces the
C30A. Also, not only is OA gain matching not required,
but also difference terms do not appear in any of the
MICHAEL AND MIKIIAEL: APPLICATIONSOF COMPOSITEOPERATIONALAMPLIFIERS
469
RouthHurwitz stability criterion, it can be easily shown
numerically that the necessaryand sufficient conditions
are well satisfied with a wide margin for the component
values (Yand p. Ideal and experimental results are shown
in Fig. 6. The experimental results in Fig. 6 illustrate the
appreciable improvement over a singleOA realization in
the MFB structure under consideration at this QP, and aP,
dB
46
W[101.
IDEAL MFE FILTER
(ideal OAs)
QPpxpsrimsntol
= 9.7
fptexperimental
= 37.2
fpfhHz
c
1
42
46
Fig. 6. Experimental amplitude response of the MFB bandpass filter
using C30A1, and its variations due to the compensating element
variations.
(+$dooI
%
10 _
C3OAl
9
II31
a=o,p=m
9=
0.2, P =
87.
6J4.
10
20
30
40
50
60
70
90
90
100
(a>
AAx
I
100
QPf
fp, kH*
IO
20
30
40
50
60
70
80
90
%
(b)
Fig. 7. Experimental results of the MFB filter designed using the proposed C30A1 and the design proposed in [13]. (a) Percentage variatrons in center frequency of the MFB filter for Q,,/ = 5, 10, 20. (b)
Percentage variations in Q,,f for Q,,, 7 5, 10, 20. * Trapezoidal oscillations were encountered. Limiting diodes were necessary.
transfer function T, coefficients, even when different OAs
are used.
A BP filter with j, = 37.9 kHz and Qp, = 10 is designed
and tested. The design values are C = 1 nF, Rj = 250 Cd,
R, = 210 a, R, = 1.3 k& and R, = 84 kQ. The OAs used
are LM747s with a GBWP ~1 MHz. Applying the
Performance comparisons are made with the recently
reported highfrequency filter design given in 1131,which
requires the same number of OAs as C30A1, namely
three.
The experimental results using the proposed C30A1
design and the design in [13] of the MFB BP filter in Fig. 5
are given in Fig. 7. No error minimization is attempted at
each operating aPr and Q,,. Rather, the compensation
remains fixed for all the results in Fig. 7. In the C30A1
design IX= 0, p = cc (compensating resistors R' and /3R
are open, while aR' and R resistorsare short).
In the design in [ 131,0 = 0.2, p = 1.
Fig. 7(a) shows the percentage deviation in &, as a
function of fPf for Qpr = 5, 10, and 20. Fig. 7(b) shows the
percentage deviation in Qpr as a function of jp, for
Q,,, = 5, 10, and 20. The comparisons in Fig. 7 show
clearly the performance improvements in the proposed
filter design.
IV. CONCLUSIONS
A general technique for the synthesisof composite OAs
(CNOAs) and for using them in extending the BW of
finitegain amplifiers has been presentedin [l].
In this paper, applications of the CNOAs in inverting
integrators and active filters are investigated. Theoretical
and experimental results of utilizing the proposed CNOAs
(1v = 2,3) in inverting integrator applications show that
the BW is extended considerably beyond present stateoftheart techniques,while other important properties such
as stability and low sensitivity are maintained.
To analyze the effectiveness of this approach, active
filters are considered to belong to one of two categories.
The first category consists of filters realized using functional building blocks (finitegain amplifiers and inverting
integrators), while in the second category the OAs are
embedded in the passivenetwork so that functional building blocks cannot be isolated from the filter structure.
Filter examples are consideredfrom both categoriesand
are constructed using the proposed CNOAs and the improved functional building blocks. The circuits thus attained are shown to perform as well as or better than the
best available stateoftheart designs with respect to ,extension of BW, stability, sensitivity, and deviation from the
ideal response.
In conclusion, an integrated technique for alleviating the
problem of BW limitation due to OA frequency dependence has been presentedwhich extends the useful BW of
linear active networks using OAs without impairing other
important properties such as stability, sensitivity, tolerance
of OA mismatch, and dynamic range. It is worthwhile
IEEE TRANSACTIONSON CIRCUITSAND SYSTEMS,VOL.CAS34,N0. &MAY 1987
470
mentioning that the technique described here and in [l] is
general and is applicable to a wide range of linear active
realizations using OAs, as well as other controlled sources.
[26]
[27]
A. Antoniou, Stability properties of some gyrator circuits, Electron. Left., vol. 4, pp. 510512, 1968.
W. B. Mikhael and B. B. Bhattacharyya, Stability properties of
some RCactive realizations, Electron. Let?., vol. 8, no. 11, pp.
288289, June 1972.
REFERENCES
PI W. B. Mikhael and S. Michael, Composite operational amplifiers:
Generation and finitegain applications, pp. 449460, this issue.
PI The Linear and Interface Circuit Data Book for Design Engineers,
Texas Instruments, Inc.
131 G. Bailey and R. Geiger, A new integrator with reduced amplifier
dependence for use in active RCfilter synthesis, IEEE Int. Symp.
Circuits Syst., 1980, pp. 8790.
141 A. M. Soliman and M. Ismail, On the active compensation of
noninverting integrators, IEEE Proc., vol. 67, pp. 961963, June
1979.
151 P. 0. Brackett and A. S. Sedra, Active compensation for highfrequency effects in op. amp.circuits with applications to activeRC
filters, IEEE Trans. Circuits Syst., vol. CAS23, pp. 6873, Feb.
1976.
161 A. Sedra and P. Brackett, Filter Theory and Design: Active and
Passive. Beaverton. OR: Matrix. 1978.
171 S. Mitra, ,4nalysis and Synthesis bf Linear Active Networks. New
York: Wiley, 1969.
PI G. C. Temes and S. K. Mitra, Modern Filter Theoty and Design.
New York: Wilev 1071
Modern Filter Design ActiveRC and
[91 M. Ghausi an ;iKTL&er,
Switched CaDacitor. Enelewood Cliffs. NJ: PrenticeHall. 1981.
[lOI L. P. Huelsman and P.E. Allen, Intioduction to the Theory and
Design of Actibe Filters. New York: McGrawHill, p. 217.
WI M. A. Reddy, R. Ravishankar, B. Ramamurthy, and K. R. Rao,
Highquality
doubleintegrator buildingblock for activeladder
filters, IEEE Trans. Circuits Syst., vol. CAS28, pp. 11741177,
Dec. 1981.
A. Reddy, S. Ravichandran, B. Ramamurthy, and
WI E: R., R;&g
An activecompensated doubleintegrator filter
without matched operational amplifiers, IEEE Proc., vol. 68, pp.
534538, Apr. 1980:
[I31 R. L. Geiger and A. Budak, Design of active filters independent of
first and secondorder operational amplifier time constant effects,
IEEE Trans. Circuits Syst., vol. CAS28, pp. 749757, Aug. 1981.
[I41 K. Martin and A. S. Sedra, On the stability of the phaselead
integrator, IEEE Trans. Circuits Syst., vol. CAS24, pp. 321324,
June 1977.
WJI A. M. Soliman and M. Ismail, A universal variable phase 3Lport
VCVS and its application in twointegrator loop filters, in IEEE
Int. Symp. Circuits Syst., 1980, pp. 8386.
1161 S. Ravichandran and K. R. Rao, A novel active compensation
scheme for activeRC filters, IEEE Proc., vol. 68, pp. 743744,,
June 1980.
P71 S. Natarajan, Active sensitivity minimization in SABs with active
compensation and optimization, IEEE Trans. Circuits Syst., vol.
CAS29, pp. 239245, Apr. 1982.
[I81 S. Natarajan, Synthesis of actively compensated doubleintegrator
filter without matched operational amplifiers, IEEE Proc., vol. 68,
pp. 15471548, Dec. 1980.
1191 A. K. Mitra and V. K. Aatre, Low sensitivity highfrequency
active R filters, IEEE Trans. Circuits Syst., vol. CAS23, pp.
670676, Nov. 1976.
WI M. A. Reddy, An insensitive activeRC filter for high Q and high
frequencies, IEEE Trans. Circuits Syst., vol. CAS23, pp. 429433,
July 1976.
VI A. K. Mitra and V. K. Aatre, A note on frequent and Q
limitations of active filters, IEEE Trans. Circuits Syst., vo Y CAS24,
pp. 215218, A r. 1977.
P21 A. S. Sedra an B J. L. Esoinoza. Sensitivitv and freauencv limitations of biquadratic act&e filters, IEEE T>ans. Circbits gyst., vol.
CAS22, pp. 122130, Feb. 1975.
1231 J. J. Friend? A. Harris, and D. Hilberman, Star: An active
biquadratic filter section, IEEE Trans. Circuits Syst., vol. CAS22,
pp: 115121,1975.
A.
Budak and D. M. Petrela, Frequency limitations of active
]241
filters using onerational amnlifiers. IEEE Trans. Circuit Theorv,
,
vol. CT19,p;. 322328, July 1972:
A.
Antoniou,
Realization
of
gyrators
using
operational
amplifiers
]251
and their use in RCactive network synthesis, Proc. Inst. Elee.
Eng., vol. 116, pp. 18381850, Nov. 1969.
Wasfy B. Mikhael (S70M73SM83F86)
was
born in Manfalout, Egypt, on November 3,1944.
He received the B.Sc. degree (honors) in electronics and communications from Assiut University,
Assiut, Egypt, the M.Sc.E.E. degree from the
University of Calgary, Calgary, Alberta, Canada,
and the D.Eng. degree from Sir George Williams
University, Montreal, Quebec, Canada, in 1965,
1970, and 1973, respectively.
From 1965 to 1968, he was an Engineer with
the Telecommunications Organization, Cairo,
Egypt. From 1970 to 1973, he taught in the Computer Science Department at Sir George Williams University, and in 1973 he taught in the
Mathematics Department, Dawson College, Montreal, Quebec, Canada.
Since May of 1973, he has been a Member of the Scientific Staff at
BellNorthern Research, Ottawa, Ontario, Canada, as well as an adjunct
Associate Professor in Electrical Engineering at Sir George Williams
University (now known as Concordia University). In August 1978, he
joined the faculty of West Virginia University, where he is now a
Professor of Electrical Engineering. His research interests are active
networks, switchedcapacitor circuits, and adaptive signal processing. He
has several patents and publications in the area of communication
networks and active filters.
Dr. Mikhael was the recipient of the Bell Northern Research Outstanding Contribution
Patent Award in 1978, the Outstanding Researcher
Award from the College of Engineering, West Virginia University, in
1982 and 1983, and the Halliburton Best Researcher Award in 1984. He
served as Chairman of the 1984 Midwest Symposium on Circuits and
Systems. Dr. Mikhael is presently Associate Editor of the Letters Section
of the IEEE TRANSACTIONSONCIRCUITSAND SYSTEMS.
Sherif Michael (S78M83)
was born in
Alexandria, Egypt. He received the B.Sc. degree
in electrical engineering (electronics and communications) from Cairo University, Cairo,
Egypt, in 1974. He received the M.Sc. degree in
industrial engineering and the Ph.D. degree in
electrical engineering from West Virginia University, Morgantown, WV, in 1980 and 1983,
respectively.
He received technical training at Philips
Industries, Eindhoven, Holland. He served asa
First Lieutenant in an Engineering Corps specializing in water well
drilling, where he conducted, as a Field Engineer, electronic measurements and supervised drilling operations in cooperation with Schlumberger Co. Dr. Michael worked as a research and teaching fellow at West
Virginia University. As a Research Engineer with the National Transportation Research Center, Morgantown, WV, he worked on designing and
implementing a new digital communications system for the Morgantown
Personal Transit System (MPRT). Since 1983, he has been an Assistant
Professor with the Department of Electrical and Computer Engineering
at the Naval Postgraduate School, Monterey, CA. His present research
interests are in the area of analog integrated circuits and active networks
design, radiation hardening, solar cells and space power applications.
Dr. Michael is a member of Eta Kappa Nu, Alpha Pi Mu, and Tau
Beta Pi and is a registered Professional Engineer.
EEL 4140
ANALOG FILTERS
LABORATORY 2
Real Zero and Pole Synthesis
I.
Objective
To study real zero and pole synthesis, and cascade design of firstorder circuits.
II. Introduction
The general form of the transfer functions can be written as the ratio of two polynomials as
follows:
a s m + am 1s m 1 + ... + a1s + a0
T (s ) = m
s n + bn 1s n 1 + ... + b1s + b0
(1)
where the numerator coefficients a0 , a1 ,..., a m and denominator coefficients b0 , b1 ,..., bn 1 are
real numbers, and n is the order of the filter.
The degree of the numerator polynomial must be less than or equal that of the denominator
polynomial for causality reasons. That is
mn
(2)
The numerator and denominator polynomials can be factored, and T (s ) can be expressed
as:
a (s z1 )(s z 2 )...(s z m )
T (s ) = m
(s p1 )(s p2 )...(s pn )
(3)
where the zeros, z1 , z 2 ,..., z m , and the poles, p1, p2 ,..., pn , can be either real number or
complex number.
Complex zeros and poles, however, must occur in conjugate pairs. In this experiment, only
real zeros and poles are discussed. In this case, the transfer function T (s ) is factored as:
T (s ) = T1 (s )T2 (s )...Tn (s )
where Ti ( s )' s, i = 1,2 ,...,n are bilinear transfer functions.
(4)
Once the transfer function is factored into multiple stages, the task is to synthesize each stage
using firstorder circuits.
Real Zero and Pole Synthesis
Synthesizing real poles and zeros is based on the cost of the energy storage elements. For
resistor, capacitor and Op Amp circuits, this requirement consists of limiting size or value of
capacitors. Capacitors remain the most expensive and sensitive components in the discrete,
hybrid, and integrated circuit design. Cost implies not only dollars but also size, weight,
parasitic sensitivity, tolerance allocation, and sensitivity to environment changes. One of the
designers main functions is to choose reasonable capacitor values.
According to pole and zero locations, we choose suitable circuits to synthesize these poles
and zeros. Then, we find a set of relations between zeros and poles, and unknown resistor and
capacitor values. The number of the unknown values is generally more than the number of
equations. This property does not represent a problem but rather an opportunity. Usually, the
design procedure begins with a selection of reasonable capacitors. Subsequently, other
components can be calculated.
Cascade Design of FirstOrder Circuits
First, we factor the transfer function T (s ) in terms of bilinear transfer functions Ti ( s )' s as
Equation (4).
Second, we synthesize the Ti ( s )' s using firstorder circuits.
Third, we connect
these circuits in the sense that each successive circuit does not load the previous circuit.
In this
way, the transfer function of this cascading circuit is T (s ) .
Design Procedure for Real Zero and Pole Synthesis
1. The transfer function T (s ) is given by:
T (s) =
(s + 5 10 )(s + 10 )
(s + 2 10 )(s + 10 )
3
Equation (5) is expressed as the product of two bilinear functions, T1 ( s ) and T2 ( s ) .
(5)
T1 ( s ) =
(s + 5 10 )
(s + 2 10 )
(6)
(s + 10 )
(s + 10 )
(7)
and
T2 ( s ) =
2. Choose proper circuits to realize poles and zeros.
For both transfer functions T1 ( s ) and
T2 ( s ) , the same firstorder circuit is used, as shown in Fig.1.
The transfer function of the
circuit in Fig. 1 is given by:
Tcir ( s ) =
R2 ( sC1 R1 + 1)
R1 ( sC2 R2 + 1)
C ( s + 1 ( C1 R1 ) )
= 1
C2 ( s + 1 ( C2 R2 ) )
3. Synthesize the transfer functions T1 ( s ) and T2 ( s ) .
(8)
For both circuits, the capacitance
values are chosen as:
C1,1 = C 2,1 = C1, 2 = C 2, 2 = 0.01uF
(9)
In the first circuit, which synthesizes T1 ( s ) , the resistor values are computed as:
R1,1 = 20k
(10)
R2,1 = 50k
(11)
and
In the second circuit, which synthesizing T2 ( s ) , the resistor values are computed as:
R1, 2 = 10k
and
(12)
R2, 2 = 1k
(13)
4. Cascade the designed two circuits to realize the transfer function T (s ) .
Since the output of
the firstorder circuit is just the output of Op Amps, the output resistance of the firstorder
circuit can be considered as zero. Thus, these two circuits can be connected directly, as
shown in Fig. 2.
Vi (s)
C1
C2
R1
R2
Vo (s )
Fig.1 The noninverting first order operational amplifier circuit
C2,1 = 0.01uF
C1,1 = 0.01uF
Vi (s )
R2,1 = 50k
R1,1 = 20k
C2, 2 = 0.01uF
C1, 2 = 0.01uF
R1, 2 = 10k
R2, 2 = 1k
Fig.2 The cascading design example, which synthesizes real poles and zeros.
(
)( )
(s + 2 10 )(s + 10 )
3
4
function is T ( s ) = s + 5 10 s + 10
3
5
Vo (s )
The resulting transfer
III. Design
The transfer function of the bandpass filter is given by:
(s + 10 )(s + 2 10 )
T (s ) =
(s + 2 10 )(s + 10 )
3
(14)
Decompose the transfer function in the factor format. Synthesize real zeros and poles
using the circuit shown as in Fig.1. Follow the above procedure, and compute the resistance
and capacitance values.
IV. Computer Simulations
1. Simulate the designed bandpass filter.
2. Plot the magnitude and phase responses of each firstorder circuit in the frequency range
from 30Hz to 40kHz.
3. Plot the magnitude and phase responses of the overall circuit in the frequency range from
30Hz to 40kHz.
Note that the magnitude response of the cascading circuit should equal the product of the
magnitudes responses of two firstorder circuits. In addition, the phase response of the
cascading circuit should equal the summation of the phase responses of two firstorder circuits
V. Experiments
1. Build this bandpass filter using two firstorder circuits.
power supply voltage of 15V.
Use LF 351 Op Amps with a split
2. Measure the magnitude response of each firstorder circuit in the frequency range from 30Hz
to 40kHz.
3. Measure the magnitude responses of the overall circuit in the frequency range from 30Hz to
40kHz.
VI. Lab Report
In the report, you need present the experiment results and compare them with the simulation
results. Comment on deviations from expected results, if any, and the reasons for these
deviations. Your report should include the following:
1. The design steps and results.
2. The simulation results.
3. The experiment results.
4. Compare the simulation results with the experiment results. Discuss any discrepancies, and
make comments.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Oxford University Press, 1982.
[2]. Dr. Robert Janes Martin, EEL 4140: Lab Manual for the Design of Analog Filters,
University of Central Florida, 1997.
EEL 4140
ANALOG FILTERS
LABORATORY 3
SallenKey Filters
I.
Objective
To study design and implementation of SallenKey filters.
II. Introduction
The term biquad is an edited form of the word biquadratic, which means the ratio of two
quadratic polynomials. Thomas C. Lee named this title several decades ago, and it is now in
common usage. All realizable polynomials used in analog filters can be factored to second
order forms and generally are of quadratic nature. Thus, the biquad is a useful and universal
building block. There are several circuits that can implement the biquad transfer functions.
SallenKey circuit is one of these circuits.
SallenKey Filters
Sallen and Key proposed a class of circuits, named as SallenKey filters, in 1955. The
basic SallenKey structure is shown in Fig. 1. This circuit incorporates a single amplifier
embedded in a passive RC network to generate any type of secondorder transfer functions:
lowpass, highpass, bandpass, and notch. In Fig.1, the box labeled secondorder passive RC
network contains resisters and two (or sometime three) capacitors.
Vo (s )
Vi (s )
is the input voltage,
is the output voltage of an amplifier having gain K , and V ( s ) is the amplifier input
voltage. If we apply superposition to the circuit, we obtain
V ( s ) = TFF ( s )Vi ( s ) + TFB ( s )Vo ( s )
(1)
where the feedforward gain of the passive network, TFF ( s ) , is defined as:
TFF ( s ) =
V ( s)
Vi ( s ) V ( s ) =0
o
and the feedback gain of the passive network, TFB ( s) , is defined as:
(2)
TFB ( s ) =
V (s)
Vo ( s ) V ( s )=0
(3)
Vi (s)
Secondorder
passive RC network
Vo (s )
V (s)
K
Fig. 1 The basic SallenKey topology, where Vi ( s) is the input voltage, Vo ( s) is the output voltage
of an amplifier having gain K , and V ( s ) is the amplifier input voltage.
Also, we have the amplifier relation as:
Vo ( s ) = KV ( s )
(4)
We obtain the relationship between the input voltage Vi (s ) and the output voltage Vo (s ) in
term of TFF (s ) and TFB (s ) as:
Vo ( s ) =
KTFF ( s )
Vi ( s )
1 KTFB ( s )
(5)
Unless the RC network is degenerate, TFF ( s ) and TFB ( s) have the same denominator
polynomial D(s) as:
D( s ) = s 2 + b1s + b0
(6)
Therefore, we can express TFF ( s ) and TFB ( s) as:
TFF ( s ) =
N FF ( s )
D( s )
(7)
and
TFB ( s ) =
N FB ( s )
D( s)
(8)
where N FF (s ) and N FB (s ) are polynomials with degree of at most two.
Using Equations (5), (7), and (8), we can derive the transfer function of the SallenKey filter
as:
V ( s)
T ( s) = o
Vi ( s )
(9)
KN FF ( s )
=
D( s ) KN FB ( s )
KN FB (s ) can modify the coefficients of the denominator polynomial of T (s ) .
It is
evident that the poles of T (s ) can be placed anywhere in the complex plane by appropriately
choosing K . According to the different K values, we can classify two kinds of SallenKey
filters. If the amplifier gain K > 0 , this kind of circuit is called as the positive SallenKey filter.
Otherwise, if the amplifier gain K < 0 , the circuit is referred to as the negative SallenKey filter.
LowPass Positive SallenKey Filter
C1
Vi (s )
R1
R2
Vo (s )
C2
Ra
Rb
Fig. 2 The lowpass positive SallenKey filter
The lowpass positive SallenKey filter is shown in Fig. 2.
From this configuration, we can
compute the feedforward and feedback gains of the passive network as:
TFF ( s ) =
N FF ( s )
D( s)
1
R1R2C1C2
=
1
R + R2
1
s2 +
+ 1
s +
R1R2C1C2
R2C2 R1R2C1
(10)
and
TFB ( s ) =
N FB ( s )
D( s )
1
s
R2C2
=
1
R + R2
1
s2 +
+ 1
s +
R1R2C1C2
R2C2 R1R2C1
(11)
Substituting Equations (10) and (11) into Equation (9), we can derive the transfer function
T (s) as:
K
R1 R2 C1C 2
T (s) =
1 K R1 + R2
1
s2 +
+
s +
R1 R2 C1C 2
R2 C 2 R1 R2 C1
(12)
where the amplifier gain K is given by:
K=
Ra + Rb
Ra
(13)
From Equation (12), we can obtain the cutoff frequency as:
0 =
and the quality factor as:
1
R1 R2 C1C 2
(14)
1
R1R2C1C2
Q=
1 K R1 + R2
+
R2C2 R1R2C1
=
(1 K )
(15)
1
R1C1
R RC
+ 1 + 2 1 2
R2C2
R1 R2C1
Design Procedure for LowPass Positive SallenKey Filters
1. There are five adjustable parameters in the circuit as shown in Fig. 2: the four passive
component values ( R1 , R2 , C1 , C 2 ) and the active parameter (the amplifier gain K =
Thus, we can arbitrary choose three values.
Ra + Rb
Ra
).
Each choice leads to a filter with different
properties. Here, we choose R1 = R2 = R and C1 = C 2 = C .
The rationale for this choice relies
on two principal aims: mathematical convenience and low element spread. The first needs no
elaboration. The second means that resistance and capacitance values should not be spread too
widely.
2. Based on the above choice, we can rewrite the expression for 0 and Q as:
1
RC
(16)
1
3 K
(17)
0 =
and
Q=
Hence, given values of 0 and Q , we can calculate K and RC .
3. We choose the proper capacitance value of C , and then compute the resistance value of R .
4. If we decide the value of Ra , we can easily obtain the value of Rb as:
Rb = ( K 1) Ra
(18)
Note that the DC gain is equal to K for the above design. If we want the DC gain to be unity,
for instance, a voltage divider can be used to replace R1 .
Although the element spread for the
equalR and equalC filter is excellent, it turns out that the quality factor is strongly sensitive to
variations in component values. Thus, it is of some interest to investigate an alternative design
in practice.
LowPass Negative SallenKey Filter
R3
Ro
Vi (s )
R1
R4
R2
Vo (s )
C1
C2
Fig. 3 The lowpass negative SallenKey filter
The lowpass negative SallenKey filter is shown in Fig. 3. We can use the same method to
derive the transfer function of the lowpass negative SallenKey filter as:
K
R1 R2C1C2
T ( s) =
1
1
1
1
1 1 + K
1
1
1
1
1
+
+
+
+
+
+
+
+
s2 +
s +
R3C1 R1C1 R2C1 R4C 2 R2C 2 R2 R3 R1 R2 R3 R4 R1 R4 R2 R4 C1C2
(19)
where the amplifier gain K is given by:
K=
Ro
R4
(20)
The cutoff frequency is given by:
1+ K
+
+
+
+
0 =
R2 R3 R1R2 R3 R4 R1R4 R2 R4 C1C2
and the quality factor is expressed as:
(21)
1+ K
1
1
1
1
+
+
+
+
R
R
R
R
R
R
R
R
R
R
1 2
3 4
1 4
2 4
2 3
Q=
1
1
1
1 C
1 C
+
+ 2 +
+ 1
R1 R2 R3 C1 R2 R4 C 2
(22)
Design Procedure for LowPass Negative SallenKey Filters
1. We choose the resistance and capacitance values as
C1 = C 2 = C .
R1 = R2 = R3 = R4 = R
and
Then, Equations (21) and (22) can be simplified as:
K +5
RC
0 =
(23)
and
Q=
K +5
5
(24)
2. Given the Q value, we obtain K as:
K = 25Q 2 5
(25)
3. We solve the following equation for the value of RC .
RC =
K +5
(26)
4. We choose the proper capacitance value of C , and then compute the resistance value of R .
5. We compute the resistance value of Ro as:
Ro = KR4
III. Design
Design example 1: the lowpass positive SallenKey filter
(27)
The filter specifications are given by 0 = 2 *1000 rad / s and Q = 5 .
Follow the above
procedure to choose the appropriate values of capacitors and resistors.
Design example 2: the lowpass negative SallenKey filter
The filter specifications are given by 0 = 2 * 4000 rad / s and Q = 1 .
Follow the above
procedure to choose the appropriate values of capacitors and resistors.
IV. Computer Simulations
1. Simulate the above two designed lowpass filters.
2. Plot the magnitude and phase responses of these two filters in the frequency range
10Hz30kHz.
V. Experiments
1. Build two designed filters, using LF 351 Op Amps with a split power supply voltage of
15V.
2. Use the Channel One of digital oscilloscope to show the input voltage waveform, and
channel Two to show the output voltage waveform. Record the input and output voltage
waveforms at the frequencies, 10Hz, 100Hz, 1kHz, and 10kHz.
Measure the magnitude and phase responses of two designed filters.
is from 10Hz to 30kHz.
3.
The frequency range
VI. Lab Report
In the report, you need present the experimental results and compare them with the expected
results. Discuss any discrepancies, make comments, and write conclusions. Your report should
include the following:
1. The complete circuit design processes and results.
2. The computer simulation results: the magnitude and phase responses of both filters.
3. The experiment results: the magnitude and phase responses of both filters and the recorded
graphs.
4. Summary and conclusions.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Oxford University Press, 1982, pp 171179.
[2]. Dr. Robert Janes Martin, EEL 4140: Lab Manual for the Design of Analog Filters,
University of Central Florida, 1997.
[3]. Artice M. Davis, SallenKey Filters, Chapter 6, in RC Active Filter Design Handbook,
Edited by F. W. Stephenson, John Wiley & Sons, Inc., 1985.
EEL 4140
ANALOG FILTERS
LABORATORY 4
StateVariable Biquads
I.
Objective
To study design and implementation of statevariable biquads.
II. Introduction
In the previous experiment, one realization of biquads, SallenKey filter, was discussed.
In the SallenKey filter, a single amplifier embedded in a passive RC network is used to
generate a secondorder transfer function. The structure of the SallenKey filter is relatively
simple. However, it is subject to great sensitivity to the constituent components values.
In this experiment, another biquad, the statevariable structure, will be studied. A
relatively high quality value can be achieved in this circuit. In addition, statevariable
biquads provide flexibility, good performance, and low sensitivity. The implementation of
the statevariable baud is based on the statevariable approach. Statevariable methods of
solving differential equations are employed in the development of the realization. In the
implementation of these realizations, three basic active building blocks are generally used:
the summer, the integrator, and the lossy integrator.
In this experiment, two kinds of statevariable biquads will be introduced: TowThomas
biquad and KerwinHuelsmanNewcomb (KHN) biquad.
TowThomas Biquad
The structure of the TowThomas biquad is shown as in Fig. 1. In this configuration, all
positive terminals of the Op Amps are grounded. The first basic building block composed
of the Op Amp U1 is a lossy integrator, while the second block is a summer amplifier, and
the third one is an integrator.
The path composed of R3 is feedback.
R4 , R5 , and R6
constitute feed forward paths to obtain the transmission zeros.
Analyzing the TowThomas biquad, we can derive the secondorder transfer function
from this realization as:
1
1 R6 R6
1
+
s 2 + s
R8
R1C1 R4C1 R7 R7 R3 R5C1C2
T (s) =
R8
1
R6
s2 + s
+
R1C1 R2 R3 R7 C1C2
From Equation (1), we can obtain the cutoff frequency as:
(1)
0 =
R8
R2 R3 R7 C1C2
(2)
R8C1
R2 R3 R7 C2
(3)
and the quality factor as:
Q = R1
Since the numerator is of general secondorder form, we can achieve any filter type by
choosing proper resistor values. For example, we can realize the secondorder bandpass
filter by choosing
R5 = R6 =
(4)
Where denotes the infinite value.
In practice, Equation (4) implies that R5 and R6 are not present in the circuit.
substitute Equation (4) into Equation (1).
Equation (1) is simplified as:
s
T (s) =
We
Therefore, the transfer function specified in
R8
R4 R7 C1
(5)
R8
1
s +s
+
R1C1 R2 R3 R7 C1C 2
2
In Equation (5), the voltage gain at the center frequency 0 is given as:
H BP =
R1 R8
R4 R7
(6)
R3
R1
C1
V1 ( s )
C2
R8
R4
R7
R2
U1
U2
+
Vo (s )
R6
R5
Fig.1 The feed forward TowThomas circuit
U3
Design Procedure for bandpass TowThomas filters
1. Given the filter specifications as , Q , and H BP .
2. Choose the resistance and capacitance values C and R such that
C1 = C2 = C
(7)
and
R3 = R7 = R8 = R
(8)
3. Define a positive constant such that
R2 = 2 R3
= R
2
(9)
4. Based on the above chosen parameter values, Equations (2), (3), and (6) can be simplified
as following:
0 =
1
RC
(10)
R
Q= 1
R
(11)
R
H BP = 1
R4
(12)
5. Compute the positive number as:
1
0 RC
(13)
6. Determine the resistance value of R2 as:
R2 = 2 R
(14)
7. Compute the resistances values of R1 and R4 as:
R1 = QR
and
(15)
R4 =
R1
H BP
(16)
KerwinHuelsmanNewcomb Biquad
The KerwinHuelsmanNewcomb (KHN) biquad is shown in Fig.2.
composed of a summer amplifier and two integrators.
paths.
This circuit is
R4 and R5 form the feedback
The three output terminals voltages, V2 ( s ) , V3 ( s ) , and V4 ( s ) , achieve highpass,
bandpass, and lowpass filter, respectively.
the same poles.
Moreover, these three transfer functions have
R5
C1
R6
C2
R2
R1
V1 ( s )
R3
U1
V2 ( s )
V4 ( s )
U2
R4
U3
V3 ( s )
Fig.2 The KerwinHuelsmanNewcomb (KHN) circuit
The highpass transfer function is obtained by the ratio of V2 ( s ) and V1 ( s ) as:
V2 ( s )
H HP s 2
=
V1 ( s ) s 2 + 0 s + 2
Q
0
(17)
Where H HP is given by:
R
1+ 6
R5
H HP =
R
1+ 3
R4
the cutoff frequency 0 is given by:
(18)
R6
R5
02 =
R1R2C1C2
(19)
and the quality factor Q is given by:
R
1+ 4
R3
Q=
R R R C
1 + 6 5 2 2
R5 R6 R1C1
(20)
From the relationship between V1 ( s ) and V3 ( s ) , the bandpass transfer function is
given as:
H BP 0 s
V3 ( s )
Q
=
V1 ( s ) s 2 + 0 s + 2
Q
0
(21)
where H BP is the voltage gain at the frequency 0 as:
R
H BP = 4
R3
(22)
The lowpass filter is achieved by the relationship between V1 ( s ) and V4 ( s ) as:
H LP 02
V4 ( s )
=
V1 ( s ) s 2 + 0 s + 2
Q
0
(23)
where H LP is the DC gain, expressed as:
R
1+ 5
R6
H LP =
R
1+ 3
R4
Design Procedure for bandpass KHN filters
1. Given the design specifications: 0 , Q , and H BP .
(24)
2. Choose the capacitance value C such that
C1 = C 2 = C
(25)
3. Let R2 , R3 , R5 , and R6 have the same resistance value as:
R2 = R3 = R5 = R6 = R
(26)
4. Define a positive constant number such that
R1 = 2 R2
= 2R
(27)
5. Equations (19), (20), and (22) are simplified as:
0 =
Q=
1
RC
(28)
R
1 + 4
2
R3
H BP = 2
(29)
(30)
6. Compute the positive constant number as:
2Q
H BP + 1
(31)
7. Compute the resistance value of R as:
R=
0C
(32)
8. Compute the resistance values of R1 and R4 as:
R1 = 2 R
(33)
and
Q
R4 = 2 1 R
(34)
III. Design
Design example 1: the bandpass TowThomas filter
Design a bandpass filter having the gain H BP = 3 , the quality factor Q = 5 , and the cutoff
frequency 0 = 2 *1000 rad / s , using the TowThomas circuit.
Choose the appropriate
values of capacitors and resistors, following the above procedure.
Design example 2: the bandpass KerwinHuelsmanNewcomb (KHN) filter
Design a band pass filter having the gain H BP = 3 , the quality factor Q = 10 , and the cutoff
frequency 0 = 2 *1000 rad / s , using the KHN circuit. Choose the appropriate values of
capacitors and resistors, following the above procedure.
IV.
Computer Simulations
1. Simulate the two designed bandpass filters.
2. Plot the magnitude and phase responses in the frequency range from 10Hz to 20 kHz.
3. Compare these two magnitude response plots, and understand the mechanism of the
quality factor Q .
V.
Experiments
1. Build above two designed biquad circuits, using LF 351 Opamps with a split power
supply voltage of 15V.
2. Use the Channel One of digital oscilloscope to show the input voltage waveform, and
channel Two to show the output voltage waveform. Record the input and output voltage
waveforms at the frequencies, 10Hz, 100Hz, 1kHz, and 10kHz.
3. Measure the magnitude response for two circuits.
20kHz.
The frequency range is from 10Hz to
VI. Lab Report
In the report, you need present the experiment results and compare them with the simulation
results. Discuss any discrepancies, make comments, and write conclusions. Your report
should include the following:
1. The complete circuit design processes and results.
2. The computer simulation results: the magnitude and phase responses for both circuits.
3. The experiment results: the magnitude responses for both filters and the recorded graphs.
Note that two bandpass filters have the same center frequency and the same voltage gain, but
the different quality factors. Compare the magnitude responses of these filters, and
understand the rule of the quality factor Q p .
4. Summary and conclusions.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Chapter 5, Oxford University Press,
1982.
[2]. Dr. Robert Janes Martin, EEL 4140: Lab Manual for the Design of Analog Filters,
University of Central Florida, 1997.
[3]. E. SanchezSinencio, Biquad I: The StateVariable Structure, Chapter 8, in RC Active
Filter Design Handbook, Edited by F. W. Stephenson, John Wiley & Sons, Inc., 1985.
EEL 4140
ANALOG FILTERS
LABORATORY 5
Single Op Amp BandPass Filters
I.
Objective
To design and implement single Op Amp bandpass biquad filters.
II. Introduction
Bandpass filters find wide use in modems, radio receivers, electrooptical systems, and
communications systems. The figure of merit for a bandpass filter is the quality factor which
is defined as the ratio of the center frequency to the 3dB bandwidth. The term quality factor
was first coined for bandpass filters in the first public communication systems. In early radio,
tuned radio frequency filters were used directly in front of the detector. These radios were very
noisy and had poor selectivity by todays standards. The sharper the transition from the pass
band to the stop band, the better the rejection of the adjacent channels. This led to better
separation of stations. A larger center frequency to bandwidth ratio was associated with higher
quality factor.
The circuit used in this experiment is the multiloop feedback filter. Its basic circuit
configuration is shown in Fig. 1. This structure is similar to SallenKey filters. There are,
however, two major differences. The first difference is that the active element is an Op Amp in
the multiloop feedback structure rather than a finitegain amplifier. The second difference is
that there are two feedback paths (rather than one) from the output of the amplifier to the RC
network. This is the reason for the name, multiloop feedback filter.
The multiloop feedback circuit can achieve the lowpass, highpass, or bandpass filter by
choosing the different RC networks. In this experiment, only the bandpass filter is discussed.
First, a basic bandpass structure without the positive feedback is introduced. Then, another
bandpass realization increases the quality factor by incorporating the positive feedback. The
design procedures for these two realizations are also given.
Vi (s )
Passive RC network
Vo (s)
Fig. 1 The basic multiloop feedback filter configuration
BandPass MultipleLoop Feedback Filters without the Positive Feedback
The bandpass multipleloop feedback filter without the positive feedback is shown in Fig. 2.
The transfer function is given by:
V (s)
T ( s) = o
Vi ( s )
s
R1C2
=
1
1
1
1
s +
+
s 2 +
+
C2 R2 C1R2 C1C2 R1R2 C1C2 R2 R3
Notice that this bandpass transfer function is an inverting one.
(1)
From Equation (1), we
obtain the quality factor Q and the cutoff frequency 0 as:
1+
Q=
R3
R1
R3C1
R3C 2
+
R2 C 2
R2 C1
(2)
and
R3
R1
R2 R3C1C 2
1+
0 =
(3)
In addition, we can get the midband gain as:
R2
R1
H BP =
C
1+ 2
C1
(4)
C2
Vi (s )
R2
C1
R1
Vo (s )
R3
Fig. 2 The bandpass multiloop feedback filter without positive feedback
Design Procedure for BandPass MultipleLoop Feedback Filters without the Positive
Feedback
1. Given the design specifications, 0 , Q , and H BP .
2. In the bandpass multiloop feedback filter, it is convenient to use the equalvalue capacitor
design.
Choose a suitable capacitance value, C , for the capacitors C1 and C 2 , as:
C1 = C 2 = C
(5)
3. Determine the resistance value of R1 as:
R1 =
Q
0CH BP
(6)
4. Determine the resistance value of R2 as:
R2 =
2Q
0C
(7)
5. Determine the resistance value of R3 as:
R3 =
Q
0 C 2Q 2 H BP
(8)
Note that in Equation (8), H BP must be less than 2Q 2 in order that R2 be finite and
positive.
BandPass MultipleLoop Feedback Filters with Positive Feedback
A modification of the bandpass filter shown in Fig.2 can be used to reduce the spread of
element values for high Q realizations. The modified circuit is shown in Fig. 3.
The positive
feedback is provided by the voltage divider consisting of two resistors Ra and Rb .
To simplify
equations describing the circuit, the effect of these two resistors is presented by defining a
constant K as:
K=
Ra
Rb
C2
Vi (s )
(9)
R2
C1
R1
Vo (s )
+
Ra
Rb
Fig. 3 The bandpass multiloop feedback filter with positive feedback
The transfer function for this circuit is given by:
V (s)
T ( s) = o
Vi ( s )
s (K + 1)
R1C2
=
1
K
1
1
+
s 2 + s
+
R2C2 R2C1 R1C2 R1R2C1C2
From Equation (10), we obtain the quality factor Q and the cutoff frequency 0 as:
(10)
1
=
Q
R1 C1
C2
+
K
R2 C2
C1
R2
R1
C2
C1
(11)
and
0 =
1
R1 R2 C1C 2
(12)
The midband gain is given as:
(K + 1)
H BP =
R1C1
1
1
K
+
R2C 2 R2 C1 R1C1
(13)
From Equation (11), it is clear that the quality factor is improved by introducing positive
feedback K . The design procedure for this filter is based on the solution of these equations, and
is summarized in following.
Design Procedure for BandPass MultipleLoop Feedback Filters with the Positive
Feedback
1. Given the design specifications 0 and Q .
2. As before, it is convenient to use an equalvalue capacitor design, in which the capacitors
C1 and C 2 have the same value.
Choose a suitable capacitance value, C , for the capacitors
as:
C1 = C 2 = C
(14)
3. Determine the ratio parameter m0 that would be required if there are no positive feedbacks
as:
m0 =
1
4Q 2
4. Select the desired resistor ratio m , which is less than one, and greater than m0 .
(15)
R
m= 1
R2
(16)
Use m to determine the amount of positive feedback K as:
K = 2m
m
Q
(17)
5. Choose a convenient value for Rb , and calculate Ra as:
Ra = KRb
(18)
6. Determine the resistance value of R2 as:
R2 =
0C m
(19)
7. Determine the value of R1 using the relation
R1 = mR2
(20)
8. The value of H BP , the gain at resonance, is determined by the relation.
H BP =
K +1
2m K
(21)
III. Design
Design example 1: the secondorder bandpass multiloop filter without positive feedback
Design a secondorder bandpass filter having the gain H BP = 2 , the quality factor Q = 2 , and
the center frequency 0 = 10 4 rad / s , using the multiloop structure without positive feedback.
Choose the appropriate values of capacitors and resistors, following the above procedure.
Design example 2: the secondorder bandpass multiloop filter with positive feedback
Design a secondorder bandpass filter having the quality factor Q = 10 and the center
frequency 0 = 10 4 rad / s , using the multiloop structure with positive feedback.
Choose the
appropriate values of capacitors and resistors, following the above procedure.
IV. Computer Simulations
1. Simulate the above two bandpass filters with the designed resistance and capacity values.
2. Plot the magnitude and phase responses, for the frequency range 10Hz30kHz.
V. Experiments
1. Build above two designed bandpass filters, using LF 351 Opamps with a split power supply
voltage of 15V.
2. Use the Channel One of digital oscilloscope to show the input voltage waveform, and
channel Two to show the output voltage waveform. Record the input and output voltage
waveforms at the frequencies, 10Hz, 100Hz, 1kHz, and 10kHz.
3. Measure the magnitude responses of two filters.
30kHz.
The frequency range is from 10Hz to
VI. Lab Report
In the report, you need present the experiment results and compare them with the simulation
results. Discuss any discrepancies, make comments, and write conclusions. Your report
should include the following:
1. The complete circuit design processes.
2. The Computer simulation results: the magnitude and phase responses for both circuits.
3. The Experiment results: the magnitude responses for both filters and the recorded graphs.
4. Summary and conclusions.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Oxford University Press, 1982.
[2]. Dr. Robert Janes Martin, EEL 4140: Lab Manual for the Design of Analog Filters,
University of Central Florida, 1997.
[3]. L. P. Huselsman, MultipleLoop Feedback Filters, Chapter 7, in RC Active Filter Design
Handbook, Edited by F. W. Stephenson, John Wiley & Sons, Inc., 1985.
EEL 4140
ANALOG FILTERS
LABORATORY 6
Two Op Amps Current Generalized Immittance
Structure (CGIC) Based Biquad
I.
Objective
To study CGIC biquads, to use this structure to design secondorder lowpass and bandpass
filters with given specifications, and to functionally tune the above circuits to get the specified
values of the cutoff frequency and the quality factor.
II. Introduction
The general CGIC structure is shown in Fig. 1.
The transfer functions between the input
and output terminals, V2 ( s ) , V3 ( s ) , and V4 ( s ) , assuming ideal Op Amps, are given below:
V ( s)
T1 ( s ) = 3
Vi ( s )
Y YY
Y5 + h( s ) Y7 1 + 6 5 8
Y2 Y2
=
D( s)
(1)
V ( s)
T2 ( s ) = 4
V1 ( s )
Y YY
Y5 1 + 8 6 7 + h( s )Y7
Y4 Y4
=
D( s)
(2)
and
V ( s)
T3 ( s ) = 2
Vi ( s )
Y + h( s )Y7
= 5
D( s)
(3)
where
YY
h( s ) = 2 3
Y1Y4
(4)
D ( s ) = (Y5 + Y6 ) + h( s )(Y7 + Y8 )
(5)
The most commonly used secondorder transfer functions can be easily generated from the
above equations, as summarized in Table 1. By using minimum sensitivity constraints in circuit
1, 3, 7, 10, and 12, possible sets of element values have been obtained, as shown in Table 2. In
this experiment, two kinds of filters will be discussed in detail: a lowpass filter using the
number 1 circuit, and a bandpass filter using the number 7 circuit.
Y1
Y3
V3 ( s )
Y2
Y4
V4 ( s)
Y7
Y5
V2 ( s )
V1 ( s )
Y6
Vi (s )
Y8
Fig. 1 The basic CGIC configuration
Design Procedure for LowPass CGIC Filters
1. Given the design specifications, 0 and Q .
2. Circuit 1 in Table 2 realizes a lowpass filter. The resistor and capacitor are chosen as
follows:
G1 = G4 = G5 = G8 = G
G3 =
G
Q
(7)
(8)
and
C 2 = C3 = C
(9)
where G is the conductance define as G =
1
.
R
3. Then, the transfer function is simplified as:
2
T2 ( s ) =
2 2
R C
1
s
s2 +
+
2
RCQ R C 2
(10)
From the definition of 0 , we obtain
0 =
4. Choose an appropriate capacitor value C .
1
RC
(11)
Then, compute the resistor value R as:
R=
1
C o
(12)
R3 = RQ
(13)
5. Consequently,
C3
R3
R1
C2
R4
+
Vi (s )
R5
Vo (s )
R8
Fig. 2 The secondorder lowpass CGIC filter
Design Procedure for BandPass CGIC Filters
1. Given the design specifications 0 and Q .
2. Circuit 7 in Table 2 realizes a bandpass filter. The resistance and capacitance values are
chosen as follows:
C 3 = C8 = C
(14)
G1 = G2 = G4 = G6 = G
(15)
G
Q
(16)
G8 = 0
(17)
G7 =
and
G is the conductance define as G = 1 .
R
3. The transfer function is simplified as:
2
RCQ
T1 ( s ) =
1
s
+
s2 +
RCQ R 2 C 2
s
4. Choose an appropriate capacitor value C .
(18)
Then, compute the resistance value of R as:
R=
1
C o
(19)
R7 = RQ
(20)
5. Consequently,
C3
R1
Vo ( s )
R4
R2
+
R6
+
Vi ( s )
R7
C8
Fig. 3 The secondorder bandpass CGIC filter
III. Design
Design example 1: the secondorder Butterworth lowpass CGIC filter
Design a secondorder lowpass filter with a quality factor Q = 0.707 , and a cutoff frequency
0 = 2 * 2 *10 3 rad / s .
Choose the capacitance value as C = 0.01uF .
Compute the
appropriate values resistors, following the above procedure.
Design example 2: the secondorder bandpass CGIC filter
Design a secondorder bandpass filter having a quality factor Q = 10 , and a cutoff frequency
0 = 2 * 3 *103 rad / s .
Choose the capacitor value as C = 0.01uF .
Compute the
appropriate values resistors, following the above procedure.
IV. Computer Simulations
1. Simulate both of the above design filter examples with the calculated resistance values.
2. Plot the magnitude and phase responses in the frequency range from 50Hz to 20kHz.
V. Experiments
1. Build above the CGIC filters designed in part III, using LF 351 Op Amps with a split power
supply voltage of 15V.
2. The circuits need to be functionally tuned to yield the specified values of the quality factor
Q and the cutoff frequency 0 .
A sinusoidal input is used during the tuning processes.
In the case of the lowpass filter, the circuit is first tuned for 0 .
Applying a sinusoidal input of
the frequency 0 , then adjusting R8 and monitoring the phase angle difference between the
output and input voltages.
The circuit is tuned for 0 when the output voltage lags the input
voltage by 90 degrees.
To tune the circuit to attain the specified Q , the gain H LP (the ratio of the output voltage
between the input voltage) at a low frequency, much lower than 0 (usually a few Hz but not
to 0Hz), is determined.
Then, a sinusoidal input at the frequency 0 is applied. R3 is
adjusted until the gain of this filter become Q times H LP .
In case of the bandpass filter, the desired value of 0 is realized by adjusting R2 until the
phase angle between the output and the input voltage equals 0 degrees at the sinusoidal input of
the frequency 0 .
Q can be attained by adjusting R7 until the output voltage advances the
input voltage by 45 degrees when the frequency of the sinusoidal input is the lower cutoff
frequency 1 .
3. Use the Channel One of digital oscilloscope to show the input voltage waveform, and
channel Two to show the output voltage waveform. Record the input and output voltage
waveforms at the frequencies, 10Hz, 100Hz, 1kHz, and 10kHz.
4. Measure the magnitude response for each of two circuits after functional tuning is complete.
The frequency range is from 10Hz to 30kHz.
VI. Lab Report
In the report, present the experimental results and compare them with the simulation results.
Comment on the deviations from the expected results, and the reasons for these deviations.
Your report should include the following:
1. Derive the transfer function for the above two circuits in the form supplied in the table from
the general form, using the specified element values.
2. Compute the quality factor Q and the cutoff frequency 0 from the measured resistance
values after functional tuning is complete.
Compare Q and 0 with the desirable
specifications.
3. The computer simulation results: the magnitude and phase responses for both circuits.
4. The experiment results: the magnitude responses for both filters and the recorded graphs.
5. Summary and conclusions.
References
[1]. Wasfy B. Mikhael, 2 OA Current Generalized Immittance Structure (CGIC) Based
Biquad, Lab Manual, University of Central Florida.
[2]. Wasfy B. Mikhael, Biquad II: The Current Generalized Immittance (CGIC) Structure,
Chapter 9, in RC Active Filter Design Handbook, Edited by F. W. Stephenson, John Wiley &
Sons, Inc., 1985.
[3]. Wasfy B. Mikhael, The Current Generalized Immittance (CGIC Biquad), Chapter 82, in
Circuits and Filter Design Handbook, Edited by F. W. Stephenson, CRC Press, 2003, pp
24952514.
Table 1 Element identification for realizing the most commonly transfer functions
Circuit
Number
Y1
Y2
Y3
Y4
Y5
Y6
Y7
Y8
G1
sC2
sC3+G3
G4
G5
G8
sC1
G2
G3
sC4+G4
G6
G7
sC8+G8
G1
G2
sC3
G4
G6
sC7
sC8+G8
G1
G2
sC3
G4
G6
sC 7
1 + sC 7 R7
0
0
Transfer Function
T2 ( s ) =
G1G5 (G4 + G8 )
s 2 C 2 C3 G8 + sC 2 G3G8 + G1G4 G5
G
G2 G3 G7 1 + 6
G2
T1 ( s ) = 2
s C1C 4 G6 + s (C 2 G4 G6 + C8 G2 G3 ) + G2 G3 (G7 + G8 )
T1 ( s ) =
T1 ( s ) =
s 2 C 3C 7 (G2 + G6 )
s 2 (C 7 + C8 )G2 C3 + sC3G2 G8 + G6 G1G4
s 2 C3C 7 (G2 + G6 )
s C3C 7 G2 + sC 7 G1G4 G6 R7 + G6 G1G4
Remarks
Lowpass
Lowpass
Highpass
Highpass
G1
sC2
sC3
G4
sC5
G6
G8
G
sC5 G1G4 1 + 8
G4
T2 ( s ) = 2
s C 2 C3 G8 + sC5 G1G4 + G1G4 G6
Band pass
sC1
G2
G3
sC4+G4
G6
sC7
G8
G
sC 7 G2 G3 1 + 6
G2
T1 ( s ) = 2
s C1C 4 G6 + s(C 7 G2 G3 + C1G4 G8 ) + G2 G3G8
Bandpass
G1
G2
sC3
G4
G6
G7
sC8+G8
sC1
G2
G3
sC4
G5
G7
sC8
T1 ( s ) =
sC3G7 (G2 + G6 )
s 2 C3C8 G2 + sC3G2 (G7 + G8 ) + G1G4 G6
T2 ( s ) =
s 2 C1G5 (C 4 + C8 ) + G2 G3G7
s 2 C1C 4 G5 + sC8 G2 G3 + G2 G3G7
Bandpass
Notch
G1
G2
sC3
G4
G5
sC7
G8
10
G1
G2
sC3
G4
G5
G6
sC7
sC8+G8
11
sC1
G2
G3
sC4
G5
G6
G7
sC8
12
G1
G2
sC3
G4
G5
G6
sC7
G8
T2 ( s ) =
s 2 C3C 7 G2 + G1G5 (G4 + G8 )
s C 3C 7 G2 + sC3 G2 G8 + G1G4 G5
Notch
2 G1G4G5 C7
s +
C3C7 G2 C7 + C8
T3 ( s ) =
G8
G G (G + G6 )
s2 + s
+ 1 4 5
C7 + C8 C3G2 (C7 + C8 )
T1 ( s ) =
s 2 C1G4 G5 sG5 C8 G3 + G3G7 (G2 + G6 )
s 2 C1C 4 (G5 + G6 ) + sC8 G2 G3 + G2 G3 G7
T1 ( s ) =
s 2 C 3C 7 (G2 + G6 ) sC3G5 G8 + G1G4 G5
s 2 C3C 7 G5 + sC3 G2 G8 + (G5 + G6 )G1G4
Notch
For allpass:
G6 = 0
G5 = G 2
For allpass:
G6 = 0
G5 = G 2
Table 2 Design values and tuning procedure
Circuit Number
Design values
Transfer function realized
(from Table 1)
1
G1 = G4 = G5 = G8 = G , G3 = G Q , C 2 = C3 = C , where C = G 0
G1 = G2 = G4 = G6 = G , G8 = G Q , C8 = 0 , C3 = C7 = C , where
C = G 0
G1 = G2 = G4 = G6 = G , G7 = G Q , G8 = 0 , C3 = C8 = C , where
C = G 0
10
G1 = G2 = G4 = G5 + G6 = G , G8 = G Q , C3 = C7 + C8 = C , where
0 = G C and n2 = 0G5 C 7
12
Tuning sequence
G1 = G2 = G4 = G5 = G , G6 = 0 , C3 = C7 = C , G8 = G Q , where
C = G 0
Note: D( s ) = s 2 + ( 0 Q )s + 02
T2 ( s ) =
2 02
D( s )
G8
G3
T1 ( s ) =
2s 2
D( s)
G4
G8
G2
G7
G2
G6
G8
G4
G4
G8
T1 ( s ) =
(2 0 Q )s
D( s)
C7 s 2 + n2
T3 ( s ) =
C D( s)
T1 ( s ) =
D( s )
D( s)
Study Guide C
EEL 4140
ANALOG FILTERS DESIGN
Biquads II: The current Generalized Immittance (CGIC) Structure
EEL 4140
ANALOG FILTERS
LABORATORY 7
HighOrder LowPass Filter Design
I.
Objective
To study the cascade design method for realizing highorder filters. Current Generalized
Immittance Structure (CGIC) Biquads are employed to build highorder lowpass filters.
II. Introduction
The previous experiments have discussed the design of firstorder and secondorder filters.
In this experiment, one design method for the highorder filter is introduced: the cascade design
method. This method is the most commonly used method because of its simplicity. It is based
on cascading firstorder and secondorder filter sections in such a manner that each section does
not interact with others.
For the cascade design method, the effect of component variations on the overall frequency
response is large. Thus, the cascade design method may be tolerated for the applications where
only a few filters are constructed and manual tuning is used, or where filter specifications are not
particularly tight. However, when filters are produced in large quantities and are subjected to
stringent specifications, the cascade design method is not recommended.
The cascade design method usually consists of three steps: decomposing transfer functions
into poles and zeros, grouping poles and zeros into firstorder or secondorder filter sections, and
synthesizing these filter sections.
Step 1: Decomposing transfer functions
In general, any form of transfer functions can be expressed in term of poles and zeros.
Poles and zeros can be either real numbers or complex numbers. Complex zeros and poles,
however, must occur in conjugate pairs.
Step 2: Grouping poles and zeros into firstorder or secondorder filter sections
In determining the order of sections and pairing of poles and zeros, it is important to
consider the occurrence of internal resonances. This refers to the phenomenon of the
largevoltage buildup at certain frequencies at internal nodes of filters (usually at the output
terminal of an Op Amp). Thus, the gain to one of these nodes from the input of the filter may
be higher than the gain to the output of the filter. Consequently, clipping may occur at output of
an internal Op Amp before the filter output voltage shows clipping. This internal clipping will
show up as a leveldependent change in the overall frequency characteristic, and will restrict the
range of input signal levels that the filter can handle.
In order to achieve maximum dynamic range, the filter should be designed such that the
clipping will first occur at the output of the last Op Amp of the filter. The following rules of
thumb are useful in maximizing the dynamic range of cascaded design filters.
Rule 1: Place the sections in the order of increasing Q values. (The Q value of the firstorder
section is assumed to be zero.)
Rule 2: Group the pole and the zero, which are closest in the S plate.
Rule 3: Distribute the overall gain equally among the sections.
Step 3: Synthesizing these firstorder and secondorder filter sections
For each filter section decided in step 2, we choose the appropriate circuit to synthesize it.
When we cascade these sections together, we must consider effects of the input and output
impedance. If the output impedance of the last section is not zero, and the input impedance of
the next section can not be assumed as infinite, a buffer is needed to insert between these two
sections. Otherwise, we can connect these sections directly.
In this experiment, the lowpass CGIC Biquad, which was studied in the last experiment, is
used as the buildup circuit block for the highorder lowpass filter. The circuit of the CGIC
lowpass Biquad is shown in Fig. 1. In this Biquad, the resistor and capacitor values are chosen
as:
R1 = R4 = R5 = R8 = R
(1)
and
C 2 = C3 = C
(2)
Based on these resistance and capacitance values in Equation (1) and (2), the transfer function is
given by:
2
R C2
TCGIC ( s ) =
s
1
s2 +
+ 2 2
R3C R C
2
From Equation (3), we can obtain the cutoff frequency 0 and the quality factor Q as:
(3)
0 =
1
RC
(4)
Q=
R3
R
(5)
and
When we use this circuit to realize a secondorder transfer function, we first choose the
capacitor value C . Then, we can compute the resistor value R using Equation (4). Finally,
we can compute the resistance value of R3 using Equation (5).
C3
R3
R1
C2
R4
+
Vi (s )
R5
Vo (s )
R8
Fig. 1 The secondorder lowpass CGIC Biquad
Design Procedure for the HighOrder LowPass Filter
1. Design a sixorder Chebyshev lowpass filter, which has a maximum passband attenuation
of 1.0dB and the bandwidth 3979 Hz. That is
0 = 25,000rad / s
(6)
2. The normalized sixorder Chebyshev lowpass transfer function with a maximum passband
attenuation of 1.0dB is given as:
Tnorm ( s) =
0.061415
(7)
s + 0.12436s + 0.99073 s + 0.33976s + 0.55772 s 2 + 0.46413s + 0.12471
)(
)(
3. From (7), it can be seen that this highorder filter consists of three secondorder sections.
Tnorm ( s ) is decomposed into three secondorder transfer functions as:
2 * 0.99073
s + 0.12436 s + 0.99073
Tnorm1(s) =
2 * 0.55772
s + 0.33976 s + 0.55772
Tnorm 2 ( s ) =
(8)
(9)
and
Tnorm 3 ( s ) =
2 * 0.12471
s + 0.46413s + 0.12471
To denormalize these three transfer functions, we replace s by s =
(10)
s
in Equations (8), (9),
and (10). Consequently, we obtain the following denormalized transfer functions as:
1.24 *10 9
T1(s) = 2
s + 3.11*10 3 s + 6.192 *10 8
(11)
6.97 *10 9
s 2 + 8.49 *10 3 s + 3.49 *10 8
(12)
1.56 *10 8
T3 ( s ) = 2
s + 1.16 *10 4 s + 7.79 *10 7
(13)
T2 ( s ) =
and
For the denormalized transfer function T1 ( s ) , the cutoff frequency 0,1 and the quality
factor Q1 are given as follows:
0,1 = 2.48 * 10 4 rad / s
(14)
and
Q1 = 8.00
(15)
In the same way, the cutoff frequency 0,2 and the quality factor Q2 corresponding to the
transfer function T2 ( s ) are given by:
0, 2 = 1.87 *10 4 rad / s
(16)
and
Q2 = 2.20
(17)
For the transfer function T3 ( s ) , the cutoff frequency 0,3 and the quality factor Q3 are given
by:
0,3 = 8.83 * 103 rad / s
(18)
and
Q3 = 0.76
(19)
Q1 > Q2 > Q3
(20)
4. Notice that
According to Rule 1, we arrange these sections in the increasing order of the quality factor.
first section is used to synthesize T3 ( s ) .
The
T2 ( s ) and T1 ( s ) are synthesized in the second
section and the third section, respectively.
5. Synthesize T1 ( s ) , T2 ( s ) , and T3 ( s ) using the CGIC lowpass Biquad as Fig.1.
For all
three sections, choose the capacitor value as C = 0.01uF .
For the first section,
1
= 4.02k
0.01 * 24.883
(21)
R3 = R * Q1 = 4.02k * 8 = 32.16k
(22)
R=
C 0,1
and
For the second section,
R=
C 0, 2
1
= 5.36k
0.01 * 18.67
(23)
and
R3 = Q2 R = 5.36k * 2.2 = 11.79k
(24)
Similarly, the following resistance values are chosen for the third section.
R=
C 0,3
1
= 11.33k
0.01 * 8.829
(25)
and
R3 = Q p,1 * R = 11.33k * 0.76 = 8.61k
(26)
6. Cascade these three filter sections into a sixorder Chebyshev lowpass filter. In this
application, the output of each section is the Op Amp output. Thus, the output resistance is
assumed to be zero. Consequently, there is no need to insert buffer here. The final schematic
circuit of this highorder filter is shown in Fig. 2.
III. Design
Design example: the sixorder Butterworth lowpass filter.
The normalized transfer function is given as:
T (s) =
(s 2 + 0.51764s + 1)(s 2 + 1.41421s + 1)(s 2 + 1.93186s + 1)
(27)
The cutoff frequency of this filter is given as f 0 = 2kHz.
Following the above design procedure, use the cascade method to design this filter, and the
lowpass CGIC Biquad as the buildup circuit block. In these Biquads, choose the capacitor
value C = 0.01uF , and compute the appropriate resistor values.
IV. Computer Simulations
1. Simulate the above design filter with the calculated resistance values.
2. Plot the magnitude and phase responses, for the frequency range 10Hz10kHz.
V. Experiments
1. Build the sixorder Butterworth filter designed in part III, using LF 351 Op Amps with a split
power supply voltage of 15V.
2. Use the Channel One of digital oscilloscope to show the input voltage waveform, and
channel Two to show the output voltage waveform. Record the input and output voltage
waveforms at the frequencies, 10Hz, 100Hz, 1kHz, and 10kHz.
3. Measure the magnitude response of this highorder filter. The frequency range is from
10Hz to 10kHz.
VI. Lab Report
In the report, present the experiment results and compare them with the simulation results.
Comment on deviations from expected results, if any, and the reasons for these deviations.
Your report should include the following:
1. The complete circuit design processes.
2. The computer simulation results: the magnitude and phase responses for this highorder filter.
3. The experiment results: the magnitude responses for the highorder filter and the recorded
graphs.
4. Summary and conclusions.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Oxford University Press, 1982.
[2]. Wasfy B. Mikhael, Biquad II: The Current Generalized Immittance (CGIC) Stucture,
Chapter 9, in RC Active Filter Design Handbook, Edited by F. W. Stephenson, John Wiley &
Sons, Inc., 1985.
[3]. Bert D. Nelin, Design of HighOrder Active Filters, Chapter 10, in RC Active Filter
Design Handbook, Edited by F. W. Stephenson, John Wiley & Sons, Inc., 1985.
[4]. Wasfy B. Mikhael, The Current Generalized Immittance (CGIC Biquad), Chapter 82, in
Circuits and Filter Design Handbook, Edited by F. W. Stephenson, CRC Press, 2003, pp
24952514.
0.01uF
11.33k
0.01uF
+
Vi (s )
8.61k
11.33k
0.01uF
5.36k
0.01uF
11.79k
5.36k
0.01uF
11.33k
32.16k
4.02k
0.01uF
5.36k
11.33k
4.02k
4.02k
5.36k
Fig.2 The sixorder Chebyshev lowpass filter
Vo (s)
4.02k
Study Guide C
EEL 4140
ANALOG FILTERS DESIGN
Biquads II: The current Generalized Immittance (CGIC) Structure
EEL 4140
ANALOG FILTERS
LABORATORY 8
Butterworth Filter Approximation
I.
Objective
To study Butterworth approximations of lowpass, bandpass, and highpass filters.
II. Introduction
The magnitude response of the lowpass Butterworth filter is expressed as:
T ( j ) =
1
1 + ( 0 )
2n
(1)
Where n is the filter order and 0 is the cutoff frequency.
From Equation (1), poles of lowpass Butterworth filters can be derived.
Poles of LowPass Butterworth Filters
From Equation (1), the following equation can be derived as:
T (s )T ( s ) =
1
2n
1 + (s j 0 )
(2)
n s
1 + ( 1)
0
2n
The poles of equation (2) are the roots of its denominator.
s
1 + ( 1)
0
2n
=0
(3)
The poles of Equation (2) are given by:
j 2k
e 2n ,
n is odd
0
,
s=
j (2 k +1)
2n
, n is even
0 e
k = 0,1,...,2n 1
(4)
As well known, poles in the right halfplane correspond to an unstable system.
in the left halfplane are selected to associate with T (s ) .
j 2k
e 2n ,
0
s=
j (2k +1)
0 e 2n ,
n is odd, k =
The poles of T (s ) is given by:
n +1 n + 3
n + (2n 1)
,
,...,
2
2
2
n is even, k =
Thus, poles
(5)
n n
n
, + 1,..., + (n  1)
2 2
2
From Equation (5), it is can be seen that the poles of the lowpass Butterworth filter are
located on the circle with the radius 0 , and are separated by n = n .
pole on the real axis. If n is even, there are poles at = n 2 .
If n is odd, there is a
As examples, the pole
locations of the 4th and 5th order lowpass Butterworth filters are shown in Fig. 1 and Fig.2,
respectively.
j
n=4
j 0
n = 45o
n 2
j 0
Fig. 1 The pole locations of the 4th order lowpass Butterworth filter
n=5
j 0
n = 36 o
n
j 0
Fig. 2 The pole locations of the 5th order lowpass Butterworth filter
LowPass, HighPass, and BandPass Filter Specifications
Filter specifications are usually given in term of attenuation characteristics.
The
attenuation ( ) is defined as:
( ) = 20 log T ( j ) dB
(6)
The specifications for lowpass, highpass, and bandpass filters are shown in Fig. 3, Fig. 4,
and Fig. 5, respectively.
1. Lowpass filter specifications: the attenuation at the passband (from = 0 to = p )
should be smaller the max ; the attenuation at the stopband (from = s to = ) should
be larger than the min .
2. Highpass filter specifications: the attenuation at the stopband (from = 0 to = s )
should be larger than the min ; the attenuation at the passband (from = p to = )
should be smaller than the max .
3. Bandpass filter specifications: the attenuation at two stopbands (from = 0 to
= s,1 , and from = s,2 to = ) should be larger than min ; the attenuation at the
passband (from = p,1 to = p,2 ) should be smaller than max .
min
passband
stopband
max
p
Fig.3 Lowpass filter specifications
min
passband
stopband
max
s
Fig.4 Highpass filter specifications
min
min
passband
stopband
stopband
max
s ,1 p ,1
p,2
s,2
Fig.5 Bandpass filter specifications
Design Procedure for LowPass Butterworth Filter Approximation
1.
The
lowpass
filter
specifications
are
by min = 20dB , max = 0.5dB ,
given
p = 1000rad / s , and s = 2000rad / s .
2.
Decide the filter order n as:
n=
Round
3.
[(
log 10 min 10 1 10 max 10 1
= 4.8321
2 log( s p )
(7)
n=5
(8)
up the next integer.
Compute the cutoff frequency 0 .
0,1 =
First, calculate the stopband cutoff frequency as:
[10
s
max
1
10
1 2n
= 1263.2rad / s
(9)
Then, calculate the passband cutoff frequency as:
0,2 =
[10
min
1
10
1 2n
= 1234.1rad / s
(10)
The actual cutoff frequency is the geometric average of the passband and stopband cutoff
frequencies as:
0 = 0,1 0,2 = 1248.6rad / s
(11)
Using this way, the excess attenuations are achieved at the frequency s and p .
4.
Obtain the normalized lowpass Butterworth transfer function, according to the filter order n .
T (s) =
(s + 1) s 2 +
s
s
+ 1 s 2 +
+ 1
0.62
1.62
5.
(12)
Denormalize the transfer function by substituting s in Equation (12) with s 0 .
The
transfer function satisfying the given specifications is given by:
T (s) =
1
s 2
s
s 2
s
s
+ 1
+
+ 1
+
+ 1
2
2 1.62 0
0
0 0.62 0
0
(13)
5.6566 * 1015
(s + 1248.6)(s 2 + 2013.8s + 1.56 * 106 )(s 2 + 770.7 s + 1.56 * 106 )
Design Procedure for HighPass Butterworth Filter Approximation
1.
The
highpass
filter
specifications
are
given
by min = 25dB , max = 1dB ,
p = 43.75rad / s , and s = 12.50rad / s .
2. Change the highpass specifications to the corresponding lowpass specifications.
and stop frequencies of the corresponding lowpass filter are given by:
p _ low =
The pass
= 0.0229rad/s
(14)
= 0.08rad / s
(15)
and
s _ low =
3.
Design the corresponding lowpass Butterworth filter.
Its specifications are as follows: the
attenuation should be at most max = 1dB at the frequency p _ low = 0.0229rad / s , and the
attenuation should be at least min = 25dB at the frequency s _ low = 0.08rad / s .
Then, the order of the corresponding lowpass filter is n = 3 , and the cutoff frequency is
0 = 0.0296rad/s .
The denomalized corresponding lowpass filter transfer function is given
by
T ( s ) low _ mod el =
1
2
s
s
s
+ 1
+
+ 1
2
0.0296
0.0296 0.0296
(16)
4.
Change the corresponding lowpass transfer function into the highpass transfer function by
1
substituting s in Equation (16) with .
s
T (s) =
1
1
1
1
+
+ 1
+ 1
0.0296 * s 0.0296 2 * s 2 0.0296 * s
(17)
s3
(s + 33.78)(s 2 + 33.78s + 1141.3)
Design Procedure for BandPass Butterworth Filter Approximation
1.
The bandpass filter specifications: min = 25dB , max = 0.5dB , p,1 = 500rad / s ,
p,2 = 1000rad / s , s,1 = 250rad / s , and s,2 = 2000rad / s . The center frequency c is
given by:
c = p,1 p,2
= s,1 s,2
= 707.1rad/s
2. Change the bandpass specifications to the corresponding lowpass specifications.
pass and stop frequencies of the corresponding lowpass filter are given by:
p _ low = p,2 p,1 = 500rad / s
and
(18)
The
(19)
s _ low = s,2 s,1 = 1750rad / s
3.
Design the corresponding lowpass Butterworth filter.
(20)
Its specifications are as follows: the
attenuation must be at most max = 0.5dB at frequency p _ low = 500rad / s , and the
attenuation must be at least min = 25dB at the frequency s _ low = 1750rad / s
The order of the corresponding lowpass filter is n = 4
is 0 = 744.6rad/s .
and the cutoff frequency
The denomalized corresponding lowpass filter transfer function is given
by:
Tlow _ mod el =
4.
1
s2
s 2
s
s
+
+ 1
+
+ 1
744.6 2 744.6 * 0.54 744.6 2 744.6 * 1.31
(21)
Change the corresponding lowpass filter model into the bandpass filter by substituting s
in Equation (21) with
s 2 + c2
.
s
T (s) =
1
2
s + 707.1
2 2
s 2 + 707.12
744.6 2 +
(744.6* 0.54) + 1
1
2
2
2
2
s + 707.12
2 s + 707.1
+
744
6
.
s
s
(22)
(744.6*1.31) + 1
III. Design
1. Find the poles of the normalized 9th and 10th order lowpass Butterworth transfer function
( 0 = 1rad / s ).
2.
Compare your results with the poles given in the textbook.
Design a lowpass Butterworth transfer function.
The specifications are given
by min = 20dB , max = 1dB , p = 1000rad / s , and s = 1725rad / s .
3.
Design a highpass Butterworth transfer function.
The specifications are given
by min = 55dB , max = 0.5dB , p = 1500rad / s , and s = 300rad / s .
4.
Design a passpass Butterworth transfer function.
The specifications are given
by min = 22dB , max = 0.5dB , p,1 = 500rad / s , p,2 = 1000rad / s , s ,1 = 333rad / s ,
and s ,2 = 1500rad / s .
IV. Computer Simulations
1. For three designed Butterworth filters, plot the magnitude and phase responses according to
their transfer functions (using MATLAB, MATHCAD, or other languages).
2. For the lowpass filter, record the attenuation values at the frequencies p = 1000rad / s ,
and s = 1725rad / s .
3. For the highpass filter, record the attenuation values at the frequencies p = 1500rad / s ,
and s = 300rad / s .
4. For the bandpass filter, record the attenuation value at the frequencies p,1 = 500rad / s ,
p,2 = 1000rad / s , s,1 = 333rad / s , and s,2 = 1500rad / s .
V. Experiments
This lab is a computer simulation lab.
No actual experiment.
VI. Lab Report
In the report, present the simulation results. Comment on deviations from expected results, if
any, and the reasons for these deviations. Your report should include the following:
1. The design steps.
2. The computer simulation results: the magnitude and phase responses for all designed
Butterworth filters.
3. Compare the specification attenuations min = 20dB and max = 1dB of the lowpass
Butterworth filter with the corresponding attenuations of the approximated transfer function.
To see if the designed transfer function satisfies the requirements.
4. Compare the specification attenuations min = 55dB and max = 0.5dB of the highpass
Butterworth filter with the corresponding attenuations of the approximated transfer function.
To see if the designed transfer function satisfies the requirements.
5. Compare the specification attenuations min = 22dB , and max = 0.5dB of the bandpass
Butterworth filter with the corresponding attenuations of the approximated transfer function.
To see if the designed transfer function satisfies the requirements.
6. Summary and conclusions.
References
[1]. M. E. Van Valkenburg, Analog Filter Design, Oxford University Press, 1982.
APPENDIX
How to read Resistor Color Codes
First the code
Black
Brown
Red
Orange
Yellow
Green
Blue
Violet
Gray
White
How to read the code
First find the tolerance band, it will typically be gold (5%) and sometimes silver (10%).
Starting from the other end, identify the first band  write down the number associated with
that color; in this case Blue is 6.
Now 'read' the next color, here it is red so write down a '2' next to the six. (you should have
'62' so far.)
Now read the third or 'multiplier' band and write down that number of zeros.
In this example it is two so we get '6200' or '6,200'. If the 'multiplier' band is Black (for zero)
don't write any zeros down.
If the 'multiplier' band is Gold move the decimal point one to the left. If the 'multiplier' band
is Silver move the decimal point two places to the left. If the resistor has one more band past
the tolerance band it is a quality band.
Read the number as the '% Failure rate per 1000 hour' This is rated assuming full wattage
being applied to the resistors. (To get better failure rates, resistors are typically specified to
have twice the needed wattage dissipation that the circuit produces) 1% resistors have three
bands to read digits to the left of the multiplier. They have a different temperature coefficient
in order to provide the 1% tolerance.
Examples
Example 1:
You are given a resistor whose stripes are colored from left to right as brown, black, orange,
gold. Find the resistance value.
Step One: The gold stripe is on the right so go to Step Two.
Step Two: The first stripe is brown which has a value of 1. The second stripe is black which
has a value of 0. Therefore the first two digits of the resistance value are 10.
Step Three: The third stripe is orange which means x 1,000.
Step Four: The value of the resistance is found as 10 x 1000 = 10,000 ohms (10 kilohms =
10 kohms).
The gold stripe means the actual value of the resistor mar vary by 5% meaning the actual
value will be somewhere between 9,500 ohms and 10,500 ohms. (Since 5% of 10,000 = 0.05
x 10,000 = 500)
Example 2:
You are given a resistor whose stripes are colored from left to right as orange, orange,
brown, silver. Find the resistance value.
Step One: The silver stripe is on the right so go to Step Two.
Step Two: The first stripe is orange which has a value of 3. The second stripe is orange
which has a value of 3. Therefore the first two digits of the resistance value are 33.
Step Three: The third stripe is brown which means x 10.
Step Four: The value of the resistance is found as 33 x 10 = 330 ohms.
The silver stripe means the actual value of the resistor mar vary by 10% meaning the actual
value will be between 297 ohms and 363 ohms. (Since 10% of 330 = 0.10 x 330 = 33)
Example 3:
You are given a resistor whose stripes are colored from left to right as blue, gray, red, gold.
Find the resistance value.
Step One: The gold stripe is on the right so go to Step Two.
Step Two: The first stripe is blue which has a value of 6. The second stripe is gray which has
a value of 8. Therefore the first two digits of the resistance value are 68.
Step Three: The third stripe is red which means x 100.
Step Four: The value of the resistance is found as 68 x 100 = 6800 ohms (6.8 kilohms = 6.8
kohms).
The gold stripe means the actual value of the resistor mar vary by 5% meaning the actual
value will be somewhere between 6,460 ohms and 7,140 ohms. (Since 5% of 6,800 = 0.05 x
6,800 = 340)
Example 4:
You are given a resistor whose stripes are colored from left to right as green, brown, black,
gold. Find the resistance value.
Step One: The gold stripe is on the right so go to Step Two.
Step Two: The first stripe is green which has a value of 5. The second stripe is brown which
has a value of 1. Therefore the first two digits of the resistance value are 51.
Step Three: The third stripe is black which means x 1.
Step Four: The value of the resistance is found as 51 x 1 = 51 ohms.
The gold stripe means the actual value of the resistor mar vary by 5% meaning the actual
value will be somewhere between 48.45 ohms and 53.55 ohms. (Since 5% of 51 = 0.05 x 51
= 2.5