Chapter 1

Analog Filter: Concepts

1.1 Introduction
The study of electric network theory has two related but distinct spheres, namely network
analysis and network synthesis. The two terms are easily identified in terms of the network, the
excitation or input, and the response or output as shown in Figure 1.1. Here network means
some combination of passive elements, such as the resistor, the capacitor and the inductor,
and (not necessarily) dependent energy sources. The excitation or input is an electrical energy
source connected to the network. The response or output is observed in different forms such as
voltage across certain element(s), current through certain element(s), or energy dissipated in a
resistor. This response is observed either in the time domain or in the frequency domain. In the
time domain, the output voltage/current variation is observed/measured with respect to time,
whereas in the frequency domain, the same observation is taken in terms of frequency.

1 2
Excitation Response
or Network or
1 2
Input Output

Figure 1.1 Network: a physical entity comprising passive (and active) components.

When the response of a given network to a certain excitation is observed/evaluated, it is

known as network analysis; whereas, the process of finding a network for a given excitation and
response is known as network synthesis. A significant difference between analysis and synthesis
is that while analyzing the network, for a certain input or excitation, the response solution is
unique; however, during the network synthesis process, it may be possible to obtain a number
of solutions and even no solution for a given combination of excitation and response. It is this

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 2 Continuous Time Active Analog Filters

possibility of obtaining many solutions which has led to the development of a large number
of synthesis procedures. Classification of the various synthesis methods shall be discussed in
Section 1.5. Sensitivity concept, being an important issue in the practical utilization of filters,
is introduced in Section 1.6 and discussed in more detail in a separate chapter, Chapter 6
The basic principles behind the classification of various types of network functions that filter
signals on the basis of their frequency, and the terminologies used, are explained in Sections
1.2 and 1.3. An important stage of conversion/evolution from passive filters (networks which
use only passive elements) to active filtering (using at least one active device) is discussed in
Section 1.4.
One very important reason for changing from passive filters to active filters in the integrated
circuit (IC) form was the inability of realizing practically feasible inductors in passive filters. For
a long time, active RC (resistance−capacitance) filter structures using resistance, capacitance
and operational amplifiers (OAs) were synonymous to active filters. Though the usage of OA
is still prevalent, other active devices are also being used now in a big way.
It is assumed that the reader is familiar with OAs; hence, only a brief discussion about these
amplifiers is included in Section 1.7. The application of OA as an integrator is reviewed in
Section 1.8. A brief discussion about the characteristics of resistors and capacitors fabricated
in the IC form is also included in Section 1.9 with the assumption that a detailed study of the
realization of these passive components is undertaken in the separate subject ‘microelectronics’
or an allied subject.

1.2 Network Functions

External connections from a network to another network or excitation and response
observations are made through terminals or ports: terminals are commonly paired to form a
port. For example, in Figure 1.2, there are four terminals in the network forming two ports.

+ I1 I2 +
1 2
V1 Two-port V2
network
1 – – 2

Figure 1.2 A two-port network showing conventional direction of current and voltage references.

For a single-port network, only one voltage and one current are identified and only one
network function (and its reciprocal) is defined, which is given as:

Z(s) = V(s)/I(s) (1.1)

This function is known as the driving-point impedance function because the terminals are
connected to the driving force or energy source. Generally, it is simply known as the impedance

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 3

of a network. Reciprocal of the impedance function is known as the driving-point admittance

function or simply admittance, Y(s).
For the two-port network in Figure 1.2, two currents and two voltages must be defined
as shown. The four quantities V1(s), I1(s), V2(s) and I2(s) can be related to each other in six
different forms (and their six reciprocals), when only two quantities are taken at a time. For
example, in Figure 1.2, the driving point impedance at input terminal port 1 for a passive
termination at port 2 is given as:

Z11(s) = V1(s) /I1(s) (1.2)

Similarly, the driving point impedance at port 2 for a passive termination at port 1 is given as:

Z22(s) = V2(s) /I2(s) (1.3)

Example 1.1: Find the driving point impedance function for the network shown in Figure 1.3.

Solution: As input voltage V1 is connected across the two impedances Z1 and Z2 in parallel,
total impedance shall be the parallel combination of these two impedances.

s 3 s2 + 3 s 5 s2 + 5
Z1 = + = and Z 2 = + =
2 2s 2s 8 8s 8s

1 1 1 s 4 + 8s 2 + 15
= + → Z (s) = (1.4)
Z z1 Z 2 10 s 3 + 34 s

1/2H 8/5F

Z in 1/8H
2/3F

Figure 1.3 A simple LC circuit for illustrating definition of driving point function.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 4 Continuous Time Active Analog Filters

Driving point impedance is a ratio of polynomials expressed in the complex frequency variable s.
The degree of the numerator is one more that of the denominator as shown as an example in
equation (1.4).

1.2.1 Transfer function

Consider Figure 1.2. Four out of the possible six network functions, which are defined for the
two-port network, are such that they relate a voltage or a current of one port to the voltage or
current of the other port. These functions are defined for a given driving force at the input port
and a terminating condition at the output port; they are known as transfer functions. For the
symbolic representation of the transfer function, a convenient convention is for the function to
be subscripted first by the input and then by the output; for example, the voltage ratio transfer
function is defined by the following relation:

V2(s) / V1(s) = G12(s) (1.5)

The current ratio transfer function is given by:

–I2(s) / I1(s) = α12(s) (1.6)

The transfer impedance function and the transfer admittance function are, respectively, as
follows:

V2(s) /I1(s) = Z12(s) and –I2(s)/V1(s) = Y12(s) (1.7)

In equations (1.6) and (1.7), I2(s) has a negative sign because of the conventionally chosen
direction of the second-port current going into the port.
Since each transfer functions is a ratio of the Laplace transform of current/voltage, these
functions are quotients of rational polynomials in the complex frequency variable s. For
example, looking at Figure 1.4, the network’s description given in the time domain is stated
as follows:

∫
L ( di / dt ) + iR + (1/ C ) i dt = v1 (t ) (1.8)

L
C
2H
+ 2F
e –t V1 R 1 V2
t0 –

Figure 1.4 A simple RLC circuit for illustrating voltage-ratio transfer function.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 5

Taking the Laplace transform of equation (1.8) for the given element values of the circuit, we
get:

 1 1
 2 s + 1 +  I ( s ) = (1.9)
2s ( 1)
s +

As V2(s) = I(s) × 1, and V1(s) = 1/(s + 1), we get the following expression for the voltage ratio
transfer function:

V2(s) / V1(s) = (2s)/(4s2 + 2s + 1) (1.10)

It is important to note that in equation (1.10), no component of the driving input or excitation
contributes to the expression of the transfer function; in fact, the transfer function depends
only on the element values and their interconnections. A general form of the transfer function
may be expressed as a ratio of polynomials in the complex frequency s.

N (s) a s m + am −1s m −1 + ………. + a1s + a0

H (s ) = = m (1.11)
D( s ) bn s n + bn −1 + s n −1 + ………….. + b1s + b0

In equation (1.11), with i = 0 to m and j = 0 to n, the coefficients ai and bj are real constants,
m is the degree of the numerator and n is the degree of the denominator. Factorization of both
the polynomials N(s) and D(s) gives the following important form of the transfer function.

N ( s ) am ( s − z1 )( s − z 2 )……….( s − zm )
= (1.12)
D( s ) bn ( s − p1 )( s − p2 )……..( s − pn )

Here, the roots of the numerator, z1, z2, …, zm are called zeros, and the roots of the denominator
p1, p2, …, pn are called poles of the transfer function, respectively. Consideration of zeros and
poles of a network function is extremely important as they describe the behavior of the network
function in the frequency domain.

1.3 Basic Filtering Action

As mentioned earlier, network synthesis is the process of finding circuit elements, such that their
values and interconnections conforms to the given relation between excitation and response.
Since excitation and response are both conveniently represented by the transfer functions and y
or z functions, the first step in the design of a filter is to convert the specified magnitude, phase
or related entity as a function of frequency to a form of the transfer function. Electronic filters
are circuits that perform signal processing in such a way that it separates signals on the basis
of frequency. Ideally, the filter suppresses signals in a certain band of frequency, known as stop

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 6 Continuous Time Active Analog Filters

band, completely and passes signals in other bands of frequency, known as pass bands, without
any attenuation or gain. Practically, however, it is not possible to suppress a signal completely
in the stop band, though the suppression is not less than a permissible level, and signals may
not pass without some attenuation in the pass band (unless intentional gain is added).

1.3.1 Types of filters

There are four commonly used types of filters: they are classified based on the terminology
of pass band and stop band mentioned in the previous section. The types as shown in
Figure 1.5(a−d) and are defined as follows:

Figure 1.5 Firm lines show ideal response and dashed lines shows practical response for (a) low
pass, (b) high pass, (c) band pass, and (d) band stop filters.

Low Pass Filters: When the frequency range of the pass band is from w = 0 to w = wc and the
stop band extends from w = wc to ∞, the filter is known as a low pass filter (LPF) as shown in
Figure 1.5(a). Here, wc is called the cut-off frequency.
LPFs are widely used in many diverse applications. For example, in acoustics, LPFs are used
to filter out high frequency signals from the transmitting sound that would otherwise cause
echo at higher sound frequencies. In audio speakers, LPF reduces the high frequency hiss
sound produced in the system and inputs the clearer sound to the sub-woofers. LPFs are also

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 7

used in equalizers and audio amplifiers. An important application of LPF is as an anti-aliasing

filter placed before the analog to digital convertor (ADC) and as a reconstruction filter placed
after the ADC. In a variety of applications, including those in medical instrumentation, LPFs
are used to eliminate high frequency noise.

High Pass Filter: The complement of LPF is the high pass filter (HPF), where the stop band
ranges from w = 0 to wc and the pass band extends from w = wc to ∞, as shown in Figure
1.5(b).
HPFs can be used wherever noise at low frequencies is to be eliminated, such as in medical
instrumentation and audio systems. They find applications in loudspeakers to reduce the low-
frequency noise. HPFs are also employed in those applications where high frequency signals
are to be amplified. For example, a popular application is the treble boost. In some applications,
combinations of HPFs with LPFs yield band pass filters or band stop filters. HPFs are used
to pass high frequency signals to a tweeter and to block interfering and potentially damaging
signals to loudspeakers. They are also used extensively in crossover of audio signals.

Band Pass Filters: In a band pass filter (BPF), the pass band ranges from w1 to w2; signals
in the rest of the frequency range, which is known as the stop band, are stopped as shown in
Figure 1.5(c).
BPFs are extensively used in all types of instruments, such as seismology and medical
instruments like electroencephalograms and electrocardiograms. There is widespread use of
BPFs in audio signal processing, where signals in a certain frequency range are to be passed
(maybe with amplification) and signals in the remaining band of frequencies are to be rejected.
In communication systems, the transmitter as well as the receiver employs BPFs to avoid
interference from unwanted signals.

Band Stop Filter: A band stop filter (BSF) (or a notch filter) complements BPF with the stop
band ranging between w1 and w2 and rest of the frequency band is the pass band as shown in
Figure 1.5(d).
These filters are mainly used in public address systems and speaker systems to ensure
rejection of power supply frequency interference. Similarly, BSFs are crucial for line noise
reduction in telephonic signal transmission. These filters are also employed in many electronic
communication devices to eliminate/reduce interference from harmonics. BPFs are used in the
medical instruments, such as electrocardiogram machines, to reject unwanted signals.
In Figure 1.5, straight lines show the ranges of pass and stop bands for ideal filters. It
has been proved mathematically in literature that such ideal characteristics are not realizable
exactly with finite number of elements [1.1]. The approximated responses for the filter, which
are practically realizable, are shown as dotted lines in Figure 1.5. The transfer function for
such practically realizable filter networks are described by the real rational function given in
equation (1.11). It is important to note that the transfer function expressed as the ratio of
polynomials in the complex frequency s must satisfy certain conditions for the filter to be
realizable practically. For example, the coefficients in the two polynomials in equation (1.11),
N(s) and D(s), ai, i = 0 to m and bj, j = 0 to n are should be real numbers. In addition, the degree

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 8 Continuous Time Active Analog Filters

of the denominator n must be larger than or at least equal to the degree of the numerator, m,
that is, n ≥ m. Magnitude of the transfer function in equation (1.11) when evaluated at the
jw-axis, that is, |H(jw)|, is a continuous function of frequency w as shown by the dotted lines
in Figure 1.5(a−d), rather than the characteristics given by the solid lines. It is to be noted
that the sharpness of the curves in Figure 1.5 depends on the value of the coefficients in the
denominator D(s); hence, sharpness can be controlled by the designer.
It is a common practice in filter circuit design to represent the characteristics of the circuit
in terms of linear output signal magnitude, or gain characteristic |H(jw)| or in terms of the
logarithmic attenuation characteristic α(w). Attenuation characteristics of the main four types
of filters, corresponding to Figure 1.5(a−d), are shown in Figure 1.6(a−d). Here, logarithmic
attenuation α(w) is related to the gain magnitude as follows:

α(w) = 20 log |H(jw)| dB (1.13)

Transition Transition
band band
a (w) a min a(w) a min
dB Stop dB
band Stop Pass
Pass band band
band
a max a max

wp ws w ws wp w
(a) (b)

a (w) a min a min a (w) a min

dB dB
Stop Stop Pass Stop Pass
band Pass band band band band
band
a max a max a max

w s1 w p1 w p2 w s2 w p1 w z1 w z2 w p2
w w
(c) (d)

Figure 1.6 Ideal and practical characteristics in term of logarithmic attenuation for the (a) low pass,
(b) high pass, (c) band pass, and (d) band stop filters.

It is to be noted that in the normal course, α will be positive for |H(jw)| > 1 and negative for
|H(jw)| < 1. However, in filter design, the negative sign for α is normally not used even when

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 9

|H(jw)| < 1. It is the context which informs the analyzer whether positive α implies gain during
filtering or attenuation during loss in signal magnitude. The negative sign for α is used only
when it becomes necessary to avoid any confusion.
Filter characteristics of Figure 1.5, which are drawn again in Figure 1.6, in terms of
attenuation, show clearly that for a practical filter, the boundary of the pass and the stop band
is separated by a definite region, which is called the transition band. This is not so in the case
of the ideal filter response, where transition from the pass band to the stop band is instant and
no transition region exists. It is the width of this transition region which decides the sharpness
of the filter characteristic, an important parameter in filter design.
Characteristics of important four types of filter shown in Figure 1.6 are shown as solid lines
and continuous dotted lines. The required practical characteristics are shown by the continuous
dotted lines; however, the characteristics are not fixed, in the sense that they are governed by
two parameters αmax and αmin shown on the solid lines. Here, αmax is the maximum attenuation
and implies that the attenuation should always be less than this value in the pass band; αmin
is the minimum attenuation and implies that the attenuation should always be more than
this value in the stop band. Hence, the most common method of providing specifications for
a filter is to mention αmin, αmax and the edge frequencies of the pass band and stop band w1
and w2, respectively. The mathematical process of finding the appropriate transfer function
H(s) in the form of equation (1.11) is known as approximation. There are different methods of
approximations: classical forms, which are discussed in Chapter 3, and other methods based
on the approximating phase or delay requirements of a filter, which are discussed in Chapter 4.
Another significant point to note is that the level of attenuation α(w) has been shown
to reach the level of zero dB. In reality, the level may be at zero dB, below or above zero dB
depending on the kind of circuitry used for filter realization and whether a gain (active) device
has been used in the circuit or not. However, this level of attenuation does not violate the
specifications αmin, αmax or pass band or stop band edge frequencies. Therefore, absolute value
of the attenuation in the pass band is not a worrying factor for the designer as shall be evident
later in the examples.

1.4 From Passive to Active Filtering

Over the years, passive elements have been used extensively for building passive filters. The
filter could use all the three elements, resistance (R), inductance (L), and capacitance (C), or
only two elements like RC, LC, or RL. Excellent literature is available on the conditions of
realization, and methods that enable the realization of such passive filters [1.1]. There are quite
a few advantages that such filters offer.
i. Passive filters do not require any power supply for operation.
ii. Two-port networks employing only LC components, commonly known as lossless
networks as they did not dissipate any power, were found to be very useful as they could
provide sharp transition between stop band and pass band.
iii. One of the most common forms of passive realization that was, and is still used, the ladder
structure. LC ladder structures have been studied extensively and data, such as structural

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 10 Continuous Time Active Analog Filters

combination, element values, poles, and zero location for their transfer function, are
available for different types of approximated filters [1.2]. One of the main advantages of
the LC ladders is their low element sensitivity in the pass band. This means that realized
filter parameters are comparatively much less affected due to changes (unavoidable or
intentional) in the values of the elements used.
iv. Passive filters can and are still used at some higher frequencies where use of inductance is
not very problematic and other types of filters face limitations.
v. Study of passive filters, especially the ladder structure, is important from the point of view
of active filter realizations as well, as many active realizations use passive ladder structures
as the starting point of synthesis because of their excellent sensitivity properties.

One of the major limitations of passive filters lies in its use of inductors, which are coils of
wires on some kind of core material. Not only are the inductors bulky, radiate electromagnetic
energy, and can result in parasitic mutual inductance, they are also practically not realizable in
integrated circuit (IC) form. With the increase in the use of electronic filters in a big way and
advances in technology, the use of inductors in filters has become a big limitation. The advent
of active devices, especially the availability of operational amplifiers (OAs) in IC form, has
reduced the need of inductors. These active filters use active device(s), mainly OAs (and later
operational transconductance amplifiers, current conveyors and other active devices as well),
capacitors and/or resistors. Such filter circuits can be implemented in a very small space in IC
form, cutting down the cost of ICs heavily [1.3].
In passive filters, output power was always less than the input power; whereas, in active
filters, use of active devices provides gain, which is very helpful especially when the input signal
magnitude is small.
When OA (or some other active device) is used, and the output is taken at the output
terminal of the OA, the output impedance of the active filter is (almost all the time) low. Such
a condition is suitable for connecting these filter circuits in a chain, better known as cascade,
even if the input impedance of a proceeding cascaded filter circuit is not very high. Cascades
are commonly used in active filters for enhancing the filters’ characteristics, resulting in higher-
order filters (where n of equation (1.11) is large).
As mentioned earlier, passive filters are used at high frequencies; this needs a bit more
explanation. Electronic filters can also be classified in terms of the reactive elements employed;
these elements control the useful frequency range of operation of the filter. For example, resonant
cavities are found suitable at microwave frequencies; whereas, excellent selectivity is provided
by the piezoelectric crystal unit in the frequency range 5 to 150 MHz. The intermediate radio
frequency reception range between 100 kHz and 10 MHz utilizes mainly ceramic filters. Even
mechanical filters can provide medium to high selectivity with good tunability in the audio
band down to about 0.1 Hz. Another important frequency spectrum lies between 20 kHz and
100 kHz, where lumped inductor and capacitors are used. However, as mentioned in Sections
1.4 and 1.5, active elements were introduced, particularly in the lower and sub-audio bands
(though not limited to this frequency range). This effectively amounted to replacing passive
inductors, thereby introducing the subject of this book: namely, active RC filters. As active

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 11

RC filters became widely used, the subject matter gained considerable momentum. However,
with advancements in technology, digital and sampled data posed great challenges to active
RC filters. Communication channels were progressively digitized and the filtering operation
was done directly on digital signals. Without going into a discussion on the limitations (or
comparison) of digital and sampled data filters, it can safely be said that analog filters are still
in great demand. There are a number of applications where analog filters become essential or
are preferred over digital or sampled data filters.
Active analog filters manufactured in IC form are in great demand in the field of
communication, instrumentation, medical science and many more areas (some of the
application areas were mentioned in Section 1.3.1), and their design is a specialized subject.
However, this book provides an introduction to this topic in order to enable one to design
filters with less demanding specifications in the domain of continuous-time filters and make
the reader ready to undertake the specialized study.
Before proceeding further, let us keep in mind some of the following points while using or
designing active analog filters.
(a) Active analog filters need dc power supply for active devices; the power is dissipated and
it produces heat which needs to be removed without increasing the working temperature.
(b) The range of the operating frequency is limited by the type of active device used, its
network topology and the magnitude of the input signal. Though the whole range of
operating frequency may extend from around a fraction of Hz to a few GHz, the actual
operating frequency range needs to be evaluated for each specific type of active filter with
specific active device(s).
(c) Depending on the type of technology used for ICs, the tolerance of the values of passive
elements and parameters of the active devices are different. Tolerance affects parameters
and characteristics of the realized filter. It is for this reason that it is important to study
the sensitivity effect on filter parameters due to the tolerance of the elements used for their
practical utilization.
(d) As active analog filters provide gain, small magnitude signals can also be processed.
However, active devices do generate noise and care has to be taken so that the signal does
not get polluted by noise. In addition, rise in the level of signal due to gain should not go
above the saturation level of the active device at any stage (final or intermediate) of the
filtering; the devices operate in linear range.

1.5 Active RC Filters

To decrease the size of electronic filters in accordance with modern technology, replacement
of inductors with appropriate elements becomes essential. Utilization of active devices, along
with resistance and capacitance elements, which simulated inductance, led to active RC filters
being considered as an alternative to passive filters. In the beginning, transistors were used
as active devices; however, integration of OAs in 1970s gave active RC circuits a big boost.
Parameters of OAs being close to ideal, active RC synthesis remained synonymous with OA−

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 12 Continuous Time Active Analog Filters

RC network synthesis for a long time. Inspite of the frequency-dependent gain of OAs, and
the advent of new active devices, OAs are still used in most voltage-mode filtering circuits.
Depending on the application, either general purpose but cheaper OAs or specially designed
but costlier OAs are used.
The new devices, which are also being increasingly used, are basically current-mode devices.
They include operational transconductance amplifiers (OTAs), current conveyors (CCs)
and its variants, and other current mode devices like current differencing transconductance
amplifiers (CDTAs).
In linear active networks, the signal processing properties are largely dependent on the
time constants, that is, RC products. Therefore, in IC filters, generally high-quality R and C
elements are realized; effort is also made to obtain filter parameters in terms of element ratios
as resistance or capacitance fabricated in ratio forms are more precise[1.3].
Whether we use OA-based design (OA−RC) or OTA based design (OTA−RC) or some
other variations (these will be discussed at a later stage), the basic synthesis processes are similar.
The processes are as follows:
i. Cascade form synthesis.
ii. Direct form synthesis using element substitution or through operational simulation.
In the cascade form active synthesis, the nth order transfer function, which was given in
equation (1.11), is decomposed into second-order functions (if n is even, and one first/third-
order sections if n is odd). These second-order functions are then individually realized and the
resulting non-interactive blocks are cascaded to obtain the overall transfer function H(s). Each
second-order building block can be expressed as follows:

N (s)
Hi (s ) = (1.14)
b2 s 2 + b1s + b0

where N(s), the numerator polynomial of second-order functions, is chosen according to the
required nature of the response of the section; whereas, the denominator polynomial defines
the two important filter parameters, viz., the pole frequency wo and pole Q (Q). These are
related to the denominator coefficient as follows:

w o = (b0 / b2 )½ and Q = (b0b2 ) / b1

½
(1.15)

There are some specific advantages while realizing second-order sections, hence great emphasis
has been given to the optimum design of second-order sections. In addition, different forms of
feedback and utilization of second-order (and first-order) sections have resulted in the follow
the leader feedback type of filters as well.
In the element substitution approach of the direct form of synthesis technique, the network
is directly realized from the transfer function H(s) of equation (1.11). A number of direct form
techniques are available; the following two are the most common.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 13

(a) Inductance simulation (IS) approach.

(b) Frequency dependent negative resistance (FDNR) approach.
These techniques and their implementation shall be taken up in later chapters.

Operational Simulation of Ladders: Using passive RLC ladders as the starting point,
active RC networks can also be realized in a slightly different direct form in what is known
as operational simulation. In this approach, instead of simulating the elements, the process
simulates the operation of the ladder by appropriately modeling the circuit equations and
voltage−current relations of the elements used. Such circuit equations are represented by block
diagrams in a signal flow graph.

1.5.1 Active R and active C filters

Conventionally, the design of active RC circuits assumes OA to be ideal with a large frequency-
independent voltage gain (AÆ∞). However, OAs exhibit a frequency-dependent gain of low
pass nature. This kind of non-ideality restricts the use of OA−RC circuits to mostly audio
frequency range while using economically cheaper off-the-shelf OAs. Suitable compensating
schemes have been developed to design filters with lesser dependence on the amplifier gain
characteristics. In an alternate approach, the frequency-dependent gain of the OAs was directly
used in the design; this helped in increasing the useful frequency range of operation to the
MHz range while using commercially available OAs. The non-ideality of OAs has been used
to develop active R and active C circuits, which employ OAs and resistors and OAs and
capacitors only, respectively. However, because of certain limitations, these types of filter
structures have not been manufactured in bulk. The limitations included the large tolerance
of the bandwidth of OAs and their dependence on temperature and biasing voltage. Suggested
methods to overcome these limitations in active R and active C filters are also available.

1.6 Sensitivity Concepts

Due to the availability of a number of choices for the synthesis of a particular transfer
function, it is possible to use a number of active circuits for a single specification. Hence,
the designer/user has to decide on the best possible choice amongst the available designs. To
take this decision, certain criterion is chosen, such as number of active and passive elements
used, the useful frequency range of operation and ease of design. However, while realizing the
circuit in practice, one significant problem is that element values differ from their nominal
values due to the fabrication process tolerance. In addition, the elements’ values change with
working temperature and variation in voltage and environment. Inaccurate modeling of the
passive as well as active elements and parasitic elements also contribute towards increasing the
complexities of the circuit. Since all the coefficients of the transfer function, hence its poles
and zeros, depend on the passive and active component values, it is obvious that the gain and
characteristics of the transfer function also deviate from the specified form. It is important to
evaluate the deviations in different performance parameters caused by the possible change in

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 14 Continuous Time Active Analog Filters

circuit element values. The amount of introduced error depends on the amount of component
tolerance and the sensitivity of the circuit’s performance parameter to these tolerances. As a
result, while comparing different available active circuits to choose the best possible alternative,
one of the important considerations is the study of sensitivity.

1.7 Operational Amplifier

OA is the most prominent active device used in active filters. It is expected that the reader is
familiar with the basics of OAs. Therefore, here, the discussion will be restricted to only those
aspects of OAs that are directly related to their usage in analog filters.
Figure 1.7 shows the pin connection diagram of the most commonly used OA, type 741. It
needs dual power supply and has two terminals, one for inverting and non-inverting inputs and
one for the output. Dual OA and quad OA ICs are also available with matching characteristics.
OA is basically a high-gain differential amplifier that can be modeled in the simplest form as
shown in Figure 1.8. Output voltage of the OA is the difference of the two input voltages
multiplied by the high gain factor A; it can be expressed as follows:

Vo = A(V+ – V–) (1.16)

Output, Vo

+
Power
±VDD
– supply
8 7 6 5
741
+
1 2 3 4
–

–ve +ve
input input
V– V+

Figure 1.7 Signal input/output and power supply connections for an operational amplifier.

The differential gain A is frequency dependent in a practical OA. Therefore, it is represented

by the single-pole roll-off model given here:

Aow a B B
A (s ) = =  (1.17)
s + wa s + wa s

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 15

Here, Ao, the ratio of the single-ended output voltage to the differential input voltage, is called
open-loop gain, or dc gain (i.e., gain at w = 0), wa is the open-loop bandwidth and B = Aowa
is the gain−bandwidth product. For a practical OA, Ao ≈ 105, wa ≈ 2π × 10 rad/s, that is,
B = 2π × 106 rad/s and the approximated expression in equation (1.17) is valid for all w  w a;
input resistance at each input (Ri) is normally more than 1 MW and the output resistance (Ro)
is less than 100 W. For the purpose of analysis, initially, OA is assumed to be ideal with the
following parameters: Ao = ∞, wa = 0, B = ∞, Ri = ∞ and Ro = 0; appropriate corrections are
made at a later stage. However, during simulation, an appropriate model has to be used, which
would involve a large number of parameters depending on the type of model used for the OA.

Ro
V– Vo
+

Vin Ri A o V in

–
V+

Figure 1.8 Simple model of a practical operational amplifier.

Open-loop Gain and Bandwidth: It is generally preferred to have the value of the open-loop
gain A as high as practically feasible. Obviously, during analysis, assuming the value of gain as
infinite will create less error if its absolute value is higher. Figure 1.9 shows the typical variation
of the open-loop gain with frequency; this is a replica of the frequency response of a general-
purpose OA like type 741. Roll-off at the rate of 20 dBs/decade is due to the presence of a
compensating capacitor in the OA; the roll-off begins at a small frequency wa. Figure 1.9 also
depicts the idea of the gain bandwidth of the OA, that is, the frequency where the gain falls to
unity or 0 dB. As the gain bandwidth product is a constant, use of OA in the close-loop form
increases its 3 dBs, or half-power bandwidth, while decreasing its close-loop gain.

Slew Rate: Signal magnitude needs to be controlled so that the level of output remains
below the supply voltage (±VDD), otherwise, the signal will be clipped and distorted. Another
important reason of distortion in OAs is due to its slew rate (SR) limitation. Because of the
current driving capability of the transistors used in OAs, output voltage cannot change at a
faster rate than specified. The maximum rate of change of the output voltage in terms of SR
is as follows:

SR = |dvo(t)/dt|max (1.18)

Therefore, for undistorted output signal, the bandwidth becomes limited. Its relation with SR
is given as:

Bandwidth = SR/(πVpp) (1.19)

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 16 Continuous Time Active Analog Filters

Here Vpp is the peak-to-peak output voltage. For a typical value of slew rate for the 741 type
OA, 0.5V/µs, if maximum useful frequency is fixed at 100 kHz, Vpp shall be limited to (0.5 ×
106/π105) = 1.59 volts.

Ao

Differential open–loop
gain (dB)

(0,0) wa Frequency

Figure 1.9 Typical variation of open-loop gain of operational amplifier with frequency.

Input Currents: OAs have a differential stage and transistors at this stage require bias current
to flow so that the transistors can operate in the saturation region. For bias current to flow, a dc
path must be made available for both the inputs to either ground or to the output of the OA.
Normally, the mentioned path remains available; but in circuits involving only capacitors/
switches, precaution is needed, as shall be shown later.

1.7.1 Analysis of circuits using operational amplifiers

For accurate analysis, simulation of filters using OAs needs better and appropriate models.
However, at the first instance, manual analysis assumes ideal OA with infinite open-loop gain
and input resistance and zero output resistance. This means that both input terminals are
forced to be at the same potential and no current flows in them. Hence, when the non-
inverting terminal is grounded, which happens in a majority of circuits, the inverting terminal
is forced to be at the ground potential; the inverting terminal is now said to be at virtual
ground. This concept of virtual grounds has been found to be very useful during analysis. With
the suggested conditions of the OA, it is preferable to employ Kirchhoff’s current law (KCL)
at the input/output terminals of the OAs, making analysis simpler.
The analysis procedure mentioned here gives fairly accurate results at low frequencies (up
to a few kHz range). Deviations from the expected results start taking place when the working
frequency is increased. Hence, the next step is to consider finite values of the open-loop gain
A, resulting in the appearance of a non-zero differential voltage between the inverting and non-
inverting input terminals. The gain A can be substituted by an appropriate numerical value
or by its frequency-dependent expression in equation (1.17). In specific cases where the effect
of finite Ri and Ro needs to be analyzed, there values are taken into consideration; otherwise,

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 17

these are taken as ideal. Considering A to be infinite is very helpful in manual analysis when
conventional circuit analysis methods are used.
In the next section, two-integrator circuits are analyzed using the aforementioned method,
with AÆ• as well as A being represented by the frequency-dependent model.

1.8 Integrators Using Operational Amplifiers

A large number of simple and not so simple applications of OAs are in practice. However, the
integrator is probably the most common OA application in active filtering. The integrating
function has been employed for simulating inductors (say in ladder structure of any order) or
for the realization of second-order filter sections. Assuming an ideal OA with infinite open-
loop gain, for the circuit shown in Figure 1.10, with no signal current going into the OA,
application of KCL at the inverting terminal gives:

(Vin − V x ) = (V x − Vout ) (1.20)

R (1/ sC )

Vin R
– Vout
Vx
+

Figure 1.10 Operational amplifier as inverting integrator.

Since Vx = 0, the transfer function of the circuit, working as an inverting integrator, is obtained
as:

Vout 1
=− (1.21)
Vin sCR

However, with the single-pole roll-off model of equation (1.17) having frequency-dependent
finite OA gain, the inverting terminal potential in Figure 1.10 shall be (−Vo/A) instead of
virtual zero. Simple analysis gives the closed-loop transfer function:

Vout 1 1
=−
Vin sCR 1 1  (1.22)
1+ 1 + 
A sCR 

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 18 Continuous Time Active Analog Filters

Expression in equation (1.22) will obviously reduce to equation (1.21) for AÆ ∞. Hence, a
larger value of A is always preferred as it means lesser deviation from the ideal condition.
Another commonly used variation of the integrator is shown in Figure 1.11(a). For a non-
ideal OA, with A being finite, the transfer function is obtained as follows:

Vo R2 1
=− (1.23)
Vin R1 (1 + sCR2 ) 1 R2 
1 + 1 + 
A  R1(1 + sCR2 ) 

1
− for A → ∞ (1.24)
R1C ( s + 1/ CR2 )

R2
R2 /R1
C 3 dB

Vin R1 Gain
– Vo Vo
+ Vin

1/CR2 w
(a) (b)
R

Vin R
– Vo
+

(c)

Figure 1.11 (a) Operational amplifier as lossy inverting integrator, (b) its frequency response, and (c)
OA as inverter.

Obviously, equation (1.23) will reduce to equations (1.22) and (1.21), respectively, for
R2Æ• and further with AÆ•. At high A, the response of equation (1.24) is like a first-order
low pass filter (LPF) shown in Figure 1.11(b) with 3 dB frequency (cutoff frequency) and the
dc gain as given here:

wLP = 1/CR2 (1.25)

hdc = (R2/R1) (1.26)

A non-inverting integrator can be obtained by cascading the inverting integrator with an OA

based inverter. For equal value resistors in the inverter shown in Figure 1.11(c), gain is (−1)

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 19

when A Æ •. However, with A represented by the approximated frequency-dependent model

of equation (1.17), the circuit’s gain is given as:

1 1
Gain = − →− (1.27)
1 + 2 / A (s ) 1 + 2s / B

This gain expression implies that when an inverter is added to obtain a non-inverting integrator,
it adds a bit more non-ideality to the inverting integrator. Compensation methods are used
for both inverting and non-inverting integrators to minimize the amount of non-ideality. This
will be studied in more detail in Chapter 7.

1.9 Passive Components in Monolithic Filters

In the early stages of the development of active filters, OAs and discrete resistors and capacitors
were mounted on a printed circuit board (PCB). The next stage of miniaturization was
called hybrid technology. In hybrid technology, RC components were fabricated using the (i)
thick film, or (ii) thin film technique and the OAs and RC components were bonded in
a single package reducing the overall size considerably and improving the reliability of the
interconnections between OAs and RC components. In the thick film technique, resistors are
fabricated by depositing a composite paste of conducting material and glass on an insulating
surface. Depending on the conducting material used, sheet resistance (resistance per square of
sheet or strip forming the resistance) of the paste is in the range of 1 Ω/square and 100 MegΩ/
square. Hence, resistors in the range of 0.5 Ω to 1 GΩ could be realized with tolerance in
the range of 20% to 50%. Reduction in tolerance became possible in the thin film technique
through deposition of metal films on insulating material. Fabricated resistance using thin film
was much smaller than that realized using thick film as its sheet resistance was in the range
of 10 Ω to 10 MegΩ. Capacitances were also fabricated in hybrid technology using thin and
thick films.
With advancements in fabrication technology employing bipolar, MOS and CMOS
technologies leading to system on chip (SOC) expertise, efforts were taken in the direction
of the fabrication of monolithic filters in which passive elements could also be integrated
simultaneously with active devices. At present, further advancements are taking place in the
fabrication of passive elements in integrated form using new materials and improved physical
configurations.

Resistors: Technologies may differ, but the most common method of fabricating resistance
is through layering rectangular sheets of semiconducting material. The value of the realized
resistance is equal to the number of squares multiplied by the sheet resistance. Obviously, a
larger value resistance requires a larger chip area.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 20 Continuous Time Active Analog Filters

While fabricating a transistor, a diffusion layer has also been used to realize resistors without
adding any separate processing step. Usually, the base diffusion resistor is the most commonly
used resistor in a bipolar process. For such resistors, a matching tolerance of +0.2% between
resistors is possible. If the base diffusion area is pinched by diffusing an n+ diffusion layer
over a p-type base diffusion, sheet resistance is increased to 2–10 kΩ/square. Resistances so
obtained are called pinched resistors; they realize larger resistance, but tolerance becomes worse,
ranging up to 50%.
An advanced technique of resistance fabrication is through ion implantation. A very thin
layer of implant (0.1–0.8 mm) leads to a very high value of sheet resistance (100−1000 Ω/
square). Matching tolerance is also good, being approximately 2%.
In a different technique, resistances are fabricated through active devices. Observing
the current−voltage relation of a BJT (bipolar junction transistor) or a MOS (metal oxide
semiconductor) transistor, it is easy to see that for a certain range of operation, these devices
behave as quite a stable resistance. In MOS or CMOS (complementary MOS) technology,
transistors operating in the linear region are used to realize active resistors. Resistance at the
fabrication/design stage is controlled by the width to length (W/L) ratio of transistors, and at
a later stage through gate voltage VG.
Important advantages of active resistors compared to passive resistors are as follows: (i)
required chip area in active resistors is very small for the same value of passive resistance, and
(ii) value of the resistance is easily controllable.

Capacitors: Quite a few techniques are available for the fabrication of capacitors in monolithic
IC technology: for example, using pn junctions, MOSFETs (metal oxide semiconductor field
effect transistors) and polysilicon capacitors. In the BJT process, capacitors are formed between
semiconductor junctions. However, the obtained value of the capacitor is very small: 0.05–0.5
pF/mil2 and large chip areas are needed for even small value capacitors.
It is well known that reverse bias semiconductor junctions create a depletion region. This
region acts as an insulator sandwiched between doped silicon on two sides, resulting in a
capacitor depending on the width of the depletion region. Along with some parasitic capacitors,
a depletion region capacitance of 0.001 pF/µm2 can be realized.
In the MOS technology process, an MOS capacitor is formed between the n+ diffusion
regions, while forming the channel region, a polysilicon layer, and a thin layer of silicon
dioxide or silicon nitride between them.
Polysilicon capacitors are the most commonly used ones, as the fabrication process suits
MOS technology. In this technique, basically, the gate of the transistor is made of polysilicon.
Thin oxide is deposited on top of a polysilicon layer which acts as the insulating layer over
another bottom polysilicon plate. This type of capacitor too has parasitic capacitance and shall
be considered later while discussing switched capacitor circuits in Chapter 15.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 21

References
[1.1] Van Valkenburg, M. E. 1976. Introduction to Modern Network Synthesis. New York: Wiley
Eastern Limited.
[1.2] Zverev, A. I. 1967. Handbook of Filter Synthesis. New York: Wiley.
[1.3] Moschytz, G. S. 1974. Linear Integrated Circuit Fundamentals Part I. New York: Van Nostrand
Reinhold.

Practice Problems

1-1 Find the driving point impedance function Z(s) for the network shown in Figure P1.1 and find its poles
and zeroes.

1/2H 8/5F

Z in 1/8H
2/3F

Figure P1.1

1-2 Plot the magnitude and phase of the following transfer functions for s = jw, 0 ≤ w ≤ 10.
1 s
(a) (b)
(s + 2)(s + 4) (s + 3)(s + 6)

s2
(c) 5 (s − 2) (d)
(s + 2) (s 2 + s + 1)

1-3 Find the transfer function for the network shown in Figure P1.2.

R=4 L = 2H
+

e –2t Vm C Vout
t0
0.5F
–

Figure P1.2

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 22 Continuous Time Active Analog Filters

1-4 Sketch the following ideal responses in one figure and classify them.
(i) pass band from 0 to 10 krad/s; rest is stop band.
(ii) pass band from 15 krad/s to infinity; rest is stop band.

1-5 Repeat problem 1.4 for the following:

(i) pass band from 10 to 12 krad/s; rest is stop band.
(ii) stop band from 18 to 20 krad/s; rest is pass band.

1-6 Sketch the magnitude responses for the following values of attenuations and classify them.
(i) αmax = 2 dBs, from 0 to 10 krad/s and αmin = 40 dBs from 20 krad/s to infinity.
(ii) αmax = 1 dBs, from 40 krad/s to infinity and αmin = 50 dBs from 0 to 20 krad/s.
(iii) αmax = 0.5 dBs, from 14 to 18 krad/s and αmin = 40 dBs from 0 to 8 krad/s and from 24 krad/s to
infinity.
(iv) αmax = 1 dB, from 0 to 6 krad/s and from 14 krad/s to infinity, αmin = 50 dBs from 8 to 12 krad/s

1-7 Apply (1/s) transformation to the network shown in Figure P1.2. Determine the new element values and
fine its voltage-ratio transfer function again. Does it remain same as before in problem 1-3 or not?

1-8 Determine the input impedance of the network of Figure P1.3, (a) with OA as ideal and (b) when OA is
represented by its first-pole roll-off model. Note: Unless specified, approximated model of OA, A ª B/s
shall be used.

R1 R3
–
+

Vin R2 Vout

– –

Figure P1.3

1-9 A 741 type OA employing power supply of + 15 volt was used to construct an inverting amplifier with
voltage gain of (– 5). Sinusoidal signal input frequency was 50 kHz. Find the largest input signal before it
gets distorted due to the slew rate limitation; assume standard value of SR.

1-10 (a) Find the transfer function for the amplifier shown in Figure P1.4 (Z1 = Z2 = 5 kΩ) and (Z1 = 5 kΩ and
Z2 = 50 kΩ). The amplifier is non-ideal and is modeled with single-pole roll-off model with Ao = 105 and
the first pole wa = 2 rad/s. Determine the 3-dB bandwidth of the amplifiers.

(b) Repeat problem in part (a) for Figure P1.5.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 Analog Filter: Concepts 23

Z2
R
+
+
Z1 – A
+ +
A –
+ +
Z2
Vin Vout Vin Vout
R Z1

– – – –

Figure P1.4 Figure P1.5

1-11 Derive the transfer function of the integrator circuit shown in Figure P1.6. Find its critical frequency when
OA used is near ideal and non-ideal with unity gain bandwidth of 105 rad/s and R1 = R2 = R3 =10 k Ohm,
R4 = 5 k Ohm and C = 0.1 nF.

R3

R4
–
Vout
A
+
Vin R1 R2

Figure P1.6

1-12 Determine the transfer functions of the circuits shown in figures P1.7 and P1.8, considering OA as (a)
ideal and (b) non-ideal with unity gain bandwidth of 105 rad/s. In Figures P 1.7 and P 1.8 capacitors C1 =
C2 = 1 pF, R1 = 10 k Ohm, R2 = 5 k Ohm, R = 1 k Ohm and C = 1 nF. Find critical frequencies when OA
is considered ideal and non-ideal.

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press

 24 Continuous Time Active Analog Filters

C2

R2
– Vout
A
+
R1
Vin

C1

Figure P1.7

–
R Vout
Vin 2
– +
1
+

Figure P1.8

https://doi.org/10.1017/9781108762632.002 Published online by Cambridge University Press