Frequency Response

8.1 BODE PLOTS

In this section, we consider the frequency response of some RC circuits that

appear frequently in amplifiers. Our objective is to show how circuit functions can
be quickly and easily plotted against frequency at least for simple circuits.
These plots are known as Bode plots.

Motivation

There are several reasons that we might want to plot the magnitude and phase of a
circuit function. We can consider a signal to consist of a summation of sinusoidal
components. Audio signals contain components ranging from about 20 Hz to 20
kHz. Electrocardiograms range from 0.05 Hz to about 100 Hz, and video signals
range from dc to 4.5 MHz. In order to amplify a signal without distortion, the
amplifier gain magnitude must be the same for all of the frequency components. If
we have a plot of the gain magnitude versus frequency, we can see whether or not
the gain is constant for all of the frequency components contained in the signal of
interest.
Another application for Bode plots is in designing feedback amplifiers. In
chapter 8, we show that partial feedback of the output signal to the input terminals
can be very useful. Oscillation (the spontaneous generation of signals) is a
problem that can occur in amplifiers with feedback. Bode plots of the amplifier
gain magnitude and phase shift versus frequency are very useful in avoiding
unwanted oscillations.

Use of the Laplace Transform Variable

In our discussion, we use the Laplace transform variable s. Actually, we don’t

make much use of Laplace transform theory, but we use the notation and some of
the terminology of Laplace transforms, so you will be able to relate this discussion
to material you have studied or will study in other courses. Eventually, we
substitute s = j. If you haven’t studied the Laplace transform, for now you can
consider s to be simply a shorthand notation for j. Thus, in the analysis of
circuits, a capacitor, C, is replaced by an impedance 1 sC , and an inductance, L, is
replaced by an impedance sL. Resistors have impedances equal to their
resistances.

An Example: The Low pass RC Circuit

As a first example, we consider the voltage transfer function of the circuit

displayed in Figure 8.1. By applying the voltage divider principle, we can find the
ratio of the output voltage to the input voltage:
V 1 sC 1
Av ( s)  0 ( s)   (8.1)
Vin R  1 sC 1  RCs
Network functions for lumped linear time-invariant circuits can always be
expressed as a ratio of polynomials in s. In this case, the numerator is simply a
constant, and the denominator is a first-order polynomial. In more complex
circuits, the numerator and denominator can both be high-order polynomials.

R
+

+ 1
Vin (s)  V0 (s)
sC

Figure 8.1 Flow-pass RC filter.

Poles and Zeros

The values of s that cause the denominator polynomial to become zero are called
poles of the network function. The magnitude of the network function is
unbounded at the poles. The values of s that cause the numerator to become zero
are called zeros of the network function. Thus, the poles are roots of the
denominator polynomial, and the zeros are roots of the numerator polynomial. For
example, the voltage transfer ratio given in Equation (8.1) has a pole at
s   1 RC , but no zeros. (Sometimes we say that it has a zero at s = , because
the transfer function goes to zero as s approaches infinity.)

Break or Cut-off or Corner Frequencies

In the sinusoidal steady-state analysis of a network, we use j2f in place of the
Laplace transform variable s. If we substitute j2f into Equation (8.1), we have
1
Av ( f )  (8.2)
1  j 2RCf
This is a complex quantity that gives the magnitude and phase of the voltage
transfer ratio as a function of frequency. For example, if we evaluate the equation
for f  1 2RC , we obtain
1
Av ( f )   0.707  45o (8.3)
1  j1
The meaning of this result is that, if we apply a sinusoid of frequency
f  1 2RC to the input, the output will have an amplitude that is 0.707 times the
input amplitude. Furthermore, the output will be phase shifted by 45o with
respect to the input.
It is convenient to express Equation (8.1) as
V 1
Av ( f )  0  (8.4)
Vin 1  j  f f b 
where
1
fb  (8.5)
2RC
is called the break frequency. Alternative names for fb are the corner
frequency, cut-off frequency, half-power frequency, and 3-dB frequency.

Gain Magnitude Expressed in Decibels (dB)

Sometimes the voltage magnitude transfer gain is expressed is decibels (dB) and it
is obtained by taking 20 times the common logarithm of the gain magnitude.
Taking the magnitude of both sides of Equation (7.4), we have
1
Av ( f )  (8.6)
2
1   f fb 
Now, if this equation is expressed in decibels, we obtain
Av ( f ) dB  20 log 1   f f b 
2
(8.7)
which can also be written as

Av ( f ) dB  10 log 1   f f b 
2
(8.8)

Logarithmic Frequency Scales

In Bode plots, we use a logarithmic scale for frequency. On this kind of scale, the
variable is multiplied by a given factor for equal increments of length along the
axis. (On a linear scale, equal lengths on the scale correspond to adding a given
amount to the variable.) A logarithmic frequency scale is shown in Figure 8.2.
One One
decade octave
f (Hz)
10 20 50 100 200 500 1000
Figure 8.2 Logarithmic frequency scale.

A decade is a range of frequencies for which the ratio of the highest frequency
to the lowest is 10. The frequency range from 2 Hz to 20 Hz is one decade.
Similarly, the range from 50 Hz to 5000 Hz is two decades. (50 Hz to 500 Hz is
one decade, and 500 Hz to 5000 Hz is another decade.)
An octave is a two-to-one change in frequency. For example, 10 Hz to 20 Hz is
one octave. The range from 2 kHz to 16 kHz is three octaves.

The Magnitude Bode Plot

A Bode plot shows the magnitude of a network function in decibels versus

frequency, with the latter on a logarithmic scale. It turns out that such plots of
network functions can often be closely approximated by straight-line segments.
To illustrate, notice that the given by Equation (8.8) is approximately 0 dB for
f << fb. Thus, for low frequencies, the gain is approximated by the horizontal
straight line shown in Figure 8.3, labeled as the low-frequency asymptote.
On the other hand, for f >> fb, Equation (8.8) becomes


Av ( f ) dB  10 log  f f b 
2
 (8.9)
which is equivalent to
Av ( f ) dB  20 log  f f b  (8.10)
Evaluating Equation (8.10) for various values of f, we obtain the results shown
in Table 8.1. Plotting these values results in the straight line shown sloping
downward on the right-hand side of Figure 8.3, labeled as the high-frequency
asymptote.
 Av (dB)

fb
10 fb 10fb 100fb
0 f
3 Table 8.1. Values of the
Low-frequency high-frequency asymptote
asymptote
of Av ( f ) dB for selected
Actual response curve frequencies.
20
f Av ( f ) dB

High-frequency fb 0
asymptote
2fb 6
(20 dB/decade slope) 10fb 20
40
100fb 40
1000fb 60
Figure 8.3 Bode plot for the low-pass RC filter.

Notice that the two straight-line asymptotes intersect at the break frequency fb.
The slope of the high-frequency asymptote is 20 dB per decade of frequency.
This slope can also be stated as 6 dB per octave.
If we evaluate Equation (8.8) at f = fb, we find that
Av ( f ) dB  3 dB (8.11)
Thus, the asymptotes are in error by only 3 dB at the corner frequency. Very
likely, we would not be off by more than a fraction of a decibel by drawing a
freehand curve through this point, asymptotically approaching the straight-line
approximations for large and small values of f. The actual curve for Av ( f ) dB is
also shown in Figure 8.3.
  (deg)

fb
10 fb 10fb 100fb
0 f

Actual phase curve

45
Approximation

Figure 8.4
Bode plot for phase 90
of the low-pass RC
filter.

The Phase Plot

From Equation (8.4), we find that the phase of Av( f ) is

   tan 1 ( f f b )
Evaluating, we find that the phase approaches zero at very low frequencies, equals
45o at the corner frequency, and approaches 90o at high frequencies.
Figure 8.4 shows a plot of phase versus frequency. Notice that the curve can be
approximated by the following straight-line segments: a horizontal line at zero for
f  f b 10 ; a sloping line from zero phase at f b 10 to 90o at 10 fb. The actual
curve departs from these straight-line approximations by less than 6o. Thus,
working by hand, we can easily construct an approximate plot of phase versus
frequency.
Many circuit functions can be plotted by the methods that we have just
demonstrated for the simple low-pass RC circuit.

8.2 ALTERNATE FORM OF VOLTAGE TRANSFER FUNCTION

An alternate form of voltage transfer function, for all the circuits dealt in this
chapter, can be written as
 am s m  am-1s m-1  .........  a0
T (s )  (8.25)
bn s n  bn-1s n -1  ........  b0

where the coefficients a and b are real numbers and the order m of the numerator
is smaller than or equal to the order n of the denominator; the latter is called the
order of the network. Furthermore, for a stable circuitthat is, one that does
not generate signals on its ownthe denominator coefficients should be such that
the roots of the denominator polynomial all have negative real parts. We shall
study the amplifier stability in Chapter 9.
An alternate form for expressing T(s) is

( s  Z1 )(s  Z 2 ).........(s  Z m )
T ( s )  am (8.26)
( s  P1 )( s  P2 ).........(s  Pn )

where am is a multiplicative constant (the coefficient of sm in the numerator), Z1,

Z2, ……, Zm are the roots of the numerator polynomial, and P1, P2,…….., Pn are
the roots of the denominator polynomial. Z1, Z2,……, Zm are the transfer
function zeros or transmission zeros, and P1, P2,…….., Pn are the transfer
function poles or the natural modes of the network. A transfer function is
completely specified in terms of its poles and zeros together with the value of the
multiplicative constant.
The poles and zeros can be either real or complex numbers. However, since the
a and b coefficients are real numbers, the complex poles (or zeros) must occur in
conjugate pairs. That is, if 5 + j3 is a zero, then 5  j3 also must be a zero. A zero
that is pure imaginary (j Z) causes the transfer function T(j ) to be exactly zero
at  = Z. This is because the numerator will have the factors (s + jZ) (s jZ) =
(s2 +  Z2 ), which for physical frequencies become ( 2 +  Z2 ), and thus the
transfer function will be exactly zero at  = Z. Thus, the “trap” one places at the
input of a television set is a circuit that has a transmission zero at the particular
interfering frequency. Real zeros, on the other hand, do not produce transmission
null. Finally, note that for values of s much greater than all the poles and seroes,
the transfer function in Equation (8.25) becomes T (s )  am s m-n . Thus, the
transfer function has (n  m) zeros at s = .
First-Order Functions

All the transfer functions encountered in this chapter have real poles and can
therefore be written as the product of first-order transfer functions of the general
form
a s  a0
T ( s)  1 (8.27)
s  0
where  0 is the location of the real pole. The quantity 0, called the pole
frequency, is equal to the inverse of the time constant of the single-time constant
(STC) network. The constants a0 and a1 determine the type of STC network.
Specifically, we studied in Section 8.1 two types of STC networks, low pass and
high pass. For the low pass first-order network we have

a0
T ( s)  (8.28)
s  0
In this case, the dc gain is a0  0 , and 0 is the corner or 3dB frequency. Note
that this transfer function has one zero at s = . On the other hand, the first-order
high-pass transfer function has a zero at dc and can be written as

a1s
T ( s)  (8.29)
s  0

Bode Plots

A simple technique exists for obtaining an approximate plot of the magnitude and
phase of a transfer function given its poles and zeros has already been discussed in
Section 8.1. The technique is particularly useful in the case of real poles and
zeros. The method was developed by H. Bode, and the resulting diagrams are
called Bode plots.
A transfer function of the form depicted in Equation (8.26) consists of a product
of factors of the form s + a, where such a factor appears on top if it corresponds to
a zero and on the bottom if it corresponds to a pole. It follows that the magnitude
response in decibels of the network can be obtained by summing together terms of
the form 20 log a 2   2 , and the phase response can be obtained by summing
terms of the form tan 1 ( a). In both cases, the terms corresponding to poles are
summed with negative signs. For convenience, we can extract the constant a and
write the typical magnitude term in the form 20 log 1  ( a ) 2 . On a plot of
decibel versus log frequency, this term gives rise to the curve and straight-line
asymptotes shown in Figure 8.18. As discussed in Section 8.1, the low-frequency
asymptote is a horizontal straight line at 0-dB level and the high-frequency
asymptote is a straight line with a slope of 6 dB/octave or 20 dB/decade. The two
asymptotes meet at the frequency   a , which is called the corner frequency.
As indicated, the actual magnitude plot differs slightly from the value given by the
asymptotes; the maximum difference is 3 dB and occurs at the corner frequency.
For a = 0  that is, a pole or a zero at s = 0  the plot is simply a straight line
of 20 dB/decade slope intersecting the 0-dB line at  = 1.
20 log 1  ( a) 2 (dB)

High-frequency
asymptote
Actual curve (+20 dB/decade slope)

3 dB
0

Low-frequency
asymptote

 (log scale)
1
3dB a 

Figure 8.18 Bode plot for the typical magnitude term. The curve shown applies for the
case of a zero. For a pole, the high-frequency asymptote should be drawn with a 20
dB/decade slope.
In summary, to obtain the Bode plot for the magnitude of a transfer function, the
asymptote plot for each pole and zero is first drawn. The slope of the high-
frequency asymptote of the curve corresponding to a zero is +20 dB/decade, while
that for a pole is 20 dB/decade. The various plots are then added together, and
the overall curve is shifted vertically by an amount determined by the
multiplicative constant of the transfer function.
 Gain (dB)

(5)
60
(1)

40

20 (4)

(2) (3)
3
0.1 1 10 2
10 10 4
10 105 106 107  (rad/s)
(log scale)

Figure 7.19 Bode plots for Example 7.3.

Example 8.3 Bode Plots and Approximate Voltage Gain Determination

An amplifier has the following voltage transfer function

10 s
T (s ) 
(1  s 10 2 )(1  s 105 )

Find the poles and zeros and sketch the magnitude of the gain versus frequency.
Find approximate values for the gain at   10, 103 , and 10 6 rad/s.

SOLUTION The zeros are as follows: one at s = 0 and one at s = . The poles
are as follows: one at s = 102 rad/s and one at s = 105 rad/s.
Figure 8.19 shows the asymptote Bode plots of the different factors of the
transfer function. Curve 1, which is a straight line with +20 dB/decade slope,
corresponds to the s term (that is, the zero at s = 0) in the numerator. The pole at s
= 102 results in curve 2, which consists of two asymptotes intersecting at
  10 2. Similarly, the pole at s = 105 represented by curve 3, where the
intersection of the asymptotes is at   105. Finally, curve 4 represents the
multiplicative constant of value 10.
Adding the four curves results in the asymptotic Bode diagram of the amplifier
gain (curve 5). Note that since the two poles are widely separated, the gain will be
very close to 103 (60 dB) over the frequency range 102 to 105 rad/s. At the two
corner frequencies (102 to 105 rad/s) the gain will be approximately 3 dB below
the maximum of 60 dB. At the three specific frequencies, the values of the gain as
obtained from the Bode plot and from exact evaluation of the transfer function are
as follows:
______________________________________
 (rad/s) Approximate Gain Exact Gain
______________________________________
10 40 dB 39.96 dB
103 60 dB 59.96 dB
106 40 dB 39.96 dB
______________________________________
 
  tan 1  
a

o
0.1 a 5.7 a 10 a
 (log scale)
0
Actual curve
o
45

o
90
o
5.7
Figure 8.20 Bode plot of the typical phase term tan 1 ( a) when a is negative.

We next consider the Bode phase plot. Figure 8.20 shows a plot of the typical
phase term tan 1 ( a), assuming that a is negative. An asymptotic straight-line
approximation of the arctan function is also shown in Figure 8.20. The asymptotic
plot consists of three straight lines. The first is horizontal at  = 0 and extends up
to   0.1 a . The second line has a slope of 45o/decade and extends from
  0.1 a to   10 a . The third line has a zero slope and a level of  = 90o. The
complete phase response can be obtained by summing the asymptotic Bode plots
of the phase of all poles and zeros.

Example 8.4 Bode Plots for the Phase  = tan1(/a) of Example 8.3
Find the Bode plots for the phase of the transfer function of the amplifier
considered in Example 8.3.
 
  tan -1  
a

(4)
o (1)
+90

o
+60

o
+30

0.1 1 10 102 103 104 105 106 107  (rad/s)

(log scale)
o
30
(3)
(2)
o
60
(4)
o
90

Figure 8.21 Phase plots for Example 8.4.

SOLUTION The zero at s = 0 gives rise to a constant +90o phase function

represented by curve 1 in Figure 8.21. The pole at s = 102 gives rise to the phase
function
  
1   tan 1  2 
 10 
(the leading minus sign is due to the fact that this singularity is a pole). The
asymptotic plot for this function is given by curve 2 in Figure 8.21. Similarly, the
pole at s = 105 gives rise to the phase function
  
2   tan 1  5 
 10 
whose asymptotic plot is given by curve 3. The overall phase response (curve 4) is
obtained by direct summation of the three plots. We see that at 100 rad/s, the
amplifier phase leads by 45o and at 105 rad/s the phase lags by 45o.

An Important Remark

For constructing Bode plots, it is important to express the transfer function factors
 
in the form 1  . The material of Figures 8.18 and 8.20 and of the preceding
 a
two examples is then directly applicable.

8.3 THE AMPLIFIER TRANSFER FUNCTION

The amplifiers considered in this chapter have voltage gain functions of either of
the two forms shown in Figure 8.22. Figure 8.22(a) applies for direct-coupled or
dc amplifier and Figure 8.22(b) for capacitively-coupled or ac amplifiers. The
only difference between the two types is that the gain of the ac amplifier falls off
at low frequencies. In the following we shall study the more general response
shown in Figure 8.22(b). The response of the dc amplifier follows as a special
case.
Gain A (dB) Gain A (dB)

3 dB 3 dB

A0 AM

H  L H 
(a) (b)
Figure 8.22 Frequency response for (a) a dc amplifier and (b) a capacitively-coupled
amplifier.
The Three Frequency Bands

As can be seen from Figure 8.22(b) the amplifier gain is almost constant over a
wide frequency range called the midband. In this frequency range all
capacitances (coupling, bypass, and transistor internal capacitances) have
negligible effects and can be ignored in gain calculation. At the high-frequency
end of the spectrum, the gain drops owing to the effect of the internal capacitances
of the device. On the other hand, at the low-frequency end of the spectrum, the
coupling and bypass capacitances no longer act as perfect short circuits and thus
cause the gain to drop. The extent of the midband is usually defined by the two
frequencies L and H. These are the frequencies at which the gain drops by 3 dB
below the value at midband. The amplifier bandwidth is usually defined as

BW  H  L (8.30)

and since L  H ,

BW  H (8.31)

A figure of merit for the amplifier is its gain-bandwidth product, defined as

GB  AM H (8.32)

where AM is the magnitude of midband gain in volts per volt. As will be shown in
later sections, it is generally possible to trade off gain for bandwidth.

The Gain Function A(s)

The amplifier gain as a function of the complex frequency s can be expressed in

the general form

A( s)  AM FL ( s) FH ( s) (8.33)

where FL ( s) and FH ( s) are functions that account for the dependency of gain on
frequency in the low-frequency band and in the high-frequency band,
respectively. For frequencies  much smaller than H, the function FH ( s)
approaches unity. Thus, for  L    H ,

A( s)  AM

as should have expected. It also follows that the gain of the amplifier in the low-
frequency band, AL ( s), can be expressed as

AL ( s)  AM FL ( s) (8.34)

and the gain in the high-frequency band can be expressed as

AH (s )  AM FH ( s) (8.35)

The midband gain is determined by analyzing the amplifier equivalent circuit

with the assumption that the coupling and bypass capacitors are acting as perfect
short circuits and the internal capacitors of the transistor model are acting as
perfect open circuits. The low-frequency transfer function, AL ( s), is determined
from analysis of the amplifier equivalent circuit including the coupling and bypass
capacitors but assuming that the transistor model capacitances behave as perfect
open circuits. On the other hand, the high-frequency transfer function, AH ( s), is
determined from analysis of the amplifier equivalent circuit including the
transistor model capacitors but assuming that the coupling and bypass capacitors
behave as perfect short circuits. Figure 8.23 provides a pictorial representation of
these points.
Low-frequency High-frequency
Midband band
band

 Coupling and bypass  No capacitors in  Transistor internal

capacitors in effect effect capacitances in effect

 FH(s) 1  FL(s) 1, FH(s) 1  FL(s) 1

A(s)  AMFL(s) A(s)  AM A(s)  AMFH(s)

0 fL fH f (Hz)

Figure 8.23 The three frequency bands that characterize the frequency response of
capacitively-coupled amplifiers. For dc amplifiers, the absence of coupling and bypass
capacitors causes FL(s) = 1 and fL = 0, thus the midband gain extends to zero frequency.
The Low-Frequency Response

The function FL(s), which characterizes the low-frequency response of the

amplifier, takes the general form

( s   Z1 )(s   Z 2 ).........( s   ZnL )

FL ( s)  (8.36)
(s   P1 )(s   P 2 ).........( s  PnL )
where P1, P2, ….., PnL are positive numbers representing the frequencies of the
nL low-frequency poles and Z1, Z2, ….., ZnL are positive, negative, or zero
numbers representing the nL zeros. It should be noted from Equation (8.36) that as
s approaches infinity (in fact, as s = j  approaches midband frequencies), FL(s)
approaches unity.
The amplifier designer is usually particularly interested in the part of the low-
frequency band that is close to the midband. This is because the designer needs to
estimate, and if need be modify, the value of the lower 3-dB frequency L. In
many cases, the zeros are at such low frequencies (much smaller than L) as to be
of little importance in determining L. Also, usually one of the polessay, P1
has a much higher frequency than all other poles. It follows that for frequencies 
close to the midband, FL(s) can be approximately by

s
FL ( s)  (8.37)
( s   P1 )

which is the transfer function of a first-order high-pass network. In this case, the
low-frequency response of the amplifier is dominated by the pole at s =  P1 and
the lower 3-dB frequency is approximately equal to P1,

L  P1 (8.38)

If this dominant-pole approximation holds, it becomes a simple matter to

determine L. Otherwise one has to find the complete Bode plot for FL ( j ) and
thus determine L. As a rule of thumb the dominant-pole approximation can be
made if the highest-frequency pole is separated from the nearest pole or zero by at
least two octaves (that is, a factor of four).
 If a dominant low-frequency pole does not exist, an approximate formula can
be derived for L in terms of the poles and zeros. For simplicity, consider the case
of a circuit having two poles and two zeros in the low-frequency band; that is,
(s   Z1 )(s   Z 2 )
FL ( s)  (8.39)
( s   P1 )(s   P 2 )

Substituting s = j and taking the squared magnitude gives

( 2   Z1
2
2
)( 2   Z2
2
)
FL ( j )  2 2 2 2
(8.40)
(   P1 )(   P2 )

2 1
By definition, at  = L, FL  , and thus
2

1 (L2   Z1
2
)( L2   Z2
2
)
 2 2 2 2
2 ( L   P1 )( L   P2 )
1  (1  L2 )( Z1
2 2
  Z2 )  (1  L4 )( Z1
2 2
 Z2 )
 2 2 2 4 2 2
(8.41)
1  (1  L )( P1   P2 )  (1  L )(P1 P2 )

Since L is usually greater than the frequencies of all the poles and zeros, we
may neglect the terms containing (1 L4 ) and solve for L to obtain
2 2 2 2
L  P1   P2  2 Z1  2 Z2 (8.42)
This relationship can be extended to any number of poles and zeros. Note that if
one of the polessay, P1 is dominant, then P1 >> P2, Z1, Z2, and Equation
(8.42) reduces to Equation (8.38).

Example 8.5 Determination of Low Frequency Response

The low frequency response of an amplifier is characterized by the following

transfer function
s( s  10)
FL ( s) 
( s  100)(s  25)
Determine its 3-dB frequency, approximately and exactly.
SOLUTION Noting that the highest frequency pole at 100 rad/s is two octaves
higher than the second pole and a decade higher than the zero, we find that a
dominant-pole situation almost exists and L  100 rad/s. A better estimate of L
can be obtained using Equation (8.42), as follows:
L  100 2  252  2  10 2  102 rad/s
The exact value of L can be determined from the given transfer function as 105
rad/s. Finally, we show in Figure 8.24 a Bode plot and an exact plot for the
magnitude of the given transfer function. Note that this is a plot of low-frequency
response of the amplifier normalized relative to the midband gain. That is, if the
midband gain is 100 dB, then the entire plot should be upward by 100 dB.

FL (dB)

3 dB
0
4 Bode plot
8
12
16
Exact
20
24
28
32
36
40
44  L = 105 rad/s
 (rad/s)
5 10 25 50 100 (log scale)
Figure 8.24 Normalized low-frequency response of the amplifier in Example 8.5.

The High-Frequency Response

Consider next the high-frequency band. The function FH(s) can be expressed in
the general form
 (1  s  Z1 )(1  s  Z 2 ).........(1  s  ZnH )
FH ( s)  (8.43)
(1  s P1 )(1  s  P2 ).........(1  s  PnH )

where P1, P2, ….., PnH are positive numbers representing the frequencies of the
nH high-frequency real poles and Z1, Z2, …..,ZnH are positive, negative, or
infinite numbers representing the frequencies of the nH high-frequency zeros.
Note from Equation (8.43) that as s approaches 0 (in fact as s = j approaches
midband frequencies), FH(s) approaches unity.
The amplifier designer is usually particularly interested in the part of the high-
frequency band that is close to the midband. This is because the designer needs to
estimate, and if need be modify, the value of the upper 3-dB frequency H. In
many cases the zeros are either at infinity or at such high frequencies as to be of
little significance in determining H. If in addition one of the high-frequency
polessay, P1is of much lower frequency than any of the other poles, then the
high-frequency response of the amplifier will be dominated by this pole, and the
function FH(s) can be approximated by

1
FH ( s)  (8.44)
1  s P1

which is the transfer function of a first-order low-pass network. It follows that if a

dominant high-frequency pole exists then the determination of H is greatly
simplified:
 H   P1 (8.45)

If a dominant high-frequency pole does not exist, the upper 3-dB frequency H
can be determined from a plot of FH ( j ) . Alternatively, an approximate formula
for H in terms of the high-frequency poles and zeros can be derived in a manner
similar to that used above in deriving Equation (8.42). The formula for H is

1 1 2 2
H  1 2
 2  .....  2  2 ..... (8.46)
P1 P2  Z1  Z2

Note that if one of the poles, say P1, is dominant then P1 << P2, P3, ….., Z1,
Z2, …. and Equation (8.46) reduces to Equation (8.45).
 Example 8.6 Determination of High Frequency Response
The high frequency response of an amplifier is characterized by the following
transfer function
1  s 10 5
FH (s ) 
(1  s 10 4 )(1  s 4  10 4 )
Determine its 3-dB frequency, approximately and exactly.

SOLUTION Noting that the lowest frequency pole at 104 rad/s is two octaves
lower than the second pole and a decade lower than the zero, we find that a
dominant-pole situation almost exists and  H  10 4 rad/s. A better estimate of H
can be obtained using Equation (8.46), as follows:

1 1 2
H  1 8
 8
 10  9800 rad/s
10 16  10 10

The exact value of H can be determined from the given transfer function as 9537
rad/s. Finally, we show in Figure 8.25 a Bode plot and an exact plot for the given
transfer function. Note that this is a plot of the high-frequency response of the
amplifier normalized relative to the midband gain. That is, if the midband gain is
100 dB, then the entire plot should be upward by 100 dB.

Using Short-Circuit and Open-Circuit Time Constants for the

Approximation Determination of L and H

If the poles and zeros of the amplifier transfer function can be determined easily,
then we can determine L and H using the technique described above. In many
cases, however, it is not a simple matter to determine the poles and zeros. In such
cases, approximate values of L and H can be obtained using the following
method.
Consider first the high-frequency response. The factors in the function FH(s) of
Equation (8.43) can be multiplied out and FH(s) expression in the alternative form
can be written as

1  a1s  a2 s 2  .........  anH s nH

FH ( s )  (8.47)
1  b1s  b2 s 2  .........  bnH s nH
 FH (dB)

3 dB
0 Bode plot
4
8
12
16 Exact
20
24
28
32
36
40
44
 (rad/s)
5103 104 2104 4104 105 2105 4105 (log scale)
 H = 9537 rad/s
Figure 8.25 Normalized high-frequency response of the amplifier in Example 8.6.

where the coefficients a and b are related to the zero and pole frequencies,
respectively. Specially, the coefficient b1 is given by

1 1 1
b1    .....  (8.48)
 P1  P1  PnH

It can be shown that the value of b1 can be obtained by considering the various
capacitances in the high-frequency equivalent circuit one at a time while reducing
all other capacitors to zero (or, equivalently, replacing them with open circuits).
That is, to obtain the contribution of capacitance Ci we reduce all other
capacitances to zero, reduce the input signal source to zero, and determine the
resistance Rio seen by Ci. This process is then repeated for all other capacitors in
the circuit. The value of b1 is computed by summing the individual time constants,
called open-circuit time constants,
 nH
b1   Ci Rio (8.49)
i 1

where we have assumed that there are nH capacitors in the high-frequency

equivalent circuit.
This method for determining b1 is exact; the approximation comes about in
using the value of b1 to determine H. Specifically, if the zeros are not dominant
and if one of the polessay, P1is dominant, then from Equation (8.48)
1
b1  (8.50)
 P1

and the upper 3-dB frequency will be approximately equal to P1, leading to

1
H  (8.51)
 
 Ci Rio 
 i 

Here it should be pointed out that in complex circuits we usually do not know
whether or not a dominant pole exists. Nevertheless, using Equation (8.51) to
determine H normally yields remarkably good results even if a dominant pole
does not exists. The method of open-circuit time constants has an important
advantage in that it tells the circuit designer which of the various capacitances is
significant in determining the amplifier frequency response.
Next we outline the use of short-circuit time constants to determine the lower
3-dB frequency, L. The factors in the function FL(s) of Equation (8.36) can be
multiplied out and FL(s) expression in the alternative form can be written as

s nL  d1s nL-1  ........

FL ( s )  (8.52)
s nL  e1s nL-1  ........

where the coefficients d and e are related to the zero and pole frequencies,
respectively. Specifically, the coefficient e1 is given by

e1   P1  P2  ......   PnL (8.53)

The exact value of e1 can be obtained by analyzing the amplifier low-frequency

equivalent circuit, considering the various capacitors one at a time, while setting
all other capacitors to  (or, equivalently, replacing them with short circuits).
Thus if capacitors Ci is under consideration, we replace all other capacitors with
short circuits, and also reduce the input signal to zero, and determine the
resistance Ris seen by Ci. The process is then repeated for all other capacitors, and
the value of e1 is computed from

nL
1
e1   (8.54)
i1 Ci Ris

where it is assumed that there are nL capacitors in the low-frequency equivalent

circuit.
The value of e1 can be used to obtain an approximate value of the 3-dB
frequency L provided that none of the zeros is dominant and that a dominant
pole exists. This condition is satisfied if one of the polessay, P1has a
frequency P1 much higher than (at least four times) that of all the other poles and
zeros. If this is the case then L  P1 and from Equation (8.53) we see that e1 
P1, leading to
nL
1
L   (8.55)
i 1 Ci Ris

Of course, in a complex circuit it is usually difficult to ascertain whether or not a

dominant low-frequency pole exists. Nevertheless, the method of short-circuit
time constants usually provides a reasonable estimate of L. Such an estimate is
quite sufficient for an initial paper-and-pencil design. The method also affords the
designer considerable insight into which of the various capacitors most severely
limits the low-frequency response.