9

FILTER DESIGN

A circuit designer frequently requires filters to extract the desired frequency

spectrum from a wide variety of electrical signals. If a circuit passes all signals
from dc through a frequency ωc but stops the rest of the spectrum, it is known as
a low-pass filter. The frequency ωc is called its cutoff frequency. Conversely, a
high-pass filter stops all signals up to ωc and passes those at higher frequencies.
If a circuit passes only a finite frequency band that does not include zero (dc) and
infinite frequency, it is called a bandpass filter. Similarly, a bandstop filter passes
all signals except a finite band. Thus, bandpass and bandstop filters are specified
by two cutoff frequencies to set the frequency band. If a filter is designed to
block a single frequency, it is called a notch filter.
The ratio of the power delivered by a source to a load with and without a
two-port network inserted in between is known as the insertion loss of that two-
port. It is generally expressed in decibels. The fraction of the input power that
is lost due to reflection at its input port is called the return loss. The ratio of the
power delivered to a matched load to that supplied to it by a matched source is
called the attenuation of that two-port network. Filters have been designed using
active devices such as transistors and operational amplifiers, as well as with only
passive devices (inductors and capacitors only). Therefore, these circuits may
be classified as active or passive filters. Unlike passive filters, active filters can
amplify the signal besides blocking the undesired frequencies. However, passive
filters are economical and easy to design. Further, passive filters perform fairly
well at higher frequencies. In this chapter we present the design procedure of
these passive circuits.

Radio-Frequency and Microwave Communication Circuits: Analysis and Design, Second Edition,
By Devendra K. Misra
ISBN 0-471-47873-3 Copyright  2004 John Wiley & Sons, Inc.

333
334 FILTER DESIGN

There are two methods available to synthesize passive filters. One of them
is known as the image parameter method and the other as the insertion-loss
method. The former provides a design that can pass or stop a certain frequency
band, but its frequency response cannot be shaped. The insertion-loss method is
more powerful in the sense that it provides a specified response of the filter. Both
of these techniques are described in this chapter. The chapter concludes with a
design overview of microwave filters.

9.1 IMAGE PARAMETER METHOD

Consider the two-port network shown in Figure 9.1. V1 and V2 represent voltages
at its ports. Currents I1 and I2 are assumed as indicated in the figure. Note that I1
is entering port 1 while I2 is leaving port 2. Further, Zin is the input impedance
at port 1 when Zi2 terminates at port 2. Similarly, Z0 is the output impedance
with Zi1 connected at port 1. Zi1 and Zi2 are known as the image impedance of
the network. Following the transmission parameter description of the two-port,
we can write

V1 = AV2 + BI2 (9.1.1)

and
I1 = CV2 + DI2 (9.1.2)

Therefore, the impedance Zin at its input can be found as

V1 AV2 + BI2 AZi2 + B

Zin = = = (9.1.3)
I1 CV2 + DI2 CZi2 + D
Alternatively, (9.1.1) and (9.1.2) can be rearranged as follows after noting that
AD —BC must be unity for a reciprocal network. Hence,

V2 = DV1 − BI1 (9.1.4)

and
I2 = −CV1 + AI1 (9.1.5)

I2

V1 A B V2
Zi1 Zin Z0 Zi2
C D

Figure 9.1 Two-port network with terminations.

 IMAGE PARAMETER METHOD 335

Therefore, the output impedance Z0 is found to be

V2 DV1 − BI1 DZi1 + B

Z0 = − =− = (9.1.6)
I2 −CV1 + AI1 CZi1 + A

Note that
V1
Zi1 = −
I1

For Zin = Zi1 and Z0 = Zi2 , equations (9.1.3) and (9.1.6) give

Zi1 (CZi2 + D) = AZi2 + B (9.1.7)

and
Zi2 (CZi1 + A) = DZi1 + B (9.1.8)

Subtracting equation (9.1.8) from (9.1.7), we find that

D
Zi2 = Zi1 (9.1.9)
A

Now, substituting (9.1.9) into (9.1.7), we have

AB
Zi1 = (9.1.10)
CD

Similarly, substituting (9.1.10) into (9.1.9) yields

BD
Zi2 = (9.1.11)
AC

The transfer characteristics of the network can be formulated as follows.

From (9.1.1),

V1 I2 B
=A+B =A+
V2 V2 Zi2
or


V1 A √ √
= ( AD + BC)
V2 D

For a reciprocal two-port network, AD − BC is unity. Therefore,

V2 D AD − BC D √ √
= √ √ = ( AD − BC) (9.1.12)
V1 A AD + BC A
336 FILTER DESIGN

Similarly, from (9.1.2),

I2 A √ √
= ( AD − BC) (9.1.13)
I1 D

Note that equation (9.1.12) is similar to (9.1.13) except that the multiplying coef-
ficient in one is the reciprocal of the other. This coefficient may be interpreted
as the transformer turns ratio. It is unity for symmetrical T and π networks. The
propagation factor γ (equal to α + j β, as usual) of the network can be defined as
√ √
e−γ = AD − BC

Since AD − BC = 1, we find that

√
cosh γ = AD (9.1.14)

The characteristic parameters of π and T networks are summarized in Table 9.1;

the corresponding low-pass and high-pass circuits are given in Table 9.2.

TABLE 9.1 Parameters of T and π Networks

π-Network T-Network

Z1 Z1/2 Z1/2

2Z2 2Z2 Z2

Z1
Z1 A =1+
ABCD parameters A =1+ 2Z2
2Z2
Z12
B = Z1 B = Z1 +
4Z2
1 Z1
C = + 1
Z2 4Z22 C =
Z2
Z1
D =1+ Z1
2Z2 D =1+
2Z2

  
Z1 Z2 Z1 Z2 Z1
Image impedance Ziπ = = ZiT = Z1 Z2 1 +
1 + Z1 /4Z2 ZiT 4Z2

Z1 Z1
Propagation constant, γ cosh γ = 1 + cosh γ = 1 +
2Z2 2Z2
 IMAGE PARAMETER METHOD 337

TABLE 9.2 Constant-k Filter Sections

Filter Type T-Section π-Section

Low-pass L /2 L /2 L

C C/2 C/2

High-pass 2C 2C C

L 2L 2L

For a low-pass T-section as illustrated in Table 9.2,

Z1 = j ωL (9.1.15)
and
1
Z2 = (9.1.16)
j ωC

Therefore, its image impedance can be found from Table 9.1 as follows:
  
L ω2 LC
ZiT = 1− (9.1.17)
C 4

In the case of a dc signal, the second term inside the parentheses will be zero
and the resulting image impedance is generally known as the nominal impedance,
Z0 . Hence,

L
Z0 =
C

Note that the image impedance goes to zero if ω2 LC/4 = 1. The frequency that
satisfies this condition is known as the cutoff frequency, ωc . Hence,

2
ωc = √ (9.1.18)
LC
338 FILTER DESIGN

Similarly, for a high-pass T-section,

1
Z1 = (9.1.19)
j ωC
and
Z2 = j ωL (9.1.20)

Therefore, its image impedance is found to be

  
L 1
ZiT = 1− (9.1.21)
C 4ω2 LC

The cutoff frequency of this circuit will be given as

1
ωc = √ (9.1.22)
2 LC

Example 9.1 Design a low-pass constant-k T-section that has a nominal imped-
ance of 75
and a cutoff frequency of 2 MHz. Plot its frequency response in the
frequency band 100 kHz to 10 MHz.

SOLUTION Since the nominal impedance Z0 must be 75
,

L
Z0 = 75 =
C

The cutoff frequency ωc of a low-pass T-section is given by (9.1.18). Therefore,

2
ωc = 2π × 2 × 106 = √
LC

These two equations can be solved for the inductance L and capacitance C,
as follows:

L = 11.9366 µH
and
C = 2.122 nF

This circuit is illustrated in Figure 9.2. Note that inductance L calculated here is
twice the value needed for a T-section. Propagation constant γ of this circuit is
 IMAGE PARAMETER METHOD 339

5.9683 µH 5.9683 µH
2.122 nF

Figure 9.2 Low-pass constant-k T-section.

0.1 0.2 0.5 1 2 5 10

0

−10
Magnitude (dB)

−20

−30

−40
0.1 0.2 0.5 1 2 5 10
Frequency (MHz)

0.1 0.2 0.5 1 2 5 10

−25

−50
Phase (deg)

−75

−100

−125

−150

−175
0.1 0.2 0.5 1 2 5 10
Frequency (MHz)

Figure 9.3 Frequency response of the T-section shown in Figure 9.2.

determined from the formula listed in Table 8.1. The transfer characteristics are
then found as e−γ .
The frequency response of the designed circuit is shown in Figure 9.3.
The magnitude of the transfer function (ratio of the output to input voltages)
remains constant at 0 dB for frequencies lower than 2 MHz. Therefore, the
340 FILTER DESIGN

output magnitude will be equal to the input in this range. It falls by 3 dB if

the signal frequency approaches 2 MHz. It falls continuously as the frequency
increases further. Note that the phase angle of the transfer function remains
constant at −180◦ beyond 2 MHz. However, it shows a nonlinear characteristic
in the passband.
The image impedance ZiT of this T-section can be found from (9.1.17). Its
characteristics (magnitude and phase angle) with frequency are displayed in
Figure 9.4. The magnitude of ZiT reduces continuously as frequency is increased
and becomes zero at the cutoff. The phase angle of ZiT is zero in passband, and
it changes to 90◦ in the stopband. Thus, image impedance is a variable resistance
in the passband, whereas it switches to an inductive reactance in the stopband.
The frequency characteristics illustrated in Figures 9.3 and 9.4 are representa-
tive of any constant-k filter. There are two major drawbacks to this type of filter:

1. The signal attenuation rate after the cutoff point is not very sharp,
2. The image impedance is not constant with frequency. From a design point
of view, it is important that it stay constant, at least in its passband.

350
300
250
Magnitude (Ω)

200
150
100
50

2 4 6 8 10
Frequency (MHz)

80

60

40

20

2 4 6 8 10
Frequency (MHz)

Figure 9.4 Image impedance of the constant-k filter of Figure 9.2 as a function of
frequency.
 IMAGE PARAMETER METHOD 341

These problems can be remedied using the techniques described in the follow-
ing section.

m-Derived Filter Sections

Consider the two T-sections shown in Figure 9.5. The first network represents
the constant-k filter considered in the preceding section, and the second is a
new m-derived section. It is assumed that the two networks have the same
image impedance. From Table 9.1 we can write
     
 Z Z1
ZiT = ZiT = Z1 Z2 1 + 1 = Z1 Z2 1 +
4Z2 4Z2

It can be solved for Z2 as follows:

 

1 Z 2 − Z12
Z2 =  Z1 Z2 + 1 (9.1.23)
Z1 4

Now assume that

Z1 = mZ1 (9.1.24)

Substituting (9.1.24) in to (9.1.23), we get

Z2 1 − m2
Z2 = + Z1 (9.1.25)
m 4m

Thus, an m-derived section is designed from the values of components determined

for the corresponding constant-k filter. The value of m is selected to sharpen
the attenuation at cutoff or to control the image impedance characteristics in
the passband.

Z1/2 Z1/2 Z′1/2 Z′1/2

Z2 Z′2

(a) (b)

Figure 9.5 Constant-k T-section (a) and m-derived section (b).

342 FILTER DESIGN

For a low-pass filter, the m-derived section can be designed from the corre-
sponding constant-k filter using (9.1.24) and (9.1.25) as follows:

Z1 = j ωmL (9.1.26)

and
1 − m2 1
Z2 = j ωL + (9.1.27)
4m j ωmC

Now we need to find its propagation constant γ and devise some way to control
its attenuation around the cutoff. Expression for the propagation constant of a T-
section is listed in Table 9.1. To find γ of this T-section, we first divide (9.1.26)
by (9.1.27):

Z1 ω2 m2 LC 4ω2 m2 /ω2c

= − = − (9.1.28)
Z2 1 − [(1 − m2 )/4]ω2 LC 1 − (1 − m2 )ω2 /ω2c

where
2
ωc = √ (9.1.29)
LC

Using the formula listed in Table 9.1 and (9.1.28), the propagation constant γ is
found as follows:

Z1 2 (mω/ωc )2
cosh γ = 1 +  =1−
2Z2 1 − (1 − m2 ) (ω/ωc )2
or
ω2c − ω2 − (mω)2
cosh γ = (9.1.30)
ω2c − (1 − m2 )ω2

Hence, the right-hand side of (9.1.30) is dependent on the frequency ω. It will

go to infinity (and therefore, γ) if the following condition is satisfied:
ωc
ω= √ = ω∞ (9.1.31)
1 − m2
This condition can be used to sharpen the attenuation at cutoff. A small m will
place ω∞ close to ω. In other words, ω∞ is selected a little higher than ωc and
the fractional value of m is then determined from (9.1.31). Z1 and Z2 of the m-
derived section are subsequently found using (9.1.24) and (9.1.25), respectively.
Note that the image impedance of an m-derived T-section is the same as
that of the corresponding constant-k network. However, it will be a function of
m in the case of a π-network. Therefore, this characteristic can be utilized to
design input and output networks of the filter so that the image impedance of
the composite circuit stays almost constant in its passband. Further, an infinite
cascade of T-sections can be considered as a π-network after splitting its shunt
 IMAGE PARAMETER METHOD 343

Z1′/2 Z1′/2 Z1′/2 Z1′/2

Z2′ Z2′

(a)

Z1′

2Z2′ 2Z2′

(b)

Figure 9.6 Deembedding of a π-network (b) from a cascaded T-network (a).

arm as illustrated in Figure 9.6. Note that the impedance Z2 of the T-network is
replaced by 2Z2 , while two halves of the series arms of the T-network give Z1
of the π-network. Image impedance Ziπ is found from Table 9.1 as follows:

Z1 Z2 Z1 Z2 + [(1 − m2 )/4]Z12

Ziπ = = (9.1.32)
ZiT ZiT

For the low-pass constant-k filter, we find from (9.1.15) to (9.1.18) that

L
Z1 Z2 = = Z02
C
 2
2Z0 ω
Z12 = −ω2 L2 =
ωc
and,

 2
ω
ZiT = Z0 1 −
ωc
Therefore,
1 − (1 − m2 ) (ω/ωc )2
Ziπ =  Z0
1 − (ω/ωc )2
344 FILTER DESIGN

2.5

2.0
Normalized impedance

m = 0.8
1.5
m = 0.6

1.0

m = 0.4

0.5
0.0 0.2 0.4 0.6 0.8 1.0
Normalized frequency

Figure 9.7 Normalized image impedance of π-network versus normalized frequency for
three values of m.

or
1 − (1 − m2 )ω2
Z iπ =  (9.1.33)
(1 − ω2 )

where Z iπ = Ziπ /Z0 may be called the normalized image impedance of a π-

network and ω = ω/ωc the normalized frequency.
Equation (9.1.33) is displayed graphically in Figure 9.7. The normalized image
impedance is close to unity for nearly 90% of the passband if m is kept around
0.6. Therefore, it can be used as pre- and poststages of the composite filter (after
bisecting it) to control the input impedance. These formulas will be developed
later in this section.

Example 9.2 Design an m-derived T-section low-pass filter with a cutoff fre-
quency of 2 MHz and a nominal impedance of 75
. Assume that f∞ = 2.05 MHz.
Plot the response of this filter in the frequency band of 100 kHz to 10 MHz.

SOLUTION From (9.1.31),

 2
fc
1 − m2 =
f∞
Therefore,  
 2  2
fc 2
m= 1− = 1− = 0.2195
f∞ 2.05
 IMAGE PARAMETER METHOD 345

TABLE 9.3 Design Relations for Composite Filters

Low-Pass High-Pass

Constant-k T-filter Constant-k T-filter

√ √
Z0 = L/C Z0 = L/C
√ √
ωc = 2/ LC 2C 2C ωc = 1/2 LC
L/2 L/2 L
C L = 2Z0 /ωc L = 0.5Z0 /ωc
C = 2/Z0 ωc C = 0.5/Z0 ωc

m-derived T-Section m-derived T-Section

(Values of L and C are the same (Values of L and C are the same
as above) as above)
     
fc 2 f∞ 2
m= 1− m= 1−
f∞ fc

mL/2 mL/2 2C/m 2C/m

1 − m2 L L/m
4m
4m C
mC 1 − m2

Input and Output Matching Sections Input and Output Matching Sections

C/0.3 C/0.3
0.3L 0.3L L/0.3 L/0.3
8 8L
L 15
15 15 C 15
8 C
0.3C 0.3C 8

Using (9.1.24) and (9.1.25) or Table 9.3 and component values of the constant-
k section obtained earlier in Example 9.1, the m-derived filter is designed as
follows:

mL/2 = 0.2195 × 5.9683 = 1.31 µH

mC = 0.2195 × 2.122 = 465.78 nF

and
1 − m2
L = 12.94 µH
4m
This filter circuit is illustrated in Figure 9.8.
346 FILTER DESIGN

1.31 µH 1.31 µH

12.94 µH

465.78 pF

Figure 9.8 An m-derived T-section for Example 9.2.

As noted earlier, the image impedance of this circuit will be the same as that
of the corresponding constant-k section. Hence, it will vary with frequency as
illustrated in Figure 9.4. The propagation constant γ of this circuit is determined
from the formula listed in Table 9.1. The transfer characteristics are then found
as e−γ . Its magnitude and phase characteristics versus frequency are illustrated in
Figure 9.9. The transfer characteristics illustrated in Figure 9.9 indicate that the
m-derived filter has a sharp change at a cutoff frequency of 2 MHz. However, the
output signal rises to −4 dB in its stopband. On the other hand, the constant-k
filter provides higher attenuation in its stopband. For example, the m-derived
filter characteristic in Figure 9.9 shows only 4-dB attenuation at 6 MHz, whereas
the corresponding constant-k T-section has an attenuation of more than 30 dB at
this frequency, as depicted in Figure 9.3.

Composite Filters
As demonstrated through the preceding examples, the m-derived filter provides
an infinitely sharp attenuation right at its cutoff. However, the attenuation in
its stopband is unacceptably low. Contrary to this, the constant-k filter shows a
higher attenuation in its stopband, although the change is unacceptably gradual
at the transition from passband to stopband. One way to solve the problem is to
cascade these two filters. Since image impedance stays the same in two cases,
this cascading will not create a new impedance-matching problem. However, we
still need to address the problem of image impedance variation with frequency
at the input and output ports of the network.
As illustrated in Figure 9.7, the image impedance of the π-section with m =
0.6 remains almost constant over 90% of the passband. If this network is bisected
to connect on either side of the cascaded constant-k and m-derived sections,
it should provide the desired impedance characteristics. To verify these char-
acteristics, let us consider a bisected π-section as shown in Figure 9.10. Its
transmission parameters can be found easily following the procedure described
in Chapter 8.
From (8.4.4) to (8.4.7), we find that

Z1
A=1+ (9.1.34)
4Z2
 IMAGE PARAMETER METHOD 347

0.1 0.2 0.5 1 2 5 10

0

−2
Magnitude (dB)

−4

−6

−8

−10

0.1 0.2 0.5 1 2 5 10

Frequency (MHz)

0.1 0.2 0.5 1 2 5 10

0

−25

−50
Phase (deg)

−75

−100

−125

−150

−175
0.1 0.2 0.5 1 2 5 10
Frequency (MHz)

Figure 9.9 Frequency response of the m-derived T-section shown in Figure 9.8.

Z′1/2

2Z′2 Zi2
Zi1

Figure 9.10 Right-hand side of the bisected π-section shown in Figure 9.6.
348 FILTER DESIGN

Z1
B= (9.1.35)
2
1
C= (9.1.36)
2Z2
and
D=1 (9.1.37)

Image impedance Zi1 and Zi2 can now be found from (9.1.10) and (9.1.11)
as follows:


Z2
Zi1 = Z1 Z2 + 1 = ZiT (9.1.38)
4
and

Z1 Z2 Z1 Z2
Zi2 =   = = Ziπ (9.1.39)
1 + Z1 /4Z2 ZiT

Thus, the bisected π-section can be connected at the input and output ports
of cascaded constant-k and m-derived sections to obtain a composite filter that
solves the impedance problem as well. Relations for the design of these circuits
are summarized in Table 9.3.

Example 9.3 Design a low-pass composite filter with a cutoff frequency of

2 MHz and an image impedance of 75
in its passband. Assume that f∞ =
2.05 MHz. Plot its response in the frequency range 100 kHz to 10 MHz.

SOLUTION Constant-k and m-derived sections of this filter are designed in

Examples 9.1 and 9.2, respectively. The filter’s input and output matching sec-
tions can be designed as follows. With m = 0.6, we find that

mL
= 3.581 µH
2
mC
= 636.6 pF
2
and
1 − m2
L = 6.366 µH
2m
This composite filter is depicted in Figure 9.11.
Figure 9.12 illustrates the frequency response of this composite filter. As it
indicates, there is a fairly sharp change in output signal as the frequency changes
from its passband to the stopband. At the same time, the output stays well below
−40 dB in its stopband. Figure 9.13 shows variation in the image impedance
 IMAGE PARAMETER METHOD 349

3.581 µH 1.31 µH 1.31 µH 3.581 µH

6.366 µH 5.9683 µH 5.9683 µH 12.94 µH 6.366 µH

2.122 nF
636.6 pF 465.78 pF 636.6 pF

Bisected Π Constant-k m-derived Bisected Π

Figure 9.11 Composite filter of Example 9.3.

0.1 0.2 0.5 1 2 5 10

0

−10
Magnitude (dB)

−20

−30

−40

−50

0.1 0.2 0.5 1 2 5 10

Frequency (MHz)

0.1 0.2 0.5 1 2 5 10

150

100

50
Phase (deg)

−50

−100

−150

0.1 0.2 0.5 1 2 5 10

Frequency (MHz)

Figure 9.12 Frequency response of the composite filter of Figure 9.11.

350 FILTER DESIGN

200

150
Magnitude (Ω)

100

50

1 2 3 4
Frequency (MHz)

75

50

25
Phase (deg)

1 2 3 4
−25
Frequency (MHz)
−50

−75

Figure 9.13 Image impedance of the composite filter of Figure 9.11 versus frequency.

of this filter as the signal frequency changes. This indicates that the image
impedance stays at 75
(pure real, because the phase angle is zero) over most
of its passband.

Example 9.4 Design a high-pass composite filter with a nominal impedance of

75
. It must pass all signals over 2 MHz. Assume that f∞ = 1.95 MHz. Plot its
characteristics in the frequency range 1 to 10 MHz.

SOLUTION From Table 9.3, we find the components of its constant-k section
as follows:
75
L= H = 2.984 µH
2 × 2 × π × 2 × 106
and
1
C= F = 530.5 pF
2 × 2 × π × 2 × 106 × 75
 IMAGE PARAMETER METHOD 351

Similarly, the component values for its m-derived filter section are determined
as follows:  
   
f∞ 2 1.95 2
m= 1− = 1− = 0.222
fc 2

Hence,
2C
= 4.775 nF
m
L
= 13.43 µH
m
and
4m
C = 0.496 nF
1 − m2
The component values for the bisected π-section to be used at its input and
output ports are found as

C
= 1.768 nF
0.3
L
= 9.947 µH
0.3
and
15
C = 0.9947 nF
8
The composite filter (after simplifying for the series capacitors in various sections)
is illustrated in Figure 9.14.
The frequency response of this composite filter is depicted in Figure 9.15. It
shows that the attenuation in its stopband stays below −40 dB, and the switching
to passband is fairly sharp. As usual with this type of circuit, its phase characteris-
tics may not be acceptable for certain applications because of inherent distortion.

0.6631 nF 0.8681 nF 1.2903 nF

9.947 µH 9.947 µH
2.984 µH 13.43 µH

0.9947 nF 0.496 nF 0.9947 nF

Figure 9.14 Composite high-pass filter with 2-MHz cutoff frequency.

352 FILTER DESIGN

1 1.5 2 3 5 7 10
0

−10

Magnitude (dB)
−20

−30

−40

−50

−60

1 1.5 2 3 5 7 10
Frequency (MHz)

1 1.5 2 3 5 7 10
150
100
50
Phase (deg)

0
−50
−100
−150

1 1.5 2 3 5 7 10
Frequency (MHz)
Figure 9.15 Frequency response of the composite high-pass filter of Figure 9.14.

250
Magnitude (Ω)

200
150
100
50

5 10 15 20
Frequency (MHz)

80
Phase (deg)

60

40

20

5 10 15 20
Frequency (MHz)
Figure 9.16 Image impedance of the composite high-pass filter of Figure 9.14 as a
function of signal frequency.
 INSERTION-LOSS METHOD 353

As depicted in Figure 9.16, the image impedance of this composite filter is

almost constant at 75
. Note that its magnitude varies over a wide range in the
stopband while its phase angle seems to remain constant at 90◦ . The phase angle
of the image impedance is zero in the passband.

9.2 INSERTION-LOSS METHOD

The output of an ideal filter would be the same as its input in the passband,
whereas it would be zero in the stopband. The phase response of this filter must
be linear to avoid signal distortion. In reality, such circuits do not exist and a com-
promise is needed to design the filters. The image parameter method described
in Section 9.1 provides a simple design procedure. However, the transfer char-
acteristics of this circuit cannot be shaped as desired. On the other hand, the
insertion-loss method provides ways to shape pass- and stopbands of the filter,
although its design theory is much more complex.
The power-loss ratio of a two-port network is defined as the ratio of the power
that is delivered to the load when it is connected directly at the generator to the
power delivered when the network is inserted between the two. In other words,

power incident at port 1 1

PLR = =
power delivered to the load connected at port 2 1 − |(ω)|2
(9.2.1)
The power-loss ratio, PLR , expressed in decibels, is generally known as the inser-
tion loss of the network. It can be proved that |(ω)|2 must be an even function of
ω for a physically realizable network. Therefore, polynomials of ω2 can represent
it as follows.
f1 (ω2 )
|(ω)|2 = (9.2.2)
f1 (ω2 ) + f2 (ω2 )
and
f1 (ω2 )
PLR = 1 + (9.2.3)
f2 (ω2 )

where f1 (ω2 ) and f2 (ω2 ) are real polynomials in ω2 .

Alternatively, the magnitude of the voltage gain of the two-port network can
be found as
1 1
|G(ω)| = √ = (9.2.4)
PLR 1 + f1 (ω2 )/f2 (ω2 )

Hence, if PLR is specified, (ω) is fixed. Therefore, the insertion-loss method is

similar to the impedance-matching methods discussed in Chapter 7.
Traditionally, the filter design begins with a lumped-element low-pass network
that is synthesized by using normalized tables. It is subsequently scaled to the
desired cutoff frequency and the impedance. Also, the low-pass prototype can
354 FILTER DESIGN

G(ω)
Passband Stopband

ω
0 ωc

Figure 9.17 Characteristics of an ideal low-pass filter.

be transformed to obtain a high-pass, a bandpass, or a bandstop filter. These

lumped-element filters are used as a starting point to design the transmission line
filter. In this section we present the design procedure for two different types of
lumped-element low-pass filters. It is followed by the transformation techniques
used to design high-pass, bandpass, and bandstop filters.

Low-Pass Filters
As illustrated in Figure 9.17, an ideal low-pass filter will pass the signals below
its cutoff frequency ωc without attenuation while it will stop those with higher
frequencies. Further, the transition from its passband to its stopband will be sharp.
In reality, this type of filter cannot be designed. Several approximations to these
characteristics are available that can be physically synthesized. Two of these are
presented below.

Maximally Flat Filter

As its name suggests, this type of filter provides the flattest possible passband
response. However, its transition from passband to stopband is gradual. This is
also known as a binomial or Butterworth filter. The magnitude of its voltage gain
(and hence, its frequency response) is given as follows:
1
|G(ω)| =  n = 1, 2, 3, . . . (9.2.5)
1 + ζω2n
where ζ is a constant that controls the power-loss ratio at its band edge, n is the
order, and ω is the normalized frequency. For |G(ω)| = 0.7071 (i.e., −3 dB) at
the band edge, ζ is unity.
Note that derivatives of |G(ω)| at normalized frequencies much below the
band edge are close to zero. This guarantees maximum flatness in the passband.
Further, (9.2.5) indicates that |G(ω)| will be a fractional number for ω greater
than zero. Therefore, it is generally known as the insertion loss, L, of the network
when expressed in decibels. Hence,
   2n
1 ω
L = −20 log10  = 10 log10 1 + ζ dB
1 + ζ (ω/ωc ) 2n ω c
(9.2.6)
 INSERTION-LOSS METHOD 355

Normalized frequency

0.2 0.4 0.6 0.8 1 1.2

−0.5
n=2
−1
n=4
G(ω)(dB)

−1.5
n=7
−2

−2.5

Figure 9.18 Characteristics of maximally flat low-pass filters.

or   2n
L ω
= log10 1 + ζ
10 ωc

or  2n
ω
ζ = 10L/10 − 1 (9.2.7)
ωc

Figure 9.18 illustrates the passband characteristics of this type of filter for
three different values of n. The power-loss ratio at its band edge is assumed to
be −3 dB, and therefore ζ is equal to unity. It shows that the passband becomes
flatter with a sharper cutoff when its order n is increased. However, this rela-
tionship is not a linear one. This change in characteristics is significant for lower
values of n.
Equation (9.2.7) can be used for determining the ζ and n values of a filter
as follows. If the insertion loss at the band edge (ω = ωc ) is specified as L =
Lc , (9.2.7) yields
ζ = 10Lc /10 − 1 (9.2.8)

Similarly, the required order n (and hence, number of elements required) of a

filter can be determined for the L specified at a given stopband frequency. It is
found as follows:

1 log10 (10L/10 − 1) − log10 ζ

n= × (9.2.9)
2 log10 (ω/ωc )

Chebyshev Filter
A filter with a sharper cutoff can be realized at the cost of flatness in its passband.
Chebyshev filters possess ripples in the passband but provide a sharp transition
356 FILTER DESIGN

into the stopband. In this case, Chebyshev polynomials are used to represent the
insertion loss. Mathematically,

1
|G(ω)| =  m = 1, 2, 3, . . . (9.2.10)
1 + ζTm2 (ω)

where ζ is a constant, ω is normalized frequency, and Tm (ω) is a Chebyshev

polynomial of the first kind and degree m, defined in (7.5.2). Figure 9.19 shows
the frequency response of a typical Chebyshev filter for m = 7. It assumes that
ripples up to −3 dB in its passband are acceptable. A comparison of this character-
istic to that for a seventh-order Butterworth filter shown in Figure 9.18 indicates
that it has a much sharper transition from passband to stopband. However, it is
achieved at the cost of ripples in its passband.
As before, the insertion loss of a Chebyshev filter is found as follows:
  
1 ω
L = −20 log10  = 10 log10 1 + ζTm2
1 + ζTm2 (ω/ωc ) ωc
or
   

 −1 ω

 10 log 1 + ζ cos2
m cos 0 ≤ ω ≤ ωc
10
ωc
L=    (9.2.11)

 ω

 10 log10 1 + ζ cosh2 m cosh−1 ωc < ω
ωc

where
ζ = 100.1×Gr − 1 (9.2.12)

Normalized frequency

0.2 0.4 0.6 0.8 1 1.2

−1
−2
−3
G(ω) (dB)

−4
−5
−6
−7

Figure 9.19 Characteristics of a low-pass Chebyshev filter for m = 7.

 INSERTION-LOSS METHOD 357

Gr is the ripple amplitude in decibels. The order m (and hence, number of ele-
ments) of a Chebyshev filter can be found from its characteristics as follows:

cosh−1 (100.1×L − 1)/(100.1×Gr − 1)
m= (9.2.13)
cosh−1 (ω/ωc )

where L is required insertion loss in decibels at a specified frequency ω.

Example 9.5 It is desired to design a maximally flat low-pass filter with at least
15 dB attenuation at ω = 1.3 ωc and −3 dB at its band edge. How many elements
will be required for this filter? If a Chebyshev filter is used with a 3-dB ripple
in its passband, find the number of circuit elements.

SOLUTION From (9.2.8),

ζ = 10Lc /10 − 1 = 100.3 − 1 = 1

and from (9.2.9) we have

1 log10 (10L/10 − 1) − log10 ζ log10 (101.5 − 1)

n= × = 0.5 × = 6.52
2 log10 (ω/ωc ) log10 (1.3)

Therefore, seven elements will be needed for this maximally flat filter.
In the case of a Chebyshev filter, (9.2.13) gives
 √
cosh−1 (100.1×L − 1)/(100.1×Gr − 1) cosh−1 101.5 − 1
m= = = 3.17
cosh−1 (ω/ωc ) cosh−1 (1.3)

Hence, it will require only three elements. The characteristics of these two filters
are illustrated in Figure 9.20.

Normalized frequency

0.5 1 1.5 2
−1
−2
−3
G(ω) (dB)

−4
−5
−6
−7

Figure 9.20 Characteristics of low-pass Butterworth and Chebyshev filters.

358 FILTER DESIGN

Low-Pass Filter Synthesis

The scope of this book does not include the theory of passive filter synthesis.
Several excellent references are available in the literature for those who may
be interested in this. The design procedure presented here is based on a doubly
terminated low-pass ladder network, as shown in Figure 9.21. The g notation used
in the figure signifies the roots of an nth-order transfer function that govern its
characteristics. These represent the normalized reactance values of filter elements
with a cutoff frequency ωc = 1. The source resistance is represented by g0 , and
the load is gn+1 . The filter is made up of series inductors and shunt capacitors
that are in the form of cascaded T-networks. Another possible configuration is
the cascaded π-network that is obtained after replacing g1 by a short circuit,
connecting a capacitor across the load gn+1 , and renumbering filter elements
1 through n. Elements of this filter are determined from the n roots of the
transfer function.
The transfer function is selected according to the pass- and stopband charac-
teristics desired. Normalized values of elements are then found from the roots
of that transfer function. These values are then adjusted according to the desired
cutoff frequency and the source and load resistance. Design procedures for the
maximally flat and the Chebyshev filters can be summarized as follows.
Assume that the cutoff frequency is given as follows:

ωc = 1 (9.2.14)

Butterworth and Chebyshev filters can then be designed using the following
formulas.
For a Butterworth filter,

g0 = gn+1 = 1 (9.2.15)
and
(2p − 1)π
gp = 2 sin p = 1, 2, · · · , n (9.2.16)
2n

Element values computed from (9.2.15) and (9.2.16) are given in Table 9.4 for
n up to 7.

g1 g3 g5 gn−2 gn

g2 g4 gn−1
gn+1
g0

Figure 9.21 Low-pass ladder network prototype.

 INSERTION-LOSS METHOD 359

TABLE 9.4 Element Values for Low-Pass Binomial Filter Prototypes (g0 = 1,
ωc = 1)

n g1 g2 g3 g4 g5 g6 g7 g8

1 2.0000 1.0000
2 1.4142 1.4142 1
3 1.0000 2.0000 1.0000 1.0000
4 0.7654 1.8478 1.8478 0.7654 1.0000
5 0.6180 1.6180 2.0000 1.6180 0.6180 1
6 0.5176 1.4142 1.9319 1.9319 1.4142 0.5176 1
7 0.4450 1.2470 1.8019 2.0000 1.8019 1.2470 0.445 1.0000

For a Chebyshev filter,

g0 = 1 (9.2.17)

1 m is an odd number
gm+1 = ξ (9.2.18)
 coth m is an even number
4
2a1
g1 = (9.2.19)
χ
and
4a(p−1) ap
gp = p = 2, 3, · · · , m (9.2.20)
b(p−1) g(p−1)

where
 
Gr
ξ = ln coth (9.2.21)
17.37
ξ
χ = sinh (9.2.22)
2m
(2p − 1)π
ap = sin (9.2.23)
2m
and
pπ
bp = χ2 + sin2 (9.2.24)
m

The element values for several low-pass Chebyshev filters that are computed
from (9.2.17) to (9.2.24) are given in Tables 9.5 to 9.9.
360 FILTER DESIGN

TABLE 9.5 Element Values for Low-Pass Chebyshev Filter Prototypes (g0 = 1,
ωc = 1, and 0.1 dB ripple)

m g1 g2 g3 g4 g5 g6 g7 g8

1 0.3053 1.000
2 0.8431 0.6220 1.3554
3 1.0316 1.1474 1.0316 1.0000
4 1.1088 1.3062 1.7704 0.8181 1.3554
5 1.1468 1.3712 1.9750 1.3712 1.1468 1.0000
6 1.1681 1.4040 2.0562 1.5171 1.9029 0.8618 1.3554
7 1.1812 1.4228 2.0967 1.5734 2.0967 1.4228 1.1812 1.0000

TABLE 9.6 Element Values for Low-Pass Chebyshev Filter Prototypes (g0 = 1,
ωc = 1, and 0.5 dB ripple)

m g1 g2 g3 g4 g5 g6 g7 g8

1 0.6987 1.0000
2 1.4029 0.7071 1.9841
3 1.5963 1.0967 1.5963 1.0000
4 1.6703 1.1926 2.3662 0.8419 1.9841
5 1.7058 1.2296 2.5409 1.2296 1.7058 1.0000
6 1.7254 1.2479 2.6064 1.3136 2.4759 0.8696 1.9841
7 1.7373 1.2582 2.6383 1.3443 2.6383 1.2582 1.7373 1.0000

TABLE 9.7 Element Values for Low-Pass Chebyshev Filter Prototypes (g0 = 1,
ωc = 1, and 1.0 dB ripple)

m g1 g2 g3 g4 g5 g6 g7 g8

1 1.0178 1.0000
2 1.8220 0.6850 2.6599
3 2.0237 0.9941 2.0237 1.0000
4 2.0991 1.0644 2.8312 0.7892 2.6599
5 2.1350 1.0911 3.0010 1.0911 2.1350 1.0000
6 2.1547 1.1041 3.0635 1.1518 2.9368 0.8101 2.6599
7 2.1666 1.1115 3.0937 1.1735 3.0937 1.1115 2.1666 1.0000

Scaling the Prototype to the Desired Cutoff Frequency and Load

ž Frequency scaling. For scaling the frequency from 1 to ωc , divide all nor-
malized g values that represent capacitors or inductors by the desired cutoff
frequency expressed in radians per second. Resistors are excluded from
this operation.
 INSERTION-LOSS METHOD 361

TABLE 9.8 Element Values for Low-Pass Chebyshev Filter Prototypes (g0 = 1,
ωc = 1, and 2.0 dB ripple)

m g1 g2 g3 g4 g5 g6 g7 g8

1 1.5297 1.0000
2 2.4883 0.6075 4.0957
3 2.7108 0.8326 2.7108 1.0000
4 2.7926 0.8805 3.6064 0.6818 4.0957
5 2.8311 0.8984 3.7829 0.8984 2.8311 1.0000
6 2.8522 0.9071 3.8468 0.9392 3.7153 0.6964 4.0957
7 2.8651 0.9120 3.8776 0.9536 3.8776 0.9120 2.8651 1.0000

TABLE 9.9 Element Values for Low-Pass Chebyshev Filter Prototypes (g0 = 1,
ωc = 1, and 3.0 dB ripple)

m g1 g2 g3 g4 g5 g6 g7 g8

1 1.9954 1.0000
2 3.1014 0.5339 5.8095
3 3.3489 0.7117 3.3489 1.0000
4 3.4391 0.7483 4.3473 0.5920 5.8095
5 3.4815 0.7619 4.5378 0.7619 3.4815 1.0000
6 3.5047 0.7685 4.6063 0.7929 4.4643 0.6033 5.8095
7 3.5187 0.7722 4.6392 0.8038 4.6392 0.7722 3.5187 1.0000

ž Impedance scaling. To scale g0 and gn+1 to X
from unity, multiply all

g values that represent resistors or inductors by X. On other hand, divide
those g values representing capacitors by X.

Example 9.6 Design a Butterworth filter with a cutoff frequency of 10 MHz

and an insertion loss of 30 dB at 40 MHz. It is to be used between a 50-
load
and a generator with an internal resistance of 50
.

SOLUTION From (9.2.9), we have

1 log10 (1030/10 − 1) − log10 (103/10 − 1)

n= ×
2 log10 (40/10)
1 log10 (103 − 1) − log10 (1.9953 − 1) 0.5 × 3
= × ≈ ≈ 2.5
2 log10 (4) 0.6

Since the number of elements must be an integer, selecting n = 3 will provide

more than the attenuation specified at 40 MHz. From (9.2.15) and (9.2.16),

g0 = g4 = 1
362 FILTER DESIGN

795.77 nH 795.77 nH
50 Ω L1 L3
R0 = Rs R4 = RL
636.62 pF C2
50 Ω

Figure 9.22 Maximally flat low-pass filter obtained in Example 9.6.

π
g1 = 2 sin =1
6
π
g2 = 2 sin = 2
2
and
5π
g3 = 2 sin =1
6

If we use the circuit arrangement illustrated in Figure 9.22, element values can
be scaled to match the frequency and load resistance. Using the two rules, these
values are found as follows:
1
L1 = L3 = 50 × H = 795.77 nH
2π × 107
and
1 2
C2 = × F = 636.62 pF
50 2π × 107

The frequency response of this filter is shown in Figure 9.23. It indicates a 6-

dB insertion loss in the passband of the filter. This happens because the source
resistance is considered a part of this circuit. In other words, this represents
voltage across RL with respect to source voltage, not to input of the circuit. Since
source resistance is equal to the load, there is equal division of source voltage
that results in −6 dB. Another 3-dB loss at the band edge shows a total of about
9 dB at 10 MHz. This characteristic shows that there is an insertion loss of over
40 dB at 40 MHz. It is clearly more than 30 dB from the band edge, as required
by the example. Also, there is a nonlinear phase variation in its passband.
Alternatively, we can use a circuit topology as shown in Figure 9.24. In that
case, component values are found by using the scaling rules:

1
L2 = 50 × × 2 = 0.15915 × 10−5 H = 1591.5 nH
2π × 107
and
1 1 10−9
C1 = C3 = × ×1= F = 318.31 pF
50 2π × 107 π
 INSERTION-LOSS METHOD 363

0.1 0.5 1 5 10 50 100

0

−10

−20
Magnitude (dB)

−30

−40

−50

−60

0.1 0.5 1 5 10 50 100

Frequency (MHz)

0.1 0.5 1 5 10 50 100

150

100

50
Phase (deg)

−50

−100

−150
0.1 0.5 1 5 10 50 100
Frequency (MHz)

Figure 9.23 Characteristics of the low-pass filter shown in Figure 9.22.

1.5915 µH
50 Ω L2

R1 = Rs R4 = RL
C1 C3
318.31 pF 318.31 pF 50 Ω

Figure 9.24 Alternative circuit for Example 9.6.

The frequency response of this circuit is illustrated in Figure 9.25. It is identical

to that shown in Figure 9.23 for the earlier circuit.

Example 9.7 Design a low-pass Chebyshev filter that may have ripples no more
than 0.01 dB in its passband. The filter must pass all frequencies up to 100 MHz
364 FILTER DESIGN

0.1 0.5 1 5 10 50 100

0

−10

−20
Magnitude (dB)

−30

−40

−50

−60

0.1 0.5 1 5 10 50 100

Frequency (MHz)

0.1 0.5 1 5 10 50 100

150

100

50
Phase (deg)

−50

−100

−150
0.1 0.5 1 5 10 50 100
Frequency (MHz)

Figure 9.25 Characteristics of the low-pass filter shown in Figure 9.24.

and attenuate the signal at 400 MHz by at least 5 dB. The load and the source
resistance are of 75
each.

SOLUTION Since
Gr = 0.01 dB

and  
400
L = 5 dB
100
 INSERTION-LOSS METHOD 365

(9.2.13) gives

cosh−1 (100.5 − 1)/(100.001 − 1)
m= =2
cosh−1 (4)

Since we want a symmetrical filter with 75
on each side, we select m = 3

(an odd number). It will provide more than 5 dB of insertion loss at 400 MHz.
The g values can now be determined from (9.2.17) to (9.2.24), as follows. From
(9.2.23),
π
a1 = sin = 0.5
6
π
a2 = sin = 1
2
and
5π
a3 = sin = 0.5
6
Next, from (9.2.21), (9.2.22), and (9.2.24), we get
 
0.01
ξ = ln coth = 7.5
17.37
7.5
χ = sinh = 1.6019
6
π
b1 = 1.60192 + sin2 = 3.316
3
2π
b2 = 1.60192 + sin2 = 3.316
3
and
3π
b3 = 1.60192 + sin2 = 2.566
3
However, b3 is not needed in further calculations.
Finally, from (9.2.17) and (9.2.20),

g0 = g4 = 1
2 × 0.5
g1 = = 0.62425
1.6019
4 × 0.5 × 1
g2 = = 0.9662
3.316 × 0.62425
and
4 × 1 × 0.5
g3 = = 0.62425
3.316 × 0.9662
366 FILTER DESIGN

74.5155 nH 74.5155 nH
75 Ω L1 L3
R0 = R s R4 = R L
20.5 pF C2
75 Ω

Figure 9.26 Low-pass Chebyshev filter circuit obtained in Example 9.7.

For the circuit topology of Figure 9.26, element values are found after applying
the scaling rules as follows:

75 × 0.62425
L1 = L3 = H = 74.5155 nH
2π × 108
and
1 1
C2 = × × 0.9662 F = 20.5 pF
75 2π × 108

The frequency response of this filter is illustrated in Figure 9.27. As expected,

there is over a 20-dB insertion loss at 400 MHz compared with its passband.
Note that the magnitude of ripple allowed is so small that it does not show up
in this figure. However, it is present there, as shown on an expanded scale in
Figure 9.28.
Figure 9.28 shows the passband characteristics of this Chebyshev filter. Since
the scale-of-magnitude plot is now expanded, the passband ripple is clearly visi-
ble. As expected, the ripple stays between −6.02 and −6.03 dB. Phase variation
in this passband ranges from 0◦ to −70◦ . The band edge of this filter can be
sharpened further either by using a higher-order filter or by allowing larger mag-
nitudes of the ripple in its passband. The higher-order filter will require more
elements because the two are directly related. The next example illustrates that a
sharper transition between the bands can be obtained even with a three-element
filter if ripple magnitudes up to 3 dB are acceptable.

Example 9.8 Reconsider Example 9.7 to design a low-pass filter that exhibits
the Chebyshev response with 3-dB ripple in its passband, m = 3, and a cutoff
frequency of 100 MHz. The filter must have 75
at both its input and output
ports.

SOLUTION From (8.2.23),

π
a1 = sin = 0.5
6
π
a2 = sin = 1
2
 INSERTION-LOSS METHOD 367

1 5 10 50 100 500 1000

0

−10
Magnitude (dB)

−20

−30

−40

−50
1 5 10 50 100 500 1000
Frequency (MHz)

1 5 10 50 100 500 1000

150

100

50
Phase (deg)

−50

−100

−150

1 5 10 50 100 500 1000

Frequency (MHz)

Figure 9.27 Characteristics of the low-pass Chebyshev filter shown in Figure 9.26.

and
5π
a3 = sin = 0.5
6

Now, from (9.2.21), (9.2.22), and (9.2.24),

 
3
ξ = ln coth = 1.7661
17.37
1.7661
χ = sinh = 0.2986
6
368 FILTER DESIGN

1 2 5 10 20 50 100

−6.022
Magnitude (dB)

−6.024

−6.026

−6.028

−6.03
1 2 5 10 20 50 100
Frequency (MHz)

1 2 5 10 20 50 100
0

−10

−20
Phase (deg)

−30

−40

−50

−60

1 2 5 10 20 50 100
Frequency (MHz)

Figure 9.28 Passband characteristics of the filter shown in Figure 9.26.

π
b1 = 0.29862 + sin2 = 0.8392
3
2π
b2 = 0.29862 + sin2 = 0.8392
3
and
3π
b3 = 0.29862 + sin2 = 0.0892
3

However, b3 is not needed.

 INSERTION-LOSS METHOD 369

0.4 µH 0.4 µH
75 Ω L1 L3
R0 = Rs R4 = RL
15.1 pF C2
75 Ω

Figure 9.29 Low-pass Chebyshev filter circuit obtained in Example 9.8.

The g values for the filter can now be determined from (9.2.17) to (9.2.20),
as follows:

g0 = g4 = 1
2 × 0.5
g1 = = 3.349
0.2986
4 × 0.5 × 1
g2 = = 0.7116
3.349 × 0.8392
and
4 × 1 × 0.5
g3 = = 3.349
0.7116 × 0.8392

These g values are also listed in Table 9.9 for m = 3.

Elements of the circuit shown in Figure 9.29 are determined using the scaling
rules, as follows (see Figure 9.30 for the characteristics):

75 × 3.349
L1 = L3 = H = 0.4 µH
2π × 108

and
1 1
C2 = × × 0.7116 F = 15.1 pF
75 2π × 108

High-Pass Filter
As mentioned earlier, a high-pass filter can be designed by transforming the
low-pass prototype. This frequency transformation is illustrated in Figure 9.31.
As illustrated, an ideal low-pass filter passes all signals up to the normalized
frequency of unity with zero insertion loss, whereas it attenuates higher frequen-
cies completely. On the other hand, a high-pass filter must pass all signals with
frequencies higher than its cutoff frequency ωc and stop the signals that have
lower frequencies. Therefore, the following frequency transformation formula
will transform a low-pass filter to a high-pass filter:
ωc
ω=− (9.2.25)
ω
370 FILTER DESIGN

1 10 100
−6

−7

−8
Magnitude (dB)

−9

−10

−11

−12

−13

1 10 100
Frequency (MHz)

1 10 100

100

50
Phase (deg)

−50

−100

−150
1 10 100
Frequency (MHz)

Figure 9.30 Characteristics of the low-pass filter of Example 9.8.

Magnitude

Magnitude

1 1

ω ω
0 1 0 ωc

Figure 9.31 Transformation from a low-pass to a high-pass characteristic.

 INSERTION-LOSS METHOD 371

Thus, inductors and capacitors will change their places. Inductors will replace
the shunt capacitors of the low-pass filter and capacitors will be connected in
series, in place of inductors. These elements are determined as follows:

1
CHP = (9.2.26)
ωc gL
and
1
LHP = (9.2.27)
ωc gC

Capacitor CHP and inductor LHP can now be scaled as required by the load and
source resistance.

Example 9.9 Design a high-pass Chebyshev filter with passband ripple magni-
tude less than 0.01 dB. It must pass all frequencies over 100 MHz and exhibit at
least 5 dB of attenuation at 25 MHz. Assume that the load and source resistances
are at 75
each.

SOLUTION The low-pass filter designed in Example 9.7 provides the initial
data for this high-pass filter. With m = 3, gL = 0.62425, and gC = 0.9662, we
find from (9.2.26) and (9.2.27) that

1
CHP = F = 2.5495 nF
2π × 108 × 0.62425
and
1
LHP = H = 1.6472 nH
2π × 108× 0.9662
Now, applying the resistance scaling, we get

2.5495
C1 = C3 = nF = 33.9933 pF = 34 pF
75
and
L2 = 75 × 1.6472 = 123.5 nH

The resulting high-pass Chebyshev filter is shown in Figure 9.32. Its frequency
response is illustrated in Figure 9.33. As before, the source resistance is con-
sidered a part of the filter circuit and therefore its passband shows a 6-dB
insertion loss.

Bandpass Filter
A bandpass filter can be designed by transforming the low-pass prototype as
illustrated in Figure 9.34. Here, an ideal low-pass filter passes all signals up to
372 FILTER DESIGN

75 Ω 34 pF 34 pF

R0 = Rs R4 = RL
C1 C3
L2 123.5 nH 75 Ω

Figure 9.32 High-pass Chebyshev filter obtained for Example 9.8.

20 50 100 200 500 1000 2000

−7.5

−10
Magnitude (dB)

−12.5

−15

−17.5

−20

−22.5
20 50 100 200 500 1000 2000
Frequency (MHz)

20 50 100 200 500 1000 2000

150

100

50
Phase (deg)

−50

−100

−150

20 50 100 200 500 1000 2000

Frequency (MHz)

Figure 9.33 Characteristics of the high-pass Chebyshev filter shown in Figure 9.32.
 INSERTION-LOSS METHOD 373

Magnitude

Magnitude
1 1

ω ω
−1 0 1 0 ωl ωu

Figure 9.34 Transformation from a low-pass to a bandpass characteristic.

the normalized frequency of unity with zero insertion loss, whereas it attenuates
higher frequencies completely. On the other hand, a bandpass filter must pass
all signals with frequencies between ωl and ωu and stop the signals that are
outside this frequency band. Hence, the following frequency relation transforms
the response of a low-pass filter to a bandpass:

1 ω2 − ω20
ω= (9.2.28)
ωu − ωl ω
where √
ω0 = ωl × ωu (9.2.29)

This transformation replaces the series inductor of low-pass prototype with an

inductor LBP1 and a capacitor CBP1 that are connected in series. The component
values are determined as follows:
ωu − ωl
CBP1 = (9.2.30)
ω20 gL
and
gL
LBP1 = (9.2.31)
ωu − ωl
Also, the capacitor CBP2 that is connected in parallel with an inductor LBP2 will
replace the shunt capacitor of the low-pass prototype. These elements are found
as follows:
ωu − ωl
LBP2 = (9.2.32)
ω20 gC
and
gC
CBP2 = (9.2.33)
ωu − ωl
These elements need to be scaled further as desired by the load and source
resistance.
374 FILTER DESIGN

Example 9.10 Design a bandpass Chebyshev filter that exhibits no more than
0.01-dB ripples in its passband. It must pass signals in the frequency band 10
to 40 MHz with zero insertion loss. Assume that the load and source resistances
are at 75
each.

SOLUTION From (9.2.29), we have

 
f0 = fl fu = 107 × 40 × 106 = 20 × 106 Hz

Now, using (9.2.30) to (9.2.33) with gL = 0.62425 and gC = 0.9662 from

Example 9.7, we get

2π × 106 (40 − 10)

CBP1 = F = 19.122 nF
(2π × 20 × 106 )2 × 0.62424
0.62424
LBP1 = H = 3.3116 nH
2π × 30 × 106
2π × 106 (40 − 10)
LBP2 = H = 12.354 nH
(2π × 20 × 106 )2 × 0.9662
and
0.9662
CBP2 = F = 5.1258 nF
2π × 30 × 106

Next, values of elements are determined after applying the load and source resis-
tance scaling. Hence,

19.122
C1 = C3 = nF = 254.96 pF ≈ 255 pF
75
L1 = L3 = 75 × 3.3116 nH = 0.2484 µH
L2 = 75 × 12.354 nH = 0.9266 µH
and
5.1258
C2 = nF = 68.344 pF
75

The resulting filter circuit is shown in Figure 9.35, and its frequency response
is depicted in Figure 9.36. It illustrates that we indeed have transformed the
low-pass prototype to a bandpass filter for the 10- to 40-MHz frequency band.
As before, the source resistance is considered a part of the filter circuit, and
therefore its passband shows a 6-dB insertion loss. Figure 9.37 shows its passband
characteristics with expanded scales. It indicates that the ripple magnitude is
limited to 0.01 dB, as desired.
 INSERTION-LOSS METHOD 375

0.2484 µH 0.2484 µH
75 Ω 255 pF 255 pF

C1 L1 L3 C3

0.9266 µH 68.344 pF 75 Ω
L2
C2

Figure 9.35 Bandpass filter circuit for Example 9.10.

0.1 1 10 100 1000

0

−20
Magnitude (dB)

−40

−60

−80

−100

−120
0.1 1 10 100 1000
Frequency (MHz)

0.1 1 10 100 1000

150

100
Phase (deg)

50

0
−50

−100

−150
0.1 1 10 100 1000
Frequency (MHz)

Figure 9.36 Characteristics of the bandpass filter shown in Figure 9.35.

Bandstop Filter
A bandstop filter can be realized by transforming the low-pass prototype as
illustrated in Figure 9.38. Here, an ideal low-pass filter passes all signals up to
the normalized frequency of unity with zero insertion loss, whereas it completely
376 FILTER DESIGN

10 15 20 30

Magnitude (dB) −6.022

−6.024

−6.026

−6.028

−6.03
10 15 20 30
Frequency (MHz)

10 15 20 30
60

40

20
Phase (deg)

−20

−40

−60

10 15 20 30
Frequency (MHz)

Figure 9.37 Passband characteristics of the bandpass filter shown in Figure 9.35.

attenuates higher frequencies. On the other hand, a bandstop filter must stop all
signals with frequencies between ωl and ωu and pass the signals that are outside
this frequency band. Hence, its characteristics are opposite to that of a bandpass
filter considered earlier. The following frequency relation transforms the response
of a low-pass filter to the bandstop:
ω
ω = (ωu − ωl ) (9.2.34)
ω2 − ω20

This transformation replaces the series inductor of low-pass prototype with an

inductor LBS1 and a capacitor CBS1 that are connected in parallel. Component
values are determined as follows:
(ωu − ωl )gL
LBS1 = (9.2.35)
ω20
 INSERTION-LOSS METHOD 377

Magnitude

1
1

Magnitude
ω ω
−1 0 1 0 ωl ωu

Figure 9.38 Transformation from a low-pass to a bandstop characteristic.

and
1
CBS1 = (9.2.36)
(ωu − ωl )gL

Also, capacitor CBS2 , which is connected in series with an inductor LBS2 , will
replace the shunt capacitor of low-pass prototype. These elements are found
as follows:
1
LBS2 = (9.2.37)
(ωu − ωl )gC
and
(ωu − ωl )gC
CBS2 = (9.2.38)
ω20

These elements need to be further scaled as desired by the load and source
resistance. Table 9.10 summarizes these transformations for the low-pass proto-
type filter.

Example 9.11 Design a maximally flat bandstop filter with n = 3. It must stop
signals in the frequency range 10 to 40 MHz and pass the rest of the frequencies.
Assume that the load and source resistances are at 75
each.

SOLUTION From (9.2.29), we have

 
f0 = fl fu = 107 × 40 × 106 = 20 × 106 Hz

Now, using (9.2.35) to (9.2.38) with gL = 1 and gC = 2 from Example 9.6,

we get

2π × 106 (40 − 10)

LBS1 = × 1 H = 11.94 nH
(2π × 20 × 106 )2
1
CBS1 = F = 5.305 nF
2π × 10 (40 − 10) × 1
6
378 FILTER DESIGN

TABLE 9.10 Filter Transformations

Filter Circuit Elements

Low-pass

gL gC

High-pass

1 1
ω c gL ωc gC

Bandpass ωu − ω l
ω20 gC
ωu − ω l gL
ω20 gL ω u − ωl

gC
ωu − ωl

Bandstop ωu − ωl gL 1
ωu − ω l gC
ω20

ωu − ω l gC
ω20

1
ωu − ωl gL

1
LBS2 = H = 2.653 nH
2π × 106 (40 − 10) × 2
and
2π × 106 (40 − 10)
CBS2 = × 2 F = 23.87 nF
(2π × 20 × 106 )2

Values of required elements are determined next following the load and source
resistance scaling. Hence,

5.305
C1 = C3 = nF = 70.73 pF
75
 INSERTION-LOSS METHOD 379

0.8955 µH 0.8955 µH

75 Ω
L1 L3

C1 C3
70.73 pF C2 70.73 pF
318.3 pF
75 Ω

L2
0.2 µH

Figure 9.39 Bandstop circuit for Example 9.11.

0.1 1 10 100 1000

−6

−6.5
Magnitude (dB)

−7

−7.5

−8

0.1 1 10 100 1000

Frequency (rad/s)

0.1 1 10 100 1000

100

50
Phase (deg)

−50

−100

−150

0.1 1 10 100 1000

Frequency (rad/s)

Figure 9.40 Characteristics of the bandstop filter shown in Figure 9.39.

380 FILTER DESIGN

L1 = L3 = 75 × 11.94 nH = 0.8955 µH
L2 = 75 × 2.653 nH = 0.1989 µH ≈ 0.2 µH
and
23.87
C2 = nF = 318.3 pF
75

The resulting filter circuit is shown in Figure 9.39. Its frequency response depicted
in Figure 9.40 indicates that we have transformed the low-pass prototype to a
bandstop filter for the 10- to 40-MHz frequency band. As before, the source
resistance is considered a part of the filter circuit, and therefore its passband
shows a 6-dB insertion loss.

9.3 MICROWAVE FILTERS

The filter circuits presented so far in this chapter use lumped elements. How-
ever, these may have practical limitations at microwave frequencies. When the
signal wavelength is short, distances between the filter components need to be
taken into account. Further, discrete components at such frequencies may cease
to operate due to associated parasitic elements and need to be approximated with
distributed components. As found in Chapter 3, transmission line stubs can be
used in place of lumped elements. However, there may be certain practical prob-
lems in implementing the series stubs. This section begins with a technique to
design a low-pass filter with only parallel connected lines of different character-
istic impedance values. It is known as the stepped impedance (or high-Z low-Z)
filter. Since this technique works mainly for low-pass filters, the procedure to
transform series reactance to shunt that utilizes Kuroda’s identities is summa-
rized next. Redundant sections of the transmission line are used to separate filter
elements, and therefore this procedure is known as redundant filter synthesis.
Nonredundant circuit synthesis makes use of these sections to improve the filter
response as well. Other methods of designing microwave filters include coupled
transmission lines and resonant cavities. Impedance transformers, discussed in
Chapter 7, are essentially bandpass filters with different impedance at the two-
ports. Interested readers can find detailed design procedures of all these filters in
the references at the end of this chapter.

Low-Pass Filter Design Using Stepped Impedance Distributed Elements

Consider a lossless transmission line of length d and characteristic impedance
Z0 , as shown in Figure 9.41(a). If it is equivalent to the T-network shown in
Figure 9.41(b), these two must have the same network parameters. The imped-
ance matrix of the transmission line was determined in Example 8.4 as follows:
 MICROWAVE FILTERS 381

Z0, β
(a)

Z11 − Z12 Z11 − Z12

Z12

(b)

jXL jXL
2 2
jB

(c)

Figure 9.41 Lossless transmission line (a), equivalent symmetrical T-network (b), and
(c) elements of T-networks.

 Z0 
−j Z0 cot βd −j
 sin βd 
[Z]line =  Z0  (9.3.1)
−j −j Z0 cot βd
sin βd
The impedance matrix of the T-network, [Z]T , may be found as
 
Z11 Z12
[Z]T = (9.3.2)
Z12 Z11

Therefore, if this T-network represents the transmission line, we can write

Z11 = −j Z0 cot βd (9.3.3)

and
Z0
Z12 = −j (9.3.4)
sin βd
Hence, Z12 is a capacitive reactance. Circuit elements in the series arm of the
T-network are found to be inductive reactance. This is found as follows:
cos βd − 1 sin (βd/2) βd
Z11 − Z12 = −j Z0 = j Z0 = j Z0 tan (9.3.5)
sin βd cos (βd/2) 2
382 FILTER DESIGN

Therefore, the series element of the T-network is an inductive reactance while

its shunt element is capacitive, provided that βd < π/2. For βd < π/4, (9.3.4)
and (9.3.5) give

βd
B≈ (9.3.6)
Z0
and
XL ≈ Z0 βd (9.3.7)

Some special cases are as follows:

1. For Z0 very large, (9.3.6) becomes negligible compared with (9.3.7). There-
fore, the equivalent T-network (and hence the transmission line) represents
an inductor that is given by

Z0 βd
L= (9.3.8)
ω

2. For Z0 very small, (9.3.6) dominates over (9.3.7). Therefore, the transmis-
sion line is effectively representing a shunt capacitance in this case. It is
given by
βd
C= (9.3.9)
ωZ0

Hence, high-impedance sections (Z0 = Zh ) of the transmission line can replace

the inductors, and low-impedance sections (Z0 = Zm ) can replace the capacitors
of a low-pass filter. In a design, these impedance values must be selected as
far apart as possible and the section lengths are determined to satisfy (9.3.8)
and (9.3.9). When combined with the scaling rules, electrical lengths of the
inductive and capacitive sections are found as

R0 L
(βd)L = rad (9.3.10)
Zh
and
Zm C
(βd)C = rad (8.3.11)
R0

where L and C are normalized element values (g values) of the filter and R0 is
the filter impedance.

Example 9.12 Design a three-element maximally flat low-pass filter with its
cutoff frequency as 1 GHz. It is to be used between a 50-
load and a generator
with its internal impedance at 50
. Assume that Zh = 150
and Zm = 30
.
 MICROWAVE FILTERS 383

L1 = 1 L3 = 1
50 Ω

R0 = 1 R4 = 1
C2 = 2
50 Ω

Figure 9.42 Maximally flat low-pass filter with normalized elements’ values for Exam-
ple 9.12.

68.75°

19.1° 19.1°

Figure 9.43 Maximally flat low-pass filter for Example 9.12.

SOLUTION From (9.2.15) and (9.2.16),

g0 = g4 = 1
π
g1 = 2 sin = 1
6
π
g2 = 2 sin = 2
2
and
5π
g3 = 2 sin =1
6
If we use a circuit arrangement as illustrated in Figure 9.42, element values can
be scaled for the frequency and the impedance. Using the scaling rules, these
values are found as follows (see Figure 9.43):

1 × 50 ◦
θ1 = θ3 = (βd)L = = 0.3333 rad = 19.1
150
and
30 × 2 ◦
θ2 = (βd)C = = 1.2 rad = 68.75
50

Filter Synthesis Using the Redundant Elements

As mentioned earlier, transmission line sections can replace the lumped elements
of a filter circuit. Richard’s transformation provides the tools needed in such a
384 FILTER DESIGN

design. Further, Kuroda’s identities are used to transform series elements to a

shunt configuration that facilitates the design.

Richard’s Transformation
Richard proposed that open- and short-circuited lines could be synthesized like
lumped elements through the following transformation:

ωd

= tan βd = tan (9.3.12)
vp

where vp is the phase velocity of signal propagating on the line. Since the tangent
function is periodic with a period of 2π, (9.3.12) is a periodic transformation.
Substituting
in place of ω, the reactance of the inductor L and of the capacitor
C may be written as follows:

j XL = j
L = j L tan βd (9.3.13)
and
1 1 1
j XC = −j = −j = −j cot βd (9.3.14)

C C tan βd C

Comparison of (9.3.13) and (9.3.14) with the special cases considered in Sec-
tion 3.2 indicates that the former represents a short-circuited line with its char-
acteristic impedance as L, while the latter is an open-circuited with Z0 as 1/C.
Filter impedance is assumed to be unity. To obtain the cutoff of a low-pass filter
prototype at unity frequency, the following must hold:

π λ

= tan βd = 1 → βd = →d= (9.3.15)
4 8

Note that the transformation holds only at the cutoff frequency ωc (frequency
corresponding to λ), and therefore the filter response will differ from that of
its prototype. Further, the response repeats every 4 ωc . These transformations
are illustrated in Figure 9.44. These stubs are known as the commensurate lines
because of their equal lengths.

Kuroda’s Identities
As mentioned earlier, Kuroda’s identities facilitate the design of distributed ele-
ment filters, providing the means to transform the series stubs into shunt, or vice
versa, to separate the stubs physically, and to render the characteristic impedance
realizable. Figure 9.45 illustrates four identities that are useful in such a trans-
formation of networks. These networks employ the unit element (U.E.), which is
basically a λ/8-long line at the cutoff frequency ωc , with specified characteristic
 MICROWAVE FILTERS 385

ω-plane Ω-plane

L Z0 = L Short

1
C Z0 = Open
C

Figure 9.44 Distributed inductor and capacitor obtained from Richard’s transformation.

Z1

U.E. U.E. 1
Z2 n2Z1 n2Z2

(a)

Z1
n2

1 U.E.
U.E.
Z2 Z1 Z2
n2
(b)
1
1
n2Z2
Z2
U.E.
U.E.
n2Z1
Z1
n2 :1
(c)

U.E.
U.E. Z1
Z1 Z2
Z1 n2
n2
1:n2
(d) Z2
n2 = 1 +
Z1

Figure 9.45 Kuroda’s identities.

impedance. Inductors represent short-circuited stubs and capacitors open-circuited

stubs. Obviously, its equivalent in lumped circuit theory does not exist.
The first identity is proved below. A similar procedure can be used to verify
others. The circuits associated with the first identity can be redrawn as shown in
Figure 9.46. A short-circuited stub of characteristic impedance Z1 replaces the
386 FILTER DESIGN

Short

d d
Open

Z1 d
n2Z2
Z2 n2Z1

Figure 9.46 Illustration of Kuroda’s first identity, shown in Figure 9.45.

series inductor that is followed by a unit element transmission line of charac-

teristic impedance Z2 . Its transformation gives a unit element transmission line
of characteristic impedance n2 Z1 and an open-circuited stub of characteristic
impedance n2 Z2 in place of a shunt capacitor.
Following Example 8.10, the transmission matrix of a short-circuited series
stub with input impedance ZS can be found as follows. The impedance ZS of
the series stub is given as

ZS = j Z1 tan βd = j
Z1 (9.3.16)

   
A B 1 j
Z1
= (9.3.17)
C D S 0 1

Following Example 8.13, the transmission matrix of the unit element is found
to be
   
  cos βd j Z2 sin βd 1 j Z2 tan βd
A B
=  j sin βd  = cos βd 
 j tan βd


C D U.E. cos βd 1
Z2 Z2
or
 
  1 j
Z2
A B 1  j

=√ (9.3.18)
C D U.E. 1 +
2 1
Z2

Since these two elements are connected in cascade, the transmission matrix for
the first circuit in Figure 9.46 is found from (9.3.17) and (9.3.18) as follows:
 

2 Z1
      1− j
(Z1 + Z2 )
A B A B A B 1  Z2 
= · = √  
C D C D C D 1+
2  j

a S U.E.
1
Z2
(9.3.19)
 MICROWAVE FILTERS 387

Now for the second circuit shown in Figure 9.46, the transmission matrix for the
unit element is
   
  1 j n2 Z1 tan βd 1 j
n2 Z1
A B   1  
= cos βd  j tan βd = √  j

C D U.E. 1 1 +
2 1
n2 Z 1 n2 Z 1
(9.3.20)

Since the impedance ZP of the shunt stub is given as

n2 Z2
Zp = −j n2 Z2 cot βd = −j (9.3.21)


its transmission matrix may be found, by following Example 8.11, as
 
  1 0
A B
=  j
 (9.3.22)
C D p 1
n2 Z2
Hence, the transmission matrix of the circuit shown on right-hand side is found as
 

2 Z1
       1− j
n 2
Z 1
A B A B A B 1  Z2 
= · = √    
C D b
C D U.E.
C D p 1 +
2  j
1 1 
2
+ 1
n Z1 Z2
(9.3.23)
On substituting for n2 , it is easy to show that this matrix is equal to that found
in (9.3.19).

Example 9.13 Design a three-element maximally flat low-pass filter with its
cutoff frequency as 1 GHz. It is to be used between a 50-
load and a generator
with its internal impedance at 50
.

SOLUTION A step-by-step procedure to design this filter is as follows:

ž Step 1.
As before, g values for the filter are found from (9.2.15) and (9.2.16) as

g0 = g4 = 1
π
g1 = 2 sin = 1
6
π
g2 = 2 sin = 2
2
and
5π
g3 = 2 sin =1
6
388 FILTER DESIGN

L1 = 1 L3 = 1
50 Ω

R0 = 1 R4 = 1
C2 = 2
50 Ω

(a)

L1 L3
L1 d L3 Z0 = 1 d Z0 = 1
Z0 = 1 Z0 = 1
U.E. U.E.
Z0 = 1 Z0 = 1
d d d
d C2
C2
Z0 = 0.5 Z0 = 0.5

(b) (c)

d d Z0 = 50 Ω Z0 = 100 Ω Z0 = 100 Ω Z0 = 50 Ω
U.E. U.E.
d
Z0 = 2 Z0 = 2
Z0 = 100 Ω Z0 = 100 Ω
Z0 = 2
Z0 = 2 Z0 = 0.5
Z0 = 25 Ω
(d) d = λ/8 at ω = 1 (e)

Figure 9.47 Maximally flat low-pass filter for Example 9.13.

The circuit is drawn with normalized values as shown in Figure 9.47(a).

ž Step 2. Using Richard’s transformation, the filter elements are replaced by
stubs. Hence, short-circuited series stubs replace inductors L1 and L3 , while
an open-circuited shunt stub replaces the capacitor C2 . This is shown in
Figure 9.47(b).
ž Step 3. Since the circuit found in step 2 cannot be fabricated in its present
form, unit elements are added on its two sides. This is shown in Fig-
ure 9.47(c).
ž Step 4. Using Kuroda’s identity shown in Figure 9.45(a), series stub and
unit element combinations are transformed to an open-circuited shunt stub
and the unit element. Since Z1 = Z2 = 1, equivalent capacitance C and the
characteristic impedance of the corresponding stub are found to be C = 12 =
0.5 and Z0 = 2. Also, the characteristic impedance of each unit element is
found to be 2. This circuit is illustrated in Figure 9.47(d).
ž Step 5. For impedance and frequency scaling, all normalized characteristic
impedances are now multiplied by 50
, and the stub lengths are selected
as λ/8 at 1 GHz. This is shown in Figure 9.47(e).
 PROBLEMS 389

SUGGESTED READING

Bahl, I., and P. Bhartia, Microwave Solid State Circuit Design. New York: Wiley, 1988.
Collin, R. E., Foundations for Microwave Engineering. New York: McGraw-Hill, 1992.
Davis, W. A., Microwave Semiconductor Circuit Design. New York: Van Nostrand Rein-
hold, 1984.
Elliott, R. S., An Introduction to Guided Waves and Microwave Circuits. Englewood Cliffs,
NJ: Prentice Hall, 1993.
Fusco, V. F., Microwave Circuits. Englewood Cliffs, NJ: Prentice Hall, 1987.
Matthaei, G. L., L. Young, and E. M. T. Jones, Microwave Filters, Impedance-Matching
Networks, and Coupling Structures. Dedham, MA: Artech House, 1980.
Pozar, D. M., Microwave Engineering. New York: Wiley, 1998.

PROBLEMS

9.1. Design a low-pass composite filter with nominal impedance of 50
. It

must pass all signals up to 10 MHz. Assume that f∞ = 10.05 MHz. Plot
its characteristics in the frequency range 1 to 100 MHz.
9.2. Design a high-pass T-section constant-k filter that has a nominal impedance
of 50
and a cutoff frequency of 10 MHz. Plot its frequency response in
the frequency band 100 kHz to 100 MHz.
9.3. Design an m-derived T-section high-pass filter with a cutoff frequency of
10 MHz and nominal impedance of 50
. Assume that f∞ = 9.95 MHz.
Plot the response of this filter in the frequency band 100 kHz to 100 MHz.
9.4. Design a high-pass composite filter with a cutoff frequency of 10 MHz and
image impedance of 50
in its passband. Assume that f∞ = 9.95 MHz.
Plot its response in the frequency range 100 kHz to 100 MHz.
9.5. It is desired to design a maximally flat low-pass filter with at least 25 dB of
attenuation at ω = 2ωc and −1.5 dB at its band edge. How many elements
will be required for this filter? If a Chebyshev filter is used with 1.5 dB of
ripple in its passband, find the number of circuit elements.
9.6. Design a low-pass Butterworth filter with a cutoff frequency of 150 MHz
and an insertion loss of 50 dB at 400 MHz. It is to be used between a 75-

load and a generator with its internal resistance as 75
.
9.7. Design a low-pass Chebyshev filter that exhibits ripples no more than 2 dB
in its passband. The filter must pass all frequencies up to 50 MHz and
attenuate the signal at 100 MHz by at least 15 dB. The load and the source
resistance are of 50
each.
9.8. Reconsider Problem 9.7 to design a low-pass filter that exhibits the Cheby-
shev response with a 3-dB ripple in its passband. The filter must have 50

at both its input and output ports.
390 FILTER DESIGN

9.9. Design a high-pass Chebyshev filter with a passband ripple magnitude

below 3 dB. It must pass all frequencies over 50 MHz. A minimum of
15 dB of insertion loss is required at 25 MHz. Assume that the load and
source resistances are of 75
each.
9.10. Design a bandpass Chebyshev filter that exhibits no more than 2 dB of
ripple in its passband. It must pass the signals that are in the frequency
band 50 to 80 MHz with zero insertion loss. Assume that the load and
source resistances are of 50
each.
9.11. It is desired to design a maximally flat low-pass filter with at least 40 dB of
attenuation at f = 1.1 GHz and −3 dB at f = 1 GHz. How many elements
will be required for this filter? If a Chebyshev filter is used with 3 dB of
ripple in its passband, find the number of circuit elements.
9.12. Design a Butterworth filter with a cutoff frequency of 960 MHz and an
insertion loss of 15 dB at 1.5 GHz. It is to be used between a 50-
load
and a generator with its internal resistance as 50
.
9.13. Design a low-pass Chebyshev filter that exhibits ripples of no more than
1 dB in its passband. The filter must pass all frequencies up to 960 MHz
and attenuate the signal at 1.5 GHz by at least 15 dB. The load and the
source resistance are of 50
each.
9.14. Reconsider Problem 9.13 to design a low-pass filter that exhibits the Cheby-
shev response with 2 dB of ripple in its passband. The filter must have 50

at both its input and output ports.
9.15. Design a high-pass Chebyshev filter with a passband ripple magnitude
below 0.5 dB. It must pass all frequencies over 500 MHz. A minimum of
20 dB of insertion loss is required at 200 MHz. Assume that the load and
source resistances are of 50
each.
9.16. Design a bandpass Chebyshev filter that exhibits no more than 0.1 dB of
ripple in its passband. It must pass the signals that are in the frequency
band 500 to 800 MHz with almost zero insertion loss and provide at least
15 dB of attenuation for signals below 300 MHz. Assume that the load and
source resistances are of 75
each.
9.17. Design a high-pass filter that exhibits no more than 2.5 dB of ripple in its
passband. It must pass all frequencies over 100 MHz. A minimum of 40 dB
of insertion loss is required at 25 MHz. Assume that the load and source
resistances are of 50
each.
9.18. Design a maximally flat bandstop filter with n = 5. It must stop the signals
in the frequency range 50 to 80 MHz and pass the rest of the frequencies.
Assume that the load and source resistances are of 50
each.
 PROBLEMS 391

9.19. Design a fifth-order low-pass Butterworth filter with a cutoff frequency

of 5 GHz. It is to be used between a 75-
load and a generator with an
internal resistance of 75
.
9.20. Design a third-order low-pass Chebyshev filter that exhibits ripples of no
more than 1 dB in its passband. The filter must pass all frequencies up to
1 GHz. The load and the source resistance are of 50
each.
9.21. Design a fifth-order high-pass Chebyshev filter with passband ripple mag-
nitude below 3 dB. It must pass all frequencies over 5 GHz. Assume that
the load and source resistances are of 75
each.