Preface

Filter Synthesis Using Genesys S/Filter covers the design of practical

lumped element, distributed, and resonator based filters from the
perspective of transmission zeros. While this book is of some interest to all
filter designers, it is written for users of the synthesis program Genesys
S/Filter from the Agilent EEsof EDA division of Agilent Technologies.
In the 1920s, filters were designed by cascading k-type sections with
attenuation poles (transmission zeros) at infinite frequencies and m-type
sections with additional transmission zeros at finite frequencies. This
method was mathematically economic and suitable for the calculation tools
then available, but the filter impedances and responses were only
approximate. In the 1930s, theories of network synthesis evolved that
produced exact designs with polynomial approximations to the ideal zonal
(brick wall) response filter. By the 1950s, solutions to these more
mathematically involved methods were found and tabulated. These tables
only required simple impedance and frequency scaling to produce efficient
lowpass and highpass filters, and simple transformations for bandpass and
bandstop filters. This became known as the modern method and today it is
still the common filter design technique. A reference to this method for RF
and microwave applications is the venerable Microwave Filters,
Impedance-Matching Networks, and Coupling Structures [1]. The
popularization of the digital computer resulted in a plethora of software
programs, most of which employ the modern method.
The modern method which is based on only a subset of topologies
available from general network synthesis does not necessarily result in the
most efficient filter. Orchard and Temes solved certain synthesis problems
caused by the precision of digital computers [2], and Szentirmai is a
pioneer in the development of commercial software for the direct synthesis
of filters [3]. Filter synthesis is a beautiful example of applied mathematics,
arguably one of the better examples in engineering. However, this book is
not about synthesis theory. This subject is treated in depth by [4],[5]. Note
that this book is not a treatment of modeling, tuning, standard values,
tolerance, or Q of elements and their effects on filter performance. Those
topics are covered in depth in my other works [6],[7].
The focus of the book is common filter design problems and how to use
direct synthesis to solve those problems. The emphasis is not on the
S/Filter user environment. The software documentation provides that
information. Rather, this book covers the application of S/Filter features to
solving important and common filter problems. S/Filters strength is the
incorporation of features that help simplify the powerful, but confusing-tothe-novice, process of direct filter synthesis. S/Filter includes tools beyond
xiii

xiv

Filter Synthesis Using Genesys S/Filter

direct synthesis, including a wide variety of both exact and approximate

equivalent network transforms, methods for selecting the most desirable
structure out of potentially thousands of synthesized alternatives, and a
transform history record that simplifies design attempts requiring iteration.
Very few software programs are based on direct synthesis, and the
additional features of S/Filter make it a uniquely effective tool for filter
design. The book assumes the reader is familiar with basic RF and
microwave concepts.
The author thanks How-Siang Yap of Agilent Technologies for
suggesting the inclusion of a chapter on multiplexers and for other
comments. He also thanks Rulon VanDyke of Agilent Technologies for
leading the incorporation of improvements to S/Filter suggested during the
writing of the book. The author also acknowloges the professional staff at
Artech House for their assistance in finalizing the book.

References
[1] G. Matthaei, L. Young and E.M.T. Jones, Microwave Filters, ImpedanceMatching Networks, and Coupling Structures, Artech House, Dedham, MA, 1980.
[2] H. Orchard and G. Temes, Filter Design Using Transformed Variables, IEEE
Trans. on Circuit Theory, Vol. CT-15, December 1968, pp. 385-408.
[3] G. Szentirmai, FILSYN A General Purpose Filter Synthesis Program, Proc. of
the IEEE, Vol. 65, October 1977, pp. 1443-1458.
[4] A. Zverev, Handbook of Filter Synthesis, John Wiley, Hoboken, NJ, 1967
[5] G. Temes and S. Mitra, editors, Modern Filter Theory and Design, John Wiley,
New York, 1973.
[6] R. Rhea, HF Filter Design and Computer Simulation, SciTech Publishing,
Raleigh, NC, 1994.
[7] R. Rhea, Practical Issues in RF Design (Three CD Series), SciTech Publishing,
Raleigh, NC, 2003.

To Marilynn:
As just a boy, I made my choice,
I now look back, and proudly smile.
With wind in face, or at our backs,
hand in hand, we trod our paths,
some a thousand, some a mile.
Youthful dreams, some wise some poor,
but oh so sweet, the ones we store.
For Hera, only change endures,
not so we say, tis love evermore,
hand in hand, tis love evermore.

Contents
Preface
References
1 Transmission Zeros
1.1 Determining TZ by Inspection
1.2 Filter Degree
1.3 Canonical Realization
1.4 Influence of TZs on the Response
References

xiii
xiv
1
1
4
4
4
6

2 All-Pole Lowpass and Highpass

2.1 Initial All-Pole Lowpass Parameters
2.2 Dual Topologies
2.3 Chebyshev Approximation with Even Order
2.4 All-Pole Highpass Example
References

7
7
9
10
11
12

3 Lowpass with Finite Zeros

3.1 Introduction
3.2 Alternative Topologies

13
13
15

4 Conventional Bandpass
4.1 Bandpass Transform
4.2 Classification Symmetry or Antimetry
4.3 A 75- to 125-MHz Bandpass
4.4 A 96- to 104-MHz Bandpass Filter
4.5 Comparative Analysis of the Wide and Narrow Filters
Reference

17
17
17
18
19
19
21

5 Extraction Sequences
5.1 The Extraction Tab
Reference

23
23
27

6 Customized Bandpass Filters

6.1 Custom Filter Specification
6.2 Partial Extractions of FTZs
6.3 Inexact Extractions
6.4 Inexact Example

29
29
33
34
34

7 Norton Transforms
7.1 Norton Series Transform
7.2 Removing a Transformer with the Series Norton

39
39
40

viii

Filter Synthesis Using Genesys S/Filter

7.3 Norton Shunt Transform
7.4 Equal-Valued Inductor Bandpass
7.5 The History Tab
7.6 Equate All Ls

43
44
45
46

8 Bandpass with Resonators

8.1 Coupled Parallel-Resonator Filters
8.1.1 Exact Design of a Parallel Resonator All-Pole Filter
8.1.2 Termination Coupling Transforms
8.1.3 Find Dual Transform
8.1.4 Exact Design with Like Coupling Elements
8.1.5 The Equate All Shunt Ls or Shorted Stubs Transform
8.1.6 Termination-Coupled Bandpass
8.2 Coupled Series-Resonator Filters
8.2.1 The Basic Series-Resonator Bandpass
8.2.2 Tubular Bandpass
8.2.3 Manufacture of the Tubular Bandpass
8.2.4 Generalized Series-Resonator Bandpass
8.2.5 Tunable Constant-Bandwidth Bandpass
Reference

47
47
49
51
53
55
56
57
58
58
59
61
61
63
67

9 TEM-Mode Resonators
9.1 Filter Insertion Loss
9.2 Filter Using 50-Ohm Coaxial Resonators
9.2.1 Lumped to Distributed Equivalents
9.2.2 The Convert Using Advanced Tline Routine
9.3 Generalized Bandpass Using Ceramic Resonators
9.3.1 Creating Parallel Resonators
9.3.2 Shifting the Internal Impedance Level
9.3.3 The Pi to Tee Transform: Increasing Coupling Caps
9.3.4 Converting the Parallel L-C to Coaxial Resonators
9.3.5 Optimizing the Values
9.4 Ceramic Bandpass with Two FTZs
References

69
69
70
70
72
74
75
76
77
77
77
78
81

10 Piezoelectric Devices
83
10.1 Quartz-Crystal Device Model
83
10.1.1 Physical Form of the Quartz Crystal
83
10.1.2 Insertion Response of a Quartz Crystal
84
10.1.3 Modeling the Quartz Crystal
84
10.1.4 Calculating Model Parameters from the Response 85
10.1.5 The Quartz-Crystal Model and Filter Design
86
10.2 Quartz-Crystal Filter Approximate Design
86
10.3 Nulling the Static Capacitance
90
10.4 Design of a Lower-Sideband Crystal Filter
91
10.5 Upper-Sideband Quartz-Crystal Filter
97
10.6 Filters with TZs Above and Below the Passband
103

Contents
10.7 Wide-Bandwidth Quartz-Crystal Filters
10.8 Very Wide-Bandwidth Quartz-Crystal Filters
10.9 Ceramic-Piezoelectric Resonators
Reference

ix
107
108
111
113

11 Symmetry
11.1 Physical Symmetry
11.1.1 A Lowpass Filter with FTZ Pairings
11.1.2 A Bandpass Filter with FTZ Pairings
11.2 Response Symmetry
11.2.1 All-Pole Symmetric Response Filters
11.2.2 Generalized Bandpass with Symmetric Response
11.2.3 Symmetry by FTZ Placement
11.3 Group-Delay Equalization
References

115
115
115
117
119
120
120
123
124
127

12 Matching with S/Filter

12.1 Matching Concepts
12.1.1 Complex Conjugate Match
12.1.2 Two-Element Matching Networks
12.2 Real Terminations
12.2.1 Exploiting Extraction Sequences
12.2.2 Exploiting Resonator Filters
12.3 Complex Terminations
12.3.1 Fanos Limit
12.3.2 Example: Power Amplifier Match
12.3.3 Example: Broadband Antenna Match
References

129
129
130
130
132
132
138
139
139
140
142
144

13 Distributed Filters
13.1 Comparing Distributed and Lumped Filters
13.2 The Genesys Microwave Filter Module
13.3 Distributed Synthesis Concepts
13.3.1 TLEs
13.3.2 Richards Transform
13.3.3 Kuroda Identities
13.3.4 Ikeno Transforms
13.3.5 Kuroda-Minnis Transform
13.3.6 Half-Angle Transform
13.3.7 Interdigital Transform
13.3.8 Combline Transform
13.4 Lumped to Distributed Equivalent Transforms
13.5 Inverters
13.6 The Convert Using Advanced TLine Routine
13.7 Box Modes
13.8 Introduction to Distributed Filter Examples
References

145
145
146
149
149
150
152
155
157
159
161
161
162
164
165
166
166
167

Filter Synthesis Using Genesys S/Filter

14 Distributed Lowpass Filters

14.1 Exact Methods
14.1.1 Lowpass with Redundant UEs
14.1.2 Stub TLEs and Contributing Unit Elements
14.1.3 Lowpass with Only Contributing UEs (Stepped-Z)
14.1.4 Generalized Lowpass Filter
14.2 Approximate Methods
14.2.1 All-Pole: Equivalent Series TLE and Shorted Stubs
14.2.2 Stepped Impedance Lowpass
14.2.3 Generalized Lowpass
14.3 Size Reduction by Penetration
14.4 Radial Stub Lowpass
14.5 Hybrid Lowpass
14.6 Distributed Lowpass Summary
Reference

169
169
169
175
176
179
180
182
183
187
190
192
194
196
198

15 Distributed Bandstop Filters

15.1 All-Pole with Stubs and Contributing UEs
15.1.1 Wide Bandwidth Bandstop
15.1.2 Moderate Bandwidth Bandstop
15.1.3 Narrow Bandstop with Ikeno Transforms
15.2 Generalized Narrowband Bandstop

199
199
199
202
204
205

16 Distributed Bandpass Filters

16.1 Tutorials of Bandpass by Synthesis
16.1.1 Edge-Coupled Using Richards Transform
16.1.2 Edge-Coupled Using Inverters
16.1.3 Interdigital Using Inverters
16.2 Unique Bandpass Designs
16.2.1 Combline with Capacitive External Coupling
16.2.2 Miniature Bandpass with Contributing UEs
16.2.3 Narrow Bandwidth with UEs and an FTZ
16.2.4 Penetrating Combline
16.2.5 Minnis Class-D Bandpass
16.3 Hybrid Bandpass
16.3.1 Penetrating Combline with Capacitors
16.3.2 Generalized Combline Hybrid
16.3.3 Direct-Coupled Bandpass with Capacitors
References

211
211
211
216
218
224
224
228
233
238
245
248
248
249
252
258

17 Distributed Highpass Filters

17.1 The Hybrid Highpass
17.1.1 The All-Pole Hybrid: Distributed Synthesis
17.1.2 The All-Pole Hybrid Highpass: Lumped Synthesis
17.1.3 The Hybrid Highpass with UEs
17.1.4 The Hybrid Highpass with an FTZ
17.2 Purely Distributed Highpass

259
259
259
261
263
266
268

Contents
17.2.1 Highpass with Three TZs at DC and a UE
17.2.2 Highpass with Three TZs at DC and Four UEs
17.3 The Highpass Synthesized as a Bandpass
17.3.1 Hybrid Highpass from an Eighth-Degree Bandpass
17.3.2 Hybrid Highpass from a 10th-Degree Bandpass

xi
268
270
272
272
275

18 Multiplexers
18.1 Contiguous Multiplexers
18.1.1 Contiguous Lowpass-Highpass Diplexer
18.1.2 Contiguous LP/BP/HP Multiplexer
18.2 Noncontiguous Multiplexers
18.2.1 Noncontiguous LP/HP Diplexer with FTZ
18.2.2 Noncontiguous Distributed Combline Diplexer
Reference

277
277
277
279
281
281
284
287

19 Electromagnetic Simulation
19.1 Overview
19.1.1 The EMPower Program
19.1.2 The Momentum Program
19.1.3 The EMPro Program
19.2 Box Modes
19.3 EM Simulation of Distributed Circuits
19.3.1 EM Simulation of Penetrating Stepped-Z Lowpass
19.3.2 EM Simulation of a Combline Bandpass
19.3.3 EM Simulation of a Direct-Coupled Bandpass
19.4 Classic Method of Bandpass Design
19.4.1 Classic Method Fundamentals
19.4.2 Example: Determining K Values
19.4.3 Example: Determining Q Values
19.4.4 Filter Example Using the Classic Method
References

289
289
290
291
292
292
295
295
298
300
302
302
304
307
307
310

Appendix A: Example Summary

A.1 Lumped Examples
A.2 Distributed Examples
A.3 Hybrid Examples
A.4 Multiplexer Examples

313
313
315
316
317

Constants, Symbols, and Initialisms

319

About the Author

323

Index

325

12 Matching with S/Filter

Matching herein refers to the use of a network to cause maximum
power transfer from an arbitrary source impedance to an arbitrary load
impedance. Even with complex impedances, matching at a single
frequency, or approximately over a narrow bandwidth, is a straightforward
process requiring simple networks with closed form equations [1],[2].
With wider bandwidth, the problem can be daunting. Genesys includes
Impedance Match, a module designed to match real or complex
impedances using eight built-in network algorithms, or a custom userspecified network. One or more of these algorithms can find the optimum
solution to an impedance-matching problem. For increasing bandwidth,
the user specifies an increased quantity of elements. One of the Impedance
Match algorithms uses a direct synthesis similar to that in S/Filter. When
the problem is primarily a matching problem, Impedance Match is
recommended. This chapter addresses matching techniques used with
S/Filter when the problem is primarily a filtering problem, but the loads
are dissimilar or complex.

12.1 Matching Concepts

Impedance mismatch may be viewed from a network or a wave perspective.
For a network perspective, consider Fig. 12.1, with a 1-volt DC source with a
50-ohm source impedance and a 50-ohm load impedance.
SOURCE

LOAD

Rsource
R=50

VS1
VDC=1V

Rload
R=50

Figure 12.1 The matching problem from a network perspective.

129

130

Filter Synthesis Using Genesys S/Filter

The current is 1 volt over 100 ohms, or 10 mA. The voltage source
delivers 10 mW; 5 mW to the source resistance and 5 mW to the load
resistance. If the load is 25 ohms, the current increases to 13.33 mA, and
the 1-volt source delivers 13.33 mA. However, only 4.44 mW is delivered to
the load, while the majority is delivered to the source resistance. If the load
is 100 ohms, the current decreases to 6.67 mA, and the 1-volt source
delivers 6.67 mW. While the majority is delivered to the load, again it is
only 4.44 mW. Maximum transfer of power occurs when the load resistance
equals the source resistance.
From a wave perspective, when a signal propagating through a system
encounters an impedance shift, a portion of the signal is reflected, thus
reducing the transmitted signal. This is analogous to reflection in an optical
system. Mismatch occurs when the impedance shift is either resistive or
complex.

12.1.1 Complex Conjugate Match

When the load impedance is complex, the power delivered from a resistive
source to the resistive component of the load is reduced. When the load
impedance is R + jX, maximum power transfer is restored if the source
impedance is R - jX. This is referred to as a complex conjugate match. The
source may be purely resistive and a matching network may be used to
present an effective R - jX to the load. This network may also shift the
resistive component of the source to equal the load resistance. When the
matching network added between the source and load is purely reactive,
this network dissipates no energy and maximum power transfer occurs.

12.1.2 Two-Element Matching Networks

Two reactors may be used to match any complex source and load
impedance [1],[2]. Given in Fig. 12.2 are the matching-spaces supported by
the matching-network types one through eight. These types of diagrams
were first introduced by Smith [3]. Complex loads RL + jXL within the
enclosed spaces on the Smith chart are matched to a real source impedance
R0 using the indicated two-reactor network. Given below are the closedform equations for computing the reactor values of the matching-network
types one through eight in Fig. 12.2. At a design frequency f and with the
load admittance given by

YL = GL + jBL =
and then setting

RL

RL2 + X L2

XL

RL2 + X L2

(12.1)

Matching with S/Filter

131

GL
GL2
R0

A=

(12.2)

B = RL ( R0 RL ) X L

(12.3)

C = RL ( R0 RL ) X L

(12.4)

the network values are

L1 =

2f GL2 + A2

(12.6)

1
2f ( A + BL )

(12.7)

GL2 + A2
2fA

(12.8)

B
2f

(12.9)

C2 =

L3 =
C3 =

2f

L4 =

B + XL

+ (B + X L )2

RL2

(R

+ (C + X L )2
2f (C + X L )
2
L

C5a =

(12.10)

(12.11)

1
2fC

(12.12)

GL2 + A2
2fA

(12.13)

A + BL
2f

(12.14)

C4 =

C5b =
L6 a =

(12.5)

A BL
2f

C1 =
L2 =

2f Gl2 + A2

(12.15)

132

Filter Synthesis Using Genesys S/Filter

1
2f ( A BL )

L6b =

C7 a =
C7b =

2f

1
2fB

(12.17)

B + XL

RL2

+ (B + X L )2

L8a =
L8b =

(12.16)

C
2f

(R

+ ( X L + C )2
2f ( X L + C )
2
L

(12.18)
(12.19)

(12.20)

Consider the match of a 50-ohm source to a load of 10 ohms in series

with 6.366 pF at 100 MHz. The load impedance is 10 j250 ohms. Types
two, three, six, and eight are valid matching networks. For a lowpass type
passband, type three is selected. From Eqs. 12.3, 12.4, 12.9, and 12.10, L3 is
429.72 nH and C3 is 63.66 pF. The resulting S11 is given in Fig. 12.3.
The match is exact at only the design center frequency. If the reactance
is significant, or the ratio of Ro to RL differs significantly from unity, the
matched bandwidth is narrow. In our example, the 20-dB return loss
bandwidth is only 0.6 MHz, or 0.6% bandwidth. The problem of matching,
at least for lossless elements, is largely an issue of bandwith. Increased
bandwidth is achieved by using matching networks with an increased
quantity of elements. The remainder of this chapter covers direct synthesis
to create these matching networks.

12.2 Real Terminations

This section deals with S/Filter matching when the terminations are purely
resistive with no reactive element.

12.2.1 Exploiting Extraction Sequences

As demonstrated in examples in previous chapters, certain extraction
sequences result in a termination resistance different than the source
resistance. This is used to advantage when matching dissimilar resistance.

Matching with S/Filter

Figure 12.2 Match space for two-element matching networks.

133

134

Filter Synthesis Using Genesys S/Filter

Figure 12.3 Match response of the two-element matching network.

Fig. 12.4 shows plots of the maximum ratio of termination resistance

that are to be matched by a bandpass filter given a ratio of the upper
frequency to the lower frequency, for a match of 20-dB return loss in the
passband. The four traces represent the quantity of nonfinite transmission
zeros at DC and infinity. For example, the 2/4 trace represents a
synthesized network with two TZs at DC and four TZs at infinity. Recall that
for an all-pole network, six TZs result in a network with six elements when
the network is canonic.
Increasing the quantity of TZs increases the matching bandwidth for a
given termination resistance ratio. For example, when matching a 50-ohm
source to a 100-ohm load, the termination resistance ratio is two. The
termination resistance ratio is also two when matching a 50-ohm source to
a 25-ohm load. With two TZs at DC and at infinity, the maximum upper
frequency is almost twice the lower frequency, an octave bandwidth. With
six TZs at DC and at infinity, the bandwidth is a decade. Likewise,
increasing the quantity of TZs increases the achievable bandwidth for a
given termination resistance ratio. For example, over an octave bandwidth,
2/2 matches a resistance ratio of almost two while 6/6 matches a resistance
ratio of 100.

Matching with S/Filter

135

Figure 12.4 Matchable termination resistance ratios versus the ratio of the upper to lower
cutoff frequency for a passband return loss of 20 dB.

For matching, even quantities of TZs are more economic than odd
quantities. For example, 2/4 provides better bandwidth or higher
termination resistance ratios than 3/3. All even TZ choices of the same
quantity result in similar performance. For example, 2/6, 4/4, and 6/2
provide similar performance.
Fig. 12.4 is for the case with 20-dB passband return loss. A worse
return loss improves the bandwidth and increases the termination
resistance ratio.
With respect to matching, FTZs are not economic. For a given quantity
of elements, the specification of FTZs reduces the performance. FTZs are
employed in filters used for matching only if the stopband requirements
benefit.
The above discussion involves design to maximize the bandwidth or
termination resistance ratio. The general case of matching, however,
involves matching a source resistance to a specific load resistance. For
example, consider a required resistance ratio of 30 over an octave
bandwidth. This requirement lies between TZ placements of 4/4 and 6/6. A
number of techniques are employed:
1) If the design is purely a matching issue, the 6/6 synthesis is used and
the bandwidth is extended beyond the required bandwidth. This
reduces the sensitivity of the design to element tolerance.

136

Filter Synthesis Using Genesys S/Filter

2) If filtering is required and the stopband rejection must be retained,

the 6/6 synthesis is used and a Norton transform is applied to adjust
the load resistance.
3) If filtering is required and the stopband rejection must be retained,
the 6/6 synthesis is used and a different extraction sequence is used.
Fig. 12.4 is derived by selecting the sequence resulting in the greatest
performance.
4) The passband ripple is reduced until the extraction sequence results
in a TRF of unity.
Fig. 12.4 is provided to illustrate the relationships between network
complexity, the resistance ratio, and matching bandwidth. To illustrate how
the design process might actually proceed, consider the following example.
A 50-ohm source is matched to a 12.5-ohm load from 100 to 200 MHz. A
passband return loss of 12 dB is required. No filtering specifications are
involved.
Given in Fig. 12.5 is the Specification tab specifying the above
parameters. The load resistance is set at the required 12.5 ohms. The
passband ripple is adjusted until the response indicated a 12-dB return loss.
A review of Fig. 12.4 suggests that two TZs at DC and four TZs at infinity
should be sufficient.

Figure 12.5 Specification tab for matching example that exploits extraction sequences.

The Extraction tab is selected and the Lmax/Lmin column is canvassed

for a small inductor value ratio. The extraction sequence DC DC is
selected. Next, the Transform tab is selected and the Remove Transformer
button is clicked. The resulting matching network is given in Fig. 12.6. The
Norton transform added an inductor. Selecting the extraction sequence

Matching with S/Filter

137

DC DC foll owed by the Remove Transformer option adds a

capacitor, thus reducing the inductor count to three.
ZO=12.5

ZO=50
L2
L=25.78nH

L1
L=194.21nH

L3
L=10.02nH
C1
C=43.12pF

L4
L=26.96nH
C2
C=147.34pF

C3
C=48.34pF

Figure 12.6 A 50- to 12.5-ohm matching network designed by exploiting extraction

sequences.

Alternatively, the bandwidth is increased by decreasing the Lower

Cutoff and increasing the Upper Cutoff until the TRF ratio for the selected
extraction sequence is unity. In this case, no Remove Transformer option is
required. The resulting schematic is given in Fig. 12.7 and the responses in
Fig. 12.8. Another option is to decrease the passband ripple until a Norton
transform is not required. In this case, a passband ripple of 0.16 dB with a
return loss of 14.4 dB results in a canonic solution very similar to Fig. 12.7.
ZO=12.5

ZO=50
L1
L=24.97nH
L2
L=12.03nH
C1
C=35.85pF

L3
L=22.44nH
C2
C=122.7pF

C3
C=59.11pF

Figure 12.7 Matching network with bandwidth expanded to avoid a Norton transform.

138

Filter Synthesis Using Genesys S/Filter

Figure 12.8 Responses of the 50- to 12.5-ohm matching network with expanded bandwidth.

12.2.2 Exploiting Resonator Filters

Coupled resonator filters typically employ impedance transformations that
may be exploited to match dissimilar terminations. This is particularly
effective for narrow bandwidth filters. Consider the following example of an
88- to 108-MHz three-section series resonator filters with matching of 100
to 50 ohms with a 20-dB return loss. Both source and load terminations are
specified as 100 ohms, the cutoff frequencies are specified, and a passband
ripple of 0.0432 dB is specified. As described in Section 8.2.1, we begin
with the Series Element First extraction sequence DC and apply
shunt Norton transforms to the first shunt capacitor. A turns ratio value of
n equal to 0.21082 results in equal first and second inductors. The resulting
schematic is given in Fig. 12.9.
C1
C=4.67pF

ZO=100
L1
L=666.37nH

C3
C=4.87pF
L2
L=666.37nH

C2
C=18.22pF

ZO=100
L3
L=27.75nH

C4
C=89.29pF

T1
P=1
S=4.9

Figure 12.9 Schematic of the 88- to 108-MHz bandpass used as a matching network.

Matching with S/Filter

139

Notice that the transformer turns ratio is 4.9, thus indicating that
without a transformer, this network matches 100 ohms down to about 4.2
ohms. The application of a shunt Norton to the final shunt capacitor
supports matching any resistance from 4.2 ohms up to 100 ohms.
Therefore, the termination resistance in the Specification tab is set to the
desired 50 ohms and a Norton Shunt transform is applied to C4, selecting n
to remove the transformer. The filter matching 100 to 50 ohms is given in
Fig. 12.10.

L3
L=332.5nH

L2
L=666.37nH

L1
L=666.37nH

C5
C=10.4779pF

C3
C=5.6216pF

C1
C=4.671pF

ZO=100

ZO=50

C4
C=25.79pF

C2
C=18.22pF

Figure 12.10 Final schematic of the series-resonator bandpass used to match 100 to 50
ohms.

12.3 Complex Terminations

As illustrated by Fig. 12.4, very wide-bandwidth, even as high as a decade or
more, is readily achieved between resistive terminations. When the load is
complex, the matching bandwidth is more limited.

12.3.1 Fanos Limit

Three definitions of Q are required to define the matching problem with
complex terminations. Loaded Q is the center frequency divided by the 3dB bandwidth of the response.

Qloaded =

f upper

f0
f lower

(12.21)

It is a finite value even if the circuit is built using elements with infinite
Q. Element Q, or unloaded Q, is a measure of element quality. It is defined
as 2f times the stored energy divided by the dissipated energy in the
element. It is as high as 200 for excellent inductors. Unloaded Q increases
with physical size, so modern miniature chip inductors have low unloaded
Q. Finally, Q of the load is a property of a complex termination. For series
impedance it is given simply by

140

Filter Synthesis Using Genesys S/Filter

Qof load =

XL
RL

(12.22)

and for parallel admittance it is

Qof load =

BL
GL

(12.23)

As the Q of the load increases, the achievable match bandwidth

decreases. In a classic paper, Fano [4] offers an elegant formula for
predicting the relationship between the bandwidth and the best achievable
reflection coefficient using a lossless, infinitely complex matching network.

min = e

Qloaded
Qof load

(12.24)

where min is the magnitude of the best reflection coefficient achievable

over a bandwidth f upper f lower . The return loss of min is

RLdB = 20 log min

(12.25)

Useful algebraic derivatives of Fanos equation are

Qloaded =

Qof load ln min

Qof load =

Qloaded
ln min

(12.26)
(12.27)

An octave bandwidth is Qloaded = 1.5 . Eq. 12.25 indicates that the

reflection coefficient of 0.1 is a return loss of 20 dB. From Fanos equation,
to achieve a 20-dB return loss over an octave bandwidth requires
Qof load less than 2.047. Fanos formula provides an absolute limit that is
approached with an infinite quantity of elements. Fanos limit is used to
discover if a solution is possible, thus avoiding effort on an unsolvable
problem.

12.3.2 Example: Power Amplifier Match

A transistor power amplifier output with 9.6 ohms resistance in parallel
with 40 pF is to be matched to 50 ohms from 200 to 400 MHz with a
worse-case return loss of 12 dB, or min of 0.251. The Q of the load from Eq.
12.23 is 0.4825 at 200 MHz and 0.9651 at 400 MHz. Using the higher Q of
the load, from Eq. 12.24, for an octave bandwidth (loaded Q = 1.5), a

Matching with S/Filter

141

reflection coefficient of 0.00758 is theoretically achievable with an infinite

quantity of elements. Design begins with the Specification tab given in Fig.
12.11.

Figure 12.11 Initial Specification tab for the 200- to 400-MHz power-amplifier matching
network.

The matching network must include a shunt capacitor at the input that
is sufficiently large to absorb the 40-pF transistor output capacitance. The
extraction must therefore begin with a shunt element first. Furthermore,
the extraction must result in a topology with a transformer turns ratio less
than unity so that a Norton transform is available to remove the
transformer. The quantity of nonfinite TZs is increased until a satisfactory
extraction sequence is achieved. The lowest degree is six. Since harmonic
suppression is desirable in a power amplifier, the extraction with one TZ at
DC and five at infinity is selected with the sequence DC .
Finally, to avoid adding an additional capacitor with a Norton
transform, to improve the design margin, and to retain the filtering
functions of the network, the requested passband ripple in the Specification
tab is reduced until the transformer turns ratio in the Extraction tab is
unity. The resulting passband ripple is 0.0282 dB, or a passband return
loss of 22 dB. The schematic for the power-amplifier matching network is
given in Fig. 12.12. Capacitor C1 is replaced with a 24.84-pF capacitor with
the remaining capacitance provided by the transistor output capacitance.

142

Filter Synthesis Using Genesys S/Filter

ZO=9.6

ZO=50
L2
L=7.68nH

L1
L=6nH

C1
C=64.84pF

C2
C=38.51pF

L3
L=17.5nH
C3
C=12.47pF

Figure 12.12 Schematic of the 200- to 400-MHz power-amplifier matching network. C1 is

reduced by the transistor output capacitance.

12.3.3 Example: Broadband Antenna Match

This example illustrates a matching network for a monopole antenna. The
design is required to center the resonant frequency at 3.75 MHz, shift the
resistive component up to 50 ohms, and broaden the bandwidth of the
return loss. An R-L-C model for the antenna is 34 ohms in series with a
series L-C of 19.5 uH and 87 pF. The VSWR of this antenna in a 50-ohm
system is given as the solid trace in Fig. 12.13. The VSWR is a radial-scaled
parameter often used with antennas. It is related to the magnitude of S11 by
the following equation:

VSWR =

1 + S11
1 S11

(12.28)

The minimum VSWR in this example is not unity because of the

mismatch of the 34-ohm antenna resistance to the 50-ohm system. The
minimum occurs when the L-C components of the model resonate.
Resonance does not occur at band center because the monopole antenna is
electrically shorter than a quarter wavelength at 3.75 MHz.

Matching with S/Filter

143

Figure 12.13 VSWR responses of the antenna before (solid trace) and after matching (dashed
trace).

The goal is to achieve a VSWR under 3:1 across the frequency span of
3.5 to 4.0 MHz. Design begins with the specification of the 50-ohm source
and 34-ohm load, frequency cutoffs of 3.5 and 4.0 MHz, and a passband
ripple of 1.25 dB that corresponds to a VSWR of 3:1. The series L-C
component of the model is incorporated into the final series branch of the
matching network. The quantity of TZs is increased and an extraction
sequence is selected so that the 3:1 VSWR is achieved over the desired
band, and the matching network contains a final series L-C branch with the
inductor greater than the antenna 19.5 uH, and a capacitor smaller than the
antenna 87 pF. The extraction sequence selected is DC DC DC . Then
a Remove Transformer transform is applied to set the output resistance at
34 ohms. The final matching network could be realized by absorbing the
antenna L-C into the matching network values. However, to eliminate the
matching network capacitor and to afford margin in the design, the ripple is
reduced to 1.0776 dB to improve the VSWR maximum and the cutoffs are
broadened to 3.48 and 4.02 MHz to widen the frequency response. This
sets the capacitor C4 to exactly match the antenna capacitance. The network
is shown in Fig. 12.14. The capacitor C4 is absent in the final network and
the inductor is 20.857 uH minus 19.5 uH, or 1.357 uH. If the antenna is
physically lengthened, this inductor is eliminated as well. The final
response is given as a dashed trace in Fig. 12.13.

144

Filter Synthesis Using Genesys S/Filter

C1
C2
C=59.02pF C=32912.06pF

ZO=50
L1
L=30677.88nH

C4
C=87pF
L3
L=20857.56nH

L2
L=313.57nH

ZO=34

C3
C=7002.93pF

Figure 12.14 Matching and broadbanding network for the antenna. Refer to the text regarding
L3 and C4.

References
[1] R. Rhea, The Yin-Yang of Matching: Part 1 Basic Matching Techniques, High
Frequency Electronics, March 2006.
[2] R. Rhea, The Yin-Yang of Matching: Part 2 Practical Matching Techniques,
High Frequency Electronics, April 2006.
[3] P. Smith, Electronic Applications of the Smith Chart, SciTech Publishing,
Raleigh, NC, 1995.
[4] R. Fano, Theoretical Limitations on the Broadband Matching of Arbitrary
Impedances, Jour. Franklin Institute, January, 1950.