Filter Synthesis Using Genesys S - Filter PDF
Filter Synthesis Using Genesys S - Filter PDF
Filter Synthesis
S/Filter includes tools beyond direct synthesis, including a wide variety of both
exact and approximate equivalent network transforms, methods for selecting
the most desirable 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.
Contents Overview:
Transmission Zeros; All-Pole Lowpass and Highpass; Lowpass with Finite
Zeros; Conventional Bandpass; Extraction Sequences; Customized Bandpass
Filters; Norton Transform; Bandpass with Resonators; TEM-Mode Resonators;
Piezoelectric Devices; Symmetry; S/Filter and Matching; Distributed Filters;
Distributed Lowpass Filters; Distributed Bandstop Filters; Distributed Bandpass
Filters; Distributed Highpass Filters; Electromagnetic Simulation; Classic
Bandpass Method
Rhea
Include bar code
BOSTON LONDON
Randall W. Rhea
www.artechhouse.com
Filter Synthesis Using Genesys S/Filter
For a complete listing of titles in the
Artech House Microwave Library
turn to the back of this book.
Filter Synthesis Using Genesys S/Filter
Randall W. Rhea
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the U.S. Library of Congress.
All rights reserved. Printed and bound in the United States of America. No part of this book
may be reproduced or utilized in any form or by any means, electronic or mechanical, including
photocopying, recording, or by any information storage and retrieval system, without permission
in writing from the publisher.
All terms mentioned in this book that are known to be trademarks or service marks have been
appropriately capitalized. Artech House cannot attest to the accuracy of this information. Use of
a term in this book should not be regarded as affecting the validity of any trademark or service
mark.
10 9 8 7 6 5 4 3 2 1
To Marilynn:
xiii
xiv Filter Synthesis Using Genesys S/Filter
References
[1] G. Matthaei, L. Young and E.M.T. Jones, Microwave Filters, Impedance-
Matching 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.
1 Transmission Zeros
Central to filter design by direct synthesis is the specification of
transmission zeros (TZs) and understanding their influence on the filter
response. With the modern method refined in the 1950s, specified filter
parameters include the type of approximation, such as Butterworth or
Chebyshev, the passband frequencies, the termination impedance, and the
order of the filter. The TZs are fixed by the filter type. Fixed TZs simplify
the design process, but they restrict the characteristics of the filter. The
newer direct synthesis method described in this book supports flexible
specification of TZs, thus providing the designer with more control [1],[2].
The first step is to understand TZs and their influence on the filter.
1
2 Filter Synthesis Using Genesys S/Filter
Figure 1.1 (a) All-pole lowpass filter showing DC and Infinite TZs and (b) a generalized
lowpass with also a TZ at finite frequency.
branch is a TZ, the series inductor in the second branch is a TZ, and the
capacitor in the third branch is a TZ. This filter has three TZs at infinity.
Because it has TZ at both DC and infinity, this is a bandpass filter.
Figure 1.2 (a) Conventional three-resonator all-pole bandpass filter and (b) generalized
bandpass filter with FTZs.
Consider the 45- to 85-MHz bandpass filter in Fig. 1.2b. At DC, all of
the shunt branches are open and they play no role. Capacitor C5 is shorted
by L3, so C5 plays no role. Capacitors C1, C4, and C7 are all in series and are
replaced with one series capacitor of smaller value. This filter has one TZ at
DC. At infinity, capacitors C4 and C5 form a short that places capacitors C3
and C6 in parallel. These latter two capacitors are replaced with one
capacitor of larger value. The branch containing L2 and C2 becomes an open
because of L2 and this branch plays no role. Therefore, L1, C3 in parallel
with C6, and L4 form three TZs at infinity. L2 and C2 series resonate at 30
MHz forming a short to ground resulting in a FTZ. C5 and L3 resonate at
102 MHz forming an open resulting in a FTZ. This filter has one TZ at DC,
three TZs at infinity, one FTZ below the passband, and one FTZ above the
passband.
4 Filter Synthesis Using Genesys S/Filter
Figure 1.3 Lowpass filter responses of a five-element Butterworth (solid), a five-element 0.5-
dB ripple Chebyshev (dashed), and a five-branch Cauer-Chebyshev (dash-dot).
Figure 1.4 Lowpass filter with one TZ at infinity and two FTZ.
References
[1] R.W. Rhea, “Transmission Zeros in Filter Design,” Applied Microwave &
Wireless, January 2001, pp. 92, 94, and 96.
[2] R.W. Rhea, Filter Design by Transmission Zeros (CD), SciTech Publishing,
Raleigh, NC, 2003.
[3] G. Temes and S. Mitra, Modern Filter Theory and Design, John Wiley & Sons,
New York, 1973.
[4] A.I. Zverev, Handbook of Filter Synthesis, John Wiley & Sons, Hoboken, NJ,
1967.
2 All-Pole Lowpass and Highpass
The direct synthesis method offers no advantage over the modern
method for all-pole lowpass and highpass filters. Nevertheless, the basic all-
pole filters are important and are used in this chapter to introduce the
concept of direct-synthesis filter design.
7
8 Filter Synthesis Using Genesys S/Filter
based on the required passband return loss for an ideal filter. A passband
ripple of 0.5 dB corresponds to a return loss of 9.65 dB.
Table 2.1 Representative Values of Radial-Scaled Parameters
Snn(dB) VSWRn LA(dB)
ρn
0.010 -40.00 1.020 0.0004
0.032 -30.00 1.065 0.0043
0.056 -25.00 1.119 0.0138
0.100 -20.00 1.222 0.0436
0.158 -16.00 1.377 0.1105
0.200 -14.00 1.499 0.1764
0.251 -12.00 1.671 0.2830
0.316 -10.00 1.925 0.4576
0.333 -9.54 2.000 0.5118
0.398 -8.00 2.323 0.7494
0.500 -6.02 3.000 1.2496
0.600 -4.44 3.997 1.9365
0.707 -3.01 5.829 3.0106
0.794 -2.00 8.724 4.3292
1 + ρn
VSWR = (2.2)
1 − ρn
(
L A (dB) = −10 log 1 − ρ n
2
) (2.3)
These two forms are given in Fig. 2.3. The two topologies are the dual
of each other. The transmission characteristics of these dual forms are
identical in all respects. The port-impedance characteristics as a function of
frequency are different, which is important for the diplexer designs
considered later. In addition, element values are different, and certain
forms are advantageous for practical reasons. The designer is free to choose
the desired topology. The remaining selections in the Extraction tab are
covered later.
10 Filter Synthesis Using Genesys S/Filter
L1 L2
L=97.85nH L=97.85nH
C1 C2 C3
C=54.3pF C=80.88pF C=54.3pF
L1 L2 L3
L=135.74nH L=202.19nH L=135.74nH
C1 C2
C=39.14pF C=39.14pF
Figure 2.3 Dual forms of the lowpass filter with shunt element first (top) and series element
first (bottom).
ZO=50Ω ZO=25.13Ω
L1 L2
L=94.9nH L=66.99nH
C1 C2
C=53.17pF C=75.32pF
Figure 2.4 Even order (fourth) lowpass filter schematic (left), and the transmission and
reflection responses (right).
inductor to a series capacitor. Then the values are scaled for frequency and
termination resistance, just as with the lowpass.
Fig. 2.5 shows the schematic and responses of a 0.1-dB ripple
Chebyshev highpass with four TZs at DC. Notice that similar to the lowpass,
the even-order Chebyshev highpass requires dissimilar termination
resistance.
ZO=50Ω ZO=30.86Ω
C1 C2
C=25.08pF C=37.41pF
L1 L2
L=57.74nH L=38.71nH
References
[1] Agilent Technologies, GENESYS 2012.01 Documentation Set, www.agilent.com.
[2] R.W. Rhea, HF Filter Design and Computer Simulation, SciTech Publishing,
Raleigh, NC, 1994.
3 Lowpass with Finite Zeros
As previously stated, direct synthesis offers no advantage over the
modern method for all-pole lowpass filter design. However, this is not the
case for generalized lowpass filters with FTZs. In this chapter, instruction
on the use of S/Filter continues with the design of generalized lowpass
filters.
3.1 Introduction
The design of generalized lowpass filters will begin with a specific filter
requirement. Consider the following required filter specification:
1) Nominal termination resistances: 50 ohms
2) Passband cutoff frequency: 32 MHz
3) Maximum passband return loss: 16 dB
4) Minimum rejection at 35 MHz: 50 dB
5) Minimum rejection above 38 MHz: 20 dB
The goals are maximum economy with minimum inductor count. The
rejection requirements are indicated on the graph in Fig. 3.1 as a mark at 35
MHz and a shaded region above 38 MHz.
An all-pole 0.1-dB ripple Chebyshev filter that just satisfies the
requirements is 19th order, thus requiring nine inductors and ten
capacitors. The response is given in Fig. 3.1 as the dotted trace. A schematic
is not given here.
Next, an elliptic Cauer-Chebyshev filter is designed using the modern
filter design method. A schematic of the solution that just satisfies the
requirement is given in Fig. 3.2. It satisfies the 35-MHz rejection
requirement with frequency margin, but the next lower order filter is
insufficient. This filter requires four inductors and nine capacitors, a
significant improvement over the all-pole Chebyshev. It is incorrect to
assume that Cauer-Chebyshev filters always have superior economy to all-
pole filters. This comparison depends on specific filter specifications.
13
14 Filter Synthesis Using Genesys S/Filter
Figure 3.1 Rejection goals (mark and shaded region) and responses of a 0.1-dB ripple all-
pole Chebyshev filter (Chebyshev), an elliptic Cauer-Chebyshev (Cauer-Cheby), and a
custom filter designed by S/Filter that just satisfies the requirements.
Figure 3.2 A 0.1-dB passband ripple Cauer-Chebyshev (elliptic) filter that satisfies the above
filter specification.
C3
ZO=50Ω C=176.58pF ZO=50Ω
L1 L2 L3
L=360.59nH L=117.1nH L=360.59nH
C1 C2 C4 C5
C=119.53pF C=136.34pF C=136.34pF C=119.53pF
Figure 3.4 A 0.1-dB passband ripple custom filter designed using S/Filter to synthesize a
solution with user-specified TZs.
f0 = fu fl (4.2)
This frequency is used to scale the lowpass prototype values. Then the
fractional bandwidth is
bw = BW (4.3)
f0
The percentage bandwidth is the fractional bandwidth times 100%.
These parameters are used to calculate element values using the modern
method. The fractional bandwidth has significant repercussions in the
design of bandpass filters. Small fractional bandwidth results in extreme
element values and high dissipative insertion loss.
17
18 Filter Synthesis Using Genesys S/Filter
infinity are classified as antimetric filters. Notice that both classes have an
even-quantity sum of TZs. Filters with an odd-quantity sum of TZs are
classified as parametric filters. This class of bandpass filter requires
approximate techniques and it is not designed by S/Filter.
Figure 4.1 Specification tab for the 75- to 125-MHz Chebyshev bandpass.
L2 C2
ZO=50Ω L=174.68nH C=15.47pF ZO=50Ω
C1 L3
C=53.31pF L=50.69nH
L1 C3
L=50.69nH C=53.31pF
Figure 4.2 Schematic and responses of the 75- to 125-MHz bandpass filter.
L2 C2
ZO=50Ω L=1091.74nH C=2.32pF
ZO=50Ω
C1 L3
C=333.17pF L=7.62nH
L1 C3
L=7.62nH C=333.17pF
Figure 4.3 Schematic and responses of the 96- to 104-MHz bandpass filter.
The ratio of the insertion losses is equal to the ratio of the bandwidths.
The increase of the insertion loss with decreasing bandwidth is inherent
with bandpass filters and this loss does not depend on the filter topology or
whether the filters are designed using the modern method or direct
synthesis. Improving the insertion loss requires elements or resonators
with improved unloaded Q [1]. Achieving good resonator Q often requires a
manufacturing process other than lumped inductors and capacitors (L-C),
such as coaxial resonators and quartz resonators. This may require all
parallel or all series resonators forms. Direct synthesis of filters with all
equal-type resonators is covered in Chapter 8.
Next, consider the group-delay responses. The group delay at band
center is 8.06 nS for the wider filter and 54.54 nS, approximately inversely
proportional to the bandwidth.
Consider the symmetry of the group-delay responses and the rejection
in the transition regions. The wide filter is highly asymmetric in group
delay and the steepness of the transition region below the passband is
greater than above the passband. This asymmetry is topology-dependent,
and direct synthesis offers effective methods for dealing with these issues,
as is discussed in Chapter 11.
Finally, consider the element values. Notice that the ratio of the largest
to smallest inductor value is 3.45 with the wider bandwidth filter and the
ratio is 143.3 for the narrower filter. While the former ratio is manageable,
realizing the inductor values in the narrow filter is difficult. For lowpass
and highpass filters, the inductor and capacitor ratios are equal to the ratio
of lowpass prototype g-values. Element values are closely related to the
termination resistance for a given cutoff frequency, naturally resulting in
Conventional Bandpass 21
Reference
[1] R.W. Rhea, HF Filter Design and Computer Simulation, SciTech
Publishing, Raleigh, NC, 1994.
5 Extraction Sequences
When a filter is synthesized, TZs are extracted into inductors or
capacitors. The extraction sequence of the previous filters with three TZs at
DC and three TZs at infinity is DC ∞ DC ∞ DC ∞. However, with direct
synthesis the TZs may be extracted in other sequences. This offers
alternative topologies and element values that may be more desirable from
a manufacturing standpoint. The Extraction tab of S/Filter is explored in
this chapter using the previous all-pole bandpass as an example.
In this case there are six total TZs, no FTZs, three TZs at DC, and three TZs
at infinity, or 20 permutations. However, some permutations are not
unique. For example, the sequence DC ∞ DC ∞ DC ∞ and the sequence ∞
DC ∞ DC ∞ DC are essentially identical because they only change the order
of the elements in each shunt and each series resonator. In addition, for
symmetric filters, the left and right ports may be exchanged resulting in an
identical filter with the same element values and only exchanged positions.
For this filter, six unique permutations exist. The filter shown at the middle
left in Fig. 5.2 as DC ∞ ∞ DC DC ∞ has eight different extraction sequences
with an identical schematic because the elements may be reversed in each
branch, and because this filter is electrically symmetric. This is the all-pole
bandpass filter topology of the filter designed by the modern method. The
extraction sequence shown as DC DC DC ∞ ∞ ∞ has four different
extraction sequences with an identical schematic because the elements may
be reversed in the center branch, and because this filter has mirror
symmetry. The four remaining extraction sequences each have two
identical schematics because elements may be reversed in one branch.
These four sequences require transformers or dissimilar termination
resistances. Transformers are dealt with in a later chapter.
23
24 Filter Synthesis Using Genesys S/Filter
Figure 5.1 Extraction tab for the 75- to 125-MHz bandpass filter.
C1 L3 C2
L1
C=68.8pF L=39.3nH ZO=50Ω C=305.8pF ZO=50Ω
L=39.3nH
L1 L2 C2 C3 C1 L3 C3
L=50.7nH L=11.4nH C=237pF C=53.3pF C=53.3pF L2 L=2.6nH C=1053.9pF
L=11.4nH
T1
P=1
S=4.4
C2 C1
L2 L2
C=15.5pF ZO=50Ω ZO=50Ω C=68.8pF ZO=50Ω
ZO=50Ω L=174.7nH L=8.8nH
L1 C1 L3 C3 L1 C2 L3
L=50.7nH C=53.3pF L=50.7nH C=53.3pF L=50.7nH C=237pF L=2.6nH
C3
C=1053.9pF
T1
P=1
S=4.4
C2 C3
L3 L2
ZO=50Ω C=15.5pF ZO=50Ω ZO=50Ω C=3.5pF ZO=50Ω
L=776.7nH L=174.7nH
L1
L=50.7nH C3 C2 L3
C=2.7pF C=12pF L=1002nH
C1 L2 C1
C=53.3pF L=225.4nH C=53.3pF
L1
L=50.7nH
T1 T1
P=1 P=1
S=0.2 S=0.2
Figure 5.2 The six unique solutions to the bandpass filter with three TZs at DC and three TZs
at infinity.
C2
L2
ZO=50Ω C=4.6pF ZO=50Ω
L=592.6nH
L3
C1 L=133.3nH
C=20.3pF
L1 C3
L=171.9nH C=15.7pF
L1 L3
L=133.3nH L=133.3nH ZO=50Ω
L2 C2
L=38.7nH C=69.9pF
Figure 5.3 Duals with a series element first of the two filters in Fig. 5.2 that do not have
transformers.
Reference
[1] Genesys 2012: Synthesis, Agilent Technologies, Santa Clara, CA 2012, p. 292.
6 Customized Bandpass Filters
This chapter describes filter customization by the independent
specification of TZs at DC, infinity and finite frequencies. This chapter also
describes the tools within S/Filter used to find the optimum solution from
potentially thousands of unique topologies.
29
30 Filter Synthesis Using Genesys S/Filter
Next, the Extraction tab is utilized to select the optimum sequence. The
Customize Table button is clicked to open the Customize Permutation
Table dialog. The manufacturability goals for this filter involved inductors
and the transformer, so Maximum C Value and Minimum C Value are
selected in turn and shifted left to the Available column. Then Lmax/Lmin
Ratio and Transformer Turns Ratio are selected in turn and shifted right
to the Show in Table column and OK is clicked. The results are displayed in
Fig. 6.2.
Figure 6.3 Extraction tab for the custom bandpass filter example.
Figure 6.4 Schematic of the custom bandpass filter with transformer removed.
Figure 6.5 Amplitude transmission and reflection responses of the custom bandpass.
Figure 6.7 Specification tab for the 50- to 90-MHz bandpass filter.
The Extraction tab for the exact solutions is given in Fig. 6.8. The
Series Element First type is selected because the inductor values for its dual
are rather small. There are 720 permutations and 57 unique solutions. The
number of unique solutions is a small fraction of the total permutations
because there are only two nonfinite TZs and the elements in each of the
FTZs may be exchanged. To find the best solution, the Permutation Table is
sorted as before. One of the solutions with the minimum number of
inductors, no transformer, and the smallest Lmax/Lmin is selected. The
Customized Bandpass Filters 35
Figure 6.8 Exact extractions for the 50- to 90-MHz bandpass filter.
C1 C3 L3 C5 L5
C=33.7pF C=8.8pF L=3316.5nH C=1.7pF L=584.6nH ZO=50Ω
L7
ZO=50Ω L=140.9nH
L1 L2 L4 L6
L=261.3nH L=979nH L=86.5nH L=64.2nH
C2 C4 C6 C7
C=200.3pF C=26.9pF C=7.5pF C=19pF
Figure 6.9 Schematic of the exact solution to the 50- to 90-MHz bandpass filter.
Figure 6.10 Responses of the exact extraction (left) and the inexact extraction before
optimization (dashed traces on the right graph) and after optimization (solid traces on the right
graph).
Figure 6.11 Extraction tab with inexact permutations included in the display.
The schematic for this inexact permutation is given in Fig. 6.12. The
filter is significantly superior to the best exact solution. Although the
quantity of capacitors is increased from seven to nine, the quantity of
inductors is reduced from seven to five and Lmax/Lmin is only 3.0, 17
times lower than the exact solution, thus significantly enhancing inductor
design options.
Figure 6.12 Schematic of the inexact bandpass filter with values after optimization.
7 Norton Transforms
Filter economy and manufacturability are often enhanced by the
application of transforms. Scores of transforms are integrated into S/Filter,
thus simplifying their application to specific filters. They are accessed
through the Transform tab. Experience with transforms is essential to the
mastery of filter design. The transforms are listed in the S/Filter
documentation so a transform-by-transform description is not repeated
here. Some of the more important transforms used to solve common filter
design problems are illustrated throughout the remainder of the book. The
Norton transforms are essential and the basis of many other “macro”
transforms. For example, conversions between Pi, Tee, and L-network
conversions use Norton transforms. Therefore, this chapter is devoted to a
description of the Norton.
Z
Zc = (7.3)
n (n − 1)
where n is the turns ratio of the transformer. In S/Filter dialog windows, n
is referred to variously as n or N. Notice that when n>1 then Za is negative,
and when n<1 then Zc is negative. The obvious question is why would it be
desirable to utilize the Norton when one reactor is replaced with a negative
reactor, two positive reactors, and a transformer. The answer is best
illustrated with an example.
Figure 7.3 The S/Filter Transform tab with Swap element selected.
Figure 7.4 Norton Transform dialog for specification of the transform parameter.
42 Filter Synthesis Using Genesys S/Filter
Figure 7.5 Schematic of the 55- to 85-MHz bandpass filter after application of the series
Norton.
Next, the Simplify Circuit button in the Transform tab is clicked. This
function combines all transformers and combines like-parallel and like-
series elements. S/Filter determines the resulting transformer has a unity
turns ratio and eliminates it. In addition, S/Filter absorbs the negative
capacitor C6 into C4, thus eliminating C6. The final schematic is given in Fig.
7.6. It is unnecessary to swap C6 and the L3/C5 branch before applying the
Simplify Circuit transform. The Simplify Circuit transform is schematically
global in that it searches the entire schematic during simplification.
Compare the original schematic in Fig. 7.2 with the final schematic in
Fig. 7.6. Inspection of the Norton series diagram in Fig. 7.1 suggests that
the Norton transform replaces a single capacitor with three capacitors and
a transformer. In reality, when applied at the proper position in the filter,
the affect is to replace a transformer with a more economical capacitor.
When applied to an inductor, the Norton replaces a transformer with an
inductor. In essence, the Norton transform is an impedance
transformation.
Norton Transforms 43
Figure 7.6 Schematic of the 55- to 85-MHz bandpass filter after simplification.
1− n
Z c = 2 Z (7.6)
n
For the Norton shunt transform, if n<1 then Za is negative and if n>1
then Zc is negative.
Figure 7.8 Schematic after the application of the Norton shunt transform to the original C1.
Figure 7.9 Final schematic of the 55- to 85-MHz bandpass filter with equal-valued inductors.
47
48 Filter Synthesis Using Genesys S/Filter
ZO=50Ω ZO=50Ω
L1 L2 L3 C6
L=100nH L=100nH L=100nH C=7.89pF
C2 C4
C=7.89pF C=15.17pF
C2 C4 C6
L1 L2 L3
ZO=50Ω L=100nH C=57.86pF L=100nH C=42.56pF L=100nH C=57.86pF ZO=50Ω
C1 C3 C5 C7
C=48.95pF C=122.15pF C=122.15pF C=48.95pF
Figure 8.1 A 90- to 110-MHz top-C coupled parallel-resonator bandpass (top) and shunt-C
coupled series-resonator bandpass (bottom).
Given in Fig. 8.2 are the responses of the 90- to 110-MHz bandpass
filters designed in the Passive Filter module. The responses of the top-C
coupled bandpass are given on the left, and the responses of the shunt-C
coupled bandpass are given on the right. Inspection of the top-C coupled
bandpass schematic in Fig. 8.1 reveals that the quantity of TZs at DC is
seven and the quantity of TZs at infinity is one. Therefore, the selectivity of
the response on the left in Fig. 8.2 is much greater below the passband. In
the shunt-C coupled bandpass, the quantity of TZs at DC is one and the
quantity at infinity is seven. Therefore, it has greater selectivity above the
passband.
The return loss peaks in the passband are not uniformly 20 dB. This is
a consequence of errors in the inverter algorithms at 20% bandwidth.
Notice that the center frequency is shifted slightly downward for the top-C
coupled filter and slightly upward for the shunt-C coupled filter. The
Passive Filter module of Genesys includes correction algorithms not
included in Matthaei, Young, and Jones. Nevertheless, above 20%
bandwidth the inverter errors become significant, and the correction
algorithms begin to fail. In the next example, S/Filter is used for the exact
synthesis of a parallel-resonator bandpass filter.
Bandpass with Resonators 49
Figure 8.2 Responses of the 90- to 110-MHz top-C coupled bandpass (left) and the shunt-C
coupled bandpass (right).
C1 L3
C=135.52pF L=18.88nH
L1
C3
L=18.88nH
C=135.52pF
Figure 8.3 Conventional 0.0432-dB 90- to 110-MHz passband filter with three TZs at DC and
three TZs at infinity.
Next, two series Norton transforms are used to convert the series
resonator to a parallel resonator. First, inductor L2 is selected and a series
50 Filter Synthesis Using Genesys S/Filter
L4 L5
L1 L2 L=18.88nH L=18.88nH
L=18.88nH C1 L=-100.96nH C3
C=135.52pF C=135.52pF
T1
P=1
S=5.35
Figure 8.4 Conventional bandpass after application of a series Norton to the original L2.
L2 C3
ZO=50Ω L=82.08nH C=31.17pF ZO=50Ω
L3
L=18.88nH L4
L1 C1 L=18.88nH
L=23.22nH C=135.52pF C2 C4
C=135.52pF C=110.18pF
n 4 Rs2 + X series
2
X shunt = (8.2)
X series
C1 L2 C4 C6
C=13.9pF C=4.9875pF C=13.9pF
L=513.0034nH
ZO=50Ω ZO=50Ω
L1 L3 L4 C5
L=145.14nH L=118nH L=118nH C=6.2708pF
C2 C3
C=10.4121pF C=21.68pF
These are false TZs, and their influence on the response is dependent on the
degree of impedance shift. Notice that the response in Fig. 8.9 has
somewhat greater selectivity below the passband.
Figure 8.9 Responses of the parallel-resonator bandpass with series termination capacitors.
C1 L1 C3 L2 C4 L4
C=9.29pF C=7.55pF C=7.55pF
L=338.81nH L=338.8nH L=275.45nH
ZO=50Ω ZO=50Ω
C2 L3
C=32.83pF L=77.93nH
Figure 8.11 History tab for the exact series-resonator 90- to 110-MHz bandpass.
n 2 Rs
X shunt = (8.3)
n2 − 1
−1
1 X
X series = + 4shunt2
(8.4)
X shunt n Rs
where n is the transformer turns ratio, Rs is the source resistance, Xseries is
the reactance of the series element, and Xshunt is the reactance of the
negative shunt element. At the source, n values greater than 1 are used. The
resonator-inductor values in series-resonator filters are reasonable except
for wide bandwidth, and the termination-coupling transform is less
frequently required for this filter.
Bandpass with Resonators 55
C2 C3 C4
ZO=50Ω C=4.99pF C=98.6e-3pF C=3.27e-3pF
ZO=50Ω
L2
L=518.13nH
L1 C1
L=17.27nH C=147.41pF L3 L4
L=25883.85nH L=804909.95nH
T1
P=1
S=4.63e-3
Figure 8.13 Initial bandpass with seven TZs at DC and one at infinity.
56 Filter Synthesis Using Genesys S/Filter
Figure 8.17 Responses of the filter with like-coupling elements and termination coupling.
58 Filter Synthesis Using Genesys S/Filter
C1
L1 L2 L3
C=7.64pF ZO=50Ω
L=338.83nH L=14.73nH L=0.6nH
ZO=50Ω
C2 C3
C=173.78pF C=4127.52pF
T1
P=1
S=23.73
Figure 8.18 Initial canonic bandpass with one TZ at DC and five at infinity.
C1 C3 C5
L1 L2 L3
C=9.17pF C=12.08pF C=9.19pF
L=338.83nH L=338.83nH L=338.1nH
ZO=50Ω ZO=50Ω
C2 C4
C=36.23pF C=36.27pF
Figure 8.19 Series-resonator bandpass after application of two shunt Norton transforms to the
shunt capacitors.
C1 C3 C5 C7
C=29.9pF C=68.19pF C=68.27pF C=29.9pF
Figure 8.20 Series-resonator bandpass with termination coupling capacitors to reduce and set
the inductors at a specific value.
C2 C5
ZO=50Ω C=3.18pF C=3.19pF ZO=50Ω
L1 L2 L3
L=338.83nH L=338.83nH L=338.1nH
C1 C4
C=4.78pF C=12.58pF
C3 C6
C=12.58pF C=4.79pF
Figure 8.21 Filter topology known as the tubular bandpass created from the series-resonator
bandpass.
L3 C2
ZO=50Ω C=1.57pF ZO=50Ω
L=1624.47nH
L1
L=135.68nH C4
C=23.93pF
L2 L4
L=233.42nH L=196.94nH
T1
C1 C3 P=1
C=6.04pF C=35.73pF S=1.39
Figure 8.24 Generalized bandpass of the sixth degree with FTZs below and above the
passband.
L1 C1 L3 C3 L5 C5
L=252.39nH C=12.09pF L=260.42nH C=9.52pF L=242.28nH C=8.71pF
ZO=50Ω ZO=50Ω
L2 L4
L=116.71nH L=136.58nH
C2 C4
C=12.09pF C=51.52pF
Figure 8.28 Responses of the parallel-resonator bandpass filter with tuning with capacitive
coupling elements (left) and with inductive coupling elements (right).
Figure 8.29 Goals for the simultaneous optimization of the tunable filter at three frequencies.
Reference
[1] G. Matthaei, L. Young, and E.M.T. Jones, Microwave Filters, Impedance-
Matching Networks, and Coupling Structures, Artech House, Dedham, MA, 1980.
9 TEM-Mode Resonators
With narrow bandpass filters, insertion loss often dominates the design
process. The solution requires higher unloaded (element) Q. Element Q is
typically dominated by metal conductivity. Higher volume increases
element Q. For a given volume, solenoid inductors typically have a higher
element Q than distributed transmission line elements (TLEs). However,
parasitics limit the allowable physical size of wire-wound inductors.
Transmission lines have higher unloaded Q than inductors when they are
physically larger [1].
Hybrid filters combine L-C elements and TLEs. These filters are readily
designed using the Lumped to Distributed Equivalent group of transforms
in S/Filter. Lumped and distributed equivalents and their theory are
covered in HF Filter Design and Computer Simulation [2]. These
transforms are approximate, but over the narrow bandwidth typically
involved in this filter class, they are sufficiently accurate. In this chapter,
the parallel-resonators discussed in Chapter 8 are transformed into TLE
resonators in order to provide higher resonator unloaded Q.
69
70 Filter Synthesis Using Genesys S/Filter
f0
Ql = (9.2)
BW
At midband, the dissipative insertion loss of all-pole bandpass filters is
n=N
4.34Ql
IL( dB ) =
Qu − ave ∑g
n =1
n (9.3)
Figure 9.1 Five-resonator narrow bandpass filter to be converted to a coaxial resonator filter.
Figure 9.2 Narrow bandpass after the termination coupling transforms to increase the value of
the inductors.
The values of the internal coupling capacitors are very small. If the
coaxial resonators are to be coupled using capacitive probes, these values
are appropriate. If the coaxial resonators are to be mounted on a printed-
circuit boad (PCB), the coupling capacitors could be printed as interdigital
microstrip gaps. If chip capacitors are to be used, the values are
unreasonable. This is addressed in the next section covering microstrip
resonators.
The next step is converting the parallel L-C resonators to coaxial
resonators using the Lumped to Distributed Equivalents: Parallel LC to
Ground to Grounded Stub transform. The result is given in Fig. 9.3. The
transform in S/Filter uses a transform frequency equal to the resonant
frequency of the parallel L-C network. This is appropriate for many
applications. However, this results in a slightly different characteristic
impedance for each distributed resonator. Here, 50-ohm characteristic
impedance is desired for each resonator. This is resolved by manually
changing each resonator to 50 ohms and optimizing the response. The
72 Filter Synthesis Using Genesys S/Filter
Figure 9.3 Narrow bandpass with parallel L-C resonators transformed to coaxial resonators.
Given in Fig. 9.4 is the filter after setting each resonator to 50 ohms
and optimizing the length of the resonators.
Figure 9.4 Narrow bandpass after setting each resonator to 50 ohms and optimizing the
resonator lengths.
This routine coverts the grounded stub transmission line model with
electrical parameters into the selected physical transmission line model, in
this case coax, using the previously defined substrate material Air. The Air
material is defined by adding a Substrate in the workspace tree of Genesys.
The dielectric in this case is air. The coax outer diameter determines the
line loss and therefore the unloaded Q. Earlier, S/Filter was used to
determine that the required unloaded Q required to achieve a dissipative
insertion loss of 2.5 dB is 505. The unloaded Q of a quarter wave coaxial
resonator due to conductor loss is
b ln (b a )
Qc = 1.513 × 10 4 f ( MHZ ) (9.5)
1+ b a
Figure 9.6 Coaxial bandpass with electrical line models converted to physical line models.
Figure 9.8 Specification tab for the generalized bandpass using ceramic-loaded coaxial
resonators.
C2
C=6.74pF C4 C5
ZO=50Ω C=9.26e-3pF C=20.1e-6pF ZO=50Ω
L2
L=3.41nH
L4
C1 C3 L=1.33e+6nH
C=58.63pF C=0.96pF
L1
L=0.45nH L3
L=2888.85nH
T1
P=1
S=581e-6
Figure 9.9 Initial L-C filter for the proposed generalized ceramic-loaded coaxial resonator.
are absorbed into L1 and C1 and positive L-C values occur on the right. The
turns ratio value is not critical as the shunt inductors are equalized later.
Here, n equal to two is entered. Series Norton transforms are then applied
to the two remaining series capacitors to place capacitance in parallel with
the remaining shunt inductors. Again, n is not critical, and a value of 10 is
specified. Next, Remove Existing Transformer is applied. Finally, the
transform Compound Operations: Equate All Shunt Ls or Shorted Stubs is
applied. The schematic with parallel shunt resonators is given in Fig. 9.10.
Figure 9.10 Generalized bandpass before the termination coupling and coaxial-resonator
transforms.
The two internal series coupling capacitors are a small value. If this is
manageable, then the steps in the next section are skipped.
REM-Mode Resonators 77
Figure 9.12 Final schematic of the generalized ceramic-loaded coaxial resonator bandpass
with one FTZ.
Figure 9.14 Transform steps for the four-section ceramic-resonator bandpass with FTZs
above and below the passband.
80 Filter Synthesis Using Genesys S/Filter
Figure 9.15 Initial synthesized ideal lumped bandpass with two FTZs (right) and after
transformation to coaxial resonators (left).
References
[1] R. Rhea, “Technology Enables New Components,” Microwave Journal,
November 2006.
[2] R. Rhea, HF Filter Design and Computer Simulation, Noble
Publishing(SciTech), Raleigh, NC, 1994.
[3] G. Matthaei, L. Young, and E.M.T. Jones, Microwave Filters, Impedance-
Matching Networks, and Coupling Structures, Artech House, Dedham, MA, 1980.
10 Piezoelectric Devices
Quartz and ceramic bulk-mode vibrational resonators offer extremely
high-unloaded Q and excellent temperature and long-term stability. These
attributes are critical for very narrow bandwidth applications. However, the
equivalent electrical model parameters for these mechanical-vibration
devices pose unique issues. This chapter illustrates bandpass filters that
utilize these devices and includes a larger number of examples than
previous chapters because of the unique difficulties in synthesizing filters
with piezoelectric devices.
Figure 10.1 Diagram of quartz-crystal resonator illustrating electrodes on a quartz blank, and
the fundamental and overtone shear vibrational modes (left), and a photograph [1] of the
resonator and a metallic housing (right).
83
84 Filter Synthesis Using Genesys S/Filter
Lm Cm Rm
Co
f 2
p
C m = C0 − 1 (10.1)
s
f
and
1
Lm = (10.2)
(2πf s )2 Cm
86 Filter Synthesis Using Genesys S/Filter
IL
Rm = 2Z 0 10 20 − 1 (10.3)
where Z0 is the characteristic impedance of the measurement system and IL
is the decibel insertion loss at the series-resonant transmission peak.
Figure 10.4 Specification tab for the approximate design four-section quartz-crystal bandpass.
The initial synthesis with a series element first and the extraction
sequence ∞ DC ∞ ∞ ∞ ∞ ∞ ∞ is given in Fig. 10.5. The topology is identical
to the shunt-C coupled series-resonator bandpass filter introduced in
Section 8.2, and the design proceeds as before. Shunt-Norton transforms
are applied to the shunt capacitors to equate the series inductors and
remove the transformer. The element values are displayed with eight-digit
precision because of the extremely high Q of the design. The high stability
of quartz justifies this expectation of precision.
C1
ZO=50Ω C=0.0902371pF ZO=50Ω
L1 L2 L3 L4
L=3.0452454e+6nH L=0.0934436nH L=1.1876835e-9nH L=36.444144e-18nH
C2 C3 C4
C=2.9407491e+6pF C=231.36976e+12pF C=7.5401426e+21pF
T1
P=1
S=289.06617e+9
The results after the three shunt-Norton transforms are given in Fig.
10.6. The value of the series inductors is approximately 3 mH, far too large
to realize as wire-wound inductors but consistent with quartz crystals.
88 Filter Synthesis Using Genesys S/Filter
The next step is shifting the impedance level higher by the ratio of the
quartz-resonator motional inductance and these schematic values. In this
case, the termination resistance is increased to 198.5 ohms. The results are
given in Fig. 10.7.
Figure 10.7 Quartz-crystal bandpass after increasing termination resistance to 198.5 ohms.
X1 X2 X3 X4
R=11.6Ω R=11.6Ω R=11.6Ω R=11.6Ω
L=12.092mH L=12.092mH L=12.092mH L=12.092mH
Cm=.02273182pF C1 Cm=.02273182pF Cm=.02273182pF Cm=.02273182pF C7
Co=5.45pF C=200pF Co=5.45pF Co=5.45pF Co=5.45pF C=200pF
ZO=200Ω ZO=200Ω
C2 C4 C6
C=130pF C=200pF C=130pF
Figure 10.8 Quartz-crystal bandpass with standard values and identical quartz resonators.
Figure 10.9 Responses of the four-section quartz crystal 9.6- to 9.602-MHz filter designed by
approximate methods.
90 Filter Synthesis Using Genesys S/Filter
Figure 10.11 Responses of the quartz-crystal bandpass with inductors to null the static
capacitance. Monte Carlo analysis displays the results of multiple simulations with 5%
tolerance-nulling inductors.
Figure 10.15 Schematic of the lower-sideband filter after the termination coupling and final
dual transforms are applied.
The next step creates capacitors in series with the first parallel-
resonant branch by applying a shunt-Norton transform to capacitor C2 in
Fig. 10.15. Any transform turns ratio is acceptable that results in a left
series capacitor that can be absorbed in capacitor C1. One option is selecting
94 Filter Synthesis Using Genesys S/Filter
which in this case is 0.88789169. If n is just slightly larger than this value,
the resulting capacitor value is extremely large and is deleted
The left side part of the schematic resulting from this transform is given
in Fig. 10.16. The form of the branch with L1, C3 and C4 is referred to as a
dipole. With appropriate values, it is exactly equivalent to the quartz-crystal
model dipole given in Fig. 10.3.
C4
C=207975.150197pF
C1
C=2.2843906e+9pF
C3
C=169.928106pF
L1
ZO=50Ω L=1.319633nH
C2 C5
C=21.455715pF C=810.530659pF
Figure 10.16 Left-side portion of the schematic resulting from the application of the shunt-
Norton transform.
C1 = C4 1 + C4 (10.5)
C3
Piezoelectric Devices 95
C 2 = C3 + C 4 (10.6)
C
2
L1 = L2 3 (10.7)
C3 + C4
C2
C3 = (10.8)
C1
1 + C
C2
C4 = (10.9)
1 + C 2
C1
2
(C + C2 )
L2 = L1 1 (10.10)
C2
Substituting the values of the 9.6-MHz quartz-crystal resonator in
Section 10.2, the required values of the left dipole in Fig. 10.16 are
C1=1315.758 pF, C2=5.472670 pF, and L1=208.5877 nH. The value of L1 in
Fig. 10.14 is now 1.319633 nH. To realize this filter using quartz-resonators
similar to the 9.6-MHz quartz-crystal resonator requires that this value is
raised to 208.5877 nH. This is accomplished by increasing the termination
resistance in the Specification tab by the ratio of these values. Therefore,
the termination resistance is increased to 7903.247 ohms.
C3 C6 C9 C12
C=1315.757626pF C=1315.565905pF C=1315.114158pF C=1314.498515pF
L1 L2 L3 L4
ZO=7903.247Ω
L=208.587744nH L=208.587744nH L=208.587744nH L=208.587744nH
C2 C5 C8 C11 ZO=1481.042992Ω
C=1.389559pF C=3.245273pF C=3.353582pF C=3.018671pF
C1 C4 C7 C10
C=0.13574pF C=2.465795pF C=5.13692pF C=8.056066pF
Figure 10.18 Lower-sideband quartz-crystal filter after equating all inductors, deleting the
original C1 and absorbing the transformer into the load resistance.
and absorbing the transformer into the loaded resistance is given in Fig.
10.18. The values of C2, C5, C8, and C11 result in static capacitance in the
quartz-crystal resonator dipoles that are smaller than the desired quartz-
crystal model. Therefore, these capacitors are split into two capacitors, one
that becomes the static capacitance and one that can serve as tuning
elements. Using Eqs. 10.8 through 10.10, the required value of capacitance
in series with L1 and C3 in Fig. 10.18 that results in a static capacitance of
5.45 pF in the quartz-crystal dipole is found to be 5.47267 pF. The
transform Basic Operations: Split Series Element: Specify Left Value is
applied with a left value of 5.47267 pF. The remainder is 1.862453 pF.
These steps are applied to the remaining three resonator branches. The
results after converting the dipoles to quartz-crystal resonator models are
given in Fig. 10.19.
Given in Fig. 10.21 are the responses of the final lower-sideband filter
with unloaded inductor Q specified as 100. This Q has a small influence on
the response as it improves the return loss without adding significant
insertion loss. Unloaded capacitor Q in excess of 1000 has very little effect
on the responses.
Given in Fig. 10.22 is the initial Specification tab for the proposed
upper-sideband filter. The TZs are placed to achieve the desired responses
given in Fig. 10.23.
The initial schematic for the extraction sequence 9.599 9.5983 9.5948
DC ∞ is given in Fig. 10.24. Th e strategy is to apply a series Norton
transform to the first three series capacitors to place capacitance in parallel
with each series L-C to ground. These branches then have a topology that
conforms to the quartz-crystal model. However, applying a series Norton to
C1 with a positive capacitor to its right results in an unrealizable negative
capacitor in shunt at the input. Therefore, it is necessary to create a shunt
capacitor at the input that later will absorb the negative capacitance created
by the series Norton. To create the input shunt capacitor a Termination
Coupling: Source: Series C to Shunt C transform is applied to C1. This
transform is approximate, but at the extreme narrow bandwidth of a
quartz-crystal filter, the transform is nearly exact. The result with n=1.6 is
given in Fig. 10.25. The value of n is not critical and is selected between the
limits suggested by the S/Filter dialog.
Piezoelectric Devices 99
2) The source and load resistance in the Specification tab are increased
by the ratio of the desired motional inductance, 12.092 mH, to the
current value of L1. The resistance is 345.743738 ohms. L1 now equals
the desired motional inductance. C3 is too large to equal the static
capacitance. This is dealt with later.
3) A series Norton is applied to C5 with n equal to the square root of the
ratio of L2 and the desired motional inductance. n is 2.378502692.
4) A series Norton is applied to C8 with n equal to the square root of the
ratio of L3 to the desired motional inductance. n is 4.099802122. At
this point, each of the shunt inductors equals the desired motional
inductance, but L4 is too large in value.
5) C10 is swapped with the L3/C9 branch.
6) C10 is split, specifying the left value as the quartz-crystal static
capacitance of 5.45 pF.
7) A shunt Norton is applied to capacitor C11 with n equal to the square
root of the ratio of L4 to the motional inductance. n is 8.035702256.
8) C3 is then split with the left value specified as 5.45 pF. It is then
swapped to be adjacent to the series L-C to ground.
9) C7 is treated in a similar fashion.
10) The transformer is absorbed into the load.
The result of these transforms is given in Fig. 10.26. The shunt
resonators are all of the form of quartz-crystal resonators with the desired
static capacitance and motional inductance. The final series resonator is
also the desired motional inductance. Each of these quartz crystals is
manufactured with the series resonant frequency equal to the resonant
frequency calculated from the motional inductance and capacitance of that
branch. The range of frequencies is 9594.727 to 9599.580 kHz. The series
resonator does not have a parallel capacitor corresponding to the static
capacitance. This is dealt with shortly.
Next, each corresponding group of elements is replaced by a quartz-
crystal model with motional resistance as shown in Fig. 10.27. The static
capacitance of the series resonator is included. In addition, each discrete
capacitor is replaced with the nearest standard value. These replacements
have only a minor effect on the return loss.
Piezoelectric Devices 101
Figure 10.25 The upper-sideband filter after application of the Termination Coupling transform
at the input.
Figure 10.26 Upper-sideband filter after application of transforms preparing for replacement of
resonators with quartz crystals.
102 Filter Synthesis Using Genesys S/Filter
Figure 10.27 The upper-sideband quartz-crystal filter with standard values for the discrete
capacitors.
The responses of this final filter are given in Fig. 10.28. The motional
resistance has introduced an approximately 1.7-dB insertion loss. The
traces labeled “with Co” are with the static capacitance added to the series
quartz crystal, and the traces labeled “original” are with a 50.3-uH nulling
inductor in parallel with the quartz crystal. The designer may choose the
more desirable response. The fact that the termination resistances are not a
standard value is dealt with as in Section 10.2.
Figure 10.29 Specification tab for a quartz-crystal filter with one TZ below and one TZ above
the passband.
Figure 10.30 Desired responses of the quartz-crystal filter with two FTZs.
Piezoelectric Devices 105
Figure 10.31 Initial synthesized schematic of the quartz-crystal filter with two FTZs.
The resulting schematic is given in Fig. 10.32. L2, C4, and C5 are ready
for replacement with a quartz-crystal resonator. When the network of C8,
C9, and L3 is replaced with the quartz crystal using Eqs. 10.8 through 10.10,
the motional values are correct. The required parallel capacitance is
13.22806 pF, of which 5.45 pF is the quartz-crystal static capacitance and
7.77806 pF is an added capacitor in parallel with the quartz crystal. The
three remaining series resonators are replaced with quartz crystals, and the
static capacitance is nulled using parallel inductors as before.
The final schematic with quartz-crystal resonators and nulling
inductance is given in Fig. 10.33. In addition, discrete capacitors and the
termination resistance are set to standard values. An exception to standard
values is the capacitor C5 in parallel with X3. Unless the return loss is not
critical, this value is somewhat sensitive and possibly needs trimming in
production. The series-resonant frequency specified for manufacture of the
quartz-crystals range from 9593.848 to 9600.214 kHz.
The responses for the final quartz-crystal filter are given in Fig. 10.34.
Standard values, dissipation, and effects of the nulled static capacitance
shift the passband up about 50 Hz. This could be compensated by shifting
the resonant frequency specification of each quartz crystal down 50 Hz.
106 Filter Synthesis Using Genesys S/Filter
Figure 10.32 Schematic of the filter with two FTZs ready for quartz-crystal resonators.
Figure 10.33 Final quartz-crystal filter with FTZs below and above the passband.
Piezoelectric Devices 107
Figure 10.34 Responses of the final quartz-crystal filter with FTZ below and above the
passband.
The response of the resulting filter is given in Fig. 10.36. The finite
unloaded Q of the susceptance canceling inductor results in limited
attenuation at the TZ above the passband. Otherwise, the 10-kHz
bandwidth quartz-crystal design is successful.
Figure 10.36 Responses of the wide-bandwidth quartz-crystal filter with partial nulling of X3 to
realize a small static capacitance requirement.
L4 L5 L6
L=50.3μH L=50.3μH L=50.3μH
ZO=2914.5Ω X1 X3 X2 ZO=2914.5Ω
R=11.6Ω R=11.6Ω R=11.6Ω
L=12.092mH L=12.092mH L=12.092mH
Cm=0.022591pF Cm=0.022687pF Cm=0.022591pF
Co=5.45pF Co=5.45pF Co=5.45pF
C2 C4
C=5.306757pF C=5.306762pF
The responses are given in Fig. 10.38. The results are satisfactory from
about 9.25 to 10.05 MHz. However, well below and well above the passband
the rejection decreases. Examination of the schematic in Fig. 10.37 reveals
that when the static capacitance of the quartz crystals is considered, the
quantity of TZs at DC is zero and the quantity at infinity is only one.
Consequently, the attenuation at these frequency extremes is very poor.
Notice that the very wide-bandwidth results in a high termination
resistance. This is used to advantage in the next step by adding matching
networks with a loaded Q of 50 at the input and output to match this filter
to 50-ohm terminations. The schematic after optimization for best response
and placing elements on standard values is given in Fig. 10.39 and the
responses are given in Fig. 10.40 with an unloaded Q of 100 for all
inductors. The high Q of the matching networks results in most of the
passband insertion loss and probably requires tuning of one of the elements
in each matching network.
Although narrow bandwidth quartz-crystal filters are less susceptible to
stopband issues, whenever inductors are used to null the static capacitance,
it is wise to simulate the attenuation well into the stopbands.
110 Filter Synthesis Using Genesys S/Filter
L3 L4 L5
C2 C5
L=49.2μH L=53.3μH L=48.5μH
ZO=50Ω C=330pF C=330pF ZO=50Ω
X1 X3 X2
R=11.6Ω R=11.6Ω R=11.6Ω
L1 L=12.092mH L=12.092mH L=12.092mH L2
L=1000nH Cm=0.022565pF Cm=0.022621pF Cm=0.022565pF L=1000nH
C1 Co=5.45pF Co=5.45pF Co=5.45pF C6
C=1500pF C3 C4 C=1500pF
C=6.2pF C=6.2pF
Figure 10.39 Very wide-bandwidth quartz-crystal filter with matching networks and standard
values for discrete elements.
Figure 10.41 Schematic of the ceramic-resonator filter after application of transforms to the
original extraction.
The first and last resonators are replaced with ceramic resonators. The
static capacitance is then nulled with 6.574 uH inductors. The dipole
consisting of C5, C6, and L2 is ready for replacement with a ceramic
resonator. When the dipole consisting of C7, C8, and L3 is replaced with the
piezoelectric dipole using Eqs. 10.8 through 10.10, the values are ready for
replacement with a ceramic resonator. However, the required static
capacitance is higher than provided by the resonator. This is corrected by
placing additional capacitance, C4, in parallel with the resonator. The
ceramic resonator filter, after placing discrete capacitors on standard
values is given in Fig. 10.42 and the responses are given in Fig. 10.43.
Figure 10.42 Ceramic-resonator filter after replacing L-C dipoles with ceramic resonators.
Piezoelectric Devices 113
Reference
[1] R. Rhea, Discrete Oscillator Design: Linear, Nonlinear, Transient, and Noise
Domains, Artech House, Norwood, MA, 2010.
11 Symmetry
Symmetry has two important manifestations in electrical filters:
physical symmetry with element types and values mirrored with respect to
the input and output, and response symmetry with equal selectivity above
and below the passband [1]. Physical symmetry reduces the number of
unique elements that are modeled, ordered, tested, stocked, picked, and
assembled. For distributed element filters that are electromagnetically
simulated, physical symmetry can reduce the required execution time and
memory by a factor of 16 in simulators that take advantage of symmetry.
The conventional all-pole bandpass filter with equal quantities 0f TZs
at DC and infinity has geometric response symmetry. If the transmission
amplitude response is plotted on a logarithmic-frequency scale, the slope of
the transition regions below and above the passbands are equal. However,
more often arithmetic symmetry is desired. Arithmetic symmetry results in
group-delay symmetry.
115
116 Filter Synthesis Using Genesys S/Filter
The better solution is 11-th degree that adds only one capacitor (a total of
six), but requires only three unique values of inductors and capacitor, and
results in equal input and output impedance. An even better solution is a
symmetry preserving lowpass with three TZs at infinity and two TZs at 54
MHz. The Specification tab is given in Fig. 11.1 and the schematic for the
extraction sequence 54 ∞ ∞ 54 ∞ is given in Fig. 11.2. This solution only
requires three inductors. Given in Fig. 11.3 are the responses of this filter.
The rejection specification is achieved with considerable margin.
Figure 11.1 Specification tab for the value-symmetric lowpass created with two-paired FTZs.
C2 C5
C=32.82pF C=32.82pF
ZO=50Ω ZO=50Ω
L1 L2 L3
L=264.69nH L=400.52nH L=264.69nH
C1 C3 C4 C6
C=90.41pF C=190.03pF C=190.03pF C=90.41pF
Figure 11.4 Specification tab for the 455-kHz bandpass with paired FTZs below the
passband.
L2 C3
ZO=50Ω L=8.44e+6nH C=14.475pF ZO=50Ω
C1 C5
C=1346.458pF C=1346.458pF
L1 L3
L=280533.943nH L=280533.943nH
C2 C4
C=659.556pF C=659.556pF
C2 C4
ZO=50Ω C=2.21pF C=34.99pF ZO=50Ω
L1 L2 L3
L=73.76nH L=374.97nH L=86.82nH
C1 C3 L4
C=7.56pF C=8.93pF L=24.19nH
C5
C=29.48pF
C5
C1 C4 C=10.3pF
ZO=50Ω C=28.07pF C=23.27pF
ZO=71.56Ω
L1 L3
L=223.59nH L=155.92nH
C2 L2 L4
C=67.61pF L=129.02nH L=90.14nH
C6
C=55.99pF
C3
C=153.18pF
Figure 11.9 Zigzag (minimum-inductor elliptic) designed by the Genesys Passive Filter
program.
122 Filter Synthesis Using Genesys S/Filter
Figure 11.11 Generalized response of the symmetric 50- to 90-MHz bandpass filter.
Figure 11.12 Responses of the generalized 50- to 90-MHz bandpass with symmetric
responses.
Figure 11.13 Schematic of the symmetric generalized 50- to 90-MHz bandpass with greater
stopband attenuation above the passband.
Figure 11.14 Responses of the symmetric generalized 50- to 90-MHz bandpass with greater
attenuation above the passband.
Figure 11.15 Three-section group-delay equalizer cascaded with the all-pole symmetric-
response bandpass.
References
[1] R. Rhea, “Exploiting Filter Symmetry,” Microwave Journal, March 2001.
[2] H. Blinchikoff and A. Zverev, Filtering in the Time and Frequency Domains,
Noble Publishing(SciTech), Raleigh, NC, 2001.
[3] R. Saal and E. Ulbrich, “On the Design of Filters by Synthesis,” IRE Trans.
Circuit Theory, Vol. CT-5, December 1958.
[4] J. Rhodes, “The Design and Synthesis of a Class of Microwave Bandpass Linear
Phase Filters,” Trans. Microwave Theory and Techniques, Vol. 17, April 1969.
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 user-
specified 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.
SOURCE LOAD
Rsource
R=50Ω
VS1 Rload
VDC=1V R=50Ω
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.
GL
A= − GL2 (12.2)
R0
B = RL ( R0 − RL ) − X L (12.3)
C = − RL ( R0 − RL ) − X L (12.4)
A − BL
C1 = (12.6)
2πf
1
L2 = (12.7)
2πf ( A + BL )
GL2 + A2
C2 = (12.8)
2πfA
B
L3 = (12.9)
2πf
B + XL
C3 =
2πf (RL2 + (B + X L )2 ) (12.10)
L4 = −
(R + (C + X L )2
2
L ) (12.11)
2πf (C + X L )
1
C4 = − (12.12)
2πfC
GL2 + A2
C5a = (12.13)
2πfA
A + BL
C5b = − (12.14)
2πf
A
L6 a =
(
2πf Gl2 + A2 ) (12.15)
132 Filter Synthesis Using Genesys S/Filter
1
L6b = − (12.16)
2πf ( A − BL )
1
C7 a = − (12.17)
2πfB
B + XL
C7b =
2πf ( RL2 + (B + X L )2 ) (12.18)
C
L8a = (12.19)
2πf
L8b = −
(R 2
L + ( X L + C )2 ) (12.20)
2πf ( X L + C )
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
Figure 12.5 Specification tab for matching example that exploits extraction sequences.
ZO=12.5Ω
ZO=50Ω
L2 L4
L=25.78nH L=26.96nH C3
C=48.34pF
L3 C2
L=10.02nH C=147.34pF
L1 C1
L=194.21nH C=43.12pF
ZO=50Ω ZO=12.5Ω
L1 L3
L=24.97nH L=22.44nH C3
C=59.11pF
L2 C2
L=12.03nH C=122.7pF
C1
C=35.85pF
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.
L1 L2 L3
L=666.37nH L=666.37nH L=27.75nH
C2 C4
C=18.22pF 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.
C1 C3 C5
ZO=100Ω C=4.671pF C=5.6216pF C=10.4779pF
L1 L2 L3
L=666.37nH L=666.37nH L=332.5nH ZO=50Ω
C2 C4
C=18.22pF C=25.79pF
Figure 12.10 Final schematic of the series-resonator bandpass used to match 100 to 50
ohms.
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 2πf 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
XL
Qof load = (12.22)
RL
BL
Qof load = (12.23)
GL
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=9.6Ω ZO=50Ω
L2 L3
L=7.68nH L=17.5nH
L1 C2 C3
L=6nH C=38.51pF C=12.47pF
C1
C=64.84pF
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 C4
ZO=50Ω C=59.02pF C=32912.06pF C=87pF
L1 L3
L=30677.88nH L=20857.56nH ZO=34Ω
L2 C3
L=313.57nH 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.
13 Distributed Filters
Distributed filters use TLEs to realize the transfer function rather than
using lumped inductors and capacitors. The reactance of ideal lumped
elements is a linear function of the frequency variable ω , and practical
elements exhibit this behavior over a wide bandwidth. When electrically
much shorter than a quarter-wavelength, TLEs behave much like lumped
reactors. The behavior of longer TLEs are a periodic function tan θ , where
θ is the electrical length of the TLE. Exact synthesis techniques are
available for distributed filters by replacing the linear frequency variable ω
with the periodic frequency variable tan θ . This is referred to as Richards’
transform [1]. Minnis is an excellent review of exact synthesis for
microwave circuits using the Richards transform [2].
Therefore, two different design methods are used for distributed filters.
One assumes electrically short TLEs, and it takes advantage of tan θ ≈ θ for
small θ . Although approximate, this method should not be dismissed.
Filters constructed with short TLEs provide better stopband performance.
Several practical filter topologies are easily designed using this approach.
Furthermore, while the initial synthesis process using the Richards
transform is exact, when realized with real transmission lines,
discontinuities result in inexact realization. For any method of design,
trade-offs and tuning or optimization are required.
The second method, exact synthesis using Richards’ transform,
requires all TLEs to be of equal length, referred to as commensurate. This
synthesis often results in series TLE stubs that are realizable only in
wireline, and not microstrip or stripline. This difficulty is resolved using
transforms and an additional type of TLE, the unit element. Mastery of
these techniques requires practice. Later chapters include many examples
to illustrate these techniques.
145
146 Filter Synthesis Using Genesys S/Filter
Filter are given in Fig. 13.1. The microstrip metal patterns are
approximately 3x actual size. These are five-section 6- to 6.2-GHz all-pole
bandpass filters with a 0.0432-dB passband ripple on the Rogers TMM6
25-mil thick substrate.
Figure 13.1 Four of the eight different distributed bandpass filter topologies automatically
designed by the Microwave Filter module of Genesys.
Figure 13.2 Four different distributed lowpass topologies automatically designed by the
Microwave Filter module of Genesys.
13.3.1 TLEs
Distributed synthesis utilizes six basic TLEs as shown in Fig. 13.3. A TLE in
the transmission path is referred to as a unit element (UE). It does not add
a TZ, but it may have an effect on the transfer function.
Each specified nonfinite TZ and each UE adds one degree to the filter
transfer function and one TLE to the topology. Each FTZ adds two degrees
to the filter and three TLEs to the topology. The open wireline and shorted
wireline are not realizable in a planar process such as microstrip. This is
resolved by transformation to one of the other basic elements.
TLE elements are defined by their characteristic impedance, Z0, and
electrical length θ . It is sometimes more convenient to use TLE
characteristic admittance, Y0. The characteristic impedance and electrical
length are related to the distributed inductance and capacitance per unit
length by
Z0 = L (13.1)
C
150 Filter Synthesis Using Genesys S/Filter
LC = µε = (µ 0 µ r )(ε 0ε r ) =
µrε r (13.3)
c2
where len is the physical length of the line and c is the speed of light in a
vacuum. With µ r = 1 ,
εr
Z0 = (13.5)
cC
A parallel-coupled line (PCL) that is symmetric with equal self-
impedance lines is defined by the even-mode ( Z 0e ) and odd-mode ( Z 0o )
characteristic impedance and the even and odd-mode electrical length. For
pure TEM-mode propagation, θ e = θ o . This holds for stripline (triplate) and
coupled rods between ground planes (slabline), but is only approximately
valid for microstrip. Useful coupled-line relationships are
Z 0 = Z 0e Z 0o (13.6)
Z 0e − Z 0o
kv = (13.7)
Z 0e + Z 0o
Z 0e − Z 0o
kw = (13.8)
2 Z 0e Z 0o
series L-C network is plotted as a dashed-dot-dot trace from 0.5 to 1.5 fq.
The shorted and open stubs reverse behavioral roles at even multiples of fq.
The behavior of ideal lumped elements is a linear function of frequency.
For example, the reactance of an inductor is ωL, where ω = 2πf . The
behavior of TLEs is a function of tan θ . Therefore, it is useful to define a
new variable for synthesis, Ω = tan θ . The Richards transform relates these
two frequency variables by
πf
Ω = tan (13.9)
2 fq
Figure 13.4 Input reactance of a shorted stub (solid trace) and an open stub (dashed trace).
Also plotted near fq are the reactance of series and parallel L-C networks to ground.
Figure 13.5 Responses of a synthesized lumped (solid trace) and distributed (dashed trace)
lowpass.
Z 3a = na2 Z 0 a (13.12)
Z 0b
Z 2b = (13.14)
nb2
Z1b
Z 3b = (13.15)
nb2
Z1c
nc = 1 + (13.16)
Z 0c
Z 0c
Z 2c = (13.17)
nc
Z1c
Z 3c = Z 4c = (13.18)
nc (nc + 1)
Z 0d
nd = 1 + (13.19)
Z1d
Z 4 d = nZ 0 d (13.21)
154 Filter Synthesis Using Genesys S/Filter
Z 0e Z 2e + Z1e Z 2e
ne = (13.22)
Z 0e Z1e + Z1e Z 2e
Z 0e Z1e
Z 3e = (13.23)
Z 0e + Z1e − ne Z1e
Z 0e
Z 4e = (13.24)
ne
Z 0e
Z 5e = (13.25)
ne (ne − 1)
When the structure is simplified, Z5e and Z2e parallel combine to one
stub with characteristic impedance equal to Z3e.
Figure 13.7 Ikeno transform to resolve excessively high characteristic impedance of Z2.
156 Filter Synthesis Using Genesys S/Filter
(
Z 0e: min = Y1 + Y0 − Y1 (Y1 + Y0 ) )
−1
(13.27)
where the admittances Yn are the inverse of the impedances Xn. Specified
values for Zc1 near the minimum value result in an open-stub TLE
characteristic impedance near 0 ohms. Specified values for Zc1 near the
maximum value result in an open-stub characteristic impedance near the
high-original value, and an unrealistically low value of Z00. If the
characteristic admittance of the final open-stub TLE is Ys, then with
a + b 2 − 4ac
Z 2a = (13.32)
2a
where
a = 2 Z1 + Z 2 + Z 3 (13.33)
(
b = 2 Z12 + Z1Z 3 ) (13.34)
then
Z
Z 4 = 1 + 1 Z1 (13.36)
Z 2a
Z
Z 9 = 1 + 1 Z 2 a (13.37)
Z 2a
Z3
nT 1 = 1 + (13.38)
Z9
Z9
Z8 = (13.39)
nT 1
Z3
Z5 = (13.40)
nT 1
158 Filter Synthesis Using Genesys S/Filter
Z 2b
Zx = (13.41)
nT21
Z
Z 6 = 1 + 8 Z 8 = Z 4 (13.42)
Zx
Z 7 = Z x + Z8 (13.43)
Z1b
Z 2b = Z 3b = (13.46)
2
Z 2c = Z 3c = 2 Z1c (13.47)
Z1a
Z 2 d = Z 3d = (13.48)
2
Z 2e = Z 3e = 2 Z1e (13.49)
Z1 f
Z2 f = (13.50)
0.5 2 + 1
(
Z 3 f = Z1 f 0.5 2 + 1 ) (13.51)
160 Filter Synthesis Using Genesys S/Filter
Z 0o = Z1 (13.53)
Z1Z 2
Z 0o = (13.55)
2 Z1 + Z 2
162 Filter Synthesis Using Genesys S/Filter
Ζο
θ C1
1 Z
= 0 (13.59)
2πfC3 tan θ
Distributed Filters 163
Figure 13.12 Short TLE shorted and open stub and lumped-element equivalent model.
13.5 Inverters
An ideal impedance inverter behaves like a transmission line that is a
quarter-wavelength long at all frequencies [4]. If the line with characteristic
impedance K is terminated in an impedance Zb at one end, the impedance
seen looking into the other end is Za where
K2
Za = (13.64)
Zb
J2
Ya = (13.65)
Yb
Inverters are used to obtain more realistic element values by scaling the
impedance level of a filter. Inverters are also used to transform a parallel
resonator to a series resonator, and vice versa. This can be used to convert a
conventional bandpass filter into a topology with all parallel or all series
resonators. The process of inserting an inverter is exact and does not alter
the responses of the filter. However, ideal inverters do not exist. Inverters
are replaced with realizable structures that approximate an ideal inverter
over a limited bandwidth. For bandpass filters, the approximation is
adequate for a wide variety of applications. One realizable structure is a
transmission line of characteristic impedance K or admittance J that is a
quarter-wavelength long at the center frequency of a bandpass filter.
Certain pi, tee and “L” networks constructed using inductors or using
Distributed Filters 165
References
[1] P.I. Richards, “Resistor-Transmission-Line Circuits,” Proc. IRE, Vol. 36,
February 1948.
[2] B. J. Minnis, Designing Microwave Circuits by Exact Synthesis, Artech House,
Norwood, MA, 1996.
[3] R. Rhea, “Technology Enables New Components,” Microwave Journal,
November 2006.
[4] G. Matthaei, L. Young, and E.M.T. Jones, Microwave Filters, Impedance-
Matching Networks, and Coupling Structures, Artech House, Norwood, MA, 1980.
14 Distributed Lowpass Filters
This chapter covers the design of both all-pole and generalized
distributed lowpass filters by both exact and approximate methods. This
chapter also illustrates techniques for reducing the physical size of the
filters to potentially reduce box-mode degradation of the stopbands.
In this chapter, each example lowpass filter is fifth-degree designed
with a 0.0432-dB passband ripple and a cutoff frequency of 2400 MHz.
They are implemented on microstrip with a 50-mil-thick Rogers TMM6
substrate with quarter-ounce copper.
169
170 Filter Synthesis Using Genesys S/Filter
Figure 14.1 Specification tab for the exact lowpass filter with series and stub TLEs.
Figure 14.2 Transformation steps for the exact lowpass filter with series and stub TLEs.
The filter does not quite meet the desired minimum and maximum
characteristic-impedance limits in that TL1 and TL9 exceed the desired
maximum TLE characteristic impedance of 120 ohms. The ¼ Wave Freq
parameter could be increased to decrease the characteristic impedance of
TL1, but this has the affect of decreasing the characteristic impedance of
TL5.
The amplitude responses of the filter are given in Fig. 14.3. As can be
seen, the design is exact as evidenced by a cutoff at precisely 2.4 GHz and
an equal passband return loss of 20 dB. Also evident is a primary limitation
172 Filter Synthesis Using Genesys S/Filter
Figure 14.3 Responses of the five-section 2.4-GHz lowpass filter designed in S/Filter based
on Richards’ transform and transformed using Kuroda.
Figure 14.4 The 2.4-GHz lowpass schematic (top) as realized in microstrip with dimensions
and discontinuities automatically added by the Genesys feature Covert Using Advanced
TLine. The layout is given at the bottom (2X scale).
Figure 14.5 Responses computed by circuit-theory simulation of the of the 2.4-GHz filter
realized in microstrip.
Figure 14.6 The 2.4-GHz lowpass schematic (top) with realized in microstrip after optimization
of dimensions. The layout is shown at the bottom (2X scale).
Figure 14.7 Optimized responses of the 2.4-GHz filter initially designed using exact methods.
Figure 14.8 Specification tab for the lowpass with contributing UEs.
TL4
W=3.05mm
TL7 TL5 TL8
L=4.85mm
W=1.88mm W=0.18mm W=1.88mm
L=1.45mm L=5.91mm L=1.45mm
Figure 14.9 Design sequence for the 2.4-GHz lowpass with contributing UEs (2X scale).
178 Filter Synthesis Using Genesys S/Filter
Figure 14.10 Initial synthesized responses (left) and microstrip responses (right) of the 2.4-
GHz lowpass with three TZs at infinity and two contributing UEs.
Figure 14.11 Schematic with (a) a 6-GHz quarter-wave frequency, (b) a 6.76-GHz quarter-
wave frequency, and (c) realized in coax.
Distributed Lowpass Filters 179
Figure 14.13 Transform steps for the generalized fifth-degree 2.4-GHz lowpass.
182 Filter Synthesis Using Genesys S/Filter
Figure 14.14 Responses of the electrical schematic (left) and microstrip form (right) of the
generalized lowpass.
Consider the Series Element First form of the five-section all-pole lowpass
in Fig. 14.2a. The reason for shifting to the Series Element First form is
evident later in this section. The series-shorted TLE stubs are unrealizable,
and in the exact design, redundant UEs are added to transform these
series-shorted TLE stubs. This adds four TLEs of commensurate length,
thus substantially increasing the size of the filter. In this section, the initial
design of the 2.4-GHz lowpass filter uses approximate methods. Design
begins with the same Specification tab except the ¼ Wave Freq is adjusted
to 5858 MHz to set the maximum characteristic impedance of TL3 at 120
ohms. The result is given in Fig. 14.16a. The following transforms are then
applied:
1) Both shunt TLE open stubs are converted to shunt capacitors using
Distributed to Lumped Equivalents: Open Stub to Shunt C at the
cutoff frequency 2400 MHz. These transforms are approximate.
2) Both shunt Cs are split using Basic Operations: Split Shunt Element.
Distributed Lowpass Filters 183
Figure 14.16 Series TLE and shorted-stub lowpass designed using approximate transforms.
Distributed Lowpass Filters 185
Figure 14.17 Responses before optimization (left) and after optimization (right) of the
microstrip series TLE and open-stub 2.4-GHz lowpass.
Figure 14.18 Layout of the optimized series TLE and open-stub lowpass (2X scale).
The design begins with the 2.4-GHz lumped element filter given in Fig.
14.19a. The series inductors are converted to series TLEs using Lumped to
Distributed Equivalents: Series L to TLine with a conversion frequency of
2400 MHz and a characteristic impedance of 120 ohms. The shunt
capacitors are converted using Lumped to Distributed Equivalents: Shunt
C to TLine with a conversion frequency of 2400 MHz and a characteristic
impedance of 20 ohms. Again, 50-ohm input and output lines are added.
The electrical schematic is given in Fig. 14.19b and the schematic after
conversion to microstrip on TMM6 is given in Fig. 14.19c. The layout is
shown in Fig. 14.19d at approximately 2x actual size.
Shown on the left in Fig. 14.20 are the responses of the stepped-
impedance 2.4-GHz lowpass with electrical TLE models and on the right
after conversion to microstrip with discontinuities on 50-mil TMM6.
186 Filter Synthesis Using Genesys S/Filter
L1 L2 L3
ZO=50Ω L=3.22nH L=5.97nH L=3.22nH ZO=50Ω
C1 C2
a C=1.82pF C=1.82pF
Figure 14.20 Responses of the TLE stepped-impedance 2.4-GHz lowpass (left) and in
microstrip including discontinuity models (right).
using variables. Narrower lines would improve the stopbands, but the
widths are at the desired minimum widths, so they are not optimized. It is
anticipated that optimization would shorten the TLE lengths to increase the
cutoff frequency and that it would widen TL2 and TL4 to improve the
stopbands, and that is what happens. The optimized microstrip schematic
is given in Fig. 14.21 and the optimized response in Fig. 14.22.
L1 L3
ZO=50Ω L=3.72nH L=3.72nH ZO=50Ω
L2
C1 L=1.29nH C3
C=1.32pF C=1.32pF
C2
C=1.53pF
TL6
W=7.33mm
L=3.29mm
c
TL6
W=5.3mm [W6]
L=4mm [L6]
d
Figure 14.23 Design steps for the generalized 2.4-GHz lowpass by approximate design.
Distributed Lowpass Filters 189
The cutoff frequency is low and the passband return loss is poor.
Optimization is used to correct the passband and to improve the stopband
from 3600 to 9600 MHz. Symmetry is used to retain symmetry. The length
of each TLE and the width of the wide stubs are optimized. The responses
after optimization are given on the right in Fig. 14.24. The layout at
approximately 2x actual size is given in Fig. 14.25.
Figure 14.26 Electrical TLE schematic of the penetrating double-stub 2.4-GHz lowpass filter.
192 Filter Synthesis Using Genesys S/Filter
Figure 14.27 Penetrating stepped-impedance 2.4-GHz microstrip response before (left) and
after (right) optimization of the line lengths.
Figure 14.28 Layout of the penetrating stepped-impedance 2.4-GHz lowpass (2X scale).
Figure 14.29 Schematic and layout (2X scale) of the microstrip radial stub 2.4-GHz lowpass.
194 Filter Synthesis Using Genesys S/Filter
extra capacitance that is ignored for this simulation. The responses are
given on the left in Fig. 14.32.
Next, the length of the series TLE and the values of the lumped
capacitors is optimized, the capacitors are then fixed at the nearest
standard values, and a final optimization of the length of the series TLE is
completed. The final microstrip schematic and layout are given in Fig.
14.31. The optimized response is given on the right in Fig. 14.32.
L1 L2
ZO=50Ω L=4.55nH L=4.55nH ZO=50Ω
C1 C2 C3
C=1.29pF C=2.39pF C=1.29pF
TL1 TL2
Z=120Ω Z=120Ω
L=12.1mm L=12.1mm
ZO=50Ω K=1 K=1 ZO=50Ω
C1 C2 C3
C=1.29pF C=2.39pF C=1.29pF
Figure 14.31 Electrical and microstrip schematics and layout (2X scale) of the hybrid 2.4-GHz
lowpass.
This hybrid lowpass has small size and the good stopband performance.
The excellent stopband performance results partly from the absence of
reentrance modes in the lumped capacitors.
Further size reduction is possible by folding the series TLEs. Given in
Fig. 14.33 is the microstrip schematic resulting from manually folding the
series TLE and adding chamfered corners. An EMPower EM of the
196 Filter Synthesis Using Genesys S/Filter
compact hybrid lowpass is then used to consider the effects of the footprint
metal. The lumped elements are then optimized for the best response. The
results given in Fig. 14.34 illustrate the difficulty in maintaining stopband
performance beyond a few multiples of the cutoff frequency, even using a
hybrid approach. Electromagnetic simulation is described in detail in
Chapter 19.
Figure 14.32 Responses of the 2.4-GHz hybrid lowpass before (left) and after (right)
optimization.
Figure 14.33 Microstrip schematic and layout (2X scale) of the compact hybrid lowpass with
lumped capacitors and folded series TLE.
198 Filter Synthesis Using Genesys S/Filter
Reference
[1] F. Giannini, M. Ruggieri, and J. Vrba, “Shunt-Connected Microstrip Radial
Stubs,” IEEE Trans. Microwave Theory and Techniques, Vol. MTT-34, No. 3,
December 1986.
15 Distributed Bandstop Filters
Examination of the responses on the left in Fig. 14.10 reveals that
reentrance modes cause lowpass filters synthesized using the Richards
transform to also possess bandstop responses centered on the ¼ Wave
Freq (fq). Unfortunately, many bandstop filters require narrow stopbands.
This causes unrealizable TLE characteristic impedance. This chapter
includes techniques for managing this problem.
In this chapter, bandstop filters centered at 5 GHz are designed. The
higher frequency is best implemented on thinner substrate, so 25-mil-thick
Rogers TMM6 with half-ounce copper is used for each example.
199
200 Filter Synthesis Using Genesys S/Filter
ZO=50Ω ZO=50Ω
TL2 TL4 TL6
W=1.48mm W=3.39mm W=1.48mm
L=6.83mm L=6.49mm L=6.83mm
b
TL1 TL3 TL5 TL7
W=0.93mm W=0.12mm [W3] W=0.12mm [W3] W=0.93mm
L=0.48mm L=7.79mm [L3] L=7.79mm [L3] L=0.48mm
ZO=50Ω ZO=50Ω
TL8 TL9 TL10
W=.12mm W=.12mm W=.12mm
L=.12mm L=.12mm L=.12mm
ZO=50Ω ZO=50Ω
TL8 TL9 TL10
W=.12mm W=.12mm W=.12mm
L=.12mm L=.12mm L=.12mm
Figure 15.1 Design steps for the wide-bandwidth 2.4-GHz bandstop filter with three TLE open
stubs and two contributing UEs.
Distributed Bandstop Filters 201
Figure 15.2 Responses of the electrical (left) and microstrip (right) forms of the wide-
bandwidth bandstop filter.
Figure 15.3 Responses of the wide-bandwidth bandstop filter modified to reduce the effects of
the microstrip tee discontinuities (left) and after optimization to improve the responses (right).
not necessary. The results after optimization of dimensions are given on the
right in Fig. 15.6.
ZO=50Ω ZO=50Ω
TL2 TL4 TL6
W=0.13mm W=0.59mm [W4] W=0.13mm
L=7.63mm [L2] L=7.19mm [L4] L=7.63mm [L2]
b
TL1 TL3 TL5 TL7
W=0.93mm W=0.56mm [W3] W=0.56mm [W3] W=0.93mm
L=1mm L=7.18mm [L3] L=7.18mm [L3] L=1mm
ZO=50Ω ZO=50Ω
TL2 TL4 TL6
W=0.13mm W=0.59mm [W4] W=0.13mm
L=7.69mm [L2] L=7.4mm [L4] L=7.69mm [L2]
Figure 15.5 Design steps for the moderate bandwidth bandstop filter.
204 Filter Synthesis Using Genesys S/Filter
Figure 15.6 Responses of moderate bandwidth microstrip bandstop filter before (left) and
after optimization (right).
later step. The resulting electrical schematic, after applying the Ikeno
transforms to TL1 and TL5 and adding short 50-ohm TLEs at the input and
output, is given in Fig. 15.7b. The resulting responses are shown on the left
in Fig. 15.8.
The resulting microstrip schematic after Convert Using Advanced
TLine is given in Fig. 15.7c. As expected, the microstrip discontinuities
disturb the response. Because the open stubs are narrower than those in
lowpass filters, the tees connecting the open stubs to the through path are
less destructive. Nevertheless, optimization is used to correct the cutoff and
center frequencies of this bandstop filter. Because the width of the center
TLE and coupled lines are similar, they are set equal, and variables are used
to force symmetry during optimization. Because the automatic DisCos for
the corners may pick either adjacent line width, microstrip corner
discontinuities were added and set equal to the width of the coupled lines.
Unfortunately, this width does not equal the width of the narrow open
stubs so the corner discontinuities are geometrically incorrect. This is best
dealt with by electromagnetic simulation as described in Chapter 19. The
microstrip schematic after optimization is shown in Fig. 15.7d and the
responses are shown on the right in Fig. 15.8. The layout is shown in Fig.
15.9. The maximum stopband attenuation is limited by losses in the
microstrip realization of the filter. Folding the open stubs back toward the
filter input and output would reduce the physical height of the filter.
Figure 15.8 Electrical schematic responses of the narrow bandstop (left) and the microstrip
responses after optimization (right).
Figure 15.9 Microstrip layout of the narrow bandstop with optimized dimensions (4X scale).
Figure 15.10 Responses of the synthesized generalized narrowband bandstop filter (left) and
the microstrip form after modifications and optimization (right).
Figure 15.12 Generalized narrowband bandstop filter layout after modifications and
optimization.
211
212 Filter Synthesis Using Genesys S/Filter
Figure 16.3 Response of the three-section edge-coupled 5-GHz bandpass (left) and with
quarter- wavelength transformers to moderate even- and odd-mode impedance (right).
The top of Fig. 16.4 shows the microstrip schematic after applying
Convert Using Advanced TLine and optimization to recover the return loss.
The bottom of Fig. 16.4 shows the layout at approximately 2x scale. The
responses after optimization are given in Fig. 16.5.
Each resonator is electrically a quarter-wavelength long at 5 GHz,
resulting in the first reentrant passband at three times 5 GHz or, 15 GHz.
However, in microstrip the even- and odd-mode propagation velocities are
slightly different. This results in a spurious passband centered near two
times 5 GHz. This lower reentrance does not occur in pure TEM mode
media such as stripline.
Distributed Bandpass Filters 215
Figure 16.4 Optimized edge-coupled 5-GHz bandpass microstrip schematic (top) and layout
at bottom (2X scale).
L1 L3
ZO=50Ω L=6.78nH C1 L=6.78nH ZO=50Ω
C3
C=0.151pF C=0.151pF
L2 C2
L=0.2915nH C=3.5114pF
a
*
TL3
W=0.65mm [W2]
S=0.22mm [S2]
L=7.09mm [L2] TL4
TL1 W=0.31mm [W1]
W=0.31mm [W1] S=0.18mm [S1]
S=0.18mm [S1] L=7.28mm [L1]
L=7.28mm [L1] TL2
W=0.65mm [W2]
S=0.22mm [S2]
L=7.09mm [L2]
e
Figure 16.6 Design steps for the edge-coupled bandpass designed using inverters.
218 Filter Synthesis Using Genesys S/Filter
Figure 16.7 Responses of the electrical edge-coupled bandpass (left) and in microstrip after
optimization (right).
After adding short 50-ohm TLEs at the input and output, the layout at
approximately 2x scale is given in Fig. 16.8. The coupled-line widths are
somewhat wider in this version of the filter resulting in somewhat less
insertion loss in the passband.
Figure 16.8 Layout of the 5-GHz edge-coupled bandpass designed using inverters (2X scale).
transformer are resolved as transforms are applied to the filter. The initial
responses are given on the left in Fig. 16.11.
Figure 16.9 Specification tab for the five resonator interdigital bandpass mirrored about fq of a
highpass synthesis.
Figure 16.10 Initial extraction and transform sequence of the interdigital bandpass.
Distributed Bandpass Filters 221
Figure 16.11 Responses of the synthesized interdigital filter (left) and after optimization of the
multiple-coupled-line model (right).
The discontinuities result in poor return loss and wider than desired
bandwidth. This is corrected by the optimization of the line widths,
spacings, and the length. The results are given on the right in Fig. 16.14.
The optimized dimensions are given in Fig. 16.13b. The responses are
improved with dimensions near the initial values, but the best responses
require extensive optimization with significantly wider lines. This results in
closer line spacing, and the increased open-end capacitance reduces the
required line lengths.
Distributed Bandpass Filters 223
ZO=50Ω ZO=50Ω
W=0.4626mm
S1=0.1544mm
S2=0.7131mm
7
S3=0.8827mm
L=6.9288mm
ZO=50Ω ZO=50Ω
W=1.8676mm [W]
S1=0.1mm [S1]
S2=0.4111mm [S2]
1
7
S3=0.6113mm [S3]
L=6.3393mm [L]
Figure 16.13 Microstrip form of the interdigital filter created using (a) Advanced TLine and (b)
after optimization.
Figure 16.14 Microstrip form of the interdigital bandpass with initial dimension (left) and after
optimization (right).
224 Filter Synthesis Using Genesys S/Filter
Figure 16.15 Layout of the 5-GHz transformer-coupled interdigital bandpass (2X scale).
Figure 16.16 Specification tab for the three-section, capacitive-coupled combline bandpass.
Distributed Bandpass Filters 225
Figure 16.17 Responses of the synthesized combline (left) and after transforms with
capacitors (right).
TL1
ZOE=60Ω
ZOO=42.3471Ω
LENGTH=37.500mm
C1 C5
C=1.5061pF C=1.507pF
TL2
Z=287.8653Ω
L=37.5mm
TL3
C2 C3 C4
Z=60Ω
C=1.9159pF C=3.6679pF C=1.9166pF
L=37.5mm
Figure 16.18 Initial extraction and transform sequences for the external-capacitance-coupled
combline bandpass.
Distributed Bandpass Filters 227
C3
C=3.67pF [C3]
C2 C4
C=1.92pF [C2] C=1.92pF [C2]
C1 C5
C=1.51pF [C1] C=1.51pF [C1]
3
ZO=50.4Ω [Zo]
K1=3.474e-3 [k1]
K2=3.474e-3 [k1]
LEN=37.5mm
C3
C=3.2309pF [C3]
C2 C4
C=2.1314pF [C2] C=2.1314pF [C2]
C1 C5
C=1.2742pF [C1] C=1.2742pF [C1]
1
W=0.882mm [W]
S1=0.9092mm [S]
L=17.42mm [L]
Figure 16.19 Combline form of the bandpass as transformed (top) and after conversion to
microstrip and optimization (bottom).
Figure 16.21 Specification tab for the 2250- to 2750-MHz miniature bandpass.
Figure 16.22 Transformation steps for the 2250- to 2750-MHz symmetric miniature bandpass.
Distributed Bandpass Filters 231
Figure 16.23 Electrical schematic of the optimized miniature bandpass (top) and microstrip
schematic after optimization (bottom).
232 Filter Synthesis Using Genesys S/Filter
Figure 16.24 Miniature 2250- to 2750-MHz bandpass responses of the optimized electrical
schematic (left) and the optimized microstrip filter (right).
TL1 and TL10 are quarter-wavelength lines and TL5 and TL6 cascaded
are 750 in length. These longer length lines are folded to reduce the overall
length of the filter. A layout at 2x scale with folded lines is given in Fig.
16.25.
Figure 16.25 Microstrip layout of the miniature 2250- to 2750-MHz bandpass with folded lines
to conserve space (2X scale).
Figure 16.27 Specification tab for the narrow bandpass with one FTZ.
11) The inverters added in step are re-edited for the most reasonable set
of characteristic impedance for the filter. In this case K=85 is chosen.
The result is given in Fig. 16.28d.
In the final layout, one two-step stub is oriented up and the other is
oriented down. The responses after transformations are given on the right
in Fig. 16.29. The quarter-wave impedance transformers are approximate,
but the performance degradation is small because of the narrow bandwidth
of this filter.
The filter is converted to microstrip on Roger TMM6 50-mil substrate
using Convert Using Advanced TLine. The schematic after optimization is
shown in Fig. 16.30. The narrow 8% bandwidth and FTZs near the
passband result in high-sensitivity to line parameters. The shunt stubs are
low impedance and are wide. To avoid the disturbance of a tee
discontinuity with a wide through line, very short and narrow lines are
added between these stubs and the through lines. Optimization requires
close user monitoring to avoid divergence. To avoid divergence, portions of
the circuit are converted using Convert Using Advanced TLine and are then
optimized before additional portions are converted. The resulting layout
approximately to scale is given in Fig. 16.31. The gap created by adding the
0.2- by 0.2-mm lines result is barely visible in the layout. The final
optimized responses are given in Fig. 16.32.
236 Filter Synthesis Using Genesys S/Filter
Figure 16.28 Transform sequence for the narrow bandpass with one FTZ.
Distributed Bandpass Filters 237
Figure 16.29 Initial responses (left) and after transforms (right) of the narrow bandpass with
one FTZ.
Figure 16.30 Microstrip schematic of the 2400- to 2600-MHz bandpass with one-finite FTZ.
238 Filter Synthesis Using Genesys S/Filter
Figure 16.32 Responses of the microstrip narrow bandpass with FTZ after optimization.
Figure 16.33 Progression of the (a) combline bandpass to (b) stepped-impedance combline
(c) to penetrating combline.
where θ1 and θ 2 are the electrical lengths of the high and low-impedance
sections respectively. The total length, as a fraction of the uniform-
impedance length, is given by
240 Filter Synthesis Using Genesys S/Filter
l fraction =
4
π
tan −1 ( K) (16.3)
Figure 16.34 Specification tab for the 950- to 1050-MHz penetrating combline bandpass.
One TZ at infinity and 2 N − 1 TZs at DC are specified for this topology. The
initial synthesized filter with a shunt element first and the extraction
sequence DC DC DC ∞ DC DC DC is given in Fig. 16.35a and the responses
are given on the left in Fig. 16.36.
The follow transforms are applied:
1) Compound Operations: Equate All Shunt Cs or Open Stubs. This adds
a shunt-open stub in parallel with each shunt-shorted stub, and
equates all the open stubs.
2) Basic Operations: Split Shunt Element with the option Two Equal
Pieces is applied to TL5.
3) Basic Operations: Swap Element is applied to TL4. This prepares TL2,
TL3 and the new TL4 for conversion to a combline PCL. The result is
given in Fig. 16.35b.
4) The same operations are applied to the other internal shorted and
open-stub pair, in this case TL8 and TL9.
5) Coupled Lines: Comb Lines, Open, Open with the Symmetric PCL
option is applied to each series-open line.
6) Transformer Operations: Absorb in Rload is applied to the near-unity
ratio transformer. The result is given in Fig. 16.35c.
7) The characteristic-impedance values are unrealizable. Inverters: Scale
Source and Load by Inserting Inverters with a parameter of 130 ohms
is applied.
8) Inverters: Replace Inverter with Quarter Wavelength TLine at 1000
MHz is applied to each inverter.
The resulting schematic is given in Fig. 16.35d, and the responses are
given on the right in Fig. 16.36.
242 Filter Synthesis Using Genesys S/Filter
Figure 16.35 Transform sequence for the 950- to 1050-MHz penetrating combline filter.
Distributed Bandpass Filters 243
Figure 16.36 Initial synthesized responses of the penetrating combline filter (left) and after the
addition of the quarter-wavelength transformers (right).
Figure 16.37 Left half of the symmetric penetrating combline 950- to 1050-MHz bandpass.
Figure 16.39 Layout of the penetrating-combline bandpass (2X scale). The quarter
wavelength transformer lines extend well beyond the figure (see text).
Figure 16.40 Minnis type-D narrow bandpass filter ideal response (left) and in microstrip after
optimization (right).
246 Filter Synthesis Using Genesys S/Filter
Figure 16.41 Transform steps for the narrowband Minnis Class-D bandpass.
Distributed Bandpass Filters 247
Figure 16.42 Microstrip layout of the narrowband 1-GHz Minnis type-D bandpass (to scale).
Figure 16.44 Transform steps for the generalized combline hybrid filter.
252 Filter Synthesis Using Genesys S/Filter
Figure 16.45 Response of the ideal generalized combline hybrid (left) and after transforms
and retuning of the FTZ (right).
Figure 16.46 (a) Original combline form of the generalized bandpass, (b) after optimization of
parameters, and (c) after conversion to microstrip and another optimization.
254 Filter Synthesis Using Genesys S/Filter
Figure 16.47 Responses of the optimized electrical-model combline (left) and after conversion
to microstrip and optimization (right).
Figure 16.48 Specification tab for the 1000- to 2000-MHz hybrid direct-coupled bandpass.
Distributed Bandpass Filters 255
Figure 16.49 Ideal responses of the 1000- to 2000-MHz hybrid direct-coupled bandpass (left)
and after the transforms with quarter-wave transformers and lumped capacitors (right).
The initial schematic with a shunt element first and the extraction
sequence ∞ UE DC ∞ UE ∞ UE ∞ is given in Fig. 16.50a. The following
transform steps are then applied:
1) TL3 is split using Basic Operations: Split Shunt Element and the right
element is swapped with the open stub using Basic Operations: Swap
Element.
2) Kuroda Wireline Transfers: Specify Transformer: Shunt Sorted Left
is applied to TL6 with the parameter option n=1.846 which is the
inverse of the existing transformer turns ratio. After Simply Circuit is
applied, the transformer is eliminated and the structure is symmetric
as illustrated in Fig. 16.50b.
3) Compound Operations: Equate All Shunt Ls or Shorted Stubs
produces the schematic in Fig. 16.50c.
4) Inverters: Scale Source and Load by Inserting Inverters with an
impedance parameter of 40 is applied to improve element values.
5) Each inverter is replaced with a quarter-wavelength line using
Inverters: Replace Inverter With: Quarter Wavelength TLine. The
center frequency of 1500 MHz is specified.
6) Each shunt-open stub is replaced with a lumped capacitor using
Distributed to Lumped Elements: Open Stub to Shunt C. The resulting
schematic is given in Fig. 16.50d.
256 Filter Synthesis Using Genesys S/Filter
Figure 16.50 Transform steps for the 1000- to 2000-MHz hybrid direct-coupled bandpass.
Distributed Bandpass Filters 257
Figure 16.51 Hybrid 1000- to 2000-MHz bandpass microstrip schematic at top and layout at
bottom (2X scale).
258 Filter Synthesis Using Genesys S/Filter
Figure 16.52 Optimized responses of the microstrip hybrid 1000- to 2000-MHz bandpass.
References
[1] S. Yamashita and M. Makimoto, “Miniaturized Coaxial Resonator Partially
Loaded with High-Dielectric-Constant Microwave Ceramics,” Trans. MTT-31,
September 1983, p. 697.
[2] R. Rhea, HF Filter Design and Computer Simulation, SciTech Publishing,
Raleigh, NC, 1994.
[3] B. Minnis, Designing Microwave Circuits by Exact Synthesis, Artech House,
Norwood, MA, 1996.
[4] G. Matthaei, L. Young, and E.M.T. Jones, Microwave Filters, Impedance-
Matching Networks, and Coupling Structures, Artech House, Norwood, MA, 1980.
17 Distributed Highpass Filters
The L-C form of highpass filters is shunt inductors alternating with
series capacitors. The series capacitor is difficult to realize in distributed
form. Therefore, the highpass filter is often realized in distributed form by
utilizing the lower transition region of a bandpass filter and accepting a
finite limit to the upper passband frequency. An alternative approach is
using shunt stubs to realize the shunt inductors and using lumped
capacitors, or semilumped elements such as overlaid strips in multilayer
structures. Because of TLE reentrance modes, these hybrid highpass filters
also have a finite passband bandwidth. This chapter considers both
approaches with filters on Rogers TMM6 50-mil-thick substrate material.
259
260 Filter Synthesis Using Genesys S/Filter
Figure 17.1 Specification tab for the highpass synthesized initially as a distributed filter.
ZO=50Ω ZO=50Ω
TL2 TL4
Z=29.48Ω Z=29.48Ω
L=19.38mm L=19.38mm
a
C1 C2
C=2.32pF C=2.32pF
b
Figure 17.2 (a) Initial distributed highpass and (b) after transform of the series-open wirelines
to lumped capacitors.
Distributed Highpass Filters 261
The response of the hybrid highpass is given on the right in Fig. 17.3.
The passband degradation above approximately 2500 MHz is caused by
reentrance modes of the shunt TLEs. Synthesis produced commensurate
shunt TLE. Consequently, with the first and last TLE at the maximum
desired 120 ohms, the middle TLE characteristic impedance is more
moderate. Extreme characteristic impedance improves passband
bandwidth. Therefore, the alternative approach of starting with a lumped
design is considered next.
Figure 17.3 Responses of the initial synthesized distributed highpass (left) and after transform
of the series-open wirelines to lumped capacitors (right).
ZO=50Ω ZO=50Ω
C1 C2
C=2.32pF C=2.32pF
L1 L2 L3
L=8.19nH L=4.42nH L=8.19nH
T1
P=1
a S=1
C1 C2
C=2.32pF C=2.32pF
b
C1 C2
C=2.32pF C=2.32pF
Figure 17.4 (a) The initial lumped highpass, (b) after shunt inductors transformed to TLE, (c)
after conversion of the TLE from electrical to physical, and (d) the final layout at 2X scale.
Distributed Highpass Filters 263
Figure 17.5 Responses of the initial lumped highpass (left) and after transform of the shunt
inductors to TLEs (right).
Figure 17.6 (a) Initial synthesized highpass with UEs, (b) after replacing the wirelines with
lumped capacitors, and (c) values after optimization. (d) Further optimization allowing shorter
TL2 through TL6 further improved the passband bandwidth.
Distributed Highpass Filters 265
Figure 17.7 Initial response of the distributed highpass with UEs (left) and after transforming
the series-open TLE to capacitors (right).
Figure 17.8 Responses after optimization of the commensurate-hybrid highpass filter (left),
and extended passband bandwidth by optimization with shorter TL2 through TL6 (right).
C1 C3
ZO=50Ω C=2.83pF C=2.83pF ZO=50Ω
L2
L1 L=6.89nH L3
L=8.13nH L=8.13nH
C2
a C=8.31pF
C1 C3
C=3.04pF [C1] C=3.04pF [C1]
TL3
TL1 Z=120Ω TL2
Z=120Ω L=15.94mm [L3] Z=120Ω
L=16.71mm [L1] L=16.71mm [L1]
C2
b C=8.05pF [C2]
Figure 17.9 Initial schematic and transformed hybrid highpass with a FTZ.
Figure 17.10 Initial responses of the 1000-MHz cutoff lumped highpass filter (left) and after
transformation of the inductors to TLEs (right).
268 Filter Synthesis Using Genesys S/Filter
Figure 17.11 Specification tab for the 1000-MHz highpass with three TZs at DC and one UE.
Distributed Highpass Filters 269
ZO=50Ω ZO=50Ω
TL3
TL2 Z=125.8251Ω
Z=74.2723Ω L=31.25mm
L=31.25mm
TL1 TL4
Z=72.3222Ω Z=194.8436Ω
L=31.25mm L=31.25mm
a T1
P=1
S=0.6091
Figure 17.12 Transformations of the 1000-MHz highpass with three TZs at DC and one UE.
270 Filter Synthesis Using Genesys S/Filter
Figure 17.13 Initial responses of the 1000-MHz highpass with three TZs at DC and one UE
(left) and after optimization of the microstrip form (right).
Figure 17.14 Responses of highpass with three TZs at DC and four UEs.
Figure 17.15 (a) Initial schematic of the highpass with three TZs at DC and four UEs and (b)
after transforms that eliminate the series-open wirelines.
272 Filter Synthesis Using Genesys S/Filter
Figure 17.16 Specification tab for the highpass filter synthesized as an eighth-degree
bandpass.
Distributed Highpass Filters 273
Figure 17.17 Transformation steps for the hybrid highpass designed as an eighth-degree
bandpass.
274 Filter Synthesis Using Genesys S/Filter
Figure 17.19 Responses of the eighth-degree highpass after transformation of the wireline
TLEs (left) and optimization (right).
ZO=50Ω ZO=50Ω
Figure 17.20 Transformation steps for the 10th-degree hybrid highpass filter.
Figure 17.21 Initial responses (left) and after transformation and optimization (right) of the
10th-degree hybrid highpass filter.
18 Multiplexers
Multiplexers as considered in this chapter are three or more port
devices that route signals to alternate ports based on frequency. They are
used to either split signals and divert them to a given port, or to combine
signals from alternate ports at a single port. A three-port device is referred
to as a diplexer. Devices with four or more ports are referred to as
multiplexers. Multiplexers differ from splitters and combiners in that with
ideal elements, the signal loss is zero. Whereas with splitters, the responses
are broadband, but signal power is divided among output ports [1].
Passive linear filters are reciprocal so design concepts are the same
whether the multiplexer is used to split or to combine signals. The
individual filters are a combination of lowpass and/or highpass filters, with
any quantity of bandpass filters. Therefore, there are an infinite variety of
multiplexers. Multiplexers are created in Genesys by designing the
individual filters and pasting them into a common schematic. All of the
features of the simulator are available, such as statistical, tuning,
optimization, and electromagnetic analysis.
C1 C2
ZO=50Ω ZO=0Ω ZO=50Ω C=69.5pF C=70.64pF
L1 L2
L=4049.7nH L=3984.49nH ZO=0Ω
C1 C2 L1 L2
C=844.69pF C=1958.68pF L=333.2nH L=143.69nH
C2 C4
L2 L4
ZO=50Ω C=625.49pF C=635.72pF
L=449.97nH L=442.72nH
ZO=0Ω
L1 C1 L3 C3
L=2998.77nH C=93.85pF L=1293.23nH C=217.63pF
Port_2
ZO=50Ω
LP_Filter
Port_3
ZO=50Ω
Port_1
BP_Filter ZO=50Ω
Port_4
ZO=50Ω
HP_Filter
Figure 18.3 Lowpass (top left), highpass (top right), and bandpass (middle) components of
the multiplexer with the reused filters assembled into the multiplexer (bottom).
Multiplexers 281
Figure 18.5 Schematic of the LP/HP 30-/50-MHz noncontiguous diplexer with doubly
terminated channel filters.
To recover the responses, the left three elements of each filter are
optimized with goals of re-establishing the initial return losses and cutoffs.
The optimization converges quickly to the responses given in Fig. 18.7.
Unlike contiguous diplexers, the return loss is poor in the crossover region,
but a minimum isolation of 35 dB is maintained.
Multiplexers 283
Figure 18.6 Responses of the individual channel filters plotted on one graph (left), and
responses after parallel connection of the filters to form the noncontiguous diplexer (right).
1
TL1
W=1.7mm [WLead]
L=8.45mm [Llead]
2
TL2
C1
W=1.7mm
C=1.84pF [Cap1]
S1=0.61mm [S1]
S2=2.6mm [S2] 3
S3=2.6mm [S2]
S4=0.61mm [S1] C2
L=25mm [L1] C=1.77pF [Cap2]
TL3 4
W=1.7mm [WLead]
C3
L=8.45mm [Llead]
C=1.84pF [Cap1]
5
Port_2
Figure 18.9 Individual responses of two combline bandpass filters plotted on the same graph
(left) and responses after connection as a diplexer (right).
Figure 18.11 Channel transmission amplitude and return-loss responses of the optimized
combline diplexer (left) and the isolation from port two to port three (right).
Reference
[1] G. Matthaei, L. Young, and E.M.T. Jones, Microwave Filters, Impedance-
Matching Networks, and Coupling Structures, Artech House, Norwood, MA, 1980.
19 Electromagnetic Simulation
In earlier chapters, simulation of the responses utilized closed-form,
analytical models for both lumped and TLEs. The emphasis in this book is
synthesis and transforms. Simulation is used to illustrate the success or
failure of these processes. Other works by the author illustrate the practical
aspects of filter realization [1-3] such as models, element (unloaded) Q,
losses, parasitics, tolerance, tuning, and standard values. For lumped
elements, the Genesys environment supports convenient substitution of
user developed and commercial models.
Distributed models in Genesys use refined and accurate industry-
published models. However, these models cannot consider a multitude of
factors including box modes, radiation, coupling between discontinuities,
and unusual geometries. These factors may significantly impact the
accuracy of the circuit theory simulation. Electromagnetic (EM) simulation
is a powerful tool for dealing with these issues. In this chapter, EM
simulation is used to further refine a few of the filters synthesized in this
book and to review a classic method of filter design newly enabled by EM
simulation.
19.1 Overview
Circuit simulators rely on network theory applied to impedance and
admittance models for individual elements. In the case of lumped elements,
simple reactance models, or models using a small subcircuit of elements,
are used. For TLEs, pure-TEM mode elements such as coax are modeled as
simple trigonometric functions. Models for quasi-TEM mode elements,
such as microstrip, are modeled by more complex functions that are based
on curve-fits to numeric data. In isolation, these models are accurate.
However, their accuracy is reduced by interactions, and models are often
not available for unusual geometries.
Rather than relying on network theory, EM simulators compute the
currents and fields produced in the metal of a circuit. Fundamentally, these
calculations rely on Maxwell’s equations, but to produce practical results,
involved mathematical concepts and tricks are employed. These concepts
vary among different commercial software programs.
Two-dimensional (2-D) simulators analyze only problems that are
infinitely continuous in one direction. Uniform transmission lines and
waveguides are analyzed with great numeric efficiency using this technique.
289
290 Filter Synthesis Using Genesys S/Filter
cavity resonance, TLE discontinuities such as open ends, bends, and steps
easily couple energy into and out of the cavity, resulting in destruction of
the stop bands. The frequency of resonance in a rectangular cavity is [4]
2 2 2
v l m n
f lmn = + + (19.1)
2 µrε r a b c
Figure 19.1 A 4- by 6-cm rectangular housing with microstrip input and output lines
(approximately to scale).
Given in Fig. 19.2 is the amplitude transmission of this system from the
housing port on the left to the housing port on the right, computed using
the EMPower electromagnetic simulator module of the Genesys suite. To
resolve the narrow resonant modes, frequency steps every 20 MHz are
used, thus requiring 400 frequency steps. The simulation is completed in a
few minutes on a standard desktop PC with an Intel Core2 2.5-GHz CPU.
The TE101-mode resonance occurs at approximately 4.3 GHz. The simulated
resonant frequency is slightly lower than predicted by Eq. 13.1 because the
294 Filter Synthesis Using Genesys S/Filter
50-mil thick TMM6 substrate loads about 13% of the height of the cavity.
The presence of the microstrip metal pattern also loads the cavity.
Radiation from the microstrip open ends and/or the standing-wave
pattern on the microstrip TLE readily couples to the cavity resonances. The
sweep through 10 GHz reveals three resonances as well as antiresonances,
with degraded stopband performance between resonances. The cavity
resonances are almost lossless, and the displayed loss of all three
resonances is nearly zero when the number of frequency simulation points
is sufficiently large.
In this case, stopbands better than 40 dB are unavailable at frequencies
above one half the lowest resonant frequency. Resonant frequencies scale
directly with the housing size. It is clear that achieving good stopband
performance at higher frequency requires a small housing. Cavity
resonance is why small filter size is critical for good stopband performance.
Given on the right in Fig. 19.2 is the response of the same layout but on
an open PCB without a conductive housing, as computed by the Momentum
EM simulator module of the Genesys suite. Resonances are avoided, thus
improving the potential stopband performance of filters. Nevertheless,
surface modes and edge reflections severely limit the ultimate stopband
performance.
Figure 19.2 Transmission amplitude response of the microstrip TLEs enclosed in a 4- by 6-cm
housing computed by EMPower (left) and on an open 4- by 6-cm PCB computed by
Momentum (right).
case study because it more accurately supports open structures and the
lower Q requires fewer frequency points.
This case study also illustrates the author’s insistent preference
throughout this book for smaller filter topologies. Smaller filters naturally
require smaller housings, which increases the frequency of the first
resonant mode, and therefore the stopband performance of the filter.
Figure 19.3 General simulation properties for the penetrating, stepped-impedance 2.4-GHz
lowpass filter.
Electromagnetic Simulation 297
Figure 19.4 Layout of the original penetrating, stepped-impedance 2.4-GHz lowpass filter
overlaid with the mesh automatically generated by the Momentum simulator (left), and after
narrowing and lengthening the open stubs (right).
Figure 19.5 Responses of the penetrating, stepped-impedance 2.4-GHz lowpass filter (right
graph) by circuit theory simulation (solid) and Momentum EM simulation (dashed), and the
responses after narrowing and lengthening the open stubs (right).
Figure 19.6 Layout for the 950- to 1050-MHz combline bandpass overlaid with the mesh
automatically created by the Momentum EM simulator.
Figure 19.7 Responses of the 950- to 1050-MHz combline simulated with circuit-theory
models (light traces) and by Momentum (heavy traces). The Momentum simulation with initial
capacitor values is on the left and after optimization of the capacitors is on the right.
300 Filter Synthesis Using Genesys S/Filter
Figure 19.8 Layout of the 1- to 2-GHz bandpass after modification of the placement of the
chip-capacitor pads.
Figure 19.9 EM-simulated responses of the 1- to 2-GHz bandpass before (left) and after
tuning to correct the responses (right).
In order to correct the return loss, the center frequency, and the
bandwidth of the filter, optimization must adjust several parameters of the
filter. Because each pass is computationally expensive for EM simulation,
and because of the requirement to adjust multiple parameters, optimization
requires significant time. A difficulty with bandpass filters is the
nonindependence of parameters. For example, increased line length might
adjust the bandwidth, but it also reduces the resonant frequency.
Therefore, to correct the responses, tuning rather than optimization is
applied. Iterative actions by an intelligent user, while tedious and time
consuming, is more effective than optimization in this case. If tuning is
simply trial and error, it is ineffective. However, certain principles guide the
tuning process. Principles common to many filter structures are:
1) Lengthening lines shifts the passband down.
2) Widening the spacing between coupled TLEs decreases the bandwidth.
3) Increasing the characteristic impedance of connecting lines in direct-
coupled filters decreases the bandwidth.
4) Adjusting the characteristic impedance of lines connecting to the
terminations changes the return loss properties.
5) Steps of one parameter at a time reveals behavior.
Notice in the responses on the left in Fig. 19.9 that the filter bandwidth is
too narrow and the center of the passband is too low. The author used
principles 1), 3), 4) and 5) in this case to tune the responses. The resulting
responses are given on the right in Fig. 19.9. The final tuned values are
given in Fig. 19.10.
302 Filter Synthesis Using Genesys S/Filter
Figure 19.10 Schematic of the 1- to 2-GHz bandpass with TLE dimensions after tuning to
correct the responses.
q1 = g0 × g1 (19.3)
gN
qN = for N even (19.5)
g N +1
where fupper is the passband upper cutoff frequency, flower is the passband
lower cutoff frequency and
304 Filter Synthesis Using Genesys S/Filter
f upper + f lower
f0 = (19.7)
2
Then the actual filter couplings and loaded Qs are
K n , n +1 = bw × kn , n +1 (19.8)
q1
Q1 = (19.9)
bw
qN
QN = (19.10)
bw
Capitalized symbols refer to actual couplings and loadings while
noncapitalized symbols refer to normalized values.
Figure 19.11 Resonator pair for finding the resonator coupling (left) and tapped resonator for
finding the external coupling (right).
line width at the open end, both the inductance and capacitance are
increased, thus lowering the resonant frequency. This supports a physically
shortened resonator [1]. Using two metal layers allows the open-end
portion of the resonator to be folded under the grounded end, thus halving
the physical length and further reducing the size. The resulting resonator is
very compact.
This structure is used by EM simulation to find the coupling between
resonators. In this case, the EMPower simulator is used because the
problem is easily placed on a large grid, and the natural x-y symmetry
results is fast simulation. The grid is 0.25 mm square.
Given in Figure 19.12 is the resulting transmission amplitude response
with a gap between resonators of four grid cells, or 1.o mm. The resonator-
coupling coefficient, K, is given by
f upper − f lower
K= (19.11)
f0
where fupper and flower are the frequencies of the system transmission peaks,
in this case 1802.8 and 1780 MHz, and f0 is the aritmetic mean of those
frequencies. In this case, K=0.01273. The external coupling during this test
should be light so as not to disturb the resonant frequencies of the pair. The
coupling is sufficiently loose if the attenuation between the peaks exceeds
20 dB.
Table 19.1 shows the results for EMPower simulations of gap spacings
0.25 to 1.0 mm. Ideally, the center frequency would be constant, but in this
case, it increases approximately 2% with increasing gap spacing. Given as
data points on the dashed trace in Fig. 19.13 are the K values. A line-fit
through these points is used to find the necessary gap spacing for any
required K in this range. A feature of the classic method is that K is typically
a monotonic function of the variable parameter and a smooth curve is
easily drawn through a small number of data points.
306 Filter Synthesis Using Genesys S/Filter
Figure 19.12 Transmission amplitude response of the coupled resonators (left) and
transmission amplitude response of the tapped resonator (right).
Table 19.1 EM Simulation Data for Coupling of the Resonators Shown on the Left in Fig.
19.11
Gap (mm) f upper (MHz) f lower (MHz) f 0 (MHz) K
0.25 1844.00 1666.00 1755.00 0.1014
0.50 1824.50 1738.00 1781.25 0.0486
0.75 1811.25 1766.25 1788.75 0.0252
1.00 1802.80 1780.00 1791.40 0.0127
Figure 19.13 EM-simulated resonator coupling, K, versus gap spacing (dashed line) and
singly loaded Qext versus tap location (solid line).
Electromagnetic Simulation 307
Figure 19.14 Initial layout of the 1800-MHz stepped-impedance bandpass filter designed
using the classic method (4X scale).
Figure 19.15 Initial responses of the 1800-MHz bandpass computed by EMPower (left) and
after tuning to correct the responses (right).
Figure 19.17 Final layout of the 1800-MHz bandpass after tuning (4X scale).
References
[1] R. Rhea, HF Filter Design and Computer Simulation, SciTech Publishing,
Raleigh, NC, 1994.
Electromagnetic Simulation 311
[2] R. Rhea, Filter Techniques: 3 DVD Set, SciTech Publishing, Raleigh, NC, 2003.
[3] R. Rhea, Practical Issues in RF Design: 3 DVD Set, SciTech Publishing, Raleigh,
NC, 2003.
[4] T. Ishii, ed., Handbook of Microwave Technology, Vol. 1, Components and
Devices, Academic Press, San Diego, 1995.
[5] R. Rhea, “EM Enables Classic Filter Technique,” Microwave Journal, February,
2007.
[6] K.V. Puglia, “A General Design Procedure for Bandpass Filters Derived from
Low Pass Prototype Elements,” Microwave Journal, Part I, December 2000, Part
II, January 2001.
Appendix A: Example Summary
313
314 Filter Synthesis Using Genesys S/Filter
319
320 Filter Synthesis Using Genesys S/Filter
323
Index
Add layout, 173 Distributed, 147
Allow inexact, 34 combline transform, 164
All-pass, 125 compare lumped and distributed, 147
All-pole, 1 half-angle transform, 162
Antimetry, 17 Ikeno transforms, 155
Bandpass, 2, 17 interdigital transform, 161
Bandwidth introduction, 147
absolute, 17 Kuroda identities, 152
fractional, 17 Kuroda-Minnis transform, 157
wide or narrow, 19 reentrance, 152
Blinchikoff, 119 Richards transform, 151
Box modes, 166 TLEs, 151
Canonic, 4 Distributed bandpass, 211
Cauer-Chebyshev, 13 combline capacitive coupled, 226
Cavity resonance, 299 combline hybrid, 249
Ceramic piezoelectric, 111 direct-coupled hybrid, 258
Chebyshev, 10 edge-coupled tutorial, 214
Classic design method, 307 edge-coupled with inverters, 218
example design, 317 generalized combline, 252
finding K by EM simulation, 313 generalized narrow, 238
finding Q by EM simulation, 316 hybrid, 253
fundamentals, 312 interdigital tutorial, 221
k & q values, 312 miniature with UEs, 228
Coaxial ceramic loaded Minnis class D, 250
generalized, 74, 79 unique, 229
Coaxial resonator Distributed bandstop, 199
from parallel L-C, 71 generalized narrow, 206
unloaded Q, 73 moderate bandwidth, 203
Combline narrow bandwidth, 205
capacitive coupled, 224 wide bandwidth, 199
diplexer, 284 Distributed highpass, 259
EM example, 307 bandpass synthesis, 272
generalized hybrid, 249 distributed synthesis, 259
hybrid penetrating, 253 generalized hybrid, 269
penetrating, 243 high degree, 277
transform, 161 hybrid, 259
Complementary multiplexers, 286 hybrid 10th degree, 283
Complex conjugate match, 130 hybrid 8th degree, 279
Compound transforms, 46 hybrid with UEs, 263
Conventional bandpass, 17 lumped synthesis, 260
Convert Using Advanced TLine, 73, 168 purely distributed, 268
Coupled lines, 159 Distributed lowpass, 169
Coupling coefficient, 152 approximate, 180
Customize exact, 169
permutation table, 30 generalized, 179
Degree, 4 hybrid, 194
Diplexer. See Multiplexer penetrating, 190
Dipole transform, 94 radial stub, 192
325
326 Filter Synthesis Using Genesys S/Filter