!

404 04,0
Final Report Covering the Period 30 September to 31 December 1962 in Detail,
with a Summary from 1 January 1961 to 31 December 1962

MICROWAVE FILTERS AND COUPLING STRUCTURES

Prepared for:
U.S. ARMY ELECTRONICS RESEARCH AND DEVELOPMENT LABORATORY
FORT MONMOUTH, NEW JERSEY CONTRACT DA 36-039 SC 87398
DA PROJECT 3A99-15-002-02-06

"By: G. L. Matthaei B. M. Schifflnan E. G. Cristal

L. A. Robinson

M AE N L0
QUALIFIED REQUESTORS MAY OBTAIN COPIES OF THIS REPORT FROM ASTIA.
ASTIA RELEASE TO OTS NOT AUTHORIZED.
 *SR
February I9ýS3

Final Report Covering the Period 30 September to 31 December 1962 in Detail,

with a Summary from 1 January 1961 to 31 December 1962

MICROWAVE FILTERS AND COUPLING STRUCTURES

Prepared for:
U.S. ARMY ELECTRONICS RESEARCH AND DEVELOPMENT LABORATORY
FORT MONMOUTH, NEW JERSEY CONTRACT DA 36-039 SC 87398
FILE NO. 40553.PM-61-93-93
DA PROJECT 3A99-15-002-02-06
SCL-2101N (14 JULY 1961)

By: G. L. Matthaei B. M. Schiffman E. G. Cristal

L. A. Robinson
SRI Project No. 3527
Objective: To advance the state of the art in the field of microwave filters
and coupling structures through applied research and development.

Approved:

(...............
....
...............
G. L. MATTHAEI, MANAGER ELECTROMAGNETIC TECHNIQUES LABORATORY

D. R, SCHEUCH, DIRECTOR ELECTRONICS AND RADIO SCIENCES DIVISION

Copy N16.......
N

LIST OF ILLUSTRATIONS . . . . . . . . . . . . . . . . . . .. . . . . . . . . .. v
LIST OF TABLES ... .......... . .............. ....................... .... ix
PURPOSE OF ThE CONTRACT ....................... ........................... xi
ABSTRACT ............................. .................................... xiii
CONFERENCES AND TECHNICAL PAPERS DURING LAST QUARTER .......... ............. xv

I INTRODUCTION . ........................... .............................. 1

II A BRIEF REVIEW OF PRIOR WORK ON TiIlS CONTRACT ........ ................ 3

A. General ........................ ..... ................................ 3
B. Book on Microwave Filters, Impedance-Matching Networks, and
Coupling Structures ... .......... !
................ .............. 3
C. Design Data for Rectangular Conductors Between Ground Planes 4
D. Analysis of the Phase and Filter Properties of Arrays of
Conductors Between Ground Planes ......................... . 5
E. Interdigital Band-Pass Filters .................. ..................... 5
F. Comb-Line Filters of Narrow or Moderate Bandwidth .... 77......
G. Stepped-Impedance Transformers and Filter Prototypes ........ ......... 8
IH. Wide-Band (and Narrow-Band) Filters Consisting of Cascaded
Transmission Lines with Lumped-Reactance Couplings ....... ........... 9
I. Band-Stop Filters ................... ........................... 10
J. Multiplexers ............................ .................... 11
K. Magnetically Tunable Filters Using Ferrimagnetic Resonators 12
L. Up-Converters for Use as Electronically Tunable Filters ....... 14

REFERENCES .......................... .................................. 16

III A NOVEL MAGNETICALLY TUNABLE WAVEGUIDE FILTER, AND A REFINED VERSION

OF A PREVIOUS STRIP-LINE MAGNETICALLY TUNABLE FILTER DESIGN .. . . ..... .. 19
A. General ............................. ................................ 19
B. A Magnetically Tunable Waveguide Filter Using Two YIG Resonators . .. 19
C. Design of the Magnetically Tunable Waveguide Filter .. ..... . . .. 22
D. Measured Performance of the Magnetically Tunable Waveguide Filter • 28
E. A More Refined Version of a Previous Strip-Line Magnetically
Tunable Filter Design .................. ....................... 32

REFERENCES .......................... .................................. 35

IV BAND-STOP FILTERS ..................... ............................ 37

A. General ...................... .................................. 37
B. Parallel-Coupled-Resonator Filter . .. ........... .... ................ 38
C. Spur-Line Type of Filter .............. ........................ 46
D. Capacitively-Coupled-Stub Filter .......... .................... 51
E. Derivation of the Design Equations .... ............ . ........... 54

REFERENCES .. ........ .................. .......................... 55

iii
 CONTENTS

V MULTIPLEXERS . . .. . . . . . . . . . . . . . . . . .. . . . . . . . . . . 57
A. General .... . ......... .................. . ....... .............. 57
B. Three-Channel Comb-Line Multiplexer Response . ........................ 57
C. Construction of the Three-Channel Comb-Line Multiplexer .......... .. 62
D. Tuning the Multiplexer ................. ......................... 65
E. Multiplexers with Guard-Bands Between Channels ... ............ .. 69
F. Susceptance Formula for an Ideal Multiplexer with Guard Bands ..... .. 73

REFERENCES .... .............................. ........................ 74

VI WIDEBAND INTERDIGITAL FILTERS WITH CAPACITIVELY LOADED RESONATORS ..... ... 75

A. General ........................ ................................. 75
B. Low-Pass Prototype Filters and Use of a Low-Pass-to-Band-Pass
Transformation .................... ............................ 77
C. Parallel-Coupled Lines ................. ............... . ......... 81
D. Design Equations ..................... ........................... 85
E. Calculated Frequency Responses ............. .................... 90
F. Derivation of Design Equations ............. .................... 99
G. Method of Frequency-Response Calculation ...... ................ 108

REFERENCES ........................... .................................. 112

VII CONCLUSIONS ........................ ................................ 113

A. Magnetically Tunable Filters ............. ....................... 113
B. Band-Stop Filters .................... ........................... 113
C. Multiplexers ..................... .............................. 113
D. Interdigital Filters with Capacitive Loading .... ............. .. 114

ACKNOWLEDGMENTS .................... ................................. . ....... 115

IDENTIFICATION OF KEY TECHNICAL PERSONNEL .......... ................... 116

iLv
 ILLUSTRATIONS

Fig. III B-l(a) Sectional View of Magnetically Tunable Waveguide Filter

with Two YIG Resonators ............ ........................ ..... 20
Fig. III B-1(b) Enlarged Sectional View of Center Portion of Magnetically
Tunable Waveguide Filter with Two YIG Resonators ... ............ .. 20

Fig. III B-2 Photograph of the Filter in Figs. III B-l(a) and (b) .. ......... .. 21
Fig. III C-1 A Four-Section, 50:1., Tchebyscheff, Waveguide Impedance Transformer . . 24

Fig. III C-2 Original and Modified Section Z of Transformer, Modified so as to

Provide Adequate Space for YIG Sphere ...... ................. ... 25
Fig. III C-3 Computed VSWB ofthe 50 ' Transformer in Fig. III C-1 with Section Z1
Replaced by Secti,-.s ZA and Z. in Fig, III C-2(b) ... ........... ... 27
Fig. III D-1 Attenuation vs. Biasing Field Strength for Magnetically Tunable
Waveguide Filter ................. ........................... ... 30
Fig. III D-2 Pass-Band Attenuation vs. Swept Biasing Field Strength for the
Magnetically Tunable Waveguide Filter ....... .................. ... 31
Fig. II( E-1 A Strip-Line Two-Resonator Filter with YIG Besonators .. ........ .. 33
Fig. III E-2 Close-Up View of Filter in Fig. III E-1 Mounted in an Electromagnet . . 34

Fig. IV B-1 Parallel-Coupled Transmission-Line Section and Equivalent

Section of Basic Filter .............. .................... ... 38
Fig. IV B-2 Basic Band-Stop Filter ............... ........................ ... 40

Fig. IV B-3 An Unsymmetrical Pair of Parallel-Couple& Lines Ca, Ceb, and C6

are Line Capacitances per Unit Length ....... ................. ... 40
Fig. IV B-4 Cross Section of Unsymmetrical, Rectangular-Bar Parallel-Coupled
Lines . . ............................................ .......... 41
Fig. IV B-5 A Parallel-Coupled-Line, Narrow-Band Filter with Cover
Plate Removed . ................. ............................ .... 43
Fig. IV B-6 Sketch of a Parallel-Coupled-Lin,, Narrow-Band Filter with Cover
Plate Removed, Showing Dimensions .............. ................... 44
Fig. IV B-7 Measured and Computed Response of Filter in Fig. IV B-5 .......... ... 45
Fig. IV C-1 Spur-Line Stub Resonator and Equivalent Section of Basic Filter . . . . 46

Fig. IV C-2 Three Possible Ways of Converting a Series Stub Filter (Dual of the
Basic Filter) to a Spur-Line Filter ............ ................... 48
Fig. IV C-3 A Spur-Line Band-Stop Filter with Cover Plate Removed ... ......... ... 49
Fig. IV C-4 Sketch of a Spur-Line Band-Stop Filter with Cover Plate Removed,
Showing Dimensions ................. .......................... ... 49
Fig. IV C-5 Measured and Computed VSWH and Attenuation of Filter of Fig. IV C-3 . . 50
Fig. IV D-1 Schematic Diagram of Capacitively Coupled Filter, and Its
Measured and Computed Response ......... ..................... ... 52
Fig. IV D-2 Capacitively Coupled Narrow-Band Filter with Cover Plate Removed . . . 53

V
 ILLUSTRATIONS

Fig. IV D-3 Sketch of Capacitively Coupled Narrow-Band Filter with Cover

Plate Removed, Showing Dimensions ................ ...................... 54
Fig. V B-1 Photograph of the Three-Channel Comb-Line Multiplexer .... ............ ... 58
Fig. V B-2 Attenuation of the Three-Channel Comb-Line Multiplexer ... .......... ... 59
Fig. V B-3 Attenuation in the Pass-Band of the Three-Channel Comb-Line Multiplexer . 60
Fig. V B-4 VSWR in the Pass-Band of the Three-Channel Comb-Line Multiplexer 6 1........
Fig. V C-1 Basic Comb-Line Filter Configuration Showing Important Dimensions ..... 63
Fig. V C-2 Sketch of the Top View'of the Three-Channel Comb-Line Multiplexer
Showing One of the Comb-Line Filters with its Cover Plate Removed ..... 64
Fig. V C-3 Detailed Drawing of the Common Junction .............. ............. 66
Fig. V C-4 Detailed Drawing of the Annulling Stub Outer Conductor .. .......... . 67
Fig. V E-i Admittance of a Hypothetical Multiplexer Using 1.0-db Tchebyscheff Ripple,
Singly-Terminated Filters .............. ......................... .... 70
Fig. V E-2 Admittance of Hypothetical Multiplexer with Annulling Stub .. ......... .. 71
Fig. V E-3 Attenuation of Hypothetical Multiplexer with Annulling Stub . .... ........ 72
Fig. V F-1 Normalized Real Part of the Admittance of a Hypothetical Multiplexer
with Guard-Bands Between Channels ................ ...................... 74
Fig. VI A-1 Capacitively Loaded Interdigital Filter with Ungrounded End Resonators . . 75
Fig. V1 B-1 Definition of Prototype Filter Parameters ........ ................. ... 77
Fig. VI B-2 Equations and Parameters for Maximally Flat Response ... ............ ... 78
Fig. VI B-3 Equations and Parameters for Tchebyscheff Response .... ............ ... 79
Fig. VI C-1 Cross Section of Unsymmetrical, Rectangular-Bar Parallel-Coupled Lines • 81
Fig, VI C-2 Cross Section of an Array of Parallel-Coupled Lines Between Ground
Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .. c3
Fig. VI D-1 Various Configurations for Loading Capacitances at Ends of Line
Elements ....................... .................................. ... 90
Fig. VI E-1 Frequency Response of a Capacitively Loaded Interdigital Filter
Designed with Admittance-Scaling Parameter d = 0.25 ..... ............ ... 93
Fig. VI E-2 Frequency Response of a Capacitively Loaded Interdigital Filter
Designed with Admittance-Scaling Parameter d = 0.50 ............... 94
Fig. VI E-3 Frequency Response of a Capacitively Loaded Interdigital Filter
Designed with Admittance-Scaling Parameter d = 0.65 ............... 95
Fig. VI E-4 Frequency Response of a Capacitively Loaded Interdigital Filter
Designed with Admittance-Scaling Parameter d = 1.00 ...... ........... ... 96
Fig. VI E-5 Frequency Response over an Extended Frequency Range for the
Capacitively Loaded Interdigital Filter with Admittance-Scaling
Parameter d= 0.65 ................... ............................ ... 98

Fig. VI F-1 Capacitively Loaded, Parallel-Coupled Lines Used as Interior

Resonators in Fig. VI A-1 .................... .......................... 101
Fig. VI F-2 Capacitively Loaded, Parallel-Coupled Lines Used at the Ends of
the Filter Structure Shown in Fig. VI A-1 ........ .................. .101
Fig. VI F-3 An Open-Wire-Line Equivalent Circuit for the Capacitively Loaded
interdigital Filter Shown in Fig. VI A-i .... .......... ........ 103

,i
 ILLUSTRATIONS

Fig. VI F-4 The Circuit of Fig. VI F-3 Split into Symmetrical Interior Sections,
and the Transformer Removed .......... ...................... ... 103
Fig. VI F-S Modified Prototype for Deriving Design Equations for Capacitively
Loaded Interdigital Filters ........... ....................... ... 104

vii
 TABLES

Table III D-1 Measured Performance of the Magnetically Tunable Waveguide Filter . . . 29
Table III E-i Measured Performance of the Two-Resonator Strip-Line Filter in
Figs. III E-1 and III E-2 ............. ........................ ... 35

Table V C-1 Data for Three-Channel Comb-Line Multiplexer ...... .............. ... 64
Table VI D-1 Design Equations for Interdigital Filters of the Form
in Fig. VI A-1 ...................... ............................... 86
Table VI E-1 Line-Element Parameters for an Octave-Bandwidth, Capacitively
Loaded Interdigital Filter with Admittance-Scaling Parameter d - 0.25 91
Table VI E-2 Liiie-Element Parameters for an Octave-Bandwidth, Capacitively
Loaded Interdigital Filter with Admittance-Scaling Parameter d = 0.50 91
Table VI E-3 Line-Element Parameters for an Octave-Bandwidth, Capacitively
Loaded Interdigital Filter with Admittance-Scaling Parameter d a 0.65 91
Table VI E-4 Line-Element Parameters for an Octave-Bandwidth, Capacitively
Loaded Interdigital Filter with Admittance-Scaling Parameter d - 1.00 . 92

ix
 PURPOSE OF TIlE COMNACT

This contract continues the work of Contracts DA 36-039 SC-64625,

DA 36-039 SC-63232, and DA 36-039 SC-74862 to develop new design tech-
niques for waveguide, strip-line, and coaxial-line microwave components,
and to prepare a book on techniques for microwave filter design.

xi
 ABSTRACT

Section II of this report briefly summarizes the work .of the

two-year course of this contract. In Sec. III a two-resonator waveguide
magnetically tunable filter and a two-resonator strip-line magnetically
tunable filter are described. These filters utilize ferrimagnetic
resonance in single-crystal yttrium-iron-garnet (YIG) to achieve tuning
which is controlled by a biasing dc magnetic field. The waveguide
filter is for X-band and uses a novel coupling structure which increases
the coupling between the external circuit and the YIG resonators. This
makes it possible to use unusually small resonators (0.025-inch diam-
eter), which helps to reduce spurious response activity in the stop
bands. The strip-line filter is for S-band and is an advanced version
of a design discussed in Quarterly Progress Report 7. In Sec. IV of
this report our previous work on the exact design of band-stop filters
is extended to cover additional strip-line (or coaxial) configurations
which have practical advantages. These include a configuration where
the resonator lines are parallel to, but separate from, the main trans-
mission line, and a configuration where the resonators are parallel to
the main transmission line and attached to it. The first form is ad-
vantageous for narrow-stop-band filters, while the latter form is
advantageous for wide-stop-band filters. Design equations and the
computed and experimental results of trial designs are given. In Sec. V
the experimental results of a trial, three-channel multiplexer using
comb-line filters are discussed. The results are found to be in good
agreement with the theory previously discussed in Quarterly Progress
Report 6. This particular multiplexer has contiguous pass-bands, and
a mathematical analysis is presented for the corresponding case where
guard-bands are desired between channels. In Sec. VI design equations
are presented for wideband, interdigital band-pass filters having
capacitive loading at one end of each resonator line, and a short cir-
cult at the other. The use of capacitive loading is seen to permit even
more compact filter structures, and causes the first spurious response
to be unusually far removed from the main pass-band.

xiii
 CONFERENCES AND TECHNICAL PAPERS DURING LAST QUARTER

Conferences,

None

Technical Papers

Leo Young, "The Practical Realization of Series-Capacitance Couplings

for Microwave Filters," The Microwave Journal, Vol. 5, pp. 79-81
(December 1962).

Leo Young, "Stepped-Impedance Transformers and Filter Prototypes,"

IRE Trans. PGMTT-1O, pp. 339-359 (September 1962).

G. L. Matthaei, "Interdigital Band-Pass Filters," IRE Trans. PGMTT-1O,

pp. 479-491 (November 1962).

Leo Young, G. L. Matthaei, and E. M. T. Jones, "Microwave Band-Stop

Filters with Narrow Stop Bands," IRE Trans. PGMTT-1O, pp. 416-427
(November 1962).

Y. Sato and P. S. Carter, "A Device for Rapidly Aligning and Mounting
Ferromagnetic Single Crystals Along Any Desired Axis," IRE Trans.
PGMTT-1O, pp. 611-612 (November 1962).

xv
 I IMrD IO!U

Section II gives a brief summary of the previous work done on this

contract, while the remainder of this report describes results obtained
during the last quarter.

Various means for electronic tuning have been studied on this project,
including tuning by use of electronically tunable up-converters, and by
use of filters having ferrimagnetic resonators which are tuned by varying
a biasing magnetic field. Our latest results on magnetically tunable
filters will be found in Sec. III of this report. Our previous work on
band-stop filters is extended in Sec. IV, where some new configurations
for band-stop filters are described, and the results of some trial designs
are presented. In Quarterly Progress Report 7 on this contract an anal-
ysis of a method for multiplexer design was made, and the performance of
some trial designs was computed. In Sec. V. of this report the results
obtained from a three-channel multiplexer which was constructed, are
given. This multiplexer has contiguous pass-bands. Analysis of the case
where there are guard bands between the pass-bands is also presented. In
Sec. VI of this report our previous work on interdigital filters is ex.-
tended to cover the case of interdigital filters with capacitively loaded
resonators. The addition of capacitive loading makes possible a much
broader upper stop-band, and yields an even more compact structure.
 II A BRIEF REVIEW OF PRIOR WORK ON THIS CONTRACT

A. GENERAL

In this section the contents of the previous reports on this contract,

and of a book prepared on this contract, are: described. Also, since nu-
merous technical papers have resulted from this contract, other references
are given in which some of the results of the contract are discussed.

B. BOOK ON MICROWAVE FILTERS, IMPEDANCE-MATCHING

NETWORKS, AND COUPLING STRUCTURES

One of the major activities on this contract has been to prepare an

extensive book dealing with practical design techniques for microwave
filters, impedance-matching networks, and coupling structures.'* This
book is largely an outgrowth of our research activities for the Signal
Corps under Contracts DA 36-039 SC-63232, DA 36-039 SC-64625, DA 36-039
SC-74862, and DA 36-039 SC-87398. The preparation of this book has been
mainly supported by this contract (DA 36-039 SC-87398), though its prep-
aration has been partly supported by Stanford Research Institute and by
off-work time contributed by the authors. The book will be issued for
the Signal Corps at about the same time as this report. An abstract is
given below.

Abstract

The book prepared on this contract presents design techniques

for a wide variety of low-pass, band-pass, and band-stop
microwave filters; for multiplexers; and for certain kinds
of directional couplers. The material is organized to be
used by the designer who wants to work out a specific design
quickly, with a minimum of reading, as well as by the engineer
who wants a deeper understanding of the design techniques used
to be able to apply them to new and unusual situations.

Most of the design procedures described make use of either a

lumped-element low-pass prototype filter or a step-transformer
prototype as a basis for design. Using these prototypes,
microwave filters can be obtained which derive response

Reference@ are listed at the end of each major section of the report.

I
 characteristics (such as a Tchehyscheff attenuation ripples
in the pass-band) from their prototype. Tabulated prototype
filter designs are given as well as data relevant to the use
of prototype filters as a basis for the design of impedance-
matching networks and time-delay networks. Data for step-
transformer prototypes are also given.

The design of microwave filter structures to serve as

impedance-matching networks is discussed, and examples are
presented. The techniques described should find application
in the design of impedance-matching networks for use in
microwave devices such as tubes, parametric devices, antennas,
etc., to achieve efficient broad-band operation. The design
of microwave filters to achieve various time-delay (or slow-
wave) properties is also discussed.

Various equations, graphs, and tables are collected together

relevant to the design of coaxial lines, strip-lines, wave-
guides, parallel-coupled lines between common ground planes,
arrays of lines between ground planes, coupling and junction
discontinuities, and resonators. Techniques for measuring
the Q's of resonators and the coupling coefficients between
resonators are also discussed, along with procedures for tuning
filters. Equations and principles useful in the analysis of
filters are collected together for easy reference and to aid
the reader whose background for the subject matter of this book
may contain some gaps.

Directional filters with special properties valuable for

certain applications are treated in detail in a separate
chapter, as are high-power filters. Tunable filters of the
kind that might be desired for preselector applications are
also treated. Both mechanically tunable filters and filters
using ferrimagnetic resonators, which can be tuned by varying
a biasing magnetic field, are discussed.

C. DESIGN DATA FOR RECTANGULAR CONDUCTORS

BETWEEN GRBUND PLANES

In many structures such as directional couplers and filters which

use parallel-coupled line elements as resonators, it. is often desirable
to fabricate the line elements as rectangular bars between ground' planes.
A mathematical solution for the even- and odd-mode fringing capacitance
associated with rectangular bars between ground planes was obtained on
this project. Data was computed from these equations by use of a digital
computer, and extensive curves of data were plotted to give easy-to-use
design charts. These charts make the design of rectangular bars between
ground planes, arrays of bars between ground planes, or rectangular bars
within rectangular housings, quite simple.

4
 The results of this research were described in Sec. II of Quarterly
Progress Report 2 on this contract. 2 They have also been published in
the IRE Transactions on Microwave Theory and Techniques; 3
and the design
charts also appear in Sec. 5.05 of the book prepared on this contract.1

D. ANALYSIS OF THE PHASE AND FILTER PROPERTIES OF ARRAYS

OF CONDUCTORS BETWEEN GROUND PLANES

A number of structures were analyzed which consisted of arrays of

parallel conductors between ground planes or above a single ground plane.
These included: interdigital line, meander line, a form of helix,
"hairpin line," and reactively loaded comb line. Equations were derived
for determining the phase function 0 per section of line in terms of
voltage coupling factors K1 , which are obtained from the static capaci-
tances per unit length of the conductors. Equations for the midband
image impedance of finite arrays of conductors were obtained for three
of the five cases considered. Data for computing coupling factors were
presented for the case of arrays using small wires between ground planes,
and for arrays having conductor dimensions such that fringing capacitance
beyond nearest neighbors can be neglected. The phase characteristics
for given coupling factors were presented for examples of the various
types of structures considered. The results of the examples indicated
that interdigital lines and comb lines look especially attractive as
potential filter structures, or as slow-wave structures. This is be-
cause of the relatively efficient coupling between the resonant elements
of structures of these types, for given spacings between the elements.

A trial intexrdigital structure was fabricated to check out the

predicted cutoff frequency and midband image impedance. Good agreement
between theory and experiment was obtained. By modifying the end ele-
ments on the structure it became a well matched filter having about an
octave pass-band wi-dth.

This work is reported in Quarterly Progress Report 14 on this

contrac~t and also in the Proceedings of the IRE.5

E. INTERDIGITAL BAND-PASS FILTERS

The study of arrays of conductors between ground planes, discussed

above, showed that interdigital-line structures are of considerable

5
potential interest for use as band-pass filter structures. The study
described in Sec. II-D above was largely a performance analysis study,
and further research was conducted to find a practical synthesis pro-
cedure for interdigital filters. By use of certain approximations,
easy-to-use and reasonably accurate synthesis procedures were obtained.

In Sec. III of Quarterly Progress Report 4 on this contract,6 the

design of band-pass filters using interdigital arrays of resonator line
elements between parallel ground planes is discussed. Two approximate
design procedures are described, both of which permit des-ign directly
from lumped-element, low-pass, prototype filters. Both design procedures
will work for either narrow- or wide-band filters, but one proc.edure
gives more practical dimensions for filters having wide bandwidths (such
as an octave), while the other gives more practical dimensions for fil-
ters having narrow or moderate bandwidths. The dimensions and measured
performance curves are presented for a 10-percent-bandwidth design and
an octave-bandwidth design.

Interdigital band-pass filters were demonstrated to have a number

of very attractive features, including the following:

(]) They are very compact.

(2) The tolerances required in their manufacture are un-

usually relaxed as a result of the relatively large
spacings between resonator elements.

(3) The second pass-band is centered at three times& the

center frequency of the first pas-s'-band, and there is
no possibility of spurious responses in between. (Note
that for filters with half-wavelength, parallel-coupled
resonators, and for many stub filters, even the slightest
mistuning will result in narrow spurious pass-bands at
twice the frequency of the first pass-hand center.)

(4) The rates of cutoff and the streng-th 'of the stop-bands
are enhanced by multiple-order poles of attenuation at
dc and at even multiples of the center frequency of the
first pass-band.

(5) Interdigital filters can be fabricated in structural

forms which are self-supporting so that dielectric
material need not be used. Thus, dielectric loss can
be eliminated.

Besides being discussed in Quarterly Progress Report 4,6 interdigital

filters were discussed in a paper presented at the 1962 PGMTT National

6
 Symposium at Boulder, Colorado,7 and in a written version published in
8
the IRE Transactions on Microwave Theory and Techniques. Interdigital
filters are also discussed in Secs. 10.06, 10.07 and 10.09 of the book
prepared on this contract.' The design of interdigital filters with
capacitive loading is discussed in Sec. VI of this report.

F. COMB-LINE FILTERS OF NARROW OR MODERATE BANDWIDTH

The study discussed in Sec. II-D above also indicated that comb-line
structures should be of special interest as potential band-pass filter
structures. That study showed that comb-line filters must use capacitive*
loading on the resonators if a pass-band is to be achieved, and if suf--
ficiently heavy capacitive loading is used, relatively large coupling
can be obtained between resonator lines for given spacing between the
resonator line elements. A study of the design of comb-line filter
structures was made, and a design procedure was obtained which is suit-
able for the design of comb-line filters of narrow to moderate bandwidth.

The design of comb-line band-pass filters is discussed in Sec. II

of Quarterly Progress Report 5.9 The resonators in such a filter con-
sist of TEM-mode lines between parallel ground planes; each resonator
line is short-circuited to ground at one end, and capacitively loaded
to ground at the other. Coupling between resonators is achieved by way
of the fringing fields between the lines. To couple into the structure,
nonresonant short-circuited lines are used at each end. The design
equations given make use of a lumped-element low-pass prototype and are
suitable for use in designs of narrow to moderate bandwidths (i.e.,prob-
ably up to around 15-percent bandwidth, although this point has not yet
been fully checked). A trial filter design for 10-percent bandwidth was
constructed and tested with good results.

Comb-line filters were demonstrated to have the following attractive

features:

(1) They are compac~t (even more so than interdigital filters).

(2) They have strong stop-bands and the stop-band abov-e the
primary pass-band can be made to be very broad (in our
exper-imental model the first spurious pass-band was at
4.2 times the frequency of the primary pass-band).

(3) If desired, they can be designed to have an unusually

steep rate of cutoff on the high side of the pass-band.
* Or inductive loading.

7
 (4) Adequate coupling can be maintained between resonator
elements with sizeable spacings between resonator lines.
(This feature means that the proper couplings can be
ma intained in manufactured filters without unreasonable
tolerance requirements.)

(5) Filters of this type can usually be fabricated without

the use of dielectric support materials so that, if
desired, dielectric losses can be eliminated.

Besides the discussion of comb-line filter design in Quarterly

9
Progress Report 5 on this contract, their use in multiplexers is ana-
10
lyzed in Sec. III of Quarterly Progress Report 6, and the results of
a trial, three-channel multiplexer design using comb-line filters are
discussed in Sec. V of this report. Comb-line filters
are also discussed
in Secs. 8.13, 8.14, and 16.04 of the book prepared on this contract.1

G. STEPPED-IMPEDANCE TRANSFORMERS AND FILTER PROTOTYPES

In Sec. IV of Quarterly Progress Report 46 the theory of the

quarter-wave transformer is reviewed and extended, and the major results
are presented. The distinctions between ideal and non-ideal junctions,
homogeneous and inhomogeneous transformers, synchronous and nonsynchro-
nous tuning, are brought out explicitly. Design formulas are presented,
numerical tables are given, methods of numerical calculation are ex-
plained, and their use is illustrated by many examples. Where exact
solutions are not available, two approximate design procedures are given:
one is applicable where R is relatively small, and one where B is rela-
tively large (where R is the ratio of the terminations). To help in
obtaining accurate numerical solutions, t-he connection between antenna
arrays and "small-B" transformers is utilized, as is the connection
between lumped-constant, low-pass filters and&"large-R" transformers.

The concept of half-wave filters (derived from quarter-wave trans-

formers) is, explained, and is later applied in Quarterly Progress
Report 59 to the design of reactance-coupled filters.

Design formulas for inhomogeneous transformers are presented. A

nonsynchronous transformer which is sometimes useful, is also described.

The last section shows how to calculate the dissipation loss and
the group delay at center frequency in terms of the transmission-line
filter parameters. A formula connecting dissipation loss and group
delay in the pass-band is given.

8
 In addition to the discussion of the above topics in Quarterly
Progress Report 4,6 a similar discussion will also be found in Chapter 6
of Ref. 1. Some of this material has also been published in the IRE
Trans. PGMTT,n the Journal of the Optical Society of America, 1 and The
Microwave Journal. 13

H. WIDE-BAND (AND NARROW-BAND) FILTERS CONSISTING

OF CASCADED TRANSMISSION LINES WITH LUMPED-
REACTANCE COUPLINGS

Design procedures for reactance-coupling filters with half-wave and

quarter-wave transmission-line resonators are given in Sec. III of
Quarterly Progress Report 5.9 The resonator transmission lines are
coupled either by series capacitances or shunt inductances., Over narrow
frequency bands, such filters show characteristics similar to those of
lumped-constant filters, and their design is straightforward. The de-
sign of direct-coupled resonator filters over wide (as well as narrow)
frequency bands is presented using quarter-wave transformers as proto-
type circuits.

Each filter has the same number of discontinuities spaced along a

transmission line as has its step-transformer prototype, with correspond-
ing discontinuities of the filter and step transformer having the same.
VSWR. Both the filter and its prototype are synchronously tuned circuits,
which means that adjacent discontinuities have out-of-phase reflection
coefficients giving the maximum cancellation. This condition determines
the spacing between discontinuities. It is also shown how the line
impedances of a reactance-coupled filter may be chosen to obtain a nearly
equal-ripple response.

Design data and graphs are given to facilitate the prediction of

filter performance when the filter is based on a selected prototype
transformer. The method of design is first to select a prototype trans-
former and then to see if the predicted filter performance matches the
specifications. If necessary, another prototype transformer has to be
selected to yield another filter meeting the specifications more closely.
This is illustrated by several numerical examples having fractional band-
widths from 10 to 85 percent.

9
 The effec.t of repeating the central elements of an eight-resonator
filter is shown for a pseudo-high-pass filter. Formulas for calculating
the gap capacitance in a circular coaxial line for this type of filter
are also given.

Be-sides being discussed in Quarterly Progress Report 5, a discussion

of the above subjects will also be found in Chapter 9 of Ref. 1. Several
items from the abov-e material have appeared or are about to appear in
The Microwave Journal.14'IS

I. BAND-STOP FILTERS

A study was made on this contract to determine ways of designing

band-stop filters in strip line or in waveguide. This work is discussed
in several reports.

In Sec. III of Quarterly Progress Report 316 an approximate design

procedure is presented for use in the design of strip-line or waveguide
band-stop filters having, narrow stop-bands. In order to obtain the
desired narrow stop-band', the band-stop resonators in strip line are
constructed from a short.-circui~ted length of line which is capacitively
coupled to the main transmission line. For waveguide, an analogous
structure is used where the resonators are inductively coupled to the
main waveguide by use of inductive irises.. Equations are presented for
estimating the effects of dissipat-on loss in the circuit, and procedures
are outlined for experimentally adjusting the resonator couplings by use
of laboratory tests. A trial, three-resonator strip-line band-stop
filter was constructed, and good measured performance was obtained.

The measured characteristics of a three-resonator, waveguide, band-

stop filter are discussed in Sec. II of Quarterly Progress Report 4.6
The design of this filter had been discussed in Quarterly Progress
Report 3.16 It was found necessary to use three-quarter-wavelength
spacings between resonators (instead of quarter-wavelength spacings)
in order to avoid undesirable interaction between resonators. Using
this larger resonator spacing, excellent agreement was obtained between
the computed and measured response for the filter, although both re-
sponses had somewhat larger pass-band attenuation ripple than was called
for by the. prototype. This was found to be due to the selectivity
effect of the coupling lines between resonators.

10
 In Sec. II of Quarterly Progress Report 717 an exact method is pre-
sented for the design of band-stop filters consisting of stubs and
quarter-wavelength connecting lines. Using this method, a low-pass
prototype circuit is chosen, and the transmission-line band-stop filter
derived from the prototype will have a response that is an exact mapping
of the prototype response. In theory the stop-band can have any width,
but if the stop-band is very narrow the stub impedances become unreason-
able. In such cases it is found to be desirable to replace the stubs
by capacitively or inductively coupled resonators of reasonable impedance.
This introduces an approximation but gives very good results. The design
of filters having three-quarter-wavelength spacings between resonators
is also discussed. This latter case is of interest for the design of
waveguide band-stop filters where it has been found desirable to separate
the resonators by three quarter-wavelengths in order to avoid interactions
between the fringing fields at the coupling irises.

The exact design methods discussed above are further extended to

cover the design of band-stop filters in additional novel forms, in
Sec. IV of this report.

Besides the discussions in Quarterly Progress Reports 3, 4, and 6,

and in this report,
some of the above material was discussed in a paper
presented at the 1962 PGMTT National Symposium in Boulder, Colorado, 1
and also in a version published in the IRE Transactions on M-icrowave
Theory and Techniques. 19 Most of the material on band-stop' filters in
Quarterly Progress Reports 3, 4, and 6 also appears in modified form in
1
Chapter 12 of the book prepared on this contract.

J. MULTIPLEXEBS

A technique for the precision design of multiplexers was investi-

gated on this contract.

10
In, Sec. III of Quarterly Pirogress Report 6 a study of the design
of multiplexers having contiguous pass-bands is presented. In the design
procedure discussed, the individual filters are designed from Tchebyschef~f
low-pass prototype filters havi~ng a resistor termination at one end only.
This approach can be used for the design of either series-connected or
shunt-connected multiplexers. In either case an additional reactance-
or susceptance-annulling network is required in order to obtain optimum

S~11
performance. A four-channel, lumped-element design is worked out, and
its response is computed. Also, a three-channel design using comb-line
filters is worked out, and its response computed. In both casaes very
low-loss performance is achieved.

In 'Sec. V of this report the measured performance of a three-channel

diplexer with contiguous, pass-bands, designed by the methods described
above, is presented. Also, the use of this 'same technique for the de-
sign of multiplexers having guard-bands between the channel bands is
evaluated.. Trial designs are worked out and their responses computed.
It is found that these same techniques can be applied to.the design of
multiplexers with guard-bands between the channel bands, but a different
type of susceptance- or reactance-annulling network is then required.

Most of this work on multiplexers will be found in somewhat dif-

1
ferent form in Chapter 16 of the book prepared on this contract.

K. MAGNETICALLY TUNABLE FILTERS USING FERRIMAGNETIC RESONATORS

On this contract, work on magnetically tunable filters begun on

Contract DA 36-039 SC-74862 2 0 was continued. These filters make use of
ferrimagnetic resonators consisting of spheres of single-crystal yttrium-
iron-garnet (YIG) or gallium-substituted YIG (Ga-YIG). The resonant
frequency of such resonators can be controlled by varying a biasing dc
magnetic field.

In Sec. II of Quarterly Progress Report 316 on this. contract, design

techniques are discussed for use in the 2.0-to-4.2-Gc tuning range. A
filter was designed, constructed, and tested which used overlapping
sections of strip transmission line to couple to the resonators. The
resonators were coupled together by an iris, preferably a slot because
it gives the highest attenuation at frequencies far from the center
frequency.

Design formulas and curves for this type of tunable filter are
given in Quarterly Prog.ress Heport 3.16 Test results on the experimental
filter, mentioned above, are presented, including the following:

(1) Response curves four various center frequencies from

2.0 Gc to 4.8 Gc

(2) A comparison of responses using circular and slot irises

to couple between resonators

12
 (3) Curves showing center frequency as a function of
dc magnetic bias field

(4) The saturation power at center frequency as a function

of center frequency

(5) The effects of the conducting boundaries on the reso-

nant frequency and unloaded Q of the resonators.

The methods of carrying out calculations of the effects of temperature

are outlined. The feasibility of using gallium-substituted yttrium-iron-
garnet as a resonator material for the 1.0-to-2.0-Gc range is discussed,
and the results of measurements of the unloaded Q of this material in
the 1.0-to-4.0-Gc range are given.

9
In Quarterly Progress Report 5 on this contract, the design, con-
struction, and testing of a magnetically tunable filter that uses an
arrangement of spherical YIG resonators lying in a plane perpendicular
to the direction of the dc magnetic biasing field is described. A mini-
mum dc magnetic air-gap spacing is required for this type of filter.
Measurements of attenuation loss, bandwidth, and tuning fields of the
two-resonator prototype are discussed. The attenuation loss varies from
4.0 db at a center frequency of 2.1 Gc, to 1.65 db at 3.9 Gc. The 3-db
bandwidth is about 32 Mc and is nearly constant over this tuning range.
Tuning requirements, for the filters with three or more resonators are
discussed. A new device by mean-s of which ferrimagnetic crystals can
be quickly aligned along any desired axis is described. The results of
anisotropy measurements made on crystals of YIG and Ga-YIG using this
device are presented.

In Sec. III of Quarterly Progress Report 7 on this contract1 a re-

view of the procedure employed in the establishment of the parameters
and dimensions of magnetically tunable filters is given. Charts and
graphs are presented which enable a designer to arrive at the dimensions
of the circuit elements, including the ferrimagnetic resonators, in
order to give a specified band-pass frequency response., Circuit struc-
tures are presented which appear or have proved to be very promising for
practical applications. Two typical examples are given, of the design
of two-resonator band-pass filters, one of which employs a waveguide and
the other a strip-transmission-line coupling circuit. Finally, the per-
formances of two side-wall-coupled strip-transmission-line filters, one
containing two resonators and the other containing three, are discussed.

13
The results are compared with a previous version of the two-resonator
side-wall-coupled filter and with a previously developed overlapping
line version. As expected, the three-resonator design was found to give
"asteeper rate of cutoff, and very high off-resonance attenuation.

21
Some of this work was also discussed in a technical paper and in
"aletter to the editor, in the IRE Transactions on Microwave Theory and
Techniques. An extensive discussion on this subject also appears in
1
Chapter 17 of the book prepared on this contract.

L. UP-CONVERTERS FOR USE AS ELECTRONICALLY TUNABLE FILTERS

Another technique for electronic tuning studied on this contract

23
(as well as on the preceding Contract DA 36-039 SC-74862) is the use
of variable-capacitance diode up-converters as electronically tunable
filters. These devices achieve a, wide ;tuning range (though they are
narrow-band at any given tuning frequency) by use of a wide-band input-
impedance-matc-hing filter, a narrow-band sideband output-impedance-
matching filter, and a voltage-tunable pump oscillator such as a
backward-wave oscillator.

Defining f as the input frequency, f' as the sideband output fre-

quency, and fP as the pump frequency, for a lower-sideband up-converter
the output is at the lower-sideband frequency

P - fP - f (II L-1)

For an upper-sideband up-converter, the output is at the upper-sideband

frequency

S f- P + f (II L-2)

For either type of up-converter, tuning action can be achieved if a

narrow-band filter is used at the output so that only frequencies equal
or very nearly equal to a specific frequency, f0, can be passed. If
the pump frequency fP is varied, the input frequency which will be ac-
cepted by the amplifier will be given by

f - f, -f' (II L-3)

14
for the case of lower-sideband up-converters, and by

f P f= - f, (II L-4)

for the case of upper-sideband up-converters. In both cases the ampli-

fiers will yield gain. The lower-sideband type introdutces some negative-
resistance amplification in addition to the up-converter amplification,
however, and will therefore generally give more gain. For gain to be
achieved, the variable-capacitance diode must see proper terminations
at both frequencies, f and f0. Since f0 is a fixed frequency, it is
relatively easy to maintain proper termination at that frequency. The
input frequency, f, varies, however, and the tuning range of the ampli-
fier is determined largely by the range of f for which proper terminating
conditions can be maintained. Because of the narrow-band output circuit
used in devices of this type they are capable of tuning ranges of the
10
order of an octave or more.

A trial strip-line lower-sideband up-converter was constructed on

23
this contract using the previously developed theory. The measured 3-db
bandwidth tuning range was 38.5 percent as compared to 40 perc~ent for
the design objective, and the peak gain was 12.6 db. The input band
center was 946 Mc while the sideband output was at 4,037 Mc. The meas-
ured noise figure was in good agreement with the computed nominal noise
figure of 2.1 db. The design and the measured performance of this de-viice
2
is discussed in Quarterly Progress Reports 2 and. 3 for this contract. '1

In Sec. II of Quart-e-rly Progress Report 6 for this contract, 10 our

6
previous theoretical work231 on electronically t-unable up-converters is
further extended. The previous design theory neglected some of the
parasitic parameters of the diode, which can become quite important if
the up-converter is to operate with microwave input frequencies. Design
procedures suitable for programming on a computer are developed for the
cases of electronically tunable upper--sideband up-converters. The diode
is assumed to be resonated by a single stub connected either in series
with the diode, or in shunt, and all diode parasitic parameters are in-
cluded. Procedures for computing the gain vs. frequency and the midband
noise figure are also presented. The computed performance of a number
of trial designs in the l-to-2-Gc and 2-to-4-Gc tuning ranges are shown,
and the results discussed.

15
 Besides the discussions described above, additional discussions were
presented in a paper presented at the 1961 PGMTT National Symposium in
a paper published in the IRE Transactions on Micro-
Washington, D.C.,24 in
a letter to the editors of the
wave Theory and Techniques,25 and in

Proceedings of the IRE.26

RPEFEMMM

1. G. L. Matthaei, Leo Young, and E. M. T. Jones, Design of Microwave Filters, Impedance-

Matching Networks and Coupling Structures, a book prepared on SRI Project 3527, Contract
DA 36-039 SC-87398, Stanford Research Institute, Menlo Park, California (January 1963).

2. W. J. Getainger and G. L. Matthaei, "Microwave Filters and Coupling Structures, " Quarterly
Progress Report 2, SRI Project 3527, Contract DA 36-039 SC-87-398, Stanford Research
Institute, Menlo Park, California (July 1961).
3. W. J. Getsinger, "Coupled Rectangular Bars Between Parallel Plates," IRE Trans. iGfl7-10,
pp. 65-72 (January 1962).

4. J. T. Bolljahn and G. L. Matthaei, "Microwave Filters and Coupling Structures," Quarterly

Progress Report 1, SRI Project 3527, Contract DA 36-039 SC-87398, Stanford Research
Institute, Menlo Park, California (April 1961).

5. J. T. Bolljahn and G. L. Matt haei, "A Study of the Phase and Filter Properties of Arrays
of Parallel Conductors Between Ground Planes," Proc. IRE S0, pp. 299-311 (March 1962).

6. Leo Young and G. L. Matthaei, "Microwave Filters and Coupling Structures," Quarterly
Progress Report 4, SRI Project 3527, Contract DA 36-039 SC-87398, Stanford Research
Institute, Menlo Park, California (January 1962).

7. G. L. Matthaei, "Interdigital Band-Pass Filters," a paper presented on May 23, 1962, at

the 1962 PGMTT National Syaposiua at Boulder, Colorado.

8. G. L. Matthaei, "Interdigital Band-Pass Filters," IRE Trans. PGVIT-10, pp. 479-491

(November 1962).
9. G. L. Matthaei, Leo Young, and P. S. Carter, Jr.,"Microwave Filters and Coupling Structures,"
Quarterly Progress Report 5, SRI Project 3527, Contract DA 36-039 SC-87398, Stanford Research
Institute, Menlo Park, California (May 1962).

10. W. J. Getninger, E. G. Cristal and G. L. Matthaei, "Microwave Filters and Coupling Struc-
tures, " Quarterly Progress Report 6, SRI Project 3527, Contract DA 36-039 SC-87398,
Stanford Research Institute, Menlo Park, California (August 1962).

11. Leo Young, "Stepped-Impedance Transformers and Filter Prototypes," IRE Trans. RVfl-10,
pp. 339-359 (September 1962).

12. Leo Young, "Prediction of Absorption Loss i-i Multi-layer Interference Filters, " J. Opt. Soc.
An., 52, pp. 753-761 (July 1962).

13. Leo Young, "Inhomogeneous QUarter-Wave Transformers, " The Microwave Journal, 5, pp. 84-89
(February 1962).
14. Leo Young, "The Practical Realization of Series-Capacitive Couplings for Microwave Filters,"
The Microwave Journal, 5, pp. 79-81 (December 1962).
15. Leo Young and B. M. Schiffman, "A Useful High-Pass Filter Design," to be published in
The Microwave Journal.

16
16. P. S. Carter, Jr., Leo Young, G. L. Matthaei and E. M. T. Jones, "Microwave Filters and
Coupling Structures," Ouarterly Progress Report 3, SRI Project 3527, Contract -
DA 36-039 SC-87398, Stanford Research Institute, Menlo Park, California (October 1961).
17. B. M. Schiffmmn, P. S. Carter, Jr. , and G. L. Matthaei, "Microwave Filters and Coupling
Structures," Qiarterly Progress Report 7, SRI Project 3527, Contract DA 36-039 SC-87398,
Stanford Research Institute, Menlo Park, California (October 1962).
18. Leo Young, G. L. Matthaei and E. M. T. Jones, "Microwave Bandatop Filters with Narrow
Stop Bands," a paper presented on May 23, 1962 at the 1962 PGM7T National Sysposium in
Boulder, Colorado.
19. Leo Young, G. L. Matthaei and E. M. T. Jones, "Microwave Bandstop Filters with Narrow
Stop Bands," IRE Trans. PGTMT-1O, pp. 416-427 (November 1962).
20. G. L. Matthaei, et al., "Design Criteria for Microwave Filters and Coupling Structures,"
Final Report, Chapter 28, SRI Project 2326, Contract DA 36-039 SC-74862, Stanford Research
Institute, Menlo Park, California (January 1961).
21. P. S. Carter, Jr.,"Magnetically-Tunable Microwave Filters Using Single-Crystal Yttrium-
Iron-Garret Resonators, " IRE Trans. /CAW-9, pp. 252-260 (May 1961).
22. Y. Sato and P. S. Carter, Jr., "A D)evice for Riapidly Aligning and Mounting Ferromagnetic
Single Crystals Along Any Desired Axis," IRE Trans. PGFHFT-I0, pp. 611-612 (November 1962).

23. G. L. Matthaei, et al., op. cit. Chapter 29.

24. G. L. Matthaei, "Design Theory of Up-Converters for Use as Electronically Tunable Filters,"
a paper presented or,May 15, 1961, at the 1961 PIM77 National Symposium in Washington, I).C.
25. G. L. Matthaei, "Design Theory of Up-Converters for Use as Electronicullly Tunable
Filters," IRE Trans. PCAfRT-9, pp. 425-435 (September 1961).
26. G. L. Matthaei, "An Electronically Tunable Up-Converter," Proc. IIRE 49, pp. 1703-1704
(November 1961).

17
 III A NOVEL MAGNETICALLY TUNABLE WAVEGUIDE FILTER,
AND A REFINED VERSION OF A PREVIOUS STRIP-LINE
MAGNETICALLY TUNABLE FILTER DESIGN

A. GENERAL

During the past quarter, two additional magnetically tunable filters

have been constructed which incorporate techniques discussed in previous
reports, along with some new techniques not previously treated.
1,2 Most
of the new techniques were tested in the waveguide filter discussed in
the following pages. The strip-line filter demonstrates a simple form
of consutruction which permits a magnet air gap which is quite small for
the size of strip line used.

B. AMAGNETICALLY TUNABLE WAVEGUIDE FILTER

USING TWO YIG RESONATORS

Figures III B-I(a), (b) show sectional views of a. magnetically

tunable waveguide filte-r which was constructed on this project dur-in-g
the past quarter, while Fig. III B-2 shows a photograph of the completed
filter. This, structure has several unusual features:

(1) It uses very small YIG resonators (0.025-inch-diameter)

in order to reduce coupling between magnetostatic mode's
within the YIG material and the fieldsý external to the
YIG material.

(2) It uses a novel impedance transforming structure which

greatly enhances the coupling between the YIG spheres
and: the waveguide structure, while at the same time the
spheres are in regions which are sufficiently spacious
so that the sphere's performance is not greatly disturbed
by the surrounding metal walls.

(3) Because of the miniaturized structure around the YIG

spheres, by machining away some unnecessary metal in the
structure it is possible to use a biasing magnet with
an air gap as small as 0.20 inch.

The waveguide structure shown is

made of aluminum and was fabricated
in three pieces which were afterwards fitted together with dowels and
screws. The structure mates with standard X-band waveguide (,which has

19
 Ho

.2500 2 .9
4E-'74

0.025-
INLL DIAMTR M ENSIONAOSINNCE

FIG. MATEIA ALUM)

ECIONUM VE FMGEIALYTNBEWVGI

3. FLTERINPTE
ANIOTPU TWO
MAEITH RESONATWAVOUR D

4. SESTWV0.05-ICH
IA YI SPERE ATCENER F ITER STRUCTURREXOIEPS

200
FIG. III B-2 PHOTOGRAPH OF THE FILTER IN FIGS. III B-1:(o) AND (b)

21
inside dimensions of 0.400 inch by 0.900 inch), and the width of the
structure shown is 0.900 inch throughout, except for the 0.080-inch-by-
0.020-inch coupling slot between the resonator spheres. The coupling
slot is oriented with its length parallel to the longitudinal axis of
the guide. This causes the slot to disturb the current in the guide
very little, and causes the circuit to have very high attenuation (in
excess of 60 db) when the spheres are off ferrimagnetic resonance.
However, when the signal frequency and applied biasing magnetic field
are of the proper values to excite ferrimagnetic resonance, a component
of RF magnetic dipole moment in the spheres is parallel to the length
of the slot1''2 and good coupling between the spheres is obtained, so as
to give two-resonator filter performance.

C. DESIGN OF THE MAGNETICALLY TUNABLE WAVEGUIDE FILTER

As previously mentioned, it was decided to use 0.025-inch diameter

spheres in this X-band filter in order to reduce coupling between pos-
sibie magnetostatic modes within the YIG spheres and the fields external
to the spheres. In some prior X-band waveguide experiments with larger
spheres, we have experienced quite large spurious responses due to
magnetostatic modes. Reducing the size of the spheres should reduce
the excitation of such magnetostatic modes, at least as far as their
excitation due to direct interaction with external fields is concerned.
However, magnetostatic modes can also be excited by direct coupling be-
tween the desired uniform precession mode and magnetostatic modes. This
type of coupling to magnetostatic modes can occur, for example, if the
sphere is not perfectly round. 3 Assuming that the spheres are of good
crystal structure with good surface polish and sphericity, making them
as small as is feasible compared to a wavelength should help consider-
ably iny reducing difficulties with spurious responses due to magnetostatic
modes.

By use of equations in Quarterly Progress Report 1, or in Sec. 17.09

of the book prepared on this contract, 2 we find that 0.025-inch-diameter
Y-IG spheres next to a short-circuiting end wall in standard X-band wave-
guide should have an external Q of about Qe - 11,600. Since the unloaded
Q of X-band YIG resonators in a closed-in structure will probably not be
over Q* - 2,000 (and very likely less than that), an external Q of
* 11,600 is much too large. Standard X-band waveguide has an inside

22
height of 0.400 inch, and by decreasing this height it is possible to
reduce the external Q in direct proportion to the guide height.1,2 How-
ever, the guide height must not be reduced too far or the metal top end
bottom walls will be very close to the spheres and cause their operation
to be disrupted as a result of currents induced in the adjacent metal
surfaces.1,2 From our past experience it appears that the guide height
should be at least 1.3 times the diameter of the spheres, and larger if
possible. In this case it was decided to make the guide 0.040-inch high
in the region of the spheres, which is 1.6 times 'the diameter of the
sph-eres. In this reduced-height guide the external Q will be about
[(0.040)/(0.400)] (11,600) = 1,160.

An external Q of 1,160, for a resonator whose unloaded Q might be

of roughly the same value, is still too large to permit a reasonably
low insertion loss in the filter. [See Eq. (III B-4) of Ref. 1, or
Secs. 11.06 and 11.07 of Ref. 2 for equations for calculating the mid-
band attenuation due to dissipation.] Now it can be shown that if a
YIG resonator has an external Q of Q when it is connected to a termi-
nating guide of impedance Z 0 , if the impedance of the terminating guide
is changed to Z;, the external Q should become

Zo
QQ:
= Q. yo
z~o (III C-i)

Thus, we could lower the external Q further (i.e., increase the coupling
between the resonator and the external load) if we could lower the ef-
fective impedance of the terminating guide. Since the guide impedance
is directly proportional to the guide height, and since we do not wish
to use a guide height less than 0.040 inch in the vicinity of the sphere,
our problem resolves to this question: With a YIG sphere in a 0.040-inch-
high guide, can we obtain an effective terminating impedance like that of
a guide whose height is considerably less than 0.040 inch? As we shall
see, at least as an approximation, the answer is yes.

In order to keep the dissipation loss in the filter down, it would

be desirable to decrease the Q, - 1,160 by at least a factor of five,
which would give Q: * 232. Since the guide impedance is proportional
to its height, by Eq. (III C-i) we c-onclude that the sphere must see a
terminating impedance like that of a guide having a height of

23
 V - b (III C-2)
Q.
or
232
V - (0.•040) - 0.008 inch
1,160

Our approach will be to first obtain a step-transformer design which

will match between standard-height X-band guide (which has b 0.400 inch)
-
and reduced-height guide having b - 0.008 inch. Then, one end of the
t-ransformer will be modified so as to present an impedance like that of
0.008-inch-high guide, at the center of a short section of 0. 040-inch-
high, guide.

The required step transformer must have a transformation ratio of

0.400/0.008 - 50. In order to cover the 8.2-to-12.4-Gc band with wave-
guide which is 0.900 inch wide, the transformer must have a fractional
bandwidth wx - 0.725, on a reciprocal guide-wavelength basis [where
wAm - X g 2 )/(X•,
2(X g + k2)' and K.j and Xg 2 are the guide wavelength
values at the operating band edges]. Tabulated impedance values for
many quarter-wave-transformer designs are given in Refs. 4 and 5. A
four-section, 50:1-impedance-ratio, Tchebyscheff transformer design with
w= 0.80 was selected, and from the tabulated normalized impedances
the waveguide dimensions in Fig. III C-i were obtained. This transformer
45
has a theoretical maximum VSWR of 1.17 in its operating band. '

Zx,

XX

b-o.oo
0.0 11o.o012 b2.O.032 go
be.o10o b,-0.260 b=o.4oo INCH
ALL STEPS ;F LONG EXCEPT FOR FRINGING CAPACITANCE CORRECTIONS

(for guide width of a - 0.900 inch, and f- 10.1 Gc, L-• 0.383 inch)

FIG. ISEC-1 A FOUR-SECTION, 50:1, TCHEBYSCHEFF, WAVEGUIDE

IMPEDANCE TRANSFORMER

24
 The next step in the design was to modify the first section (having
61 0.012 inch) of the transformer so as to replace it with a short
section of line having bA ' 0.040 inch, plus another section of line
having a height and length such that the performance of the new line
sections will be similar to that of the original uniform bI -0.012-inch-
high line section.

10-0.00 51-0.012

(a)
ORIGINAL SECTION ZI

•MMODIFIED
SECTION

r .. . • .-//////••////• b2 Z2

(b)
MODIFIED SECTION Zt-s

FIG. III C-2 ORIGINAL AND MODIFIED SECTION Z1 OF TRANSFORMER,

MODIFIED SO AS TO PROVIDE ADEQUATE SPACE FOR
YIG SPHERE

Figure III C-2(a) shows the original first section of the trans-
former. At midband the length of this Z 1, section is X.0/4, and the
impedance Z. normalized with respect to Z0 is

(z\ 2
ZSN
(III C-3)

We now wish to replace this line section of impedance Z, and length 41

25
by two line sections, one being of impedance ZA (corresponding to a
,height of bA - 0.040 inch) and of length tA, and the other being of
impedance Z. and of length t., as is indicated by the solid line in
Fig. III C-2(b). The length 'tA was chosen to be 0.050 inch, in order
to provide sufficient room in the longitudinal direction so that the
metal walls will not unduly disturb the resonance of the sphere. The
height bB and length tB were chosen so that the normalized impedance
Z./Z0 in Fig. III C-2(b) at the midband frequency will be the same as
Z./Z0 for the original structure in Fig. III C-2(a).

The values of b. and t were determined by trial and error pro-

cedures using'a Smith chart, and making use of the fact the guide
impedance Z, is directly proportional to guide height bk. By
Eq. (I1I C-3), for Fig. III C-2(a)

Z20 b/ 2.25

Next, Z in Fig. III C.-2(b) was computed in normalized form. At

midband, X 0 . 1.530 inch-es, so that tAl/kso0 0.050/1.530 - 0.0326.
Normalizing with respect to impedance ZA, ZO/ZA - bo/bA - 0.200. Using
a Smith chart, moving 1A/Xgo towards the generator from Zo/ZA gives
Z Y/ZA - 0.21 + jO.205. Several values of b. were tried until the value
bB . 0.006 inch was arrived at. For that value, Z /Z8 .(Z Y/ZA)(bA/bB) -
1.4 + j].36. After moving towards the generator a distance tB/k 0 0.605,
a purely real value Z./Z, . 3.05 was obtained. Renormalizing with re-
spect to Z. gives Z/Z 0 - (Z/ZB)(b8 /bO) - 2.29, which was deemed to be
satisfactorily close to the desired value of Z,/ZO0 2.25 computed above.
In this manner the dimensions b. . 0.006 inch and 'C . 0.096 inch were
arrived at.

Within the accuracy of the above calculations, the two lines ZA and
Zs in Fig. III C-2(b) give exactly the same midband performance as does
the single line of impedance Z, in Fig. III C-2(a). Note that the
extra-high impedance of the ZA line section is compensated for by an
extra-low impedance in the Z5 line section. Though the circuit shown
at (b) can be expected to deviate in performance from that at (a) as
the frequency deviates from the midband. value, the circuit at (b) should
be reasonably broadband in its performance. This point is verified by

26
 0Gc 12.5 Gc

1.4

1.2 ___

0.5 0.6 0.7 0.6 0.9 1.0 1.1 1.2 L3 1.4 1.5

FIG. III C-3 COMPUTED VSWR OF THE 50:1 TRANSFORMER IN FIG. III C-1 WITH SECTION Z,
REPLACED BY SECTIONS ZA AND Z8 IN FIG. III C-2(b)

the computed response in Fig. III C-3, which shows the computed VSWR of
the 50:1 impedance transformer in Fig. III C-1 with its Z, line section
replaced by line sections ZA and Z shown in Fig. III C-2(b). Note that
the maximum VSWR in the band of interest is about 1.55 which, though
larger than the 1.1 maximum VSWR value for the original transformer, is
reasonably good. However, this design is obviously not optimum, and
further s:tudy to provide techni-ques for optimum des-ign of imped'ance
transforming and coupling circuits of this type is contemplated.

In the actual filter circuit the Z0 line of height b 0 . 0.008 inch

in Fig. III C-2(b) was replaced by the YIG sphere, as indicated by dashed
lines in the figure. Also, a short-circuit wall was placed a distance
tA " 0.050 inch to the left of the center of the sphere, as is also
indicated by dashed lines in the figure. This distance was again gov-
erned by the need to keep the metal end wall a reasonable distance from
the sphere in order to prevent induced currents in the wall from dis-
turbing the sphere. The additional length of waveguide, will introduce
an additional, small self-reactance in the coupling to the sphere, but

27
this reactance should be small enough to be ignored without serious con-
sequences. After corrections for fringing capacitances (as discussed in
Refs. 4 and 5), the final dimensions of the input and output impedance
transforming sections were as shown in Figs. III B-l(a), (b).

The dividing wall between the two spheres was made to be 0.020 inch
thick, which resulted in a separation between the centers of the spheres
of approximately 2.4 diameters. From past experience, this separation
is believed to be adequate so that the spheres will not greatly distort
each other's biasing fl-field. Using the results of previous designs as
a guide, trial values of 0.020 inch by 0.080 inch were estimated for the
dimensions of the coupling slot between the spheres. It was expected
that this slot would later be widened a little until the desired degree
of coupling between spheres was obtained. However, initial tests indi-
cated that the degree of coupling was about right with the dimensions as
they were.

In working out the mechanical design of the filter, many of the con-
tact surfaces of the joints in the device were partially recessed [not
shown in Figs. III B-I(a), (b)] so as to leave a pressure lip along the
edge of joints where good contact was important. In addition, at the
joints in the short-circuit walls beside the YIG spheres a little silver
paint was used in assembly to further insure a good contact.

D. MEASURED PERFORMANCE OF THE MAGNETICALLY

TUNABLE WAVEGUIDE FILTER

The measured performance of the filter bore out quite well the
validity of the design principles described above. Table III D-1 shows
the measured values of midband attenuation, 3-db bandwidth, 30-db band-
width, midband VSWR, and the required biasing magnetic field strength
for different pass-band frequencies. Assuming that the response of this
filter at 10 Gc corresponds to that of a two-resonator filter designed
from an equal-element prototype, 6 for a midband attenuation of 2.4 db
and a 30-db bandwidth of 276 Mc (as is indicated in Table III D-1 for
f0 - 10 Gc), then the computed unloaded Q of the resonators is Q.= 1,040,
while the external Q of the resonators is Q. - 289. This value of Q, is
seen to agree reasonably well with the design value which was Q, = 232,
especially since it is known that the presence of the adjacent coupling
slot will cause the Q, value to be raised typically by about 20 percent

28
or somewhat more. 1,2 The estimated 1,040 value of unloaded Q for the
spheres was lower than was hoped for. Because of a shortage of time,
the linewidth of these spheres was not measured before inserting them
in the filter, and it is probable that spheres with better unloaded Q's
could be obtained.

Table III D-1

MEASURED PERFORMANCE OF THE MAGNETICALLY
TUNABLE WAVEGUIDE FILTER

fO Le Af3db Af30db VSWIR N0

(Gc) (db) (Mc) (Mc) -t 0 k-oerted)
8.0 2.1 -- 2.43
1.5 2.1 45.7 249 1.22 2.86
9.0 2.2 ... .. .. 2.90
9.5 2.1 3.00
10.0 2.4 49.8 276 1.06 3.20
10.5 2. 4 .. .. .. 3.25
11.0 2.2 .. .. .. 3.61
11.5 2.4 .. .. .. 3.68
12.0 2.9 4.22
112.5 12.6 1 47.1 1 25 1 1.06 4.30

f O Tuning frequency
Le Minimum attenuation at tuning frequency

&f3db=Bondwidth. for 3 db attenuation with- respect to Lc

W30db Bandwidth for 30 db attenuation with respect to Lc

Figure III D-1 shows the measured attenuation characteristics of

the filter vs. biasing magnetic field. Here again the results are very
much in line with what was hoped for. Note that there is only one stop-
band spurious response, and its minimum attenuation is 38 db, a high
value for a two-resonator magnetically tunable filter. Also note that
,the off-resonance attenuation is everywhere greater than the measuring
range of the test equipment which in this case was from around 52 to
,59 db. This very high off-resonance: attenuation results from the very
low impedance in the vicinity of the coupling slot between the two wave-
guides, and also from the small size of the coupling slot (which was
made possible by the use of small YIG spheres).

One unexpected problem that became apparent after testing this

filter was that though the spurious response activity in the stop-band
of this filter was quite low, so that the stop-band performance was
very good for a YIG filter with only two resonators, there were various
small spurious responses which appeared in the p.ass-band of the filter.

'29
 TO 70 I ' I I I: 7 1 -, i•
7o I w I ' I I
60 OF601
.I,.MI.T O.DETECTIOLIMIT 6- OF DETECTION

150- 50
z 40 - ,o- 5s - _40o- 1,0- .Oec
2 S30 ~30,- X
z
W WI 2 0
1-20
I-
- t--
10 2.:db - 0- f 2.4 db
0 1000 2000
II3000 4000 5000
0 1o
0 1000 2000 3000 4000 5000
BIASING FIELD STRENGTH-osrstods BIASING FIELD STRENGTH -orsteds

70 1 I I I I I I I
60 LIMIT OF DETECTION

50-

5400
130K
S I
0-20 F

10 2.6 db
o BIASIN I
0 1000 2000 3000 400 5000
BIASING FIELD STRENGTH-oersteds

FIG. III D-1 ATTENUATION vs. BIASING FIELD STRENGTH FOR

"MAGNETICALLY TUNABLE WAVEGUIDE FILTER
These curves were obtained using a sweeping magnet
power supply and a recorder

30
 These can be seen from the responses shown in Fig. III D-2 which were
taken using, a sweeping magnet power supply, along with a recorder. These
spurious responses are quite different in character from those that we
have observed in our strip-line S-band filters. 1,2 In our S-band filters
the stop-band spurious responses were generally more prominent than those
in this present filter, and there were no spurious responses observable

30

20

8.0 Ge 9.0 Ge 10.0 GC

I0

I 00I --- II
z
0

30

IO-
2O

11.0 Ge 12.0 Gc 12.5 Ge

20

BIASING H-FIELD

FIG. III D-2 PASS-BAND ATTENUATION vs. SWEPT BIASING FIELD STRENGTH
FOR THE MAGNETICALLY TUNABLE WAVEGUIDE FILTER
Those curves wore obtained using a sweeping magnet power supply
and a recorder

31
•m •' . -_-_••• .=. .=••: m . . . M .=: -__. .. .

in the pass-band except for typically around 4.5 Gc where one of the
stop-band spurious responses would tend to merge with the pass-band
response. In the case of the data in Figs. III D-1 and III D-2 the
stop-band spurious response never merges with the pass-band, and though
there is movement of the pass-band spurious responses, as they move out
of the pass-band. they never grow into stop-band spurious responses, or
at least not into stop-band spurious responses of any magnitude.

Though it will take further investigation to prove this point, it

appears likely that these spurious responses which are occurring only in
and around the pass-band of the spheres are responses due to direct
coupling between the uniform precessional mode and the magnetostatic
modes as a result of imperfections in the spheres (as contrasted with
spurious responses excited by non-uniform external RF fields). The most
likely imperfection t-o be giving this trouble is that the spheres may
not be sufficiently spherical in shape.3

E. A MORE REFINED VERSION OF A PREVIOUS STRIP-LINE

MAGNETICALLY TUNABLE FILTER DESIGN

One type of filter discussed previously1,2 is a strip-line, two-

resonator filter for the 2-to-4-Gc band, with the resonator spheres
lying in a plane perpendicular to the biasing H-field. As a two-resonator
filter this device had more stop-band spurious response activity than
some other strip-line configurations we have investigated, 2,7 but it had
the advantages that it was very easy to convert into a filter with three
or more resonators, and it required an unusually small magnet air gap..
When this filter was tested using three resonators, the spurious response
activity which was apparent was greatly reduced.1,2

A more refined two-resonator version of the filter described above

was constructed for use as one of the experimental models to be sent to
the Signal Corps. This filter is shown in Fig. III E-1 with its upper
half raised up. The significant electrical dimensions of this filter
are the same as those given in Fig. III F-i of Ref. 1 or in Fig. 17.08-3
of Ref. 2, but in this case the filter is fabricated in quite a differ-
ent manner, and it is fabricated so as to minimize the magnet air gap
required. In this case the filter is designed to split in half along
a horizontal plane, the upper and lower housings being machined out of
solid blocks of brass. After the housings were machined from brass, the

32
 3:
0 I 2 3

FIG. III E-1 A.STRIP-LINE TWO.RESONATOR FILTER WITH YIG RESONATORS

The upper half is raised so that the YIG spheres and the strip lines
may be seen

strip 1lies were soldered in place in one half of the structure, while

the YIG spheres were mounted in the other half. Dielectric was used
along the part of the strip lines in the regions next to the connectors
in order to prevent possible flexing of the strip lines by forces in the
connectors. Using this construction the magnet air gap required is only
0.260 inch. The housing structures for this filter could probably be

made quite cheaply by die-casting methods. Figure III E-2 shows a close-
up view of this filter mounted in an electromagnet purchased from a local
manufacturer. Of course, by using a combination electromagnet and per-

manent magnet the size of the magnet could be reduced considerably.

33
FIG.. III E-2 CLOSE-UP VIEW OF FILTER IN FIG. III E-1 MOUNTED IN AN ELECTROMAGNET

34
 The measured performance of this filter is much like that of the
earlier experimental model 1 ' 2 so the performance curves will not be re-
peated here. However, Table III E-1 shows the measured values of midband
attenuation, 3-db bandwidth, 30-db bandwidth, midband VSWR, and biasing
H-field required for resonance.

Table III E-1

MEASURED PERFORMANCE OF THE TWO-RESONATOR
STRIPLINE FILTER IN FIGS. III E-I ANDIII E-2

f0 L, N3db Af 3 0db VSWR N0

(Gc) (db) (Mc) (Mie) at fo (k-oersted)
2.0 2.3 18.71 93.0 1.02 0.66
2.5 1.4 0.84
3.0 1.1 18.88 99.0 1.3, 1.00
3.5 1.2 __
- - 1.20
4.0 1.2 18.24 100.9 1.6 l.4

REFERENCES

1. B. M. Schiffman, P. S. Carter, Jr., and G. L. Matthaei, "Microwave Filters and Coupling

Structures," Quarterly Progress Report 7, SRI Project 3527, Contract DA 36-039 SC-87398,
Stanford Research Institute, Menlo Park, California (October 1962).

2. G. L. Matthaei, Leo Young, and E. M. T. Jones, "Design of Microwave Filters Impedance

Matching Networks and Coupling Structures," a book prepared on SRI Project 3527, Contract
DA 36-039 SC-87398, Stanford Research Institute, Menlo Park, California, Chapter 17
(January 1963).
3. P. C. Fletcher and I. H. Solt, Jr., "Coupling of the Magnetostatic Modes," J. Appl. Phys.,
Vol. 30S, pp. 181S-182S (April 1959).

4. Leo Young, and G. L. Matthaei, "Microwave Filters and Coupling Structures," Quarterly
Progress Report 4, Sec. IV, SRI Project 3527, Contract DA 36-039 SC-87398, Stanford Research
Institute, Menlo Park, California (January 1962).
5. G. L. Matthaei, Leo Young, and E. M. T. Jones, Op. cit., Chapter 6.

6. G. L. Matthaei, Leo Young, and E. M. T. Jones, Ibid., Sec. 11.07.

7. P. &. Carter, Jr., Leo Young, G. L. Matthaei, and E. M. T. Jones, "Microwave Filters and
Coupling Structures," Quarterly Progress Report 3, SRI Project 3527, Contract DA 36-039
SC-87398, Stanford Research. Institute, Menlo Park, California (October 1961).

35
 IV RAM}-STOP FILTERS

A. GENERAL

An exact method for the design of band-stop filters has been given
in some detail in Quarterly Progress Report 7.1 The main feature de-
scribed there is a table of formulas (derived with the aid of aprocedure
given by Ozaki and Ishii 2 ) for directly transforming a low-pass lumped-
element prototype filter into a band-stop transmission-line filter that
consists of alternating sections of quarter-wavelength connecting lines
and stubs. The stubs are in shunt with the main line and are open-circuited
at their outer ends, yielding a configuration that is here called the basic
band-stop filter. Although this basic form is conceptually simple and
easy to design it is not suitable for filters requiring very narrow stop-
bands. The reason for this limitation is the difficulty of constructing
stub lines with the high value of characteristic impedance required in
such a filter. To overcome this difficulty, and also to allow greater
freedom to the designer in general, other configurations of the band-stop,
filter are described here, together with design equations for converting
the basic configuration to one of the newer types. Since the method of
designing the basic filter (and other associated forms of band-stop
filters) has been fully covered in Quarterly Progress Report 7,2 it will
not be reviewed here.

Two of these newer forms, both of which employ parallel-coupled lines,

are completely consonant with the exact design procedure. A filter sec-
tion of the first type, which is described in Sec. IV-B, has a stub reso-
2 3
nator that is open-circuited on one end and short-circuited on the other, '
and is separate from the main line. This type is particularly useful for
filters with very narrow stop-bands because the stop-band width is es-
sentially a function of the degree of coupling between thestub andmain line.

The second type of filter employing coupled lines is suitable for

wide bandwidths. This type of filter, described in Sec. IV-C, is similar
to the first except that the stub is open-circuited on one end and direct.ly
connected to the main line at the other end. Because of its form it is
here called a spur-line type of filter. In both types there is electro-
magnetic coupling between the stub and the main line along the full length

37
of the stub. Exact design equations and test result-s on filters designed
by both the above methods are given herein.

In Sec. IV-D, test results are given on a filter configuration that

uses short-circuited stubs capacitively coupled to the main line. 4' 5 In
the latter case the design formulasI (which were described in Quarterly
Progress Report 7), although not exact, are a closer approximation to
the exact design than earlier methods 4 from which they evolved. This
form is particularly well suited to narrow stop-band filters.

In each of the three described cases it is required that a basic

band-stop filter be designed as a first step. Then from this basic
circuit (or its dual) one of the three alternative forms can be designed
with the aid of the given equations.

B. PARALLEL-COUPLED- RESONATOR FILTER

A single section of the parallel-coupled-resonator type filter

suitable for narrow stop-bands is shown in Fig. IV B-i. Also shown

Y12 " Z
12

1 b - I 2

2 r

WYUCASE~ 16-
, . . Zoo
,,2
2
2
.2•

.ith CG arbitrary (.ee text)

f 2 Y 2(1 d

FIG. IV B-1 PARALLEL-COUPLED TRANSMISSION-LINE SECTION

AND EQUIVALENT SECTION OF BASIC FILTER

38
there, are the corresponding portion of the basic filter and design
equations that enable the basic filter to be converted, section by sec-
tion, into the parallel-coupled type. The coupled lines are completely
specified by their even- and odd-mode impedances Z%, Zb.., Z%, and Z0b
67
(or their odd- and even-mode admittances). - The superscripts a and b
pertain to lines a and b in Fig. IV B-i.

Before the formulas can be applied it is necessary to partition

the basic filter, shown in Fig. IV B-2, so that each section is complete
according to Fig. IV B-I. Since there are more stubs (one more) than
connecting lines, one quarter-wavelength section of connecting line must
be inserted between the filter and one of the terminations. The charac.-
teristic impedance of the inserted section must of course equal the
terminating impedance; although the filter amplitude response will be
unaltered by this change, some time delay will be added to the signal.

The conversion formulas yield values for the distributed capaci-

tances of the coupled lines as illustrated in Fig. IV B-3, These
distributed capacitances C., C,, and C,., pertaining to lines a and b
shown in the form of a general 77-network, are directly related to the
67
even- and odd-mode admittances (or impedances) of the coupled lines. '
The formulas give distributed capacitances directly because these
quantities are most useful in designing coupled-rectangular-bar strip
lines. The coupled lines: may be symmetrical (C. - Cd)or asymmetrical
(C. 9 Cb).

If it is desired that the coupled lines be symmetrical, the lower

set of equations of Fig. IV B-I is used. If the coupled lines are made
unsymmetrical, the value of:Cb is arbitrarily chosen (subject to the
condition that positive values are obtained for C. and Cd)and the upper
set of equations is used.

The relationship of the capacitances of the 77-network of Fig. IV B-3

to the component capacitances (i.e., parallel-plate and fringing capaci-
tances) of parallel-coupled rectangular bars is easily seen in Fig. IVB-4.
Since the capacitances C. and C, are not necessarily equal, the two
coupled lines of each section need not be identical in cross section.
However, as explained below, by restricting the heights of the two rec-
tangular bars to be equal and further limiting the strip-line proportions,
the coupled lines of both the symmetrical and asymmetrical type can be
easily designed with the aid of Getsinger's graphsof fringing capacitance.

39
 2
/,.-I,II n Z,

(a) ALL STUBS AND CONNECTING LINES ARE \0/4 LONG

•I
-AO
-v/o
v I IIi
I
L
II
I I I
i

LINE V TLINE b

C.Cb

A-SU7-273

FIG. IV B-3FAN UNSYMMETRICAL PAIR OF PARALLEL-COUPLED

LINES C C A Cb ARE LINE CAPACITANCES
PER UNN'LENGTH

4,
 ELECTRIC WALL FOR OOD MODE
MAGNETIC WALL FOR EVEN MODE

L a cb Cb I L N

cap
T, cf 0 ,f Cfo c c.

A-3527-274

FIG. IV B-4 CROSS SECTION OF UNSYMMETRICAL, RECTANGULAR-BAR

PARALLEL-COUPLED LINES

8
Getsinger has derived equations for the fringing capacitances Ce,
C;0, and C;, of rectangular bars shown in Fig. IV B-4, and has prepared
convenient charts which relate C;e/E, C. 0/E, and Cý1e to rectangular-bar
strip-line dimensions. Here E is the dielectric constant of the medium
of propagation, so that the above ratios are dimensionless and of mod-
erate size. Getsinger gives equations for the design of symmetrical,
parallel-coupled, rectangular strip-lines, and here we adapt his equations
6
to fit the unsymmetrical as well as the symmetrical case.

Note that the shape of the strip lines in Fig. IV B-4 is specified
in terms of the dimensions t, b, s, w., and w.. To design a pair of lines
such as those in Fig. IV B-4, to have odd- and even-mode admittances-or
impedances as determined implicitly by the calculated values of C IE,
C., E, and Cb/e, first select a convenient value for t/b. Then, noting that

SiC C
AC
(IV B-l)

8
use Getsinger's chart of AC/e and C;. vs. s/b to determine s/b, and also
C;./6. Using t/b and Getsinger's chart of C. vs. t/b, determine C /E,
and then compute

b 2 L2

'41
 b *
2 - b { 2ic) : a3
(IV'B3

When the ground plane spacing b is specified, the required bar widths,
w and w,, are determined. This procedure also works for the thin-strip
case where t/b - 0. If either w./b or w,/b is less. than 0.35 (1 - t/b),
the width of the bar should be corrected using the approximate formulas

, 0.07 1 - t + b
0.07- ( i b) , (IV B-4)
b 1.20

providing that 0.1 < (w'/b)/(1 - t/b) < 0.35. In Eq. (IV B-4), w is the
uncorrected bar width and w' is the corrected width. The need for this
correction arises because of the interaction of the fringing fields at
opposite sides of a bar, which will occur when abar is relatively narrow.

The strip-line filter shown in Fig. IV B-5 was designed by the

methods outlined here. The stop-band center frequencyjis 1.6 Gc. The
prototype filter has two elements and a maximally flat response. The
element values are go - g 3 = 1, and g,= g 2 = 1.414. The characteristic
impedance values of the basic two-stub filter designed with the aid of
the formulas in Quarterly Progress Report 7 for a 5 percent width of the
stop-band (between 3-db points) and S0-ohm terminations are Z, = 949.9 ohms
and Z2 ' 899.9 ohms for the shunt-stub characteristic impedances, and
Z12 ' 52.8 for the connecting-line impedance. A quarter-wavelength sec-
tion of S0-ohm impedance line was inserted between Stub 2 and the 50-ohm
termination, thus yielding two complete L-sections ready to convert to
two sections of parallel-coupled lines. The symmetrical type of coupled-
line construction was chosen for the filter. The values of distributed
capacitance (normalized to the permittivity of the medium) were then
calculated with the aid of the formulas of Fig. IV B-i and found to be:
(1) for Section 1 consisting of Stub 1 and the connecting line, C 4 /E =
Cb/e = 5.80 and C.,/1 = 1.73, and (2) for Section 2 consisting of Stub 2
and the added 50-ohm section, C./E - Cb/E = 6.12 and C.=/E - 1.82. With
the aid of Getsinger's graphs the strip-line dimensions shown in the
sketch Fig. IV B-6 were obtained.

"42
N3

P-352-6

FIG. IV &-5 A PARALLEL-COUPLED-LINE, NARROW-BAND FILTER

WITH COVER PLATE REMOVED

43
 COAX-TO-STRIPLINE ADAPTER AIR SUPPORT AND GROUND CONNECTION
AND GROUND PLANE SPACER DIELECTRIC FOR PARALLEL STUB.

UG_ I187/U T10

colx / 1.2z* I.oo-~

N 0.156 o0156

(M50 OHMDI) to !2 035901

CENTER CONDUCTOR AND PARALLEL STUBS -- 0.250 THICK

GROUND PLANE SPACING - 0.500

FIG. IV B-6 SKETCH OF A PARALLEL-COUPLED-LINE, NARROW-BAND FILTER

WITH COVER PLATE REMOVED, SHOWING DIMENSIONS

The parallel-coupled resonator and its grounding wedge were milled

from a solid piece of aluminum. Each resonator length is slightly less
than K0 /4, as determined experimentally, to allow for fringing capacity.
The measured and computed responses of the filter are shown in Fig. IVB-7.
The stop-band center frequency was found, after the final adjustments
of the lengths of the resonators, to be Jo 1.608 Gc, and the curves
of Fig. IV B-7 are so plotted. The measured attenuation loss and VSWR
are aeen to be in excellent agreement with the computed values over a

very wide frequency range, even including the second stop-band at 3f0.

44
 40

*MEASURED RESPONSE
30 - COMPUTED RESPONSE
f 0. j~01.64k
I 20
J

10

0
0 2 3 4
WO

(a)

STOP SAND

• : o
1.0 1 o 0
0.4 0.6 0.8: 1.0 1.2 IA Is 1. 2.0

'At.

50 I I I

404

30 - MEASURED RESPONSE
0COMPUTED RESPONSE

VA 1.6 Ge
I 20-

0
0.04 0.94 0.96 1.0 1.02 104 lOGS I.0o 10

WO

(b) K.M,.,

FIG. IV B-7 MEASURED AND COMPUTED RESPONSE OF FILTER IN FIG. IV 8-5

45
C. SPUR-LINE TYPE OF FILTER

A second type of band-stop filter section employing coupled lines

is shown, together with its design formulas, in Fig. IV C-I. Because
of the direct connection (and the resulting strong coupling) of the stub
to the line, type of filter is similar to the basic type.
t:h~is In the
limit, as the coupled lines are separated until there is very little

Zoo , .zo* 1
Yg

] -y22 yo o

r 2V '4-C7-

MA-55Z7-686

FIG. IV C-1 SPUR-LINE STUB RESONATOR AND EQUIVALENT SECTION OF BASIC FILTER

distributed coupling, the structure becomes almost identical to the basic

shunt-stub filter section. The spur-line type of construction is thus
suitable for filters with wide stop-bands. When the coupled lines are
physically symmetrical, the energy stored in the resonant structure is
determined mainly by the electromagnetic fields of the odd mode of propa-
gation. This energy can be reduced and the bandwidth narrowed by reducing
Z..-that is, by bringing the coupled lines close together-but this
process has a practical limit. Furthermore, copper losses would tend
to increase and thereby reduce the maximum attainable attenuation in the
stop-band.

'46
 A two-resonator spur-line-type band-stop filter was designed with
the aid of the formulas of Fig. IV C-1 for 60-percent stop-band bandwidth
and 1.6-Gc stop-band center frequency. Here again, as for the narrow
band filter described in Sec. IV-B, symmetrical coup.led lines we-re used.
The: prototype circuit was chosen to have maximally flat response and its
element values are, as before, go = g 3 - 1, g1 - g 2 - 1.414.

The basic filter elcment values were found to be Z1 . 119.3 ohms

and Z2 . 69.3 ohms for the shunt stubs, and Z12 ' 86.05 ohms for the
:connecting line. A quarter-wavelength section of 50-ohm line was then
connected to one end of the filter so that it could be properly parti-
tioned. The two possible circuits thus obtained and their methods of
partitioning are shown in Fig. IV C-2(a) and (b). Trial paper designs
based on those circuits yielded coupled rectangular bar lines of such
proportions that it was feared that the discontinuities at the junctions
of the two sections and main line would be large enough to badly degrade
the performance. The circuit of. Fig. IV C-2(c) was found superior in
this respect. This circuit, which was used in the design of the filter,
was derived from the basic circuit as follows: First the dual of the
basic circuit was calculated by the equation

z20
ZD - Z (IV C-l)

where • ZAZ• (50)2, Z is the characteristic impedance of a stub or

connecting line section and ZD is the. impedance of the dual of that sec-
tion of line. (The dual of a shunt s~tub is a series short-circuited stub
and the dual of a connecting line is again a connecting line.) Then a
quarter-wavelength section of 50-ohm line was added to each end of the
filter, and Kuroda's. transformationI was used to move the two stubs out-
ward, each to its end of the network, and at the same time transform the
series stubs back to shunt stubs with the circuit of Fig. III C-2(c) as
a result. This circuit was then partitioned so that it could be convertled
to a filter structure with two spur-line sections separated by a length
of line of impedance Z, 3 - 29.1 ohms. The finished strip-line filter is
shown in Fig. IV C-3. [Note that the two spurs point toward each other,
and the spur-to-line junctions are separated by 3,\ 0 /4. An alternative
realization, in which the spurs point, toward each other and their con-
necting points are separated X0 /4, could be obtained from a series-stub

47
type filter in which the stubs are separated 3k 0 /4. This latter circuit
can be derived by applying Kuroda's transformation to the separate parts
of the original shunt-stub basic circuit.) A sketch of the spur-line type
filter with essential dimensions is given in Fig. IV C-4. The coupled
lines were designed by the method described in Sec.. IV-B.

Hý-SECTION I --+-SECTION 2-4

ZOl
ZA
50O
Z
1
2 .66.I
.Z
119.3 Z269'.3

(a)

SECTION I-+-SECTION 2-4

Z,z 0Z SO-5
8
6e.
12 -
ZZ
Z, - 19.3 /Z 2 .69.3

SECTION I -4SECTION 2-4+SECTION 3-4

(b)
Z12- 70.9 Z2 3 .-a2.1 Z3 4 'S66'

Z, - 169.3 Z4 * 119. 3

AA-35Z?-"l

FIG. IV C-2 THREE POSSIBLE INTERMEDIATE STEPS

IN CONVERTING A BASIC FILTER TO
A SPUR-LINE FILTER

'48
 00

P-3527-678

FIG. IV C.3 A SPUR-LINE BAND-STOP FILTER WITH COVER PLATE REMOVED

9
COAX-TO-STRIP-LINE ARDIELECTRIC
TRNIT IONAR
MODFIED [
UG 17UI
CONNECTOR 1.4 1.oo~
.845 0.209

00 '.250 0 r 10011, 029 .

GRUD 0LNESACN:0.0

FHIG.NISSCOF SKTCH-LN COFNDSUCTR-LNS

AITPFLTRWT OE LT

REMOVED, SHOWING DIMENSIONS

49
 The lengths of the spurs were individually adjusted to resonate
at stop-band center as follows: The gap between the end of one spur
and the center section of the main line was bridged with adhesive alu-
minum foil, thereby making that spur-line section non-resonant. The
frequency of maximum attenuation of the remaining resonant spur line
was then measured, and the unbridged spur was reduced in length about
half the amount calculated to bring it to resonance at fo when considered
to be a simple open-circuited shunt stub. (Half. the calculated reduc-
tion in length was used as a safety precaution because the capacitance
of the spur end to main line is reduced at the same time, and this latter
value is not included in the calculation.) A few attempts brought the
spur to resonance at the d'esired frequency and the same process was
repeated for the other spur.

The measured values of attenuation ltoss and VSWR are shown to agree
fairly well with the computed v-slues, in Fig. IV C-5. The anomalous
departure of measured values from computed values that does occur, mainly
near the upper edge of the first stop-band and the lower edge of the
5.0
0
40

S3.0 0

2.0 %
o 0

°... "00 00

4C ,75
-L 5 db i

LA. 75|,O o MEASUJRED RESPONSE

30- A COMPUTRED RESPONSE

• .6Gseo

20
2

gO 0 0

0 o, 0 0 0

06 0.5 1.0 1.8 2.0 2.s 3.0 3.W 4.0

FIG. IV C-5 MEASURED AND COMPUTED VSWR AND ATTENUATION

OF FILTER OF FIG. IV C-3

s5
second stop-band, may be due to the remaining discontinuities between
sections, which could not be eliminated entirely.

D. CAPACITIVELY-COUPLED-STUB FILTER

A method of designing an easily realizable narrow-band filter by

modifying the exact design procedure was fully described in Quarterly
Progress Report 7.1 Although this design method is not exact, it yields
a response curve that follows the desired response over a broad band.
The modification consists of replacing each open-circuited high-impedance
shunt stub with a short-circuited capacitively coupled stub of medium
impedance-i.e., the same order of magnitude as the connecting-line im-
pedance, and of length slightly less fthan X0 /4. The stubs are thus
easily realized in coaxial or strip line, whereas by-the exact method
they would (in the case of very-narrow-band filters) be difficult or
impossible to realize. In Quarterly Progress Report 7, a modified
three-stub filter having ideally a 5-percent stop-band width and 0.1-db
ripple in the pass-band was analyzed. The response in the frequency
range 0 to l.5f0 was shown to deviate no more than a negligible amount
from the exact response, while the filter was usable over a still
broader range.

This section describes a two-stub filter derived from an exact

maximally flat design that is constructed in strip line. The stop-band
width is 5 percent of the stop-band center frequency, and the basic filter,
which is identical with the basic filter of Sec. IV-B, has a maximally
flat response. The impedance of each stub of the capacitively coupled
filter was arbitrarily set. at 75 ohms for high unloaded Q. (This pro-
cedure allows an arbitrary choice of one design parameter for each
resonator.) The other design parameters given in the schematic diagram
in Fig. IV D-1(a)- the electrical length of each stub 0oi, the gap
susceptance w 0 Ci, and the impedance of the connecting line--were calcu--
lated by design formulas which have been previously given.1,3

A photograph of the filter is shown in Fig. IV D-2. Tuning screws

are visible in the bottom plate under the main line end of each resonator.
Similar screws were also put in the top plate. The short-circuit end
supports of the resonators were so constructed that the stub lengths and
capacitive gaps between the stubs and main line could be separately
adjusted. The first step in adjusting the otherwise finished filter

51
 so
LA (MEASURED) • 71 db LA(MEASURED)
AT w/w0 - 1.00 59db AT PEAK

25 0- MEASURED RESPONSE
- COMPUTED RESPONSE
fo" -" 1.6l Gc

20
0D

* 52.8
15 - 50 SOohms

801z-73.3 =573.71

WoC,0JD03M9 woC •0.003896

5)

0 -
0 1.0 .0. 3.0

wo
3.00

o MEASURED
COMPUTED FROM' NETWORK
2.50 INSERTION LOSS FUNCTION
fo 2.1.6 GC

~,2.00

1.50

00

0.50 0.750 1.00 1.25 1.50

-w
(wo

FIG. IV D-1 SCHEMATIC DIAGRAM OF CAPACITIVELY COUPLED FILTER,

AND ITS MEASURED AND COMPUTED RESPONSE

52
 0 0

S~~P-357-9

FIG. IV D-2 CAPACITIVELY COUPLED NARROW-BAND FILTER

WITH COVER PLATE REMOVED

53
was the setting of the capacitive gaps of the two resonators for the
proper loaded Q, and the second step was the tuning of each resonator
to center frequency. This procedure, described more fully in Quarterly
Progress Report 3,5 p. 57, for a similar filter, was facilitated by the
split-block construction of the resonator support that allowed Separate
adjustment of resonator length and capacitive gap. The measured response
of the filter is seen to be very close to the computed response in
Fig. IV D-](a), (b), wh-ich shows attenuation and VSWH respectively.
Figure IV D-3 gives the essential dimensions of this filter and reflects
the final adjustments.

0.500WINCH GROUND PLANE RESONATOR SUPPORT

S/SPACING,AIR DIELECTRIC AND GPOUND

oT
0o o 0 COPPER
COAX-TO- STRIP-LINE 0TI
TRANSITION

- J

0.050w 1.I54 1 0.35

6r
6ALUM.CENTER CONDUCTOR
0,200 THICK -

FIG. IV D-3 SKETCH OF CAPACITIVELY COUPLED NARROW-BAND FILTER

WITH COVER PLATE REMOVED, SHOWING DIMENSIONS

E. DERIVATION OF THE DESIGN EQUATIONS FOR FILTER SECTIONS

HAVING PARALLEL-COUPLED LINES

Design equations for symmetrical parallel-coupled band-stop filter

'sections of the type described in Sec. IV-I1 have been previously published
(Ref. 7, p. 215); the equations for the unsymmetrical case are given
herein. These latter equations were derived by applying the proper

54
terminal conditions to the general four-port consisting of two coupled
transmission lines, and then solving the set of simultaneous equations
obtained by expanding the matrix equation V a 'IZ based on the 4 x 4 im-
2
pedance matrix Z. The general solution for the spur-line type filter
was obtained in a similar way.

REFERENCES

1. B. M. Schiffman, P. S. Carter, Jr. , and G. L. Matthaei, "Microwave Filters and Coupling

Structures," Quarterly Progress Report 7, Section II, SRI Project 3527, Contract
DA 36-039 SC-87398, Stanford Research Institute, Menlo Park, California (October 1962).
2., H. Ozaki and J. Ishii, "Synthesis of a Class of Strip-Line Filters," IRE Trans. PGCT-5,
pp. 104-109 (June 1958).

3. E. M. T. Jones and J. T. Bolljahn, "Coupled-Strip-Transmission-Line Filters and Directional

Couplers," IRE Trans. PGMTT-4, 2, pp. 75-81 (April 1956).

4. Leo Young, G. L. Matthaei, and E. M. T. Jones, "Microwave Band-Stop Filters with Narrow
Stop Bands," IRE Trans. PGIfl-lO, 6, pp. 416-427 (November 1962).
5. P. S. Carter, Jr., Leo Young, G. L. Matthaei, and E. M. T. Jones, "Microwave Filters and
Coupling Structures," Quarterly Progress Report 3, SRI Project 3527, Contract DA 36-039
SC 87398, Stanford Research Institute, Menlo Park, California (October 1961).

6. Leo Young and G. L. Matthaei, "Microwave Filters and Coupling Structures," Quarterly Progress
Report 4, SRI Project 3527, Contract DA 36-039 SC-87398, Stanford Research Institute,
Menlo Park, California (January 1962).

7. G. L. Matthaei, Leo Young, and E. M. T. Jones, Design of Microwave Filters, Impedance Matching
Networks, and Coupling Structures, Vol. 1, pp. 170-193, a book prepared on SRI Project 3527,
Contract DA 16-039 SC 87398, Stanford Research Institute, Menlo Park, California (January 1963).
8. W. J. Geteinger, "Coupled Rictangular Bars Between Parallel Plates," IRE Tranm. iPaIUT-O,
pp. 6S-72 (January 1962).

55
 V NULTIPLEMS

A. GENERAL

Theory of the design of comb-line multiplexers having contiguous

channels, a three-channel multiplexer trial design, and the calculated
transducer loss per channel of the trial multiplexer design are discussed
in Ref. 1, Secs. III-E and F. The design of Sec. III-F (Ref. 1), except
for two minor changes,* was used as a prototype for a UHF three-channel
comb-line multiplexer that is discussed in this report. The construction
and illustrations of important components of the UHF three-channel comb-
line multiplexer are described in Sec. V-C of this report. The tuning
of the filters of the multiplexer and the staggering of the center fre-
quencies of the filters is reported in Sec. V--D. Section V-E discusses
the theory of designing, multiplexers having guard-bands between channels.

B. THREE-CHANNEL COMB-LINE MULTIPLEXER RESPONSE

A photograph of the completed three-channel comb-line multiplexer is

given in Fig. V B-1. The view is that seen when looking obliquely at the
top of the multiplexer. The common input may be seen in the center of the
photograph; the three remaining type-N connectors visible in the photograph
are the output connectors of the three comb-line filters. Four resonator
fine-tune screws are also visible on the top ground plane of each comb-
line filter. The susceptance-annulling network cannot be seen in this
particular view of the multiplexer.

The multiplexer attenuation, after final tuning and adjusting of the

annulling network for an optimum response, is given in Fig. V B-2.
Figure V B-3 gives the multiplexer attenuation in the pass-band using an
expanded ordinate. The attenuation in the pass-band of the lower- and
middle-frequency channels is typically 0.5 db. The attenuation in the
pass-band of the highest-frequency channel is less by approximately 0.1
to 0.2 db. The VSWI of the multiplexer (measured a-t the common input) is
shown in Fig. V B-4 and is less than 1.6 throughout t•he pass-band except
at the multiplexer band edges.

Tle mormelised slope perseeter, usI/A

wa cheaed to 0.918 is order to'optimise the resonator Q9. r.
end YA were node equal to 0,02 ik..

57
FIG. V B-I PHOTOGRAPH OF THE THREE-CHANNEL COMB-LINE MULTIPLEXER

58
35I II , II

O0X 0 MEASURED VALUES

30

25

20

Is

10-

5 1
• x
, X--X -/
0 . .
1.2 1.3 1.4 5 1.6
1.1
.. 1:.? 1.6 1.9 2.0 2.1
FREQUENCY- 6c

FIG. V B-2 ATTENUATION OF THE THREE-CHANNEL COMB-LINE MULTIPLEXER

59
 3.53I II

6 0 X0 MEASURED VALOIS
3.0 -

2.5

2.0
I .

-l 1.34 1.60.4

.0 L

.0.
0 I I I I I I I I-
1.1 1.2 1.3 1.4 1.5 ,.6 L.? 1. i.9 2.0 2.1
FREQUENCY - Gc

FIG. V B-3; ATTENUATION IN THE PASS BAND OF THE THREE.CHANNEL

COMB-LINE MULTIPLEXER

60
 •*I I
3.2
0 MEASURED VALUES
3.0

2.6

2.6

2.4

a.2
2.0,

1.0

1.6

1.4

1.2

1.0
12 1.3 1.4 1.5 1.6 I.? I.$ .9 2,0
FREQUENCY-Ge
A-3S2lT-,

FIG. V B.4 VSWR IN THE PASS-BAND OF THE THREE-CHANNEL

COMB-LINE MULTIPLEXER

61
 The 3-db percentage bandwidths of each channel, defined as the differ-
ence of the frequencies at which the attenuation is 3 db, divided by the
arithemetic mean of those frequencies, were determined and found to be 17,
16, and 13 percent for the lowest-, middle-, and highest-frequency channels,
respectively. These percentages are larger than the corresponding values
of the prototype multiplexer whose channel 3-db percentage bandwidths were
found to be each 11 percent. Bandwidth spreading in actual filters relative,
to the design prototype has been noted before. 2 However, in the case of a
comb-line filter having a design bandwidth of 10 percent, the bandwidth
spreading was less than in the present cases. Bandwidth spreading is
believed to be due to coupling effects beyond nearest neighboring line
elements, which were neglected in the derivation of the design equations
for comb-line filters.

C. CONSTRUCTION OF THE THREE-CHANNEL COMB-LINE MULTIPLEXER

The basic comb-line filter configuration is shown in Fig. V C-1. The

mutual and self capacitances of the resonators, the lumped-capacitance
resonator loadings, and the impedances of the large Z0 coupling wires were
calculated using the formulas of Table III E-l, Ref. I. The relative
spacings and dimensions of the resonators of the comb filters were deter-
mined using the design graphs of Getsinger.3 The dimensions of the reso-
nator widths and spacings in Fig. V C-1 are those independent of frequency,
while the lengths of the resonators, coupling wires, and output coupling
bars, denoted by L, s, and t, respectively, depend on the center frequencies
of the filters and are therefore tabulated in Table V C-1. Also in
Table V C-1 are the AWG wire sizes of the large Z0 coupling wires, the de-
sign center frequencies, and the experimentally measured center frequencies
of the filters of the multiplexer. The deviation of the mea-sured center
frequencies from the design values results from the bandwidth spreading
discussed in Sec. V-B.

A sketch of the top view of the multiplexer is given in Fig. V C-2,

in which one of the filters is shown with its cover plate removed. The
filters are arranged at approximately 120 degree intervals about a common
coaxial line input and are electrically coupled to the common input by a
direct connection of the large-Z 0 coupling wires to the inner conductor
of the coaxial line. The common coaxial line is located at the center of
the multiplexer by means of a permanently situated triangular section of

62
 T LARGE-Z,0
CUPLIN& WIRE
IMPEDANCE

r FROM BELOW -
4 PLACES

TO COTROL0.440

CAAPACIIVANCES

0.806 0.394

C0.37

CISD UG-1167/U
0.565 MODIFIED

0.112 -- E1Z
0.345

0.577 (4P= 0.156

L
NOTE:
1. Dissni on Iniches
. etable V-I I fo olfimefsl..s of L, -e, 3.dA ,.

FIG. V C-1 BASIC COMB.LINE FILTER CONFIGURATION

SHOWING IMPORTANT DIMENSIONS

63
 Table V C-1
DATA FOR THREE-CHANNEL COMB-LINE MULTIPLEXER

DESIGN CWME MEASURED CElflR* L, RES014ATO ~

XWf ,LRE 0 LARGE.Z 0
FIL1%R FRZQUERC FREQUENCY LEVUh COUPLING BAR COUPLING COUPLING WIRE
(Gc) (Gc) (s)AVGi WIRE SI]ZE

Low-Frequency 1.50 1.34 0.983 0.863 0.655 No. 38

Channel (3.96 mils),
iddle -Frequency 1.67 1.60 0.802 0.738 0.580 No. 44
&Real (2.00 mile)
igh.Frequency 1.86 1.85 0.791 0.674 0.527 No. 41
hinnel (2.75 oil*)

The meassuredcenter frequencies differ from the design center frequencies because of bandwidth spreading,
which is discussed in See. V.B.,

0 (TOP GROUNDPLANE REMOVED-

SHOINGONEOF
HE OM-LIE FLTESWTHTSCOVBLIER PILATE REMOVED

64REZoCULIGWR
ground plane (visible in Figs. V B-1 and V C-2). Use of the triangular
plate permits the removal of any filter cover plate without disturbing the
common junction or the other filters. Three small, circular cylindrical
rods: connect the triangular section of ground plane to the bottom ground
plane.

A detailed view of the common junction with the annulling network is

given in Fig. V C-3. The center conductor of the common coaxial line is
slightly longer than the outer conductor and is terminated in a small disc
in which very small holes are drilled in order to facilitate soldering the
coupling wire to the inner conductor. The inner conductor of the annulling
stub screws into the inner conductor of the common input line. The annul-
ling stub inner conductor is rounded at the junction end to provide clear-
ance for the coupling wire and also to decrease the junction capacitance.
The outer conductor of the annulling stub (see Fig. V C-4 for a drawing of
the outer conductor of the annulling stub) fits into three longitudinal
slots of the outer conductor of the common coaxial line input. Good elec-
trical contact between the outer conductor of the annulling stub and both
the bottom ground plane and the outer conductor of the common coaxial line
is assured by the locking plug which screws into the bottom ground plane.
This design of the common junction (Fig. V C-3) permits minimum distance,
and almost symmetrical paths to both ground planes for the currents flowing
in the inner wall of the outer conductor of the common coaxial line. The
electrical length of the annulling stub is varied by moving the sliding
short,. while the admittance is varied by changing the diameter of the inner
conductor.

D. TUNING THE MULTIPLEXER

The filters of the multiplexer were tuned using the alternating short-
circuit and open-circuit procedure discussed by Dishal4 and described in
its application to the comb-line filter in Ref. 2. First, the large-Z0
coupling wires of the lowest- and middle-frequency channels were discon-
nected from the common input while the highest-frequency channel of the
multiplexer was tuned. During the tuning the inner conductor of the annul-
!ling stub was removed from the common junction. However, the outer conduc-
'tor of the annulling stub was left intact in order to provide a minimum-
distance path to the bottom ground plane for currents on the inner surface
of the outer conductor of the common coaxial line input.

65
TRIANGULAR SECT-ION SA O
OF CVEN LATEUG-1I07lU MODIFIED

LARGE 2IMPEDANCE
COUPLNG WIRE

OUTER CONDUCTOR OF EOI 122

COMMON COAXIAL INPUT LINE RXLT
INNER CONDUCTOR OF
COMMON COAXIAL INPUT LINE012

LARGE Z0 IMPEDANCETWE
CULNWIEPLANES SECTION A-A

OUTER CONDUCTOR OF
ANNULLING STUB

SLOT IN OUTE R CONDUCTOR

LOCKINGTo RELIEVE~ SLIDING

OUTER CONDUCTOR OF
INNE
OFANNULLING
CONUCTO STUB

NOEIESLIOINGSHARE- INE SLNCNSHRT

FIG. V C.3 DETAILED DRAWING OF THE COMMON JUNCTION

66
 C

120*

/ No•o _
- 0.5000

4.- - 0.4160

•--4- 0.175
.29101
0.100
RAO. FROM
REAMER

7 RAO.
0.400 REQ.

] 0.100

3.175 T-

I
;I.

I I

DIMENSIONS IN INCHES SE T O- C
SECTION C-C

FIG. V C4 DETAILED DRAWING OF THE ANNULLING STUB OUTER

CONDUCTOR

67
 After the highest-frequency channel was tuned, a load was placed on
the output of that channel (the other channels remained decoupled during
this time) and the reflection coefficient was obierved on an oscilloscope
using a reflectometer and an electronically swept signal generator. The
reflection-coefficient trace on the oscilloscope was adjusted to the proper
shape by varying only the tuning screws of the resonator to which the
large-Z 0 coupling wire was attached. In this case the coupling of the end
resonator was much different from the other resonators, and the bandwidth
of the filter was sufficiently large so that the alternating short-circuit
and open-circuit procedure did not give entirely satisfactory results in
the pass-band response. However, it was known from the experience of
tuning the comb-line filter described in Ref. 2 that this could be correc-
ted by readjusting the tuning screws of the end resonator. Next, the at-
tenuation of the highest-frequency channel was recorded using a pen
recorder and a mechnically swept signal generator. Several frequencies
at various points on the attenuation curve were then determined using a
wave meter. From this data the band spreading and percentage ba-ndwidth
were calculated.

Using the data of the highest-frequency channel as a guide, the center

frequency of the middle-frequency channel was computed such that the cross-
over of the attenuation curves of the middle- and highest-frequency channels
would be at 4 db. A value of 4 rather than 3 db was used to allow for the
effect of the susceptance-annulling network, which was to be added after all
channels were tuned. It is this fact--i.e., that the filters of the multi-
plexer must be tuned individually because of interaction, and the annulling
network be added to the multiplexer only after all the filters are tuned--
that makes tuning the multiplexer. relatively complicated.

After the highest-frequency channel was tuned, its coupling wire was
disconnected and then the middle-frequency channel was tuned using the same
procedure. The center frequency of the middle-frequency channel had to be
shifted and the filter retuned several times in order to obtain the proper
crossover point in the attenuation curves. It was found that the middle-
frequency channel had slightly more bandwidth spreading than the highest-
frequency channel, so that the center frequency of the middle-frequency
channel had to be decreased from the original calculated value.

Repeating the above procedure, the lowest-frequency channel was next

tuned. The location of its center frequency was also found to be critical,
and the filter was re-tuned several times in orderto obtain the proper crossover.

68
 During the testing of the three channels for VSWR and attenuation,.
several changes were made in the sizes of the various coupling wires in
order to adjust the impedance levels of the three filters. The. final
values for the sizes of the large:-Z 0 couplingwires are given in
Table V C-1.

Next, the center conductor of the annulling stub was added, loads
were placed on all channels, and all coupling wires were connected to the
common coaxial-line input. A reflectometer was connected to the common
input and the reflection coefficient was observed on an oscilloscope using
an electronically swept signal generator whose sweep-width exceeded the
multiplexer bandwidth. The sliding short-circuit of the annulling stub
was positioned for an optimum response, and final adjustments of the tuning
screws of the resonators to which the coupling wires were attached were made.
Several times during the final adjustments the diameter of the inner con-
ductor of the annulling stub was decreased slightly to further reduce the
multiplexer susceptance. The fine tuning and susceptance annulling adjust-
ments were stopped upon obtaining the response given in Figs. V B-2 and
V B-3.

E. MULTIPLEXERS WITH GUARD-BANDS BETWEEN CHANNELS

The design of multiplexers having guard-bands between channels, rather

than contiguous channels, differs primarily in the design of the susceptance-
annulling network. It has been showni that the susceptance in the pass-band
of a multiplexer with contiguous channels may be partially cancelled by a
properly designed annulling network. This is possible because (assuming the
filters of the multiplexer to be designed from singly-terminated prototypes)
the susceptance takes a somewhat monotonic form in the pass-band, going
from capacitive to inductive near the center frequency of the multiplexer,
and therefore is of a form that may be partially annulled by a properly de-
signed resonant circuit. If the channels of the multiplexer are not con-
tiguous, however, the susceptance assumes a much different form and, th-e-re-
fore, the susceptance-annulling network requires a different design.

For example, the admittance of a hypothetical multiplexer, using

1.O-db Tchebyscheff ripple, singly-terminated filters, is shown in
Fig. V E-1. The frequency scale of Fig. V E-1 has been normalized with
respect to oj, the center frequency in radians per second of the lowest-
frequency channel. The admittance has been normalized so that the minimum

69
. .. . ...... . . S.. . . . . . . . .. . .. . I III I I III , i ,

-I. [i'j

.II I

S~II
II
SI I

S.
I I
I I

I €I

eac have nomlie

I , I• ,
badidh of W Ii,1 sh1adwdhi
l.,whr
1I$

I.IsI.I I.2 1.c I.e ITI 0 .1

HIMALIZfO
FEOSEOCY,•).I

FIG. V E-1 ADMITTANCE OF A HYPOTHETICAL MULTIPLEXER USING l:.Odb

CHEBYSHEY RIPPLE, SINGLY-TERMINATED FILTERS

value of its real part in the pass-band is 1.0. The multiplexer channels
each have normalized bandwidths of W = 0.1, where W is the bandwidth in
radiasn per second normalized with respect to the center frequency of the
lowest-frequency channel. The normalized guard-bands are each 0.1. The

susceptance of the multiplexer in the region of each of its channels re-

sembles closely that of an individual filter in the absence of the others.
This is because the guard-band is sufficiently large to limit significant
contributions of susceptance from the other channels. If the guard-band
were increased from its present value, the susceptance of the multiplexer
(in the vicinity of each of its channels) would resemble still closer that
of a given filter acting in the absence of the others. In that instance
the design of the susceptance-annulling network is made easier. (The
reasons are discussed in succeeding paragraphs.) On the other hand, if
the guard-band were decreased, the susceptance of the other channels would
contribute significantly to the total susceptance of the multiplexer (at a
given frequency) and the form of the susceptance would depart appreciably
from that of Fig. V E-1. In that instance,, annulling the multiplexer aun-
ceptance in its pass-bands may be very difficult.

71
 In the example of Fig. V E-1 the susceptance may be reduced by using
an open-circuit transmission line as the annulling network. Because of the
guard-bands, it is necessary to use a transmission line several wavelengths

long at the multiplexer pass-band frequencies. Depending on the frequencies

of operation and the guard-band width, this may be accomplished
conveniently
by using an open-circuited stub immersed in a dielectric which
has a large
relative dielectric constant. In the example of Fig. V E-l, the susceptance
was reduced by adding to it the susceptance of an open-circuit transmission
line having a normalized characteristic admittance of 1.427, and which was
2.5 wavelengths long at a - 1.0. The resulting admittance is given in
=lwo

Fig. V E-2. Because we ar6 using a hypothetical ideal lossless transmission

line, there are poles of ausceptance at the normalized frequencies 0.9, 1.1,
1.3, etc. Note that the susceptance is generally reduced but retains two
large peaks in the pass-bands of the lowest- and highest-frequency channels.

One reason for this result is the fact that the contributions of sus-
ceptance from other channels have shifted the zeros of the susceptance of
the three channels. Note that in Fig. V E-1 there are susceptance zeros

a.. I - I .

1 IA 1.4 1 .5

El I II 111.1LIE EQU[TT, --

II | I ,•

.5.0.11.2 1.3 1,4 1.5 1.6

SORUIALIZ[S
FRIOUEKC,
("•0)

ýFIG. V E-2 ADMITTANCE OF HYPOTHETICAL MULTIPLEXER WITH ANNULLING STUB

71
at 1.02, 1.185, and 1.378 rather that at 1.0, 1.2, and 1.4, which would
be their locations had there been no susceptance contributions from other
channels. Important, also, is that the frequency difference between aus-
ceptance zeros is 0.165 and 0.193 rather than 0.2. Thus, because the
poles of the annulling stub repeat at a fixed frequency interval (0.2 in
our example), it should be expected that the annulling of the multiplexer
susceptance in its pass-bands will not be completely satisfactory.

The driving source was assumed to have an internal normalized admit-

tance of 1.13 mho (the mean value of the real part of the multiplexer ad-
mittance in the pass-bands), and the values of input admittance were used
as given in Fig. V E-2, to calculate the multiplexer attenuation. The re-
sult is given in Fig. V E-3. In the pass-bands of the three channels the
attenuation is typically less than 0.3 db. However, in the low- and high-
frequency channels the attenuation does rise to approximately 0.6 db be-
cause of insufficient reduction of susceptance in the pass-bands of these
channels.

__PASS-Sail PAM-SAISS Pm-Sa-ss}

SCIH¢KAdLE 1101TSCALE 5151li MALE

41

s S

10
h. 1.1 1.2 I.) .4 I.E

UORNALIZEO (, -)
FRIEUENCY.

FIG. V E-3 ATTENUATION OF HYPOTHETICAL MULTIPLEXER WITH ANNULLING STUB

72
F. SUSCEPTANCE FORMULA FOR AN IDEAL MULTIPLEXER WITH GUAID-BANDS

In the investigation of designing multiplexers having contiguous

channels,' Bode's integral formulas relating real and imaginary parts of
passive admittances of minimum susceptance type was applied to the real
part of the admittance of a hypothetical multiplexer. The normalized
real part of the multiplexer admittance was supposed to be unity in the
pass-band and zero outside. The analytical expression for the susceptance
that resulted from Bode's formula was useful for sketching quickly the
susceptance, and, particularly, for estimating the parameters of the anul-
ling network.

We will generalize the expression of Ref. 1, which relates the real

and imaginary parts of the hypothetical multiplexer having contiguous
channels, to account for the case of an N-channel multiplexer with guard-
bands between channels. The bandwidth of the N channels of the multiplexqer
may vary as well as the guard-band widths, so that the expression to be
stated is general and includes the formula for contiguous channels as a
special case.

A graph of the normalized real part of the input admittance of the

hypothetical multiplexer with guard-bands between channels is given in
Fig. V F-1. There areN channels with guard-bands between channels. The
first channel begins at wi and ends at w2 . The second channel begins at
W3 and ends at WV' and so on. For each channel the normalized real part
of the admittance is unity in the pass-band and zero outside.

For all w, the expression for the susceptance is

B(w) I (W-w1 )(-+co 2 )(+-c 3 )(•+_W 4 ) ... (+-.I )(W Wi)

(V-F-I)

73
W
U
z
I-

U)
h-

I- 1.

,-I

-J
L4 MCHANNELEXER WI. GURDBAD CHANNEL
SI 2 3 N/2
IhJ
N

WI wZ 4*3 W4 W5 W•
6 "N-I "''N '4

FREQUENCY

FIG. V FI NORMALIZED REAL PART OF THE ADMITTANCE OF A HYPOTHETICAL

MULTIPLEXER WITH GUARD-BANDS BETWEEN CHANNELS

REFERENCES

1. W. J. Getainger, E. G. Cristsl and G. L. Matthaei, "Microwave Filters and Coupling

Structures," Sixth Quarterly Progress Report SRI Project 3527, Contract DA-36039
SC-97398, Stanford Research Institute, Menlo Park, Calif. (August 1962).
2. G. L. Matthaei, L. Young, and P. S. Carter, Jr., "Microwave Filters and Coupi-ing
Structures," Fifth Quarterly Progress Report SRI Project 3527, Contract DA-36-039
SC-97398, Stanford Research Instituter Menlo Park, Calif. (May 1962).

3. W. G. Getainger, G. L. Matthaei, "Microwave Filters and Coupling Structures,"

Second Quarterly Progress !1eport SRI Project 3527, Contract DA-36-039 SC-9739a,
Stanford Research Institute, Menlo Park, Calif. (July 1961).
4. M. Diahal, "Alignment and Adjustment of Synchronously Tuned Multiple-Resonator-Circuit
Filters," Proc. IRE 39, pp. 1448-1455 (November 1951).
5. H. Bode, Network Analysis and Feedback Amplifier Design, p. 335, (D. Van Nostrand
Company, Inc., Now York, N.Y., 1945).

74
 VI WIDEBE•ID INTEIDIGITAL FILTERS WITH CAPACITIVELY
LOADED ISOONAtS

A. GENERAL

Some interdigital structures suitable for use as band-pass filters

have been discussed in Sec. III of Quarterly Progress Report 4.1 As de-
scribed there, the coupled-line elements are short-circuited at one end
and open-circuited at the other end, and were X/4 long at the center fre-
quency in the pass-band. In the filters to be considered in this section,
the coupled-line elements are made less than X/4 long at resonance, and
capacitive loading to ground is added to the open-circuited end of each
resonator. The specific arrangement of coupled-line elements, loading
capacitances, and terminating admittances that will be considered here is
illustrated in Fig. VI A-1. Note that the input and output lines are con-
nected to loading capacitors. With this arrangement the structure will be
most suitable for filters having moderate to wide bandwidths.

LINE NUMBERS 1 2 3 4 . . . n-I

CS C CS

TERMINATING TERM
LINENE
OI"TNCE ADMITTANCE
ADMITTANCE
vA YI

FIG. VI A-i CAPACITIVELY LOADED INTERDIGITAL FILTER

WITH UNGROUNDED END RESONATORS

These capacitively loaded interdigital filters have a number of very

attractive features, including the following:

75
 (1) They are very compact. They are even more compact in the
dimension parallel to the coupled-line length than are the
interdigital filters without capacitive loading. The cal-
culated frequency response curves presented in Sec. VI-F
demonstrate that capacitive loading can reduce the length
of the coupled-lines by a factor of at least two.

(2) The second pass-band is centered at greater than three

*:,mes the center frequency of the first pass-band. The
more the resonators are shortened by capacitive loading,
the higher in frequency will be the second pass-band.
For example, if the resonators are X/8 long at the center
of the first pass-band, the second pass-band will be
centered at between 4 and 4.5 times the center frequency of
the first pass-band.

(3) The tolerances required in their manufacture a-re relatively

relaxed as a result of the relatively large spacings between
resonator elements.

(4) For a given bandwidth, number of resonators, and length of

resonators, the spacing between resonators will be greater
than for the comb-line filters described in Sec. II of
Quarterly Progress Report 5.2 This would be an advantage
in maintaining tolerances during manufacture. This difference
between capacitively loaded interdigital filters and comb-
line filters becomes insignificant, however, as the resonators'
are made very short in terms of wavelength.

(5) The rates o-f dutoff and the strength of the stop-bands are
enhanced by multiple-order poles of attenuation at zero
frequency, and within the stop-bands above the first pass-band.

(6) These filters can be fabricated in structural forms which are

self-supporting so that dielectric material need not be used.
Thus, dissipation loss that would occur within the dielectric
can be avoided.

If all possible couplings between the elements of an interdigital

filter are taken into account, then even analysis of the filter behavior
involves unwieldy mathematics. Synthesizing a structure to have a pre-
scribed response would be even more difficult; thus it does not appear
practical to attempt to devise an exact synthesis procedure. The synthesis
procedure given here involves a number of simplifying approximations that
permit straightforward, easy-to-use design calculations. Although the
design formulas are approximate, the calculated frequency response pre-
sented in Sec. VI-E indicates that usable designs can be obtained for
many practical applications.

76
 For the benefit of readers who have filter design requirements but
who are not particularly interested in filter theory as such, the design
equations for capacitively loaded interdigital filters, and some calcu-
lated frequency responses, will be presented first. Readers who are also
interested in the source of the design equations will find a discussion
of their derivation in Sec. VI-F, and a discussion of the method of cal-
culating frequency response in Sec. VI-G.

B. LOW-PASS PROTOTYPE FILTERS AND USE OF A LOW-PASS-

TO-BAND-PASS TRANSFORMATION

The design procedures to be described are based on the use of low-

pass prototype filter element values to give the band-pass filter its de-
sired characteristics. Figures VI B-1(a), (b) show typical low-pass
prototypes of the type under discussion, and they also define the manner
in which the element values gk are to be specified for the purpose of this
discussion. The left side of Figs. VI B-2 and VI B-3 show typical maxi-
mally flat and Tchebyscheff responses such as can be achieved by filters
of this type. Element values for prototype filters having such responses
have been tabulated. 3 ,4 Notice that n is the number of reactive elements

(a)

ILfgl L3*g 3 nI

(b)

FIG. VI B-I DEFINITION OF PROTOTYPE FILTER PARAMETERS

go, 9gl' g2 ... gn' gn+ 1 - [A prototype circuit is shown
at (a) an its dual is shown at (b). Either form will give
the same respona.. An additional prototype parameter
wl1 is defined in :Figs. VI B-2 and V I B-3.1

77
 PROTOTYPE RESPONSE $ANO-PASS FILTER RESPONSE

LA LAO
- * ---- "

Gu L~r IL~r~
4
W,. .W %iI
Wo W2W

, 10 log1 o 1 + db L1) - L(w') db

where where

a
-ntilog,, I- Co -

Wo2 + W
oo I
2
and
'2 -W
' as

Wo0

To determine n required for given values o-fco/Co0 , LAr,r, C/w

0 and LA.- find
smallest integer n value which satisfies

lo
01 [antilog, 0(LA)]-

01
where

I __I
___________
lol
FIG. VI B.2 EQUATIONS AND PARAMETERS FOR MAXIMALLY FLAT RESPONSE

75
 PROTOTYPE RESPONSE BAND-PASS FILTER RESPONSE

* z

0 fo0 hWoW b
&8--

#w
-w 2w, w

L (o'W) - 10 loglo + C co 82 nco a-1 db L A(co) . L'A(cj') dh

for .. where

db W
LA(w')- 10 log, 0 fl+C cosh 2 n cosh"1

for C&Y CO~"&'2 Wo " - COJ

e -i . 2
0 0~nioj(ý-
a a
Le')*1 ntioglO{ 0 oh2f- ] l(..)] C oo-
where
10)_ I___

To determine n required for give2 values of A)1160 LAr W,/o and LA.

find smallest integer n, value which satisfies

CantilOg a
l#O W~
I W1 I
cosh-1 - -

cosh• •

where
{[nioj(grJ-i ,o

FIG. VI&I3 EQUATIONS AND PARAMETERS FOR TCHEBYSCHEFF RESPONSE

79
and that, including the resistor terminations, the element values range
from g0 to 9.+1' A low-pass prototype with n react-ive elements leads to
a band-pass filter with n resonators.

The right sides of Figs. VI B-2 and VI B-3 show band-pass filter
responses which correspond to the given low-pass- prototype responses.
The band-pass filter response will have the same type of pass-band char-
acteristic as the prototype, but the width of the band-pass filter pass-
band can be specified arbitrarily. Filters constructed with distributed-
cons:tant elements will also have other pass-bands centered at frequencies
above w0" On the left sides of Figs. VI B-2 and VI B-3 will be found
equa-tions for determining the attenuation characteristic of the low-pass
prototype as a function of the radian frequency variable w', for specified
a) and n. On the right side of Figs. VI B-2 and VI B-3 will be found an
approximate low-pass-to-band-pass transformation from which the attenuation
characteristic of the low-pass prototype as a function of co' can be mapped
to the band-pass filter attenuation chatacteristic (centered at wo) as a
function of the band-pass filter radian frequency variable W. Since the
attenuation is the same for both the low-pass and band-pass filters at
frequencies w' and, n), respectively, which are related as given by the
mapping, the band-pass filter attenuation characteristic can be predicted
by use of these data. At the bottom of Figs. VI 8-2 and VI B-3 are equa-
tions for determining the value of n required in order to achieve a given
amount of attenuation, LAa db, at a given frequency, w..

The low-pass-to-band-pass transformation given in Figs. VI B-2 and

VI B-3 is an approximate one, and is usually considered to be useful only
for filters of narrow or moderate bandwidth. However, it appears to work
fairly well for capacitively loaded interdigital filters even in cases of
qu-ite large bandwidth, as is verified by the nearly octave bandwidth de-
sign discuss-ed in Sec. VI-E. One limitation of the low-pass-to-band-pass
transformation is that it assumes the pass-band to have arithmetic sym-
metry about o0, whereas the capacitively loaded interdigital filters have
a pass-band shape that is slightly narrower above w. than below w0 " The
parameter w, which appears in the mapping equation, is the fractional
bandwidth of the band-pass filter. This parameter will also be used in
the filter design equations that are about to be presented. Our experi-
ence indicates- that, because the design equations are based on various
approximations, the actual fractional bandwidth of the filter will be
about 6 to 10 percent smaller than the value of w used in the equations.

80
Thus, to allow for this shrinkage in bandwidth, it is suggested that a
value of w whic~h is about 6 to 10 percent larger than is actually desired
be used in the design. equations in Secs. VI-E. However, the actual de-
sired value of w should be used. in the mapping in Figs. VI B-2 and VI B-3.

C. PARALLEL-COUPLED LINES

As previously mentioned, interdigital filters consist of arrays of

parallel lines between ground planes. The design equations for such
filters to be given herein will yield the various line capacitances per
unit length for the various lines in the filter array. From these values
of capacitance per unit length, the actual dimensions of the lines can
be determined. Definitions for these various line capacitances will now
be given, and also equations for obtaining the line dimensions from the
specified capacitances. In order to relate the design methods under dis-ý
cussion to existing design data that are extremely useful when adapted
for the design of arrays of parallel lines, we will first consider the
case of two unsymmetrical parrallel-coupled lines. These methods will
then be extended to the case of an array of parallel-coupled lines.

Case of Two Unsymmetrical Parallel-Coupled Lines--Some unsymmetrical

parallel-coupled lines that are quite easy to design are shown in
Fi~g. VI C-1. Note that the capacitance per unit length of each line has
been separated into component parts. The capacitance C* is the parallel-
p
plate capacitance per unit length between one side of Line a and the ad-
jacent ground plane, while C6 is the c~orresponding parallel-plate capac-
P
itance of Line b. The capacitance C; is the fri-ng~ing capacitance per unit
length for any of the outer corners of t-he strip lines. The fringing

ELECTRIC WALL FOR ODD MODE

CCIO Cf C

PARALLEL-COUPI ED LINES
capacitances per unit length at the inner corners of each strip are des-
ignated Cf'
1 when the strips are excited in the even mode (i.e., with
equal voltages of the same phase); they are designated C; when the strips
are excited in the odd mode (i.e., with voltages having equal amplitudes
'and opposite phases). Both bars have the same height, and both are as-
sumed to be wide enough so that the interactions between the fringing
fields at the right and left sides of each bar are negligible, or at least
small enough to be easily corrected for. On this basis the fringing
fields are the same for both bars, and their different total capacitances
C. and C. to ground are due entirely to different parallel-plate capac-
itances CA and C6, For the structure shown in Fig. VI C-1,
p P

C, = 2(C; + C; + c' )

C (C' - C.,)
10

C a- 2(Cb + C; + C .) (VI C-l)

Getsingers has derived equations for the fringing capacitances

C;., Cfo, and C>, and has prepared convenient charts which relate CfE,
C,0 /6, and CY/E to rectangular-bar strip-line dimensions. Here E is the
absolute dielectric constant of the medium of propagation, so that the
above ratios are dimensionless and of moderate size. Getsinger gives
equations for the design of symmetrical, parallel-coupled, rectangular
strip-lines, and here we will adapt his equations to fit the unsymmetrical
case.

Note that the shape of the lines in Fig. VI C-1 is fixed in terms of
the dimensions t,b,s,w. and w,. To design a pair of lines such as those
in Fig. VI C-1 so as to have specified odd- and even-mode admittances or
impedances, first use equations such as those in Table VI D-1 to compute
C/' 6 , Ca 4 /e, and C,/E. Select a convenient value for line thickness t
and plate spacing b, which fixes the ratio t/b. Then, noting that

tAd Co
A = -
(VI C-2)
E E

use Getainger's charts of AC/c and C' vs. s/b to determine s/b, and also
C;./I. Using t/b and Getainger's chart of Cus. tb, determine

12
and then compute

. ( ) ) (VI C-3)

b:~~ -1(t2 ) C ;. (VI C-4)

When the ground plane spacing b is specified, the required bar widths,
. and wv, are determined. If either w0 /b or v4 /b is less than
0.35(1 - tib), the width of the bar should be corrected using the approxi-
5
mate-formula

' 0.07 -- +
b 1.20 '(VI C-5)

provided that 0.1 < (w'/b)/(l - t/b) < 0.35. In Eq. (VI C-5) w is the
uncorrected bar width and w' is the corrected width. The need for this
correction arises becausc of the interaction of the fringing fields at
opposite sides of a har, which will occur when a bar is relatively narrow.

Case of an Array of Parallel-CoupledLines--Figure VI C-2 shows an

array of parallel-coupled lines such as is used in an interdigital filter.
In the structure shown, all of the bars have the same t/b ratio and the
other dimensions of the bars are easily obtained by generalizing the pro-
cedure just described for designing unsymmetrical parallel-coupled lines.

k , -.C2

FIG. V! C-2 CROSS SECTION OF AN ARRAY OF PARALLEL-COUPLED

LINES BETWEEN GROUND PLANES

83
"In the structure in Fig. VI C-2 the electrical properties of the struc-
ture are characterized in terms of the self-capacitances, C,, per unit
'length of each bar with respect to ground, and the mutual capacitances,
per unit length between adjacent bars k and k + 1. This repre-
sentation is not necessarily always highly accurate; it is conceivable
that a significant amount of fringing capacitance could exist between a
giv-en line element and, say, the line element beyond the nearest .neighbor.
However, a-t l-east for geometries such as that shown, experience has shown
this representation to have satisfactory accuracy.

For design of the capacitively loaded interdigital-filter structures

discussed herein, equations will be given for the normalized self and
mutual capacitances, C,/E and C k,k+ 1 /1, per unit length for all the lines
in the structure. Then the cross-sectional dimensions of the bars and
spacings between them are determined as follows. First, choose values for
t and b. Then, since

C= 1 (VI C-6)
E C

Getsinger's chartsS can be used to determine sk.k1/b. In this manner, the

spacings sk,k 4 1 between all the bars are obtained. Also, using Getsinger's
charts, the normalized fringing capacitance (CGe ,),k.i/E associated with
the gaps sk,,, 1 between bars are obtained. Then the normalized width of
the kth bar is

-l .-. . (VI C-7)

b 2k b/L2\ E E

In the case of the bar at the end of the array (the bar at the far left
in Fig. VI C-2), C;e/E for the edge of the bar that has no neighbor must
be replaced by C;/e, which is also obtalned from Getainger's charts. Thus,
for example,, for Bar I in Fig. VI C-'2:,

1 --
1~ --c (C (VI C-8)
b 2 7 [\ -l2(CI)

If %/b < 0.35 (1 - t/b) for any of the bars, the width correction given
in Eq. (VI C-5) should be applied to the affected bars.

84
D. DESIGN EQUATIONS

The approximate design equations for the capacitively loaded inter-

digital filter illustrated in Fig. VI A-1 are summarized in Table VI D-l,
Based on previous experience with interdigital filters without loading
capacitances, these design equations are thought to be most practical for
filters having moderate to wide bandwidth. The calculated frequency
responses presented in Sec. VI-E demonstrate that the design equations
are usable for bandwidths up to an octave wide, and perhaps these design
equations -would be usable down to bandwidths' on the order of 30 percent.
These design equations are generalized to the case of unequal terminating
admittances at the ends of the filter, and the lumped-element prototype
on which designs are based need not be either symmetrical or antimetrical.*

To use Table VI D-1 for the design of capacitively loaded interdigital

filters, first use Fig. VI B-2 or V1 R-3 to estimate the number, n, of re-
active elements required in the low-pass prototype in order to give the
desired rate of cutoff with the desired pass-band characteristics. When
the prototype cutoff frequency a, and the element values go, gl"
9.-1 have been obtained from tables 3' 4 or by computations, 6
these values
can be inserted into the equations of Table VI D)-I. It is suggested that
the value of filter fractional bandwidth, w, used with Eq. (1) of
Table VI D-1 be 6 to 10 percent larger than is actually desired, since ex-
perience has shown that there will be some bandwidth shrinkage due to the
approximate nature of the design equations. In the equations in
Figs. VI B-2 and VI 13-3, however, the desired value of w should be used.
It is also suggested that the resonance frequency c,0 be chosen 5 to 9 per-
cent above the desired arithmetic average of the band-edge frequencies,
since the capacitively loaded interdigital filters do not give a frequency
response that has arithmetic symmetry about u0"

The parameters, d, hA, and h. are essentially admittance scale

factors. For theinterdigital filterswithout capacitive loading discussed
in Quarterly Progress Ieport 4, the frequency response of the filter did
not depend on the parameters d, hA, or h0 when- di-s-sipati-on loss isneglected.

A network is antimstrical if the impedance-vs. -freqaency function looking into, one end of the network
in the reciprocal of that looking into the other end of the network. Prototypes with reeistor termi*
nations at both aede. having either maximally flat- or Tchebyetheff renponsesn with one or more frequenciea
w where zero tranducer losn occurs, are either symmetrical or antinetrical.

85
 Table VI D-1
DESIGN' EQUATIONS FOR INTEBDIGITAL FILTERS OF THE FOIR IN FIG. VI A-i

Use the equations in Fig. VI B-2 or VI B-3 to select a low-pass prototype with the re-
quired value of n Z 4. The input and output elements in this filter count as resonators, ad
that there are n elements for an n-reactive-eleaent prototype. Choose 0, 2=i 0Xo - the elec-
trical length of the elements at the resonance wavelength 0 in the medium surrounding the
line elements. Compute:

8. 0 Wo '9 (1)

i d~
tA 90 1 9 A (2)

s = (i w0cot 80 tan~i , T =S(. cot. e0 * elm)8 (4)

Nh41&
toa-2." A)(3)~i) 0 ~ gW00

N0 h-,
A - gd)S tan (o-92,o 8,1 (5)

" " AA(

Y• 121-d~
jf M'"tan 01 J2.3

"to -- go + 5*"n.- • 9in (7)

•A h.4 NiI1b Nk,k+I)S' _ A A/-*÷J

* sn8

k[3 t o )m-21

(9)

86

_______________ ______
 Table VI D-1 (continued)

Z t.8/
(10)

bk=+l "h- . (11)

n=2
k=2 to

The normalized aelf-capacitancea per unit length for the coupled-line elements
are:
C 1-v7E
376.7_• • A 1)
" " A (ZIZA.)'

37 6 7
.L2 _ .

o o,

CC
- A(15)

where e and (p are the absolute and relatie dielectric constants, respectively, of
the medium surrounding the reaonator line elements. The d, hA, and A8 are dimension-
lea. admittance scale factors whose values should be chosen to give a convenient ad-
mittance level in the filter (see text).

The normalized nutual capacitances per unit length between adjacent coupled-line
elements are:
C- 376.7

lb=2 to n-2 ,v•- A ,'1A8

e 376.7 YA (Z7I ) (19)

67
 Table VI D-1 (cosnclud)

The lumped loading capacitances are:

YACL
ao. 9 (20)

'YA ot 6 (0 Y Y
2 WOo_ A A(21)

A - ..- (2
1 t M/2
n , cot0

"-! A Co a %-2n-1 (23)

Csn Cot
WO(Z4/ZA) (24)

where the resonance frequency. W00is in radians per second, admittances are in mhos, and im-
pedances are in ohms to give capacitances in farads.

For the capacitively loaded interdigital filters, however, the value of

d influences the amplitude of the ripples in the pass-band.* The calcu-
lated frequency responses presented in Sec. VI-E indicate that d should
be between 0.5 and 1.0 for octave-bandwidth filters, with a value of
d - 0.65 giving best results. The choice of d is not extremely critical,
however, and the optimum value can readily be determined by a few trial
designs,. That is, values are assumed for d, hA, and h,, and if the re-
suiting frequency response is not satisfactory, another value of d is
tried.

:For the design equations presented in Table III E-1 of Quarterly Progress Report 4,1 d has been set equal

to 0.5. A single parameter h = hA = hB appears in Quarterly Progress Report 4, since equal terminatinag
'admittances were assumed in that report. The h-parameter was also introduced in Quarterly Progress
Report 4 at a different stage in the derivation of the equations. Except for theme, differences, the
dasign equations presented here in Table VI D-1 reduce to those is Table Ill E-1 of Quartatiry Progress
Report 4 for the case of so capacitive loading.

88
 The prime consideration in the choice of the admittance scale factors
';hA and h. is that the dimensions of the coupled lines be such that the res-
0onators have high unloaded Q in order that dissipation loss be low. The
dimensions that optimize the unloaded Q of interdigital filters are not
known. For air-filled coaxial-line resonators, however, it is known that
a characteristic impedance of 76 ohms gives the highest unloaded Q.
Various approximate studies suggest that the optimum impedance for strip-
line resonators of nearly square cross section, such as are shown in
Fig. VI C-2, is not greatly different than 76 ohms. Thus, it is suggested
that hA and h. be chosen such that Eq. (VI DVI) is satisfiýed (at least when
air dielectric is used):

2C&-l' +.+ 5.4.

E E E
k=-K/I for n even
=(n4-1)/2 for n odd. (VI D-1)

Equation (VI D-1) corresponds to making the coupled-line impedance 70 ohms

for the resonators in the center of the filter, under the conditions that
the adjacent lines are excited with voltages of equal amplitude but of
opposite phase (this is a generalized odd-mode admittance condition). The
trial designs mentioned above for determining the best value of d will also
provide values of odd-mode admittance at the filter center for.a given com-
bination of d, hA, and h . The parameters hA and h can then be adjusted
to satisfy Eq. (VI D-1).

Once values for d, hA, and hB have been determined, the remaining cal-
culations to determine the capacitances per unit length of the line elements,
C,/e and Ci k../E are straightforward. From these capacitances, the line
dimensions are determined as discussed in Sec. VI-C. The lumped loading
capacitances at the ends of the resonators can be constructed in many con-
figurations, such as those in Fig. VI D-1. Each configuration would require
estimates of the fringing fields in order to determine the dimensions for
the loading capacitances. Data for estimating.the fringing fields associated
with the loading capacitances are available 5 7 11
in other works., Since the
estimates of fringing capacitances will be approximate, tuning screws should
be incorporated into the first model of a given filter to trim the loading
capacitances. It is anticipated that the fringing capacitances associated
with Fig. VI D-l(c) will be discussed in a.Ufture report on
'Contract DA 36-039-AMC-00084(E), after a capacitively loaded filter is
constructed and tested.

I9
 (1a,NARROW GAP AT END OF LINE (b) TAB THAT IS LARGER THAN CROSS
ELEMENT SECTION OF LINE ELEMENT

(C) LINE ELEMENT PROJECTING INTO (d) COAXIA, CYLINDER AT END OF LINE
GROOVE ELEMEI1'1' RA-352T-667

FIG. VI D.1 VARIOUS CONFIGURATIONS FUR L!,AOING CAPACITANCES

AT ENDS OF LINE ELEMENTS

E. CALCULATED FREQUENCY RESPONSES

Some trial designs for capacitively loaded interdigi'al filters have

been worked out using an n = 8 reactive-element, Tchebysciioff prototype
with ripple amplitude LA, " 0.10 db. The element values for this prototype
are go = 1, g1 = 1.1897, g2 = 1.4346, g3 - 2.1199, g4 = 1..6010, gs = 2.1699,
6 = 1.5640, g 7 = 1.9444, g8 = 0.8778, g 9 = 1.3554, and w,' = 1. The line
elements were chosen X/8 long at the resonance frequency a0, so 00 = 7T/4.
The ratio o~i/', was chosen equal to 0.65, which would give 2.08-to-1.0 band-
width if the frequency response had arithmetic symmetry about coo. Equal
terminating admittances YA = YB = 0.02 mho were used, and the admittance
scale factorsh A - hB 0.18 were chosen to be the same as an interdigital
filter without capacitive loading whose measured performance was presented
in Sec. III-E of Quarterly Progress Report 4.1 The relative dielectric
constant, E,, was set equal to unity. The parameter d was given the values
0.25, 0.5,, 0.65, and 1.O, and: the relsulting filter parameters calculated
by Table VI D-1 are summarized in Tables VI E-1 through VI E-4, respectively.

90

- - -
 Table VI E-1
LINE-ELEMENT PARAMETERS FOR AN OCTAVE-BANDWIDTH, CAPACITIVELY LOADED
INTERDIGITAL FILTER WITH ADMITTANCE-SCALING PARAMETER d = 0.25

cb h + 1 y ., k+1 c OCr y
e sho *A_.h
I and 7 1.412 --- 1 and 8 1,:916 0.00884 ---
2 and 6 0.558 0.0740 2 and 7 0.758 0.00565 0.2085

3 and 5 0.373 0.0496 3 and 6 0.560 0.00396 0.0144

4 0.369 0.0490 4 and 5 0.550 0.00343 0.0729
z z 37 6.7//'
-= - 2.264 .6. 7 = 186 ohms
_t' +-2:,V

Table VI E-2
LINE-ELEMENT PARAMETERS FOR AN OCTAVE-BANDWIDTH, CAPACITIVELY LOADED
INTERDIGITAL FILTER WITH ADMITTANCE-SCALING PARAMETER d = 0. 50

k Ck,k 4 .1 Yk k+1 kCk WOec Yk

ýA Rho, TA

I and 7 1. 412 1 and 8 1.916 0. 00884 --

2 and 6 0.789 0. 1047 2 and 7 0.593 0.00583 0.1866
3 and 5 0.747 0.0991 3 and 6 1.098 0.00699 0.1458
4 0.738 0.0980 4 and 5 1.099 0.00686 0.1459

g
z 376.7/lve'
4L= ='A 2.264 2 - . = 93 ohms
A A 2 :.ýW -+ 2 ,

Table VI E-3
LINE-ELEMENT PARAMETERS FOR AN OCTAVE-BANDWIDTH, CAPACITIVELY LOADED
INTERDIGITAL FILTER WITH ADMITrANCE-SCALING PARAMETER d - 0._65

k S• 111
II
k___ •Rho
k'~k. ) k
-7-A
1 and 7 1.412 --- 1 and 8 1.916 0.00884
2 and 6 0.899 0.1194 2 and 7 0.504 0.00588 0.1748
3 and 5 0.971 0.1289 3 and 6 1.431 0,00876 0.1900
4 0.960 0.1274 4 and 5 1.429 0.00892 0.1896

z
-=
z
2.264 ,
~376.71- , a 71 ohms
2 C.4 *4 2 4.S

91 -7 -,
I
Table VI E-4
LINE-ELEMENT PARAMETERS FOR AN OCrAVE..ANBIIDTH, CAPACITIVELY*.UMD
INTERDIGITAL FILTER WITH ADMIITrAN.-SCALING PARAITER. d * 1.00

Ck ,k+1 yk•,•l CiYk

7A 4 wh Yr.

1 and 7 1.412 -- 1 and 8 1.916 0.00884 --

2 and 6 1.116 0.1481 2 and 7 0.318 0.00596 0.1501

3 and 5 1.493 0.1982 3 and 6 2.230 0.01285 0.2960
4 1.476 0.1959 4 and 5 2.198 0.01372 0.2916
Z Z 376.7,/V
2.24 " 4ohum
'k A 2S4+4+2C1

The capacitances in Table VI E-1 through VI E-4 could be used to deter-

mine physical dimensions for use in constructing the filters, as was ex-
plained in Secs. VI-C and VI-D. The admittances listed in Tables VI E-1
through VI E-4 are useful for calculating the frequency response of each
filter, as is explained in Sec VI-G.

For each of the filter designs presented in the paragraph above, the
transducer loss has been calculated as a function of frequency. These
calculations are based on open-wire-line equivalent circuits of the filters,
as is discussed in Sec. VI-G. Dissipation loss was neglected.
Figures VI E-1 through VI E-4 show the frequ.ency response for d = 0.25,
0.50, 0.65, and 1.0, respectively. Taking d - 0.65 seems about the best
choice, since this value keeps all the ripples below 0.35 db. However,
the value of d is not critical; the highest ripp-les for d = 0.50 and 1.0
are 0.4 db and 0.5 db, respectively.

If an exact synthesis procedure had been followed, these eight-

resonator filters would becapable of producing an equal-ripple response
with eight frequencies of zero transducer loss, neglecting the effects of
dissipation loss. The fact thatthese filters do not duplicate the0.l-db,
equal-ripple characteristics of the low-pass prototype from which these
were designed, is a result of the approximations made in derivation of the
design equations. The approximate synthesis procedure gives best control
of the frequency response at the resonance frequency, &)0' and at the lower-
b-and-edge frequency, wl. Note that Figs. VI E-1 through VI E-4 have exactly
0.10 db transducer loss at co,, and the lower edge of each pass-band is near
the specified value of 0.65 WO. If the frequency response had arithmetic

92
 V• I

70 •O"-"-" -

S- C0.6 0. '0.8 90.9

0 1.1 1.2 1.3 1.4
0

60
J ~SPECIFIED
0
w
EDGE
ItSBAND

30 •it
5-J

30C

2c

'C0
SEE ENLARGED
PLOT ABOVE
0 1 I 1- W I
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.$ 2.0
SNORMALIZED FREQUENCY
OC-352?-14

FIG. VI E-1 FREQUENCY RESPONSE OF A CAPACITIVELYrLOADED INTERDIGITAL

FILTER DESIGNED WITH ADMITTANCE-SCALING PARAMETER d - 0.25

93
 !Do Tm------ _ _

I .i I

sO 1.0 -

To- 0.- :

S600
0 -
0 61 , 0.8 0.9 1.0 1.1 1.2 1.3 1.4
SPECIFIE[D
BAND EDGE o

u 50 _

z
,0

S0'

• SEE ENLARGED
20o-- ~PLOT ABOVE-

30 0.- 0.4 0._ 0. _ 1.0 1.2 1.4 14 IB 2.0

wJ NORMAL IZEDO FREQUENCY

FIG. VI E-2 FREQUENCY RESPONSE OF A CAPACITIVELY LOADED INTERDIGITAL

FILTER DESIGNED WITH ADMITTANCE-SCALING PARAMETER d - 0.50

94
so A1.0I

-- TO 0.5 - - -|

0.%oO. .O. .7 0. 0; L 1 1.2 1. 1.4

IOC******~.-.. I I ....
-N SPECIFIEDE RQ N
BAND EDGE

50.

04 07-.-. . . L . . . .
4W

-67 *CC--

U SO

0.2 TE
0.4 io
NDWTHAMTACESAIGPAAEEi 0.6501

95
 90 1.5 I I T T1 I - _ _

s0
4
1.0-
I
70 0.5 9

. 10.7
0.66 0.6 0.9 1.0 1.1 1.2 1.3 1.4
U,

i SPECIFIED W0
,t BAND EDGE
ui
50

40-

30•

20

SEE
SEENLARGED
IPLOT ABOVE
o W W I W . W i W
O' 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.@ 2.0

-G , NORMALIZED :FHEQUENCY
c -3S27-671

FIG. VI E-4 FREQUENCY RESPONSE OF A CAPACITIVELY LOADED INTERDIGITAL

FILTER DESIGNED WITH ADMITTANCE-SCALING PARAMETER d , 1.00

96
symmetry about on, the upper edge of the pass-band would be at 1.35 o0"

The upper band-edge ia actually as much as 9 percent below 1.35 coo An

additional departure from the ideal response is that there are only two
frequencies of zero transducer loss above o0, rather than the four pos-
sible frequencies.

The departures noted above of the frequency response of capacitively-

loaded interdigital filters from the desired response are probably due to
two factors. One factor is that the susceptance of each loading capaci-
tance is a linear function of frequency, while the susceptance of each
short-circuited, coupled-line element is proportional to a cotangent func-'
tion of frequency, cot (810/wo
0 ). The other factor is that this cotangent
function is not symmetrical about the resonance frequency, 00, when
60 < 7T/2 due to capacitance loading. It is found, for example, that the
image admittance and image phases of the individual sections in the filters
are not symmetrical about wo. In addition, the image admittance as a func-
tion of frequency does not have zero slope at a)0, but has a negative slope
of cot (00 /W0) evaluated at Wc. Interdigital filters without capacitive
loading, on the other hand, have image parameters and frequency responses
with arithmetic symmetry about c%" The frequency response of the inter-
digital filter without capacitive loading described in Sec. III-E of
Quarterly Progress Report 41 was calculated, and exact arithmetic symmetry
was confirmed. The ripples all fell within 0.063 to 0.105 db, as compared
to the 0.i-db ripple of the low-frequency prototype f:rom which the filter
was: designed.*

To demonstrate the wide-stop-band potentialities of the capacitively-

loaded interdigital filter, the transducer loss is shown in Fig. VI E-5
over an extended frequency range for the filter having d - 0.65. The first
spurious pass-band does not occur until nearly 4.5 times the resonance
frequency ao within the main pass-band. If the coupled-line elements were
made even shorter than X/8 at co•0 , the stop-band would extend over an even
larger frequency range. For example, if the coupled-line elements were
X/16 at w0, the first spurious pass-band would occur at slightly less
than 8.5 c0 .

* The close correspondence between this calculated frequency response end the measured date presented in
Quarterly Progress Report 4 also demonstrates the validity of the method described in Sec. VI- of this
report for calculating frequescy response. The calculated bandwidth was 2.06:1 as compared with the
measured 1.95:1. and as compared to the design goal of 2.0:1. The calculated freqsency response had
eight frequescies of perfect input match, which is the maximum possible mamber for am bight-resometor
filter. The measered frequency response, however, had only six frequencioe of beat input match.

97
 180

120

,.J

U
0

W' C -loo_,_"

I,---

20__

WO ,NRAIE RQEC

FIG. V1 E-5 FREQUENCY RESPONSE OVER AN EXTENDED FREQUENCY RANGE

FOR THE CAPACITIVELY LOADED INTERDIGITAL FILTER
WITH: ADMITTANCE-SCALING PARAMETER d - 0.65
 That these predicted wide stop-bands can actually be obtained in
practice is indicated by the following arguments. Consider the case where
the capacitive loading is chosen to make the coupled-line elements X/8
long at the resonance frequency w•, as is the case for the filters whose
frequency responses are shown in this report. For these same coupled-line
lengths, in order for a pass-band to exist at 2ov0 , there would have to be
no reactive loading on the ends of the line elements. The capacitive
loading that exists completely detunes any pass-band that might otherwise
exist at 26, 0 . For a pass-band to exist at 3wv0 , the reactive loading at the
ends of the line elements should be inductive rather than capacitive. At
4w", the line elements are X/2 long, and the sum of the magnetic coupling
and electric coupling between adjacent line elements is zero.* Thus, a
pass-band cannot exist at 4600 . The first spurious pass-band would occur
when the length of each line element is X/2 longer than the original X/4--
that is, at 4.5 w0--if the susceptance of each loading capacitance were the
same at 4.5 wo, as it is at wo. Since the susceptance of each loading ca-
pacitance is actually greater at the higher frequency, the first spurious
pass-band will be tuned lower than 4.5 co•, but not as low as 4 coo. This
behavior of the frequency response is what is observed in Fig. VI E-5.

Although the capacitively loaded interdigital filters discussed in

this report do not yield as good pass-band characteristics as do inter-
digital filters without capacitive loading, Figs. VI E-2, VI E-3 and
VI E-4 demonstrate that filter characteristics can be obtained that are
very satisfactory for many practical applications. It is to be expected
that if the same number of resonators were used in a capacitively loaded
filter having less than octave bandwidth, the ripples within the pass-band
would more closely approach the ripple amplitude of the low-pass prototype
used in the design.

F. DERIVATION OF DESIGN EQUATIONS

The design equations for capacitively loaded interdigital filters

'presented in Table VI D-1 are based on the use of low-pass, lumped-element
prototypes whose element values are known to give prescribed low-pass fre-
quency responses. The parameters of the capacitively loaded interdigital
filters are. related to the prototype element values by a.method due. to
13
Matthaei wherein the image admittance and image phase of sections of the

This can he sees, for exmple, by the fact that the image impadses for filter 2 in Fig. 2(a) of ef. 12
goes to infinity when the lime elements are X/2. long (01- i,).

'9
filter are made equal to the image admittance and image phase of corre-
sponding sections ofthe prototype at certain important frequencies.
Splitting the low-pass prototype into sections involves modifying the
prototype from a ladder network of alternating series reactances and shunt
ausceptances, to a network of shunt capacitances separated by ideal admit-
tance inverters. This modification of the prototype is similar to a con-
cept introduced by Cohn, 6 and enlarged upon by Matthaei. 1 3 The image
admittance and image phase of the filter sections are found by means of
open-wire-line equivalent circuits for the coupled-line elements forming
interdigital filter.

The open-wire-line equivalent circuit for the capacitively loaded

interdigital filter can be drawn by considering the two types of coupled
sections in Fig. VI A-1. Each pair of coupled lines from Lines 2 and 3
through Lines (n - 2) and (n - 1) is of the same configuration as the two
lines on the left in Fig. VI F-1, and has the open-wire-line equivalent
circuit on the right in Fig. VI F-1. Figure VI F-1 is simply obtained
from Fig. III F-4 of Quarterly Progress Rleport 41 by adding the lumped
loading capacitances at the ends of the coupled lines. The Y% and Y:.
in Fig. VI F-1 are the odd-mode and even-mode characteristic admittances,
respectively, for Line a. The Yb and Yb are the odd-mode and even-mode
characteristic admittances, respectively for Line b. The capacitances
are as given by Eq. VI C-i, and v is the velocity of propagation in the
medium surrounding the coupled lines.

The end pair of coupled lines in Fig. VI A-1 are of the same config-
uration as the two lines, and have the open-wire-line equivalent circuit,
shown in Fig. VI F-2. Figure VI F-2 is an extension of Fig. III F-7 of
Quarterly Progress Report 4.1 In this equivalent circuit, the location
of the tuning capacitance was not considered as obvious as for the circuit
in Fig. VI F-1. Therefore, the open-wire-line equivalent circuit of
Fig. VI F-2 was derived directly from the open-circuit impedance and short-
circuit admittance matrices that apply in general to any two parallel-coupled
transmission lines. The elements of these matrices are given by Jones and
Bolljahn 1 2 for the case of equal-size coupled lines, and their results can
readily be generalized to the case of unequal coupled lines.*

Combining the two types of open-wire-line circuits on Figs. VI F-1

and VI F-2, the opev-wire-line equivalent circuit for the entire filter

N B. The1diuttaoMC-matrix elements givo by Eq. (9) of: Jones @EdBellja l1 have inmorreet algebraic
alga to Yl Y l. Y34' Y ' Y14' X4t' Y3 ad1Y32'

100
 0
a ~ I v8 Y0
0

4-. b 0A2 I I- b

Y~b.Ybe lb Y04voYO.-vCb

Cob+~b

FIG. VI F.1 CAPACITIVELY LOADED, PARALLEL-COUPLED LINES

USED AS INTERIOR RESON4ATORS IN FIG. VI A-1

cl.
+ Y e0 v(Co+Cab)
Clz::oco

I N.TURNS RATIO
Co"

2l 'CoT li

FIG. VI F-2 CAPACITIVELY LOADED, PARALLEL-COUPLED LINES

USED AT THE ENDS OF THE FILTER STRUCTURE
SHOWN IN FIG. VI A-i

101
 is as shown in Fig. VI F-3. To relate this equivalent circuit to the
prototype, Fig. VI F-3 is broken into sections as in Fig. VI F-4, the
interior sections S, 3 through S.,_ 1 being made symmetrical. The

transformers have also been removed to increase the resemblance between

Fig. VI F-4 and the modified prototype to be presented later in this
section.

The first condition applied to the interior sections of Fig. VI F-4

is the resonance condition. This condition is that when a purely real
admittance terminates any of the interior sections, the input admittance
is also purely real at the frequency wo. The tuning capacitances that
satisfy this condition are:

cot 60
Ct kkLk2 to
R
n-2
c h 0 (*
W 0h A -_
1 + Yk' . 4
l k~
) (VI F-1)

where 00 = 2t/k0 is the electrical length of the coupled-line elements

at COO, t being the physical length of the coupled lines, and K0 being the
wavelength within the medium surrounding the coupled lines at 600 . The
admi-ttance scale factor hA is defined in fig. V1 F-4. Other conditions
to be applied to the interior sections of Fig. VI F-4 involve the image
admittance of these sections, which is:

S k. cSC9 Y- . t k.kl
h tan9

-
(Ys)
lb..!L
to a-2 Aco, /A \k
tto .-

tan 00/co
Co

(VI F-2)

where & - 90col/w0 is the electrical length of the coupled-line elements at

any frequency w. Equation (VI F-2) was obtained by the usual procedure
of taking the geometric mean of the short-circuit admittance parameter of
one end of the section divided by the open-circuit impedance parameter of
the same end.* It is assumed in Eq. (VI F-2) that the tuning capacitances
satisfy Eq. (VI F-0).

* A worthwhile reduction in the complexity of the allehraiec maitlatioes reesults from teking advantee of
the eyumatry of the settimes. ach eaction cam be divided in half. nd the open-cireuit and the short.
circuit placed at the midplame.

102
 c" c"
CII i

v~~c' + Ci,2) Yn-2,n-I vCn-2,n-IVc +t-'

Y2 . V(I c,.c,) Y ,. -vC y v, C. Y. , . 1 v (Co +, c n)

c.2-.. .- . ,,.,2. C c., 1,c1)

hA \•C 1 +CI.2/ \cC-.+c- ,1

FIG. VI F-3 AN OPEN-WIRE.LINE EQUIVALENT CIRCUIT FOR THE CAPACITIVELY

LOADED INTERDIGITAL FILTER SHOWN IN FIG. VI A-i
All lines are of electrical length 0.

S ~-Cn hC
A

: 4 (. - Zn.

t I h\

Sl ;A S;• YS3 , 4 __n*C-I n-ln

O F,2ThA h- h"
A

hAý ha h# Ah

k-,5 3,4 k,' fhl-l2k .1,

52S 2 ,3 S3, S Z Sn-I~n

S-~ S 'I -!-iC4OF FIG.=F-

FOR k 2 THROUGH n-~ Vs *Yk-1kY Yk OF FIGM F -3Ct,h , ,hn ha
*I-IS2? -I?l

FIG. VI F-4 THE CIRCUIT OF FIG. VI F-3 SPLIT INTO SYMMETRICAL INTERIOR
SECTIONS, AND THE TRANSFORMERS REMOVED

103
 Consideration will now be given to the modified prototype shown in
Fig. VI F-5. Modification of the prototype in the forms shown in
Fig. VI B-1 to obtain Fig. VI F-5 in which admittance inverters" appear
has been described by Matthaei. 13 Figure VI F-5 is slightly more general
than Fig. i9 of Matthaeil3 in that unequal terminating admittances are
permitted, but slightly less general in that all the element values have
been renormalized so that the terminating admittances are the same as in

- Is
g,.

,A C2 2., heye

S• ----
SZ$ s'5 , 4 s- 2 ,.-,

JhCe.I T GoC
________

2 i
Jk * ,lag,,,I I 5 ____ __

kJ23 TO .0 k *-3SU?-SP7 4-3

FIG. VI F.5 MODIFIED PROTOTYPE FOR DERIVING-DESIGN EQUATIONS

FOR CAPACITIVELY LOADED INTERDIGITAL FILTERS

.A JIs
n-2,n-I I t og -,L - /

Fig. VI F-4. In order for unequal terminating admittances to be possible,

Lthere must be at least one admittance inverter in Fig. VI F-5, which re-
* quires nt >_ 4. !",xcept for the above-mentioned re-normalization, the el-
ements L,, C 2 , 1C Nl, and have the same
C values
9-1. that they had innig. V

,The other shunt capacitances have been chosen such that each interior sec-
tion, s;, 3 through S,' ,is symmetrical. The parameter d specifies
what fraction of C 2 is split int~o C.2 to be part of $2,3 and similarly
for C,_. These choices completely specify' 3 the valuesL f-or the edmittance-
inverter parameters, Js,, , as is shown at the bottom of Fig. VI F-5.

Ad invertera are defined a herine

VItt-ace u aconstant iege admittance J and a cofstant imagephaos of
tgo degreei at all frequeaes. (Thee they operitn like a qn erter-mavelength line of charecteristic ad-
mirteacn J at all freqoeacier.) Thea invert-era itroduced
nre here omly for mathematical conveniemte,
aed mo attempt mill he made to realis e zcte l that behave aa ad iittaace invertera.
scircuita

1-04
 The image admittance of each of the interior sections of the modified
prototype is readily found to be::

: ., / 'w'dg~h'Y' 2

+ Ik2 to n-2 ( .. .. (VI F-3)*

Both the modified prototype and the open-wire-line equivalent cir-

cuit of the filter are now in forms where the end sections consist of a
series reactance and a shunt susceptance, and the interior sections are
symmetrical. Attention will now be turned to relating the equivalent-
circuit parameters in Fig. VI F-4 to the prototype parameters in
Fig. VI F-5, with the interior sections being considered first. The first
condition applied requires that the image admittance of each interior
filter section, Eq. (VI F-2), evaluated at the resonance frequency W0 be
equal to the image admittance of the corresponding prototype section,
Eq. (VI F-3), evaluated at w' = 0. This condition gives the characteristic
admittance of the connecting open-wire-lines in Fig. VI F-4 in terms of the
admittance-inverter parameters:

ykk+II t - Jk,+lhA' sin 00 (VI F-4)

I ku2 to *-2

which is Eq. (11) in Table VI D-1. Equations (VI F-I) and (VI F-4) are
sufficient to also ensure that the image phase shift of each interior
section in Fig. VI F-4 is the same at co - co0 as the image phase shift of
the co~rresponding prototype section at co' = 0.

The second condition applied requires that the image admittance of

each interior filter section, Eq. (VI F-2), evaluated at the lower band-
edge frequency w be equal to the image admittance of the corresponding
prototype section, Eq. (VI F-3), evaluated at the band-edge frequency
w' - -(o4. These frequencies are as deifined. in Figs. VI B-2 and' VI 133.
After some algehraic manipulation, this condition produces Eq. (VI F-5):

Ykk*4 'L2 to n S hAYANh, - J.k, 4 lh A sin 0' (VI F-5)

105
where we define

( -- cot 80 tan 6

hto n-2 YAo ( Cos

The parameters S and N,.,- are introduced only to shorten the form of
some of the design equations, and have no particular physical significance.
Combining Eq. (VI F-5) with the relationship at the lower left-hand corner
of Fig. VI F-4, the characteristic admittances of some of the shunt stubs
in Fig., VI F-3 are found' to be:

Y1. s S hA'YANA-.h + S hAYANk.k+l

- -,k + Ji~ +,)hA sin 6 .0 (VI F-6)

This is Eq. (8)

in Table VI D-l. Equations (VI F-1) and (VI F-S) are also
sufficient to ensure that the image phase shift of each interior section in
Fig. VI F-4 is the same at w a w as the image phase shift of the corre-
sponding prototype section at w' -w

The end sections S 1 , 2 and S in Fig. VI F-4, and S, and S'

in Fig. VI F-S are sufficiently simple that only reactances and susceptancea
need be considered, instead of considering image admittances. For the series
stub at the left-hand end of Fig. VI F-4, its input reactance at w W 0 is
required to be the same as the reactance of the inductance L in Fig. VI F-5
at - 0-that is, zero reactance.
*' This specifies the end-loading capac-
itance in terms of the characteristic impedance and length of the end stub:

cot 0
G1 = (VI F-7)

.The characteristic impedance Z1 is determined by making the input impedance

of the series stub at the left-hand end of Fig. VI F-4, evaluated at w cl,
w

106
be the same as the reactance of the inductance LI, evaluated at w' *-w

which gives.:

Z, - 0-- T (VI F-6)

where T is defined as

T S(L-° cot 0 + tan 01

Equations (VI F-7) and (VI F-8). appear in Table VI D-1 as Eqs. (20-Y and
(6), respectively.

For the shunt stub at the left-hand end of Fig. VI F-4, its auscep-
tance in parallel with the tuning capacitance C,1.2' evaluated at o - (01
is required to be the same as the susceptance of the capacitance Cs in
Fig. VI F-5 at w' a 0-that is, zero shunt susceptance. Therefore:

C'
-1.
Y,1.2 cot 0(VI F-9)
hAW 0

Substituting Eq. (VI F-9) along with Eq. (VI F-l) evaluated for k w 2
into the relationships at the bottom. of Fig. VI F-4, we have:

Y, cot 00 Y2 Y ,2 3 )
C 2 Ws
0o - + LIYAI (VI F-10)

which is the same as Eq. (21) of Table VI D-1. The shunt susceptance of
stub Y1. 2 in Fig. VI F-4 in parallel with C.I2 evaluated at w - a) is now
made equal to the susceptance of C; of Fig. VI F-5 evaluated at w0 =-co.
This condition, along with Eq. (VI F-9), gives the stub characteristic
admittance:

hAYAw"(1 - d)g 2 S tan 61

Y ',2 (VI F-11)
go

107
Equation (7) of Table VI D-l results from taking the sum of Eq. (VI F-i1)
and Eq. (VI F-5) evaluated for k = 2.

The end section S.-,,, in Fig. VI F-4 is related to end section

in Fig. VI F-5 following the same procedure given above for sec-
tions S 1 , 2 and Sj, 2. The resulting equations are given in Table VI D-1
as Eqs. (9), (10), (23), and (24).

Equations (12) through (19) of Table VI D-I follow directly from the
definitions of the characteristic admittances and charateristic impedances
of the open-wire transmission lines in Fig. VI F-3. Use is made of the fact
that the velocity of propagation in the medium surrounding the coupled-line
elements can be written:

' - • (VI F-12)

AJA 4ie e T_ 376.7 e

where

- the permeability of free space - 47r x

*o 1,07" henry/meter
0 .= the absolute permittivity of free space - 8.85 x 10I'farad/
meter - 0.225 picofarad/inch
6 a the absolute permittivity of the medium surrounding
the coupled-line elements
e, - the relative permittivity (dielectric constant) of
the medium surrounding the coupled-line elements
1 *
a 376.7 ohms - chsracteristic impedance of free space.

Equation (VI F-12) applies for the usual case where, in order to minimize
dissipation loss, only nonmagnetic materials would be used in constructing
the filter.

G. METHOD OF FREQUENCY-RESPONSE CALCULATION

The frequency responses presented in Sec. VI-E were calculated using

the open-wire-line equivalent circuit of Fig. VI F-3. The validity of such
open-wire equivalent circuits is based on the fact that the pairs of coupled
lines on the left-hand side of Figs. VI F-l and VI F-2 have exactly the same
short-circuit admittances as functions of frequency as do the open-wire-line
equivalent circuits on the right-hand aide of the same figures. In order

108
to obtain a valid equivalent circuit for an entire filter by cascading
several of the equivalent circuits as in Figs. VI F-I and VI F-2, the
coupling beyond adjacent coupled-line elements of the filter must be
negligible. That coupling beyond nearest neighbor is insignificant for
coupled-line proportions similar to Fig. VI C-2 is demonstrated by the
expe-ime'ntal results obtained with the interdigital filter of Sec. III in
Quarterly Progress Report 4,1 with the comb.-l-i-ne filter- of-Sec- -I1in
Quarterly Progress Report 5,2 and with comb-line filters used in the multi-
plexe:r of Sec. III in this report.

As a specific check on the digital-computer program outlined in this

section, the frequency response was calculated for the interdigital filter
without capacitive loading whose measured performance was presented in
Sec. III of Quarterly Progress Report 4.1 The calculated frequency response
was in close agreement with both the design objectives for the filter, and
with the measured performance. The calculated bandwidth was 2.06:1 com-
pared with the design objective of 2.08:1, and compared with the measured
bandwidth of 1.95:1. The amplitude of the calculated VSWR peaks within
the pass-band fell between 1.27:1 and 1.36:1, compared with the design ob-
jective of 1.35:1, and compared with the measured values of 1.35:1 to 1.5:1.
The number of frequencies of minimum input-VSWR was eight for the calculated
data, and six for the measured data.

The frequency response of the open-wire-line equivalent circuit of

Fig. VI F-3 was calculated using the general-circuit parameters, A, B, C,
D, of the circuit exclusive of the terminating admittances. The general-
circuit parameters of the entire equivalent circuit was found by taking
the matrix product of the various filter parts that are cascaded to form
the entire circuit:

VB IjX
[TAi0
(VI G-1 continued)

109
 n-2 0coo 69 V. .

4-2jo C.-
1 0 Yj. 0
[: :j
L we -1 L(V .IG2

L._ cot L J "

(VI G-1 concluded)

where

O o • -( VI G- 2 )
I ) A A V0-)

z YA coto
1Z ZA (-cI/o 0)(o C0•)

ZA (-/Wi0) OWC81)

ohms (VI G-4)

Comparing the matrices with the parts of Fig. VI F-3, it can be seen that
the first matrix in Eq. (VI G-1) represenLs the input reactance of the
left-hand series stub, which has characteristic impedance Z and is loaded
1

at one end by capacitance C1. the second matrix in Eq. (VI G-1) represents
ther left-hand transformer. The first matrix within the continued-product

11e
sign in Eq. (VI G-1)represents the shunt stub of characteristic admit-
tence in parallel with the loading capacitance C;. The second matrix
Within the continued-product sign in Eq. (VI G-1) represents the con-
necting line of characteristic admittance Y2 , 3' and so on throughout
Eq. (VI G-1). The digital computer-a Burroughs 220--was programmed to
carry out the matrix multiplication of Eq. (VI G-J) for n < 15.

Using the general-circuit parameters obtained from Eq. (VI G-1), the
computer was also programmed to find the ratio of power available from the
generator, P .. i., to the power P2 actually delivered to the conductance
Y when. the filter is inserted between the generator and Y'. This ratio
is:

Pe~i Y / \ / i
[(F[A ++D)y +
+A (VYA + 2] (VI G-5):
P A YB B

where use has been made of the fact that for a lossless network B and C
are purely imaginary, by setting B = jb and C = jc, j = V-l. (Also, A
and D are purely real.) Finally, the. transducer loss LA was calculated by
the computer using:

LA 10 1og10
1 Pa decibels (VI G-6)

Ul1
 REFICES

1. Leo Young and G. L. Matthaei, "Microwave Filters and Coupling Structure,." auarterly
Pro6ress Repart 4, SRI Pro4 ect 3527, Contract DA 36-039 SC-87398, Stanford asearch
Institute, Menlo Park, California (January 1962).

2., G. L. Matthaei, Leo Young, and P. S. Carter, Jr., "Microwave Filters and Coupling
Structures," Quarterly Progreas Report 5, SRI Project 3527, Contract DA 36-039 SC-87398,
Stanford Research Institute, Menlo Park, California (May 1962).

3. L. Weinberg, "Additional Tables for Design of Optimum Ladder Networks," JI. Franklin
Institute 246. pp. 7-23 and 127-138 (July and August 1957).

4. G. L. Matthaei, et aL:, "Design Criteria for Microwave Filters and Coupling Structures,"
Final Report, SRI Project 2326, Contract DA 36-039 SC-74862, Stanford Research Institute,
Menlo Park, California (January 1961).
5. W. J. Getainger, "Coupled Rectangular Bars Between Parallel Plates," IRE Trans.
PFGMT-10, pp. 65-72 (January 1962).

6. S. B. Cohn, "Direct-Coupled-Resonator Filters," Proc. IRE 45, pp. 187-196(February 1957).

7. J. R. Whinnery and H. W. Jamieson, "Equivalent Circuits for Discontinuities in Transmissiom
Linee,"Proc. IRE 32, pp. 98-114 (February 1944).
8. J. R. Whinnery, H. W. Jamieson, and T. E. Bobbins, *Coaxial-Line Discontinuities,"
Proc. IRK 32, pp. 695-709 (November 1944).
9. N. Marcuvitz, Wavoguide Ifandbook, MIT Red. Lab. Series, Vol. 10 (McGraw-Hill Book Ce.,
Inc., New York, N.Y., 1951).
10. H. M. Altshuler and A. A. Oliner, "Discontinuities in the Center Conductor of Symmetric
Strip Transmission Line," IRE Trans. PWIfT-8, pp. 328-339 (May 1960).

11. J. F. Cline, et al., "Design Data for Antenna-Multicoupler Systems," Final Report,
SRI Project 2183, Contract AF 19(604)-2247, Stanford Research Institute, Menlo Park,
California (September 1959).

12. E. M. T. Jones and J. T. Bolljahn, "Coupled-Strip-Transmission-Line Filters and

Directional Couplers," IRE Trans. !G1T-4, pp. 75-81 (April 1956).

13. G. L. Matthaei, "Design of Wide-Band (and Narrow-Band) Band-Pass Microwavea Filters on

the Insertion Loss Basis, IJ8r Trans. PG0r1-, pp. 580-593 (November 1960).

112
S

VII CONCLUSIONS

A. MAGNETICALLY TUNABLE FILTERS

The new technique used in the two-resonator waveguide filter design

for gett'ing tighter coupling to a YIG sphere was demonstrated to be
successful. Using this technique, it should be possible to use smaller
spheres or spheres with lower saturation magnetization than would gen-
erally be practical otherwise. Using smaller spheres has the advantage
of reducing the tendency towards the excitation of magntetostatic modes
(which cause spurious responses). Using'ferrimagnetic waterials with
lower saturation magnetization makes possible operation to lower frequencies.

B. BAND-STOP FILTERS

The procedure for the exact design of the basic shunt-stub type of
band-stop filter that was described in Quarterly Progress Report 7 has
been extended to cover two types of filters that employ parallel-coupled
lines, with no loss of exactness in the designs. One of these types was
shown to yield a practical filter where the basic shunt--stub type would
have been impractical (namely, for the design of narrow-stop-band filters).

An advantage of the exact design procedure is that the filter respons.e

can be accurately determined over a very wide band (theoretically without
limit) once the prototype response is known. The test filters that were
constructed with parallel-coup:led lines bore out this feature in general,
even for frequencies in the second stop-band (three times the frequency
of the first stop-band).

Certain additional parallel-coupled types of filter sections other

t-han those investigated under th~is contract appear to be applicable to
the exact design of band-stop filters and would therefore be suitable for
further study along these lines.

C. MULTIPLEXEBS

The measured performance of the three-channel comb-line multiplexer

having contiguous channels compares well with the design prototype. Band-
width spreading in-the filters of the multiplexer caused the multiplexer

113

.~..-.--~---.--.------
pass-band to be larger than the design prototype. It is recommended that
in future designs the percentage bandwidth be understated to account for
bandwidth spreading. Tuning of the multiplexer is facilitated by using
an electronically swept frequency oscillator. Without the sweeper, tuning
is difficult and tedious.

The design principles for multiplexers with contiguous channels may

be extended to multiplexers with guard-bands between channels. However,
cancelling the susceptance in the pass-bands of multiplexers with guard-
bands by using a suaceptance-annulling network in the form of an open- or
short-circuited transmissicn line may be difficult to'do very completely.
However, susceptance cancellation can be obtained whinh is sufficiently
good for most practical purposes.

D. INTERDIGITAL FILTERS WITH.CAPACITIVE LOADING

Because of the complicated nature of capacitively loaded interdigital

filters, the use of an approximate filter model was necessary in order to
represent the filter structure in a form simple enough to permit practical
design. That the model used is reasonably valid is demonstrated by the
fact that an analogous model used for interdigital filters without capac-
itive loading gave good agreement between theory and experiment. By
working out trial designs and computing the responses of the corresponding
approximate models, it has been demonstrated that a practical design pro-
cedure for capacitively loaded interdigital filters with wide bandwidths.
has been obtained. The use of capacitive loading is seen to have the
useful features of permitting an even smaller filter'structure, and of
causing the first spurious pass-band to be unusually far removed from the
main pass-band. The capacitive loading feature may also be of interest
with regard to the design of devices such as traveling-wave parametric
amplifiers where the capacitive loading elements would consist ofvariable-
capacitance diodes. An experimental model of capacitively loaded inter-
digital filter for laboratory tests is now being worked on.

114
 ACKNOWLEDGMENTS

Mr. York Sato worked out many of the details of the mechanical
design of the two magnetically tunable filters discussed in Sec. III,
and made the laboratory tests on the filters.

Mr. H. B. Larrick and Mr. P. H. Heznek made the laboratory tests

on the band-atop filters discussed in Sec. IV.

Mr. P. R. Reznek made the laboratory.tests on the three-channel

multiplexer discussed in Sec. V.

The programming of the Burroughs 220 computer for determining the

responses of trial filter designs such as those discussed in Sec. VI
was done by Mr. V. H. Sagherian.

115
 IDENTIFICATION OF KEY TEClINICAL PERSONNEL

HOURS CHARGED
TO PROJECT
DMIING QUARTER

DR. P. S. CARTER, JR. 30.5

Senior Research Engineer

DR. E. G. CRISTAL 172

Research Engineer

DR. G. L. MATTHAEI 386

Manager, Electromagnetic Techniques Laboratory
(Project Leader)

MR. L. A. RoBeIsoN 166

Senior Research Engineer

MR. B. M. SCHIrFMAN 273

Senior Research Engineer

DR. LEo YOUNG 164

Senior Research Engineer

In addition, Stanford Research Institute and the authors have contributed

additional work time toward the microwave filter book being written on
this project.

116
 DISTRIBUrION LIST

NO. or NO. or
ORGANIZATION COPIES ORGANIZATION COPIE
OASD (IME) Corps of Engineers Liaison Office
AMTT: Technical Library U.S. Army Electronics Research and-
Room 3E1065 Development Laboratory
The Pentagon Fort Monmouth, New Jersey'
Washington 25, D.C."
AFSC Scientific/Technical Liaison
Chief of Nesearch and Development I Office
OCS, Department of the Army' U.S. Naval Air Development Center
Washington 25, D.C. Johnsville, Pennsylvania
Commanding General USAELHD Liaison Office
U.S. Army Materiel Cmemasnd Rome Air Development Center
ATIN: R&D Directorate ATIN: IW)L
Washington 25, DC. Griffiss Air Force Base, New York
Commanding General 1 Commending Officer
U.S. Army Electronics Command U.S. Army Electronics Material
ATIN: AMSEL-AD Support Agency
Fort Monmouth, New Jersey ATTN: SEIMS-AI)
Fort Monmouth, New Jersey
Director, U.S. Naval Research I
Laboratory Marine Corps Liaison Office
ATIN: Code 2027 U.S. Army Electronics Research and
Washington 25, D.C. Development Laboratory
ATIN: S•ILRA/!LN R
Commander, Aeronautical Systems I Fort Monmouth, New Jersey
Division
ATIN: ASAPHL Commanding Officer
Wright-Patterson Air Force Baas, Ohio U.S. Army Electronics Research and
SDeveloment Laboratory
H Electronic Systems DiviioATIN: Director of Research or
L.N: HSco Engineering
L.G. Haenscom Field Fort Monmouth, New Jersey
Bedford, Massachusetts
Commanding Officer
Commander, Air Force Cambridge 1 U.S. Army Electronics Research and
Research Laboratories Development Laboratory
ATIN: CHO ATI" : Technical Documents Center
L.G. Hanscom Field Fort Monmouth, New Jersey
Bedford, Massachusetts
Coimmanding Officer
Commander, Air Force Command & U.S. Army Electronics Research and
Control Development Division Development Laboratory
ATIN: C11C ATTN: SEIWA/ADJ (FU No. 1)
L.G. Hanscom Field Fort Monmouth, New Jersey
-Bedford, Massachusetts
Advisory Group on Electron
Devices 2
Commander, Rome Air Development Center 1 346 Broadway
ATIN: RAALD New York 13, New York
Griffisa Air Force Base, New York
Commanding Officer 3
Commander, Armed Services Technical 10 U.S. Army Electronics Research and
Information Agency Development Laborstory
ATIN: TISIA ATIN: SLIDA/TNH
Arlington Hall Station Fort Monmouth, New Jersey
Arlington 12, Virginia ('FOR RETRANSmITTAL TO ACCREDITED BRITISH
AND CANADIAN GOVERNMENT REPRESENTATIVES)
Chief, U.S. Army Security Agency 2
Arlington Hall Station Commanding General
Arlington 12, Virginia U.S. Ar Cobat Developments Command
ATTN : C&};-E
Deputy President I Fort Belvoir, Virginia
US. Army Security Agency Board
Arlington Hall Station Commanding Officer
Arlington 12, Virginia U.S. Army Communications-Eleetronics
Combat Development Agency
Commanding Officer I Fort Huachuca, Arisons
Harry Diamond Laboratories
ATIN: Library, Room 211. Building 92
Washington 25, D.C.

/
 DISTRIBUTION LIST

NO. OF NO. OF
ORGANIZATION COPIts ORGANIZATION COPIIn
Director, Fort Monmouth Office 1 General Electric Microwave I
U.S. Army Communications-Electronics Laboratory
Combat Development Agency ATIN: Technical Library
Building 410 601 California Avenue
Fort Monmouth, New Jersey Palo Alto, California
AFSC Scientific/Technical Liaison I Dr. K. L. Kotzebue
Office Watkins-Johnson Company
U.S. Army Electronics Research and 3333 Hillview Avenue
Development Laboratory Stanford Industrial Park
Fort Monmouth, New Jersey Palo Alto, California
Commanding Officer and Director 1 Dr. Griemmuman
U.S. Navy Electronics Laboratory Polytechnic Institute of Brooklyn
San Diego 52, California Route 110
ELAIBS Library I Farmingdale, Long Island; New York
Stanford Industrial Park National Bureau of Standards 1
3300 Hillview Avenue Engineering Electronics Section
Palo Alto, California ATIN: Mr. Gustave Shapiro, Chief
Airborne Instruments Laboratory I Washington 25, D.C.
ATIN: Mr. R. Sleven Chief
Deer Park, Long Island, New York Bureau of Ships
ATIN: Mr. Gumina, Code 68182
Commanding Officer Department of the Navy
U.S. Arm Sinal Reearch Unit Washington 25, D.C.
ATIN : SIGIJU-3
Mountain View, California Physical Electronics Laboratory
ATIN: Philip Carter
Stanford Electronics Laboratory 1 2493 Pulgas Avenue
Stanford University E. Palo Alto, California
ATIN: Applied Electronics Laboratory
Stanford, California Loral Electronics Corporation
ATIN: Mr. Magid
Convair 1 825 Bronx River Drive
Pomona, California New York 72, New York
Commander, Pome Air Dovel bmet Catr 1 Commanding Officer
ATIN: WIiM-2, Mr. Patsy A."Pomenelli U.S.A. Electronics Research &
Griffiss Air Force Base Development Laboratory
Rome, New York Fort Monmouth, New Jersey
ATIN: SELItA/SiJ I
Rantec Corporation 1 ATIN: SWIJA/PIU (Mr. Reingold) 1
ATIN: S. B. Cohn, Technical Director ATIN: SELAA/SEA (B. Ellis) 1
Calabasas, California ATIM: SEILA/PE (Dr. E. Both) 1
ATIN: SEUA/PE (Division Director) I
U.S. Naval Ordnance Test Station 1 ATIN: SB/SIE (Mr. I. Dattilo, 10
ATIN: Code 4021, Mr. Hobert Corzies
China Lake, California