Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Synthesis of Time-Var y ing Passive Networks
B. D. 0. Anderson
March 1966
Reprinted October 1966
Technical Report No. 6560-7
This work was entirely supported by the Air Force Office of Scientific Research under Grant AF-AFOSR 337-63 except for report publication under Nonr-225I83)
SVSTEIllS THEORV LRBORRTORV
STsRllFORD ELEbTROnlCS ItRBORRTORIES STRIFORO URIUERSITV STRIFORD, CALIFORRIA
SYNTHESIS OF TIME-VARYING PASSIVE NETWORKS
by
B. D. 0. Anderson
March 1966
Reprinted October 1966
Reproduction i n whole o r i n p a r t i s permitted f o r any purpose of t h e United S t a t e s Government
Technical Report No. 6560-7
This work was e n t i r e l y supported by the A i r Force Office of S c i e n t i f i c Research under Grant AF-AFOSR 337-63 except f o r r epor t publ ica t ion under ~onr-225(83)
Systems Theory ~ a b o r a t o r ~ Stanford Elec t ronics Labora tor ies
Stanford Universi ty Stanford, Ca l i fo rn ia
CONTENTS
I . INTRODUCTION
A. A i m s and Philosophy o f t h e Research
B. Review of Previous Work
C. Review of t h e Research Reported
D. Mathematical Techniques Employed
11. REVIEW, DEFINITIONS AND NOTATION
A. Po r t Variables and t h e Sca t t e r ing Matrix
B. P a s s i v i t y and Losslessness
C. Other Network Matr ices
D. Examples of t h e Preceding Concepts
111. PROPERTIES OF PASSIVE SCATTERING MATRICES
A. Antecedal Property
B. Existence of Adjoint
C. X Mapping Property -2
D. Loss less Condition on s -
IV. FINITE NETWORKS
A. Equivalences
B. Ef fec t of F in i t eness on Matrix Descript ion
C. Existence of t h e Antecedal Adjoint
D. The Concept of Degree of a Sca t t e r ing Matrix
E. Cascade Loading
F. Examples of Cascade Loading
Page - 1
1
1
2
4
V. NONDISSIPATIYE NETWORKS 29
A. Loss less and Nondissipat ive Networks 2Y B. Quas i loss less Sca t t e r ing Matr ices 30
C. Sca t t e r ing Matrices of Nondissipat ive 31- Networks--Quasilossless Condition
iii
D. Scattering Matrices of Nondissipative Networks--Mathematical Form
E. Examples of Quasilossless Scattering Matrices
F. Quasilossless Immittance Matrices
VI. FACl'ORIZATION OF SCATTERING MATRICES
A. The Concept of Factorization
B. Conditions for Factorization
C. Conditions for Factorization with Degree Reduction
VII. NONDISSIPATIVE SYNTHESIS
A. Factorization of Quasilossless Scattering Matrices
B. Realization of a Product of Scattering Matrices
C. Realization of Quasilossless Scattering Matrices of Degree Zero and Degree One
D. Synthesis Example
VIII.MSSY SYNTHESIS
A. Statement of Problem and an Outlined Attack
B. Solution Procedure
C. The Multiport Case
D. Example of One Port Synthesis
IX. RELATION TO OTHER WORK
A. Spaulding's Lossless Immittance Synthesis
B. A New Viewpoint of Lossless Time- Invariant Synthesis
C. Relation to Two Problems of Automatic Control
Page - 34
X. CONCLUSIONS
A. Summary of Preceding Work
B. Directions for Future Research
C. Possible Applications
APPENDIX
Verification of Admittance Matrix Formulation
Page - 79 79 79 80
REFERENCES
ILLUSTRATIONS
Figure Page - 1 I l l u s t r a t i o n of S o l v a b i l i t y 6
2 Network wi th Sca t t e r ing Matrix of t h e 18 Form of Equation (3.13)
3 Time-Varying C i r c u i t Equivalences 20
4 I l l u s t r a t i o n of Cascade Loading 25
5 Representat ion of Network a s Cascade Loading 26
6 A Nondissipat ive Network 3
7 Rea l i za t ion of One Por t Sca t t e r ing Matrix 53 Product
8 Rea l i za t ion of Mult iport Sca t t e r ing Matrix Product
54
9 Rea l i za t ion of Mult iple Product 54
10 Proto type Network f o r Degree One S c a t t e r i n g Matrix
58
11 Pass ive Quas i loss less Sca t t e r ing Matr ix 59 Rea l i za t ion
1 2 Reduction i n t h e Gyrators f o r Product Real iza t ion 60
13 Network Dual t o t h a t of Figure 10 61
14 Example I l l u s t r a t i n g Real iza t ion of Quas i loss less 63 Sca t t e r ing Matrix
15 A Lossy Synthesis 71
16 Equivalent Networks with t h e Transformer 77 Turns Rat io ~ ( t ) = cos a t s i n a t
s i n at -cos a t t I
PRINCIPAL SYMBOLS
Symbol f o r a network
Augmented network
Por t vol tage vec to r
Por t current vec to r
Por t inc ident vol tage
Por t r e f l e c t e d vol tage
Exci ta t ion of augmented network.
Unit matr ix (of dimension n )
Zero matr ix (square, of dimension n)
Impedance matr ix
Shunt augmented network impedance matr ix
Admittance matr ix
S e r i e s augmented network admittance matr ix
Sca t t e r ing Matrix
Adjoint of s -
Antecedal ad jo in t of s - +m
Short f o r -m - s( t77)vi (7)d7 - +m
Short f o r J-. zl(t ,A) s 2 ( A , ~ ) d 3
Transpose i n t h e matr ix sense of X - Set membership and i n t e r s e c t i o n symbols
Set of infinitely differentiable vector functions of compact support
Set of infinitely differentiable vector functions with support bounded on the left
Set of vector distributions wlth support bounded on the right
Set of square integrable vector functions.
Unit impulse
Derivative of unit impulse
Unit step function, zero for t< 7, unity for t >T.
Typical vectors
Typical matrices
Typical scalars t
Short for T ( k ) g ( h ) d h -m -
Same as < f, f >t - -
Operator norm.
Energy input up till time t.
Nonnegative definite matrix kernels.
Matrices with elements polynomial in derivative operator.
viii
ACKNOWLEDGMENTS
I t is a most p leasant t a sk t o acknowledge t h e generous Support
provided by t h e Services Canteens Trust, Fund, an agency of t h e
Aust ra l ian Government, over t h e pas t eighteen months, and a l s o t h e
United S t a t e s Educational Eoundation i n Aus t ra l i a f o r a Fulbr ight
Grant. Support by t h e A i r Force Of f i ce of S c i e n t i f i c Research under
Grant A F - A F O S R - ~ ~ T - ~ ~ i s a l s o g r a t e f u l l y acknowledged.
The author would a l s o l i k e t o express i n t h e s t ronges t poss ib le
terms h i s apprec ia t ion and g r a t i t u d e t o Professor Robert W. Newcomb,
who throughout t h e course of t h e research cons tant ly provided ex-
cept ional ly cons t ruc t ive guidance.
The i n t e r e s t displayed by Professors Thomas Cover and John Moll
i n t h e i r reading of t h e manuscript i s a l s o appreciated.
1 I. 1NTIU)DUCTION
A. AINS AND PHILOSOPHY OF THE RESEARCH
Huch of engineering i s concerned with the replacement of physical
e n t i t i e s by models, upon which calculations may be performed; of ten the
aim i s t o interpret physical conditions of the physical en t i ty a s mathe-
matical conditions on some part of a mathematical model, o r vice versa.
This i s precisely the aim here. The physical e n t i t i e s that we consider
are time-varying e l ec t r i ca l networks, which we shal l often assume t o be
imbued with physical properties such a s l inear i ty , passivity, and f i n i t e -
ness, a l l properties well known i n more c lass ica l studies; the mathe-
matical models we consider are pr incipal ly described by network matrices,
such as the s c ~ t t e r i n g matrix o r impedance matrix. We may seek necessary,
suff ic ient , o r necessary and suf f ic ien t conditions on ( for example) a
scat ter ing matrix f o r the associated network to be l inear, o r passive, o r
f i n i t e , o r t o possess any combination of these and other properties; or
we may seek means f o r passing from the network to a matrix description
' (analysis) , and means for passing from a matrix to the network (synthesis) ,
In point of f ac t t h e . l a t t e r i s the principal aim here.
Some of these problems may have well-known answers (e.g. those con-
cerned with l i n e a r i t y ) o r have answers which a re easy to obtain (e.g. those
concerned with reciproci ty) . It is possible t o say for example tha t a net-
work N is l i nea r (and reciprocal i f and only i f i t s scat ter ing matrix s - 1 ' - possesses property a (and p ), where a and @ are two simple mathe-
matical properties [Ref. 11. In t h i s d i sser ta t ion we venturkaway from
the beaten path i n a discussion primarily concerned with networks composed
0% interconnections of possibly time-varying, l inear, nondissipative c i r -
cu i t elements. While section C discusses i n more de ta i l what is attempted,
l e t it be s ta ted a t t h i s stage tha t the object is again to pa i r off physical
and mathematical properties, as much as possible i n a one-to-one fashion.
B. REVIEW OF PREVIOUS WORK
A search of the available l i t e r a t u r e shows a dearth of material con-
sidering time-varying networks from a generalized point of view. Although
Zadeh i n a 1961 review a r t i c l e [Ref. 21 was able t o quote 130 references
concerned wi th time-varying c i r c u i t s i n Some way, only a bare handful of
t h e s e r e fe rences could be described a s applying t o general time-varying
c i r c u i t s , t h a t is, c i r c u i t s where only a minimum number of physical con-
s t r a i n t s were assumed, as opposed t o c i r c u i t s where f o r example s inuso ida l
o r "slow" element v a r i a t i o n s were assumed. Among such we could quote f o r
example two o t h e r a r t i c l e s by Zadeh [Refs. 3>4] and t h a t of Darlington
[ ~ e f . 51, Since 1 9 6 1 t h e s i t u a t i o n has been l i t t l e changed save f o r t h e ap-
pearance of a work by Newcomb [ ~ e f . 61, which l e d on t o f u r t h e r develop-
ments by Newcomb, Spaulding and t h e present author, ac t ing separa te ly and
i n conjunction, e.g. [Ref s. 7-15]. Reference 6 approaches time-varying
network theory axiomatical ly; t h e only p r o p e r t i e s t h a t a r e pos tu la ted
(not necessa r i ly simultaneously) a r e l i n e a r i t y , pas s iv i ty , s o l v a b i l i t y
(equiva lent t o ex is tence of a s c a t t e r i n g matrix, o r t h e p resc r ip t ion of a
unique response t o a c e r t a i n e x c i t a t i o n ) , l o s s l e s sness , rec iproc i ty , and
time-invariance. Prom t h e pos tu l a t ed p rope r t i e s , and these alone, i t is
poss ib l e t o deduce derived p r o p e r t i e s of networks. Of t h e sum t o t a l of
Newcomb and Spaulding 's work, Spaulding's d i s s e r t a t i o n [ ~ e f . 91 probably
is most successfu l i n pursuing a given l i n e of deduction. Af ter an i n i t i a l
cons idera t ion of s c a t t e r i n g and immittance d e s c r i p t i o n s of t h e b a s i c bui ld ing
blocks of networks, i . e . r e s i s t o r s , capac i tors , inductors , gy ra to r s and
mul t ipor t t ransformers, Spaulding deduces t h e mathematical form of t h e im-
mit tance of a l o s s l e s s network, and shows how t o synthes ize a network given
an immittance of t h i s form.
C. REVIEW OF THE RESEARCH REPORTED
The p r i n c i p a l but not t h e s o l e r e s u l t presented he re is concerned
wi th networks composed of nondiss ipa t ive elements. Somewhat paradoxical ly,
such networks a r e not necessa r i ly l o s s l e s s ( i n a sense t o be described) ,
and t h e r e s u l t s on l o s s l e s s s c a t t e r i n g ma t r i ces of Spaulding a r e not d i r -
e c t l y appl icable . A new mathematical condi t ion on t h e s c a t t e r i n g m a t r i x o f such
a network i s described (chapter v), and t h e converse problem i s solved, t h a t of
synthes iz ing a network given a s c a t t e r i n g mat r ix s a t i s f y i n g t h e condition
(chapters V I , V I I ) .
A s a note of caut ion t o t h e reader, we remark t h a t except i n t h e most
important cases f u l l proofs of t h e der iva t ion of these r e s u l t s and o the r s
may not be given here. The d e s i r e f o r brevi ty, t h e a v a i l a b i l i t y of these
proofs i n referenced papers of t h e author, and hopeful ly heightened con-
t i n u i t y of t h e t e x t through t h e use of o u t l i n e proofs a r e our apology f o r
a l l omissions.
I n chapter I11 a s a pre lude t o t h e main mater ia l , we apply methods
of func t iona l analys is t o t h e s tudy of passive s c a t t e r i n g matr ices t o
conclude new and apparently fundamental proper t ies . S imi lar mathematical
techniques y ie ld o t h e r r e s u l t s , most of which a r e d e t a i l e d i n [Ref.l5],
which we draw on t o g ive new i n s i g h t i n t o a e cons t r a in t of loss lessness ,
(chapter 111) and t h e d e s c r i p t i o n of t h e interconnect ion of networks
(termed cascade loading)(chapter IV) . There is a l s o t o be found i n chap-
ter I V a discussion of network equivalences, cu l l ed from Anderson, Spaulding
and Newcomb, [Ref. 81 which i s important i n the sequel.
Following t h e nondiss ipa t ive synthesis , t h e syn thes i s of lo s sy sca t -
t e r i n g matr ices i s considered. While not i n the n i c e form of t h e nondis-
s i p a t i v e synthesis ,
whose so lu t ion , i f it e x i s t s , immediately y ie lds a syn thes i s of a lossy
network s c a t t e r i n g matr ix. (chapter VI I I ) .
Lest system t h e o r i s t s f e e l neglected, we hasten t o mention t h a t a
time-varying equivalent of t h e degree of a matr ix of r a t i o n a l t r a n s f e r
funct ions i s discussed i n chapter IV, while i n chapter I X t h e r e i s mention
of two con t ro l systems r e s u l t s achieved by Newcomb and Anderson [Refs. 16,171 which depend on t h e s o r t of ana lys i s reported here.
Natural ly any time-varying synthes is technique when applied i n t h e
t ime-invariant case should s t i l l go through; the consequences of t h i s a r e
examined i n chapter IX, where r e s u l t s without t h e a s soc ia t ed mathematical
d e t a i l s a r e mentioned. A s t hese r e s u l t s a r e s t r i c t l y pe r iphe ra l t o t h e
main ma te r i a l of t h i s d i s s e r t a t i o n , we do not appologize f o r t h e omission
of purely technica l ma te r i a l .
While it should be c l e a r from t h e t e x t a s t o what ma te r i a l is new, we
poin t ou t here t h a t most of chapter I1 i s not new, t h a t which i s new being
merely r e i n t e r p r e t a t i o n of o l d r e s u l t s . A similar remark can be made con-
cerning sec t ions A and D of chapter I11 and s e c t i o n s B, C, and F of
chapter I V . Other ma te r i a l however was developed by t h e author, though
o f t e n i n conjunction wi th o the r s .
-3-
D. : MATHEMhTICAL TECHNIQUES W U ) Y E D
I t i s probably appropr ia te a t t h i s s t age t o mention t h e s o r t of
mathematics t h a t i s used. Di s t r ibu t ion Theory a s developed by L.
Schwartz [Refs. 18, 1 , 0 i s required t o formulate theorems on net-
works which a r e not necessa r i ly f i n i t e , though f o r f i n i t e networks
heavier demand i s probably placed upon l i n e a r d i f f e r e n t i a l equation
theory.
Functional analys is , i n p a r t i c u l a r r e s u l t s from t h e theory of
Hi lbe r t space, is used t o i l luminate severa l poin ts , a s t h e mathe-
mat ics here enables one t o see "what is r e a l l y going on."
Final ly, t h e no ta t ion and ideas of networks and system theory,
a s developed by many e l e c t r i c a l engineers who found t h e s tandard
mathematics i n s u f f i c i e n t o r unsuitable, is drawn upon from time t o
t i m e .
11. REVIEW, DEFINITIONS AND NOTATION
Most of the notation t o be used is described i n t h i s chapter; a
f u l l l i s t of symbols however appears before chapter I .
A. PORT VARIABLES AND THE SCATTERING MATRIX
For greater de ta i l , Reference 6 should be consulted. A l inear
n-port network N can be described through the allowable port voltage - and current vectors v and , i which are n-nectors i n r9 i . e . , 1 - - -+ ' and i are vector functions which are in f in i te ly dif ferent iable and - zero u n t i l some f i n i t e time. The f u l l reasoning behind the mathematical
r e s t r i c t i ons on v and i is discussed i n Reference 6, but i n brief - - these r e s t r i c t i ons are associated with the fac t that a l l excitations of
a physical system must have commenced a t some f i n i t e time, and only very
unusual physical systems exhibit discontinuit ies i n t he i r responses.
Given port vectors v and i i t i s then possible t o define inci- - - i r
dent and ref lected voltage vectors v and v where - -
A scat ter ing matrix is an n x n matrix of dis t r ibut ions i n two vari- i
ables [Refs. 18, 19, 201, o r kernel, - s(t,'r), mapping arbi t rary - v r
in to v 6 _+ - fi through
o r i n br ie fe r notation,
r i v = s o v - - -
Linear networks possessing a scat ter ing matrix are termed solvable
[Ref. 61. This is because t h e existence of a scattering matrix for - N
i s equivalent t o the series-augmented network N formed from N i n -a
Fig. 1 possessing a unique current response i t o an arbi t rary ex- - c i t a t i on e e &, see Anderson and Newcomb [ ~ e f . 141. Plainly solvabi l i ty - w i l l be a property normally possessed by real-world networks
FIGURE 1
ILLUSTRATION OF SOLVABILITY;
1 denotes n un i t res i s tors ; -n
ports of N a r e not shown separately. -
In addition to the notation of Eq.(2.2b) whichis shorthand f o r an
operator acting on a function, it is convenient to define a notation for
a composite operator defined by l e t t i n g one operator act on another. For
example i f s ( t , ~ ) and s 2 ( t , 7 ) a r e two kernels which are matrices of -1 dis t r ibu t ions i n two variables t and T and which map $ in to g+, then we may define s ( t , ~ ) , the composition of s and s by
-3 -1 -2'
r i v = s . v = s . ( s 2 e x i ) - -3 - -1 (2.3)
i . Because s maps g+ into $I s v is i n the domain of s and in -2 3' -2 - r -1 '
f ac t we w i l l have v - e $ as a resu l t .
For convenience we write
o r (as may be. logical ly ver i f ied from Eq. (2.3))
A s is discussed i n References 15 and 20, the composition of more than
two operators i s not necessari ly associative. In the case where the op-
erators map. &! in to &! however, associat ivi ty does apply. 3 4
We remark tha t when s and s are scat ter ing matrices we w i l l -1 -2
of ten term s the product of s and s and say tha t s possesses -3 -1 -2' -3
the factorization s o s -1 -2.
B. PASSIVITY AND MSSLESSNESS
We introduce the notation f o r two n-vector functions of time f and - g -
and
assuming the in tegra l s always ex is t . The superscript t i l d e denotes matrix
transposition.
Suppose now f and g are taken as the port voltage vector v and - - - simultaneous current vector i of a network N, with 1, 5 assumed to - be i n &&. Thus we form
which has the physical meaning of the energy input t o the network up till
time t. Reference 6 defines N as passive i f t h i s quantity i s non- - negative for a l l possible choices of simultaneous pa i r s v and 2, and - times t . An a l te rna t ive expression for ~q' ,(2.6a), following direct ly
from ~ q . (2.1), i s
i Suppose f o r the moment we r e s t r i c t N t o being passive. I f v can - -
i be taken a s square-integrable, i . e . v e B n g then &(a) i s wel l - 4 -2'
i defined, being nonnegative and bounded above by / v i f . It f u r t h e r -
r fo l lows from Eq. (2.6b) and t h e nonnegativity of &(a) t h a t v - e $2, and thus t h e por t vol tage and current vec tors v and i a r e a l s o - -
i r square integrable, being l i n e a r combinations of v and v by - - ~ q . (2.1).
Lossless N [Ref.6] a r e now defined a s those N which a r e - - pass ive and solvable, and have one addi t ional property, v i z . t h a t f o r
i every v g+ nJ2, &(m) = 0 . The physical motivation f o r t h i s re- - quirement i s t h a t t h e energy de l ivered i n t o the c i r c u i t up t i ll t = m
should be zero, provided t h a t t h e exc i t a t ions f a l l t o zero f a s t enough
a s t approaches i n f i n i t y . We comment t h a t p a s s i v i t y impl ies t h a t i
~ ( m ) is f i n i t e , while s o l v a b i l i t y permits the choice of any v - r € 5
with t h e knowledge t h a t t h e r e w i l l be a corresponding v . - To c lose t h i s sec t ion we remark t h a t some of t h e c o n s t r a i n t s
imposed by p a s s i v i t y and los s l e s sness on s c a t t e r i n g ma t r i ces w i l l be
discussed i n t h e next chapter .
C. OTHER NETWORK MATRICES
We excuse t h e f a c t t h a t our t h e o r e t i c a l d iscuss ions a r e normally
i n terms of the l e s s known s c a t t e r i n g matrix r a t h e r than t h e b e t t e r
known immittance matr ices on t h e grounds t h a t t h e l a t t e r have more
r e s t r i c t e d a p p l i c a b i l i t y . None t h e l e s s the r e l a t i o n of s c a t t e r i n g
t o immittance matr ices should be understood i f f o r no o t h e r reason than
t h a t t h e l a t t e r w i l l be en te r ing i n t o subsequent d iscuss ions on occasions.
The time va r i ab le impedance z( t ,T) and admittance y(t ,T) of a net- - - work N a r e defined i n a completely na tu ra l way. I f v and i a r e - - - p o r t vol tage and current vectors , then
and +m
i ( t ) = J Y ( ~ , T ) v ( T ) ~ T = y . V - - - - - -m
The mere inscr ipt ion of Eq.(2.7) of course does not guarantee the I
existence of t h e relevant matrices; though we cap usually, but not al-
ways, expect a scat ter ing matrix t o exis t , thJs i 6 not the case for
immittance matrices. A transformer f o r example has neither an impedance
o r an admittance matrix.
Equations ( 2 . 1 , (2.2) and (2.7) allow deduction of the inter-
re la t ions between s, ? and y, assuming existence. We have -
Here 61 i s shorthand f o r the un i t matrix multiplying the impulse -n
6(t-?); thus 6 1 @ w = w for any n-vector w. -n - - - Two other network matrices are sometimes of interest , the augmented
admittance matrix y ( t , ~ ) and augmented impedance matrix z' ( t , ~ ) . The -A -A
network of Fig.1, consisting of N with a un i t res i s tor i n se r ies with - each port, has admittance y , and N with a uni t r e s i s to r i n pa ra l l e l
-A - with each por t has impedance z' Reference 14 shows tha t the existence
-A' of an; one of y z ' and s guarantees the existence of the other two,
-A' -A - and tha t t he following equations r e l a t e the matrices:
D. EXAMPLES OF THE PRECEDING CONCEITS
To illustrate some of the preceding concepts we shall consider the
description by network matrices of the time-varying inductor, which has
the differential equation description
Here q t ) is the instantaneous inductance. We observe immediately from
Eq. (2.10) that
(where the prime denotes differentiation) and thus
From Eq.(2.10) it also follows that
t
(where u(t-7) is the unit step function, 0 for tq, 1 for t%)
Thus:
i r By substituting 2v = v+i and 2v = v-i in Eq.(2.10) the dif-
ferential equation relating the incident and reflected voltages is
found to be
(where the superscript dot denotes differentiation)
This f i r s t order equation is readily solvable, and has the impulse re-
sponse. ' [~ef. 131
The augmented admittance could .be formed from s(t, T ) using Eq. (2.9)
o r by writing down the d i f f e r en t i a l equation re la t ing e and i for the
network of Fig.1. One obtains
The impedance z l ( t ,T) i s determined the same way, &uta t ig routandfs, or A
from 2': - 61-y see Eq. (2.9). . A - A '
111. PROPERTIES OF PASSIVE SCATTERING MATRICES
In t h i s chapter we examine t h e cons t ra in t s on a sca t t e r ing matrix
imposed by p a s s i v i t y and loss lessness .
A. ANTECEDAL PROPERTY
Denoting t h e n x n zero matrix by 0 w e can show f o r a passive -n3 network
s ( t , 7 ) = Cn - f o r t < T (3.1)
I n essence t h i s equation s t a t e s t h a t N i s i n some sense causal, which - i s a property t h a t i s f a m i l i a r i n real-world systems. More prec ise ly s - is antecedal [ ~ e f .21]. But although E ~ . (3.1) might indeed be assumed on
physical grounds, it i n f a c t follows, a s i s now shown, from our e a r l i e r
de f in i t ion of pass iv i ty . i
Let to be f ixed and take - v ( t ) = 0 f o r tito. Equation (2.6b) - . shows - v r ( t ) = - 0 f o r t<to s i n c e e(t)zO by assumption. Therefore,
with an obvious extension of t h e < > nota t ion t o cope with double in te -
g r a l s
n n
whenever y(t)=O - f o r t>t and - X ( T ) = ~ f o r ~ < t with 9, if e & the s e t 0 0'
of i n f i n i t e l y d i f f e r e n t i a b l e functions zero ou t s ide some i n t e r v a l I.
A s t h i s equation i s t r u e i r r e s p e c t i v e of and - x f o r tCto<7, equation
(3.1) follows.[Ref. 8 p.26, p.1081.
B. EXISTENCE OF AD3OINT a .
Associated wi th s t h e r e is defined an ad jo in t transformation, 5 , - through
f o r a l l 5, 2 s. A s i s discussed i n Ref. 17, it follows read i ly from
-12-
a t h i s that s is rela ted simply to s through - -
where, it w i l l be recalled, the t i l d e denotes matrix transposition. It a is not t rue tha t s maps A+ in to &9 but merely, see Ref. 20, that -
a 4' s maps 5 - in to A where A i s the s e t of dis t r ibut ions zero
-4 ' 4 af te r some f i n i t e time. ( In fact , i f the direction of time flow is re-
a a versed, s is antecedal) I t follows tha t a product of the form s o s - - - is not a p r i o r i defined, see section A of chapter I1 o r Ref. 15, and
a a certainly tha t one may not a p r io r i wri te s o (s o s) = s o s ) o s . - - (- - -
I t should also be noted that the inner product on the r ight s ide of
(3.2) is well defined, as it is essent ia l ly an inner product of a d i s t r i - a
bution, o cp, and a tes t ing function, 5. [Ref. 18, p.241 This use - - of the inner product notation w i l l appear i n future pages also.
C. - g2 MAPPING PRaPERTY
The most fundamental property of s is associated with the f ac t that - i
for a passive network, a square integrable v is mapped into a square - r
integrable v . From Eqs. (2.6) and the passivi ty of N - -
i Choosing v c n g2 and l e t t i n g t = m shows tha t s is a bounded l i nea r - - continuous transformation on a subset of g v iz &I+ n ; we can there- - 2' fore make an extension [Ref. 22, p.2981 defining - s i n a bounded manner
f o r a l l vi - =J2. The def ini t ion of the adjoint v i a the inner product a
of Eq.(3.2) means tha t s w i l l be the adjoint of s when s i s re- - - - garded as an operator mapping the Hilbert space g2 into i t s e l f ; thus -
a s i s a lso a bounded l i nea r continuous transformation on S Then noting - 2'
tha t s and sa as .& operators have the same operator norm [ ~ e f .22, - - p.2011, E ~ . (3.4) yields
Equation (3.5) is a necessary condition for the passivity of N; - essen t ia l ly it is also suff ic ient i f N i s l inear and solvable: -
Theorem 3.1. A l inear solvable network N is passive i f - and only i f (1) A scat ter ing matrix s ex is t s mapping v
i - - € % in to
vr c 5 and - (2) s maps 22 in to g 2 and
(3) z(t,r) = gn for t < ~ and
(4) Ik114 1.
Proof: The only i f portion has been proven by the reasoning leading
t o Eq. (3.5). To show the i f portion we f i r s t observe that i f conditions
(I), (2) and (3) are not s a t i s f i ed then N must f a i l t o be e i ther l inear - o r solvable o r passive.
A s we are assuming N l inear and solvable we are led t o consider the - existence of an s sat isfying 1 but which is not passive. Then -
i there is some port voltage Y and current i associated with v =
-1 -1 -1 $ (zl + i ) 6 6 and some f i n i t e constant T such that < v i > = 8 ( T ) < 0. -1 4 -1' -1 T 1
i Let a second excitation v 8 be defined as
-2 4
i . f o r a rb i t r a r i l y small C- 0 (v is defined i n an in f in i te ly dif ferent iable
-2 i manner i n T < t < T + E ) . Then 2v2 = v + i . by solvabi l i ty v - v
-2 -2' -2 - -1 and i2 = Al f o r t S T. Then -
where Y ( G ) can be made a rb i t r a r i l y small by properly choosing E since i r
E ( t ) i s evaluated as an integral . But x2eL2 and thus also v2€Z2 by
1 1 1 1 . Equation (3.6b) then shows, by llsll < 1,
Choosing 0 < y(c) < - E1(T) shows that the assumption of E ~ ( T ) < 0 i s
violated, and hence N must be passive. - Q. E. D.
A negative r e s i s t o r of -2 ohms i s act ive under usual definit ions and
should be so under the above definit ions. As may be easi ly found, it has
s = 3S(t-7). Conditions (l), (2) and (3) of the theorem are sat isf ied, but - condition (4) f a i l s as 1k11 = 3.
We remark tha t it i s now possible t o wri te down a well defined operator a
s o 2, provided i ts domain i s res t r ic ted t o S, functions. For i f x E Z2, - -2 - a a s Q x E X2, and (s o s ) 0 x = s 0 ( s a x) i s then well defined. - - - - - - - - -
Any 2 sat isfying the conditions of Theorem 3.1 i s termed passive;
as an easy consequence of the theorem we have
Theorem 3.2. I f and s are passive then s o s i s passive. -2 -1 -2
Proof: (Outline only) The f i r s t three conditions of theorem 3.1 are easi ly shown t o be sa t i s f ied . For the fourth, we have lkl o s211 lklll 1k211 < 1. For detai ls , see Ref. 15.
We sha l l l a t e r show how a network with scat ter ing matrix 2l o s2 can be constructed from networks of scat ter ing matrices s and s2. -1 Scattering matrix products and factor izat ions w i l l be of considerable
importance f o r the nondissipative synthesis of chapter V I I .
A convenient a l ternat ive formulation of the passivi ty c r i te r ion can
now be given. By definit ion, a rea l dis t r ibut ional kernel - k(a,p) w
which i s self-adjoint , ~ ( p , a) = - k (a, p), i s cal led nonnegative def in i te
writ ten k 2 0, i f f o r a l l x E 8 - - -
Considering Eq.(2.6b), w e can write
The in t roduct ion of u ( t ) allows replacement of t h e i n t e g r a l s over
(-a, t$ by i n t e g r a l s over ( -m, a). Manipulations d e t a i l e d i n Ref.15
then show t h a t we can w r i t e
where
Q (a,@) = u ( t - a ) 6 ( m @ ) l - %,a) 0 {u(t-X)s(h,@)I (3.9d) -t -n -
and Q is a s e l f ad jo in t bounded operator , mapping J2 i n t o J2. -t
Theorem 3.3. A l i n e a r so lvable network N is pass ive i f and only i f condit ions (I), ( 2 ) and (3) of thGorem 3.1 a r e s a t i s f i e d and f o r a l l f i n i t e t
Corollary: By (4) of theorem 3 .1 t h e following version of ( 4 ' ) i s s u f f i c i e n t :
One advantage of t h i s formulation i s t h a t it shows t h e na ture of re-
s u l t s when z o r y, Eqs. (2.7) , e x i s t . Thus, using ~ ( t ) = < v , i > - - - - t = < i , v >t, t h e manipulations of Eq.s (3.9), and a r e s u l t s i m i l a r t o Eq.(3.3), -- we s e e t h a t an equivalent statement of condi t ion ( 4 ' ) is: f o r each t t h e
form
I n genera l R t w i l l not be a bounded opera tor on J i n con t ra s t t o Q . -2' -t
For example, i f z i s chosen t o be u(t-T), corresponding t o a 1 fa rad - capaci tor , R ( q p ) = ~ ( t - a ) u ( a - p ) + ~ ( t - p ) u ( p - a ) . Because of t h e i n t e r - t g r a t i o n s involved, it w i l l be impossible f o r ((R ( c I , ~ ) c i (p) l I t o be bounded
t by ~ l l i ( p ) I I f o r a f ixed M i n t h i s example.
D. LOSSLESS CONDITION ON s - Suppose t h a t N i s a l i n e a r , solvable, passive network which i s loss- -
i l e s s . Then &(m) is zero f o r any inc ident vol tage v - E 5 n L ~ , and thus i n t h e language of t h e previous sect ion, from Eq.(3.9c)
from which it read i ly fol lows ( s e e Ref. 15) by applying known r e s u l t s
from t h e theory of Hi lber t spaces [Ref. 23, p.2671 t h a t
Q = o -Q) -
a Observing Eq. (3.9d) we see t h a t C&= 6Ln - o S. Thus w e have -
Theorem 3.4. A pass ive s i s l o s s l e s s i f and only i f -
a s 0 s = 6 1 - - a (3.12)
An important example of a l o s s l e s s s c a t t e r i n g matr ix is
where cp(t) i s square in t eg rab le over [c,co) f o r a l l f i n i t e c. The
network t o which t h i s s c a t t e r i n g matr ix corresponds ac tua l ly c o n s i s t s 1
Of a t ransformer with time-varying t u r n s - r a t i o ~ ( t )
which i s loaded at i ts secondary i n a u n i t capaci tor , see Fig.2.
FIGURE 2
NETWORK WITH SCATTERINGMATRIX OF THE FORM OF EQUATION (3.13)
That ~ q . (3.12) is i n f ac t s a t i s f i ed by the expression of Eq. (3.13) is
not hard t o verify.
Though t h i s network w i l l be loss less for any cp square integrable a on [c,m) for a l l f i n i t e c, i . e . s o s = 61 , it w i l l not necessarily
a be the case that a t the same time s o s = 61. In f ac t d i rec t cal-
culation shows tha t a necessary and suff ic ient additional condition for a
s o s = 6 1 i s that cp not be square integrable on (-m,m). I t should thus - a
be noted that s i n Eq.(3.12) is not necessarily a r ight inverse of - s i n general. -
We also point out that by s ta r t ing with ~ q . ( 3 . 1 0 ) i t i s not possible
t o conclude that %= 0. Hilbert space theory cannot be applied here, - and indeed such a resu l t would clearly be wrong, since R (a,'$) =?(a,@) -a
f o r W. > p, and there are c lear ly non-zero loss less impedances, e.g.
z=u(t-T), a 1 farad capacitor. Although Spaulding [ ~ e f .g] derives the
loss less constraint on s, the importance of the d: mapping properties - -2 i s not made clear ; the preceding remark should therefore emphasize t h i s
importance.
The re la t ion (3.12) can be likened t o a similar resu l t of time-invar- - i an t theory, vie _S(-p)S(p) - = _1 (where S(p) i s the Laplace-transformed
n - s ( t ) ) . For time-invariant systems, the composition becomes convolution, - which i n turn becomes multiplication on taking Laplace transforms.
IV. FINITE NETWORKS
A. EQUIVaENCES
Fini te networks a r e those networks consisting of an inter-
connection of a f i n i t e number of res is tors , capacitors, inductors,
gyrators and (multiport) transformers, any of which may be time-
variable. For l a t e r calculations, it i s convenient t o replace
certain time-variable elements by equivalent c i r c u i t s where the
only time-variable elements appearing i n these equivalent c i rcu i t s
are time-variable transformers.
A time-varying r e s i s to r r ( t ) has an equivalence of the form
[ ~ e f . 81 shown i n Fig. 3(a).
The gyrator, inductor and capacitor have the equivalences shown i n
Pig, 3(b), (c) , and (d), provided the i r time var ia t ion i s "sufficiently"
smooth.
Several features a r e common to these real izat ions . F i r s t , a l l time-
variation i s i n the transformers, Second, the transformer turns r a t i o s
are functions of r( t) , c ( t ) , etc, as the case may be. For example, i f
r ( t) 2 0,then tl = (r(t), t2 = 0, while i f r ( t ) < 0, tl = 0 and
t2 = J-r(t) . The precise formulae for the t . a re i n Ref. 8. Third, 1 -
i f the time-variable element i s passive, then no ac t ive elements w i l l
appear i n i ts equivalent c i rcu i t [ ~ e f , 81. This means f o r example tha t
i n the case of the inductor, t and t w i l l be zero. 2 4
Consequently we may assume for any f i n i t e network a l l time var ia t ion
c m by the use of c i r cu i t equivalences be included i n transforaners, and
i f the time-varying elements are passive, a l l elements i n the c i rcu i t
equivalence a re passive.
In the passive case, element scattering matrices are known t o
exis t . We have already noted tha t of the inductor, see ~ q . (2.14). That
of the capacitor may be obtained by replacing &(t) by @(t) i n Eq.(2.14),
and changing the sign of so Eqa(2.8b) yie lds quick calculation of S
for the r e s i s t o r and gyrator.
To simulate a time-varying ci rcui t , it is evident tha t the time-
variable transformer is going to require simulation. This problem is
discussed i n Ref. 12, where it is pointed out t ha t variable autotrans-
FIGURE 3
TIME-VARYING CIRCUIT EQUIVALENCES
formers can approximate t h e d e s i r e d behavior. The cascade of two gyra to r s
a l s o behaves l i k e a transformer, and thus i f a t ime-variable gyra to r can be
constructed, a simulation is a v i l a b l e . The idea l behavior is hardest t o
approximate, whatever means a r e used, i n the v i c i n i t y of zero tu rns - ra t io .
This seems t o be because, while leakage inductance i s t o be kept minimal, - t h e mutual inductance must be a s large+%s p o s s i b l 9 yet change s ign a s t h e
t u r n s - r a t i o goes through zero. A t some s t age the mutual inductance must
cease t o be f a r l a r g e r than t h e leakage inductance, and i d e a l behavior is
l o s t . I f t h e time i n t e r v a l over which t h i s occurs is small however, i ts
e f f e c t on t h e c i r c u i t may be n e g l i g i b l e i n some s i t u a t i o n s .
Since t h e e f f e c t i v e list of c i r c u i t elements includes; t ime-invariant
(posi t ive-valued) r e s i s t o r s , capac i to r s and inductors, t ime-invariant
gy ra to r s and time-variable t ransformers, we choose t o descr ibe such net-
works with no r e s i s t o r s a s nondiss ipa t ive . I t is of course known t h a t
t ime-invariant inductors, capac i to r s and gyra tors a r e l o s s l e s s . The time-
v a r i a b l e transformer is a l s o l o s s l e s s [ ~ e f . 123; i f zl, x2, 51, i2 a r e
t h e primary voltage, secondary voltage, primary current , secondary cur- - r e n t , it i s known t h a t vl = 2 v2, A2 = -2 il where N is t h e turns- - r a t i o matrix. Then '; i + i t h e net power i n t o t h e transformer, i s
-1-1 -S2' i d e n t i c a l l y zero.
The reason f o r naming networks containing no r e s i s t o r s nondiss ipa t ive
r a t h e r than l o s s l e s s w i l l become c l e a r a t t h e s t a r t of t h e next chapter .
B. EFFECT OF FINITENESS ON MATRIX DESCRIPTIONS i r
For a f i n i t e n-port network N v and v a r e r e l a t e d by an ordinary -7 - -
d i f f e r e n t i a l equation.
where C and D a r e n x n ma t r i ces whose elements a r e polynomial i n - - t h e d e r i v a t i v e operator p d/dt wi th time-varying c o e f f i c i e n t s . I f
s i s passive, then i t can be shown [Refs. 9, 131 t h a t -
Here A ( t ) is an n x n matrix, and @ and Y a r e n x r matrices, - - - depending on the order of t h e d i f f e r e n t i a l equation (4.1) .
Without explicit mention, we shall henceforth assume that the
element values of N are always infinitely differentiable. This - i
means that an infinitely differentiable v will yield an infinitely - r
differentiable v , and that the entries of A(t), g(t) and y(t) are - - - all infinitely differentiable.
Passivity, as applied to the case of a finite network, requires
inter alia the following:
1. The eigenvalues of A(t) are bounded in modulus by unity. - 2 . The elements of @(t) are functions square integrable on -
LT,~) for all finite T.
3. The elements of y(t) are functions square integrable on - (--,TI for all finite T.
Statement (1) can be shown to follow from Eq.(3.5) while (2) and
(3) are consequences of Eq.(3.1) and the L mapping property of s. -2 - Similar analysis applied to passive immittances of finite net-
works shows they have the form [~efs. 9, 111
The matrix entries are again all infinitely differentiable, with L - and M n x p matrices, (p depending on the order of the defining - differential equation). The matrix T is q x n, q s n. -
One consequence of passivity is that Z f is nonnegative de- -0 -0
finite.
C. EXISTENCE OF THE ANTECEDAL ADJOINT
Given a scattering matrix s(t,'T) we have described the existence - of an adjoint - sa(t,~) in Eq. (3.3):
sa(t,7) = Z(7,t) - -
For s of the form of ~q.(4.2), i.e. -
s(~,T) = a(t)s(t-T) + 2(t)z(~)uft-~) - -
The presence of U(T-t) means that t h i s sa is not antecedal, and the - d i f f i c u l t i e s associate& with including sa i n a composition product - have been discussed. To eliminate t h i s problem and for other reasons,
we define an antecedal adjoint:
a a We note tha t s i s formed from s by replacing U(T-t) with
-a - a a
-u(t-7); t h i s means tha t s and s sa t i s fy the same d i f fe ren t ia l -a -
equation. I f s however i s not determined from a d i f fe ren t ia l - a a
equation it may not be possible to define s (though s can always -a -
be defined) . a
Naturally a f i n i t e z defines a z . For the z of (4.3) one - -a - has the antecedal solution of the d i f fe ren t ia l equation sa t i s f ied by a
Z -
D. THE CONCEPT O F DEGREE O F A SCATTERING MATRIX
The matrices @ and \Y i n the equation f o r 2, (4.2), are n x r ; - n is the number of por ts of the network N, while r is a measure of
the complexity of the network, termed i ts degree. (S t r i c t l y speaking,
the columns of both - @ and 2 are required t o be l inear ly independent on
(-a, m ) . I f t h i s i s not the case they may readi ly be made so. I f rl
< r, and r columns only of @ are l inear ly independent, m ( t ) = ml(t)c 1 - -
where - @ i s n x r C is constant and r x r ; then m ( t ) Y ( T ) = m,(t)Yl(7) 1: - 1 where lyl(t) = l y ( t ) ~ and i s also n x r ).
1
The concept of t h e degree of a t ime-invariant network o r a matrix
of r a t i o n a l t r a n s f e r funct ions has been known i n network theory f o r
some time, e .g. [Refs. a4, 251. We r e c a l l t h a t t h e degree of a passive
s c a t t e r i n g matr ix - s ( ~ ) is t h e minimal number of energy s torage elements
i n a network synthesizing t h e matrix. I f such a s c a t t e r i n g matrix i s
w r i t t e n i n time-domain notat ion, a s i n Eq.(4.2), t h e elements of t h e
@ and matr ices become exponentials, and t h e number of l i n e a r l y - - independent columns over (-a, a) equals t h e degree of t h e matrix.
I t i s the re fo re p leas ing t o note t h a t t h e syn thes i s procedures
given f o r s c a t t e r i n g matr ices of (passiv4,non-dissipative and then
lossy networks l a t e r i n t h i s work a r e minimal degree ones, i n the
sense t h a t t h e number of r e a c t i v e elements used i n t h e synthes is i s
equal t o t h e number of l i n e a r l y independent columns of @ and Y. - - Abstract arguments contained i n ~ e f . 2 6 and discussed i n Ref.27 show
t h a t synthes is wi th fewer r eac t ive elements i s not poss ib le . Ref. -
should a l s o be consulted f o r a f u l l e r discussion of t h e degree concept
i n t h e present context, including its extension t o immittances.
Denoting t h e degree of s by deg [?I, f u r t h e r p a r a l l e l s with t h e - t ime-invariant theory a r e exhib i ted by such r e s u l t s a s deg [s] =
deg [s + ~ ? 5 ] where K i s a time varying matrix, and deg ~ s - l ] = deg [?I, - - - - provided the inverse e x i s t s [ ~ e f . 281.
E. CASCADE LOADING
In order t o o b t a i n somewhat more s p e c i f i c r e s u l t s f o r networks of
considerable importance we t u r n t o a usefu l method of combining networks,
t h a t of cascade loading. I n p a r t i c u l a r w e c a l c u l a t e t h e input s c a t t e r i n g
matr ix s when a c e r t a i n inverse e x i s t s and show t h a t under such a con- - d i t i o n s i s pass ive when t h e subnetworks are . -
Consider an ( n + m)-port N whose v a r i a b l e s a r e p a r t i t i o n e d a s t h e - -i %r -r
por ts , t h a t is, v = [ yi 1, v = [!I, v2] with t h e subsc r ip t s 1 and -c -1' -2 -z 2 respect ive ly denoting n- and m-vectors. Then an m-port 3 i s s a i d
t o cascade load N i f -z i r r i 2 = x2 and 3 =
i r where 3 and 3 a r e incident and r e f l e c t e d vol tages f o r N This 4 connection def ines a new n-port 2, a s i l l u s t r a t e d i n Fig. 4.
- 2k-
FIGURE 4 ILLUSTRATION OF CASCADE LOADING
The incident and reflected voltages of N are -
i i r r v = v a n d v = v - -1 - -1 (4.7b)
subject to the constraints placed on the coupling network N by loading -z it with Partitioning the scattering matrix z of N according to - -C its port partition and defining and s as the scattering matrices - of and N lead to -
r i where Eqs.(4.7) have been used in v = z o . In order to gain
. -z some insight into various of the following manipulations we write these
out fully as
i We wish t o so lve f o r vr i n terms of 1 , t o equate t o ~ ~ . ( 4 . 8 c ) by -
i el iminat ing v Doing t h i s when t h e ind ica ted inverse e x i s t s y i e lds &'
Unfortunately t h e inverse requi red i n (4.10) may not always e x i s t
even though N and 3 a r e pass ive . I n con t ra s t t o t h e time-invar- -x i a n t case [Ref. 291 t h e r e is not always a way around t h i s d i f f i c u l t y .
A transformer with tu rns - ra t io ~ ( t ) with secondary open-circuited does
not have a wel l defined s c a t t e r i n g matr ix i f T ( ~ ) = o f o r some t CRef.151, and exemplif ies t h e d i f f i c u l t y . I n f u t u r e appl ica t ions of these r e s u l t s
we s h a l l normally be assuming exis tence; f o r a discussion of t h e d i f f i -
c u l t i e s involved, see Ref. 15.
One app l i ca t ion of t h e idea of cascade loading i s t o represent any
network N a s a combination of N cons i s t ing of t h e c i r c u i t elements - 4' of N unconnected t o one another, and N a network cons is t ing of con- - -2 necting wires alone, a s shown i n Fig. 5 . The interconnect ion of - any two
networks N and 2 2
can a l s o be viewed a s a cascade loading, [Ref .29]. -1
I n Fig. 5 %would cons is t of N and N2 uncoupled t o one another. -1 -
and Shor ts
FIGURE 5 REPRESENTATION OF NETWORK AS CASCADE LOADING
This technique has been f r u i t f u l l y applied i n t h e examination of
t ime-invariant network problems, [ ~ e f s . 29, 301, and it w i l l be used i n
t h e next chapter i n a considerat ion of nondiss ipa t ive networks.
F. EXAMPLES OF CASCADE MADING
We examine here the e f fec t of connecting an orthogonal transformer
a t the ports of a network 3 with scattering matrix s ( t , ~ ) . An 4 orthogonal transformer i s defined by
where T is an orthogonal matrix, and the subscripts r e f e r to primary - and secondary .ports.
Ref. 6 out l ines the calculation of the scat ter ing matrix of a trans-
former with the turns-ratio matrix ~ ( t ) being an arbi t rary m x n matrix. - The resu l t i s
~ ( t , 7 ) = 6(t-7) -
- -1 (r, +E) (Am - *) L
(4.12)
I f now T is r e s t r i c t ed t o being orthogonal (with m=n), considerable - simplification ensues with
Equation (4.10) now allows calculation of the scat ter ing matrix s ( t , ~ ) - of the transformer loaded a t i t s secondary with the network N The re-
4' su l t is
and i s of some importance. I f T i s not orthogonal then the cascade - loading calculation using (4.10) w i l l c lear ly not be as straightforward.
A s a second example, w e t a k e N t o be an a r b i t r a r y network, and -C -
t o cons is t of m u n i t r e s i s t o r s , one f o r each por t , s o t h a t 2.=
Then i f
d i r e c t app l i ca t ion of t h e cascade loading formula (4.10) y i e l d s
Effec t ive ly , we d e l e t e rows and columns of 2 by matching a t some of - i ts por t s . This technique w i l l be used i n t h e sequel, chapters V I I
and V I I I .
V. NONDISSIPATIVE NETWORKS
A. IDSSLESS AND NONDISSIPATIVE NETWORKS
In t h i s section we draw a d i s t inc t ion between loss less networks a
( i . e . those networks with scat ter ing matrices satisfying s o s = - - 6 1 ) and f i n i t e nondissipative networks ( i . e . those composed of f i n i t e -n
interconnections of loss less elements). That there is a difference i s
seen by exhibiting a non-lossless nondissipative network, which is i n
f ac t an interconnection of loss less subnetworks.
Specifically we consider the s i tua t ion shown i n Fig.6
FIGURE 6 A NONDISSIPATIVE NETWORK
Calculations outlined i n Ref. 15 show tha t
(5.1)
Here the rea l f i n i t e constant a is arbi t rary.
The adjoint is found i n the usual fashion: f 7
a (5.2)
The composition product s o s can now be formed; it consists of the - - sum of a term 6(t-7), a term Y ( ~ , t ) u ( ~ - t ) and a term F ( t , ~ ) u ( t - T ) , where
F ( ~ , T ) i s given by
a If ( P(A)d), i s not divergenl, then i t i s impossible t o have s o s
= 6; on t h e o t h e r hand i f T (~ )d ) , is divergent , then d i r e c t cal- a Jcx '
cu la t ion shows t h a t s o s = 6 . Thus t h e network of Fig. 6 is loss-
less i f and only i f T (h)dh = m f o r any a. This conclusion is a l s o L 2 reached bv an a l t e r n a t i v e anaLysis elsewhere [ ~ e f . 3]: where the pro - p e r t i e s of s i m i l a r networks a r e more f u l l y described. The physical
reasoning suggesting t h i s r e s u l t is a l s o given i n t h i s reference; i n
b r i e f we may say t h a t i f ~ ( t ) vanishes f a s t enough as t -t m, it i s
poss ib le f o r t h e capaci tor t o t r a p charge and r e t a i n it t h e r e a f t e r . Then
&(m) f o r t h e network is nonzero.
The aim of t h i s chapter i s t o f i n d condi t ions on t h e s c a t t e r i n g
matr ix of a network which correspond t o t h e network being nondiss ipa t ive .
I t i s p l a i n t h a t we must appeal t o some more involved concept than t h e
e ( m ) = O property, which i s evident ly not preserved under interconnect ion.
W e in t roduce t h i s concept i n t h e next sec t ion .
B. QUASILOSSLESS SCATTERING MATRICES
Suppose w e a r e given a s c a t t e r i n g matr ix of t h e form of Eq. (4 .2) :
s ( ~ , T ) = A(t)6(t-?) + $ ( t ) y ( ? ) u ( t - r ) - - -
Direct ca l cu la t ions then y i e l d
a - -., s o s = ~ ( t ) ~ ( t ) b ( t - r ) + F ( ~ , T ) u ( ~ - T ) + ! ( ~ , t ) u ( ~ - t ) (5.4) \ - - - - -
where
t a
We may also form s o s, which we reca l l is w e l l defined as s and -a - -
a s both map bQ+ in to XI+ . Again by d i rec t calculation -a
By analogy with the loss less condition on scat ter ing matrices, we cal l
a scat ter ing matrix quasilossless i f
We note tha t i f s is lossless, then it i s automatically quasilossless. - - a
For i f s i s lossless, so tha t s o s = 6 1 , Eq.(5.4) immediately yields - - - - -n ~ ( t ) ~ ( t ) = Ln and F ( ~ , T ) = s. When these are subst i tuted i n ~ ~ . ( 5 . 6 ) , - - - equation (5.7) follows. The converse resu l t does not hold however.
Thus the time-invariant capacitor, inductor and gyrator, and the
time-varying transformer a r e a l l quasilossless. So is any interconnecting
network, a s N of Fig.5; t h i s i s because such networks are loss less -E
[Ref. 291. It is the quasilossless property which w i l l be shown i n succeeding
sections t o be retained on interconnecting quasilossless networks.
Henceforth, unless exp l i c i t assumption i s otherwise made, quasiloss-
l e s s scat ter ing matrices w i l l be taken as being also passive.
C. SCATTERING MATRICES OF NONDISSIPATIVE NETWORKS -- THE QUASIIDSSLESS CONDITION
We apply the cascade loading methods of chapter I V t o demonstrate t ha t
the scat ter ing matrix ofanondiss ipat ive network is quasilossless.
More specifically, we represent a nondissipative network N a s a - nondissipative network 3 of uncoupled c i r cu i t elements loading a net-
work N consisting en t i re ly of interconnections, as i n Fig. 5. Then -C'
though N need not be lossless, both N and are, the former be- - -z cause it contains no c i r cu i t elements a t a l l , t he l a t t e r because the
c i r c u i t elements t h a t c o n s t i t u t e N a r e los s l e s s , and a r e disconnected 4 from one another. Put another way, % i s t h e " d i r e c t sum" of a number
of l o s s l e s s networks, one corresponding t o each c i r c u i t element of 3 Following t h e n o t a t i o n of sec t ion C, t he s c a t t e r i n g matr ix s of -
N i s given i n terms of t h e s c a t t e r i n g matr ices C and 3 of N and - - -z a s ( ~ ~ . 4 . 1 0 ) ,
where we assume t h e ind ica ted inverse e x i s t s . Then, using t h e eas i ly a
derived r e s u l t ( h o k )a = ka o h f o r two opera tors h, k mapping - - a -a -a -t
i n t o 8+
a The product s o s can then be formed; a l l composition products w i l l
-a - be well defined and as soc ia t ive , s ince a l l f a c t o r s map i n t o s.
4 a
I n t h i s r e su l t ing expression f o r s o 2, it is poss ib le t o make -a
s impl i f i ca t ion using
These equations a r e consequences of t h e l o s s l e s s cha rac te r of N and -z % (and the r e s u l t i n g quas i los s l e s s charac ter of t h e i r s c a t t e r i n g matr ices) .
When a l l s impl i f i ca t ions a r e made, one f inds
That is, s is quas i los s l e s s . The d e t a i l e d ca lcu la t ions , though not - hard, a re avai lable i n Ref. 27. This reference a l s o e s t a b l i s h e s t h e
r e s u l t f o r t h e case when [gl -1 - o ] does not necessar i ly e x i s t .
I n f u l l , we have
Theorem 7.1. Let N be a nondissipat ive n-port network with - s c a t t e r i n g matr ix s. Then s is quas i los s l e s s , t h a t is, - -
We comment t h a t N must be solvable f o r t h e theorem t o hold; - t h e theorem does not e s t a b l i s h t h e exis tence of 2, but a condition
-1 on s when i t i s known t o e x i s t . The assumption t h a t [g& - Zp2 o 3,] - e x i s t s i s s u f f i c i e n t (but not necessary) f o r t h e exis tence of 2.
a In con t ra s t t o t h e l o s s l e s s case, where s o s = 6 1 does - - -
a a not imply s o s = 61, it is t r u e t h a t s o s = Q implies - - -
a -a - s o s = 61. One way o f e s t ab l i sh ing t h i s i s t o observe t h a t f o r a - -a s i n g l e element of t h e network % ca lcu la t ions r e a d i l y e s t a b l i s h t h a t a
s f o r t h i s element i s a r i g h t and l e f t inverse. I t follows t h a t the -a a a same is t r u e of s being a d i r e c t sum of t h e s e elemental s . I t can +' -a a l s o be e a s i l y seen t h a t i s a r i g h t and l e f t i nve r se f o r x; t h i s
-a - is because i s a constant matr ix mult iplying 6(t-7) . Using these -
a f a c t s , t h e c a l c u l a t i o n s used t o e s t a b l i s h s o s = 6 1 may be mod-
-a - - a
i f i e d only slightly t o show s o sa = 61. I n s e c t i o n F of t h i s chapter, - we indica te an a l t e r n a t i v e der iva t ion r e ly ing on knowledge of the quasi-
l o s s l e s s condit ion f o r immittances. Thus
Theorem 5.2. With t h e same condit ions a s requi red i n Theorem 5.1,
a a s 0 s = s o s = 6Ln -a - - -a
One of t h e p r i n c i p a l t a sks s t i l l ahead of u s ( t o be covered i n
Chapter VII ) i s t o demonstrate t h a t a s c a t t e r i n g matr ix s a t i s f y i n g Eq.
(5.10) possesses a syn thes i s a s a nondiss ipa t ive network. This r e s u l t
i s t h e converse of t h a t e s t ab l i shed i n t h i s sec t ion .
D. SCATTERING MATRICES OF NONDISSIPATIVE NETWORKS -- MATHEMATICAL FORM
In t h i s section we derive more specif ic re la t ions between the
matrices A, 9, 2 of ~ q . (4.2) -
which are consequences of the quasilossless character of s . We recal l , - see section B, Eqs. (5.5) and (5.6),
where
e ~ t , ~ ) = y t ) ~ c t ) ~ ( ~ ) + get) [Jrl(~i~(i)a] I(T) - t
If - s i s quasilossless, sa t isfying ( 5 . l O ) , we obtain immediately
w , - A ( t ) ~ ( t ) - = 1
-n (5.11)
F ( ~ , T ) - F ( T , ~ ) = o - - -n (5.12)
For convenience se t
m
o ( t ) = y t ) i ( t ) + y( t ) [ J z(A)J(A)~A] - t (5.13)
Then (5.12) becomes
Using the l inear independence of the columns of y, which as explained - ea r l i e r we are f ree t o assume, i t follows from (5.14), see Ref. 27, that
there ex is t s a constant matrix C such that -
with
This reference also establishes that passivity is ensured if and only if
C is nonnegative definite, written -
The method used to establish (5.17) is to show that corresponding to an i
incident voltage v , the associated &(a) is -
i The independence of the columns of y means that v can be selected - - to yield any value for the integral terms. Then for e ( m ) 2 0, (5.17) is necessary and sufficient.
The preceding manipulations have thus established
Theorem 5.3. The scattering matrix
is quasilossless if and only if [see E~S. (5.11), (5.13)) (5.15) (5.16)) and (5.17)1
w ~(t)~(t) = 1 - - -n (5.19.)
where C is a symmetric nonnegative definite matrix. -
An alternative formulation of the quasilossless condition can be
derived from either (5.19) or the second half of Eq. (5.10)--hitherto a
unused--vie. s o s = 61.The symmetry between Theorem 5.3 and the - -a -n
following theorem should be clear.
Theorem 5.4. The scattering matrix
s(t,~) = ;(t)6(t-~) + m(t)Y(~)u(t-~) -
is quasilossless if and only if
- ~(t)~(t) = 1 - - -n (5.20a)
where D is a symmetric nonnegative definite matrix. -
The two equations (5.19) and (5.20) can be written in more com- pact notation, as may be directly verified. Respectively, they are
and
To conclude this section, we point out modifications of these
results which apply to a subclass of quasilossless scattering matrices,
namely the lossless ones.
From the equation (5.4) of section B,
it follows that an additional condition to be fulfilled for losslessness
is
F(~,T) = 0 - -n (5.22)
In place of ~ ~ . ( 5 . 1 4 ) we obtain @ ( t ) Y ( T ) = 0, where @(t) is given by - - - - ~ q . (5 .13) .~ne l i nea r independence of the columns of y then requires - tha t @(t) =g, which i s ~ ~ . ( 5 . 1 5 ) with C = 0. That C = 0 also follows - - - - - from E ~ . (5.18), since f o r loss less s we require ~ ( m ) = 0 for a l l -
i square integrable v . Accordingly we have -
Theorem 5.5. The Scattering matrix
i s loss less i f
and
o r equivalently,
Contrary to what might be expected, it i s not necessarily t rue that
s e y = 2. This is i n f ac t because i t is not necessari ly t rue that - - a
5 0 s = & A n . - -
This concludes our discussion of the detai led mathematical con-
di t ions on quasilossless scat ter ing matrices. In summary, the ortho-
gonality condition on the coefficient matrix of the 6( t -T) term i n 2, a
and the resul t s a 3 = -y C (with C = 0 i f s i s lossless, and - - - - -- - symmetric nonnegative de f in i t e 0therwise)are the conclusions of the
analysis.
E. EXAMF'LES OF QUASIIOSSLESS SCATTERING MATRICES
A s a f i r s t example, w e consider
where A is orthogonal; l a t e r (chapter VII) it w i l l be shown t h a t t h i s - matrix can be synthesized from transformers and gyra to r s alone. We
have a t once
and a a
S 0 S = S 0 S = 6( t -7) - - -a -
a a by t h e orthogonali ty. I n t h i s instance, not merely s but a l s o s i s
-a - a r i g h t inverse f o r s ; t h e l a t t e r need not always be t h e case, a s seen by - t h e example of sec t ion D, chapter 111.
A s a more complicated example we consider t h e capacitor-loaded
transformer of Fig. 6. We found i n sec t ion A t h a t with ~ ( t ) t h e tu rns
r a t i o
s ( t , ~ ) = -6(t-7) + . T ( T ) ~ X P
(5.1) - Thus here the coe f f i c i en t of 6 ( t -T) is c e r t a i n l y orthogonal, being -1.
The matr ices @(t) and y ( ~ ) have become s c a l a r funct ions cp and $. - - We ca lcu la t e
m a
s ocp = - 2 ~ ( t )
where
He observe tha t c i s nonnegative, and i s zero i f P ( h ) d k
diverges; t h i s has already been established i n section A as the loss-
l e s s condition.
It may happen tha t s is not given i n the form above where ~ ( t )
appears expl ic i t ly , i . e . we may be given s purely as
Then we see using Eq.(5.19) that s w i l l be quasilossless i f $ i s
related t o by
where c i s a nonnegative number. Then the s of Eq.(5.26) is the
scattering matrix of a capacitor loaded transformer with turns r a t i o
2 This follows by comparing (5.1) with (5.261, (observe that T ( t ) =
v( t )q ( t ) ) .
F. QUASILOSSLESS IMMITTANCE MATRICES
There i s no a p r io r i reason to suppose tha t the existence of a
quasilossless condition on scattering matrices requires the existence
of one f o r immittance matrices. Indeed by analogy with the loss less
case, where t h e r e appears t o he no meaningfully simple l o s s l e s s con-
d i t i o n on immittance matr ices , one might expect t h e same f o r the quasi-
l o s s l e s s case. Spaulding [Ref. 91 has given a condit ion which is neces-
sary f o r lo s s l e s sness of f i n i t e networks i n terms of an impedance matr ix
z which i s -7
The network of Fig. 6 which has
is such t h a t Eq.(5.29) i s s a t i s f i e d i r r e s p e c t i v e of whether o r not - m
J a ~ ~ ( h ) d ? , diverges. Thus ~ q . (7.29) i s c e r t a i n l y not a s u f f i c i e n t con-
d i t i o n f o r lo s s l e s sness .
Spaulding g ives a syn thes i s procedure f o r impedances s a t i s f y i n g
(5.29) which always y ie lds a nondiss ipa t ive network. Thus we may wel l
ask, i s Eq.(5.29) both a necessary and s u f f i c i e n t condit ion f o r t h e
immittance of a nondiss ipa t ive network? The answer i s yes.
Equation (5.29) can be shown t o follow from t h e condit ion a
s 0 s = i3An i n t h e following way. From Eq.(2.8b), -a -
w e have
and thus
with the composition products a s soc ia t ive . Multiplying on t h e r i g h t by a
( z - + 6Ln), on t h e l e f t by ( z + 61 ), and evalua t ing t h e r e s u l t i n g pro- -a -n ducts leads a t once t o
W e have thus concluded t h a t a solvable nondiss ipa t ive network N - with an impedance matr ix z has z sa t i s fy ing (5.29). The r e s u l t can - - be extended t o nonsolvable networks; any nonsolvable f i n i t e network N - wi th an impedance matr ix z can be s p l i t i n t o t h e s e r i e s combination - of transformer-coupled inductors , and a solvable network N [Ref. 111. -1 Now N i s such t h a t (5.29) is s a t i s f i e d ; the transformer coupled
-1 inductors have an impedance s a t i s f y i n g Eq. (5.29) a lso . Then t h e im-
pedance of N s a t i s f i e s (5.29), being t h e sum of two such impedances. - Consequently w e have t h e fol lowing theorem, i n p a r t due t o Spaulding:
Theorem 5.6. Let N be a network with impedance matr ix z. - Then i f - N is non&ssipative,
and i f z s a t i s f i e s t h i s quaBilossless condition, t h e r e is a nondissiFat ive network synthes iz ing N. -
A dual theorem holds f o r admittances.
A s remarked i n sec t ion C, w e can now suggest new reasons why
sa i s a r i g h t inverse of s i f it i s a l e f t inverse. I f t h e network -a -
a t o which s corresponds possesses an impedance matrix, then z + z = 0 - - -a -7
a and by t h e symmetry of t h i s equat ion e i t h e r of s o s = 6 1 and - -a -
a s o s = 61 implies t h e o the r . I n t h e case where no impedance matr ix -a - - e x i s t s , minor adjustments d e t a i l e d i n Ref. 27 reduce t h e problem t o t h e
case where t h e impedance matr ix does e x i s t .
VI. FACTORIZATION OF SCATTERING MATRICES
A. THE CONCEPT OF FACTORIZATION
We have already mentioned the idea of factorization in section
C of Chapter 111. In this chapter we are concerned with some of the
conditions under which a passive scattering matrix s can be repre- - sented as the composition product of two passive scattering matrices
s and s a -1 -2'
The motivation for this is strong. The matrix s can be syn- - thesized, using an algorithm to be described in the next chapter, if
the matrices s and s can be synthesized. While the synthesis -1 -2
of an arbitrary s may not be immediately possible, syntheses of s - -1 and s may be immediately possible, since these matrices may have a
-2 simpler form than s Again, taking as a yardstick of the simplicity -1' of a scattering matrix its degree, a particularly pertinent problem
for investigation is the determination of conditions guaranteeing s -1
in ~q~(6.1) to have degree one, with s to have degree one less than -2
s. Then although a synthesis of s is not generally immediately - -2 possible, a synthesis for s is perhaps immediately possible, especially
-1 if we can ensure that s is quasilossless. Thus although with this -1 factorization a synthesis of s is not immediate, the problem of syn- - thesizing s has been reduced to a simpler problem, namely that of - synthesizing s of degree less than s.
-2 - The next section will consider the problem of obtaining a factor-
ization as in (6.1) with s quasilossless of degree one and s not -1 -2 -
necessarily of degree less than 2, while the following section will con- sider the requirements necessary to achieve degree reduction (i.e., s -2 of degree less than s). The results obtained will be used in the next - chapter to give a complete synthesis of any quasilossless s, -
B. CONDITIONS FOR FAClVRIZATION
The pr incipal resu l t i s as follows:
Theorem 6.1. Let s be a passive scat ter ing matrix of the - form of Eq. (4-2) :
s ( ~ , T ) = A ( t ) 6 ( t - ~ ) + ~ ( ~ ) Y ( T ) U ( ~ - T ) - -
and l e t s be a degree one scat ter ing matrix of the form -1
Then a decomposition
is possible f o r a l l c > 0 i f is a square integrable on [T,-) f o r a l l f i n i t e T, and s a t i s f i e s
where 1 i s a constant r-vector ( r being the degree of 5, or the number of columns i n y, assumed l inear ly independent).
Before outl ining the proof of t h i s theorem, we comment that
s i s a quasilossless scattering matrix, t h i s fac t following from an -1 application of theorem 5.3 with the simplifications A ( t ) = I.., 2 - and
y vectors ra ther than general matrices, and C a scalar . There are - - certainly quasilossless degree one s without A ( t ) = In, (e.g. -s ), - - - -1 but it is convenient t o assume the form of Eq. (6.2).
The theorem above precludes the case c=O, corresponding to s -1
lossless, a s d i s t i n c t from quasilossless. This is deal t with i n
Theorem 6.2.
I t should be noted that the theorem makes no claims as to the
passivity o r otherwise of s . s however is passive, since y i s -2' -1
square integrable on [T,m), fo r a l l f i n i t e T, and c > 0 (see Theorem
5.3). The passivi ty of s2 w i l l be discussed l a t e r .
Proof: (Outline only) With sl and s as given, we can - a form the well defined transformation s o s. Direct calculation
-la - ( fo r de t a i l s see Ref. 27) yields
(6.4a)
We now define
Then the quasilossless character of s means tha t s o s = s o a -1 -1 -2 -1
O S a ) O S = 6 A n O S = S. (zla O S) = ( s ~ -la - - Associativity of the products follows from the fact tha t a l l factors map $3 in to --+ %*
For notational convenience, we se t
where
(h )~ (h )dh , f tg(h(IH~)dh+ c 1
Although our def ini t ion of s w i l l not always make it passive, -2
w e sha l l t r y t o make i t so by choice of appropriate Cp; one condition - that must be f u l f i l l e d for passivi ty i s tha t y2(7) have elements
which are square integrable over (-a, TI fo r a l l f i n i t e T (chapter I V ,
section B). Although - y ( ~ ) has t h i s property ( s being passive) we must - i n addition require
t o be square integrable over (-=,TI. In order t ha t a t the same time
x 2 ( ~ ) have no greater a number of l inear ly independent columns than y(7) - ( fo r otherwise s would have greater degree than s), we sha l l suppose
-2 - that
where i s a constant vector. Direct calculation shows t h i s i s the - same as
Assuming ~ q . ( 6 . 3 ) , o r equivalently Eq.(6.5), we then obtain
from Eqs.(6.4) the following equation for s . -2'
where i t w i l l be noticed that the dimensions of the matrices corres-
ponding t o @ and y i n s a re the same as the dimensions of $ and - - - - y, namely n x r. This completes the proof. -
We comment tha t there do ex is t cp which sa t i s fy equations of - t he form (6.3). We sha l l be exhibiting such i n the next chapter.
The passivity of s may be examined with the a id of the fol- -2
lowing relation, which i s established in Ref.27. The re la t ion w i l l be
used i n the next chapter t o es tabl ish passivity i n our quasilossless
synthesis technique.
This must be nonnegative def in i te t o ensure passivity (Theorem 3.3, corollary).
Reference 27 also car r ies out the calculations which es tabl ish
the following theorem closely re la ted to Theorem 6.1; we now deal with
the factoring out of a loss less s -1'
Theorem 6.2. Let s be a passive scattering matrix of the - form of E ~ . (4.2):
and l e t s be a degree one scattering matrix of the form -1
then a passive decomposition
is possible i f i s square integrable on [T,") f o r a l l f i n i t e T, and s a t i s f i e s
We comment tha t i n t h i s case the passivi ty of s follows -2
automatically by v i r tue of the following eas i ly established resu l t
[Ref - 271,
The matrix s is i n fac t given by (6.6), with c and Y -2
both zero.
C. CONDITIONS FOR FACTORIZATION WITH DEGREE REDUCTION
In section A we pointed out the des i r ab i l i t y of achieving a
factor izat ion with factors of l e s s degree than the original scat-
ter ing matrix s, while i n section B we detai led a factorization
where one of the fac tors had the same degree as s. Evidently some - additional condition(s) must be s a t l s f i e d i f degiee redugtion i s t o
occur. These conditions are given i n the following theorem:
Theorem 6 . 3 Let the hypotheses of Theorem 6.1 o r Theorem 6.2 hold, with the conditions for factor izat ion being sa t i s f ied . Then the fac tor s has degree one l e s s than s i f the fol- - lowing additional-gonditions a r e f u l f i l l e d :
9 i s a column of 5 - (6.11)
and when c > 0, c is the i - t h entry of y i f i s the i - t h column of $. - Proof: From Eq. (6.6) we have
* '3)T)uit-T)
(with c=O and y=O i n the loss less case).
For brevity we sha l l write t h i s as
Here @ and y are n x r matrices, and s w i l l have degree r -2 - -2
i f g2 and y have r l inear ly independent columns. Now y - - certainly has t h i s property, but ~q . (6 .11) can now be used t o show
that @ has one column zero (and thus fewer than r l inear ly in- -2
dependent columns).
I f T is the i - th column of @, then immediately the i - th - - column of @ i s -2
This completes the proof.
In summary, Theorem 6.1 has shown tha t factor izat ion of s is - -
a possible i f there ex i s t s a square integrable such tha t s 2 =
-Y Y j for some constant vector y; the more useful factor izat ion with - - degree reduction occurs i f T is also a column of 3, with a pos i t iv i ty - condition on an element of y then being required (Theorem 6.3). - Questions of passivi ty when a quasilossless, a s d i s t i nc t from lossless,
s i s factored out can i n theory be decided by Eq.(6.7). -1 In the next chapter we shal l be applying these resu l t s t o the
case where s is quasilossless. One interest ing feature is that we - shal l be able t o show tha t the result ing s2 i s (passive and) quasi-
lossless.
VII. NONDISSIPATIVE SYNTHESIS
In t h i s chapter we present a synthesis of quasilossless scat-
t e r ing matrices. The synthesis proceeds by factoring a quasilossless
scat ter ing matrix into the product of degree zero and degree one
quasilossless scat ter ing matrices, exhibiting syntheses f o r these in-
dividual matrices, and a synthesis f o r a scattering matrix product.
A. FACTORIZATION OF QUASILOSSLESS SCATTERING MATRICES
We commence with a scat ter ing matrix of degree r, assumed
quasilossless, of the form of Eq. (4.2).
The scat ter ing matrix
w i l l then be quasilossless, i n f ac t lossless, by the orthogonality of
~ ( t ) ; moreover s o s must then be quasilossless for - --O -
This matrix is also passive, f o r as may readily be verified,
a " o ? ) ~ o ( Z O O ) = 6 1 - s o s 6Ln - (s -n - -
Since s = s o (s o s), the problem of factoring s is now a - -0 -0 - - - problem of factoring s o 2, which we have pointed out above i s s t i l l
-0
quasilossless, but, as direct calculation shows, has as the fj(t-7) co-
e f f i c i en t matrix 1 rather than an arbi t rary orthogonal A ( t ) . Ac- -n -
cordingly, consider now quasilossless r-th degree s ( t , ~ ) , given by -
Suppose now the f i r s t column of @ i s designated as Cp. - - Equation (5.21a) which requires f o r any quasilossless s tha t -
f o r some nonnegative def in i te 5, then yields
where Y i s the f i r s t column of C. The conditions of Theorem 6.3 - - are now d i rec t ly applicable t o yield
where
Here c is the f i r s t entry of y and thus the 1, 1 entry of C . 11 -
I t i s therefore nonnegative since C is nonnegative def ini te . The - vector y(t) i s square integrable on [ ~ , m ) , because the elements of
@(t) have t h i s property, (chapter IV, section B) . The matrix 6 - -1 i s thus f i r s t degree, ( ~ a s s i v e ) and quasilossless by Theorem 5.3.
An important resu l t of t h i s factor izat ion i s that the matrix
ol, of degree (r-1), is also (passive and) quasilossless. To observe a
the quasilossless property f i r s t , Eq. (6.4b) yields S - -1 - Sla O 2' a a a a Then S o S = s o s o s o s = s 0 6 & ~ 0 5 = 6 L ~ .
-la -1 -a -1 -la - -a
To demonstrate passivity, we r eca l l tha t from Eq.(6.7), c f 0,
a Since &An - - s o - s = - ~ ( t , r ) u ( t - 7 ) - - ? ( r , t )u ( r - t ) - from Eq. (5 .4) ,
Eqs. (5.5) and (5.19) together yield
a 6Ln - S o - s = - y ( t ) $(r)[u(t-T) + ~ ( r - t ) ] (7.7)
and thus
1 This w i l l be nonnegative def in i te i f and only i f - C - - c - - y 7 i s non-
negative. But recal l ing the def ini t ion of - y as t he f i r s t column 1
of C, we see tha t C - 1 is indeed nonnegative def in i te by - - vi r tue of the eas i ly ver i i i ed re la t ions
and - T C T = c @ D - - - 11 - (7.8b)
where @ denotes the d i rec t sum and
In the case where c ,la= 0, passivity of S follows immediately -1
a - as 6Ln - El 0 g1 - 6Ln - 2 o - s, see ~ q . ( 6 . 1 0 ) .
The coefficient matrix of the &(t-7) term of S i s orthogonal by -1
the quasilosslessness of S I t i s in f ac t Ln a s d i rec t calculation of -1 '
a S = s o s shows. -1 -la -
Accordingly S is, l i k e s, passive, quasilossless and has also i t s -1
6(t-7) coefficient matrix 1 . It must therefore possess a factor i - -n
zation of its own
Here now s i s (passive), quasilossless, of degree one and S i s -2 -2
(passive), quasilossless, of degree (r-2)) and with i ts 6( t -T) co-
e f f ic ien t matrix 1 . This process can then be repeated with S -n -3'
S etc . un t i l S a t which point the process terminates as we reach 4 -r a matrix of degree zero. The number of degree one fac tors is thus r, the degree of s. -
The preceding arguments collected together then es tabl ish the
following theorem, which i s the fundamental theorem of t he thesis:
Theorem 7.1. Let s be a passive quasilossless scat ter ing - matrix of the form
s ( ~ , T ) = ~ ( t ) 6 ( t - ~ ) + m ( t ) x ( ~ ) ~ ( t - ~ ) - -
then there ex i s t s a factorization of s of the form - S = s 0 s 0 s O . . . S - -0 -1 -2 -r (7.10)
where s i s of degree zero, s s -1' -2' ' ' s are of degree one, and a l l 3 h e s . are (passive and) quasilossless. The scat- tering matrice3 sl, s~,...s, are of the form
wllere 'P. i s square integrable on [ ~ , m ) f o r a l l f i n i t e T, -1
ci2 0; the scat ter ing matrix s i s of t he form -0
where A ( t ) i s orthogonal. -
B. REALIZATION OF A PRODUCT OF SCATTERING MATRICES
We consider f i r s t the rea l iza t ion of s o s when s and 1 2 1
s a re scat ter ing matrices of one port networks N and N2. With 2 1
the gyrator i n Fig. 7a possessing the scat ter ing matrix
6 ( t - ~ ) , the network of Fig. 7b i s readily found to have scattering
matrix
FIGURE 7 REALIZATION OF ONE PORT SCATTERING MATRIX PRODUCT
The scattering matrix of the network i n Fig. 7c can be found
using the cascade loading resu l t of ~ ~ . ( 4 . 1 0 ) , and is [ ~ e f . 321
To r ea l i ze the matrix s o s i n the n-port case, we use n un- . -. -1 -2
coupled g y r a t k s (so that the scat ter ing matrix of t h i s network con-
sists of the direct sum of n matrices s ). The i - th ports of El -g
and N are connected to the i-th gyrator, (i = 1, 2, .. .n), as in -2
Fig. 7. Calculations for this case are almost as simple as those for
the one port. Schematically, we represent the realization by Fig. 8 below.
FIGURE 8 RIWLIZATION OF MULTIPORT SCATTERING MATRIX PRODUCT
Extension to the multiplication by more than one scattering
matrix is simple. Figure 9 shows by way of example two realizations for s o s o s
-1 -2 -3'
FIGURE 9 REALIZATION OF MULTIPLE PRODUCT
C. REALIZATION OF QUASILOSSLESS SCATTERING MATRICES OF DEGREE ZERO AND DEGREE ONE
Hitherto we have shown how t o factor an arbi t rary quasilossless
scattering matrix into the prpduct of quasilossless scat ter ing matrices
of degree zero and d~gree-on*; i n the preceding section we have also
shown how t o rea l ize scattering matrix products, given realizations of
the factors. This section s e t s out realizations of the factors, thus
completing the synthesis of arbi t rary quasilossless scattering matrices.
There is more than one nontrivial variant realizing s ( t ,7 ) = A(t) - - ~ ( t - 7 ) where A is orthogonal, see Ref. Zi' for two. We w i l l content - ourselves with presenting one such real izat ion here.
Since - A(t) is orthogonal, there ex is t s an orthogonal T ( t ) Such -0
tha t [Ref. 33, p.2851
where hi is +1 o r -1.
Scattering matrices +&(t- 'r) and -6 ( t -7 ) can b e synthesized by an
open c i r cu i t and a short c i r cu i t respectively.
Now consider i n i t i a l l y the matrix, closely re la ted t o one of the
blocks of (7.14), . . . .
. .
s in e j ( t )
Since t h i s m a t r i x is symmetric, there ex is t s an orthogonal 2 x 2 matrix
I n chapter IV, sec t ion F, i t was es tabl i shed t h a t an orthogonal
transformer of t u r n s r a t i o ~ ( t ) connected a t t h e p o r t s of a network - N of s c a t t e r i n g matr ix s y ie lds a network N of s c a t t e r i n g -2 -2 -1 matrix ( see Eq. 4.14)
We note that .
can be synthesized a s a two por t of t h i s type, with ~ ( t ) = ~ . ( t ) , a n d -J
N an uncoupled s h o r t - c i r c u i t and an open-circui t . -1
By connecting a u n i t gyra tor ac ross t h e f i r s t p o r t only of t h i s
network, we o b t a i n a network of s c a t t e r i n g matr ix
Physical ly, t h i s i s because t h e gyra to r causes a 180 degree phase
s h i f t i n waves t r a v e l l i n g through it i n one d i r e c t i o n only. The
mathematical v e r i f i c a t i o n i s not d i f f i c u l t .
By simple juxtaposit ion, the sca t t e r ing matrix
- s i n B . ( t )
s i n B . ( t )
can then be synthesized, and, f i n a l l y , by connecting a transformer
of turns r a t i o matrix T ( t ) a t the input ports, it follows from -0
Eq.(7.14) and another appl ica t ion of t h e r e s u l t of chapter I V ,
sec t ion F, t h a t we achieve a network of scat ter ing matrix
The r e a l i z a t i o n of
i s physical ly more straightforward. This - s is of t h e form of the
quas i loss less passive degree one sca t t e r ing matrices r e s u l t i n g from
t h e fac to r iza t ion of Theorem 7.1, when we take c 2 0 and - cp t o
be square in tegrable on [T ,~J ) f o r a l l T.
Consider the network of Fig. 10
FIGURE 10
PROTOTYPE NETWORK FOR DEGREE ONE SCATTERING MATRIX
Easy calculations yield that the admittance of t h i s network is
given by
y ( t , ~ ) = n(t)u(t-T);(T) - - - (7.19)
when n is an n-vector whose i - th entry is n. . We take - 1
Th& t h i s network i s precisely a real izat ion of ~q . (7 .18) , as may
be proved by showing tha t ( 6 1 + y)-l 0 (61 - y ) is precisely the -n - -n -
s of ~ ~ ~ ( 7 . 1 8 ) . The actual calculations a r e detailed i n the appendix. -
We note t h a t t h i s r e a l i z a t i o n of a degree'one s c a t t e r i n g matrix
uses only one energy s torage element.
The work of t h i s and t h e preceding two sec t ions then y ie lds the
following theorem.
Theorem 7.2. A passive quas i los s l e s s n-port s c a t t e r i n g matrix may be r e a l i z e d a s a nondissipat ive network. The number of energy s to rage elements used equals t h e degree of t h e sca t t e r ing matrix.
The forms of t h e r e a l i z a t i o n a r e many. One poss ib le one i s shown
i n Figure 11 below.
FIGURE 11
PASSIVE QUASILOSSLESS $CATTERING MATRIX REALIZATION
Here N i s a degree zero network, and N through N a re degree -0 -1 -r
one networks of t h e type of Fig. 10. Considerable v a r i a t i o n s a r e pos-
s i b l e , and many a r e discussed i n Ref. 27. Among t h e more s ign i f i can t ,
w e would mention:
1. An a l t e r n a t i v e r e a l i z a t i o n of N . -0
2. The replacement of t h e banks of n gyra to r s required t o
r e a l i z e any one product with one gyra to r and o the r cir-
c u i t r y . More s p e c i f i c a l l y , t h e r e exists an equivalence
of t h e f o m shown i n Fig. 12, where ~ ( t ) i s a mult iport - transformer.
REDUCTION IN THE
A possible reduction i n the number of gyrators required
for the gyrator bank t o which N i s connected, and -Q
N i t s e l f ; and a possible reduction i n the complexity -0
of the transformer realizing N . -Q
The use of a network dual t o Ni i n place of N . . The -1
dual network has scattering matrix which is the negative
of tha t of Ni, and can always be used by multiplying
fac tors i n Eq. (7.10) by -1; i f N . has the rea l iza t ion -1
of Fig. 10, then its dual network has the rea l iza t ion of
Fig. 13.
FIGURE 13
NETWORK DUAL TO THAT OF FIGURE 10
D. SYNTHESIS EXAMPLE
We consider the scat ter ing matrix of degree 2:
The quasilossless character of t h i s scattering matrix can be
established by d i rec t computation; a l ternat ively we sha l l be able t o
observe that the matrix has a factor izat ion into quasilossless matrices.
The f i r s t column of the - @ matrix is [2=] and t h i s generates
the matrix
Note tha t [e7, 01 comes from [2e-', o]{ rk?2'M)-1. The constant c 11 'T
a 0 of equation (7.>) calculates t o zero, as s 0 [2e , 01 = 0. By - -
a calculating s o s = s we derive the degree one scattering matrix
-la - -2
which i s readily ver i f ied as quasilossless. The matrix - s is now
factored into s o s where s and s are both quasilossless -1 -2 -1 -2
and degree one.
The turns-ratio vector of the transformer i n Fig. 10 which
rea l izes s i s -1
Fig. 14 below real izes s For s the turns r a t i o vector i s -1 ' -2
Figure 14b real izes s In accordance with the formation of -2'
products of scattering matrices, Fig. l 4c real izes - s. Figure l4d
gives a realization using networks dual to those of Fig. 14a and l4b.
1- 2 ..i 1 HENRY
'-
(a)
1 HENRY
1 HENRY
..
1 HENRY
(dl
FIGURE 14
EXAMPLE ILLUSTRATING REALIZATION OF QUASIIBSSLESS SCATTERING MATRIX
V I I I . LOSSY SYNTHESIS
A. STATNENT OF PROBLEM AND AN OUTLINED ATTACK
The problem we consider can be posed simply: given a s c a t t e r i n g
mat r ix s ( t , ~ ) of a f i n i t e network, wi th s assumed passive, how may - - s be synthesized? -
When s i s not quas i lo s s l e s s , we term s lossy. For convenience - - we s h a l l r e s t r i c t cons idera t ion i n i t i a l l y t o t h e lossy one p o r t ; t h e
techniques t o be described a r e extended t o t h e mul t ipor t case i n sec t ion
C.
We r e c a l l from s e c t i o n A of chapter I V t h a t every f i n i t e network
can be represented by an equiva lent network containing t ime-invariant
r e s i s t o r s , capaci tors , inductors , and gyrators , and t ime-variable
t ransformers. By s u i t a b l e u s e of t ransformers we can assume t h e time-
inva r i an t elements, i n p a r t i c u l a r t h e r e s i s t o r s , t o be normalized t o
one ohm, so t h a t t h e network may then be regarded a s being composed of
a nondiss ipa t ive network N i n conjuc t ion with a number of u n i t re- -C
s i s t o r s . The p o r t s of t h e nondiss ipa t ive network w i l l be t h e p o r t s of
t h e o v e r a l l network, toge ther wi th terminal p a i r s across which u n i t
r e s i s t o r s a r e connected i n t h e o v e r a l l network.
I f we assume t h a t t h e nond i s s ipa t ive network possesses a sca t -
t e r i n g matr ix C, then we may w r i t e -
The b a s i s f o r tak ing t h e 1, 1 e n t r y of C a s s was e s t ab l i shed - i n sec t ion F of chapter I V ; when t h e p o r t s of
NC corresponding t o
t h e 2, 2 terms i n t h e p a r t i t i o n of C a r e terminated i n u n i t re- - s i s t o r s ,
= s must be t h e s c a t t e r i n g matr ix of t h e r e s u l t i n g
network.
For a one port we make the assumption that s can be synthesized
using one resistor at most, as is possible for the time-invariant case,
see e.g. Ref. 32. Then the C . . are scalars, with I J
By the finiteness, we know that s is of the form
and accordingly we assume the following form for C -
which is of the same degree as s. Since - C is quasilossless (cor-
responding to a nondissipative network) the coefficient matrix of 2
6(t-7) must be orthogonal. Accordingly we suppose (noting a S 1 by
passivity)
The choice of the particular orthogonal matrix multiplying 6(t-7) is not
critical.
At this stage we are now faced with the following problem: determine
the unknowns Cp+ and * in Eq.(8.5) so that C is quasilossless. Ob- - - serve that all other terms in C are known, via ~~.(8.3), the stated -
form f o r s.
Once having found ifx and J* we have a synthes is f o r s, - f o r we can synthesize C by t h e h i the r to e s t ab l i shed methods, and - then simply obta in s by terminating the second p o r t of t h e network
synthesizing C i n a u n i t r e s i s t o r . - One poss ib le way t o go about f inding p and $+ is de ta i l ed -
i n t h e next sec t ion .
B. SOLUTION PROCEDURE
Our aim w i l l be t o der ive equations s a t i s f i e d by t h e unknowns a ifx and $* which a r e p o t e n t i a l l y more use fu l than merely C o C = 6 1
-a - -2' Theorem 5.3 o u t l i n e s equations which must be s a t i s f i e d t o
make the - G of ~ q . ( 8 . 5 ) quas i loss less . When applied t o t h e case i n hand,
we have
Examining j u s t t h e f i r s t column of t h i s matrix equation, we obta in an
equation t h a t is independent of I)*: -
where
From Eq.(8.7) we conclude t h a t knowledge of ~ ( t ) i s s u f f i c i e n t t o - determine <pic and 2, t h e l a t t e r by evaluat ing l i m ~ ( t ) , t h e former - t- - by f i r s t computing ~ ( t ) . From (8.7a) we have -
while from (8.7b)
Consequently we have on equating these two equations
The determination of P i s thus equivalent t o solving t h i s matrix - Riccati d i f fe ren t ia l equation [ ~ e f .34]. We comment that whereas we
require P to be defined over (-m,m) a d i f f e r en t i a l equation of the - form of Eq.(8.9) need not necessarily possess a solution defined over
2 an in f in i t e interval. We note also tha t the [L-a ( t )] denominators
appearing i n Eq.(8.9) indicated that possible s ingular i t i es can occur 2
i n P; a ( t ) can have values lying between -1 or +1, and thus 1-a ( t ) - may be zero everywhere, over an interval, o r merely a t an isolated
point.
Once P has been found, (and with it cpx) the function $* can - - immediately be obtained from Eq. (8.6), as
Accordingly, by assuming a form f o r the network synchesizing s,
we are led t o the nonlinear d i f fe ren t ia l equation (8.9). I f t h i s can
be solved over (-m,m) then a synthesis f o r s is achievable.
C. THE MULTIPORT CASE
By analogy again with t h e t ime-invariant case, we seek a non-
d i s s i p a t i v e 2n-port network N such t h a t , when the l a s t n p o r t s of -C
N a r e terminated i n u n i t r e s i s t o r s , a network N of prescr ibed -C - pass ive s c a t t e r i n g matr ix s i s obtained. -
We suppose
s = ~ ( t ) 6 ( t - 7 ) + @ ( t ) y ( ~ ) u ( t - 7 ) - - - - and
Here t h e A. . ( t ) a r e matr ices chosen t o make t h e coe f f i c i en t matr ix -13 w w
of 6( t -T) orthogonal. This i s p o s s i b l e because 1 -A A and 1 -A A -n - - -n - - a r e nonnegative d e f i n i t e by p a s s i v i t y . h he n x n matrix A ( t ) may be f i r s t
-12 chosen t o make t h e n rows of [A(t)A&t)] - orthon0:rniai; then [A ( t ) _ ~ ~ ~ ( t ) ] i s
-21 s e l e c t e d t o complete these n rows t o a set of 2n orthonormal row vec to r s . )
Theorem 5.3 allows us t o w r i t e down t h e following r e l a t i o n in-
volving t h e unknows @* and t h e constant nonnegative d e f i n i t e matr ix -
where
A s f o r t h e one port , knowledge of P serves t o determine @* and - - C. Again a s f o r t h e one port , it fol lows t h a t the re is a R icca t i -
d i f f e r e n t i a l equat ion f o r g,. provided t h a t w e assume t h e exis tence -1
Of +21 (corresponding t o [l-a']-' i n t h e one p o r t case) . This
equation i s
The remaining unknown i n C i s Y*, which may be determined from - - t h e following equation, r e s u l t i n g from an app l i ca t ion o f Theorem 5.3:
D. EXlUlPLE OF ONE PORT SYNTHESIS
A s an example of t h e preceding theory, consider t h e one por t
s c a t t e r i n g matr ix
We seek a quas i los s l e s s C of t h e form -
+ 2e
(8.171
The matr ix P of Eq.(8.7b) becomes a s c a l a r funct ion; it s a t i s f i e s - t h e d i f f e r e n t i a l equat ion (8.9):
@(e-2t+ e -t 2 @(e-2t+ e-t * + 2 ~ 2L-4. + 2e -2. ] - P e-'Le-.t + 2e -2.3 + 2e-2t =
A so lu t ion of t h i s equat ion is
-2t which y ie lds CEO and cpn = -e . Equation (8.10) then y i e l d s
$* = (-e 2 -2t) and thus
The methods of chapter VII can then he applied to synthesize C. - First we synthesize the matrix
which has the form shown in Fig. l5a. The matrix
is then synthesized, and has the f o m of Fig. l5b. Note that
C = s o s Then it is clear that Fig. l5c synthesizes s. - -0 -1' -
1 ' -5) -1:n 1 HENRY
2 '
FIGURE 13
A LOSSY SYNTHESIS
Several comments can be made here. F i r s t , ca lculat ions of the
type out l ined i n the previous chapter y i e l d
Second, because of t h e shor t c i r c u i t and open c i r c u i t appearing i n
Fig. l 5 a and again i n Fig. l5c, t h e a s soc ia t ed transformer is cap-
a b l e of s impl i f i ca t ion . The transformer may be replaced by a s i n g l e
two-port t ransformer of tu rns r a t i o (1+J2): 1, t h e primary i n s e r i e s
wi th 1 - l', t h e secondary i n s e r i e s with 2 - 2 ' . Third, we observe
t h a t p r e c i s e l y one r e a c t i v e element i s used f o r t h e ove ra l l synthesis ,
corresponding t o t h e degree one na ture of t h e prescr ibed s(t ,r) .
Fina l ly , may be synthesized a s t h e product of two sca t t e r ing matrices,
-s and -s The networks of Figs. 15a and l 5 b a r e then replaced by -0 -1 '
t h e i r duals . I n t h e case of Fig. l5a, t h e shor t c i r c u i t and open c i r -
c u i t a r e simply interchanged, while Fig. 13 i l l u s t r a t e s t h e dual of
Fig. l5b.
IX. RELATION TO OTHER WORK
A. SPAULDING'S MSSLESS IMMITTANCE SYNTHESIS
Spaulding's immittance synthesis of lossless and more generally
quasilossless immittances [~ef .g] can be applied to yield a quasiloss-
less scattering matrix synthesis. Suppose the given quasilossless
scattering matrix is
We can construct s by synthesizing -
S = A(t)6(t-T) - - (9.2a) and
s = 1 6(t-T) + x(t)@(t)Y(-i)u(t-T) -1 -n - - - (9.2b)
As is discussed in Ref.11, it is possible to form (6Ln + S1)-l because
the term multiplied by 6(t-T) in 61 + S is a nonsingular matrix. -n -1
Then the admittance
evidently exists, and is quasilossless because S is. Spaulding's -1 synthesis can then be used on Y -1 '
TWO remarks should be made. First, in the synthesis the number
of reactive elements used is equal to the degree of Y1 (by Spaulding's results) which is in turn equal to the degree of - s and S (by the -1' results of section D of chapter IV). Second, since s = (-S) - o (-El), we can use an impedance rather than admittance quasilossless synthesis
if we wish, with the impedance possessing scattering matrix -S1.
The converse problem of using t h e s c a t t e r i n g matr ix synthes is
given a quas i los s l e s s immittance can a l s o be considered. Suppose
the re fo re we a r e given a pass ive quas i los s l e s s immittance without
l o s s of g e n e r a l i t y assumed an impedance Z. I t i s known [Refs. 9, - 111 t h a t Z must be of the form -
(9.3)
where r i s a skew symmetric matr ix. Physically, t h i s corresponds - t o t h e s e r i e s connection of transformer-coupled inductors, t r ans -
former-coupled gyra tors , and transformer-coupled capaci tors .
Sca t t e r ing matr ices need not e x i s t f o r such Z ( ~ , T ) ; it i s known 3 f o r example t h a t t 6 ' ( t - 7 ) ~ ~ has no s c a t t e r i n g matrix CRef.351.
However it i s a l s o known t h a t Z can be synthesized a s t h e s e r i e s - combination of an inductor-terminated transformer with impedance
matr ix - y ( t ) 6 ' ( t -T)T(T) - and a pass ive impedance Z with Z given -1' -1
by
Moreover [Ref. 111, Z1 has a s c a t t e r i n g matr ix S which i s of course -1'
quas i los s l e s s i f Z is. Consequently, although Z i t s e l f may not -1 -
be suscept ib le t o t h e s c a t t e r i n g matr ix synthesis , we can, by ex t rac t ing
t h e induct ive p a r t of Z, arrange t o synthesize t h e remainder by t h e sca t -
t e r i n g matr ix synthes is .
B. A NEW VIEWPOINT OF LQSSLESS TIME-INVARIANT SYNTHESIS
Since time-invariant networks a r e merely spec ia l cases of time-
varying networks, and l o s s l e s s t ime-invariant networks s p e c i a l cases
of nondissipat lve time-varying networks, both the quas i los s l e s s sca t -
t e r i n g matr ix and immittance matr ix synthes is should work f o r time-
inva r i an t matr ices. Indeed they do, and i t is i n s t r u c t i v e t o consider
t h e r e s u l t s .
The time-invariance p laces a cons t r a in t f i r s t upon t h e mathematical
form of s o r z, and then upon t h e network r e a l i z i n g the matrix. - - I n t h e case of
j - a t ~ ( t ) is a cons tant ; t h e columns of @ a r e of t h e form u t e cos b t - - - -
j - a t . o r u t e s i n b t , where u i s a constant vector, a and b a r e - -
j +a t r e a l with a p o s i t i v e ; t h e columns of Y ( t ) a r e of t h e form v t e , - -
j +a t j +at . v t e cos b t o r v t e s i n b t . I t should be noted t h a t these - - r e s t r i c t i o n s a r e not s u f f i c i e n t , but merely necessary, f o r time-
invarianee.
For a quas i lose le s s impedance of t h e form of ~ q . ( 9 . 3 )
t h e time-invariance causes T and t o be constant , and ~ ( t ) t o - - - be of t h e form N(0)exp Ct where C i s a skew matrix. When t h i s - - - s u b s t i t u t i o n is made one der ives
w N(t)u( t -T)N(T) = F ( o ) ~ @ cos a i ( t -T) - s i n a . ( t - T ) - - - -"[ 1
s i n a ( t - T ) cos a . ( t - T ) i 1 I
where @ denotes d i r e c t sum, P i s a constant orthogonal matrix, and - t h e ai a r e r e a l nonzero constants .
When t h e cascade synthes is is applied t o t h e - s of (9.4) , two
c l a s s e s of f i r s t degree networks r e s u l t . Corresponding t o t h e s e l e c t i o n j - a t
of a column of @ of t h e form t e - t o genera te a f i r s t degree f a c t o r ,
a t ime-invariant network i s ext rac ted . On t h e o t h e r hand a column Of t h e
j - a t j - a t . form t e cos b t o r t e s l n b t leads t o t h e ex t rac t ion of a
time-varying subnetwork. A t t h e next ex t rac t ion however, one may
e x t r a c t a second time-varying degree one network, which i n combination
wi th t h e f i r s t , behaves l i k e a t ime-invariant degree two network. Thus
t h e degree two networks t h a t appear i n t h e c l a s s i c a l cascade synthes is
[ ~ e f . 321 may be regarded a s cascades of time-varying degree one net-
works (with r e a l parameters) .
A s f a r a s t h e impedance syn thes i s goes, we note from Eq. ( 9 . 5 )
t h a t t h e use o f a t r a n s f o r m e r of t u r n s r a t i o P N(0) reduces t h e pro- - - blem t o t h a t of synthesizing
a ( t -T) - s i n a ( t - T ) ~ ( t - T )
a ( t - T ) cos a ( t - T ) 1 -I
When Laplace transformed, t h i s i s
which has the synthes is of Fig. 16a below. But a l s o
a t - - s i n a ( t - T )
a( t- ' r ) cos a ( t - T ) 1 s i n a t -cos a t
u ( t -T) p s a7 s i n aT1
1 s i n a'r -cos aT J and t h e synthes is with time-varying transformers of Fig. 16b app l i e s .
Thus a time-varying transformer terminated i n capaci tors r e a l i z e s a
time-invariant network which, using time-invariant elements only,
normally requires a gyrator i n t he real izat ion.
( a ) (b)
FIGURE 16
EQUIVALENT NETWORKS, WITH THE TRANSFOF&!ER
TURNS RATIO - M( t ) = rS at s in a]
s i n a t -cos a t
A s indicated by Eq.(g.5), time-invariant transformers combined with
blocks of the type shown i n Flg. 16 w i l l r ea l ize any loss less time-
invariant Z, once removal has been made of se r ies inductors, cor- - - responding t o - ~ ( t ) 6 ' ( t - T ) T ( T ) , - gyrators, corresponding t o - T ( t ) ,
and capacitors, corresponding t o ( 0 - - - r - - 1 P ~ ( 0 ) . The time-variable
synthesis procedures lead however t o dif ferent forms f o r the blocks.
W e comment that fo r a reciprocal network, there w i l l be no se r ies
gyrators extracted, while when the blocks of Fig. 16a are connected
together correctly, e i ther 1-1' or 2-2' w i l l remain open-circuited or
short-circuited for each block. This w i l l allow the removal of the
gyrators, and the replacement of some of the capacitors by inductors
[ ~ e f . 361.
C. RELATION TO TWO PROBLEMS OF AUTOMATIC CONTROL
We make brief mention here of how the material tha t has been developed
has shed l igh t on some problems of contvol systems which a re reported i n
Refs. 16, 17.
One of t h e aims of applying feedback around a system is t o i m -
prove t h e system's performance by decreasing i t s s e n s i t i v i t y t o var i -
a t ions i n t h e p lan t parameters. For t ime-invariant systems, a re-
i n t e r p r e t a t i o n of known r e s u l t s es tabl i shed t h a t t h e inverse of t h e ,f return-difference" matr ix of t h e system (which is a matr ix c lose ly
r e l a t e d t o the loop ga in ) must be t h e s c a t t e r i n g matrix of a l i n e a r ,
passive, t ime-invariant network, i f improvement i n t h e system's per-
formance was t o be achieved. Using t h e mathematical c r i t e r i a contained
here in f o r pass ive tine-varying s c a t t e r i n g matrices, t h i s r e s u l t was
extended t o time-varying systems, thus enabling the wr i t ing down of a
s p e c i f i c mathematical c r i t e r i o n f o r improvement i n system performance
given a mathematical desc r ip t ion of t h e system.
The design of l i n e a r r egu la to r s f o r time-varying systems using
optimal cont ro l theory has l e d t o a f u r t h e r usefu l r e s u l t .
I n t h e l i n e a r r egu la to r problem, t h e ob jec t i s t o design a feed-
back law which w i l l r e t u r n t h e system t o t h e zero s t a t e when i t i s
d isp laced from t h e zero s t a t e , i n an optimal fashion. By " i n an
optimal fashion" w e o f t e n mean by minimizing t h e sum of ( i ) t h e energy
required t o r e t u r n t h e system t o t h e zero s t a t e and (ii) t h e i n t e g r a l
squared e r r o r ( i .e . t h e e r r o r being t h e displacement of t h e s t a t e vec tor
from zero) . Using such a c r i t e r i o n , ( involving a quadra t i c l o s s funct ion) ,
i t tu rns out t h a t a l i n e a r feedback law provides minimization, t h a t is,
t h e instantaneous input vector . Reference 17 e s t a b l i s h e s t h a t t h e
feedback law computed v i a t h e optimal cont ro l theory w i l l automatical ly
y i e l d a system whose performance i s improved with r e spec t t o p l a n t para-
meter s e n s i t i v i t y . The der iva t ion of t h i s r e s u l t depends on t h e e x p l i c i t
formulation mentioned above of t h e c r i t e r i o n f o r improvement i n system
performance; t h e r e s u l t is a use fu l l i n k between modern and c l a s s i c a l
cont ro l concepts.
X. CONCLUSIONS
A. S W R Y OF PRECEDING WORK
I n r e t rospec t it i s poss ib le t o say t h a t much of t h e work s e t
f o r t h has been a f o r m a l i s t i c p resen ta t ion of p roper t i e s of time-
varying networks, where t h e r e has been l i t t l e o r none before. But
it would be wrong t o say t h a t t h i s formalism represented t h e s o l e
content of t h e t h e s i s . Of paramount i n t e r e s t has been t h e iden t i -
f i c a t i o n of t h e concepts of nondiss ipa t ive networks and quasi loss-
less s c a t t e r i n g matr ices. By chapter V, a n a l y t i c a l tebhniques were
a b l e t o e s t a b l i s h t h a t every so lvable nondiss ipa t ive network possesses
a q u a s i l o s s l e s s s c a t t e r i n g matrix. How s a t i s f y i n g it was then t o s e e
t h a t i n chapter V I I , syn the t i ca l techniques could e s t a b l i s h t h a t t h e r e
i s a nondiss ipa t ive network corresponding t o a quas i los s l e s s matrix.
Of less but not minor importance, t h e r e a r e two concepts we would
mention: t h a t of t h e pass ive s c a t t e r i n g matrix, examined v i a Hi lber t
space techniques i n chapter 111, and t h a t of t h e lossy synthesis , pre-
sented i n chapter V I I I . Here we were a b l e t o reduce t h e physical pro-
blem f i r s t t o a mathematical problem, t h a t of solving a fuc t iona l eq-
uat ion, and then t o a more usual mathematical problem, t h a t of solving
a d i f f e r e n t i a l equation.
Perhaps w e could mention t h e cascade loading concept too, though
without decrying i t s importance it i s poss ib ly of more use i n time-
inva r i an t s i t u a t i o n s than time-varying ones.
Throughout t h e web of a l l t h e preceding theory, one golden th read
i s t o be seen: t h e cen t ra l aim i n t h i s s tudy has been, and w i l l cont inue
t o be, t h e t r a n s l a t i o n of physical concepts i n t o mathematical concepts
and v i c e versa . Without exception, a l l t h e ma te r i a l presented i n t h i s
t h e s i s se rves t h i s aim.
B. DIRECTION FOR FUTURE RESEARCH
One outs tanding problem remaining i s a f u l l so lu t ion of t h e l o s s y
syn thes i s problem. Insofar a s we found i n chapter VII an equation whose
so lu t ion y i e l d s a lossy synthesis , one a t t a c k on t h e lossy synthes is
problem would be t o demonstrate t h e exis tence of a s o l u t i o n of t h i s
equation. Hopefully t h i s ex is tence would he a consequence of pas-
s i v i t y .
While t h e r e s u l t s here have been concerned with passive networks,
no doubt many could be applied t o the study of a c t i v e networks. The
s o r t of study which might be rewarding, i n terms of both t h e o r e t i c a l
and p r a c t i c a l r e s u l t s , would be one embracing networks composed of
passive t ime-invariant elements and time-varying transducers (e.g.
i dea l ampl i f i e r s whose ga ins a r e funct ions of time).
A s i s apparent from t h e theory, a l l time v a r i a t i o n s can be
lumped i n t o time-varying transformers. A s t h e p r i n c i p a l time-varying
element therefore , a more de ta i l ed study of such devices, p a r t i c u l a r l y
from a p r a c t i c a l poin t of view, seems warranted. A spec ia l problem is
involved where t u r n s r a t i o s change s ign during t h e passage of time.
The f a c t o r i z a t i o n theorems have presumably wider a p p l i c a b i l i t y
than t h e purpose f o r which they were used here, i . e . t h e quas i los s l e s s
synthesis . Two problems could p r o f i t a b l y be examined, t h e f ac to r ing
of degree one los sy s c a t t e r i n g matrices, and the p o s s i b i l i t y of fac-
to r ing renormalized quas i los s l e s s s c a t t e r i n g matr ices t o achieve a
lossy synthesis , a s may be done i n t h e t ime-invariant case, [ ~ e f . 321. Although many more suggestions could be made, s e e Ref. 37, we
content ourse lves with one more, which c e r t a i n l y has general system
t h e o r e t i c overtones. For obvious reasons we a r e never in t e res t ed i n
behavior a s f a r back a s t = -m; perhaps a shade l e s s obviously, we a r e
never i n t e r e s t e d i n behavior a s f a r ahead a s t = +m. We would not he
doing p o s t e r i t y an i n j u s t i c e by l imi t ing our cons idera t ion of devices 6 t o say 10 years pas t t h e present time. What then a r e t h e consequences
of r e s t r i c t i n g our considerat ions t o a f i n i t e r a t h e r than i n f i n i t e time
in terva l?
C. POSSIBLE APPLICATIONS
Venturing s t i l l f u r t h e r in to the world of speculat ion, we hesi-
t a n t l y o f f e r suggestions as t o the app l i ca t ion of t h i s theory.
Its use i n t h e study of commonly encountered time-varying c i r -
c u i t s seems possible, though more so f o r pass ive devices such a s a
varac to r diode, than f o r a c t i v e ones such a s a frequency converting
heptode. Control systems problems may be a t tacked wi th some of t h e
techniques used i n our study, and have already been successfu l ly so
a t tacked i n two instances, a s repor ted i n chapter I X .
Just a s many t ime-invariant physical systems have been modelled
by e l e c t r i c a l c i r c u i t s because t h e problems as soc ia t ed wi th the l a t t e r
have been b e t t e r understood, so w e can suggest t h e same might happen
with time-varying phys ica l systems. While not f a r removed from t h e
idea of e l e c t r i c a l c i r c u i t s , t h e ionosphere with i t s time-varying
propagation c h a r a c t e r i s t i c s can perhaps be examined t h i s way. Again,
an i n t e r e s t i n g study would be a search f o r phys ica l systems which
could be modelled by nondiss ipa t ive networks and described by quasi-
l o s s l e s s s c a t t e r i n g matr ices.
The problems a r e obviously many, but t h i s is only r i g h t ; i f we
ever th ink t h a t we have run out of problems t o solve, we w i l l have
l o s t t h e a b i l i t y t o so lve any problems a t a l l .
Ah Love! Could you and I but wi th Fate conspire
To grasp t h i s so r ry scheme of th ings e n t i r e ,
Would not we s h a t t e r it t o bits--and then
Remould it nearer t o t h e Hear t ' s Desire .
Omar Khayyam, The Rubaiyat
REFERENCES
1. B. D. Anderson and R. W. Newcomb, "On Reciprocity and Time-Variable ~etworks," Proceedings of the IEEE, vol. 53, No. 10, October 1965, p. 1674.
2. L. Zadeh, "~ipe-Varying Networks, I," Proceedings of the IRE, vol. 49. No. 10, October 1961, pp. 1488-1503.
3. L. Zadeh, "The Determination of the Impulsive Response of Variable Networks, "Journal of Applied Physics, vol. 2 1 July 1950, pp.642-645.
4. L. Zadeh, " c i r cu i t Analysis of Linear Varying-Parameter ~etworks," Journal of Applied Physics, vol. 21, November 1950, pp. 1171-1177.
5. S. Darlington, "~ime-variable Transducers," Proceedings of the Brooklyn Polytechnic Symposium on Active Networks and Feedback Systems, vol. 10, pp. 621-633, 1960.
6. R. W. Newcomb, "The Foundations of Network Theory," The In s t i t u t i on of Engineers Australia, E lec t r ica l and Mechanical Engineering Trans- actions, vol. EM-6, No, 1, May 1964, pp. 7-12.
7 . D. A. Spaulding and R. W. Newcomb, "Synthesis of Lossless Time-Varying ~etworks," ICMCI Summaries of Papers, Part 2, Tokyo, September 1964, PP. 95-96.
8. B. D. Anderson, D. A. Spaulding and R. W. Newcomb, "useful. Time-Variable Circuit-Element Equivalences," Electronics Let ters , vol. 1, No. 3, May 1965, PP. 56-37.
9. D. A. Spaulding, "passive Time-Varying Networks," Ph.D. Dissertation, Stanford University, January 1965.
10. R. W. Newcomb, "on the Definition of a ~etwork," Proceedings of the IEEE, vol. 53, No. 5, May 1965, pp. 547-548.
11. B. D. Anderson, roperti ties of Time-Varying N-Port Impedance Matrices, 11
Proceedings of the Eighth Midwest Symposium on Circuit Theory, June 1965, pp. 8-0 - 8-13.
12. B. D. Anderson, D. A. Spaulding and R . W. Newcomb, "The Time-Variable ~ransformer," Proceedings of the IEEE, vol. 53, No. 6, June 1965, p.634.
9 ,
13. D. A. Spaulding and R. W. Newcomb, ne he Time-Variable Scattering Matrix, Proceedings of the IEEE, vol . 53, No. 6, June 1965, pp. 651-652.
14. B. D. Anderson and R. W. Newcomb, "on Relations Between Series-and Shunt- Augmented ~etworks," Proceedings of the ,-, vol. 53, No. 7, Ju ly 1965, P. 725.
- 15. B. D. Anderson and R. W. Newcomb, "Functional Analysis of Linear Passive Net- works," International Journal of Engineering Science, ( t o ' be published).
B. D. Anderson and R. W. Newcomb, "An Approach to the Time-Varying Sensi t ivi ty problem," Report No. SEL-66-06, (TR No. 6560-l), Stanford Electronics Laboratories, Stanford, California.
B. D. Anderson, " sens i t i v i ty Improvement Using Optimal Design, I,
Proceedings of the IEE, ( t o be published).
L. Schwartz, Thgorie des D i s t r i b u t i o n ~ ~ v o l . I, Hermann, Paris, 1957.
L. Schwartz, ~he*orie des Distributions, vol. 11, Hermann, Paris, 1959.
L. Schwartz, "Thgorie des noyaux, "proceedings of t he International Congress of Mathematicians, Cambridge, Mass., 1950, pp.220-230.
M Bunge, "causali ty, Chance and ~aw," American Scient is t , vol. 49, NO. 4 December 1961, pp. 432-448.
F. Riesz and B. Sz-Nagy, Functional Analysis, Ungar, New York, 1955.
G. F. Simmons, Topology and Modern Analysis, McGrawHill, New York, 1963.
B. Mdil lan, "Introduction t o Formal Realizabil i ty ~heory," Bell System Technical Journal, vol. 31, NO. 2, March 1952, pp. 217-279; and vol. 31, No. 3, May 1952, pp. 541-600.
9 , R. J . Duffin and D. Hazony, h he Degree of a Rational Matrix Function,
Journal of the Society f o r Industrial and Applied Mathematics, vol. 11, No. 3, September 1963, pp. 645-658.
R. E. Kalman, "Algebraic Structure of Linear Dynamical Systems," ( t o be published).
B. D. Anderson, "Cascade Synthesis of Time-Varying Nondissipative N-~orts ," (submitted f o r publication).
R. W. Newcomb and B.D. Anderson, "The Degree of a Matrix of Separable Time- Variable Impulse Response Functions," ( i n preparation).
B. D. Anderson and R. W. Newcomb, "on the Cascade Connection f o r Time- Invariant N-Ports," Proceedings of the IEE, to ,be published.
B. D. Anderson and R. W. Newcomb, "On Reciprocity i n Linear- Time-Invariant Networks," Report No. SEL-65-101 (TR No. 6558-3), Stanford Electronics Laboratories, Stanford, California, November 1965.-
B. D. Anderson and R. W. Newcomb, "Apparently Lossless Time-Varying Net- works," Report No. SEL-65-094 (TR No. 6759-1)) Stanford Electronics Laboratories, Stanford, California, September 1965.
32. V. Belevitch, '"Factorization of Scattering Matrices with Applications t o Passive Network Synthesis, " Phil ips Research Reports, vol. 18, No. 4,. ~ u g u s t 1963, pp. 275-317.
33. F. R. Gantmacher, The Theory of Matrices, Chelsea Publishing Company, New York, 1959.
34. J. J. Levin, "On the Matrix Riccati Equation," Proceedings of the American Mathematical Society, vol. 10, 1 , pp. 519-524.
35. R. W. Newcomb and B. D. Anderson,"Degenerate Networks," Prooeedings of the IEEE, to be published.
36. R. W. Newcomb, Linear Multiport Synthesis, McGraw-Hill, 1966.
37. Proposal to A i r Force Office of Sc ien t i f ic Research f o r "Functional Network Analysis and Synthesis," Stanford Electronics Labra tor ies , SEL 8-65, December 1965.
APPENDIX. VERIFICATION O F ADMITTANCE MATRIX FOWlJLATION
The object i s t o identify t he scattering matrix and admittance
matrix of Eqs. (7.18) and (7.19) as corresponding t o the same network.
By v i r tue of the resu l t of Eq.(2.8a)
i t i s suf f ic ien t to show that
where s and y are given by Eqs.(7.18) and (7.19). We have by - - di rec t calculation:
Now,
and thus,
y o s = fe( t)
u( t -7) 3 7 )
- - j r 1-
= - Y + (6An - z) - 22
That is,
JOINT SERVICES ELECTRONICS PROGRAM STANFORD ELECTRONICS LABORATORIES
REPORTS DISTRIBUTION LIST October 1'966
DEPARTMENT OF DEFENSE
D r . Edward M. R e i l l e y A s s t . D i r e c t o r ( ~ e s e a r c h ) O f f i c e of Defense Res & Eng Department of Defense Washington, D.C. 20301
O f f i c e of Deputy D i r e c t o r ( ~ e s e a r c h and In fo rmat ion Rm 3 ~ 1 0 3 7 ) Department of Defense The Pentagon Washington, D.C. 20301
D i r e c t o r Advanced Resea rch P r o j e c t s Agency Department of Defense Washington, D.C. 20301
D i r e c t o r f o r M a t e r i a l s S c i e n c e s Advanced Research P r o j e c t s Agency Department of Defense Washington, D.C. 20301
N a t i o n a l S e c u r i t y Agency A t t n : R4-James T i p p e t
O f f i c e o f Resea rch F o r t George G. Meade, Maryland 20755
C e n t r a l . I n t e l 1 i g e n c e Agency At tn : OCR/DD P u b l i c a t i o n s Washington, D.C. 20505
I n s t i t u t e f o r Defense Analyses 400 Army-Navy D r i v e A r l i n g t o n , V i r g i n i a 22202
O f f i c e of t h e D i r . o f Defense Resea rch and E n g i n e e r i n g Advisory Group on E l e c t r o n Devices 346 Broadway, 8 t h F l o o r New York, New York 10013
DEPARTMENT OF THE AIR FORCE
Headquar te r s AFRSTE Defense Communications Agency (333) Hqs . USAF The Pentagon Room ID-429, The Pentagon Washington, D.C. 20305 Washington, D.C. 20330
Defense Documentation C e n t e r AUL3T-9663 A t t n : TISIA Maxwell AFB, Alabama 36112 Cameron S t a t i o n , B ldg . 5 Alexandr ia , V i r g i n i a 22314 ( 2 0 c y s ) AFFTC (FTBPP-2)
T e c h n i c a l L i b r a r y D i r e c t o r Edwards AFB, C a l i f o r n i a 93523 N a t i o n a l S e c u r i t y Agency A t t n : L i b r a r i a n C-332 Space Systems D i v i s i o n F o r t George G. Meade, Maryland 20755 A i r F o r c e Systems Command
A t t n : SSSD Weapons Systems E v a l u a t i o n Group Los Angeles A i r F o r c e S t a t i o n A t t n : Co l . F i n i s G . Johnson Los Angeles , C a l i f o r n i a 90045 Department of Defense Washington, D.C. 20305
NOTE: 1 CY-TO EACH ADDRESSEE UNLESS OTHERWISE INDICATED
Page 1 O N R ( S ~ ) 10/66
SSD (SSTRT/L~. s t a r b e c k ) AFETR (ETLLG-1 ) AFUPO STINFO O f f i c e r (For ~ i b r a r y ) Los Angeles, C a l i f o r n i a 90045 P a t r i c k AFB, F l o r i d a 32925
DET #6, OAR (LOOAR) A i r Force Uni t P o s t O f f i c e Los Angeles, C a l i f o r n i a 90045
Systems Engineer ing Group (RTD) Techn ica l I n f o r m a t i o n Refe rence
Branch Attn: SEPIR D i r e c t o r a t e of E n g i n e e r i n g S tandards
& Techn ica l I n f o r m a t i o n Wright-Pat terson AFB, Ohio 45433
ARL (ARIY) Wright-Pat terson AFB, Ohio 45433
AFAL (AVT) Wright-Pat terson AFB, Ohio 45433
AFAL (AWE/R.D. L a r s o n ) Wright-Pat terson AFB, Ohio 45433
O f f i c e of Research A n a l y s i s At tn : Techn ica l L i b r a r y Branch Holloman AFB, New Mexico 88330
Commanding Genera l At tn : STEWS-WS-VT White Sands Missile Range New Mexico 88002 (2 CYS)
RADC (EMLAL-1) Attn: Documents L i b r a r y G r i f f i s s AFB, New York 13442
Academy L i b r a r y (DFSLB) U.S. A i r Force Academy Colorado 80840
FJSRL USAF Academy, Colorado 80840
APGC (PGBPS-12) E g l i n AFB, F l o r i d a 32542
AFETR Techn ica l L i b r a r y (ETV, MU-135) P a t r i c k AFB, F l o r i d a 32925
AFCRL (CRMXLR) AFCRL Research L i b r a r y , S top 29 L.G. Hanscom F i e l d Bedford, Mass. 01731
Department of t h e A i r Force AFSC S c i e n t i f i c and Technical
L i a i s o n O f f i c e 111 E a s t 1 6 t h S t r e e t New York, N e w York 10003
ESD (ESTI) L.G. Hanscom F i e l d Bedford , Mass. 01731 (2 c y s )
AEDC (ARO; 1 n c . ) A t t n : ~ i b r a r ~ / ~ o c u m e n t s Arnold AFS, Tenn. 37389
European O f f i c e of Aerospace Res. S h e l l B u i l d i n g 4 7 Rue C a n t e r s t e e n B r u s s e l s , Belgium ( 2 c y s )
L t . C o l . R. K a l s i c h C h i e f , E l e c t r o n i c s Div. (SREE) D i r e c t o r a t e of E n g i n e e r i n g S c i . A i r F o r c e O f f i c e of S c i . Res. 1400 Wilson Boulevard A r l i n g t o n , V i r g i n i a 22209 ( 5 c y s )
Department of t h e A i r Force AF F l i g h t Dynamics Labora to ry At tn: FDCC Wrigh t -Pa t t e r son AFB, Ohio 45433
DEPARTMENT OF THE ARMY
U.S. Army Research Off i c e At tn : P h y s i c a l S c i e n c e s D i v i s i o n 3045 Columbia P i k e A r l i n g t o n , V i r g i n i a 22204
Research P l a n s O f f i c e U.S. Army Research O f f i c e 3045 Columbia P i k e A r l i n g t o n , V i r g i n i a 22204
Page 2 o ~ R ( 8 3 ) 10/66
Commanding General U.S. Army Mater ie l Command At tn : AMCRD-RS-PE-E Washington, D.C. 20315
Commanding General U.S. Army S t r a t e g i c Communicatiorls
Command Washington, D.C. 20315
Commanding O f f i c e r U.S. Army Ma te r i a l s Research Watertown Arsenal Watertown, Massachusetts 02172
Commanding O f f i c e r U.S. Army B a l l i s t i c s Research Lab At tn : V.W. Richards Aberdeen Proving Ground Aberdeen, Maryland 21005
Commandant U.S. Army A i r Defense School At tn : Mis s i l e Sciences Div.
C & S Dept. P .O. Box 9390 F o r t B l i s s , Texas 79916
Commanding General U.S. Army Mis s i l e Command Attn: Technical L ib ra ry Redstone Arsenal , Alabama 35809
Commanding General Frankford Arsenal At tn : SMUFA-L6000 ( ~ r . S idney ROSS)
Ph i l ade lph i a , Pa. 19137
U.S. Army Munitions Command At tn : Technical Informat ion Branch P i ca t i nney Arsenal Dover, New J e r s e y 07801
Commanding O f f i c e r Har ry Diamond Labora tor ies At tn : Microminiatur izat ion Branch Connect icut Ave. and Van Ness S t r e e t
N . W . Washington, D .C . 20438
Commanding O f f i c e r U.S. Army Secu r i t y Agency Ar l ing ton Ha l l Ar l ing ton , V i rg in i a 22212
Commanding O f f i c e r U.S. Army Limited War Laboratory Attn: Technica l D i r e c t o r Aberdeen Proving Ground Aberdeen, Maryland 21005
Commanding O f f i c e r Human Engineer ing Labo ra to r i e s Aberdeen Proving Ground, Maryland 21005
Di r ec to r U.S . Army Engineer Geodesy, I n t e l l i g e n c e and Mapping Research and Development Agency Fo r t Be lvo i r , V i r g i n i a 22060
Commandant U.S. Army Command and General
S t a f f Col lege Attn: S e c r e t a r y Fo r t Leavenworth, Kansas 66270
D r . H . Robl, Deputy Chief S c i e n t i s t U.S. Army Research O f f i c e ( ~ u r h a m ) Box CM, Duke S t a t i o n Durham, North Ca ro l i na 27706
Commanding O f f i c e r U . S . Army Research Off ice ( ~ u r h a m ) Attn: CRD-AA-IP ( ~ i c h a r d 0. ~ l s h ) Box CM, Duke S t a t i o n Durham, North Ca ro l i na 27706
Superintendent U .S . M i l i t a r y Academy West P o i n t , New York 10996
The Walter Reed I n s t i t u t e of Res. Walter Reed Medical Cente r Washington, D . C . 20012
Commanding O f f i c e r U.S. Army Engineer ing R & D Lab. Attn: STINFO Branch Fo r t Be lvo i r , V i r g i n i a 22060
Dr. S . Benedict Levin, D i r ec to r I n s t i t u t e f o r Exp lo ra to ry Research U.S. Army E l e c t r o n i c s Command Fo r t Monmouth, New J e r s e y 07703
Page 3 O N R ( S ~ ) 10/66
D i r e c t o r I n s t i t u t e f o r Explora tory Research U.S. Army E l e c t r o n i c s Command At tn : M r . Robert 0 . Parker
Execut ive Sec re t a ry , JSTAC (AMSEL-M-D)
F o r t Monmouth, New J e r s e y 07703
Commanding General U.S. Army E l e c t r o n i c s Command F o r t Monmouth, New J e r s e y 07703 At tn : AMSEL-SC
RD-D RD-G RD-GF RD-MAF-1 RD-MAT XL-D XL -E XL-C ILL -S HL-D HL-L HL-J HL-P HL-0 HL-R NL-D NL-A NL-P NL-R NL-S KL-D KL-E KL-1 KL-S KL-T VL-D WL-D WL -C WL-N WL-S
NOTE: 1 CY TO EACH SYMBOL LISTED INDIVIDUALLY ADDRESSED
Commanding O f f i c e r U.S. Army Combat Development Command
Communications-Electronics Agency CDC-ECOM Liaison O f f i c e F o r t Monmouth, New J e r s e y 07703
DEPARTMENT OF THE N A W
Chief of Naval Research Attn: Code 427 Washington, D.C. 20360 ( 3 c y s )
Chie f , Bureau of Weapons Department of t h e Navy Washington, D.C. 20360 ( 3 cys )
Commanding O f f i c e r Off ice of Naval Research Branch
Off i c e Box 39, Navy No. 100 F.P.O. New York, New York 09510 (2 c y s )
Of f i ce of Naval Res. Branch Of f i ce At tn : Deputy Chief S c i e n t i s t 1076 Mission S t r e e t San Franc isco , C a l i f o r n i a 94103
Commanding O f f i c e r O f f i c e of Naval Research Branch
O f f i c e 219 South Dearborn S t r e e t Chicago, I l l i n o i s 60604 ( 3 c y s )
Commanding O f f i c e r O f f i c e of Naval Research Branch
Off i c e 1030 E a s t Green S t r e e t Pasadena, C a l i f . 91101
Commanding O f f i c e r Of f i ce of Naval R e s . Branch O f f i c e 207 West 24th S t r e e t New York, New York 10011
Commanding O f f i c e r Of f i ce of Naval R e s . Branch O f f i c e 495 Summer S t r e e t Boston, Massachusetts 02210
D i r e c t o r , Naval Research Laboratory Technical Informat ion O f f i c e r Washington, D . C . Attn: Code 2000 ( 7 c y s )
Code 5204
Commander Naval A i r Development and Mater ia l
Cente r J o h n s v i l l e , Pennsylvania 18974
Page 4 O N R ( S ~ ) 10/66
L i b r a r i a n U.S. Naval E l e c t r o n i c s L a b o r a t o r y San Diego, C a l i f o r n i a 95152 (2 cys)
Commanding O f f i c e r and D i r e c t o r U.S. Naval Underwater Sound Lab. F o r t Trumbull New London, C o n n e c t i c u t 06840
L i b r a r i a n U.S. Navy P o s t Graduate School Monterey, C a l i f o r n i a 93940
Commander U.S. Naval A i r Missile T e s t C e n t e r P o i n t Mugu, C a l i f o r n i a 93041
D i r e c t o r U.S. Naval Observa to ry Washington, D.C. 20390
Chie f of Naval O p e r a t i o n s OP-07 Washington, D.C. 20360 ( 2 c y s )
D i r e c t o r , U.S. Naval S e c u r i t y Group At tn : G43 3801 Nebraska Ave. Washington, D.C. 20390
Commanding O f f i c e r Naval Ordnance L a b o r a t o r y White Oak, Maryland 21502 ( 2 c y s )
Commanding O f f i c e r Naval Ordnance T e s t S t a t i o n China Lake, C a l i f o r n i a 93555
Commanding O f f i c e r Naval Avion ics F a c i l i t y I n d i a n a p o l i s , I n d i a n a 46206
Commanding O f f i c e r Naval T r a i n i n g Device Cen te r Or lando , F l o r i d a 32813
U.S. Naval Weapons Laboratory Dahlgren, V i r g i n i a 22448
Weapons Systems T e s t D i v i s i o n Naval A i r T e s t C e n t e r P a t u x t e n t R i v e r , Maryland 20670 A t t n : L i b r a r y
U.S. Naval Ordnance L a b o r a t o r y D i e l e c t r i c s & Semiconductors Branch Corona, C a l i f o r n i a 91720
Commander Naval S h i p Systems Command Main Navy B u i l d i n g Department o f t h e Navy Washington, D.C. 20360
Department of t h e Navy Naval E l e c t r o n i c Systems Command Semiconductors & M i c r o e l e c t r o n i c s
Uni t B a i l e y s C r o s s r o a d s , V i r g i n i a 22041
OTHER GOVERNMENT AGENCIES
M r . C h a r l e s F. Yost S p e c i a l A s s i s t a n t t o t h e D i r e c t o r
of Resea rch N a t i o n a l Aeronau t i c s and Space Admin. Washington, D.C. 20546
D r . H. H a r r i s o n , Code RRE C h i e f , E l e c t r o p h y s i c s Branch N a t i o n a l Aeronau t i c s and Space Admin. Washington, D.C. 20546
Goddard Space F l i g h t C e n t e r N a t i o n a l Aeronau t i c s and Space Ad. A t t n : L i b r a r y , Documents S e c t i o n
Code 252 G r e e n b e l t , Maryland 20771
NASA Lewis Research C e n t e r At tn : L i b r a r y 21000 Brookpark Road Cleve land , Ohio 44135
N a t i o n a l S c i e n c e Foundat ion At tn : Dr. John R. Lehmann
D i v i s i o n of E n g i n e e r i n g 1800 G S t r e e t , N.W. Washington, D.C. 20550
U.S. Atomic Energy Commission D i v i s i o n of Tech. I n f o r m a t i o n E x t . P.O. Box 62 Oak Ridge, Tenn. 37831
Page 5 oN~(83) 10166
The U n i v e r s i t y of Oklahoma Col lege of Engineer ing School of E l e c t r i c a l Engineer ing Norman, Oklahoma 73069
Hughes A i r c r a f t Company Hughes Research Labora tory 3011 Malibu Canyon Road Malibu, C a l i f o r n i a 90265
The Un ive r s i t y of Wisconsin Laboratory of Neurophysiology Medical School 283 Medical Sc i ences Bui ld ing Madison, Wisconsin 53706
Fede ra l S c i e n t i f i c Corpora t ion 615 West 131s t S t r e e t New York, New York 10027
The Un ive r s i t y of San t a C l a r a The School of Engineer ing Dept. of E l e c t r i c a l Engineer ing San ta C l a r a , C a l i f o r n i a 95053
Sprague E l e c t r i c Company Research and Development Labo ra to r i e s North Adams, Massachuset ts 01247
Worcester Po ly t echn i c I n s t i t u t e Dept. of E l e c t r i c a l Engineer ing Worcester , Massachuset ts 01609
Westinghouse E l e c t r i c Corporat ion Research Labo ra to r i e s Chu rch i l l Borough P i t t s b u r g h , Pennsylvania 15235
I n t e r n a t i o n a l Business Machines Corp. Monterey & C o t t l e Roads San J o s e , C a l i f o r n i a 95114
Un ive r s i t y of Notre Dame Dept . of E l e c t r i c a l Engineer ing Notre Dame, Ind iana 46556
Texas Ins t ruments , I n c . Technica l Reports Se rv i ce P.O. Box 5012 D a l l a s , Texas 75222
Radio Corpora t ion of America RCA Engineer ing L ib r a ry Moorestown, New J e r s e y 08057
Lockheed-Georgia Company Department 72-34, Zone 400 Research Informat ion Mar i e t t a , Georgia
Varian Assoc ia tes Technica l L i b r a r y 611 Hansen Way P a l o A l to , C a l i f o r n i a 94303
P r ince ton Un ive r s i t y School of Engineer ing and
Applied Sc ience
P r ince ton , New J e r s e y 08540
Vice P re s iden t B e l l Telephone Labo ra to r i e s , I nc . Murray H i l l , New J e r s e y 07971
Un ive r s i t y of I l l i n o i s Dept. of E l e c t r i c a l Engineer ing Urbana, I l l i n o i s 61801
M r . John J . S i e , P r e s iden t Micro S t a t e E l e c t r o n i c s Corp. 152 F l o r a l Avenue Murray H i l l , N e w J e r s e y 07971
Un ive r s i t y of Alber ta Department of Phys ics Edmonton, A lbe r t a Canada
S t a f f Vice P r e s i d e n t M a t e r i a l s and Device Research Radio Corpora t ion of America David Sarnoff Research Center P r i nce ton , New J e r s e y 08540
Informat ion Technology L ib r a ry AUERBACH Corpora t ion 121 North Broad S t r e e t P h i l a d e l p h i a , Pennsylvania 19107
Page 10 ONR( 83) 10166
P h i l c o Corpora t ion Module Engineer ing Department Mic roe l ec t ron i c s D iv i s i on 504 South Main S t r e e t Sp r ing C i t y , Pennsylvania 19475
Motorola, I n c . Government E l e c t r o n i c s D iv i s i on 8201 E a s t McDowell Road S c o t t s d a l e , Arizona 85252
L i b r a r i a n Microwave Assoc i a t e s , I n c . Northwest I n d u s t r i a l Pa rk Bur l ing ton , Massachuset ts 01803
P r o f e s s o r A . S . Vander Vors t Labo ra to i r e d l E l e c t r o n i q u e U n i v e r s i t e de Louvain 94, Kardinaal Merc ie r laan Hever lee - Leuven, Belgium
General E l e c t r i c Company Research and Development Center E l e c t r o n i c s Phys i c s Labora tory P.O. Box 8 Schenectady, New York 12301
United A i r c r a f t Corpora t ion Norden Div i s ion Technica l L i b r a r y Norwalk, Connec t icu t 06851
Watkins-Johnson Company 3333 Hi l lv iew Avenue S t an fo rd I n d u s t r i a l Park P a l o A l to , C a l i f o r n i a 94304
The Ohio S t a t e Un ive r s i t y D e p t . of E l e c t r i c a l Engineer ing 2024 Nei l Avenue Columbus, Ohio 43210
Un ive r s i t y of F l o r i d a Col lege of Engineer ing Dept. of E l e c t r i c a l Engineer ing G a i n e s v i l l e , F l o r i d a 32603
Sylvan ia E l e c t r o n i c Systems Western Operat ion Advanced Technology Labora tory P.O. Eox 2.05 Mountain View, C a l i f o r n i a 94042
The Un ive r s i t y of Western A u s t r a l i a Dept. of E l e c t r i c a l Engineer ing Nedlands, Western A u s t r a l i a
Radio Corpora t ion of America Manager of Engineer ing 415 South 5 t h S t r e e t Ha r r i son , New J e r s e y 07029
General Telephone & E l e c t r o n i c s Labo ra to r i e s , I nc .
208-20 Willets P o i n t Boulevard Bayside, New York 11360
The Boeing Company O f f i c e of t h e Vice P r e s i d e n t Research and Development P.O. Box 3707 S e a t t l e , Washington 98124
P ro fe s so r R.E. Vowels, Head School of E l e c t r i c a l Engineer ing The Un ive r s i t y of N.S.W. Box 1 P.O. Kensington, N.S.W. A u s t r a l i a
The George Washington U n i v e r s i t y School of Engineer ing and Applied S c i . Department of E l e c t r i c a l Engineer ing Washington, D.C. 20006
Georgia I n s t i t u t e of Technology P r i c e G i l b e r t Memorial L ib r a ry A t l an t a , Georgia 30332
Sandia Cor9ora t ion Technica l L i b r a r y P.O. Box 5800, Sandia Base Albuquerque, New Mexico 87115
Corne l l Aeronaut ica l Laboratory, I n c . of C o r n e l l U n i v e r s i t y P.O. Box 235 Buf fa lo , New York 14221
The Un ive r s i t y of Michigan E l ec t ron Phys ics Laboratory Dept. of E l e c t r i c a l Engineer ing 3505 Eas t Engineer ing Bui lding Ann Arbor, Michigan 48104
Page 11 O N R ( ~ ~ ) 10/66
AUTONETICS Research and E n g i n e e r i n g D i v i s i o n T e c h n i c a l In fo rmat ion Cen te r P.O. Box 4173 3370 Miraloma Avenue Anaheim, C a l i f o r n i a 92803
Mar t in -Mar ie t t a C o r p o r a t i o n Mart in Company, B a l t i m o r e D i v i s i o n Chief of L i b r a r i e s Ba l t imore , Maryland 21203
S p e r r y Gyroscope Company D i v i s i o n of S p e r r y Rand Corpora t ion Head, E n g i n e e r i n g Department Countermeasures E n g i n e e r i n g Grea t Neck, New York 11020
R e n s s e l a e r P o l y t e c h n i c I n s t i t u t e Dept. of E l e c t r i c a l Eng ineer ing R u s s e l l Sage L a b o r a t o r y Troy, New York 12181
Genera l E l e c t r i c Company L i a i s o n Eng ineer Computer L a b o r a t o r y P.O. Box 1285 Sunnyvale, C a l i f o r n i a 94088
U n i v e r s i t y of I l l i n o i s Dept . of Computer S c i e n c e D i g i t a l Computer Labora to ry Room 230 Urbana, I l l i n o i s 61803
U n i v e r s i t y of Chicago I n s t i t u t e f o r Computer Research 5640 South E l l i s Avenue Chicago, I l l i n o i s 60637
S c i e n t i f i c Advisor Lockheed E l e c t r o n i c s Company M i l i t a r y Systems P l a i n f i e l d , New J e r s e y 07060
C a l i f o r n i a I n s t i t u t e of Technology J e t P r o p u l s i o n L a b o r a t o r y L i b r a r y S u p e r v i s o r 4800 Oak Grove Dr ive Pasadena, C a l i f o r n i a 91103
E l e c t r o - O p t i c a l Systems, I n c . Micro-Data Opera t ions 300 North H a l s t e a d S t r e e t Pasadena, C a l i f o r n i a 91107
Manager Systems Engineer ing Radio Corpora t ion o f America Communications Systems D i v i s i o n F r o n t and Market S t r e e t s , Bldg. 17C-6 Camden, New J e r s e y 08102
E n g i n e e r i n g L i b r a r i a n Var ian A s s o c i a t e s LEL D i v i s i o n Akron S t r e e t Copiague, New York 11726
D r . Alan M. Schne ide r U n i v e r s i t y of C a l i f o r n i a , San Diego Department of t h e Aerospace and
Mechanical Eng ineer ing S c i e n c e s P.O. Box 109 La J o l l a , C a l i f o r n i a 92038
Page 1 2 O N R ( ~ ~ ) 10166