Microelectronics and Reliability Pergamon Press 1967.

Vol. 6, pp. 113-124. Printed in Great Britain

A M I N I M A L CAPACITOR CASCADE SYNTHESI S FO

R I N T E G R A T E D CIRCUIT S
R. W . NEWCOMB, T. N . RA O a n d ]'. W O O D A R D

Stanford Electronics Laboratories, Stanford, California

Abstract--Using a theory of non-reciprocal cascade synthesis based upon removal of even part
admittance zeros a synthesis of resistively terminated filters is developed with reference to
integrated realizations. The resulting structure is a cascade of capacitor-gyrator sections, with all
gyrators grounded, and using a minimum number of capacitors all of which can be chosen equal.

1. INTRODUCTION to the structures useful for integrated circuit syn-

ALTHOUGH fabrication and design techniques are thesis. I n particular this synthesis employs three
fairly well-developed for the basic building blocks terminal gyrators with a common ground, (the only
(transistors, resistors, capacitors, etc.) used in configuration physically available), and uses the
integrated circuits, m there are relatively few syn- minimum possible number of capacitors. I f so
thesis techniques available which are particularly desired, all capacitors can be chosen identical, by an
pertinent to integrated circuits. T h e problems of appropriate parameter choice in the gyrators.
linear integrated circuit synthesis are of course a s s o Similarly an arbitrary ratio of terminating resistors
c i a t e d w i t h o b t a i n i n g i n d u c t o r l e s s c can be accommodated.
o n - figurations; thus most of the results of active RC W e assume as given a rational positive-real input
synthesis ~2~ appear at first sight to be applicable. admittance y(p). Customarily we would be more
However, a closer look ~s~ reveals that, as with interested, for example, in the synthesis of a
lumped circuits, active RC techniques turn out to be p r e s c r i b e d t r a n s f e r v o l t a g e r a t i o m a g n i t
uneconomical because of the tolerances involved. u d e [V2(jto)/Vs(jto)] 2, under given resistive load and
Recognizing this latter fact in the lumped case, source terminations, as shown in Fig. 1. But,
Orchard ~*~ has proposed that the most practical of following standard theories (Ref. 12, p. 425) y(p) can
the inductorless synthesis tech-niques should result be obtained from [V2/Va [ through
by making gyrator-capacitor replacements of
inductors~S. 6~ in "classical" pas-sive designs. This 1-4o,
practicality is expected since good passive designs [~(jto)+a, ~ IV,(/to)l (1)
are relatively insensitive to changes in element values
and the introduction of the gyrator does little to where G t and G2 are the source and load con-d u c t
change this insensitivity.~" ances.Equation(1)assumesthat1
Because of the availability of gyrators for inte- >/4G2] V2/Vsl2/G1 for all real to, but, if such is
grated circuits ~s-m the idea of inductor replace-ment not the case a constant gain amplifier can be inserted
by gyrator-capacitor equivalents seems attractive for at the output at least in the cases of interest when
integrated circuits. However, mere replacements can V2(jto)/Vs(jto ) is finite for all real to. In a similar
often by quite wasteful of elements, so that it seems manner, if transducer gain, or the insertion power
more appropriate to develop new techniques which ratio, is prescribed y(p) can be obtained.
naturally fit the integrated philosophy. Consequently, Given then y(p) the philosophy is to realize zeros
we develop here a cascade RC-gyrator synthesis of the even part ofy(p ) by extracting factors from
appropriate y(p) by means of a cascade of capacitor
113
114 R. W. NEWCOMB , T, N, RAO and J. W O O D A R D

grounded-gyrator sections of degree one or two. chain matrices and conversion to Y through (2b)
The development follows that of Hazony (Ref. 13, gives the equivalence with Fig. 2(b), which is
pp. 130-135) but with emphasis on a structure of described by
primary interest to integrated circuits.

2. P l t E L I M I N A R I ~
As a preliminary to the synthesis, we review If, as shown in Fig. 3, a load of admittance Yz~) is
results to be used in the presented cascade method, connected at port two of a general 2-port, then
these results also serving to define the notation 0 r'~

t[ ii
used. We begin by considering a general 2-port
described by its admittance matrix Y(p) and its
chain (or transmission) matrix q(p), which, with . Yl l Y12
variables as illustrated in Fig. 1, are defined by
° [ LYe, ~*J I ^
II _ . - 12 Fxo. 3. Determination of y from Yl.

vs v~ 2 - pOR t vz one finds, through y=yxx--yl~(y;+yv,)-ly21, that the

O - input admittance y(p) is given by
Fro. I . T e n m n a t ~ structure for Av+ YnY~
(4a)
IT,/Vs. Y-- Yz+Y22

where Av=yny22--ylo, . . Solving for Yz gives

[ ~ = Y [vV~I, [Mx] = q [ V / J (2a) (4b)
and related through Av--ymy
y~ - - -
Y--Ylt
-- L--ill2 -I 'ql'~-l'qll J
LY21 y - m

,,= y ----~
[", I ",i, 1 = [ - y , - ' y . , - Y " - ' - ' ]( 2 c ) O O

Lmmt mz~l Lyl~-ytly21-1y+z yttyzl-iJ Fro. 4. Cascade gyrator removal.

For example, a gyrator is described by
Y=[Og g]
where g, the gyration conductance, is a real para-
meter. If such a gyrator is cascaded with an
inductor as shown in Fig. 2(a), multiplication of
For example the gyrator connection of Fig. 4 yields
(3a)
y=g~/y~ , y~=g~/y (5)
which shows that a preliminary gyrator extraction
can be used to change impedances into
admittances as well as inductors into capacitors.

! g g
0..----, * * - 0

c,~t

C c 0 O

to) (b]
Fro. 2. Useful gyrator equivalence.
A MI NI MAL CAPAC I T O R CASCADE SYNTHESIS
FOR I NTE G RATE D CIRCUITS 115

T o continue, we let a subscript asterisk, ,, denote 3. BASIC S E C T I O N - - G ~ REAL ZEROS T h e

replacement of p by -- p (Hurwitz con-jugation), a
superscript tilde, ~, denote matrix transposition, and
define the even part of a function y(p) through
2 Ev y=y+y° (6a)
2-port of Fig. 5 represents the section upon
which cascade synthesis is based. Its advantages for
integrated circuitry lie in the grounding of the gyrator
and the presence of a single capacitor.

It is important to note that if z=l/ y then y(k)lk

2 Evy=yy . Ev z (6b) o I(

Thus zeros of the even part of z are generally zeros of

the even part o f y and vice versa.
If ¥(p) is the admittance matrix of a lossless 2-port
then (Ref. 13, p. 1557
y ___ -- ¢ / . (7) Fro. 5. Basic cascade section.

For a lossless 2-port loaded as in Fig. 3, (4a) gives,

since Av----Av. By adding admittance matrices the 2-port of Fig.
--Y12Y2x Ev y~
5, called a basic section f o r y at p=k, is found
Ev y = (yz+Ym) (yz, +Y2~.) (8a)
to be
while in a similar manner (4b) gives
Y = Y(-~kk) [ P --P--k~ (9a)
--Y1~21 Ev y (8b) --p+k p
Ev y~ = (Y--Y11) (Y*--Yu*)
For comparison purposes where for now k is a positive and real number.
we note, from Applying (2c) we find, the chain matrix as
G2V,. = --13 and GtVs=II+G1Vx, that for Fig. 1 1[p k[y(k)]
v2 -G,y~I (8c) ~ = p_--Z--g (9b)
V, -- [yu+Gd [Y22+ G~]--Y12Y2t ky(k p d
Since zeros of Y2z appear in the numerator of (8a), )
(8b) and (8c) we conclude that zeros of
V2(p)/V~(p), zeros of transmission, are generally If this basic section is used as the 2-port in Fig. 3,
zeros of Evy(p). Consequently, recalling (1), a then (4b) shows, since Av=y~(k),
synthesis of lV2(jo~)/V,(jo)] can proceed by a yz(P) ky(k)--py(p)
cascade synthesis of y(p) based upon the removal of
zeros of Ev y. Equation (Sb) shows that original zeros y(k) - ky(p)--py(k) (10)
of E v y not extracted in a given cascade section
At this point we observe that yz(p), as given in (10),
remain available for extraction in later sections, thus
is positive-real when y(p) is since (10) defines a
giving the designer some freedom in the choice of the
Richards' transformation (Ref. 15, p. 779). Further, if
final structure through the order chosen for even part
k is chosen as a zero of Ev y(p) then the
zero removal. We note, from (6a), that the extraction
degree ofyl(p) is one less than that ofy(p) (Ref. 15, p.
of a right half p h n e zero of Ev y(p) also removes a
779). That is, if we denote the degree (Ref. 16, p.
left half plane zero of Evy(p). However, since we are
543) by 8[ ] then
treating non-reciprocal synthesis right half plane zeros
of transmission need not be left half plane zeros of a[y,(p)l=a[y(p)]-I if Evy(k)----0 (11)
transmission or vice versa, necessitating the
development of separate extractions for left and right Wefurtherobservethatwhenksatisfie
half plane zeros. s y(k) = -- y( -- k ) then so does --k, and, hence,
(11) still holds since yt(P) in (10) is still positive-
116 R. W. NEWCOMB, T. N. RAO and J. WOODARD

real. Thus, the section shown in Fig. 5 is realizable Yu(P) k, yz(kl)--PY~(P)

with the capacitor positive valued when k is either
a positive- or a negative-real zero of Ev y(p).
We conclude that given a rational positive-real
(12)
yl( kl) - - k~y~(2)--pyz( k~)
After substituting (10) into (12) and considerable,
but straightforward, manipulation one finds

kk t ky(k)--kly(kl) ' , ] y(k)y(kl) k2--kt ~

pa ky(k~)_kly(k) "v l Y(P) - - p ky(kl)--kty(k) (13)
Yu(P) = [kkt ky(k~)_kly(k ) .., ] y(p) k2_k12
[p2 ky(k)_lhy(k~) "t-~ p ky(k)--k~y(k~)

Similarly, by multiplying chain matrices as given by

y(p), if k is a real finite non-zero zero of Evy(p) we (9b) and converting to Y through (2b), one finds for
can extract the gyrator-capacitor section of Fig. 5 to the 2-port of Fig. 6, again after suitable manipulation,
1 Yl(kl)
lkklt,kl, kly,k
, kly,kl,[ ::]
m

Y=p ~ +~ k2--kt ~ y(k)

+ k__kl Yt(kl)

: :] y(k) [yt(kl)-J.•

(14)

At this point we let k be a non-real zero of Ev y, with

obtain the load admittance yt(P) as given by (10).
This extraction uses one capacitor and decreases the
degree of the admittance by one. Since con-neetion of
a realization for ys(P) as a load on the basic section
yields y(p) at the input, a repetition of the process
yields a cascade synthesis. However, for all-inclusive
synthesis we need to investigate other than real finite
and non-zero zeros of transmission.
Re k > 0 temporarily. The n we note that yz(P) is
positive (Ref. 17, p. 278) but not real. Still
(12) holds and thus yz is one degree less than y. But
if k is a zero of Ev y then so is the complex conjugate
k* since Ev y has real coefficients when y is positive-
real. Thus, by (8b), k ° is a zero of Evyz (which does
not have real coefficients). Choosing
kl=k * (15)

4. G E N E R A L C O M PLEX then guarantees that y u of (12) must decrease in

~ROS degree by one below that ofyt , or
T o proceed to the more general case let us make a
second extraction of a basic section for Yl at p-.-~kI, ~[yu(P)] = ~[Y(P)]--2 (16)
this to follow a basic section for y at p = k , as Further, by the above arguments Yu is positive. Using
shown in Fig. 6. Equation (I0) still holds, even (15) in (13) we find, recalling the positive-real
though we are going to allow complex k, and in a constraint y(k*)=y*(k),
similar way the load for the second section is
A M I N I M A L C A P A C I T O R C A S C A D E S Y N T H E S I S F O R I N T E G R A T E D C I R C U I T S 117

y(k)Iki( o Y{ (kl~Ik'

,o ,o
Y

FIG. 6. Basic section cascade for complex k.

[[k [ 2 Imky(k) ] [y(k)[ ' Imk~

p2 I m k y * ( k ) + 1 y(p) p Im ky*(k)
Ytz(P) = [[ k I' Im ky*(k) ] y(p) Im k' (17)
L P~ I m ky(k) + 1 p Im ky(k)

which shows that y u has real coefficients; y u is

then positive-real and of degree two less than y.
But with k complex the gyrators and capacitors
of Fig. 6 become complex valued and hence un-
realizable. Fortunately, however, Fig. 6 can be
replaced by a physically obtainable structure by
evaluating its admittance matrix, (14) in terms of k
t=k*
T o make Fig. 7 useful for integrated circuits
one notes that the inductor-transformer sub-2-port
can be replaced by the convenient capacitor
grounded-gyrator equivalence of Fig. 8.(b, 6) The
equivalence of Fig. 8 is easily checked by noting
that a transformer of turns ratio m.'l is equivalent

Y= P Imk2
ii] k Im ,k 1 Im ky(k)
Im ky'(k)

+p s
Im ky(k) J i m ky(k)] '

Im ky*(k) [Imky*(k)J
(18)
+ Imk

This admittance is realized by the two reactive

element circuit of Fig. 7, which by a somewhat
involved argument (Ref. 13, p. 137) is seen to have
positive inductance and capacitance. Figure 7 is
therefore a realizable equivalent for the cascaded
basic sections of Fig. 6. Figure 7 represents a g e n
eralnon-reciprocalDarlingtonsec
t i o n developed on an admittance basis (Ref. 13,
p. 134). It should be observed that Yzt of (17)
remains unchanged and Y of (18) is replaced by its
trans-pose when k is replaced b y its negative.
Thus zeros of transmission with Re k < 0 can
equally well be extracted by this method.
to a cascade of two gyrators as shown in Fig. 9 with
m=gl/g 2. Figure 9 is readily checked by com-
paring the chain matrices for (a) and (b), the latter
being obtained b y chain matrix multiplication.
Letting the capacitance of Fig. 8 be free to be
chosen in any convenient manner as K, Fig. 2
applies to fix gx~=K/l, thus giving the element
values of Fig. 8.
In summary, when k, with Re k=~0, is a zero of
the even part of y the degree two extraction of Fig.
7 can be made, with the replacement of Fig. 8
allowing a capacitor grounded-gyrator network
 118 R . W . N E W C O M B , T . N. R A O a n d J. W O O D A R D

lm ky(kl
Im k 2
o \ \ I( / /

Y(P)

o , , °

Fxe. 7. Realizable equivalent for Fig. 6.

.!

Ikl 2 lm ky*(k)
0 0

to) (b)
g> O FREE TO BE CHOSEN
OI • Ikl~/[K Im ky*(k))/tlm k2]"
gZ= [gl Im ky*(k)]/[Irn ky(k)]

FIO. 8. G r o u n d e d gyrator replacement for inductor and transformer of Fig. 7.

where the prime denotes differentiation, that

1,I
B
-
o
a+

Y pa[ I 1
m

(a) a-

(bl
FIG. 9. Transformer gyrator realization, m=gx/g2.
suitable for integrated circuit implementation. Th e
capacitance K can be chosen to insure that all
capacitors have identical value, as will later be
discussed.
S. IMAGINARY ZEROS
If k is purely imaginary the results of the last
section become indeterminate. However, a Brune
type structure can be obtained by differentiating
numerators and denominators of the various terms in
(14).Withk l = k * = - - k = ~ O satisf
y i n g y(kl) = --y(k) we find, on defining
2a+=y'(k) -t- y_~ ,.(k~2a-=y'(k) -- y(k)k
(19a)
a+
a_
(19b)
In this case Y is symmetric and positive-real since
a+ and a- are positive by Takahasi's Theorem (Ref.
18, p. 58). Equation (19b) is realized by Fig. 7,
without the gyrator and with a suitable change in
element values; the replacement of Fig. 8
therefore applies to give a realization suitable for
integrated circuits. T h e final structure, which is
practically useful for obtaining imaginary axis
zeros of transmission, is shown in Fig. 10. As in Fig.
8 we insert a parameter K > 0 which can freely be
chosen, say to equalize capacitance values, while
requiring the turns ratio, from (19b),
 A M I N I M A L C A P A C I T O R C A S C A D E S Y N T H E S I S F O R I N T E G R A T E D C I R C U I T S 119
o J(
y(p )
o o
K•0 ',
g.2 .- Klkl2a ;
2u+:y(k)
FIG. 10. Brune type section for imaginary k.
to equal the gyration conductanc e ratio, a+/a_ =
gJga.
Performin g the sam e differentiations of n u m e r
a - tors and denominator s on (13) we find tha t Fig.
10 has
a÷+z y(p)-
yu(P) ~___kI' a - ] y(p )
Lp~ ~++1 pa+
which, following previous arguments , is positive-real
an d o f degree two less tha n 8[y(p)].
I n the cases wher e y ( k ) = 0 these previous
argument s ru n into som e difficulty, as witness (10),
but, nevertheless, the final results remai n valid and
correspon d to removin g poles at 4 - k fro m z = 1/y.
U n d e r this condition, y ( k ) = 0, we not e tha t 2a+
= 2a_ = y ' ( k ) , whic h is still positive b y T a k a h a
s i ' s T h e o r e m , such tha t ga = g~ in Fig. 10.
Consequently, the transforme r in an equivalent
similar to Fig. 8 is 1 : 1 ; hence Fig. 10 is identical to
the capacitor, c = a+, in parallel with an induc - tor, l
= l/[k[ 'a _ . Further , f l y ( k ) = 0, (20a) gives
Y(P)Y'(k)[p+ I k ~
yzz(P) = [ p ' +[ k ]Z]y'(k)--2py(p)
and, since
y(p) = ( p + k ) y ' ( k ) + . . . . (p--k)y'(k)+
...,
one can check tha t ( P + I k ]92 cancels f r o m n u m
e r a t o r and denominato r such tha t 8[ylz] = 8[y] --
2 , as expected. I t should also be pointe d out tha t if
z(k) = 0 the n a preliminar y gyrato r can be extracted,
b y (15), to tur n the impedanc e into an admittanc e
(for which y(k) = 0).
6. ZEROS AT ~ ' ~ O AND INFINIT Y
Zeros of transmission, or zeros o f E v y, at zero
and, especially, infinity require somewha t special
04-
o
/ . ,,,,° ,
FREE TO BE CHOSEN
+ ~ki - g2: g,o_/o.
--~- ; 20_: y ( k ) - - ~
attention since the pertinen t cascade sections m a y
not have an admittance matrix, or the zeros m a y
com e fro m the denominato r of (8c). T h u s it is
importan t to note tha t such zeros are zeros of y(p )
or its inverse z ---- 1/5,. I f p = 0 is a zero o f y it is a
pole of z and can be extracted as a series capa - citor,
Fig. ll(a) . I f p = 0 is a zero of z it can be mad e a
(20a) pole of z b y a gyrator extraction and then the pole
can be remove d b y a series capacitor, as shown in
Fig. l l ( b ) . I n Fig. l l ( b ) the gyration
conductance g is arbitrary and can be chosen to
obtain a convenient capacitance value. I f p ---= oo is
a zero of z, it is a pole o f y and can be remove d b y
a shun t capacitor, Fig. 12(a). Finally a preliminar y
gyrato r extraction can change a zero o f y at infinity
into a pole of y yielding again a shun t capacitor
extraction, Fig. 12(b) where, as before, any
convenient value o f g can be chosen. Not e the
applicability of Fig. 2 to this extraction.
7. S U M M A R Y OF PROCEDURE
At this point we can summariz e the procedure .
Give n y(p ) one removes, in an y desired order, the
zeros of E v y(p) throug h the sections of Figs. 4-12. I
(20b) f y(p ) has been derived fro m a filter characteristic
for a circuit as shown in Fig. 1, the n removals are
mad e of zeros of E v y(p ) which are zeros of trans -
mission, it being observed that the negative of the
zero of E v y(p) wiU automatically be extracted in
the process. Eac h remova l need use only grounded -
gyrators and capacitors and each section reduces the
degree of the admittance b y the n u m b e r of
capacitors used. Repetition of the procedur e even -
tually leads to a zero degree function, which is a
resistor. Henc e the final structure is a cascade of
capacitor grounded - gyrato r sections terminate d in
a resistor and using the m i n i m u m n u m b e r , 8[.y],
of capacitors. ~zs~ T h e structure as such is ideal for
120 R. W. NEWCOMB, T. N. RAO and J. WOODARD

integrated circuits where it is of special interest to

note that all capacitors can be made identical by T o illustrate the details let us normalize the Brune
appropriate admittance normalizations. section of Fig, 10 such that both capacitors are of unit
capacitance. As shown in Fig. 13, we first extract the
8. NORMALI/ATION gyrator, such that we wish to apply the theory to y t =
T o introduce another parameter at almost any g2/y. Since k is assumed an imaginary zero of Ev y it
stage in the procedure a gyrator of gyration con- is also an imaginary zero of Ev yz and Fig. 10 applies
ductance g can be extracted as in Fig. 4. By (5) this directly, upon inserting subscript l's, to give a Brune
converts an admittance to an impedance and scales section pertaining to yv Since y ' t ( k ) = --
the impedance level. Since zeros of E v y are gZy'(k)/y2(k) > O, we have
generally zeros of Ev z, (6b), synthesis can still g2 (_..k . . . . g2
proceed from the admittance as described to this 2.at+ ----- -- ~ [y'(k) -- Y---))],,= y ' ( k ) 2a_
point. However, the inserted parameter allows the (21a)
scaling of one capacitor per section while a free
choice of K in Figs. 8 and 10 allows the scaling of the g2 + Y~)] = g2
second capacitor when two are present. In the case of 2a,_ ----- -- ~ [y'(k) - - y~(k) 2a+
~
Figs. ll(a) and 12(a), a preliminary gyrator extraction (21b)
requires a second gyrator extraction with Fig. 9
showing that the two gyrators are merely a gz2 a~_ a+ gl (21c)
transformer, clearly allowing capacitor sealing. In
conclusion, we see that by a simple preliminary gll al+ a_ g~
gyrator extraction per section, all capacitors can be g~lz = K [ hl2a~_. (21d)
chosen of equal capacitance.
Choosing unit capacitors requires az+ = K = 1 or

y,(O) g g2z,(O)
g~ = -- y~(k)/a - (22a)
l,t t/ g ~ , x = t k [ ~ a t - = - - ] k [2 g2a+ (22b)
y~=]k['~_
(el (b)
T h e final section is as shown in Fig. 13.
y(O): 0 z(O): 0
Such normalizations are especially convenient for
FIG. 11. Extraction of (transmission) zeros at zero. integrated circuits, since identical elements are quite
preferable for circuit layout and electronically
g adjustable gyrators are readily obtainable, c9) In
" O 0 0
actual fact one could set up standard sections, as Fig.
10 for real frequency zeros of transmission, which
can be electricaUy adjusted, by gyrator variation, to
(o ) (b ) obtain desired zeros of transmission (and with fixed
z(OO) = 0 y(m) = 0 capacitors). An interesting appli-cation, for example,
would be to the synthesis of variable bandwidth
FIG, 12. Extraction of (transmission) zeros at infinity. filters.

g=lyck)l/~
o

y---...

0 i'o
y~: oz/y
+'Pq
FIo. 13. Normalization of a Brune section to unit capacitors.
A M I N I MA L CA PAC I T O R CA S CAD E S Y N T H E S I S F O R I N T E G RAT E D C I R C U I T S 121

9. EXAMPI.~'~ Fig. 8, where K = 0.2308 is chosen to give identical

Here we consider three examples which illus-trate capacitors, is then as shown in Fig. 15.
almost all the points of the theory. Example 3 (imaginary zeros). Let it be
Example 1 (real zeros). T h e admittance desired to design a low-pass, doubly-terminated,
third-order Cauer filter with an effective passband
3p~+12p+9 ripple of 0.18dB. Such a filter is dimensioned by
y(p) -- 9p2+25p+19 (23a) standard techniques by Saal and Ulbrich (Ref. 20, p.
has zeros of 304) from which we can derive, for (1), either
Ev y(p) = 913p4 - 1 8 p 2 + 19]/[81p4--2183f-+
2pa+ 0.992009p~+ 1-035852p+ 0.276463
361] at p = +2.1524, and ±1.1692. Choosing, arbi-
trarily, or perhaps for a desired zero of trans-mission, y(p) = 0.992009p~+ 0.489186p + 0.276463
(253)
k -- --2-1524 (23b) or y as the reciprocal of the right side of (25a). This
yields low-pass filter has a zero of transmission at infinity,
y(k) ----- --0.4255 and by construction there are zeros of transmission at
and, from (10), p ---- -4-j1"912730. Proceeding from
1"2766 p+1"2521 (23c) (253) the zero of transmission at infinity (pole of y at
Yz(.P) = 3-8~81Y oo) is easily extracted yielding

Repeating with k = + 1.1692, which has yt(k) = 0- y~(p) = y(p)--2.0161 lp --

3570 and Y n ( P ) = 0.3333 yields the circuit of Fig. 0.005756p2+ 0.478472p+ 0.276463
14, which then illustrates the extraction of = 6.992009p=+0.489186p+0.276463. (25b)
0.1977f 0-5053f

/\ =

o
\ )l

o
0,~570U 0.3333ff

FIG. 14. Real zero example circuit.

positive and negative real zeros of Ev y to produce Then Yx(P) has a zero of its even part at
transmission zeros at these frequencies.
Example 2 (complex zeros). T h e admittance k = k*1 = - - k I = j1"912730 (25c)

3p=+12p+9 from which we find y l ( k ) = --j0.272959,

y(p) = 6p=+13p+10
has zeros of
Evy(p ) = 18[p4--4p2+5]/[36p'--49p2+ 100]
at p = q-[1.4553+j0.3436]. Arbitrarily choosing
k---- 1'4553--j0"3436
requires the choice of k t = k* = 1.4553+j0.3436 and
yields y(k) = 0.7864+j0.0271 for which we have k 2
= 2 - - j l , Im ky(k) = --0.2308, Im ky*(k)
- --0.3096, and l y(k) [~ - - 0.6192. Equation (17)
is then yu(P) ---- 0-5000. Figure 7, combined with
(24a) y'l(k)
= 0.1662409 giving, by (193) applied to Yl, a+ =
0.011768, a_ = 0.154474, and, by (203), gu(P) =
0.005803. Figure 16 then results from Fig. 10, with K
> 0 any desired value. Th e con-venient cascade
(24b) structure of the synthesis which uses a minimum
number of capacitors should be observed; note that
one reactive element is saved over the conventional
design (Ref. 20, p. 306).
Since all the capacitors of Fig. 16 are not of equal
size it is of interest to illustrate a normali-zation
procedure. We could insert a gyrator after the left-
hand capacitor of Fig. 16 and apply (21)
122 R.W. NEWCOMB, T. N. RAO and J. W O O D A R D

0"2308 f

-1_ °'}:'~9-363L--/ o , ooou

, /
0-6717tt
Fro. 15. Complex zero example circuit,

1 O.OIISf
÷

9.es9JR u

+F
vz

Y Yl Y

FIG. 16. Realization of a third-order Cauer filter.

to obtain any desired values, by a choice of K and g,
for the right-hand two capacitors. I f the left-hand
capacitor is desired to be changed in value it can be
preceded by a transformer, which also scales all
impedance levels to the right and which can be realized
by Fig. 9. A more appealing approach, since it saves one
gyrator, is to begin with
0.992009p ~+ 0.489186p+ 0.276463
y(p) = 2p3 + O.992009p~+ l.O3 5852p+ 0.276463
(26)
(the inverse of (25a) which yields the same I
V2(jo)/Vs(jo) ~. This aUows the extraction of Fig.
12(b) with g a parameter which can be freely chosen. At
this point the remaining admittance is g~Yl(P), where
Yl(P) is given by (25b). Another gyrator can then be
extracted to finally yield
Fig. 17 with the element values found by applying
(21) with g replaced by gn. Note, then, that with the
three parameters g, gn and K free to be chosen one
can obtain any desired values of capacitance, in
particular all capacitors can be chosen equal and of
any size readily available. I f the final load resistor is
desired to be equal to the source resistance this can
be also obtained by inserting a gyrator before G~
(since a gyrator loaded in a resistor is another
resistor).
Figures 16 and 17 are normalized such that a
cut-off frequency of tOc = 0.59863 results with a
zero of transmission at toa = 1-91273 and a 1 ohm
source resistance. By standard frequency and
impedance level scaling any desired cut-off
frequency and source resistance can of course be
obtained.

"4
1 +

G2f V2
Ci= 2'016gz C3= Kgn2/g2 g2 = 0"076'gI
¢2 = 2.073gn2'tJ gl2= 0.577BKg4/g ` G2= '72"3g2/g

Fzo. 17. Equivalent realization of Fig. 16 with arbitrary capacitance values.

A M I N I MALCAPAC I T O R CAS CADE
T E G R A T E D C I R C U I T S 123
10. DISCUSSION
By revising, and extending portions of, the cas-cade
synthesis of Hazon y a synthesis of driving-point
admittances, and, with these, voltage-transfer filters, is
obtained. Because the synthesis uses in a natural wa y
only capacitive reactive elements, and of these a m i n i
m u m n u m b e r all of which can be chosen identical,
the method is ideally suited to the synthesis o f linear
integrated circuits. Of special interest are the sections o f
Figs. 10 and 11 which yield real frequency (p ----j~ )
zeros o f transmission, such zeros being the ones of most
interest in usual filter designs.
T h e techniques rest u p o n the availability of
gyrators in integrated form. But, when one places the
restriction that all gyrators have a c o m m o n groun d
(for the purpose of biasing transistors in available
gyrator realizations), gyrators in integrated form seem
readily available at least for frequencies below the
middle megacycle range. Indee d one is fortunate in
having electronically adjustable gyrators making it
possible to easily accommodate the identical capacitors
so convenient for integrated
use.
I f one investigates the calculations involved one sees
that for any really useful filter the number s need to be
carried to m a n y significant figures. Yet this is c o m m
o n to any accurate filter design and points to the
advantages o f a computerized society. I t should be
observed, however, that if y(p) is originally lossless, then
y ( p ) + y ( - - p ) ~ 0 and all p are zeros of the even part.
Consequently, any convenient k can be chosen for the
development with k = 0 or oo leading to the standard
Caner forms u p o n capacitor-gyrator replacements of
inductors. I n other than the lossleas case the main
freedom available to the designer is the choice of order o
f removal o f the sections, with a judicious choice wort h
searching for.
T h e method given is of course valid only for scalar
y(p). I f it is desired to synthesize matrices then it m a y
be well-worth developing the present n - port scattering
matrix cascade methods tx~, ~x~ to fit integrated
circuits.
Likewise, since recent results in time-variable
synthesis ~2z~ call heavily u p o n the time-variable
gyrator, which should be readily available in integrated
form, extensions of the theory unde r present
investigation should yield practical means for time-
variable circuit design.
SYNTHESIS FOR IN
Acknowledgements -- The authors are indebted to
BARBARA SES~A~O for her careful preparation of the
manuscript and to the critical discussions carried on by
the Stanford integrated circuits class EE292y. This work
was supported by the U.S. National Science Foundation
under Grant NSF GK-237.
REFERENCES
1. R. M. WAR,mR, JR. (Editor), Integrated Circuits.
McGraw-Hill (1965).
2. K. L. Su, Active Network Synthesis. McGraw-Hill
(1965).
3. M. OrvN~a~, Integrierte Resonanzverst~rker ohne
Induktivit~ten. Microelectron. & Relish. 5, 57 (1966).
4. H. J. ORCHARD,Inductorless filters. Electron. Lett.
2, 224 (1966).
5. B. D. ANDva~so.% W. NEw and R. N~CCCOMB,
Proposed adjustable tuned circuits for micro-
electronic structures. Proc. IEEE 54, 411 (1966).
6. A. G. J. HOLT and J. TAYLOR,Method of replacing
ungrounded inductors by grounded gyrators.
Electron. Lett. 1, 105 (1965).
7. J. WOODARDand R. W. NEWCOMB,On the
sensitivity improvement of inductorless filters.
Stanford Electronics Laboratories Technical Report
No. 6558-11, June (1966).
8. W. NEw and R. W. NEWCOMB,An integratable
time-variable gyrator. Proc. IEEE 53, 2161 (1965).
9. T. N. RAO and R. W. N~VCOMB, A direct coupled
gyrator suitable for integrated circuits and time-
variation. Electron. Lett. 2, 250 (1966).
10. D. F. SHEAI~ANand H. J. ORCHARD,A high quality
transistorized gyrator. Electron Lett. 2, 274 (1966).
11. P. GARY,An integratable direct coupled gyrator, and
J. H. HAWLEY, JR., A gyrator realization using
operational amplifiers, contained in Papers on Inte-
grated Circuit Synthesis (compiled by R. NEWCOMB
and T. RAo). Stanford Electronics Laboratories
Technical Report No. 6560-4, June (1966).
12. M. E. VAN VAL~BURO, Introduction to Modern
Network Synthesis. John Wiley (1960).
13. D. HAZON¢,Elements of Network Synthesis.
Reinhold (1963).
14. Y. OONO and K. YASUURA, Synthesis of finite
passive 2n-terminal networks with prescribed
scattering matrices. Mera. Fac. Engng Kyushu Univ.
14, 125 (1954).
124 R. W. N E W C O M B , T . N. R A O a n d J. W O O D A R D
15. P. I. RICHARD$,A special class of functions with
positive real part in a half-plane. Duke Math. J. 14,
777 (1947).
16. B. M c M m L ~ , Introduction to formal realizability
theory -- II . Bell ,Syst. Tech. J. 31, 541 (1952).
17. V. BELEVlTCH,Factorization of scattering matrices
with applications to passive-network synthesis. Philips
Res. Rep. 18, 275 (1963).
18. H. T ~ H A S I , Theory of network synthesis, .7.
Phys.-Math. Soc. Japan 14, 51 (1940).
19. R. E. KALMAN', Irreducible realizations and the
degree of a matrix of rational functions..7. Soc. Ind.
Appl. Math. 13, 520 (1965).
20. R. SAALand E. ULBRICH, On the design of filters by
synthesis. I R E Trans. Circuit Theory 5, 284 (1958).
21. D. C. YOULA, Cascade synthesis of passive n-ports.
Polytech. Inst. Brooklyn, Technical Report Numbe r
RADC - TDR - 64 - 332, August 1964.
22. B. D . O. A_~o~aSON, Cascade synthesis of time-
varying nondissipative n-ports. (Submitted for
publication.)