IRE TRANSACTIONS—MICROWAVE THEORY AND TECHNIQUES

Planar Transmission Lines*

DAVID PARK~

Swnmary—This paper derives formulas for the transmission attenuation in the conductors is determined by P,, the
properties-characteristic impedance and attenuation—in the prin- power dissipated in them per unit length:
cipal mode of a transmission line consisting of one or two long strips
of metal foil embedded in a dielectric material between two long a = PC/2P, (2)
metal strips considerably wider than the central ones. The width and
spacing of the central strips is arbitrary, and it is also necessary to where
take account of their thickness in computing the attenuation. A
graphical method is given for evaluating the characteristic impedance
in general, and analytic approximations are given for a number of
P+: J] I E ‘ds, q = w’(mjW’ac) (3)

special cases. Finally the question of the leakage of power from be-
tween the outer strips is considered briefly. and the integration is over the boundaries of all the con-
ductors; f is the frequency, and ~c is the conductivity.
INTRODUCTION P represents the power flowing in the line. and is equal .
TTENTIOiN has been drawn recently to the possi- to
bility of constructing a transmission line in the P = v’/(2zJ, (4)
A
form of a sandwich, consisting of one or two
central conductors of metal foil between slabs of dielec- where V is the maximum instantaneous voltage across

tric, the whole inclosed by two broad plates, as shown the line. We shall not consider loss in the dielectric,

in cross section in Fig. 1. Where there are two strips though it is easy to do so. L

the voltage is applied between them and the top and In order to perform the calculations we shall carry

bottom plates are grounded, whereas when there is one, out a conformal mapping of the z-plane, with co-

the strip will be one side of the line and the two plates ordinates x and y (z =x+jy), which is the plane of Fig. 1,

will be the other. We shall assume that the strips are onto the w-plane, with w = u +jv =~(z). (The first step in

very thin. this mapping is shown in Fig. 2, opposite.) In the w-plane,

the lines of constant u are lines of force and those of con-
stant v are equipotentials. It is clear that Aq, the charge
per unit length contained between two points on a con-
ductor, is given in terms of the corresponding difference
in v by
-=++= Aq = eAv. (5)

To compute the attenuation we need the field strength

E, which in the z-plane is

r-----w --J E = –
———
dw/dz, (6)

where the bar denotes the complex conjugate. Then the

integral in (3) becomes

Fig. l—Cross sections of two transmission lines.

(7)

A compact formalism for performing the required

analysis in the complex plane has recently been given where ] d.. ] = <(d.Y2+dy2) is an element of the boundary
by Assadourian and Rimai, 1 to which the reader is re- curve C of the conductor, and this can be written as
ferred for further details. We shall determine 20, the
characteristic impedance, in terms of C, the capacitance (8)
per unit length of the line, by the relation, valid in the
TEM mode, where ] dw ] = ti(dua+dvz) is an element of the image

z, = d(/.K)/c. (I) curve C!.

(We use rationalized MKS units throughout.) The THE CONFORMAL TRANSFORMATION

We shall first consider the two-strip arrangement of

* Supported by the Sprague Electric Co., North Adams, Mass.
j’ Williams C@lege, Williamstown, Mass. Fig. 1(a); that of Fig. 1 (b) follows easily by letting d
1 F.. Assadourlan and E. Rlmal, “Simplified theory of microstrip
become infinite. A transformation corresponding to
transmission systems, ” PROC. IRE, vol. 40, pp. 165 1–1657; Decem-
ber, 1952. such an arrangement without the bounding planes was
 1955 Park: Planar Transmission Lines 9

W-PLANE I

z ‘/==’’A/vL--

“T’
,,
-)(-zJR’ +’ Z“f ‘L A NE
+ -1 -2iK’ o K-2jK’ 1- +1
t h=~rb 4

Z -F’L41rvE

)(

Fig. 2—’The Conformal Transformation. In the w-plane, numbers to

Fig. 3—The second mapping. Points labeled by
the left of points give the values of u+jzI, numbers to the right
Greek letters correspond.
give the corresponding values of sn w. Points labeled by Greek
letters on the two diagrams correspond.
the capacitance per unit length is
given long ago:z

(11)
z=asnw (9)

where sn w is the Jacobian elliptic function.3 The re- and the characteristic impedance, by (1), is
aping is shown in Fig. 2. The equipotentials and lines of
force form a rectangular grid in the w-plane where the
(12)
two sides of the right-hand strip in the z-plane are
mapped into the single line in the w-plane running from
But this is not the transmission line we want. We still
(~, –~.K’) to (~, +~~’), and similarly for the other
have to introduce the top and bottom plates, andl this
strip. Two lines of force, A and B, have been drawn on
can be done either by carrying out a new mappiug or
the left and their images given on the right. K and K’
by using Maxwell’s method of images.4 To do it the first
are the complete elliptic integrals of the first kind,
way, let us write instead of (9)
formed with the complementary modrdi k and k’ re-
spectively, where z’=jz=~asnw, (13)
~z + /2/2 = 1,
(lo) which rotates the z-plane counterclockwise by 9(II de-
grees, but in Fig. 3(a) we draw the configuration in the
and k is determined (cf. Fig. 2) by the width of the
z’-plane as a single strip and an infinite grounded plane,
strip. The total charge per unit length on one strip is e
the second strip now being merely the image of the first.
times the difference between the values of v at the be-
A simple Schwarz-Christoffel transformation
ginning and end of the image of the strip—clearly it is
2e K’. The potential difference is the difference in the
-!z jr
values of u belonging to the two strips, or 2K. Thus, z“ = b
( Cosh–l # —
) (14)

‘ J. J. Thomson, “Recent Researches in Electricity and Mag- now maps the upper half z’-plane into the inside of the
netism, ” The Clarendon Press, Oxford, Eng., p. 237; 1892.
s A convenient r6sum6 of these functions will be found in R. S.
Burington and C. C, Torrance, “Higher Mathematics, ” McGraw- 4 i(Treati~e on Electricity and Magnetism, ” Oxford LTniv(XsitY
Hill Book Co., Inc., New York, N. Y.; 1930. Press, New York, N. Y., 3rd cd., p. 310; 1904-1946.
10 IRE TRANSACTIONS—MICROWAVE THEORY AND TECHNIQUES April

polygon of Fig. 3(b). It also maps the strip as shown, In Fig. 4 we have plotted In I/q =$(k’) from the data

and this arrangement is clearly exactly equivalent to in the Jahnke-Emde tables, and from (12) and (19) we

that of Fig. l(a), so the problem is solved. From (14), now have two formulas for 20 in ohms:

!)
z 120f(k’2) 1207r2
sinh — = — jz’ Zo = (21)
b <; = f(k’)di ‘

or by (13) where we have set /(p/c) equal to 1207r//~, K being

Jr the dielectric constant of the central material. (The
.
sinh — = a sn w. (15) modification of this and following formulas required
b
when the dielectric has a magnetic permeability differ-

Comparing Fig. 3(b) with Fig. 1(a), we have by (9) ent from that of free space consists in writing ~~/Ko in-

and (15) stead of K throughout. ) In general, it is most convenient

to find k’ from the dimensions of the line and then read
da d’
a=~inh~, ~=sinh—~ h = :rb (16) 20 off the plot of Fig. 4, In special cases, however, the
b k Y 2

so that we can determine k from

sinh djb
k= (17)
sinh d’jb ‘ f(kz)

with which the solution is formally completed.

THE CHARACTERISTIC IMPED~~C~

It will often be convenient in what follows to work 0 0. I 0.20.30.40.50.60.7 12$ 0.9 1.0
p
with the ----
LULL1~LC1l
1 ,,-,.m.n,-.,
LC1l LCL1 j
_-J..l,,.-
L1lUUUIUC!
l,! , ~LVC1l
fi
-:.,fim h.,
U) (17) as
Fig. &—Plot off (hz) = In 1/g against kz.
d+d’ Note that ~(kz) = rr’/f (k”).
sinh ~ sinh —
b b
~~2=~_k2=— (18) expansion (20) can be used to advantage. Consider for
d’ “
sinhz — example what happens when the strips are very narrow,
b so that w<<b.8 From (18) we have

Making use of the complementarily of K and K’, we

#2
can now write (12) as =2; coth+,

K’

K)
P
ZIJ= --= (k’), (19) so that the first term of (20) gives [with (16)]

where
k’ as the
the sign
modulus
on the right
throughout.
means
The
that
quotient,
we are to
however,
take
‘a+(%anh$) (w<< L).

is known from the theory of &functions5 to be given by As a second example, suppose that the strips are close

K’ together (d<<b) but rather wide (d’>> b). Eq. (17) gives
~ in q = ~f(k’) (k),
7=–3r 7r d
k =— e-”/’ (d<< b<< d’) (22a)
where q is given6 by b

so that again the logarithm suffices, and we have

so that, taking the logarithm of this,’

K’

_iF=7r
~
(ln~–~k2–$k4–~kG–. . .
) . (20)

line
As a final
of Fig. l(b).
special
In (17)
case,
we let
let us consider
d+m, so that
the one-strip

5 E. T. Whittaker and G. N. Watson, “Modern Analysis, ” Cam-

bridge University Press, Cambridge, Eng., 4th ed.; 1927. The proof
of this formula, given as a problem on p. 479, follows immediately
from the definitions above it.
8 E. T. Whittaker and G. N. Watson, ibid. This follows at once s There will be a number of these “double inequalities” in what
from p. 486. The function g(k) is tabulated and plotted in Jahnke- follows. Here, w<<b means that zw/b is small enough so that higher
Emde, “Tables of Functions, ” Dover Publications, New York, N. Y., terms in it can be neglected. Since in nearly every case ratios of this
ch. IV; 1943. kind appear in exponential or hyperbolic expressions, it usually
7 The series is given in the “Encyklopadie der mathematischen suffices for purposes of ordinary accuracy to read “a<<fJ” as “a< ~b”
Wissenschaften, ” vol. II, p. 293. or “a>>b” as “a >2b. )’
? 955 Park: Planar Transmission Lines JY

In this line the potential difference is only half that in To find the power absorbed in the two strips we must,
the case of two strips, so that 20 is given by half its for the first time in this work, introduce the thickness
value in (21). In the extreme cases we have as before of the strips, for if we take it to vanish, the field inten-
sity, and hence the dissipation, will be infinite at the
edges. To avoid further complexity we shall assume that
the strip follows the equipotential v = K – e, where
e/K<<l. From (24a) we have, evaluated along the
strip, with dw =e and dz =jdy =jh,
The sign >> in the second formula is actually unneces-
sary, for the error is already very small when w = lz. g

TI+IZ .\ TTENUJITION

In order to calculate the attenuation from (8) we need

to know dw/dz. From (15) (dropping the now unneces-
To find the approximate half-thickness of this surface,
sary primes) we find, in terms of z and w,
we write A as a function of c and x, and then find its
dw a cosh z / b maximum value. This turns out to be
—— = .—

‘z‘5 d[(a’-sinh’:)(a’-sinhn:)l)l ‘24’) A,,,ZY = be

cosh d’/b

sinh
— cosh d/’b

d’/b –
1 ~(l+a’ sn2 w)
. —— ______ (24b) so that if we take t to be the (total) thickness 0[ the
ab cnwdnw
strip, the corresponding equipotential sunface has
The second of these is not hard to integrate. Let us begin
t sinh d’/b
by calculating Jl, the contribution to the dissipation in- (27)
tegral (~) due to the top and bottom plates. The contour ‘ = % cosh (i’/b – cosh d/b “
in (8) corresponding to half the top plate, is by Fig. 3,
that for which — w <z’ S – 1. On this plate u=O, so This artifice of introducing the strip by means of an

that w =.jv, sn w =j tn (v, k), and (13) becomes3 equipotential amounts to replacing square edges by
rounded ones. An actual strip of foil, however, has edges
.! =
e —atnv (k’), which are neither square nor round, and the point is of
little importance.
so that (8) is

——
~Tow we are to integrate (24 b). For the top side Off one
~.
~:J,=L
ab s ~,,–111~
~11–a’tn’v]

dn v
cn2 vdv (k’).
strip we
(O STJ ~ K-’).
E we have,
integrate
Making
to sufficient
along
use of the
the

accuracy,
equipotential
approxima.tions
w = K+
for
c +jv
small

This is evaluated by letting s = <(l +az) sn v to give

sn v .
cn IJ
cn(~”+c+jv)=–jk
( —
dn v –
—
‘e dn2 ,;
) “
(k’).

cn v snvl
a

-1 d}z(K+e+jv)=k — + jek” —
— ~anh—l ——
)
( dn v dnv, j
<(l+a’)

4 d’ With this, the contribution to (8) from one strip becomes

– b2#2
( d’ coth ~ – d coth : . (26a)

s ‘nvcnv[l+i;(R-k’’’:)21
) 2 1< ~(a’+dn’ v)dn vdv
J2=—— _ ——
When k’ is close to zero this form is awkward, but then abk2

‘k’)
o

we can do (25) directly to get

2d/b
Jp: l–———— (k’ <<1) (26b) for the dissipation in one side of one strip,. The substitu-
( .&& 2d/b ) tion s = dnZ v now reduces this to
and similarly the limiting form when d and d’ are large is

s
1 <(a’ + s)ds
J2 . ~ ___ –-(k)
(26c)
‘b ‘: (1 – S)(S – k2) +5 [(1+ k’).r2 – 2k’]’
s
~ These formulas have been given by R. M. Barrett in ‘{Micro-
wave Printed Circuits—Preliminary Memo and Technical Note, ” The result of doing this integration, dropping all tmrns
.\F Research Center, Cambridge, Mass.; 1951, together with the
results of extensive experiments verif}-ing them. which vanish with e, is, by (27),
 )2 IRE TRANSACTIONS—MICROWAVE THEORY AND TECHNIQUES April

::::6%)
LEAKAGE FROM THE LINE

“=i-HcOth:lnr:cOsh: One way to reduce the leakage of power from the edges

+ cot’ ~ In
b
?
( t
k ~o~h ~
b
sinh w/2b
cosh d“/2b )1
(28)
of

the
the
somewhat
chief
plates
the
is to

objection
electrical
close

to it
them

lies in
in.
characteristics
This,

the
of

increased
course,
of the
complexity
changes
line, but

where d” is d’ +d. This can be evaluated readily enough of the operation of making the ‘(sandwich.’7 Although to

in any particular instance. If we assume that d’/b and d/b calculate exactly the power which leaks from the line is

are large enough so that e–zdlb and e–ad[b can be neg- beyond the scope of this paper, we can at least make a

lected, it simplifies considerably, and we find that relative estimate of it by calculating the intensity of the
electric flux at points in the central plane of the system.
One would at once think that where there are two strips
J, = $,(1 + e-2’/b)
[’n(?sinha-:l ‘2’) with opposite charge the lines of force would predomi-

nantly run from one to the other, so that there would

and that for the two conductors
be less flux out to the sides than in the single-strip line.
We shall see that this is true, though the effect is not so
J= J,+2J2=;
pronounced as one might have thought.
[ “+e-’d’b’%sinh:)
If we square (24a) and let z be a real number x, much
w+-2d (30)
— e–2d/b . greater than b, we find that
2b 1
4a2
e–2z/b4
E2=— (34)
Here we shall give only two further limiting forms. The k2b2

first is that corresponding to (22a) where, for the whole

system, (8) is But throughout this analysis we have assumed that the
voltage applied across the line is given by K, which de-
pends on the line geometry. What is relevant here is,
J= J1+2J2=: ln~+~(w–2b) (d<< b<<d’)
[ 1 say, the field produced when one volt is applied. From
(34), this is
so that, with (2), (3), (4) and (22 b), the attenuation is
4a2
e–2xlb
Elz = —
K2k2b2 “

For the double-strip line this is, by (16)

4
El’ = — sinh’ ~ e-”fb (double strip). (35)
Ktb2
where we have used the approximation
For the single-strip line x should be measured from the
(k<< 1). center of the strip. Therefore we replace x in (34) by
x+d+&o, and further, the potential difference between

The other limiting case which we shall consider is that the strip and the envelope is only half that between the

in which d becomes infinite, corresponding to the single- two strips. Thus

strip line: 4
E12 = — e-zf’~f’) /’ (single strip). (36)
K2b2
Jz=; in ~ – e-”lb) (w>> b). (31)
( Supposing that, as would generally be the case, d’>>b

In these formulas, t must here be taken small enough so in (35), we find that

that the second term in the numerator is not larger than 1

(or nearly equal to) the first. To find the total dissipa- E12 = ——— e-2 f’-d’J 1’ (double strip). (37)
K2b2
tion in this case we use (26c) and (29) to form
These two special cases will in general have k2<<l [see

J= Jl+Jz=&
( ln~+~–e–”lb
) (W>> b). (32) (22a)
both
that
and
cases.
in each
(23a)
Further,
the
]. If this

factor
is true
comparing
in parentheses
then
(37)
K is close
with
is the
(36)
to 7r/2
we
distance
in
see

This, with (2), (3), (4), and (23b) gives for the attenua-
measured from the outside of a strip, so that if the one
tion in this case.
or two strips are to occupy a given width, the two-strip
arrangement will, for a given width of plate, have one-
ln~+~
~~K fourth the leakage of the one-strip system. These equa-
a— (W>> b). (33) tions also make it clear why the most compact and eco-
= 1207rk
ln2+~ nomical construction will have the dielectric sheets made
as thin as possible.