Transmission-Line Properties of A Strip On A Dielectric Sheet On A Plane

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

IEEE TRANSACTIONS ON MICROWAVE THEORY ANDTECHNIQUES,

VOL.MTT-2?5,
NO.8, AUGUST1977 631

Transmission-Line Properties of a Strip on a


Dielectric Sheet on a Plane
HAROLD A. WHEELER, FELLOW, IEEE

Abstract—The subject is a strip Iiue formed of a strip and a A peculiarity y of the half-shielded line is the mixture of two
parallel ground plane separated by a dielectric sheet (commordy different dielectrics. One is the material of the sheet between
termed “microStrip”). Building on the author’s earlier papers [1],
the strip and plane. The other is the air above the sheet. The
[2], all the significant properties are formulated in explicit form for
practical applications. This may mean synthesis and/m analysis.
simple rules of conformal mapping are restricted to a
Each formula is a close approximation for all shape ratios, obtained uniform dielectric or to some discrete boundaries that are
by a gradual transition between theoretical forms for the extremes of different from the subject configuration. Various other
narrow and wide strips. The effect of thickness is formulated to a approaches have been directed to this problem.
second-order approximation. Then the result is subjected to numeri-
The first close approximation for this strip line with mixed
cal differentiation for simple evaluation of the magneticloss power
factor from the skin depth. dielectric was published by the author in 1965 [2]. It is based
The transition formulas are tested against derived formulas for on some rigorous derivations for a thin strip by conformal
overlapping narrow and wide ranges of shape. Some of these formulas mapping. These are supplemented by some logical concepts
are restated from the earlier derivations and others are derived for interpolation between the extremes of dielectric. The
herein. The latter include the second-order approximation for a
uncertainties of interpolation are small enough to meet
narrow thin strip, and a close approximation for a narrow or wide
square cross section in comparison with a circular cross section. design requirements within practical tolerances. The result
Graphs are given for practical purposes, showing the wave resist- is a collection of formulas and charts which are complete for
ance and magnetic loss for a wide range of shape and dielectric. For the wave resistance of a thin strip.
numerical reading, the formulas are suited for programming on a The loss power factor (PF = l/Q) in a strip line has
digital pocket calculator.
components of electric loss in the dielectric and magnetic
I. INTRODUCTION loss in the conductor boundaries. These were not treated in

o
the early paper but have been addressed by some other
NE FORM of strip line is naturally suited for the
authors in the meantime.
simplest fabrication in a printed circuit. It is the
In the frequency range where a strip line may have a
familiar type made of a dielectric sheet with a shield-plane
length comparable with the wavelength, the magnetic loss is
conductor bonded on the bottom side and a pattern of strip
usually the dominant component. It is largely dependent on
lines on the top side.
the strip thickness, so the formulas for a thin strip do not
The purpose of this paper is to present some improved
suffice. This loss PF can be evaluated from knowledge of the
formulas and graphs, including not only the wave resistance
inductance of the line, which is independent of the dielectric.
but also the losses. The effect of strip thickness is simply
This evaluation can be made with the aid of the
formulated to, enable the evaluation of magnetic loss.
“incremental-inductance rule,” published by the author in
In the vernacular, this type of line is termed “microstrip,”
1942 [3]. Other authors have applied this rule to the
a term which is avoided in this scientific article because it is
formulas of the early papers [13], [17] with the first-order
commonly used without a clear definition and is not self-
thickness effect stated therein.
descriptive. Apparently it was intended to be a short desig-
In the sandwich line, it has been simpler to evaluate its
nation for “microwave strip line.” The “microwave”
properties, for various reasons. First, the homogeneity y of the
description is ambiguous and only partially relevant. Fur-
dielectric avoids the problem of mixed dielectric, which is
thermore it does not distinguish from the “sandwich” form
relevant for wave resistance. Second, the symmetry and
of a microwave strip line.
two-sided shielding cause much greater decay of a field with
Here also the descriptive term “wave resistance” is used in
distance. The symmetry simplifies the evaluation of the
preference to the nondescriptive term “characteristic
thickness effects, so those have been published, including the
impedance.”
magnetic-loss PF [8]. These give an indication of trends in
The subject strip line may be described as half-shielded, by
the subject line, but not quantitative values.
the ground plane on one side, as distinguished from the
As in most of the previous articles, only the lowest mode of
sandwich type, which is fully shielded, by ground planes on
wave propagation in the line shall be considered, and,
both sides. The half-shielding is adequate for some practical
furthermore, only at frequencies so low that there is negli-
purposes, because the external field is relatively weak and
gible interaction between the electric and magnetic fields.
does decay with distance.
This is valid if the transverse dimensions are much less than
half the wavelength in the dielectric. This mode may be
Manuscript received October 29, 1976; revised February ’23, 1977.
termed the “quasi-TEM mode, ignoring second-order
The author is with the Hazeltine Corporation, Greenlawn, NY 11740. effects of dispersion and surface-wave phenomena.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
632 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

After the following list of symbols, the configuration will


be defined and the scope of this article will be indicated. ‘~ IJ ~
——————t–8 t
STRIP
————— ~
II. SYMBOLS
fi\\\\\\\N\\\\Y \ 1111~1’+ I I t I (
The units are MKS rationalized (meters, ohms, etc.).

k= dielectric constant of the sheet of material


separating the strip and the ground plane.
k{ = 1 + q(k – 1) = effective dielectric constant
of all space around the strip.
q= (k’ – 1)/(k – 1)= effective filling fraction of (a)
the dielectric material.
RC = 377 = 120rc = wave resistance of a square
area of free space or air.
R= wave resistance of the transmission line
formed by the strip and the ground plane (of
perfect conductor) separated by a sheet of
dielectric k.
RI = R without dielectric (k= 1).
R6 = RI subject to skin depth d in a real conductor. /
(b) (c)
R/Rl = I/@ = L~/& = speed ratio in mixed dielec-
Fig. 1. A strip line parallel to a plane. (a) Rectangular cross section. (b)
tric k relative to free space or air.
Cross-section square or inscribed circle. (c) Cross-section small square
w= width of the strip conductor. or equivalent circle.
h= height (separation) of the strip from the
ground plane.
h
. thickness of the dielectric sheet. R of the asymmetric model (single strip and ground plane)
t . thickness of the strip conductor. and the descriptive dimensions w,h,t. Here the thickness is
w’ . effective width of a strip with some thickness. featured, and the equivalence between a practical strip and a
, . wider theoretical thin strip (a perfect conductor with a
width of an equivalent thin strip (t+ O).
A: . w’ – w = width adjustment for thickness. thickness approaching zero). This equivalence is described
Aw’ . width adjustment with mixed dielectric k. in terms of the width adjustment Aw.
d . skin depth in the conductor. For evaluation of the magnetic-loss PF, the skin effect is
P . l/Q = magnetic PF of the strip line. indicated in dashed lines. These boundaries are recessed by
Pk
. electric PF of the dielectric material k. one half the skin depth (6/2) so they indicate the actual
= effective PF of mixed dielectric k’. center of current. The actual boundary is the theoretical
P’
P —
— current center in a perfect conductor. The change between
P + 8/h = ph/6 = normalized p.
— rate of attenuation (nepers/meter). one and the other is involved in the computation of the
A: — wavelength in free space or air. magnetic PF. It is assumed that all conductive boundaries
/lg — guide wavelength in mixed dielectric k’. are nonmagnetic and have equal conductivity and skin
e . 2.718 = base of natural logarithms. depth.
exp x . c? = natural exponential function. As an extreme case of strip thickness, a square cross
in x . log. x = natural logarithm. section is introduced, as shown in Fig. 1(b) and (c). These are
acosh x — anticosh x = cosh” 1 x. related to a circular cross section in either of two ways, the
asinh x . antisinh x = sinh -1 x. inscribed circle (b) or the equivalent circle (c). Each is found
asin x . antisin x = sin – 1 x. to be helpful in some studies, mainly because the circle yields
to simple exact formulation for comparison with an approx-
The following table translates some symbols from the
imation for the square. Both will be used for reference.
author’s earlier papers.

IV. SCOPE
[1] [2]
The thrust of this article is to enable explicit synthesis of a
w, h? t, Aw 2a, b, Ab, 2Aa
line to meet some specifications. This is achieved for various
R of 1 strip R of 2 strips (twice as great)
(A- )(B- ) ( )( ) formulas sequences. The wave resistance R is related to the dielectric k
and the shape. On the other hand, the magnetic PF can be
decreased by increasing the size, while the shape has a lesser
III. A STRIP LIF.TEON A DELECTRIC effect. The PF is usually a tolerance rather than a requisite.
SHEET ON A PLANE The wave-speed ratio is taken not to be specified, but rather
Fig. l(a) shows the cross section of the subject line. It evaluated after synthesis of a design of a cross section.
corresponds to the 1965 article [2] except for the translation Some graphs are introduced here, for reference in various
to “practical” parameters. The latter are the wave resistance sections. They present the relations needed for the purposes

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIESOF STRIP ON DIELECTRIC SHEET 633

R (OHMS)
M“o

200

100

50

40

30

20

10
-0.1 0.2 0.3 0.4 0.5 1 2 345 10
w’/ b

Fig. 2. The wave resistance of a thin strip on a dielectric sheeton a plane.

of practical design, and can be read close enough for thickness ratio of the conductive strip and the dielectric
ordinary purposes. The formulas to be given are intended as sheet.
an alternative to the graphs, and also to give further insight New formulas are presented here in the main text without
into the relations. The formulas are designed for program- derivation. Most of them are empirical formulas providing a
ming in a small digital calculator such as the HP-25 or gradual transition between narrow and wide extremes.
HP-65. These are tested against the derived close approximations
Fig. 2 is a graph of the wave resistance of a thin strip, the for overlapping narrow and wide ranges, which are reviewed
same as previously published [2]. It is made with closest in Appendix VI. Some derivations, not previously available,
approximation by complete computation for overlapping are given in Appendixes IV and V. Special emphasis is
ranges of narrow and wide strips. An alternative to this placed on some formulas which are “reversible” in the sense
graph is the simple empirical formulas to be given for the that a formula can be expressed explicitly in a simple form
entire range of width. The wave-speed ratio (relative to air or for either analysis or synthesis.
free space) for any width ratio is equal to the ratio of wave
V. A THIN STRIP WITHOUT DIELECTRIC
resistance with and without dielectric (R/R1 ).
The effective filling fraction q of the dielectric is also The 1964 paper [1] gave the derivation for a wide thin strip
graphed on Fig. 2 for a mean value of the dielectric constant without dielectric, and, incidentally, also gave formulas for a
(k= 3). It enables an alternative computation of the effective narrow thin strip. These together covered any width. Explic-
dielectric constant k’ and the resulting speed ratio (1/~~). it formulas were given for both purposes, analysis and
Fig. 3 is a graph of the thickness effect on the wave synthesis.
resistance without dielectric. The relative effect is less with Recent studies yielded the discovery that the “narrow”
dielectric, so the indicated effect is an upper bound. It is a formula could be put into a form which would also be
small effect with respect to wave resistance but has a greater asymptotic to the “wide” formula. This is accomplished
effect on the magnetic PF. This is generally similar to the while retaining its principal features for “narrow” approxi-
first-order effect of thickness as previously statecl [2] but is mation. Furthermore, this has been so arranged that the
refined and extended to include the second-order effect in formula is “reversible.” By this is meant that an explicit
some degree. formula for either analysis or synthesis can be converted to
Fig. 4 is a graph of the normalized magnetic PF (P = p + an explicit formula for the other. This conversion is per-
c5/h) as evaluated from the thickness effect. The magnetic PF mitted no complication beyond the solution of a quadratic
is independent of the dielectric and its normalized value is equation. The resulting formulas are empirical in the sense
independent of the size. The thickness parameter t/h is that they must be tested against derived formulas in the
chosen as being a property of the laminate, specifically the “wide” range and in the overlap of “wide” and “narrow.” For

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
634 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

R1(OHMS)

bw-l&
u+
k=l R h M

--0.1 0.2 0.3 0.4 0.5 1 2 345 10


w/h

Fig. 3. The wave resistance of a strip without dielectric, showing the effect of thickness.

---
0.1 0.2 0.3 0.4 0.5 1 2 345 10
wlh

Fig. 4. The magnetic power factor of a strip, showing the effect of thickness.

the thin strip without dielectric, such derived formulas are The following are the asymptotic forms for the narrow
the subject of the 1964 paper. The resulting formulas herein and wide extremes: narrow:
are based on “narrow” derivations which are relatively
w’/h = 8(exp – R1/60)[1 + 1.73(exp – R1/60)2]
simple although the derivation of the second term has not
been published before. It is based on a two-wire second- R, =60 In (81z/w’ + 1.73(w’/8h)) (3)
order approximation for a narrow thin strip. The overlap of wide:
the “narrow” and “wide” derivations is indicated, by stating
the small error of either at the transition, w’/h = 1207rjR ~; R ~ = 120z(h/w’). (4)
The new formulas are to be generalized for dielectric, but
In the “narrow” formula, the second-order term has the
they are first given here in simplest form to show some
proper form but its coefficient is compromised (1.73 instead
features. Reversible formulas are here shown first for syn-
of 2) to accomplish asymptotic “wide” behavior.
thesis and then for analysis:
The asymptotic behavior at both extremes could be
accomplished by any of several variants, yielding somewhat
~1/h = 8 ~(exP R1130 - 1) + rc2/4 (1)
(exp R1/30 – 1) different behavior in the transition region. The form chosen
was found to give close enough approximation with the
RI = 30 in {1 + -$(8h/w’)[(8h/w’) + ~(8h/w’)2 + rc2]} minimum number of terms.
The form of these transitional approximate formulas
(2)
shows some points of similarity to the exact formulas for a
where the error is < O.OIR ~ round wire near a plane, which are to be given here.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 635

VI. SQUARE OR CIRCULAR CROSS SECTION These formulas are intended mainly for the magnetic-loss
PF, for which there is no effect of dielectric, and only the
As an extreme departure from the thin strip, a square or
“analysis” form (R of w/h) is used. The synthesis form (w/h of
circular cross section is considered, still without dielectric.
R) is shown mainly for academic interest, since it formed the
Fig. l(b) shows a square or an inscribed circle as the cross
basis for the empirical formulas for the square wire over the
section, with the description in the same terms as Fig. l(a)
entire range of width ratio.
(t/w = 1). It is noted that the distance from the plane is
In Fig. 1(c), in addition to the equivalent circular and
described by the separation height h, not by the distance to
square cross sections, there are shown some equivalent thin
center (which is h + t/2). Hence it is compatible with
strips. A round wire far from the piane has an equivalent
separation by a dielectric sheet.
concentric thin strip whose width is double the wire
For the narrow case, simple formulations for a square
diameter (2.36w). If not so far from the plane, there is a thin
wire and the equivalent round wire are known. See Appen-
strip of lesser width (w’) which is equivalent by the following
dix III. Fig. l(c) shows this relation and the radius (2h + w)
two tests:
of the outer circle equivalent to the plane.
For the wide case, the exact formula is known for the a) height above the plane equal to that of the lower side of
round wire but not for the square one. Therefore a close the square;
approximation for the square wire has been derived and is b) equal wave resistance.
presented in Appendix IV.
The indicated geometric proportionality of the two strip
For the round wire, the exact formula for any width ratio
widths is of interest in kind but not in degree, because their
is known in simple reversible form. By modifying this form, a
difference becomes substantial for a square so wide that the
reversible empirical formula has been derived for the square
simple rules of equivalence are failing.
shape. These formulas are presented here. (1?~ without
The lesser thin strip, compared with the square, deter-
dielectric is here simplified to R, because here there is no
mines the width adjustment here associated with the
need for this distinction. )
thickness of the square.
For round wire without dielectric the exact formulas are
as follows: VII. A THIN STRIP WITH DIELECTRIC
2 1
w/h = The 1965 paper [2] gave the derivation for a thin strip with
cosh R/60 – 1 = (sinh R/120)2
dielectric. Two sets of formulas covered wide and narrow
2 strips, with close agreement in the transition region. The
2
=
(exp R/120 – exp – R/120 )

4
reversible formulas given above are here adapted to dielec-
tric. Asymptotic behavior is achieved for the following
— conditions:
– exp R/60 + exp – R/60 – 2 a) narrow strip, low-k and high-k extremes, with a logical
4 exp – R/60 4 exp R/60 interpolation therebetween;
(5) b) wide strip, all k.
= ~p – R/60)2 = (exp R/60 -17
The resulting empirical formulas are found to track the
R = 60 acosh (2h/w + 1)= 120 asinh ~h/w
derived for~ulas- over the entire range of width and
=60 in [(2h/w + 1)i-~(2h/w + 1)2 -- 1] dielectric:

— 120 In (~h/w + ~h/w + 1), (6)
w,h=8/[expt&@)-llW+*
For square wire without dielectric the approximate for-
mulas are as follows:

w/h = —
1/0.59 2 + exp – R/60
[exp(~=b] ‘9)
exp R/60 – 0.21 – exp – R/60


1 2 + exp – Rj60
(7)
+1’(3[(14YkK)
‘=&
0.1185 exp R/60+ exp – R/60 -%
‘m+211 ’10)
‘=601nKA+1+0’ The error is < 0.02R (or < O.OIR over most of range).
The analytic form gives R (fork) and R ~ (for k = 1), from

‘J(==T7
which the speed ratio is
(8)
R/Rl = l/@< 1. (11)

The relative error is <0.025 or < (0.005R + 0.5 Q). If R ~ O, Therefore no other formula is needed for the speed ratio. The
w/h ~ 38 l/R (near 377/R). Each of these formulas is asymp- simpler formula (2) for R ~ may be used, but that is no
totic in the first- and second-order terms for “narrow” and advantage if the more general formula is recorded in a
the first-order term for “wide.” program for numerical computation.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
636 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

While the effective filling fraction q [2] of the dielectric is The resulting formula is expressed in terms of the actual
not required in the procedures given here for design compu- width w or the equivalent-thin-strip width w’. As mentioned
tations, it is a matter of some interest. Particularly, it is a above, these relations are based on free space, without
factor in the electric-loss PF to be formulated. Schneider dielectric:
[14] has given an ingenious simple empirical formula, based
Aw=~ln
on [2], which is close enough for practical purposes:

(
‘=~l+&” ) (12)

or
‘ z * ’13)

Compared with the derived formulas for narrow and wide


ranges (for a mean value, k = 3) the departure is <0.02. It is “n* ’14)
a simple transition between the bounds (~,1 ). It lacks the
This adjustment enables a width conversion either way
shape that is peculiar to either extreme, which is contained in
between equivalent strips with or without thickness.
the derived formulas.
The development of this formula for the wide and inter-
mediate regions has been enabled by complete computation
VIII. STRIP THICKNESS AND THE Loss POWER FACTOR
of a few examples (Appendix II). These were accomplished
The earlier papers did not make any attempt to evaluate by the technique of conformal mapping. Specifically, a few
conductor loss, because it is not determined in the limit of a shapes (w,h,t) of rather small thickness were evaluated by
thin strip. However, there was given a width adjustment for numerical integration of the space gradient. This process is
the edge effect of a small thickness. From this adjustment, laborious and required some ingenuity near some bounds of
some other authors have formulated the losses to be ex- integration.
pected, and their reduction by thickness [13], [17]. Three examples so evaluated were sufficient to indicate
This subject has been reviewed. The width adjustment has two features implicit in this formula.
been verified for small thicknesses of a narrow strip, and has
a) For a wide strip, the previous formula (1965) is refined
been formulated more closely for a wide strip. A single
in respect to its second-order effect. The ratio
formula is given here for the entire range of width. Also it is
previously included as 2h/t is here changed to 4h/t. The
adapted to moderately large thicknesses (up to a square
former ratio was based on unlimited width, and the
cross section for a narrow strip).
change is an adaption to the limited width.
The loss PF (PF = I/Q) of the magnetic field (bounded by
b) The “narrow” and “wide” formulas appear to be upper
the conductors) is evaluated by the rule proposed by the
bounds, as would be expected. Furthermore, the qua-
author in 1942 [3]. This “incremental-inductance rule” is
dratic sum of the two inverse ratios fits the sample
based on differentiation of the inductance relative to the skin
points.
depth 6 in the conductor boundaries, as indicated in Fig. 1.
In a transmission line with perfect boundaries and no The adaptation of this formula for a greater thickness has
dielectric material, the inductance is proportional to the been enabled by derivations for a square cross section. The
wave resistance. Only the relative change is significant, so extra numbers (+1. 10 or – 0.26) are chosen to match the
the rule is here applied to the wave resistance R ~.This avoids square condition (t = w) for a narrow strip. The formula is a
the nuisance of magnetic units and surface resistivity. close approximation for moderate thicknesses (t< h) of a
A great simplification is now available by numerical wide strip. (Another formula has been derived for a wide
differentiation. This was not available in the slide-rule strip of square cross section, Appendix IV.)
computations of earlier days so analytical differentiation For loss computation, the actual width and thickness (w,t)
was necessary, however cumbersome. It was used by the are converted to the width of an equivalent thin strip
other authors. It is no longer needed. What is needed is an (w’ = w + Aw). Then the thin-strip formula (R, of w’) can be
analytic formula giving the wave resistance in terms of all used for differentiation with respect to the actual dimensions
dimensions but without dielectric. (w,h,t).
The edge effect related to the strip thickness is here As indicated in Fig. 1, each dimension is incremented by
described in terms of the extra width Aw of a thin strip ~ d and the same formula is used again to obtain R& Then
having equal wave resistance RI without dielectric. This is the (small) loss PF is computed by the incremental-
indicated in Fig. 1. inductance rule:
The first-order effect of a small thickness is given in the
R6 – RI
1965 paper, for the extreme cases of narrow and wide strips. p= = 1 –R1/R~= In Rb/Rl < L (15)
D

Three advances are here presented:


A normalized form for loss PF is proposed, which gives
a) a refinement for the wide strip (Appendixes I and II),
the effect of shape, independent of the size, frequency, and
b) a unified formula for the entire range of width,
conductor material. It is normalized to the height h:
c) a second approximation for greater thickness, within
some restrictions. P = p + (d/h)= p(h/c5) p = P(ti/h). (16)

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 637

& Aw’ . ~
The reference (d/h) is the nominal PF of a very wide strip. h x- 2

In computing the normalized PF P, the value of the skin _ A:’ _


RI
depth is immaterial if it is sufficiently small to approach the _ Aw _
h
limiting behavior of the skin effect (which is usually of
3 4 2
interest). Also it must not approach the sensitivity of the FIG.2
* ~ \ P z
computer. In a computer giving ten decimal places, a fair R,’ FIG.3
R, -— —__ --
compromise is 6/h = 0.0001. Then the skin effect is well
d \ L L ‘
represented if all dimension ratios exceed 0.001. R o
b
For evaluation of a resonator made of a strip line, the loss
1
PF (or dissipation factor or l/Q) is usually the most
t/h
J
significant factor. The wave R is incidentally relevant in the 1
\ P
circuit application of the resonator. The loss PF of the R

magnetic field is evaluated by the simplest formulas (R ~and - d \

Aw without dielectric). For any shape, the value of P enables o


a computation of the size of the cross section to realize a k
w“/h wlh w’/h
value of p: (THIN) t/h
(REF) (REAL) +1

Fig. 5. Relations which determine the speed ratio.


h = Pti/p = P8Q. (17)

The graphs in Fig. 4 show the loss PF in terms of P for a


wide range of shapes. The common reference is the height h section is one which is bounded by these two dimensions and
and the thickness ratio t/h because they may be fixed by a has a certain thickness (t/h < 1). The extreme optimum is a
dielectric sheet and a conductive sheet bonded thereto. peculiar rounded shape bounded by these dimensions.
For small thicknesses, the loss PF exceeds the reference
IX. STRIP THICKNESS WITH DIELECTRIC
value, as would be expected. Also the amount of excess is
greater for lesser thickness, as a result of the current The effect of strip thickness is formulated above, but
concentration at the edges. For example, reducing the without the effect of dielectric. A width adjustment for
thickness from square to t/h= 0.02 may double the PF (in thickness may be made in the synthesis for a specified wave
the moderately narrow range). resistance with dielectric.
An unexpected result is the loss PF being less than the The width of an equivalent thin strip is defined as one
reference value for a wide strip of substantial thickness. This which is wider by the amount which yields an equal value of
happens because part of the magnetic energy is beyond the wave resistance. This involves both inductance and capaci-
region bounded by the height. This part has boundaries tance. The width adjustment for the former is independent of
further apart, and hence a lesser value of loss PF. dielectric. That for the latter is less for a greater dielectric
In Fig. 4, the two lowest curves give the loss PF for square constant, because the thickness of the edge is somewhat
and circular cross sections of the same width. It is less for the spaced from the dielectric.
latter, the lower bound for the wide extreme being one half To approximate this effect, the width adjustment is
the reference value (P+ ~). In the narrow region, it is less divided in two equal parts, and one part is decreased by the
because of the following. factor l/k. The modified value becomes:

a) The two shapes are known to have equal skin resist- Aw, = 1 + I/k Aw
w = w’ — Aw’. (21)
ance [9]. 2’
b) The circular shape has greater reactance. The propor-
The entire width adjustment Aw is effective for wave
tionate wave resistance is greater by 60 in 1.18 = 10 ~;
resistance without dielectric (k = 1) or for inductance alone.
this is denoted, “the rule of 10 !2”
For capacitance alone, the entire width adjustment would be
If the thickness is comparable with the height, the relevant decreased by the factor l/k.
restriction may be the overall height (h + t), perhaps for Fig. 5 shows the behavior of the width adjustment without
reasons of clearance space. Also the width maybe restricted. or with dielectric. Especially it shows its graphic determina-
Then the thickness ratio has an optimum value. This is tion from Figs. 2 and 3. The full value of Aw is effective
found by minimizing a related normalized PF defined as without dielectric, decreasing the wave resistance R ~equally
follows : by decreasing inductance and by increasing capacitance. Its
amount is represented by the horizontal separation of the
Pkt=Ph; — t = P(l + t/h)= (p/c$)(h + t) (18) upper pair of curves, both shown in Fig. 3. The effect of
dielectric with a thin strip is represented by the separation of
Square: w/h near 0.55, min P~, == 1.65 (19) the upper curves (R ~,R) in the upper and lower pairs, both
shown in Fig. 2. The reduced amount of width adjustment
Circular: w/h near 0.50, min P~, == 1.56. (20)
with dielectric is constructed and projected downward to
Within specified bounds of the overall height and width (not give the lesser horizontal separation of the lower pair of
less than the overall height), the optimum rectangular cross curves.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
638 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

The speed ratio for a thin strip is R/R ~ from Fig. 2. The bound. Therefore the electric PF is only slightly less than
interpolation for mixed dielectric, taking account of that of the material, so the complete formulation is not
thickness, gives a greater speed ratio from this construction. critical and may be unnecessary. If desired, it can be
The latter locates a point on the lower curve of the ‘lower computed (as above) from R/R ~ and pk.
pair, not shown elsewhere. Either attenuation or PF may be deduced from the other.
As will be seen in a procedure and example to be given, the However, it is preferable to evaluate the magnetic PF
indicated numerical sequence yields all the quantities from directly, because it is independent of the dielectric and the
the graphs in Figs. 2 and 3. Or they maybe computed in this speed ratio. In a wide range of situations, it represents nearly
sequence by these formulas: all of the loss PF.

Sequence Formulas
XI. PROCEDURES FOR COMPUTATION
1 (9)
2 (2) The formulas are intended for useful applications, which
3 (14) may be theoretical and/or practical. As brought out in the
4 (21) earlier papers, “synthesis” and “analysis” are the alternative
5 (l~~f) objectives, the former for practical design and the latter for
6 evaluation of a configuration (the classical textbook
approach). Both are needed here for a practical design to
In this sequence, 2-3 is the width adjustment downward meet some specifications and tests. Therefore a few
from the upperbound for a thin strip. A parallel dashed line procedures and examples will be outlined to show the use of
shows also the width adjustment upward from a strip with these formulas in arriving at a practical design.
thickness. The former will be used in a synthesis procedure, The first few procedures start with the synthesis of a line to
the latter in analysis. The amount of the adjustment is meet a specification of wave resistance. The subsequent
designated alike in both (Aw/h) although it may differ evaluation of speed ratio and skin effect are inherently
slightly (too little for any practical significance). analysis, but the procedures build on the synthesis.
First Procedure: On a specified printed-circuit board, find
X. ATTENUATION
the width for a 50-f2 line:
The rate of attenuation with distance in a transmission
a) specify properties of a dielectric sheet with metal faces:
line is simply expressed in terms of the average PF (magnetic
k=2.5, h=lmm, t=O.l mm;
p and electric p’) and the wavelength 1~ in the line:
b) specify wave resistance: R = 50 Q;
c) width of thin strip by (9) or Fig. 2: w’/h = 2.85;
d) width adjustment (without dielectric) by (14) or Fig. 3:
(22) Aw/h = 0.15;
e) effect of dielectric by (21): Aw’/h = O.10;
In words, this is the average PF (nepers) per radian length.
f) width by (21): wjh = w’/h – Aw’jh = 2.75; w = 2.75
The magnetic PF is evaluated by the skin effect, as described
mm.
above.
Second Procedure: For the same line, evaluate the speed
The electric PF p’ of the effective dielectric in the line (k’)
ratio, referring to Fig. 5:
can be expressed in terms of the various parameters
involved: g) no. 1 in sequence, c) above: w’/h = 2.85;
speed ratio: h) no. 2, find R; of thin strip by (2) or Fig. 2 or 3:
R; = 71.5;
(23)
i) no. 3, d) above: Aw/h = 0.15, w“/h = w’/h –
filling fraction: Awjh = 2.70;
kr–l
j) no. 4, e) above: Aw’/h = 0.10, w/h= w’/h –
~ [(R,/R)’ – 1] (24) AwfJh = 2.75 ;
a=k–l=k–l k) no. 5, by Fig. 3: RI = 71, or can be computed by the
electric PF: “fourth procedure”;

~, = ~k
1) speed ratio= R/Rl = 50/71 = 0.70.
(25)
l/q – 1 Third Procedure: For the same line, evaluate the magnetic
l+T
PF and the attenuation from this cause, referring to Fig. 4,
bounds: Appendixes VII and VIII:

The electric
dielectric-material
PF
‘k’p’> u qpk

Pk

1+1/k
p’ is seen to be within
PF pk, and usually
‘~pk”
(~+
(26)

~) of the
it is nearer the upper
m) find the normalized

n)
be computed
numerical
PF by Fig. 4: P = 1.10; or it may
by (16) using a nominal
differentiation;
specify the frequency
20 = 0.3 m;
small d and

(or wavelength 2.): ~= 1 GHz,

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 639

o) specify the conductivity (or material) of the metal (which may have a square or circular cross section). ,
boundaries: copper; The latter &ffers a les~er PF.
p) evaluate the skin depth by (62) or [7]: d =: 2.1 flm; b) If a thick strip is to be afforded, specify the bounds of
q) compute the PF by (16): p = 0.0023 ❑ = 2.3 roil, the space (overall height and width, h + t and w).
Q = 440; c) Specify whether the conductor (strip or whatever) is to
r) compute the attenuation rate from PF, speed ratio, be supported in contact with a dielectric sheet. If so,
etc., by (22): a = 0.034 Np/m or 0.30 dB/rn. specify the height of the latter (h).
d) Subject to these restrictions, choose a cross section
If used for a long line, the attenuation rate may be
giving near-minimum P~, (18).
significant. If used for a resonator, the PF and speed ratio are
e) If using a strip of small thickness t,a lesser PF is
relevant.
obtainable by greater width w and greater height h.
In the third procedure, if one is concerned with only one
f) If using a square cross section in contact with a
example (size, shape, materials, frequency) the actual skin
dielectric sheet (t/h = w/h), the least PF is obtainable
depth may be used directly, then the procedure assumes this
by a moderately wide shape (say w/h near 3).
order: (n,o,p) (m,q,r).
If using a round wire in contact, a lesser PF is
A lesser PF may be required, or it may be desired to
obtainable by greater width (diameter), but little re-
explore the compromise between the loss PF and the height
duction is obtainable beyond a moderate width (say
and/or thickness. The first-order relation gives the PF
w/h near 3).
inversely proportional to size (h,t,w). A closer evaluation
h) If using a square or round wire with no need for
may require complete computation of various examples,
contact, the least PF is obtainable by a width near one
then interpolation.
third the overall height.
The following example and procedure are mlodified to
i) If a rectangular space is specified, with the width not
develop from analysis only. In particular, the width adjust-
less than the overall height, the least PF obtainable
ment corresponds to the dashed line in Fig. 5.
with a rectangular cross section requires some
Another Example: Design a resonator to be made of a
thickness less than one third the overall height.
square wire bonded to a printed-circuit board Similarly
lettered items refer to the foregoing procedures: There is usually not available an explicit formula for the
synthesis to realize a specified value of the loss PF. The
a) k=2.5, h=lmm, t=w;
graphs in Fig. 4 can be applied to this problem. Knowing
m) from Fig. 4: near-minimum P = 0.8 for w/h = 2,
the skin depth 6 and specifying the material (h,t), a value of
w=t=2 mm;
the loss PF p requires the P computed from (16). In Fig. 4,
p) 6 =2.1 pm;
this value of P determines the shape ratio w/h and hence the
q) p = 0.0017= 1.7 roil, Q = 590.
width w. If this P is lower than a practical curve, the size (h,t)
The speed ratio can be evaluated by the following procedure. may be increased to permit a greater value of F’.
It is found to be 50/67= 0.75.
Fourth Procedure: For any configuration, fincl the speed XII. CONCLUSION
ratio. For a thin strip, see (11) and Fig. 2 for the simple rule.
The transmission-line properties of a strip parallel to a
The following gives the effect of thickness:
plane, with or without an intervening dielectric sheet, are
a) specify configuration: k = 2.5, w = 2.75 mm, h = 1 evaluated in simple formulas, each one adapted for all shape
mm, t= 0.1 mm, w/h = 2.75, t/h = 0.1; ratios. The formulas relating the width/height ratio with
b) width adjustment (without dielectric) by (14) or Fig. 3: wave resistance are stated explicitly for both analysis and
Awjh = 0.15, wflh = 2.90; synthesis, with or without dielectric. The wave-speed ratio
c) wave resistance (without dielectric) by (2) or Fig. 2 or and the magnetic-loss PF are stated from the viewpoint of
3: R1 =71; analysis, which is usually what is needed.
d) effect of dielectric by (21): Aw’/h = 0.7, Aw/h = 0.10, The advance over previous publications appears mainly
w’/h = 2.85; in two areas:
e) wave resistance (with dielectric) by (10) or Fig. 2:
a) a relation is expressed explicitly by a single simple
R = 50;
formula for the entire range of the shape ratio;
f) speed ratio: R/R ~ = 0.70.
b) the width adjustment for thickness is formulated and
The speed ratio is slightly greater than that for a thin strip of used for evaluation of the magnetic loss.
the same width.
Each formula is an empirical relation obtained by designing
If resonance (small PF or high Q) is the principal objective
a gradual transition between known simple formulas for
(rather than wave resistance) a different procedure maybe
both extremes of narrow and wide shapes.
indicated. The following outline gives some relevant
All formulas are designed for ease of programming on a
considerations.
pocket calculator such as the HP-25 or HP-65. Particularly,
a) Choose between a specified printed-circuit material the digital calculator enables the numerical differentiation
(h,t) and the alternative of an attached thick strip (for loss evaluation) which is here used to realize a great

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
640 ISEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

simplification. While beyond the scope of this article, the I -In %3 (w/t+f) I
5 In 47e (w/t) UB
writer would welcome inquiries relating to programs for the ~~ . In 4h/t (NEW)
t 4 — — In 2eh/t (OLD)
HP-25, some of which may be available on request.
The subject line, formed by a strip parallel to a plane, has (LINEAR 3 COMPUTED
SCALE) EXAMPLES
2
presented problems of evaluation which are much more
difficult than those of the strip between two planes (sand- 1
In 2.1
wich line). That configuration is symmetrical and the dielec- .
n M
It 11
—— w/h (LOG SCALE)
T
tric is homogeneous, so even the thickness effects have 4re h
❑ NARROW —WIDE
yielded to straightforward formulation [8]. The asymmetri-
G?
cal strip line is here formulated in a manner that is competi-
tive, although necessarily involving mixed dielectric. I I
Wlt h/t
While there is always room for further progress, the (NM~;)W (wIDE
THIN)
graphs and formulas presented here are complete in that
(LINEAR
they offer the option of graphical or numerical reading for SCALE ) 1
1
the all numerical values that may be needed for design
purposes. Preliminary estimating is usually aided most by
the graphs. _l_l w/h (LOG SCALE)
r
XIII. ACKNOWLEDGMENT (b)
Fig. 6. The behavior of the width adjustment for thickness. (a) Transi-
This study has been ‘stimulated by the attempts of other tion between narrow and wide. (b) Family of transitions.
authors, building on the writer’s early papers. The stimula-
tion has come partly from a perception of some deficiencies
extreme, the edge-field pattern being influenced mainly by
in progress, but more from an appreciation of the construc-
the proximity of the two edges.
tive efforts and interesting results of a few of the intervening
A smooth transition at the knee is provided by a quadra-
workers. The final impetus was provided by the advent of the
ture combination in formula (13). This is validated by some
HP-25, which offered the computational power best suited
computations to be described in Appendix II, indicated as
for the “close support” essential to such a development.
three points on the curve. This validation requires a change
APPENDIX I in the wide formula, from the “old” in [2] to the “new” in (13).
BEHAVIOR OF THE WIDTH ADJUSTMENT The computed points indicated that the level UB should be
FOR THICKNESS raised by a factor of two under the logarithm, as seen. (This
factor is not exactly determined, but two appears to be the
Formulas (13) and (14) for the width adjustment are based
nearest and simplest number that might be indicated, and it
on some asymptotic relations and a transition therebetween.
may have an exact basis.) This is regarded as a refinement of
Asymptotic formulas for narrow and wide extremes were
the previous rule, whose derivation ignored the second-
given in the early papers [1], [2]. Here a revision of the wide
order interaction between the edges far apart. It is noted that
formula and an integrated formula with a simple form of
the transition occurs in the vicinity of a width ratio some-
transition were presented. This appendix is a graphical
what less than unity (w/h = l/z).
description of the behavior of this adjustment, for the
The asymptotic relations are based on the limiting condi-
purpose of visualizing the transition and some associated
tion of a thin strip. Formula (13) includes an adaptation
relations.
(w/t + 1.1) which extends the close approximation to the
Fig. 6(a) shows a graph foT a constant ratio of
square condition. This introduces another curved transition
thickness/height (t/h). This may be the practical situation
at the foot of the graph, raising the curve from the “square”
when designing for a printed circuit to be made by etching a
point (t/w = 1). While beyond the present scope, it is noted
conductive sheet bonded to a dielectric sheet. The width
that the curve has a minimum near the foot and approaches
adjustment ratio zAw/t is plotted on the width ratio w/h. The
a higher level (z) in the narrow extreme (w/t < 1).
scales are, respectively, linear and logarithmic, to give
Fig. 6(b) is a diagram showing a family of such graphs. For
straight lines for the sloping graphs.
greater thickness, the knee’ is closer to the foot of the graph,
The normalized form for the width adjustment takes out
so the two curved transitions would merge as the thickness
the principal dependence on thickness, so one can see the
approaches the square shape. Then their separate descrip-
variations of the coefficient which is dependent on shape. A
tions become indefinite, so the validity of formula (13) is
higher value indicates a greater coefficient (responsive to
further tested on square shapes, as evaluated in Appendixes
thinness) but the amount of the adjustment is still nearly
III and IV.
proportional to thickness.
APPENDIX 11
There are two upper bounds (UB’S) for this coefficient,
SMALL-THICKNESS EXAMPLES BY
based on the narrow and wide asymptotic behavior. The
CONFORMAL MAPPING
level upper line is based on the wide extreme, the edge-field
pattern being influenced mainly by the proximity of the Formula (13) gives the width adjustment for thickness. It
shield plane. The sloping lower line is based on the narrow is an empirical transition between the narrow and wide

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 641

Y
TABLE I
COMPUTED EXAMPLES

~ ~ —2 J_

188.5 133.7 88.5


‘1
1.414 1.1 1.01
‘4
h II 0.7368 0.7774 0.8412
X+,y=z U12
u. 0.8867 0.8720 0.8988
{/ ,@ L
0.5267 0.6520 0.7648
~lz —x ‘1
o 0.348 0.879 2.020
w’/h
(a)
w/h 0.237 0 .754 1.894

“J \
tlh

Aw/h
0.0755

0.111
0.0816

0.125
0.0838

0.126

Aw/t 1.48 1.53 1.51

h \ (14) 1.47 1.55 1.56

dif. +.01 -.02 -.05

—— “

o U12 U2 u~=l ‘J4


%
(b) rules have been devised for closer convergence near the
Fig. 7. Conformal mapping of the cross section of the strip line. (a) singular points which correspond to the angles of the
Contour in space. (b) Space gradient. (Each area equals dimension in contour. The result is a close approxirpation in cases where
space.) the singular points are not too closely spaced.
The wave resistance is determined by the gaps in the
extremes of the asymptotic behavior in the limit of a thin straight line, both sides of center. For the upper half-plane,
strip. The first-order relations for these extremes have been ~~ I-kz
1 + 1.14k2
known [2] but not the behavior
Also there is found a second-order
for the wide extreme.
in the transition
effect requiring
regions.
a revision
R = *RC K(k)
K’(k)
=
( n
1 + 1.14(1 – k’)ln
1P p
)
16
,, (28)

The validation of this formula, especially in the transition ( rc l–kz


‘)
region, is provided by some complete computations ofa few
in which k = I/uA.
examples by a procedure based on conformal mapping. The
The latter (empirical) formula has a relative error <0.005.
contour of the cross section is mapped on a straight line. The
It has the correct center value, skew symmetry, and asymp-
space gradient on this line is integrated to evaluate
totic behavior at both extremes. A closer simple iorm~la for
the dimensional ratios on the contour. Rather than implicit
a wide strip is
elliptic integrals, numerical integration is used. Even that is
confronted by difficulties of integration where there is an 607r2 607c2
R= (29)
infinite value and/or slope at either bound (co,@, or ~@. 8 ‘~”
In — In —
Fig. 7 shows the essentials of the conformal mapping of in I/k in u+
the cross section of the strip line. The actual contour, Fig.
7(a), is described on the space plane (x + jy = z) in terms of If l/k = U4 <1.4, the relative error is < 0.00IR.
the shape dimensions w,h,t whose ratios determine the Three examples have been computed. They are sum-
properties. This contour is mapped on a straight line, Fig. marized in Table I, numbered in order of increasing width.
7(b). On the scale of this line u is graphed the space gradient The first is a critical shape (Rl = 377/2) while the others are
(or inverse field gradient) on the contour. The area under the chosen to give a range of widths in the transition region.
space gradient in each interval is equal to the dimension on These three examples have comparable values of the
the contour. thickness ratio (t/h near 0.08). This ratio is small enough to

The space gradient is formulated by inspection, as be representative of small thicknesses (t/h <1 and w/h < 1).
follows : Its value is the basis for the graph in Fig. 6(a), and the three
[1 – (u/u, )’][1 – (L@,)”] ‘1’ (27) points are plotted in relation to the curve of formula (13).
z= lc?z/dul = The close agreement is regarded as confirmation of that
[1 - (u/u3)’][1 - (u/u4)q
formula (for small thicknesses), especially in the transition
Only the area ratios are significant, so the scale is arbitrarily region which does not have a clear theoretical basis. This
chosen for simplicity. result was the objective of the complete computation of
The analytic integration would involve elliptic integrals. these few examples.
There is a constraint that precludes an explicit solution. The
APPENDIX III
upper and lower faces of the strip must have equal width
EQUIVALENT SQUARE AND CIRCULAR
(WO= w), to be realized by proportioning one of the critical
CROSS SECTIONS
values on the straight line.
Numerical integration is simple in concept and has been The extreme thickness of a narrow strip line is taken to be
found useful in computing a few examples. Some special a square cross section (t/w = 1). Therefore the formula for

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
642 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

width adjustment contains a constant which assures a close t--- \


w+h
-. \
approximation up to this thickness. Its derivation is based ‘\ \ @
\\
on the relations among three equivalent concentric cross \\ \

sections, the square, the circle, and the thin strip, shown to \ \
\ \ ,.,8 ~,2 EOUlVALENT
scale in Fig. 1(c). Their dimensional ratios are such as to give
{ CIRCLE
equal values of capacitance and inductance (assuming a :+ h ,-— L
/;/ ---
\
small skin depth) and the resulting wave resistance (all in all \
\
free space). *) (’ ~
\
\
Starting with the square (of width = w), the equivalent \
\/
/’
@ ‘\
circle has a diameter which is greater in the ratio: L h + /=’ \ : \

@ ‘>.@, \
— r(5/4)
4912n —
r(7/4)
= 1.1803 = 1/0.8472. (30) ~“ “ ~
2 I
w+h

;+h
The equivalent strip has a width which is double this (a)
diameter.
In Fig. l(c), the large dashed arc is the circular boundary
equivalent to the ground plane (radius = 2h + w).
Based on these equivalents, a narrow strip of square cross
section has the following wave resistance:

R= 60 ln~~ww
— = 60 in 1.70(2h/w + 1)= 60 in 8h/w’

(31)

in which L ~(i+n 2) L ~in ~

(b)
2.36w Fig. 8. Derivation for a wide square cross section. (a) The four regions in
“=l+w/2h” one quadrant. (b) Analysis of the corner region.

In terms of width and thickness, the corresponding adjust-


ment is The third region, 3, is taken to be filled with logarithmic
field, which is described by radial lines and concentric

W’-W=W(RL-1)=W’(+.3
‘W= (32)
circles:

C,= ~ln (1 + w/2). (35)

In formula (13) or (14) forAw, the constant+ 1.1 or –0.26 The second region, 2, is the excess in the transition
is inserted to give the correct value for a narrow strip of between 1 and 3, taking into account the nearby distortion in
square cross section. both of those regions. The upper and lower boundaries
From another viewpoint, Fig. l(b) shows cross sections of (corner and straight line) are mapped on parallel straight
a square and an inscribed circle (having equal width). In the lines. Then the relative displacement of far points evaluates a
narrow case, these have wave resistances differing by “stretch which represents the excess in the transition. This
is divided in two parts for the two directions from the corner.
60 in 1.18 = 9.93 (near 10 Cl). (33) Fig. 8(b) shows this result diagrammatically. Region 2 is
This is denoted, the “10-f2 rule” for these two cross sections. represented by 1A and 3A, the respective extensions of the
They are known to have equal skin resistances [9] so the loss adjacent regions. The validity of this viewpoint resides in the
PF of the circular wire is less in the inverse ratio of its greater fact that the distortion from the transition decays rapidly in
reactance and wave resistance. either direction, and also tends to average out. The resulting
value of the transition region is
APPENDIX IV
WIDE SQUARE CROSS SECTION C2=~(l–ln2)+~ln~
The wide square cross section is here evaluated in simple
terms by invoking a variety of techniques in four regions of 2
=— 1 – In ~ = 0.483. (36)
each of the two active quadrants. These regions are 7C() n
described in Fig. 8(a). Each is to be evaluated first in terms of
The fourth region, 4, is closely related to an inscribed
normalized capacitance C, which is the simplest concept for
the boundaries involved. The dimensions are referred to the circle, as shown. The region around the inscribed circle
height (h= 1). would contribute
1
The first region, 1, is taken to be filled with uniform field:

c1., = W12. (34)


.,

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 643

For ease of computation, this ratio can be approximated


z PLANE
by

(c.) =
() ~
1 + 2/w
1’3<1.

Between the inscribed circle and the equivalent circle (of


(38)
I
o
I
1
I
2 3
I I
4

1.18 times the radius) the nominal capacitance is that of one (a)
ciuadrant: W= z’ PLANE

[c.] = * = 9.49= 1/0. 105. (39)


in 1.18
~“’”oo’
This is,used to increase (C3) to approximate the capacitance
o 1+%2
of the square in this quadrant: HALF–CURRENT CENTER J
1 (b)
cd= <1.12. (40)
l/(c4) – l/[c4]
*“’*
The resulting wave resistance of the two quadrants
(restoring w to w/h) is ~+,~:
3’77 “T
—. —+)+”+-2–
R=
2C1 + 2c~ + 2C3 + 2C4
1 /1! I
. 377
:2 “
w/h + 0.966 + ~ in (1 + w/2h) +
(1+ 2h/w)l/3 -0.1
Fig. 9. The pair of small wires equivalent to a thin strip. (a) Thin strip
(41) and equivalent round wire. (b) Thin strip and equivalent pair of round
wires. (c) Thin strip and its image in a ground plane.
This formula is best for a wide square cross section. The
best for narrow is formula(31) for the equivalent round wire
mean distance between one pair of current centers and the
shown in Fig. l(c). Their effective overlap is indicated by
other pair is increased to
their close values for a transition shape (w/h= 1):
2h’ = ~2h~(2h)2 + w2/2 = 2h[l + ~(w/h)2]1i4
1) wide (41) above: 94.91;
2) narrow (31): 95.11 (close lower bound); = 2h[l + &(w/h)2 – . “ “]. (42)
3) all (13) (2): 95.32.
The narrow-strip formula becomes
The intermediate value is believed to be the closest approxi-
mation for this case. R = 60 in 8h/w
=30 in [(8h/w)~(8h/w)2 + 8] (43)
APPENDIX V
NARROW THIN STRIP This is a reversible formula, giving the following for
synthesis:
For a narrow thin strip (without dielectric) there is here
derived the second-order approximation stated without ~exp R/15 + 16 + 4
proof in the early papers [1], [2]. It is based on a pair of small w/h = 8 (44)
exp R/30 “
wires equivalent to the strip. It forms the basis for the simple
formulas (l), (2) for any width. For a narrow strip, the 4 term is of the second order
Fig. 9(a) shows a single round wire of unit radius and its (exp R/30) and the 16 is of the fourth order (exp R/15)
known equivalent thin strip whose width is 4 units. It is relative to the first order (exp R/60).
described on the z plane. It is to be transformed to another A modification of the above formula gives a simpler form
plane, z’= ~~. which has a linear slope for R a O, while retaining the
This transformation is here performed about one end of second-order approximation:
the strip cross section, and the result is seen in Fig. 9(b). An exp R/30 -t 2
equal strip survives but the wire becomes a pair of smaller w/h = 8 (45)
exp R/30 – 1 “
wires. This pair provides a second-order approximation to
the far field of the strip. (This simple equivalence has not In the limit, wfh + (@)240/R = 416/R.
been seen by the author in any of the many published This is not far from the desired 377/R. The latter result can
exercises in conformal mapping.) It is noted that the smaller be obtained by substituting a slightly lesser value (n2/4 –
wires are not strictly circular in cross section, but that is 1 = 1.467) for the constant 2. Asymptotic behavior for a
irrelevant in the use of the concept herein. wide strip is then realized at the cost of a slight deficiency in
Fig. 9(c) shows the thin strip (or equivalent pair of wires) the second-order term. The result is the simple formula (1)
and its image in a ground plane. From this g~ometry, the giving a close approximation for any width.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
644 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

The two preceding formulas are extremely close (the The two preceding approximations give a large overlap.
relative difference is <0.0002) in the narrow range They are closest near R ~ = 126 or w/h = 1, where the
(w/h < 2). The first term of error is proportional to (w/h)’. relative difference is 0.0005R ~. For the graphs in Fig. 2, the
Their average has a relative error <0.003 if w/h <4. The computation of any one point is made with the formula
second formula is closer for greater widths, and therefore judged to be the closer of the two; if so, its relative error is
probably for all widths. The corresponding formula for less than this amount.
analysis is (2) except change nz to 12. A comparison of these two formulas can be made in
‘1’hn exercise is a striking example of the technique of explicit form by the following sequence:
higher order approximation and its application to obtain a
1) wide (49): wlh from R ~~;
simple and versatile empirical formula with support from
various theoretical relations. 2) narrow (2)
(modified): RI. from w/h;
APPENDIX VI
PREVIOUSLY DERIVED FORMULAS 3) ratio: R1n/RIW = 1 + relative
difference. (50)
Any empirical formula must be validated by comparison
with derived formulas. These are typically more complicated The relative difference of RI is the significant comparison.
and/or restricted as to the range of the width ratio, A thin strip with dielectric likewise has different formula-
Here are some such formulas for a thin strip, selected from tions for narrow and wide. The effective dielectric constant k
the earlier papers, (A- #) referring to the first [1] and (B-#) depends on the shape w/h and on the dielectric k. One
referring to the second [2], They are stated in a form that is sequence can be used for explicit formulations in any case:
convenient for computation, to provide a comparison test
1) specify: Rl;
for the more recent empirical formulas covering the entire
range of the width ratio. They are converted to the dimen- 2) compute (45): w/h ;
sions used herein (w, h, etc.).
3) specify: k;
Every one of the derived formulas is essentially the first
4) compute
few terms of a series converging in the extreme of a narrow or
(53), (57), (52): q, k, R = R1/@;
wide shape (and small thickness). Therefore a “narrow” or
“wide” identity is necessary. The transition between the two 5) graph: R for k, wfh. (51)
occurs for a shape which may be “borderline” for close
The “effective filling fraction” [2], defined as follows,
approximation by either. Hopefully the two kinds will
depends mainly on the shape and less on the-dielectric:
overlap to give a covq-age for all shapes. As stated in the
early papers and as supported by more recent studies, the //-1

transition occurs near w/h = 1. The more sophisticated k=l+q(k– 1), (52)
‘=k–l’
formulas give substantial overlap.
Because it has only second-order dependence on the dielec-
For a narrow thin strip without dielectric, the second-
tric, a simple formula for a mean value of k is sufficient for
order approximation was not supported by a derivation.
practical purposes. A mean value (k = 3) is chosen because it
One is given here in Appendix VI. Formula (45) and the
places the effective dielectric k’ midway between the ex-
corresponding modification of (2) are presented as the
tremes (for 1< k < co) and within the midrange of practical
closest approximation known to date. It provides overlap of
values. Some formulas will be stated for this mean value,
the wide range. The relative error is < 0.003R if w/h <2.
with a supplemental term which may be ignored, having a
For a wide thin strip without dielectric, the first paper
factor (l/k – ~). It is graphed in Fig. 2 in terms of w/h directly
yields a remarkably close approximation with overlap of the
narrow range. The synthesis form is an explicit formulation. and R ~ indirectly.
The shape dependence of the filling fraction was derived in
Specify
terms of the wave resistance without dielectric (R-l) and is
RI < 60n = 188 most simply expressed in those terms. This R ~and the actual
shape w/h are related by various formulas. The filling
(A-l), (A-45) d = ~RC/Rl = 592/Rl > rc (46) fraction is here expressed in very simple form from the
previous derivations for narrow and wide.
(A-67) d = d’ + (2d’)2 exp – (2d’) > z (47) For a narrow thin strip with dielectric, the effective
dielectric constant is formulated as follows. The shape is
(A-1O) C=~(d– 1)2 – 1 =~d(2d – 1) (48)
introduced in terms of the wave resistance without dielectric
(Rl):
(A-68) w/h = ~ [c – acosh (d – 1)]

‘B-32)( B-44) ‘=~+:(lnf+M


=~[c–ln (c+d-1)]>0.3. (49)

The relative error is < 0.00IR,.

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIELECTRIC SHEET 645

~,=k+l+~k–l TABLE II
(B-32), (B-45) —
2 RI –(’nf”+n$
2“

(54)
Identification
COMPARISON OF FORMULAS FOR ONE

Derivation
EXAMPLE

Rl (OhltlS) _Error

(s) [14] unrestricted 126.553 0


The relative error is <0.01 of E if R’ is > 70; w/his <3; q is (W.1) (43) narrow 126.533 -.020
<0.72. (w.2) (2) (modified) narrow 126.528 -.025
For a wide thin strip with dielectric, the effective dielectric (w.3) [11 (A-68) wide 126.473 -.042

constant is formulated in terms of parameters defined above (w.4) [11 (A-66) narrow 126.641 +.088

and here: (W.5) (2) narrow-wide 126.310 -.243

(K) [16] wide 127.857 +1.304

(w.6) [1] (A-71) wide 124.424 -2.129

(B-8) d = :Rc/R, = 592/R, > IT, R, <188


(s) Schneider’s example is derived rigorously
$:A:) (A-16), from elliptic integrals and is taken to be
“exact” for purposes of comparison. Its rela-
S’ = 0.732[aco:h (d – 1) – acosh (0.358d + 0.598)] tion to the other items tends to confirm its
validity.
(55)
(W.1) This is the derivation based on the pair of
(B-25) s“ = in 4-1- l/(2d - 1) wires equivalent to a narrow thin strip. It is
the closest approximation (the relative error
= 0.386 – l/(2d – 1) (56)
is <0.0002).
(W.2) This is similar to (W. 1) but modified to a form
(B-4) q = 1- ~ acosh (d – 1) - s“ + ~
[ 1
. (57)

(W.3) (W.4)
suitable for matching the wide extreme. It is
the reverse of formula
These are the closest approximations
(45).
given in
The relative error is <0.01 (estimated).
the 1964 paper. They bracket the correct
For the mean case (k = 3), this result is approximated
value within a relative difference of +0.0007.
very closely by the simple formula
Their computation is much easier than (S).
~=l_!!ilnl!!. (58)
(W.5) This is the only item providing a rather close
592 RI approximation over the entire range of shape.
(K) Kaden’s “wide” formula is an approximation
The overlap between the two simple formulas (53) and (58) to his derivation from elliptic integrals. The
occurs near RI = 100 or w/h = 1.5. error (about 1 percent) indicates that this
The narrow and wide simple formulas for the mean case shape is “borderline” for his approximation.
can be integrated and supplemented by an adjustment for It is comparable with (W.6) in its explicit form
any k, as follows: and in simplicity, and gives a closer
approximation.

This concludes a summary of the earlier formulas,’and


some more recent, as required for the above procedure. They
are adequate for a set of close computations for a thin strip
with any dielectric. These may be used for checking any
empirical formula such as those proposed herein. They are
used for the graphs in Fig. 2.
The relative error is <0.01 of k. The first term is altransition
APPENDIX VII
between the narrow and wide extremes. The second term is a
COMPUTATION OF Loss BY
very close adjustment for the intermediate range. The last
NUMERICAL DIFFERENTIATION
term is negligible in the practical effect on k, so it serves
mainly to indicate the weakness of the dependence on k. In practical applications of a strip line, the PF of conduc-
One simple example is here reviewed in Table II as a test tor loss is usually determined by the skin effect. Some simple
of various derivations for a thin strip without dielectric. It is rules are applicable if the skin depth d is much less than the
a shape (w/h = 1) which is in the region of transition least transverse dimension. One is the “incremental-
between narrow and wide approximations. The wave resist- inductance rule” stated by the author [3]. It relates the skin
ance R ~is based on free space (1207c). The items are listed in loss with the inductance, by a formula based on
order of increasing error from the first. The derivation is differentiation.
described with respect to its development from the extreme In a transmission line made of perfect conductors, the
of narrow and/or wide strip. The following notes give further wave resistance without dielectric (R ~) is uniquely related to
comments. the inductance, so that formula may be used instead. Then

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
646 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, AUGUST 1977

DIRECT ( p FROM 8). NORMALIZED (P),


the loss PF of the skin effect may be expressed as follows:

Rb – RI

&
8 =SMALL VALUE
p= ALfL= R = 1 –R1/R8~ 1. (60) (SAY .0001)
Ls
The incremental-inductance rule is here represented by the

47 D
ENTER
relative increment of inductance (AL/L) that would be ENTER
w IN RO w/h IN RO
caused by removing a thickness (6/2) from the face of every h IN RI h=l IN RI
t IN R2 t/h IN R2
conductor bounding the field. The wave resistance (R ~)for a
perfect conductor would be increased in the same ratio if the
RUN
boundary were modified in the same manner. Then this

&
change (from RI to R6) is used to compute the PF.
The elegant application of the incremental-inductance
rule was stated in terms of the analytic differentiation of an
inductance formula in terms of a simple continuous func-
tion. In its more general application, such a formula may not

&
YES NO
be available. The rule is still useful if the inductance varia- R7=0

tion is formulated continuously over a range of dimensions. (1) (2)


RECYCLE FINISH
Such a formula can be subjected to analytic differentiation,
CALCULATE
but the resulting expression may be much more complicated RI IN R?
p=l-R4/R8
than the inductance formula from which it is derived. This

Ea%l I
Q
CALCULATE
has been the experience of some workers who have taken P=p+8/h
this approach in evaluating the loss PF in a strip line [13],
[17].
In the meantime, the advent of the digital calculator has I
—-r---’ I
”-’
opened up a new opportunity for the differentiation of a
Fig. 10. F1OW chart for computing the loss power factor by numerieal
complicated formula. It enables a close approximation of
differentiation.
the derivative by computing a small finite difference. The
basic formula is used twice, which requires little more effort
in a programmable calculator. The versatility of this
the dimension be changed by 6, either increased or de-
procedure reaches a peak in the Hewlett-Packard pocket
creased. The increment is entered by arithmetic in the
models, HP-25 and HP-65. The convenience and availabil-
register.
ity of the HP-25 provided the author with the tools and
Conditional branching is required at the end of each
incentive to prepare this paper.
execution of the R ~ subroutine. One register (R7) is vacated
Having stated the objective of numerical differentiation
until the end of the first execution, then occupied by R ~,
by finite differences, the programming is routine. Fig. 10
which signals the eqd of the second. This serves also to retain
shows the flow chart of one such program. It serves to bring
R ~ for comparison with R&
out the application of some features available in the HP-25.
If the program storage is inadequate for the principal
It includes as the principal subroutine, some formula for R ~
subroutine and also the transitional subroutines, one or
in terms of the transverse dimensions. After incrementing
more of the latter is easily performed manually.
the dimensions in accord with the skin effect, this subroutine
In Fig. 10, the right-hand column of notes describe the
is traversed a second time for R& The relative increase is
program changes for the normalized form.
interpreted as the PF in direct or normalized form. The
following features of the program are notable.
The skin depth 6 may be evaluated and then utilized to
APPENDIX VIII
give the PF p for any example. The more versatile nor-
FORMULAS FOR THE SKIN DEPTH
malized PF P may be obtained by arbitrarily choosing a
small difference (say d = 0.0001). Then the result Here are some formulas for the skin depth in nonmagnetic
approaches the analytic derivative. The value of the differ- conductors [3], [7]:
ence cancels out in the normalized form. The number of
decimal places in the small difference maybe somewhat less
than one half the number available in the computation. The
skin depth or small difference is entered once in one register
(R4) where it need not be renewed unless a change is desired.
The dimensions w,h,t are entered in assigned registers
R0,R1,R2. For the second computation, each dimension is
incremented by ~ c5.Each dimension is between two con- = ().()291~A/ci = & -. (61)
dfictor faces so the removal of 6/2 on each face requires that

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.
WHEELER : TRANSMISSION-LINE PROPERTIES OF STRIP ON DIEIJ3CTRIC SHEET 647

In copper (a= 58 Mmho/m): The small-thickness examples reported in Appendix II are


in the range of marginal approximation by Chang, so they
4.36 kHZ
6,= 3.81~~ ~m = mm have not been compared,
r f

‘r’”
REFERENCES

[1] H. A. Wheeler, “Transmission-line properties of parallel wide strips


by a conformal-mapping approximation: IEEE Trans. Microwave
Theory Tech., vol. MTT-12, pp. 280-289, May 1964.
=~mm==mm. (62) [2] — “Transmission-line properties of parallel strips separated by a
diele&ric sheetj” IEEE Tran$. Microwave Theory Tech., vol.
MTT-13, pp. 172-185, Mar. 1965.
Symbols used in (61) and (62) are defined below. [3] — “Formulas for the skin effect: Proc. IRE, vol. 30, pp. 412-424,
Sept.’ 1942. (Skin loss by the “incremental-inductance rule.”)
(s). = 2nf = radian frequency (radians/second). [4] F. Oberhettinger and W. Magnus, Applications of Elliptic Functions
~ = wavelength in free space (meters). in Physics and Technology. New York: Springer, 1949. (Wave resist-
RO = l/GO = wave resistance of a plane wave in a ance of coplanar strip, p. 63. Tables of K’/K, p. 114.)
[5] E. Weber, Electromagnetic Fields—Theory and Applications—
square area in free space (ohms). Mapping of Fields. New York: Wiley, 1950. (Thickness, p. 347.)
Vo = magnetivit y (magnetic permeability y) in free [6] H. A. Wheeler, “Transmission-line impedance curves; Proc. IRE,
space (henries/meter). vol. 38, pp. 140&1403, Dec. 1950:
[7] ———— “Universal skin-effect chart for conducting materialsfl Elec-
G = I/p= conductivity in copper (mhos/meter). tronics, vol. 25, no. 11, pp. 152-154, Nov. 1952.
R. = 377 = 1207CQ. [8] S. Cohn, “Problems in strip transmission lines; IEEE Trans.
Microwaue Theory Tech., vol. MTT-3, pp. 119-126, Mar. 1955.
Po = 0.47t = 1.257 pH/m.
(Sandwich line, thickness and loss.)
[9] H. A. Wheeler, “Skin resistance of a transmission-line conductor of
APPENDIX IX polygon cross section: Proc. IRE, vol. 43, pp. 805-808, July 1955.
RECENT ARTICLE ON WIDE STRIP [10] D. S, Lerner (Wheeler Labs., Inc.), unpublished notes, Nov. 1963.
(Very wide strip, width adjustment for thickness. Half-shielded type
WITH THICKNESS
compared with sandwich type.)
Subsequent to the preparation of this paper, the author [11] M. Caulton, J. J. Hughes, and H. Sobol, “Measurements on the
properties of microstrip transmission lines for microwave integrated
has seen a recent article related to the subject [19]. W. H. circuits,” RCA Rev., vol. 27, pp. 377-391, Sept. 1966.
Chang has described an ingenuous and powerful approxi- [12] H. Sobol, “Extending IC technology to microwave equipment: Elec-
mation based on conformal mapping. To that extent, it has tronics, vol. 40, no. 6, pp. 112–124, Mar. 1967. (A simple empirical
formula, remarkably close for a practical range.)
something in common with the author’s 1964 paper [1]. [13] R. A. Pucel, D. J. Masse, and C. P. Hartwig, “Losses in microstrip~
Some thickness is accommodated at the expense of some IEEE Trans. Microwave Theory Tech., vol. MTT-16, pp. 342-350,
refinements in other respects. The result is a very useful June 1968. (By analytic differentiation.)
[14] M. V. Schneider, “Microstrip lines for microwave integrated cir-
approximation for wide strips with thickness. To yield this in cuits,” Bell Syst. Tech. J., vol. 48, pp. 1421–1444, May 1969.
analytic form is a major achievement. Moreover, it appears [15] —, “Dielectric loss in integrated microwave circuits; Bell Syst.
that his result may be closely bracketed by further apprecia- Tech. J., vol. 48, pp. 2325-2332, Sept. 1969,
[16] H. Kaden, “Advances in microstrip theory; Siemens Forsch. u.
tion of his approximation.
Entwicld. Ber., vol. 3, pp. 115-124, 1974. (Computation of wave
Relevant to the present paper, Chang gives a table of resistance and loss for a wide strip of small thickness.)
examples computed from his formula and also by a numeri- [17] R. Mlttra and T. Itoh, “Analysis of microstrip tr%smission lines; in
Advances in Microwaves, vol. 8. New York: Academic Press, 1974, pp.
cal approximation from W. J. Weeks [18]. The agreement is 67-141. (Latest review, many references.)
very close. Most of those examples fall within the range of [18] W. J. Weeks, “Calculation of coefficients of capacitance for multi-
validity of the present paper, formulas (1), (2), (13), (14) conductor transmission lines,” IEEE Trans. Microwave Theory
Tech., vol. MTT-18, pp. 35-43, Jan. 1970. (Numerical method for a
without dielectric. The agreement is well within 0.0 lR. The rectangular cross section with a parallel plane.)
formulas herein offer a close approximation for any width. [19] W. H. Chang, “Analytical [C metal-line capacitance formulas; IEEE
They are based on a thin strip with width adjustment for Trans. Microwave Theory Tech., vol. MTT-24, pp. 608-611, Sept.
1976. (ConformaI mapping approximation for a wide strip with
thickness. Chang’s formulas for a wide strip are remarkable thickness. Examples compared with numerical method of Weeks [18]
for including the width and the thickness in one formula. using his program. )

Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 06,2022 at 11:32:42 UTC from IEEE Xplore. Restrictions apply.

You might also like