ACCURATE MODELS FOR MICROSTRIP

COMPUTER-AIDED DESIGN

E.Hammerstad and @.Jensen

ELAB
7034 Trondheim-NTH, Norway

Summary

Very accurate and simple equations are pres The accuracy of this model is better than 0.2%
ented for both single and coupled microstrip at least for cr~ 128 and 0,!31< u<IOO.
lines electrical parameters, i.e. impedances,
effective dielectric constants, and attenua-
Com~~redto earlier equations, t;ese give much
better accuracy, in the order of 0.1% for koth
tion including the effect of anisotropy inkhe
substrate. For the single microstrip the
impedance and wavelength. They ate also
effects of dispersion and non-zero strip complete with respect to range of strip width,
thickness are also included. while earlier ones where given in two sets.
Strip thickness correction
Introduction

In correcting the above results for non-zero
In microstrip design it is often observed strip thickness, a method described by Wheeler3
that the physical circuit performance differs is used. gowever, some modifications in his
significantly from that theoretically calcu- equations have been made, which give better
lated. This is due to factors such as atten accuracy for narrow strips and for substrates
uationr dispersion, and discontinuity effect% with low dielectric constant.
which again are functions of physical para- For homoge~ous media the correction is
meters and the actual circuit lay-out. The
4exp(l)
conclusion which may be drawn from this obs- Aul=~ ln(li ) (6)
ervation, is that optimum microstrip design t coth2~6.517 U
should be based directly upon physical dimen-
where t is the normalized strip thickness. For
sions. This approach requires accurate
mixed media the correction is
models and for computer-aided design the
models must also be in an easily calculable
Aur=#l+l\cosh~) Aul (7)
form.
By defining corrected strip widths, U1=U+AUL
The models presented in this paper represent
and ur=u+Au=,the effect of strip thic~ness may
partly completely new models, but also revi-
be included in the above equations:
sions and expansions of some which have been
given earlier 1,2. It has been a goal to
z@(u,t,Er)=zol(ur)/4Ee (Urrsr) (8)
obtain an accuracy which gives errors at
least less than those caused by physical )=Ee(urrcr)[zol(ul)/zol (ur)12 (9)
Eeff(urtrEr
tolerances.
theoretical he
data 1:;; odes
wh~le are dlscont~nuity
the based n
models also are based on experimental data. Dispersion

As microstrip propagation is not pukely TEM,

Sgqle microstrip
both impedance and effective dielectric con-
stant vary with frequenoy. Getsinger4 has
The model of the single microstrip line is
propsed the following model for dispersion in
based upon an equation for the impedance of
effective dielectric constant
microstrip in an homogeneous medium, ZOI , and (o)
E
an equation for the microstrip effective r - eff (10)
dielectric constant, Ee. eff = r - 2
l+G(f\fP)

where f =Zo/(2uoh) may be regarded as an app-

(1) roximation to the first TE-mode cut-off freq-
uency, while G is a factor which is empirically
determined. Getsinger gave anexpression for G
f(u)=6+(2n-6)exp
[
_(30.666
----R)
0.7528
1 (2) which
substrates,
fitted
but
experimental
this did
data
not
for
qive
alumina
correct
results for other substrates. The following
Here v is wave impedance of the medium expression for G has been shown to give very
(376.7?fl in
vacuum) and u is the strip wicltih good results for all types of substrates now
normalized with respect to substrate height in use:
(w/h) . The accuracy of this model is better
then 0.01% foru<l and 0.03% for u < 1000. (1.1)

Er+l &r-l There is today not general agreement on a mdel

~l+lO)-a(u) b(&r)
ee(u,cr)=~ + ~ (3)
for microstrip impedance dispersion. Based on
u
a parallel-plate model and using the theory of
1 u4+(u/52)2 dielectrics, the following equation was found.
a(u)=l% ln U4+0-432 +*ln 1+(*1)3;(4)
[

K-g 0.053
(5)
Zo(f)=zo(o) & eff
e
(o)
. eff
eff
(f)
(o) -1
-1
(12)
b(&r)=0.564(~)
r
407
While we would accept theoretical objections ~. 20+g2
LJ(urg) = g exp(-g) (26)
to this model on the qrounds that microstrip
1 o+g2
impedance is rather arbitrarily defined, it
may be pointed out that it agrees very well
exp[p(g)!n u + q(g) .
with the calculations of Krage & Haddad ? and
that in the limit f+- it is agreement with
three ;f the definitignsdiscussed by Bianco (2-z)
& al. . It may also be noted that the model
predicts a rather small but positive increase
in impedance with frequency and that this p(g)=exp(-O.745 g0.295 \cosh(g0.68) (28)
would seem to fit experimental observations. (29)
q(9) = eXp(-1.366-g)
0.15
Couwled microstri~s fol(g,FR)=l-exp{-0.179 9---

The equations given below represent the first r(g,sr)

generally valid model of coupled microstrips _ 0.328 g
}
with an acceptable accuracy. The model is 9mExp(l)+(g\7)28] (30)
based upon perturka~ns of the homogeneous
microstrip impedance equation and the equat- exp(l-(Et-l)2jB.2)
ions for effective dielectric constant. The r(9,sr)=l+0.15[l- 1 (31)
equations have been validated against theore- l+g-6
tically calculated data in the range
When the development of the above equations
O.1<U<1O and g >0.01 (g is the normalized
was done (1) and (2) were not available. In-
gapw~dth (s/h)~, a range which should cover
stead a simpler equation for homogeneous
that used in practise.
microstrip impedance with a maximum error of
The homogeneous mode impedances are 0.3% for u >0.06 was used:

(u,9)=zol(u)/[1-z (), (u) om(ur9)\no)l (13) 0.172,

Olin OI(U)= no\(u+l.98 U
(32)

where the index m is set to e for the even

The errors in the even and odd mode impedances
mode and o for the odd. The effective dielee where then found to be less than 0.8% and
tric constants are less than 0.3% for the wavelengths. The above
Er+l Erl
model does not include the effect of non-zero
Eem(urg,sr)=~ +7 Fm(u,g,cr) (14) strip thickness or assymetry. It is a goal at
ELR.B to include these factors, although the
parameter range probably will have to-5e re-
10 l-a(~) b(Er)
Fe(u,g,&r)=[l+ (15) stricted to the strictly practically useful in
LI(U,9) regardof the complexity of the above equations.
10 -a(u)b(sr)
Fo(u,g,er) = fo(U,g,Er) (1+-) (16) Dispersion is also not included presently.
Getsinger7 has proyosed modifications to his
The modifying equations are as follows single strip model
dispersion , but unfortu-
nately it is easily shown that the results
o~(u;9=Q(ulA4
(g).la(9)um(g) +[1-a(9)lu-m(g)}} are
of
asymptotically
gap width. However,
wrong
it
for
is
extreme
possible
values
to
(17)
modify the Getsinger dispersion mdel=thatthe
0.172
Q(u)=O.8645U (18) asYmp-tati&: values are satisfied for coupled
2.09 lines, but the details remain to be worked out
lJJ(g)=l+ ~ + 93 95 (19)
1.45 . Attenuation

a(g)=O.5exp(-g) (20) The strip capacitive quality factor is

m(g)=O. 2175+[4.113+ (%)7251 + (l-q)+q&r

(33)
10 d = (l-q)/QA+q&r/Qs
1 9
+ 323 ln 10 (21)
1+(* ) where q is the mixed dielectric filling frac-
. . tion, QA is the dielectric quality factor of
-n(9)in ~1 (22) the upper half space medium, and Qs is the
00(w9)=0e(wg)-* exp[f3(9)u
substrate quality factor.
The strip inductive quality factor is approxi-
6(9)=1.729+1.175 JLn(l+ (23) mated by 8:
~+o.:;;2:2 -17)
10 Qc = n Zolh f . u
(34)
6(g)=0.2306+ ~ h g lo + R5 c E
1+ (~) where R is the skin resistance and K is the
1 1.175, currentsdistribution factor. Surface ~ugh-
JLn(l+O.646 g (24)
+ 5.3 ness will increase the skin resistance .

Rs.(A)=RsJO) [l+ # Arc. tan 1.4(+)21 (35)

n(g)={~+exp[ -6.424-0.76 ing-(~3)5]}.

. ~n ~0+68.3 g2
(25)
1+32.5g3.093

408
where 6 is the skin depth and A is the rms tropic substrate, may be directly calculated
surface roughness. with the above equations, while the substrate
quality factor has to be modified:
For the single microstrip current distribution
factor we have found
Qs = 2(+++)-1 (45)
zolyo.7 Sx SY
K = exp(-l.2( ). (36)
no

to be a very good approximation to the currerk Discontinuities

distribution factor due to PwzeJ et al. 9
provided that the strip thickness exceeds Due to the illnesr modelling of microstrip
three skin depths. discontinuities could unfortunately not be
finished to meet the digest deadline. The
The mictrostrip quasi-TEM mode quality factor, final paper presented at the symposium will
Qo, is now given by give models for the microstrip open end,tk gap,
compemated right-an@e bend, and the T-junction.
1
-. ~+~ (37)
Q. Qd Qc Acknowledgements

This work has been supported by the European

and the attenuation factor becomes
Space Agency under contract no. 3745/78.

References
(38)
1. E.O.Hammerstad & F.Bekkadal: Microstrip
Handbook ELAB-report, STF44 A74169,
The above loss equations are also valid for
Feb. 1975, Trondheim.
coupled microstrips 8, provided that the di-
electric filling factor, homogeneous impe-
2. E.O.Hammerstad: Equations for Microstrip
dance, and current distribution factor of the
Circuit Design. Conference Proceedingsr
actual mode are used. Presently, no equations
5th. Eu.M.C., Sept. 75, Hamburg,
are available for the odd and even mode cunmt
distribution factors. Except for very tight
3. H.A.Weeler: Transmission-Line Properties
coupling, however, the following approxi-
of a Strip on a Dielectric Sheet on a
mation gives good results
Plane. IEEE-trans. Vol. MTT-25, No. 8,
z Aug. 1977, p 631.
ole+% O. 7
Ke = K. = exp[-1.2( ~ ) 1 (39)
no 4. W.J.Getsinger: Microstrip Dispersion Modd
IEEE-trans., Vol. MTT-21, No.1, ,Jan. 1973,
Anistropy
p. 34.

From recent work 10111 it may be shown that 5. M.K.Krage & G. I.Haddad :Frequency-DepenCkxt
the above models are also valid with aniso- Characteristics of Microstrip Transmission
tropic substrates by defining an isotro- Lines IEEE-trans, Vol. MTT-20, No. 10,
pyized substrate where Oct. 1972, P. 678.

h= h- (40) 6. B.Bianco et al.: Some Considerations About

eq the Frequency Dependence of the Character-
. ~xq (41) istic Impedance of Uniform Microstrip.
E eq
IEEE-trans, Vol. MTT-26, No. 3, March 1978,
P. 182.
This method requires that one of the su~tes
prinCi@at axes is parallel to the substrate 7. W.J.Getsinger: Dispersion of Parallell-
(x-axis) and the other normal to the substrate Coupled Mictostrip IEEE-trans. Vol. MTT-
(y-axis). The strip dimensions remain un- 21, No. 3, March 1973, p. 144.
changed, but are normalized with respect to
The procedure for utilizing the above 8. M.?.kesel: A Theoretical and Experimental
mo
% els then Droceed through calculation of Investigation of Coupled Microstrip Lines.
the mode capacitances, Cm, ELAB-report TE-168, April 1971, Trondheim.

cm = l/(v z (42)
eqm eqm) 9. R.E.Pucel et al.: Losses in Microstrip.
IEEE-trans. , Vol. MTT-16, No. 6, June 1968,
Here v is the mode phase velocity and
z P. 342.
t!%!mrnode impedance on the isotropyized
sl%~trate. Calculating the homogeneous mode
capacitance, C 10. fl.Jensen: Single and Coupled Microstrip
the impedance and effective
dielectric con~~;nt on the anistropic sub- Lines on Anistropic Substrates ELAB
Project Memo No. 3/79, Project no.441408.04
strate are then
June 1979, Trondheim.
= l/(C/Cm Cmi) (43)
Om 11. M.Kobayashi & R.Terakado :Accuratl y Appro-
ximate Formula of Effective Filling
Fraction for Microstrip Line with Iso-
= cm/cml (44)
effm tropic Substrate and Its Application to
the Case with Anistropic Substrate, XEEE-
The inductive trans. ,Vol.MTT-27,No 9,Sept.1979, p.776.
quality factor of the aniso-

. 409