Beruflich Dokumente
Kultur Dokumente
IEEE
TRANSACTIONS
ON
MICROWAVE
THEORY
AND
TECHNIQUES,
VOL.
MTT-23,
NO.
8, AIJGWST975 1
(Special Issue on Microwave FiZters), vol. MTT-13, pp. 676-692, Sept. 1965. accurate calculat~on of uniform Sr., An [10] H. E. Stinehelfe\, microstrip transmwsion lines, IEEE am Tech. (Special Issue on Microwave Integrated W?WV: Czrcutts), hmy vol. MTT-16, pp. 439-444, July 1968. M. A. Earle and P. Benedek, Characteristic impedance of dielectric supported strip transmission line,) IEEE Trans. M~woTheorv Tech. (Corres~.), VO1. M77J-16, PP. 884-885) Oct. wave 1968. Calculation of coefficients of capacitance of [12] W. T. Weeks. rnulticoriducto; transmission lines in the presence o~ a dielectric inter face, IEEE Trans. Microwave Theory Tech., vol. MTT-18, pp. 3543, Jan. 1970. and J. A. Weiss, Parameters of microstrip trans[13] T. G. Bryant [11]
..-.
[16]
wave Theory Tech. (Lett.), vol. MTT-21, pp. 155156, Mar. 1973. A. Farrar and A. T. Adams, A potential theory method for covered microstrip, IEEE Trans. Microwave Theory Tech. (Short Papers), vol. MTT-21, pp. 494-496, July 1973. [17] P. Silvester, TEM wave propertied of mlcrostrip transmission lines, Proc. Inst. EZec. Eng., vol. 115, pp. 43-48, Jan. 1968. and A. T. Adams, Characteristic impedance of [18] A. Farrar microstrip by the method of moments, IEEE Trans. Microwave Theory Tech. (Corresp.), vol. MTT-18, pp. 6566, Jan. 1970. A network analogue method for computing [191 B. L. Lennartsson, the TEM characteristics of planm transmission lines, IEEE Trans. Microwave Theory Tech., vol. MTT-20, pp. 586-591, Sept. 1972. [20] N; Levinson, The Wiener RMS (Root Mean Square) Error Criterim Filter Dewgn and Prediction. in Cambridge, Mass.: ~l&.TI8Press, 1964 (appendix: time series by N. Wiener), pp. [21] R. T. Gregory and D. L. Karney, Testing Computational Algorithms. pp. 134-142. IC technology [22] H. Sobol, Extending Microwaves, Mar. 20, 1967. A Collection New York: to microwave of Matrices for Wiley, 1969, equipments,
Calculation
ALISTAIR
of Microstrip
F. THOMSON
Discontinuity
GOPINATH,
Inductances
AND ANAND
MEMBER, IEEE
AbstractInductive equivalent formulation number and steps experiment. circuits and changes of numerical are method
of the
agreement
with
experiment.
The magnetic
wall model
has
symmetric These
T junctions,
been used for triplate lines which are wide and homogeneous and have confined fields, but its extension to microstrip lines which parison are inhomogenous and much narrower is not completely in comjustified. and are open structures
are presented.
results
compare
INTRODUCTION
Quasi-static calculation of inductance by the moment method [7] has provided results which show reasonable agreement with experiment. However, the disadvantage of discontinuities in these circuits by Deenthis method by Farrar requirements for these three-dimensional and Adams [1] even for modest problems as shown store of some interest. is the very large computer discretization.
HE
of microstrip is currently
equivalent
tailed,
of the parameters
The alterna-
ables easy implementation of paper designs without tedious cut-and-try methods. While the published literature [l] [4] provides curves for the capacitive components of these circuits, little is available for their inductive components. The method of calculation suggested by Horton [5], [6] is not rigorous since the inductance calculation is based on charge estimates and the results obtained are not in
tive is to use a finite-element method as in the skin-effect formulation [8]. The results of discontinuity inductance from this method obtained previously were inaccurate as they were arrived at by subtracting two nearly equal numbers. Also the method was limited to finite-length strips and thus could not represent the actual situation in which the strips extend so far from the discontinuity y that they may be considered as semi-infinite. The present paper is an extension of this finite-element method which
Manuscript received November 19, 1974; revised March 27, 1975. The work of A. F. Thomson was supported by a Science Research Counril Studentship. The authors are with the School of Electronic and Engineering Sciepce, University College of North Wales, Bangor, Caerns., LL57, lUT, Wales, U. K.
overcomes these difficulties, and the results obtained for right-angled bends, step-width changes, and symmetrical T junctions are presented with in the form of curves. experiment [9] Comparison of the results shows reasonable
THOMSON
AND
GOPINATE
: MICROSTRIP
DISCONTINUITY
INDUCTANCES
649
agreement nately,
with
low-frequency alumina
values,
valid
tuting
from
up to 5 GHz
Unfortudepend-
these inductive
components
ent, and quasi-static calculation low-frequency values. However, rigorous method starting time-dependent and associated point program data solution, results
can provide only their in the absence of any it is hoped here provide for this work of discontinuities that the given a firm is cap% pro-
which
Since u is
we are only
vided that these have Manhattan-t ype geometries. This limitation is due to a singularity in the Greens function which is easily integrable However, this limitation pense of computation while at the present The data presented perimentally experiments ation of the symmetric calculation geometries. FORMULATION The imply formulation the following: the current assumes OF PROBLEM quasi-static retardation are involved, conditions effects which is small can be only over a rectangular element. may be circumvented at the exbut was not considered worthexthe which trates is the high-frequency the conductors. of (5) gives v#=o i.e., the impressed potential satisfied condition This is equivalent to saying that the impressed
electric
time, time.
field in the strip conductor is balanced by the magnetically induced fields, and the ohmic drop is negligible. The curl of (5) gives
in this paper by
(6)
performed number
and the evaluof the the value y on the strip conductors. Thus the governing problem two equations were (7) Laplac es equations for this when zero flux pene-
of one example
of measurements.
The divergence
here provides
components
of discontinuity
equations
to be sollved
as it is inconvenient function
compared neglected;
to use (6) because the intein (6) is comof the relaof (7) for d, and then from the following
Greens
Since inductive
The solutions
of the dielectric substrate (provided it is nonmagnetic) may be disregarded, and only the discontinuity y structure and its image in the ground discontinuity semi-infinite The magnetic uniform vector lines. potential ~ due to the current denis given by (1) plane need be considered. The is assumed to be the junction of two or more
to be obtained
and (9)
function
G is 1
the strip
*O) +
(y ?Jo)+
Maxwells structure
equations, is
the electric
therefore the current density ~ is uniform across the thickness of the strip r. Thus ~ is a two-dimensional vector, and volume integrals become surface integrals Equation and surface ones become line integrals. (5) thu~~ becomes (lo)
(2)
jwpor where : G~
Gi] dS = V~
impressed
potential.
Ohms
law re-
(3) Substi-
thickness of the strip conductor; angular frequency; modified Greens function which includes
the ground
650
IEEE
TRANSACTIONS
ON MICROWAVE
THEORY
AND
TECHNIQUES,
AUGUST
1975
plane images 1 i = 4T[(X ~)L + (y YO)Z]112 1 47r[(x h thickness XI))2 + (y yll)z + (2h)2]l/2
region uniform
by
the
planes
PPt
and that
QQ, with
the
constraint
the solution
matches
solution
at these bounding
planes.
netic vector potential due to the semi-infinite lines terminating at these planes PP1 and QQ also requires to be taken into consideration (12) and (11) in (11). gives the current distribution in Solving
of the substrate. @
Define
a new variable
the region bounded by PP and QQ in Fig. 1. The total inductance LT of this region, bounded by PP and QQ in Fig. 1, can then be obtained from (8) and (9), and the required tracting form discontinuity line y inductance L2.. Thus L. is obtained the inductance of uniform planes by subuniof the infinite MM the inductance from of a 21 length of equivalent
infinite L.,
up to the
reference
(7) becomes
172@ =
(12) over the whole diswhere L. L. LT at infiEquation nearly equal (13) gives L. from the subtraction of two numbers; hence the error in L. is approxi1 discontinuity plane AIM inductance long line; total inductance of PP of the section and QQ from between MM PP and reinductance and NN; per unit length between the reference infinitely
The problem
requires
to be solved
continuity y structure
including
the semi-infinite
lines, and
this becomes impractical without further However, the effects of the discontinuity rapidly Laplaces tion and perturbed from the junction how the far as can from the field equation noting from
of the uniform
QQ;
distance spectively. and NN,
some distance
the current
distributions
nite uniform lines. Reference planes are defined at these positions and for a right-angled symmetrical bend discontinuity; distance and NN. these 1 from are PP and QQ as shown reference be solved in Fig. MM 1, the discontinuity need only planes
mately that of the absolute value of error LT, hence accuracy of L. becomes a problem. This is overcome using the excess-current [10]. The (charge) technique of Silvester current-density and Benedek distribuand
M&f
The equation
in the strip
uniform
infinite-line
tion ~m is assumed to exist from co up to MM, from NN onwards to co ; and in the discontinuity
I
sem
(me (1)
sect
I.mfin
ion
ta
1
Section
to NN a known current distribution ~d is defined to preserve continuity of current. Excess circulating current ~. is added in the entire region between PP and QQ so that
section 3
(s2)
(s3)
distributions
(s1)
Iy I
equation.
Tx-xsectmn 4
the solution
The gradient
of the impressed
Fig. 1. Plan view of a symmetric righ+sngle bend in microstrip showing the discontinuity reference planes MM and NN and
the boundary planes PP and QQ. Also of the structure in rectangular elements shown is the 81 to S5. subdivision
!t . Q sema.infmta line
(s4 )
potential V@ which is the right-hand side (RHS ) of this equation can be obtained by the solution of the Laplaces equation on the strip re@ons between PP and QQ with appropriate boundary conditions. Thus (11) becomes
(2)
section
(S5)
where ~d is the sum of ~d and ~. in this region is known. Now ~zm and ~Va can be obtained by solving
THOIkON
AND
OOPINATH
: .MICROSTRIP
DISCONTINUITY
INDUCTANCES
651
(12)
infinite
line with
which A=po
simplifies
il
to
xl
of the strip
nessof the substrate. Since these currents (~~~ and~,w) vary only with respect to Y and x, respectively, the integration with respect to xandy, respectively, can be performed analytically. We also note that ~~m extends from x = w to MM, and Jum extends from y = NN1 to co, thus (14) becomes
/ ffo
Gl~z.w dy +
PO /
Zo
Gz~Udx + M
G,~. dfi.
,!~T
21
PO
/ %0
G2jUW d.
plane MM functions
+ Po
(20)
SZ ST
and
(16) and
(?/0
(x Zo) +
(y. 3/) ) /2
GZ = loge [
Y) + (Yo
{ (!/0 Y) +
Y)2 +
(ZO
~)+
(2h)2)/2
{(YO
Y)+
(ZO
Z) ) l
1
is to
and L_ll~ is the difference in inductance between a length 1 m long from the termination point of semi-infinite line and that of a length with so that 1 m long of infinite from RHS line. This term added to and subtracted it combines each other Also note that (20) : in the addition
(17) Note (15) that these Greens functions method, are obtained plane. for a
semi-iniinite
Equation
first term of (20) to give lL,n. since ~zw and ~U~ are perpendicular
21Lm.
/ S2
Siilarly, simplifications are made for the tlhird (18). Thus the whole equation now becomes
term
in
Consider the first term on the RHS of this equation. (1) we have
From + I(lLm
L_l/2cJ + m //
+/
~-~.dS S3 LT from
LCA~
(22)
652
IEEE TRANsACTIONS
ON MICROWAV2i THEORY
AND
TECHNIQUES,
AUGUST
1975
12-LCAL
=H
PO ~,
~, G,],
d~~z.
dtl
AL?
ALI
w
1-
__
+ No / Although L. is still
/ G,jedSju. sb sT obtained
dS +1 Sa
~~. two
dfi
w
by subtracting
values, (a)
ALI ALZ
The alternative
scheme is to define
V*. in an infinite
(24) uniform
AL , ALI + ALZ
where
obtained
, HII
w
W2
lime and V+. is the excess gradient Also from ~. which RHS
(b)
Gi~dS=~
= _V@~ in (18).
_(V@m+ The
V@@)
T
ALZ I
second
term
of
(c)
2112L.
/ S2 from
dS (25),
! S4
V@UW~U_dS.
(25)
Substituting 1 L.=~ [/
Fig. 2. Equivalent circuits of three microstrip discontinnities with the positions of the reference planes. (a) Symmetric right-angled bend, strip w, wide. (b) Step-width change from strip w, width to wz width. (c) Symmetric T junction, straight arms WI wide; vertical, wz wide. Substrate thickness kept constant at h = 1. The inductances given in these circuits are plotted in Figs. 3-6.
m
w +
W2
ST V@..j.
dS +
/ sT
method
the
term
L-ll,m
is calculated
previously
(26]
a large number of Gaussian points to obtain the accurately, whereas in the latter method, V* must very J. dS accurately is (O. 1 percent) clear cut; to ensure that Thus the Jv+.. accurately determined.
be obtained
functions
these terms
V@. ~m) have all been evaluated It would on first sight that no
in the solution
the latter method would be better since it involves subtraction. However, it is easily shown that V@.. ~. dS = PO Gi~. dS.~=m dS HS2 ST
The method of numerical implementation is by the Galerkln method and is similar to that used by Gopinath and Silvester only difference nondivergent, tor action [8] and thus will not be included density here. The to be a vecis that ~ the current so that Vxw=l ~ requires
+ /Jo
is often negative, and
s4 sT
by defining
potential
I sT
V+.. ~e dS +
I Sr
V*W . ~. dS =
I ST
~.~.dS
and hence V.?i = o is satisfied. Since ~ is a two-dimensional vector, need only have a z-directed component. ing equations are recast, replacing J, being zero, W Thus the preced(V X W.) and
the subtraction of the two numbers. in the first method, where the result difference or sum of two numbers. Also in the latter method the ratio
~ by
to I is important since terms are divided by 12, and thus any error will appear in the results. However, in the former method, all terms involve the product of two currents, and thus any error in magnitude cancels out. In the
~. by (V X W.e) in the numerical solution. In this case, it was necessary only to ensure continuity of normal component of current between elements. or potential W at common boundaries This was implemented using the gen-
THOMSON
AND
GOPINATH
: MICROSTRIP
DISCONTINUITY
INDUCTANCES
653
0 6r
05
03 -
11
.
experimental expemmental
bars # results
//
calculated results / /
02 -
02 -
d~ 1:0
1
05
,
20
-01 -
-02 %
Pi /
right-angled
bend
Comparison is with
//
calculated result forwlh, 10
/
O
I
0.
experimental
bar
for
wlh.1
d z aj1= c
0.(
1.0
Fig. 4. Normalized inductance for w,/h = 0.5 and 1.0 for for one value me marked.
1
20 s
h
1
4.0
3.0
matrix
inverse
tectilque
also
discussed
previ-
method
of images
[11].
Althougll~
this
test
is
RESULTS Preliminary
AND
DISCUSSION to check the inductlines segmented in the These results were in from Silvesters
operation
ante of straight uniform center for cliff erent w/h excellent agreement with
Subsequently, a large number of runs were performed on the step-width change, symmetric right-angled bend, and symmetric T junction, for a number of different w/h. These results compared favorably with experimental resuits due to Easter [9] as shown in Figs. 26. It is apparent
those obtained
654
123EETRANsACTIONS
ON MICROWAVE
THEORY
AND
TECHNIQUES,
AUGUST
1975
w2/h
oh
05
10
15
20
-o
-01 -
\ \ \ \
-o
Z8 a-i
F
\, \.
\. \
\x \ \ \ \ \ \
-o.2-
i ~: 12
-o
w#7=10
expermwnt exp.mnwntaf calculated A bars results result$
\
Y
-03-
\
\ \
. ._w+. \
\ \ \ \ % W,/h= . . W*.1O v+@
o 5
-o
VOQ,21 results
15 20
(a)
w2/h
-c
(a)
-01
05
10
15
2 o
-C32
\ \ I \ \
07 t
\
0s
1. \
09
1
I
y/hz10
\
\ bars r.zsultz results ~,
\,
\ \ \ ?
-09
3o
. .
vogd
resu n~
(b)
-11
w,lh wl/h
= 20 = 1 5 1 0
\x
. . ..__
\.\
\. \. \. \ \. \%
~1 /~=
- . -
W,h=
0 5
-1.2 L
Fig. 5.. (a) Normalized inductance of the straight arm of the T junction AL1/(Lov,h) for w,/h = 1.0 for different wt/h. Comarisen is with the experimental results of Easter [9] and calcu~ated due to Vogel [12]. (b) Normalized inductance of the vertical leg of the T, &Lt/(L~wjh) for w,/h = 1.0 for different wt/h. Corparison is with experiment [9] and Vogel [12].
(b) Fig. 6. (a) Normalized inductance of the straight arm of the T, AL1/(L_,h) for w,/h = 0.5, 1.0, 1.5, and 2.0 for different wz/h. n(b)Normalizwf inductance of the vertical leg of the T, AL/(L_&) for W, /h = 0.5, 1.0, 1.5, and 2.0 for different w2./h.
that
the
agreement
worsens
when
the
experimental
un-
straight
arms
a few percent
of the
ducta,nce values m-e highly frequency dependent, the estimate falling outside the experimental error range. It is expected that the results for st~p-width changes and the
actual values. For the bend and the vertical leg of the T, the results are somewhat worse and the reason for thk are not determined. Fig. 5(a) and (b) shows comparison ~f
IEEE
TRANSACTIONS
ON MICROWAVE
THEORY
AND
TECHNIQUES,
VOL.
MlW-23,
NO. 8, AUGUST
1975
655
results.
that
shows that
line lengths
decrease
increasing
frequency. CONCLUSION
[5]
A method
of calculating
quasi-static
microstrip
discon-
[6]
ti&ity inductance has been outlined. The computer program written for these calculations has given results which compare favorably with experiment. Curves for some widely used discontinuities have been provided. ACKNOWLEDGM13NT The authors cussions during wish to thank B. Easter for valuable dis-
[7]
[8]
[9]
[10]
[11] [12]
[1]
and A. T. Adam:, cOmpUtfitiOh of lumped microstrip capacitances by matrix methodsRectangular sections and end effects, IEEE Trans. Microwave Theory Tech. (Gorresp.), vol. MTT-19, pp. 495-496, May 1971. [2] , Matrix method for microstrip three-dimensional prob-
A. Farrar
lems, IEEE Trans. Microwave Theiny Tech., vol. MlT-20, pp. 497-504, Aug. 1972. D. S. James and S. H. Ike. Miwostri~ . end effects. Eka!ron. Lett., vol. 8, pp. 46-47, 1972. I. Wolff, Statischie IGipazitaten von Rechteckigen und KreisArch. Eklc. fprmigen MikrostripScheibeukcmdiensatorem, ~bertragung., vol. 27, pp. 44-47, 1973. R. Horton, Electrical characterization of a right-angled bend in microstrip line, IEEE Trans. Microwave Theory Tech, (Short Papers), vol. MTT-21? pp. 427-429, June 1973. , Electrical representation of an abrupt impedance step in microstrip line, IEEE Trans. Microwave Theory Tech. (Short Papers), vol. MTT-21, pp. 562-564, Aug 1973. A. Gopinath and B. Easter, Moment method of calculating discontinuity inductance of microstrip righi%angled bends, IEEE Trans. Microwave Theory Tech. (Short Papers), vol. MTT-22, pp. 880-883, Oct. 1974. A. Gopinath and P. Silvester, Calculation oj inductance of finite-lenxth striws and its variation with freauenrw. IEEE Trans. Microwa~e Theory Tech., vol. hITi-2i, pp~ 380-386, June 1!373. B. Easter, An equivalent, circuit of some microstrip discontinuities, this issue, pp. 655-660. P. Silvester and P. Benedek, Equivalent capacitances of microstrip open circuits, IEEE Trans. Microwave Theory Tech:, vol. MTT-20, pp. 511-516, Aug. 1972. ,P. Sdvester, TEM wave properties of microstrip transmission lines, Proc. Inst. Elec. Eng., vol. 115, pp. 43-48, Jan. 1968. R. W. Vogel, Effects of the T-junction discontinuities on the design of microst,rip directional couplers, IEEE Trans. Microwave Theory Tech. (Short Papers), vol. MTT-2 II, pp. 145146, Mar. 1973.
Discontir)uities
of common
synunetricaf
equivalent circuit data fall broadiy into two categories. First, there is a growing body of information [1]--[4] on the quasi-static In due capacitance more recently course this of microstrip structures, provide but, planesupplemented [5]-[7]. accurate by studies approach of the inductance should of interest,
uniform better
requires plane-wave
it is generally
quasi-static
model.
no indication derives
effects. A second
on the properties of uniform rnicrostrip trast, the circuit designer is often without formation comprised on the discontinuity and in typical practical circuits.
HERE
IS NOW
available
a substantial
body
of data
wave parallel plate model of the microstrip cross section as described by Vogel [14] and others, following the method applied by Oliner [8] to symmetrical (( l.rihas the great advautage plate ) strip lines: This approach of reducing the problem to two-dimensional complexity. Closed-form expressions are available in classical texts [9] for several quasi-static elements, and i!ome studies [10], [11] have directly tackled the dynamic situation. In addition, reference to Babinet equivalence enables the use of rectangular waveguide data. Against these advantages, there remains the difficulty of any estimation of the error associated with the use of the uniform plane-wave model. It can be noted that not only is the proportion of fringe
Manuscript received November 19, 1974; revised March 27, 1975. This work was carried out as a part of a study supported by the United Kingdom Scien6e Research Council. The author is with the School of Electronic Engineering Science, University College of North Wales, Bangor, Caerns., LL57 lUT, Wales, U. K.