Implementation of Single and Coupled Microstrip

Lines in APLAC
Luis Costa∗ and Martti Valtonen∗

CT-33
December 1997

Abstract
The transmission line model for the single and coupled microstrip line com-
ponents implemented in APLAC are presented. A detailed description of the
algorithm used to calculate the characteristic impedance and the effective per-
mittivity is given for the single microstrip, the symmetrical and asymmetrical
pair of coupled microstrip lines, and the n symmetrical coupled microstrip
lines. The effects due to shielding, and conductor and dielectric loss are also
accounted for.

Indexing terms: microstrip line, symmetrical coupled microstrip lines, asym-

metrical pair of coupled microstrip lines, microwave device, transmission line.

ISBN 951-22-3881-0
ISSN 1239-8233
ESPOO

∗
The author is with the Helsinki University of Technology, Faculty of Electrical Engineering,
Circuit Theory Laboratory, P.O. Box 3000, FIN-02015 HUT, Finland.
Contents
1 Introduction 1

2 Transmission line model for microstrip components 2

2.1 Ideal transmission line . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Microstrip line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.1 Static approximation of the characteristic impedance and ef-
fective permittivity including effects due to shielding . . . . . 4
2.2.2 Effect of finite strip thickness . . . . . . . . . . . . . . . . . . 6
2.2.3 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.4 Attenuation loss . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.5 Effective parameters . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.6 Open-ended microstrip line . . . . . . . . . . . . . . . . . . . 12
2.3 Coupled transmission lines . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Solution to the eigenvector problem of symmetric n coupled trans-
mission lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Impedance of symmetric n coupled microstrip lines . . . . . . . . . . 16
2.6 Symmetric pair of coupled microstrip lines . . . . . . . . . . . . . . . 18
2.6.1 Static parameters including effects due to shielding . . . . . . 18
2.6.2 Correction due to finite strip thickness . . . . . . . . . . . . . 23
2.6.3 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6.4 Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.7 Asymmetric pair of coupled lines . . . . . . . . . . . . . . . . . . . . 28
2.7.1 Equivalent capacitances . . . . . . . . . . . . . . . . . . . . . 29
2.7.2 Dispersion model . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Specific functions implemented in APLAC 33

3.1 C-level implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Input-file level functions . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Conclusion 36

Acknowledgements 37

References 38

i
1 Introduction
The objective of this report is to provide a comprehensive documentation on the
symmetric and asymmetric coupled microstrip line models implemented in APLAC,
an analog circuit simulator [16]. Since the models for the symmetrical and asym-
metrical coupled microstrip lines use equations governing the single microstrip line,
the models for the single microstrip line, proposed by Hammerstad and Jensen [4],
and Jansen and Kirschning [7], are also discussed here. The discussion on the single
microstrip does not cover the APLAC microstrip model entirely and excludes the
effects of a microstrip short-circuited at one end and the case when a ground-plate
is placed alongside the microstrip (coplanar waveguide case). Also, microstrip con-
figurations like the step, gap, bend, taper, T- and X-junctions are excluded from
this discussion.
A complete description on the syntax and use of these components is given in
[16]. The following discussion describes the closed form empirical equations and
numerical solutions used to model the said components. All equations here have
been taken from the literature, and no attempt is made to derive them.
The symmetrical and asymmetrical pair of coupled microstrip lines had earlier
been implemented in APLAC but did not function as expected. The basic problem
was that the coupled microstrip line components in pre-6.25 versions did not show
correct behaviour as conductor thickness was increased from the ideal zero-thickness
case. As a result, a thorough investigation into the problem was begun which led
to the complete reconstruction of the model for the symmetric coupled pair of mi-
crostrip lines, and is now based on the model proposed by Kirschning and Jansen [9].
The model for the asymmetric pair of coupled microstrip lines is based on the model
proposed by Sellberg [14] and effects due to dispersion in this model are according
to Tripathi [15]. The model for symmetric n coupled transmission lines is based on
[13]. Some of the changes made, especially those affecting the functions governing
the effects due to finite thickness, affect the model for the single microstrip line. All
the equations required to construct these models are reproduced in this report.

1
2 Transmission line model for microstrip compo-
nents
The physical dimensions of lumped elements may be larger than the wavelength of
the propagating microwave voltage and so they cannot be used to model circuit
elements in the microwave frequency region that ranges from 300 MHz to 30 GHz.
Instead, the transmission line (figure 1), with distributed elements, is used for this
purpose.
Dispersion results when the frequency components of a voltage pulse propagating
along the transmission line propagate with different phase velocities, so distorting
the pulse. Dielectric and conductor losses grow rapidly with increasing frequency so
also causing distortion. These phenomena should be accounted for when designing
microwave circuits.
This section begins with a brief review of the relations governing the ideal trans-
mission line required to get an understanding of the microstrip component models
described later.

2.1 Ideal transmission line

Z0 , the characteristic impedance, τ , the propagation delay that depends on the
length of the line, and α, the attenuation constant, are the parameters used to
model the transmission line. The transmission line is considered to consist of a large
number of infinitely short segments of lumped elements of length dz, with a series
impedance

Zs = R + jωL,

where R and L are the distributed resistance and inductance, respectively, and a
shunt admittance

Yp = G + jωC,

where G and C are the distributed capacitance, respectively, distributed along the
entire line. One such segment is shown in figure 1.
A pair of first-order differential equations, called the telegrapher’s or telegraphist’s
equations are used to describe the transmission line mathematically:


 ∂U (z)

 = −Zs I(z)
∂z (1)

 ∂I(z)

 = −Yp U (z) .
∂z
U and I are the voltage and current, respectively, of a differential segment of the
transmission line as shown in figure 1(b), and z denotes distance along the line. The

2
 I1 I2 I Zs I + dI

U1 Z0 , τ, α U2 U Yp U + dU

(a) (b)

Figure 1: (a) Transmission line. (b) Equivalent circuit of a differential length dz of

a two-conductor transmission line.

telegrapher’s equations can be combined to give the wave equations

∂ 2 U (z)
2
= γ 2 U (z)
∂z (2)
∂ 2 I(z)
= γ 2 I(z) ,
∂z 2
where the propagation constant γ (1/m) is
q
γ= Zs Yp = α + jβ .
The real part α of the propagation constant is called the attenuation constant
(Np/m) and the imaginary part β (rad/m) is the phase constant, though neither
α nor β are really constants but are complicated functions of angular frequency ω.
The solution of the wave equations can be thought to consist of two waves propa-
gating in opposite directions. The impedance of the line is the ratio of the voltage
to the current in a given direction. For an infinitely long line, the impedance is
independent of z and is called the characteristic impedance of the line; it can be
written as
s
Zs γ Zs
Z0 = = = .
γ Yp Yp
The phase velocity of the propagating wave is obtained from
ω
v = λf = .
β
For a lossless line, R = G = 0 and so also α = 0. This means that
Zs = jωL Yp = jωC (3)
s
L
Z0 = (real) (4)
C
√
γ = jβ = jω LC (5)

3
 6
h2  w - ?
? t
6 6 ǫr
h
?

Figure 2: Cross-section of a microstrip line.

s
1
v= . (6)
LC

This approximation is reasonable for frequencies of up to about 4 GHz since losses

are small in this range.
√
The quasistatic approximation, where the phase velocity v = 1/ µǫ, is also very
useful up to the range of about 4 GHz, beyond which effects due to dispersion have
to be accounted for. Here µ is the magnetic permeability and ǫ the permittivity of
the medium in which the transmission line conductors are placed.
The microstrip models described below are all frequency-domain models.

2.2 Microstrip line

The microstrip line is composed of a thin conducting metal strip of width w, thick-
ness t and length l placed on a nonmagnetic dielectric substrate that is in turn
placed on a conducting metal ground-plane. The substrate has thickness h and rela-
tive permittivity ǫr . The structure is shown in figure 2. Above the conducting strip
is a second dielectric material, air. In addition, the structure may have a cover plate
at a height h2 from the substrate surface so introducing effects due to shielding.

2.2.1 Static approximation of the characteristic impedance and effective

permittivity including effects due to shielding
The microstrip line static model gives accurate results for low frequencies (up to
about 4 GHz). It uses physical dimensions of the microstrip line as parameters from
which static i.e. zero frequency, impedance and effective permittivity are calculated.
The equations here are from [4] and equations governing shielding effects are taken
from [11].
The physical dimensions of the microstrip are normalized with respect to the
substrate height h. The following symbols are used to denote the normalized di-
mensions:
w t h2
u= , th = , h2h = . (7)
h h h

4
 The model first calculates the impedance of the microstrip line in a homogeneous
medium. This means that the substrate material and the material above the con-
ductor are the same. The equations here use air, whose relative permittivity ǫr = 1,
as the homogeneous material. The impedance is
 s 
 2
η0  f (u) 2 ,
Z0∞ (0, u) = ln + 1+ (8)
2π u u

where
"  0.7528 #
30.666
f (u) = 6 + (2π − 6) exp −
u

and η0 = 376.73 Ω is the wave or intrinsic impedance of free space. The subscript 0
indicates a homogeneous air dielectric and the argument 0 refers to zero frequency,
i.e. static values. Subscript ∞ indicates that the cover height h2 = ∞, and so
equation (8) does not include effects due to shielding.
Shielding effects are accounted for by subtracting a correction term ∆Z0h2 (0, u)
from Z0∞ (0, u) [11]. Hence the effective impedance with shielding accounted for is

Z0 (0) = Z0∞ (0, u) − ∆Z0h2 (0, u) , (9)

and the correction term is given by

∆Z0h2 (0, u) = P Q (10)

  q 
1.389
P = 270 1 − tanh 1.192 + 0.706 1 + h2h −
1 + h2h
!
0.012u + 0.177u2 − 0.027u3
Q = 1.0109 − artanh .
(1 + h2h )2

APLAC is coded in the C programming language which does not have a built-in
artanh function. For convenience the function is defined below:
1 1+x
artanh(x) = ln .
2 1−x
In order to account for effects due to shielding, the development of the concept
of the filling factor q was necessary. According to March [11]

q = (q∞ − qt )qc , (11)

where qc is the correction for a finite cover height h2 given by

 
1.164
qc = tanh 1.043 + 0.121h2h − , (12)
h2h

5
qt is the shielding correction due to finite conductor thickness given by
2 ln 2 th
qt = ·√ , (13)
π u

and q∞ is the filling factor for an open microstrip (infinite cover height, h2 = ∞)
with zero conductor thickness given by
 −a(u) b(ǫr )
10
q∞ = 1 + (14)
u
 
u 2 "
u + 4   #
1 52 1 u 3
a(u) = 1 + ln 4 + ln 1 +
49 u + 0.432 18.7 18.1
 0.053
ǫr − 0.9
b(ǫr ) = 0.564 .
ǫr + 3

The effective relative permittivity is, thus, given by

ǫr + 1 ǫr − 1
ǫeff (0, u) = + q. (15)
2 2
Again the zero in the argument means zero frequency indicating a static effective
relative permittivity.
In order to get the characteristic impedance of the microstrip line placed on
a dielectric with relative permittivity ǫr we need to divide the impedance of the
microstrip line in a homogeneous material by the square root of the effective relative
permittivity, i.e.
Z0 (0)
ZL (0) = q . (16)
ǫeff (0, u)

The above equations are accurate to within 0.5 percent in the range ǫr < 60,
0.01 ≤ u ≤ 60 and h2h > 1. The accuracy of the impedance and effective relative
permittivity without shielding is better than 0.2 percent in the range ǫr ≤ 128 and
0.01 ≤ u ≤ 100.

2.2.2 Effect of finite strip thickness

The effect of finite conductor thickness t is calculated according to the method
proposed in [4]. In the first step, a normalized width correction term accounting for
the finite strip thickness is calculated for a homogeneous microstrip structure from
√ !
t 4e tanh2 6.517u
∆ut,1 = ln 1 + (17)
π t

6
from which the resulting effective width is obtained as

ut,1 = u + ∆ut,1 . (18)

e in the expression for the correction term is the Neperian base 2.71828. This
effective width ut,1 is then used in the place of u in equation (8) to calculate the
resulting impedance Z0∞ (0, ut,1 ) of the homogeneous microstrip structure.
In the next step, the effective width for the structure having a substrate with a
relative permittivity ǫr is determined using
!
∆ut,1 1
ut,ǫr =u+ 1+ √ . (19)
2 cosh ǫr − 1

The effective relative permittivity of this structure is then determined using the
expression
" #2
Z0∞ (0, ut,1 )
ǫeff,t (0) = ǫeff (0, ut,ǫr ) , (20)
Z0∞ (0, ut,ǫr )

where ǫeff (0, ut,ǫr ) is the effective relative permittivity obtained from equation (15)
using ut,ǫr instead of u, as indicated by the argument ut,ǫr , and Z0∞ (0, ut,1 ) and
Z0∞ (0, ut,ǫr ) are the impedances similarly obtained from equation (8) using ut,1 and
ut,ǫr , respectively.
Finally, the characteristic impedance of the microstrip structure having finite
conductor thickness is calculated from
Z0∞ (0, ut,ǫr )
Z0∞,t (0) = q . (21)
ǫeff,t (0)

2.2.3 Dispersion
Both the characteristic impedance and the effective permittivity of the microstrip
line change with frequency due to dispersion. The effective relative permittivity
increases with frequency and asymptotically approaches ǫr . Dispersion causes a small
increase in the characteristic impedance of the microstrip line for high frequencies.
The effects due to dispersion should be taken into account for high frequencies (above
about 4 GHz). In the equations below ǫeff (0) is either ǫeff (0, u) or ǫeff,t (0), depending
on whether the effect of finite strip thickness is ignored or not.
The dispersion model proposed in [4] is used in LEVEL 1 calculations in APLAC.
The effective permittivity is calculated from

ǫr − ǫeff (0)
ǫeff (f ) = ǫr − !2 . (22)
f
1+G
fp

7
fp is an approximation of the cut-off frequency of the first transverse electromagnetic
(TEM) mode and is given by
ZL (0)
fp = ,
2µ0 h
where µ0 = 4π × 10−7 H/m is the permeability of vacuum, and G is an empirically
determined factor that is sufficiently accurate for all substrates in use, given by
s
π 2 ǫr − 1 2πZL (0)
G= .
12 ǫeff (0) η0
Argument f indicates a frequency dependence.
The LEVEL 2 dispersion model is based on [7], and is a sequence of expressions
that are easily programmable. The expressions for the effective permittivity are [6]
ǫr − ǫeff (0)
ǫeff (f ) = ǫr − (23)
1 + P (f )

P (f ) = P1 P2 [(0.1844 + P3 P4 )fn ]1.5763

" #
0.525
P1 = 0.27488 + u 0.6315 + − 0.065683 · exp(−8.7513u)
(1 + 0.0157fn )20

P2 = 0.33622 [1 − exp(−0.03442ǫr )]
  !4.97 
 fn 
P3 = 0.0363 · exp(−4.6u) 1 − exp − 

38.7
( "  8 #)
ǫr
P4 = 1 + 2.751 1 − exp − ,
15.916

where fn GHz·mm is the frequency normalized with respect to the substrate height,
fh
fn = .
106
The effective characteristic impedance due to dispersion is [7]
 R17
R13
ZL (f ) = ZL (0) (24)
R14

R1 = 0.03891ǫ1.4
r

R2 = 0.267u7
 
R3 = 4.766 · exp −3.228u0.641

8
 R4 = 0.016 + (0.0514ǫr )4.524
!12
fn
R5 =
28.843

R6 = 22.2u1.92

R7 = 1.206 − 0.3144 · exp(−R1 ) [1 − exp(−R2 )]

  !2.745 
 fn 
R8 = 1 + 1.1275 1 − exp −0.004625R3 ǫ1.674
r


18.365

R5 exp(−R6 ) (ǫr − 1)6

R9 = 5.086R4
0.3838 + 0.386R4 1 + 1.2992R5 1 + 10(ǫr − 1)6

R10 = 0.00044ǫ2.136
r + 0.0184
!6
fn
19.47
R11 = !6
fn
1 + 0.0962
19.47

1
R12 =
1 + 0.00245u2

R13 = 0.9408ǫeff (fn )R8 − 0.9603

R14 = (0.9408 − R9 )ǫeff (0)R8 − 0.9603

!1.097
fn
R15 = 0.707R10
12.3
( "  6 #)
u
R16 = 1 + 0.0503ǫ2r R11 1 − exp −
15
   
R12
R17 = R7 1 − 1.1241 exp −0.026fn1.15656 − R15 .
R16

In the above expressions the terms R1 , R2 and R8 should be restricted to nu-

merical values less than or equal to 20 in order to prevent problems due to overflow
in the computer.

9
2.2.4 Attenuation loss
Losses due to dissipation in a microstrip structure are made up of dielectric losses
and conductor losses. The total loss or attenuation is given by the attenuation
coefficient in dB/m by

α = αc + αd , (25)

where αc and αd are the attenuation coefficients due to conductor and dielectric
losses respectively.
The dielectric loss in dB/m is given by [2]
20π f ǫr ǫeff (0) − 1
αd = q tan δd , (26)
ln 10 c0 ǫeff (0) ǫr − 1

where c0 is the velocity of light, tan δd is the loss tangent of the dielectric material
and ǫeff (0) is calculated from equation (15). In APLAC the loss tangent tan δd is
given by the user using the identifier TAND [16].
Conductor loss in dB/m is derived from [4] and is given by
q
20π f ǫeff (0)
αc = . (27)
ln 10 c0 Qc
The factor Qc is the strip inductive quality factor and is approximated by
πZ0 (0)hf u
Qc = , (28)
Rs c 0 K
where Z0 (0) is the static impedance of the microstrip line in a homogeneous medium
from equation (8), Rs is the surface resistance (due to the skin effect) and K is the
current distribution factor. Rs is an increasing function of surface roughness ∆,
( "  2 #)
2 ∆
Rs (∆) = Rs (0) 1 + arctan 1.4 , (29)
π δ

Rs (0) is the skin resistance for a smooth surface given by

1
Rs (0) = , (30)
σδ
and σ and δ are the conductivity and skin depth, respectively, of the conductor.
The argument zero in equation (30) indicates a smooth surface, ∆ = 0, and not
frequency as before. The skin depth, defined as the depth below the conductor
surface at which the current density decreases to 1/e of its maximum value at the
conductor surface, is obtained from
1
δ=√ . (31)
πµ0 f σ

10
The current distribution factor K is
 !0.7 
Z0 (0)
K = exp −1.2  . (32)
η0

The conductor loss (αc ) calculations above are valid for a minimum conductor
thickness of t ≈ 3δ.
The dielectric loss αd is usually very small compared to the conductor loss αc .

2.2.5 Effective parameters

The phase velocity of the wave propagating in the microstrip line is

c0
vp = q , (33)
ǫeff (f )

where c0 is the velocity of light. The propagation delay is easily found once the
phase velocity is known. The capacitance and inductance per unit length can be
obtained by manipulating equations (4) and (6) as

1
C= (34)
vp ZL (0)

ZL (0)
L= . (35)
vp

For low frequencies the attenuation constant α is given by [5]

R
α= Np/m .
2ZL (0)

Since α is known in dB/m, the series resistance per unit length is obtained as

ln 10
R= αc ZL (0) . (36)
10

Similarly, the shunt conductance per unit length is obtained from

ln 10 αd
G= . (37)
10 ZL (0)

11
2.2.6 Open-ended microstrip line
The open-ended microstrip line is modeled by a small extension ∆l in the length of
the strip that is added to the physical length. Two models have been implemented
in APLAC. The model proposed by Hammerstad [3] is used in LEVEL 1 simulation
and the more accurate and elaborate model proposed by Kirschning, Jansen and
Koster [10] [5] in LEVEL 2.
Hammerstad’s model is given by
 
∆l u + 0.106 ǫr + 1
= 0.102 1.166 + [0.9 + ln (u + 2.475)] . (38)
H u + 0.264 ǫr
Kirschning et al. have presented a sequence of equations to model the extension
in length. The equations are
∆l
= ACE/D (39)
H

(ǫeff (0)0.81 + 0.26) (u0.8544 + 0.236)

A = 0.434907
(ǫeff (0)0.81 + 0.189) (u0.8544 + 0.87)

u0.371
B =1+
2.358ǫr + 1

0.5274  
1.9413/B
C =1+ arctan 0.084u
ǫeff (0)0.9236
 
D = 1 + 0.0377 {6 − 5 · exp [0.036(1 − ǫr )]} + arctan 0.067u1.456

E = 1 − 0.218 · exp(−7.5u) .

2.3 Coupled transmission lines

The model for n coupled lossless transmission lines shown in figure 3 consists of n
uncoupled lines and coupling transformation network banks at the input and the
output ports. It is based on the assumption that the TEM mode is the only mode
of propagation in the lines. The voltages and currents on a lossless n-line system
are given by the generalized telegrapher’s equations [13]
" # " #" #
v z (z, t) 0 L v t (z, t)
=− , (40)
iz (z, t) C 0 it (z, t)

where vectors v(z, t) and i(z, t) denote the voltages and currents, respectively, and
z and t denote distance and time, respectively. Superscripts z and t denote differ-
entiation with respect to space and time, respectively. L and C are the inductance

12
 i1 (0, t) i1 (l, t)
r r
î1 (0, t) î1 (l, t)
v1 (0, t) v1 (l, t)
r vˆ1 (0, t) c , τˆ
Z vˆ1 (l, t) r
1 1

TRANS- TRANS-
. FOR- FOR- .
MA- . MA-
. . .
TION . TION
. NET- în (0, t) în (l, t) NET- .
WORK WORK
in (0, t) in (l, t)
r vˆn (0, t) c , τˆ
Z vˆn (l, t) r
n n

vn (0, t) vn (l, t)
r r

Figure 3: Transmission line model for the n coupled line system.

and capacitance matrices whose elements represent self and mutual parameters per
unit length of the lines. Matrix C is symmetric and is given by
 
C1,1 −C1,2 · · · −C1,n
 .. .. .. 
C=
 . . ··· .  ,
 (41)
−Cn,1 −Cn,2 · · · Cn,n

where the diagonal

n
X
Ci,i = Ci,0 + Cij , (42)
j=1, j6=i

Ci,0 is the capacitance per unit length of line i with respect to ground and Ci,j is
the capacitance per unit length between line i and line j.
For structures of interest, with either a single or multilayered dielectric medium
whose magnetic properties are the same as those of free space,

L = L0 = µ0 ǫ0 C 0 −1 , (43)

where C 0 is the capacitance matrix of the same set of transmission lines with the
dielectric replaced by air. In other words, the inductance matrix L is calculated
by determining the capacitance matrix for the set of transmission lines with the
dielectric medium replaced by air.

13
 In order to map the voltage vector v onto v̂ and the current vector i onto î the
following is done:

v = M V v̂

i = M I î ,

where M V and M I are the right eigenvector matrices of the LC and CL matrices
respectively. Due to the structure of the physical problem, matrices LC and CL
have the same eigenvalues and

M −1 ⊤
I = MV ,

where superscript ⊤ means the transpose operation of the matrix. As a result the
inductance and capacitance matrices map as follows:
h i⊤
b = M −1 L M −1
L V V

c = M ⊤ CM .
C (44)
V V

Thus, in order to compute the transformation network and the parameters for
the system of n coupled transmission lines, the matrix of right eigenvectors of matrix
LC (or CL) needs to be computed. For a non-symmetric matrix, the eigenvector
matrix is not easily computable, if at all.

2.4 Solution to the eigenvector problem of symmetric n cou-

pled transmission lines
The discussion proceeds with the case for symmetric coupled transmission lines
where it is assumed that identical lines are equally spaced. It is also assumed that
only the mutual capacitances between adjacent lines are considered, which results
in an nth order symmetric tridiagonal capacitance matrix
 
C1,1 −C1,2 0 0 ··· 0
 


−C1,2 C1,1 −C1,2 0 ··· 0 


C= .. .. .. 
 . (45)
 . . ··· . 
 0 ··· 0 −C1,2 C1,1 −C1,2 
 
0 ··· 0 0 −C1,2 C1,1

The eigenvector matrix of a symmetric matrix is easily computed using the method
suggested in [13].
The first step is to compute the eigenvalues of the n-order matrix T defined by
Tij = 1 , |i − j| = 1,
i, j = 1, · · · , n (46)
Tij = 0 , otherwise,

14
 whose characteristic polynomial φ(µ) is given by the recursive expression

φk (µ) = µφk−1 (µ) − φk−2 (µ) , k = 2, · · · , n (47)

φ0 (µ) = 1

φ1 (µ) = µ .

The eigenvalues µ are the solutions to the equation φn (µ) = 0 and are given by

iπ
µi = −2 cos , i = 1, · · · , n. (48)
n+1
This step is dependent on the number of lines only and so needs to be calculated
only once.
The next step is to compute the matrix M of right eigenvectors of T from

φi−1 [µj (T )]
Mi,j = , i, j = 1, · · · , n (49)
γj

n
X
γj2 = {φi−1 [µj (T )]}2 .
i=1

Now with the eigenvector matrix computed, the voltages or currents for each de-
pendent source in figure 4 may be computed from
n P
vi (z, t) = j=1 Mi,j v̂j (z, t) − v̂i (z, t)
Pn (50)
ii (z, t) = j=1 Mi,j îj (z, t) − îi (z, t) ,

where z = 0 or z = l, l being the length of the transmission line.

ei (0, t) ei (l, t)
iˆi (0, t) iˆi (l, t)
ii (0, t) ii (l, t)
vi (0, t) ji (0, t) vˆi (0, t) c , τ̂
Z v̂i (l, t) ji (l, t) vi (l, t)
i

Figure 4: Complete model for the ith transmission line segment.

The following sections describe methods used to compute the L and C matrices,
from which the impedance matrix Z c may be computed.

15
2.5 Impedance of symmetric n coupled microstrip lines
The symmetric n coupled microstrip line structure is made up of n identical metal
conductors of width w, thickness t, length l, and with spacing s between them placed
on a nonmagnetic dielectric of height h that is in turn placed on a metal ground
plane.
In addition to the earlier assumptions, it is assumed that the only significant
coupling capacitances are between adjacent strips and all other side effects are con-
sidered negligible. Hence, only the parameters for a symmetrical pair of coupled
microstrip lines need to be calculated. The static even- and odd-mode impedances
are first calculated from equations (69) and (70) in section 2.6, from which the
impedance matrix for the system of n coupled lines is then computed. The circuit
parameters (per unit length) of the symmetric pair of coupled microstrip lines are
shown in figure 5.

b
L

d
L d
M CM Cb

b
L
Cb

Figure 5: Circuit constants per unit length for a pair of symmetric coupled mi-
crostrips.

The following expressions for the phase velocities of both the even and odd modes
and consequently the equivalent inductances and capacitances of the pair of coupled
lines are from [5]. The phase velocity is given by
c0
vp,m = q , (51)
ǫeff,m (0)

where subscript m is either e or o for the even and odd modes, respectively, and
c0 ≈ 3 · 108 m/s is the velocity of light. The equivalent strip-to-ground and mutual
capacitances of the structure (see figure 5) are calculated from
1
Cb = (52)
vp,e ZL,e (0)
!
d = 1
C
1
−
1
. (53)
M
2 vp,o ZL,o (0) vp,e ZL,e (0)

16
ZL,e (0) and ZL,o (0) are the even- and odd-mode static characteristic impedances of
a symmetric pair of coupled microstrip lines and their significance is explained in
section 2.6.
The capacitance matrix for a symmetric pair of coupled lines is of the form
" #
Cb + C
d
M
d
−CM
C= d b d . (54)
−CM C + CM
For the system of n coupled lines with couplings between adjacent lines only the
capacitance matrix is of the form
 
Cb + C d
M −Cd
M 0 0 0 ··· 0
 
 d b
−CM C + 2CM d d
−CM 0 0 ··· 0 
 
 d b d d 


0 −CM C + 2CM −CM 0 ··· 0 

C =

.. ... ... ... .. 
.(55)
 . . 
 d b d d 


0 ··· 0 −CM C + 2CM −CM 0 

 0 ··· 0 0 d
−CM b d
C + 2CM −C d 
 M 
0 ··· 0 0 0 d
−CM b
C + CMd

It has been observed from analysis of practical circuits that more accurate results
are obtained when the matrix diagonal is obtained from

c =C
b+ 2n d
C1 CM (56)
2+n
c =C
for an n strip structure instead of C b + 2C
d (see equation (42)). As the number
1 M
of strips increase the factor approaches the value two, which implies coupling only
between two adjacent strips.
The equivalent strip inductance L b and the mutual inductance Ld between the
M
two strip inductances are given by
!
b = 1
L
ZL,e (0) ZL,o (0)
+ (57)
2 vp,e vp,o
!
d = 1
L
ZL,e (0) ZL,o (0)
− . (58)
M
2 vp,e vp,o

The impedance Z c and the propagation delay τ̂ for each line i are then calculated
L
using the relations [13]
v
u b d
u L + µi (T )L M
ZL (0) = t c
c
d
(59)
C − µ (T )C
1 i M

q
τb = b + µ (T )L
(L d )(C
c − µ (T )C
d) . (60)
i M 1 i M

17
2.6 Symmetric pair of coupled microstrip lines
A symmetric pair of coupled microstrip lines is composed of two parallel microstrip
lines of equal width w situated close together with a spacing s and of length l
mounted on a nonmagnetic dielectric substrate of thickness h and with a relative
permittivity ǫr . This structure is, in turn, mounted on a metallized common ground-
plane as shown in figure 6. The structure may have a cover plate mounted at a height
h2 above the surface of the substrate that affects its effective impedance if sufficiently
close.

6
h2  w - s - w - ?
? t
6 6 ǫr
h
?

Figure 6: Cross-section of a symmetric coupled microstrip structure.

The symmetric coupled microstrip line structure implemented in APLAC is

called Mclin, and accepts more than two strips to model a coupled structure [16].

2.6.1 Static parameters including effects due to shielding

The model first calculates the even- and odd-mode effective permittivities and
impedances of the structure. Even-mode parameters are obtained by applying a
voltage of equal magnitude and phase to both conductors. This produces an electric
field that has components only tangential to the axis of symmetry that divides the
structure such that one conductor lies on either side of the axis. Therefore, the axis
of symmetry can be replaced by a magnetic wall, whose electric field has its normal
component equal to zero and magnetic field its tangential component equal to zero
(Enorm = 0, Htan = 0). The circuit may thus be divided into two halves, each half
having a characteristic impedance ZL,e and effective permittivity ǫeff,e . Odd-mode
parameters are obtained by applying a voltage of equal magnitude but with op-
posite phase to the two conductors. This results in an electric field that has only
components normal to the axis of symmetry. So, the axis of symmetry may now
be replaced by an electric wall with the tangential component of its electric field
and normal component of its magnetic field equal to zero (Etan = 0, Hnorm = 0).
Again the circuit may be divided into two halves, with each half having a character-
istic impedance ZL,o and effective permittivity ǫeff,o . The wave propagating on the
coupled line pair is described by the superposition of the even and odd modes [5].
The expressions that follow describe the effective dielectric constants and the
power-current characteristic impedances of a pair of coupled microstrip lines as

18
shown in figure 6, and are discussed in detail in [9]. The physical dimensions of the
structure have again been normalized with respect to the substrate height as
w s t h2
u= g= th = h2h = . (61)
h h h h
The results are accurate to within 1-percent in the range

0.1 ≤ u ≤ 10, 0.1 ≤ g ≤ 10, 1 ≤ ǫr ≤ 18 , (62)

where ǫr the relative permittivity of the substrate material.

The even-mode static (frequency = 0) effective permittivity is calculated accord-
ing to Hammerstad and Jensen [4] and is given by
ǫr + 1 ǫr − 1
ǫeff,e (0) = + q, (63)
2 2
where q is the filling factor from equation (11),
 −ae (ν)·be (ǫr )
10
q∞ = 1+ (64)
ν
u(20 + g 2 )
ν= + g · exp(−g)
10 + g 2
   
ν 2 "
ν + 4   #
1   52

 1 ν 3
ae (ν) = 1 + ln  + ln 1 +
49  ν 4 + 0.432  18.7 18.1

 0.053
ǫr − 0.9
be (ǫr ) = 0.564
ǫr + 3
and
!
1.733
tanh 1.626 + 0.107h2h − √ for h2h ≤ 39
qc,e = h2h (65)
1 for h2h > 39 .

Kirschning and Jansen [9] remodeled Hammerstad and Jensen’s equations for
the odd-mode static effective permittivity, so improving accuracy, to get
 
ǫr + 1
ǫeff,o (0) = + ao (u, ǫr ) − ǫeff (0) q + ǫeff (0) (66)
2
 
q∞ = exp −co · g do (67)
 
ǫr + 1
ao (u, ǫr ) = 0.7287 ǫeff (0) − [1 − exp(−0.179u)]
2

19
 0.747ǫr
bo (ǫr ) =
0.15 + ǫr

co = bo (ǫr ) − (bo (ǫr ) − 0.207) · exp(−0.414u)

do = 0.593 + 0.694 · exp(−0.562u) .

The odd-mode shielding correction term is

 
9.575
tanh − 2.965 + 1.68h2h − 0.311h22h for h2h < 7
qc,o = 7 − h2h (68)
1 for h2h ≥ 7 .

The quantity ǫeff (0) is the effective permittivity of a single microstrip of width w
and zero conductor thickness and is calculated using equation (15). The subscripts
e and o refer to the even and odd modes, respectively, and the argument 0 implies
static parameters. In both, the even- and odd-mode cases, qt is obtained from
equation (13). The correction terms for finite cover height, qc,e and qc,o , are from
[11].
The next step is to calculate the even- and odd-mode characteristic impedances
of the coupled microstrip pair. This is done using the equations presented in [9].
For the even mode, the static characteristic impedance is
s
ǫeff (0)
ZL (0)
ǫeff,e (0)
ZL,e (0) = , (69)
ZL (0) q
1− ǫeff (0) Q4
η0

where ZL (0) and ǫeff (0) are the static characteristic impedance and effective permit-
tivity, respectively, of a single microstrip of width w and zero conductor thickness,
η0 is the intrinsic impedance of free space and Q4 is given by

Q1
2
Q2
Q4 =
exp(−g)uQ3 + (2 − exp(−g)) u−Q3
 
 !6 −0.387
8.4  1  g 10 
Q3 = 0.1975 + 16.6 + + ln 
 

10 
g 241  g 
1+
3.4

Q2 = 1 + 0.7519g + 0.189g 2.31

Q1 = 0.8695u0.194 .

20
Similarly the odd-mode characteristic impedance is given by
s
ǫeff (0)
ZL (0)
ǫeff,o (0)
ZL,o (0) = , (70)
ZL (0) q
1− ǫeff (0) Q10
η0
where
 
Q2 Q4 − Q5 · exp ln(u)Q6 u−Q9
Q10 =
Q2
 
1
Q9 = ln(Q7 ) Q8 +
16.5
"  5 #
g
Q8 = exp −6.5 − 0.95 ln(g) −
0.15

10 + 190g 2
Q7 =
1 + 82.3g 3
 

1  g 10  1  
 
Q6 = 0.2305 + ln   10  + ln 1 + 0.598g 1.154
281.3  g  5.1
1+
5.8
!
0.638
Q5 = 1.794 + 1.14 ln 1 + .
g + 0.517g 2.43

Effects due to shielding are, again, accounted for using the expressions in [11].
The correction due to shielding in the even- and odd-mode characteristic impedances
of a symmetric pair of coupled microstrip lines in a homogeneous air dielectric
(ǫr = 1) is calculated from equation (9). The correction term ∆Z0 (0) is calculated
from a different set of expressions for the even and odd modes. For the even mode,
the correction is

∆Z0,e (0) = fe (u, h2h ) · ge (g, h2h ) (71)

fe (u, h2h ) = 1 − artanh [A + (B + Cu)u]

−4.351
A=
(1 + h2h )1.842

6.639
B=
(1 + h2h )1.861

21
 −2.291
C=
(1 + h2h )1.90
  q 
F
ge (g, h2h ) = 270 1 − tanh D + E 1 + h2h −
1 + h2h
0.747
D= 
π

sin x
2
 
π
E = 0.725 sin y
2
log10 F = 0.11 − 0.0947g
log10 x = 0.103g − 0.159
log10 y = 0.0492g − 0.073 ,
and for the odd mode the correction is
∆Z0,o (0) = fo (u, h2h ) · go (g, h2h ) (72)
fo (u, h2h ) = uJ
!
(1 + h2h )1.585
J = tanh
6
  q 
L
go (g, h2h ) = 270 1 − tanh G + K 1 + h2h −
1 + h2h
G = 2.178 − 0.796g
log10 (20.492g 0.174 ) g > 0.858
K=
1.30 g ≤ 0.858
2.51g −0.462 g > 0.873
L= .
2.674 g ≤ 0.873
In order to get the corrections for the impedances of the coupled lines on a dielectric
substrate with relative permittivity ǫr , the corrections obtained for the air dielectric
need to be divided by the square root of the respective effective permittivity, i.e.
∆ZL,e (0)
∆ZL,e (0) = q (73)
ǫeff,e (0)
and
∆ZL,o (0)
∆ZL,o (0) = q . (74)
ǫeff,o (0)
As in section 2.2, the subscript L in the above equations implies a dielectric with
relative permittivity ǫr .

22
2.6.2 Correction due to finite strip thickness
For the coupled microstrip lines the correction required due to finite strip thickness
is according to Jansen [6]. The effective widths for the even and odd modes are
calculated from
  
1 −0.69∆u
ut,e = u + ∆u 1 − · exp
2 ∆t
  
1 −0.69∆u
ut,o = u + ∆u 1 − · exp + ∆t (75)
2 ∆t
th
∆t = ,
gǫr
where the subscript t refers to finite strip thickness and subscripts e and o refer to the
even and odd modes respectively. The above equations are valid for technologically
meaningful geometries and s > 2 t. ∆u in equation( 75) is the normalized width
correction for the case of a single microstrip.
The normalized width correction of Hammerstad and Jensen for the single strip
discussed earlier is not used here as it tends to exaggerate the effect of finite strip
thickness [9]. Instead, the width correction for the single strip ∆u is taken from [1]
and is given by
  
4πu − 2
1.25  2
th  
∆u = th  
1 + ln  + 

  (76)
π   th 1 
1 + exp −100 u −
2π
The normalized effective even- and odd-mode widths from equation (75) are used
to calculate the effective permittivities and impedances from equations (63), (69)
and (71), and (66), (70) and (72), respectively.

2.6.3 Dispersion
The dispersion equations by Kirschning and Jansen [9] are elaborate and their ac-
curacy in the range of applicability (equation (62)) is better than 2.5 percent up
to a normalized frequency fn = 20. The accuracy is better than 1.5 percent for
ǫr ≤ 12.9 and fn ≤ 15. Frequency fn (GHz·mm) is normalized with respect to
substrate thickness as
fh
fn = . (77)
106
Experimental evaluation ([18] [12]) shows that this dispersion model is consistent
over a wide range of frequency, strip width and dielectric constants.
The even and odd-mode effective permittivities are obtained from
ǫr − ǫeff,m (0)
ǫeff,m (f ) = ǫr − , (78)
1 + Fm (f )

23
where the subscript m refers to mode and may be either e or o for the even and odd
modes, respectively. As in section 2.2 argument f implies frequency dependence.
For the even mode
Fe (f ) = P1 P2 [(P3 P4 + 0.1844P7 ) fn ]1.5763 (79)
" #
0.525
P1 = 0.27488 + u 0.6315 + − 0.065683 · exp(−8.7513u)
(1 + 0.0157fn )20

P2 = 0.33622 [1 − exp(−0.03442ǫr )]
  !4.97 
 fn 
P3 = 0.0363 · exp(−4.6u) 1 − exp − 

38.7
( "  8 #)
ǫr
P4 = 1 + 2.751 1 − exp −
15.916
"  3 #
ǫr
P5 = 0.334 · exp −3.3 + 0.746
15
 !0.368 
fn
P6 = P5 · exp − 
18
 
P7 = 1 + 4.069P6 g 0.479 · exp −1.347g 0.595 − 0.17g 2.5 ,

and for the odd mode

Fo (f ) = P1 P2 [(P3 P4 + 0.1844) fn P15 ]1.5763 (80)
!
1.076
P8 = 0.7168 1 +
1 + 0.0576(ǫr − 1)
  !1.424  "
   0.946 #
fn ǫr
P9 = P8 − 0.7913 1 − exp −  · arctan 2.481

20 8

P10 = 0.242(ǫr − 1)0.55

"  1.629 #
u
P11 = 0.6366 [exp(−0.3401fn ) − 1] · arctan 1.263
3

1 − P9
P12 = P9 +
1 + 1.183u1.376
1.695P10
P13 =
0.414 + 1.605P10

24
   !3.215 
 fn 
P14 = 0.8928 + 0.1072 1 − exp −0.42 

20
  


0.8928 1.092 
P15 = 1 − (1 + P11 ) P12 · exp −P13 g  .
P14
√
A TEM wave propagates in the substrate with a velocity of c0 / ǫr , where c0 is
the velocity of propagation of light. The time required for the wave to propagate in
the pair of coupled lines in the even- and odd-modes is
√
l ǫeff,m
τm = . (81)
c0
The even-mode impedance dispersion is obtained from
 Qo
0.9408ǫeff (f )Ce − 0.9603
ZL,e (f ) = ZL,e (0) (82)
[(0.9408 − de ) ǫeff (0)Ce − 0.9603]Qo

  !2.745 
 fn 
Ce = 1 + 1.275 1 − exp −0.004625pe ǫ1.674
r

 18.365 

− Q12 + Q16 − Q17 + Q18 + Q20

re exp(−22.2u1.92 ) (ǫr − 1)6

de = 5.086qe · ·
0.3838 + 0.386qe 1 + 1.2992re 1 + 10(ǫr − 1)6
 
pe = 4.766 · exp −3.228u0.641

qe = 0.016 + (0.0514ǫr Q21 )4.524

!12
fn
re =
28.843
" #
0.3
Q11 = 0.893 1 −
1 + 0.7(ǫr − 1)
 !4.91 
fn
 
 20 
Q12 = 2.121 



!4.91  · exp(−2.87g)g

0.902

 fn 
1 + Q11
20
 5.1
ǫr
Q13 = 1 + 0.038
8

25
  
ǫr 4
Q14 = 1 + 1.203 15
 4
ǫr
1+
15

1.887 · exp(−1.5g 0.84 )g Q14

Q15 = !3
fn u2/Q13
1 + 0.41
15 0.125 + u1.626/Q13
" #
9
Q16 = 1+ Q15
1 + 0.403(ǫr − 1)2
( "   !1.87 
 0.672 #)  
u −4.25
fn 
Q17 = 0.394 1 − exp −1.47 
1 − exp 
7 20
"  1.593 #
u
1 − exp −2.13
8
Q18 = 0.61
1 + 6.544g 4.17

0.21g 4
Q19 =  !3 
fn
(1 + 0.18g 4.9 ) (1 + 0.1u2 ) 1 + 
24
" #
1
Q20 = 0.09 + Q19
1 + 0.1(ǫr − 1)2.7
 
 42.54g 0.133 · exp(−0.812g)u2.5 

Q21 = 1 −  .

1 + 0.033u2.5

Once again, ǫeff (f ) is the effective permittivity of a single microstrip line from equa-
tion (15). The quantity Qo also refers to the single microstrip and is the exponential
term R17 in equation (24). Similarly the odd-mode impedance is
!Q
ǫeff,o (f ) 22
ZL,o (0) − ZL (f )Q23
ǫeff,o (0)
ZL,o (f ) = ZL (f ) + (83)
1 + Q24 + (0.46g)2.2 Q25
!1.536
fn
0.925
Q26
Q22 = !1.536
fn
1 + 0.3
30

26
 0.005fn Q27
Q23 = 1 +  !1.9 
1 + 0.812
fn  (1 + 0.025u2 )
15
" #4.29
fn
(1 + 1.3u)
99.25
Q24 = 2.506Q28 u0.894
3.575 + u0.894
" #
0.3fn2 2.333 (ǫr − 1)2
Q25 = · 1 +
10 + fn2 5 + (ǫr − 1)2
   
ǫr − 1 12
 
Q26 = 30 − 22.2  13 
    − Q29
 ǫr − 1 12 
1+3
13
" #
0.84 2.5(ǫr − 1)1.5
Q27 = 0.4g 1+
5 + (ǫr − 1)1.5

0.149(ǫr − 1)3
Q28 =
94.5 + 0.038(ǫr − 1)3
15.16
Q29 = .
1 + 0.196(ǫr − 1)2

ZL (f ) is the frequency-dependent power-current characteristic impedance of a single

microstrip with width w calculated from equation (24).
This rigorous calculation gives the even- and odd-mode impedances from equa-
tions (82) and (83) respectively, the effective relative permittivity from equation (78)
and propagation delays from equation (81).

2.6.4 Losses
Losses due to dissipation for the coupled pair of microstrip lines are evaluated in the
same way as for a single microstrip line. The dielectric loss αd and the conductor loss
αc are calculated separately for the even and odd modes. The necessary equations
are rewritten here for convenience.
The dielectric loss in dB/m is calculated according to [2]
20π f ǫr ǫeff,m (0) − 1
αd,m = q tan δd , (84)
ln 10 c0 ǫeff,m (0) ǫr − 1

where the subscript m is either e or o referring either to the even or the odd mode
respectively, c0 is the velocity of propagation of light and tan δd is the loss tangent
of the dielectric material.

27
 Conductor loss in dB/m is from [4] and is given by
q
20π f ǫeff,m (0)
αc,m = . (85)
ln 10 c0 Qc
Qc , the strip inductive quality factor is
πZ0,m (0)hf u
Qc = , (86)
Rs c 0 K
where Z0,m (0) is the static impedance of the microstrip line pair in a homogeneous
medium for the even and odd modes and is obtained by multiplying the static
even- and odd-mode impedances by qthe square root of their respective effective
permittivities, i.e. Z0,m (0) = ZL,m (0) ǫeff,m (0). Rs , the surface resistance which is
an increasing function of surface roughness ∆, is
( "  2 #)
2 ∆
Rs (∆) = Rs (0) 1 + arctan 1.4 , (87)
π δ
where Rs (0) is the skin resistance for a smooth surface given by
1
Rs (0) = , (88)
σδ
and σ and δ are the conductivity and skin depth, respectively, of the conductor.
The skin depth δ is obtained from
1
δ=√ . (89)
πµ0 f σ
The current distribution

factor K for the two modes

is
!0.7
ZL,e (0) + ZL,o (0)
Ke = Ko = exp −1.2  . (90)
2η0
Again, the conductor loss, αc , calculations above are valid for a minimum con-
ductor thickness t ≈ 3δ.
The total loss is given by the attenuation coefficient in dB/m by
α = αc + αd . (91)

2.7 Asymmetric pair of coupled lines

Like the symmetric pair of coupled lines, the asymmetric pair of coupled microstrip
line structure is mounted on a metallized common ground-plane and is composed of
two parallel microstrip lines situated close together with a spacing s and of length
l mounted on a nonmagnetic dielectric substrate of thickness h and with a relative
permittivity ǫr . The widths of the two microstrip lines, however, are not equal
and are denoted by w1 and w2 . The asymmetric coupled line structure is shown in
figure 7.
The APLAC model of the asymmetric coupled line structure is called Maclin
and may contain more than two strips in its structure. The discussion that follows
is limited to the two strip case.

28
 6
h2  w1 - s - w2 -
?t
?
6 6ǫ
h r
?

Figure 7: Asymmetric coupled microstrip line structure.

2.7.1 Equivalent capacitances

The Maclin model is based on the model presented in [14] which calculates the
equivalent capacitances and inductances per unit length of the structure as shown
in figure 8.
Referring to figure 8, C10 is the capacitance formed by strip w1 and the ground-
plane, C20 is the capacitance formed by strip w2 and the ground-plane, C12 is the
capacitance formed by the two strips, L1 is the inductance of strip w1 , L2 is the
inductance of strip w2 and L12 is the mutual inductance between the two strips. In
order to calculate the strip-to-ground and strip-to-strip capacitances of the asym-
metrical pair of coupled lines, the strip-to-ground and strip-to-strip capacitances for
the symmetrical pair of microstrip coupled lines of widths w1 and w2 , spacing s and
relative permittivity ǫr need to be known. The inductances are found by calculating
the capacitances of the same asymmetrical coupled line structure but with an air
dielectric, i.e. ǫr = 1, as described in equation (43).

L2

L12 C12 C20


L1
C10

Figure 8: Circuit constants per unit length for a pair of asymmetric coupled mi-
crostrips.

The model described below is accurate to within one percent for the case of a
symmetrical pair of coupled microstrips in the range 1 < ǫr < 20 and

(10u)2 for u < 0.1

29
 1
0.001 + ≤g≤ 1 for 0.1 ≤ u ≤ 10 . (92)
(200u)2
(10/u)4 for u > 10

As before, the dimensions of the asymmetric microstrip structure are normalized

with respect to the substrate height as
w1 w2 s
u1 = u2 = g= . (93)
h h h
The first step in modelling the asymmetrical pair of coupled lines is to obtain
the even- and odd-mode characteristic impedances and effective permittivities for a
symmetrical pair of coupled lines for widths w1 and w2 from equations (69), (70),
(63) and (66) for a homogeneous air dielectric as well as for a dielectric substrate
with relative permittivity ǫr , thus giving four sets of impedances and permittivities.
The strip-to-ground and strip-to-strip capacitances for each of the four cases of
symmetrical lines are then evaluated from [5]
q
ǫeff,e (0)
C0,sym (u, g, ǫr ) = ǫ0 η0 (94)
ZL,e (0)
q q 
ǫ0 η0  ǫeff,o (0) ǫeff,e (0)
 .
C12,sym (u, g, ǫr ) = − (95)
2 ZL,o (0) ZL,e (0)

The mutual capacitance C12 for the asymmetrical pair of coupled lines is then
determined by averaging the values for two symmetric couplers with normalized
widths u1 and u2 as
!1/Km
2
C12 (u1 , g, u2 ) = , (96)
C12,sym (u1 , g, ǫr )−Km + C12,sym (u2 , g, ǫr )−Km
where Km has been obtained by fitting as

Km = 0.95 + 0.33 ln(u1 u2 ) − 0.4 ln g .

The strip-to-ground capacitances are given by

C10 = C0,sym (u1 , g) + [1 − Ks (ǫr )] [C12,sym (u1 , g, ǫr ) − C12 (u1 , g, u2 )] (97)

C20 = C0,sym (u2 , g) + [1 − Ks (ǫr )] [C12,sym (u2 , g, ǫr ) − C12 (u1 , g, u2 )] , (98)

where Ks (ǫr ) is obtained by modelling as

Ks (1)
Ks (ǫr ) = √
1 + 0.58(1 − 1/ ǫr )

Ks (1) = [0.21 − 0.023 ln(u1 u2 )] · exp {−g [1.56 + 0.22 ln (u1 u2 )]} .

30
Care must be taken not to confuse the capacitances for the symmetrical and asym-
metrical coupled lines. The capacitances in equations (96), (97) and (98) should
be evaluated for the dielectric substrate case as well as for the air dielectric case
when ǫr = 1. In the latter case, the term Ks (1) should be used instead of Ks (ǫr ) in
equations (97) and (98).
The quasi-static line constants per unit length for the coupled microstrips are
the elements of the capacitance and inductance matrices below:
" #
C11 −C12
C= (99)
−C12 C22

" #−1
C110 −C120
L = µ 0 ǫ0 . (100)
−C120 C220

2.7.2 Dispersion model

The dispersion model used here is according to Tripathi [15]. The frequency-
dependent effective parameters are easily obtained from the capacitances calculated
earlier.
The capacitances referred to in the following discussion are the capacitances
modelling the asymmetrical coupled line pair. Capacitances C11 , C12 and C22 are
the capacitances for the structure with a dielectric substrate, and capacitances C110 ,
C120 and C220 are the capacitances for the structure with an air substrate, ǫr = 1.
The following effective static relative permittivities and widths are first calcu-
lated:
C11 − C12
ǫeff,1 (0) = (101)
C110 − C120

C22 − C12
ǫeff,2 (0) = (102)
C220 − C120

C12
ǫeff,3 (0) = (103)
C120

h(C110 − C120 )
weff,1 (0) = (104)
ǫ0

h(C220 − C120 )
weff,2 (0) = (105)
ǫ0

sC120
weff,3 (0) = . (106)
ǫ0

31
 The next step is to calculate the inflection frequency fp , that corresponds to the
cut-off frequency of the first higher mode of the structure. This is done by finding
the lowest-order solution, or, in other words, the smallest zero of
g 1 1
0 = q +q +
ǫeff,3 (0) tan(β3 weff,3 (0)) ǫeff,1 (0) tan(β1 weff,1 (0))
1
q , (107)
ǫeff,2 (0) tan(β2 weff,2 (0))
where
q
βj = 2πf µ0 ǫ0 ǫeff,j (0) , j = 1, 2, 3.
The required zero of equation (107) is found using the Newton-Raphson algo-
rithm [17]. The algorithm is given by
f (x)
xk+1 = xk − ′ , (108)
f (x)
where the superscript k is the iteration number. The variable x in this case is the
frequency f and f (x) is the right hand side of equation (107). The derivative f ′ (x)
of the right hand side of equation (107) is
√ √ √
−g2π µ0 ǫ0 2π µ0 ǫ0 2π µ0 ǫ0
− − . (109)
sin2 (β3 weff,3 (0)) sin2 (β1 weff,1 (0)) sin2 (β2 weff,2 (0))
Using a suitable initial value for f , fp , the smallest zero, is found using the iteration
routine in equation (108).
The frequency-dependent effective relative permittivity and width are then found
from
ǫr − ǫeff,j (0)
ǫeff,j (f ) = ǫr −  2 j = 1, 2, 3 (110)
1 + G ffp
G = 0.6 + 0.009(2µ0 hfp )
wj − w eff,j (0)
weff,j (f ) = wj −   j = 1, 2, 3. (111)
f 2
1+ fp

Substrate height h is used for w3 when evaluating the above equations.

The last step is to calculate the capacitances resulting due to dispersion and are
calculated from
ǫeff,1 (f )weff,1 (f ) ǫeff,3 (f )weff,3 (f )
C11 (f ) = + (112)
h s
ǫeff,3 (f )weff,3 (f )
C12 (f ) = (113)
s
ǫeff,2 (f )weff,2 (f ) ǫeff,3 (f )weff,3 (f )
C22 (f ) = + . (114)
h s
The corresponding values for the air dielectric must also be calculated in order
to get the required inductance matrix in equation (100).

32
3 Specific functions implemented in APLAC
This section documents the various functions implemented in APLAC at the ‘C’
level in order to facilitate modifications in the code and, also lists the functions
available to users at the input file level [16].

3.1 C-level implementation

The impedance of a single microstrip in a homogeneous air dielectric, including
effects due to shielding, is calculated in function mlin Z01. If necessary mlin Z01
calls function delta u, which computes the required correction in width, in order to
account for finite strip thickness. Function mlin epse returns the effective relative
permittivity, taking into account finite strip thickness. The characteristic impedance
of a microstrip on a dielectric substrate is then calculated in mlin Z0 by calling the
above mentioned functions.
The two methods used to calculate effects due to dispersion have been imple-
mented in mlin dispersion (Hammerstad) and mlin level2 (Kirschning), and
may be chosen by the user by specifying LEVEL 1 or LEVEL 2, respectively. However,
computation of the LEVEL 2 effective relative permittivity using the equations given
by Kirschning and Jansen [8] is carried out in a separate function mlin dispersion2,
and is called by mlin level2 when calculating the corresponding impedance. Equa-
tions for the dielectric and conductor losses are in mlin losses.
Zl even odd calculates the static even- and odd-mode impedances and effec-
tive permittivities of a pair of coupled microstrip lines, and includes effects due to
shielding. Effects due to finite thickness are evaluated in mclin thick. However, the
effective width correction due to finite strip thickness for the single microstrip that
is required in mclin thick is now calculated using the equations in [1] in function
mlin thick. Dispersion is accounted for in mclin dispersion.
CalcCm calculates the static strip-to-ground and strip-to-strip capacitances of
the asymmetrical pair of coupled microstrips. The static model is implemented in
AclinValues. The frequency dependence is calculated in Aclin Disp. eff parf
calculates the effective widths and permittivities from which eff cap calculates the
frequency-dependent strip-to-ground and strip-to-strip capacitances per unit length
of the asymmetrical pair of coupled microstrip lines. The eigenvector matrix is
calculated in mlayer eigenv.

3.2 Input-file level functions

Several functions are available to the user of APLAC from which various microstrip
parameters may be calculated [16]. The ones concerning the microstrip components
under discussion are Mclin Z, Mclin Zf, Mlin epse, Mlin u, Mlin w and Mlin Z01.
Mclin Zf calculates the characteristic impedance of a symmetrical coupled pair
of microstrip lines for given physical specifications. It differs from Mclin Z in that

33
it calculates the characteristic impedance at a given frequency and so accounts for
dispersion effects. Its syntax is
Mclin_Zf(rW,rT,rS,rEr,rH,rH2,rF,vrZ)
where rW, rT, rS, rEr, rH and rH2 are the physical dimensions of the structure (see
figure 6), and rF is the frequency at which the impedance is to be calculated. The
function returns vector vrZ whose four values are

vrZ[0], the even-mode impedance,

vrZ[1], the odd-mode impedance,
vrZ[2], the even-mode effective permittivity and
vrZ[3], the odd-mode effective permittivity.

Impedance and effective εr

APLAC 7.20 User: HUT Circuit Theory Lab. Thu Dec 4 1997
70 10
Z/Ω
εr
60
8
50

40 6

30 4
20
2
10

0 0
1 3 10
f/GHz
Ze Zo
εr,e εr,o

Figure 9: Even- and odd-mode impedances and effective permittivities of a sym-

metric coupler on an FR4 board.

As an example,
Declare VECTOR z REAL 3

Sweep "Impedance and effective \\epsilon_r"

+ LOOP 200 FREQ LOG 1000MEGHz 25GHz
+ Y "Z" "\\Omega" 0 75
+ Y2 "\\epsilon_r" "" 0 10
+ MULTX="G"

34
 Call mclin_zf(3mm, 20um, 1mm, 4.5, 1.5mm, 20mm, f, z)

Display Y "Z_e" z[0]

+ Y "Z_o" z[1] MARKER=1
+ Y2 "\\epsilon_{r,e}" z[2] MARKER=2
+ Y2 "\\epsilon_{r,o}" z[3] MARKER=3

EndSweep

plots the even- and odd-mode impedances and effective dielectric-constants as a

function of frequency of a symmetric coupler on an FR4 board. The results of the
example above are shown in figure 9.

35
4 Conclusion
All the equations used to model the above mentioned microstrip components in
APLAC have been thoroughly documented in this report. In addition, a little back-
ground was provided to facilitate the comprehension of the working of the models.
However, a detailed background is not given.
The transmission line model for single and coupled microstrips has been pre-
sented. It was seen that for the single microstrip line, its characteristic impedance,
attenuation constant, and propagation delay need to be computed. The closed form
equations required for this purpose were presented. For the coupled microstrip
lines, the eigenvector matrix has to be determined in addition to the characteris-
tic impedance, attenuation constant, and propagation delay. Detailed closed form
numerical expressions were presented for the symmetrical and asymmetrical pair of
coupled lines. It was seen that the n symmetrical coupled microstrip lines could be
modeled by considering one pair of coupled lines at a time.
Although no comparison results have been presented in this report, various
APLAC simulations were compared with those using the MDS (Hewlett-Packard)
circuit simulator and results were found to be comparable. Good agreement was
also found with measurement results presented in the literature.

36
Acknowledgements
The author would like to thank professor Martti Valtonen for providing the oppor-
tunity to work on this project and for creating a relaxed and amicable atmosphere
to work in at the Circuit Theory Laboratory. The author would also like to thank
the entire staff at the laboratory for their support, and especially Hannu Jokinen
and Sakari Aaltonen, who also proof-read the report, for their patience with his
numerous questions.

37
References
[1] H. A. Atwater, “Simplified Design Equations for Microstrip Line Parameters,”
Microwave Journal, pp.109-115, November 1989.

[2] E. J. Denlinger, “Losses of Microstrip Lines,” IEEE Transactions on Microwave

Theory and Techniques, Vol. MTT-28, No. 6, pp. 513-522, June 1980.

[3] E. Hammerstad, “Computer-Aided Design of Microstrip Couplers with Accu-

rate Discontinuity Models,” IEEE MTT-S Symposium Digest, pp. 54-56, June
1980.

[4] E. Hammerstad and O. Jensen, “Accurate Models for Microstrip Computer-

Aided Design,” IEEE MTT-S Symposium Digest, pp. 407-409, June 1980.

[5] R. Hoffmann, Handbook of Microwave Integrated Circuits, Artech House, Nor-

wood, MA 02062, 1987.

[6] R. H. Jansen, “High-Speed Computation of Single and Coupled Microstrip

Parameters Including Dispersion, High-Order Modes, Loss and Finite Strip
Thickness,” IEEE Transactions on Microwave Theory and Techniques, Vol.
MTT-26, No. 2, pp. 75-82, February 1978.

[7] R. H. Jansen and M. Kirschning, “Arguments and an Accurate Model for the
Power-Current Formulation of Microstrip Characteristic Impedance,” AEÜ,
Vol. 37, No. 3-4, pp. 108-112, March-April 1983.

[8] M. Kirschning and R. H. Jansen, “Accurate Model for Effective Dielectric Con-
stant of Microstrip with Validity up to Millimetre-Wave Frequencies,” Electron-
ics Letters, Vol. 18, No. 6, pp. 272-273, March 1982.

[9] M. Kirschning and R. H. Jansen, “Accurate Wide-Range Design Equations for

the Frequency-Dependent Characteristic of Coupled Microstrip Lines,” IEEE
Transactions on Microwave Theory and Techniques, Vol. MTT-32, No. 1, pp.
83-90, January 1984.

[10] M. Kirschning, R. H. Jansen and N. H. L. Koster, “Accurate Model for Open

End Effect of Microstrip Lines,” Electronics Letters, Vol. 17, No. 3, pp. 123-124,
1981.

[11] S. March, “Microstrip Packaging: Watch the Last Step,” Microwaves, Vol. 20,
No. 13, pp. 83-94, December 1981.

[12] J. L. Medina, A. Serrano and F. J. Mendieta, “Microstrip Effective Dielectric

Constant Measurement and Test of CAD Models up to 20 GHz,” Microwave
Journal, pp. 82-93, March 1993.

38
[13] F. Romeo and M. Santomauro, “Time-Domain Simulation of n Coupled Trans-
mission Lines,” IEEE Transactions on Microwave Theory and Techniques, Vol.
MTT-35, No. 2, pp. 131-137, February 1987.

[14] F. Sellberg, “Simple and Accurate Algorithms to Include Lange and Asymmet-
ric Microstrip Couplers in CAD-Packages,” Proceedings of the 18th European
Microwave Conference, pp. 1157-1162, Stockholm, Sweden, 12-16 September
1988.

[15] V. K. Tripathi, “A Dispersion Model for Coupled Microstrips,” IEEE Trans-

actions on Microwave Theory and Techniques, Vol. MTT-34, No. 1, pp. 66-71,
January 1986.

[16] M. Valtonen et al., APLAC – An Object-Oriented Analog Circuit Simulator

and Design Tool, Helsinki University of Technology, Circuit Theory Laboratory
and Nokia Corporation, Research Center, 6.2 User’s Manual and 7.1 Reference
Manual Vols. I, II & III, 1997.

[17] J. Vlach and K. Singhal, Computer Methods for Circuit Analysis and Design,
Van Nostrand Reinhold Company, New York, 1983.

[18] R. A. York and R. C. Compton, “Experimental Evaluation of Existing CAD

Models for Microstrip Dispersion,” IEEE Transactions on Microwave Theory
and Techniques, Vol. 38, No. 3, pp 327-328, March 1990.

39