3

MODELLING OF LOSSY MICROSTRIP LINES WITH FINITE THICKNESS

J. F. Kiang, S. M. Ali, and J. A. Kong

3.1 Introduction

3.2 Integral Equation Formulation

3.3 Numerical Solution of Charge Distribution 3.4 Magnetostatic Dual Problem

3.5 Calculation of G and R

3.6 Transmission Line Analysis 3.7 Results and Discussions Conclusions

References

3.1 Introduction

For microwave integrated circuit applications, the characteristics of interconnects have been investigated for propagation modes [1,2], time response [3], crosstalk [4], coupling [5], delay [6], etc. In these analyses, it is assumed that quasi- TEM modes are guided along the multiconductor transmission line. In [1], the analysis was performed for two asynunetric transmission lines. In [2] and [3], an arbitrary number of transmission lines were analyzed. In [3], the load and the source conditions were presented in terms of the modal reflection and transmission coefficient matrices.

To perform the quasi-TEM analysis, the capacitance matrix for the multiconductor transmission line has to be obtained first. Both the spectral and the spatial domain methods have been proposed to

- 85-

86

3. Modelling of Lossy Microstrip Lines

,L.

(0)
%--4
(1)
z- -d.
% - -4'_2
(1-1)
% - -4'_1
(I)
z--rl.
(1+1) f fi'##'#/~/~~ Itrip3

h, mip2
1 ~I
•
•
•
h,_, "-1
+ ~
f ~~!i IUipft
h, fJ
, fw//,?##J'/,$#..¢'l1
t I ~'+I
h'+1
t oI- ~.~-------. _ -4'+1

•

•

"f-==:--;;;;;;;;;;;;;:--~===--' - -4"'_1

la' - *4 JZnl Wd mamA (N)

'" ." . .• mi M

J""";,;;,;;""",,;,~;;,,,

FIgure 3.1 GeometrIcal configuratIon of M microstrip lines with finite thickness embedded In arbItrary layers of an IsotropIc stratified medium.

calculate the capacitance matrix. In the spectral domain methods, two side walls are used to enclose the whole transmission line structure, and the thickness of the strip lines has not been considered [7,8]. In using the spatial domain method [9], the structure has to be truncated to a finite extent to make the numerical implementation feasible. In [10], the infinite extent of the structure was incorporated, but only a two-layer medium was considered.

In practical microwave integrated circuits, the dielectric loss due to the substrate and the conductor loss due to the metallic strips have been studied in the analysis of circuit performances[1l-13].

In this chapter, we present a quasi-TEM analysis of coupled lossy microstrip lines of finite strip thickness embedded in different layers of a lossy isotropic stratified medium as shown in Fig. 3.1. First, a spectral domain scalar Green's function in a lossy isotropic stratified medium is derived. In the formulation, no side walls are introduced, the transmission structure is not truncated, and the analysis is valid for arbitrary number of dielectric layers.

3.2 Integral Equation Formulation

87

Based on the scalar Green's function, a set of coupled integral equations is obtained for the charge distribution on the strip surfaces. The method of moments is then applied where pulse basis functions and a point-matching scheme is used to solve numerically the set of integral equations for the charge distribution, and hence the capacitance matrix. The duality between the electrostatic formulation and the magnetostatic one is applied to calculate the inductance matrix. The conductance matrix is obtained by using the duality between the electrostatic problem and the current field problem. A perturbation method is used to calculate the resistance matrix.

Finally, a transmission line analysis is derived to obtain the transfer matrix for multiconductor line, which significantly reduces the effort in treating the load and the source conditions. Transient responses are obtained by using the Fourier transform. The results for two coupled lines are presented.

3.2 Integral Equation Formulation

We first formulate the scalar Green's function in the spectral domain in a homogeneous medium of permittivity e with a ground plane located at z = 0 . Thus, we consider a uniform line charge of unit amplitude to be located at (:c', z') along the y direction, and evaluate the electrostatic potential at (;z:, z) •

The scalar Green's function g(r, r') is the solution of the following Poisson equation :

1

V~g(r, r') = - -5(r - r')

€

(1)

where V~ = (j2 /8;z:2 + 82/ 8z2, r = z:c + h, and r' = z;z:' + h' . The scalar Green's function can be expressed in the spectral domain as

(2)

where k. = zkz + zkz ,and U(k.) is the Fourier transform of g(r, r') with respect to :c and z , Substituting (2) into (1), and using the image theory, we get

(3)

88

3. Modelling of Loss)" Microstrip Lines

Z'-l

Z, t .... <_'_-_1_)_ Z, = 0 = Z'-l

(1)

Figure 3.2 Definition of the local refiection and transmission coefficients at a dielectric interface between medIa (I) and (I - 1).

The second term is the Fourier transform of the Green's function due to the image line charge at r[ = zz' - h' which is the image of r' with respect to the ground plane at z = O. The scalar Green's function can then be written as

(4)

By contour integration in the complex kz plane, we have

(5)

The integrand is regular at kz = 0 , and the integration is well defined as r goes to infinity.

Next, we introduce local reflection and transmission coefficients at a boundary between two dielectrics. Consider a dielectric interface between media (I) and (l- 1) as shown in Fig. 3.2. We assume that the spectral domain potential in each region can be expressed as

(6a)

S.2 Integral Equation Formulation

89

(1- 1)

A RU(I-i)

z = -d'-2

(I)

ARuI

z

z = -d,

Figure 3.3 Definition of the global reflection coefficients at the upper boundaries of dielectric layers (I) and (l - I).

(6b)

where ZI = ZI-1 = Z + dl-1, RI(I-l) and TI(I-l) are constants to be determined. The Z component of the electric field in each region can be obtained as

(7a) (7b)

By imposing the boundary conditions that the potential and the normal component of the electric flux density are continuous at z = -dl-1 , we obtain

R _ fl - fl-l

1(1-1) -

ei + fl-l

2£1 T,(I_l) = --fl + fl-1

(Sa)

(8b)

where R'(l-l) and T,(I-l) can be interpreted as local reflection and transmission coefficients, respectively.

Next, consider the two source-free layers (I) and (I - 1) of the stratified medium as shown in Fig. 3.3, where the potential and the z

90

3. Modelling of Lossy Mierostrip Lines

component of the electric field in each layer can be expressed as

(9a) (9b)

where Zj = Z + dj and j = I, I - 1. We introduce the global reflection coefficient RUj defined at the upper boundary of layer (j),

z = -dj-l, as

(10)

where hj is the thickness of layer (j). By imposing the boundary conditions at Z = -d,-l , we obtain the following recursive relation

1= 2,···,N

(11)

and RUl = RlO' Similarly, the global reflection coefficient Rnj is defined at the lower boundary of layer (j), Z = -dj, as

A·

R . - J

nJ - B.' J

(12)

By imposing the boundary condition at Z = -d, , we obtain

1= N -1", ·,0

(13)

and RnN = -1.

With the use of the above reflection coefficients, the expression of the scalar Green's function in a layered medium can be obtained in a simple way. For the case where both the line charge and the observation point are located in the upper half-space, that is layer (0), we have

3.2 Integral Equation Formulation

91

where Zo = z + do, z~ = z' + do, the uniform line charge of unit amplitude is located at 1" = (Z/, z&), and the observation point is located at l' = (:I:, zo) . The first term inside the brackets is the direct term, and the second term can be interpreted as due to the reflection from the lower boundary at z = - do .

For the case when both the line charge and the observation point are both located in an arbitrary layer (I) with 1 1:- 0, the Green's function can be expressed as

where ZI = z + dl, zt = z' + dl, A( kz) and B( kz) are the unknowns to be determined. From the definition of RUI and Rnl, we have

RUI[e-11e.1(.z:.-.z:D + A(kz)e-11e.1.z:.} = B(kz)e11e•I.z:., Zl = hi (16a)

Rnl[e-11e.1(.z::-.z:.) + B(kz)e11e.1.z:.] = A(kz)e-11e.1.z:., ZI = 0 (16b)

By solving (16), we get

Substituting (17) into (15), we obtain the explicit form of the Green's function gil (1', 1") in the layer (I) as

911(1',1") =_1_ /00 dkz eile.(tIt-z') {e-IIe.II.z:.-.z::1 + 1_---=-o~~

41r€1 J-oo Ikzl 1 - RU1Rnle-211e.lh.

[RUle-IIe.I(2h.-zl-.z:D + Rnle-11e.1(.z:·+.z:D

+ RUlRnle-I·.I(~-·;+2h,) + Ru,Rnlo-I'.I(- .. +.:+ .. ,)] } (18)

92

3. Modelling of Lossy Microstrip Lines

(0)
(1)
·
·
·
r /) , (I)
z=-d,
(I + 1)
1''' •
·
·
<> z = -d .. _1
(m)

(N)
1///i/17iJYIIi/H/lIiYHP/l/ln»/)"H)q (0)
(1)

r" /)~ (m)
z= -don
(1-1)
z = -d,_,
(I) (N)

>nllmm,;mJn;m;;,nm»»»;

Figure 3.4 Determination of the scalar Green's function where the source is in a layer (m) and the observation point is in the layer (1). (a) Layer (1) Is above layer (m). (b) Layer (1) Is below layer (m).

The first term inside the brace is the direct term, the other terms are the summation of the multiple reflections between the two boundaries at z = -d'-1 and z = -d, . It is observed that (14) is a special case of (18) with 1=0 and Rul = O.

Next, we consider the case when the source and the observation points are in different layers. As shown in Fig. 3.4(a), the source is in a layer (m) and the observation point is in a layer (1) which is above layer (m). We assume that each spectral field component is transmitted upward from the source layer (m) to the layer (1) with the upward transmission coefficient XUI,m' The Green's function can thus be expressed as

elle.I%~ + Rnme-IIe.I(2h ... +%~) 1 - RumRnme-211e.lh ...

(19)

S.2 Integral Equation Formulation

where Zl and z:n are the local coordinates defined as Zl = z + dl and z:n = Zl + dm-1 , respectively. Similarly, when the observation point is located in the layer (I + 1) , we have

1 /00 eik.(z-z')

g(l+l)m(1'>, r~) = -4 - dkz Ik I XU(!+l).m

1r€m -00 z

[e-1k•I.I'l+1 + RU(I+1)e-lk.I(2hl+1-.I'1+1)]

elk.lz~ + Rnme-I1e.1(2hm+z~) 1 - RumRnme-2Ik.lh ...

where ZI+1 = z + dl+1 • Imposing the boundary conditions at the interface z = - dl, we get

(20)

1 = m - 2,·· ,,0

(21)

and for 1 = m - 1, we have

X = 1 + Rum

U(m-l),m 1 + R -21k Ih 1

U(m-l)e • m-

(22)

For the case when the layer (m) is above layer (I) as shown in Fig. 3.4(b), the Green's function may be expressed as

e-11e.lz~ + RUme-IIe.I(2hm-z~)

1 - RumRnme-211c.lh ...

where Xnl,m is the downward transmission coefficient, ZI and z:n are the local coordinates defined as Zl = Z + dl-1 and z:n = Zl + dm, respectively. Similarly, when the observation point is located in the layer (I - 1), we have

(23)

1 100 ei". (o:-z')

g(l-l)m(r<,~) = 4-- dkz Ik I Xn(l-l),m

1r€m -00 z

[elle.lzl-l + Rn(l_l)e-IIc.I(2hl-l+ZI-d]

e-IIe.lz~ + Rume-IIc.I(2h ... - .. ~)

1 - RumRnme-211c.lh",

(24)

94

8. Modelling of Lossy Microstrip Lines

where ZI-1 = Z + d,-2, Imposing the boundary conditions at the interface z = -d,-1 , we have

-11e.lh'-l 1 + Rn(l-l)

Xnl,m = Xn(l-l),me 1 + Rnle-2Ilc.lh,'

1 = m + 2" ··,N

(25)

and for I = m + 1 , we have

x _ l+Rnm

n(m+l),m - 1 + R -21lc Ih +1

n(m+l)e • m

(26)

Next, consider M microstrip lines embedded in arbitrary layers of an isotropic stratified medium as shown in Fig. 3.1. All the conductor strips are of finite thickness, the permittivity of layer (j) is Ej, and a ground plane is put at z = -dN as the potential reference.

Using the scalar Green's function, the potential in layer (1) can be represented as

(27)

where r" is the cross section contour of the p th microstrip line, p,,(r') is the charge density on the p th microstrip surface, L(p) is the layer where the p th microstrip line is embedded. The cross section of the microstrip lines can be arbitrary in general. For practical applications, only the rectangular cross sections are considered.

The potential distribution in (27) satisfies all the boundary conditions that the potential and the normal electric flux density are continuous across the interfaces between adjacent dielectric layers. To obtain the charge distribution on each micros trip surface, we impose the boundary condition that the potential on each micros trip is equal to the impressed voltage. Thus we have

(28)

where q = 1"" ,M.

In the next section, the method of moments is applied to solve (28) for the charge distributions.

3.3 Numerical Solution of Charge Distribution

95

3.3 Numerical Solution of Charge Distribution

By applying the method of moments to solve (28), we first choose a set of pulse basis functions for the surface charge density, hence

Np

pp(r') = L apiPpi(r'),

i=l

p = 1,. ",M

(29)

where api and Ppi(r') are, respectively, the expansion coefficients and the basis functions on the i th section of the p th microstrip surface; and Np is the total number of basis functions on the p th micros trip surface. The pulse function is defined as

{ 1,

Ppi(r) =

0,

(30)

elsewhere

where r pi is the i th section on the p th microstrip surface.

Substituting (29) into (28), we have

ron rq

(31)

where q = 1"" ,M.

Next, we choose the center point at each pulse basis as the testing point to test (31), thus we have

M s,

L L Zqj,piapi = {3qj,

p=li=l

1 s q s M, 1 s i s Nq

(32)

where

(33)

(34)

and rqj = (Zqj,Zqj) is the center coordinate of rqj. By solving (32) for api '5, the charge density can be obtained.

The capacitance matrix C of M micros trip lines can be obtained by solving (28) M times. For the nth time, all the microstrip lines are

96

3. Modelling of Lossy Microstrip Lines

grounded except that one volt is imposed on the nth line. By solving (32), the charge distribution is obtained. The total surface charge per unit length on the m th microstrip line is equal to the mn th element of the capacitance matrix, Cmn • This can be observed from the definition of the capacitance matrix

M

LCqpVp = a; p=l

(35)

By setting Vp = 6p,n, we have Q~n) = Cqn with 1 s q s M. Here, 6p,n is the Kronecker's delta function and the superscript (n) in Q~n) is the index of the micros trip where one volt is imposed.

3.4 Magnetostatic Dual Problem

In this section, we briefly review the magnetostatic dual problem.

For nonmagnetic materials, the magnetostatic problem for the geometrical configuration shown in Fig. 3.1 is equivalent to the one where all the dielectric media are replaced by the free space. The magnetostatic potential t/;(1') due to a uniform line current at 1" can be derived in a way similar to that in Section 2 as

(36)

where q = 1"", M; gtrq)L(p)(1',1") is the scalar Green's function; Jp(r') is the surface current on the p th microstrip surface; and t/;q is the magnetostatic potential measured on the q th microstrip surface with respect to the ground plane, which is equal to the magnetic flux linkage between the q th microstrip line and the ground plane per unit length.

Instead of solving (36) directly, a dual electrostatic problem for the same geometrical configuration in free space is solved. It can be shown that

M(- -I) E(- -')

g r, r = J.Lof.og r, r

(37)

where gE(1',1") is the scalar Green's function derived in Section 2 with all f.l 's replaced by the dielectric constant in the free space.

a.5 Calculation of G and R

97

Imposing the same boundary conditions that Vq = 1/Iq for 1 ~ q ~ M, we have

(k)(_) J(k)(1') = Pp r ,

p J.loEo

1 ~p,k s M

(38)

where p~k)(1') is the surface charge distribution on the pth microstrip surface when we impose one volt on the k th strip, and zero volt on the other strips; J~k)(1') is the surface current distribution on the p th microstrip surface when we impose one tesla-m on the k th strip, and zero tesla-m on the other strips.

In general, the surface current on the p th micros trip surface can be expressed as

M 1 M

Jp(1') = L 1/IkJ~k)(1') = - L 1/JkP~k)(1') (39)

k=1 P.oEo k=1

Integrating (39) over one unit length on the pth micros trip surface, we have

1 M

Ip = -- L Co,pk1/lk

P.OEO k=1

where Ip is the surface current on the p th micros trip line, and Co,p" is the pk th element of the capacitance matrix Co with all the dielectrics replaced by free space. _!!ence, we obtain the conventional result that the inductance matrix L is proportional to the inverse of the capacitance matrix in free space as

(40)

- =-1

r = P.oEOCO

(41)

3.5 Calculation of G and R

In this section, the dielectric loss and conductor 108s are considered.

The transmission line equations for M coupled microstrip lines are

(42a)

(42b)

98

3. Modelling of Lossy Microstrip Lines

where the conductance matrix G accounts for the dielectric loss, and the resistance matrix R accounts for the conductor loss.

The nth eigen solution to (42) can be represented as

1 = Ine-"fnJl V = Vne-"fnll

(43a) (436)

where In is the eigenvalue of the nth mode; and In and V n are the corresponding eigenvectors. Substituting (43) into (42), we have

InIn = [G - iwC] . Vn

I'nVn= [R-iwL] .t.

(44a) (44b)

and hence

I!In = [G - iwC] . [R - iwL] . In I'!V n = [R - iwL] . [G - iwC] . V n

(45a) (45b)

There are in general M eigen solutions to the above equations. The eigenvalues solved from (45a) and (45b) are the same; and the eigenvectors In's and V n 's are related by (44).

For the nth eigenmode, the time-average Poynting's power PnT can be represented as

1 [-t -]

PnT = 2Re In' V n

(46)

where 1~ is the transposed, complex conjugate of In. The timeaverage power loss per unit length PnL along the transmission lines can be calculated as

(47)

where

T1 = -

Pnc = (1/2)In . R . In = 2ancPnT

Trt = -

PnD = (1/2)V n • G . V n = 2anDPnT

(48a) (48b)

1.5 CalculatIon of G and R

99

where Pnc and PnD are the time-average conductor loss and dielectric loss per unit length, respectively; anT, anc, and «-» are the corresponding attenuation constants for the nth mode.

To calculate the conductance matrix which accounts for the dielectric loss, we make use of the duality between the electrostatic problem and the current field problem. Assume that the geometrical configuration of the current field problem is the same as in Fig. 3.1, and the conductivity in layer (1) is designated by (Tl with 0 $ 1 $ N . If the conductivities in the current field problem and the dielectric constants in the electrostatic problem satisfy the following relation

(To (Tl (TN

-=-= .. ·=-=a

EO El EN

(49)

then, we have

G=aG

(50)

The perturbation method is used to _!olve for the attenuation constants anC and the resistance matrix R. We start from the Iossless eigenvalue equations :

= = - 1-

e·L·lon= 210n 11n

= = - 1- L·G·YOn= 2YOn 11n

(51)

where 1n = -if3n; 11n = W / f3n is the phase velocity for the nth mode; and G is the capacitance matrix with all materials lossfree. The timeaverage guided power can be approximated by the power guided along lossless lines as

(52)

The power loss per unit length of the n th mode due to imperfect conductor can be calculated by using the surface current obtained for the perfect conductor lines. Thus, we have

(53)

where R1cS is the surface resistance per unit area on the Ie th microstrip surface, JnJ.(r) is the surface current distribution of the nth

100

3. Modelling of Lossy Microstrip Lines

mode on the k th microstrip surface. Here M' is the total number of conductors including the ground planes. As defined in (29), p~i)(f) can be expanded in terms of the pulse basis functions as

N.

p~i)(f) = L a~i~Plcm(f)

m=1

(54)

where Plem(f) is the basis function defined in (30). By using (39), the surface current on the k th microstrip surface can be represented as

1 M N. .

JIc(f) = - L,pi L a~%Plcm(f)

POEO i=1 m=l

(55)

where ,pi is obtained by solving

M

,pi = L LijIOn,i i=1

(56)

where Lij is the ij th element of the inductance matrix L, Ion,i is the j th element of the eigenvector IOn. The conductor loss can then be calculated as

1 M N. M . 2 r

PnC = '2 L RIcS(POEo)-2 L L ,pia~% Jf IPlcm(r)12 dr + PG

Ie=l m=1 ,=1 r.m

(57)

where PG is the conductor loss due to the ground planes. To calculate PG, we first solve for the magnetic field Hz as

(58)

where gE(kz, z, z') is the Fourier transform with respect to z of gE(r, r') as defined in (37). Thus, we have

3.6 Transmission Line Analysis

101

where Rs is the surface resistance per unit area of the ground planes, ZG is the z coordinate of the ground planes.

Now, set G = 0 in (45b), we have

Making the approximation that f3n ~ f3on, V n ~ V On, and utilizing the relation (44) with G = 0, the imaginary part of (60) becomes

M

2anc VOn,m = L RmleIOn,'u 1e=1

1 s n,m s M

(61)

where IOn,1e is the k th element of IOn; and VOn,m is the m th element of VOn. The elements of R can be o~ained by solving (61).

After obtaining the matrices C, L, G, and R: the matrix equations (44) and (45) can be solved for the eigenvalues and eigenvectors.

3.6 Transmission Line Analysis

The voltage and current along the transmission lines can be represented in terms of the eigenvectors obtained in the last section as

M M

V(y) = LanVne-"Y,,1I + LbnVne"Y"1I

n=1

n=1

= Bv' A(y)· A + Bv' A( -y). B

M M

I(y) = L an1ne-"Y"1I - L bnIne"Yftli

(62a)

n=1

n=1

(62b)

102

s. ModeWng of Lossy Mlcrostrlp Lines

~here 5v is the matrix with the eigenvector V n as its nth column; 51 is the matrix with the eigenvector In as its nth column; A and B are the mode coefficient vectors in the forward and backward directions, respectively; A(y) is a diagonal matrix with exp( -'"fnY) as its nth element; namely,

where the superscript T represents the transpose of a row vector.

The characteristic impedance matrix can be defined as

_ _ =-1

Zc = Sv· SI

(64)

From (62), the mode coefficients A and B can be represented in terms of the line voltages V(O) and line currents 1(0) as

A = ~ [8;1 . V(O) + 8;1 .1(0)] B = ~ [8;1 . V(O) - 8;1 .1(0)]

(65a)

(65b)

Next, the line voltages V(/) and the line currents 1(1) can be expressed in terms of V(O) and 1(0) through (65) as

[:(1)] = [AVV(I) AVI(I)]. [:(0)] = A(I). [:(0)] (66)

1(1) AIV(/) AII(/) 1(0) 1(0)

where A( I) is called the transfer matrix for the uniform transmission

I -

lines of length ,; and the explicit form of the submatrices of A( I) are

Avv(l) = ~Sv . [A(/) + A( _I)] .8;1 AV1(1) = ~8v . [A(/) - A( -I)] .8;1 AIV(I) = ~81. [A(l) - A( _I)] .8;1 AII(I) = ~81. [A(I) + A( -I)] . 8;1

(67a) (67b) (67c) (67d)

S.8 Transmission Line Analysis

lOS

If more than one transmission line section with different characteristics are cascaded, the transfer matrix of each section is multiplied to obtain the overall transfer matrix.

Imposing a voltage source V s in series with an impedance matrix Z s at y = 0, andJmposing a voltage source V L in series with an impedance matrix Z L at Y = I , we have

yeO) = V s - zs' 1(0) V(I) = V L + ZL .1(1)

(68a) (68b)

The line voltages at all ports can thus be calculated as

[z~, AVI{l)· z~: - AVV(I)]. [:«1))]

L AIl(I) . Z s - AIV(I) 0

[ Av l( I) . Z / . V s ]

(69)

AII(I).Z;l.VS+Z~l.VL

The transient response can be obtained by using Fourier transform.

For the case of a single microstrip line, we first solve for the capacitance and the inductance per unit length C and L, respectively. Next, solving the eigenvalue equations in the lossless medium, we have

[2 = _f32 = _w2 LC 10 = 1

Vo =..fL!C

(70a) (70b) (70c)

The magnetic flux linkage is 1/J = LID = L, the time-average power is PT = (1/2)..fL!C j and the conductor loss per unit length can be calculated by (57) and (~hence we have OI.c = Pc /2PT' Using (61), we have R = 201.c';L/C. The conductance per unit length can be calculated by using (50). With L, C, R, and G, the eigenvalue and the eigenvector can be solved from (44) and (45) as

[ = V(G - iwC)(R - iwL) 1=1

V = ZcI = . / R - iwL V G - iwC

(71a) (71b)

(71c)

104

3. Modelling of Loss)' Mlcrostrip Lines

The transfer matrix A( I) for a transmission line of length 1 can be obtained from (67) as

[cosh -yl -Zc sinh -Y/]

A(l) =

-Zc1 sinh-yl cosh-yl

For the case of two symmetric microstrip lines, we first solve the eigenvalue equations in the Iossless medium for the even and the odd modes, respectively. Thus we have

(72)

-y; = -13; = -w2(L11 + L12)(CU + C12) (73a)

I •• = [ :]. v •• = ~:: : ~:: [ : 1 = ~:: : ~:: I •• (736)

-y! = -f3! = -w2(Lu - L12)(CU - C12) (73c)

100 = [ 1 1 ,V 00 = / L11 - L12 [ 1 1 = . / L11 - L12 100

-1 V Cu - C12 -1 V Cu - C12

(73d)

The magnetic flux linkage and the time-average power can be calculated as

fije = L • leo = (L11 + L12)leo Lu + L12

PeT =

Cll + C12

,po = L .100 = (L11 - L12) 100

p. - JLu - L12 oT -

Cll - C12

(74a) (746) (74c) (74d)

The conductor loss per unit length can be obtained by using (57) and (59), thus we have

(75)

The resistance matrix can be solved from (61) as

Lll + L12 J i« - L12

Ru = OleC + oso

Cll + C12 Cu - C12

(76a)

Lll + L12 J Lll - L12

-OloC

Cll + C12 Cll - C12

(766)

8.6 Transmission Line Analysis

105

The conductance matrix £.er unit _!ength can be calculated by using (50). Substituting these I, C, R, and G into (44) and (45), we obtain the eigenvalues and eigenvectors for the even and the odd modes as

1: = (Yll + Y12)(Zll + Z12) Y. = [ :]. V. = Z.Y. I! = (Yll - Y12)(Zll - Z12)

Y. = [~J V. = z.t,

(77a)

(77b)

(77c)

(77d)

where

Zll + Z12 Yll + Y12'

Zll - Z12 Yll - Y12

(7S)

Yll = Gu - iwCll, Zll = Ru - iwLll,

Y12 = G12 - iwC12 Z12 = R12 - iwL12'

(79)

The characteristic impedance matrix can be obtained from (64) as

= = =-1 [(Ze + Zo)/2 (Ze - Zo)/2]

ZC=SV,SI =

(Ze - Zo)/2 (Ze + Zo)/2

(SO)

The transfer matrix oflength I can be calculated from (67) as

= 1

AVI(I) = 2

[ -Ze sinh 1e1- z; sinh10l -Ze sinh lei + z, sinh 101] (S1b) -Ze sinh lei + Zo sinh "}'ol -Ze sinh1el- Zosinh10l

106

3. Modelling of Lossy Microstrip Lines

= 1

AVI(I) = 2

[_Z;l sinh"Y"I- Z;;l sinh "Yo I _Z;l sinh"Y,,1 + Z;;l sinh "Yo1]

-z;» sinh"Y,,1 + Z;;l sinh"Yol _Z;l sinh"Y"I- Z;;l sinh"Yol (SIc)

= 1 [COSh "Y"I + cosh "Yol cosh "Y"l - cosh "Yo1]

AIl(I) = -

2 cosh "Ye1 - cosh "Yol cosh "Yel + cosh "Yol

In the next section, we demonstrate how the transfer matrix is applied to solve the coupled transmission line problems.

(SId)

3.7 Results and Discussions

First, we check the results of our method in calculating the capacitance matrix using the spectral domain Green's function and compare it with other methods using the spatial domain Green's function. In Tables 3.1 to 3.4, we present the results of capacitance matrix for different micros trip line configurations. In Table 3.1, the difference between our results and those in [9] is about 1% for the self-capacitance, and is about 0.03% for the mutual capacitance. The difference from those in [10] is about 0.1% for the self-capacitance, and about 1% for the mutual capacitance.

In Table 3.2, the difference between our results and those in (9] is less than 0.7% for the self-capacitance, and is less than 4% for the mutual capacitance. The difference between our results and those in [10] is about 0.01% for the self-capacitance, and is less than 0.4% for the mutual capacitance.

The capacitance matrix for two microstrip lines embedded in the same and different layers of a two-layered medium are presented in Tables 3.3 and 4, respectively. The difference between our results and those in [9] is about 3% for the self-capacitance and about 1% for the mutual capacitance.

In Fig. 3.5, we present the charge distributions on the broad sides of a microstrip line with t/h = 0.02 and w/h = 0.1, 1.0,2.0. We use 24 pulses on each broad side, and 2 pulses on each narrow side. It is observed that the charge density on the bottom side of the microstrip

S.7 Results and Discussions

107

Comparison of Capacitance Matrix Elements (F 1m).
element present work [9] [10]
ell 0.6301 x 1O-1D 0.6233 X 10-10 0.6307 X 10-10
eu -0.5929 x 10-11 -0.5931 X 10-11 -0.5866 X 10-11
c, 0.6301 x 10-10 0.6233 X 10-10 0.6307 X 10-10 %

Table S.l Comparison of capacitance matrix for two symmetrical striplines of finite thickness.

is about an order larger than that on the top side because the electric field between the micros trip and the ground plane is stronger than the electric field above the micros trip. Also, as the width of the microstrip increases, the edge effect becomes less significant.

In Table 3.5, we present the results of resistance calculation for a micros trip line compared with those in [12]. The discrepancy is 7.6% for w/h = 0.1, and 3.4% for w/h = 2.0. The calculation of the resistance depends on the square of the charge density distribution which possesses edge effect as shown in Fig. 3.5. Hence, even when two different methods can predict close capacitance results, it is possible that the resistance results can deviate by a higher percentage.

Next, we present the frequency response and the transient response

108

3. Modelling of Lossy Microstrip Lines

Comparison of Capacitance Matrix Elements (F 1m).
clement present work [9] [10)
Gil 0.9225 x 10-10 0.9165 X 10-10 0.9224 X 10-10
Gil -0.8539 X 10-11 -0.8220 X 10-11 -0.8504 X 10-11
G22 0.9225 X 10-10 0.9165 X 10-10 0.9224 X 10-10 z

:z:

0' -. 00

Table 3.2 Comparison of capacitance matrix for two symmetrical mlcrostrip lines of finite thickness.

of two symmetrical microstrip lines. The driving voltage is assumed to have a sinusoidal pulse waveform V1(t) with duration T = 200 PICO seconds as

{ (1/2)[1- coS(211't/T)], 0:::; t s T V1(t) =

0, elsew here

(82)

The frequency response and the transient response are presented in Fig. 3.6(a) and Fig. 3.6(b), respectively, with all the four load impedances equal to the characteristic impedance of the even mode ZeO. In Fig. 3.6(a), V1(0) is the voltage at z = 0 along line 1, Vl(l) is the voltage at the receiving port, V2(0) is the near-end coupling, and V2(1) is the far-end coupling. The increase of V2(1) with frequency shows that the high frequency components are responsible for the far-end coupling[16].

3.7 Results and Discussions

109

Comparison of Capacitance Matrix Elements (F 1m).
element present work [9J
Cll 0.3827 x 10-10 0.3720 X 10-10
Cu -0.6884 X 10-11 -0.6889 X 10-11
ell 0.2245 x 10-10 0.2169 X 10-10 z

1.1 1.0

·F~~~:;

----------~--~~0.5--~----------

6.8fo

-0.3 -0.1 0.1 0.3

Table S.S Comparison of capacitance matrix for two microstrip lines of finite thickness embedded in the same layer of a two-layered medium.

In Fig. 3.6(b), the even and the odd modes propagate in different velocities, and the odd mode propagates faster than the even mode. Hence, the waveform V1(l) becomes broader than V1(0) , and V2(1) shows the split of the odd and the even modes. The waveforms V1(0) and V2(0) after t = 800 ps are due to the reflections by the load impedances at y = I. Since the load impedances are chosen to be the same as the characteristic impedance of the even mode, only the odd mode is reflected.

In Fig. 3.7, the transient response is presented with all the four

110

8. Modelling of Lossy Mlcrostrlp Lines

Comparison of Capacitance Matrix Elements (F 1m).
element present work [9]
011 0.3772 x 10-10 0.3651 X 10-10
012 -0.1583 X 10-11 -0.1562 X 10-11
On 0.2152 x 10-10 0.2099 X 10-10 z

(0 0.7
0.6
0.5
0.4
6.8(0
-0.1 0.1
cr - 00 Table 8.4 Comparison of capacitance matrix for two mlcrostrip lines of finite thickness embedded in different layers of a two-layered medIum.

load impedances equal to the characteristic impedance of the odd mode ZoO' The split of the odd and the eV«fn modes is observed again. The waveforms V1(0) and V2(0) after t = 800 ps show that only the even mode is reflected, and the reflected signal arrives later than that in Fig. 3.6(b) because the velocity of the even mode is slower than that of the odd mode.

In Fig. 3.8 and Fig. 3.9, we present the transient responses of two symmetrical microstrip lines with 8 = 0.25 mm and 8 = 0.375 nun, respectively. The load impedances are chosen the same as the characteristic impedance of the even mode. Compared with Fig. 3.6(b), it is observed that the coupling signals V2(0) and V2(1) become weaker

S.7 Results and Discussions

111

0.10
0.08
... 0.06
0
...
X 0.04
-
Ii
-
~ 0.02
0.00
to -0.02
0
....
X -0.04
-
H
- -0.06
...
CI..
I
-0.08 -0. ]0 .................. _.__ ......... _L.-.....L..._L..,.._-1....__JL....__._....L.J

-0.5-0.4-0.3-0.2-0.10.0 0.10.2 0.3 0.4 0.5

z/w

Figure a.5 The charge distribution on the broad aides of a microstrip nne with finite thickness, E., = 11.7, h = 2cm, t/h = 0.02, potential on the microstrip surface is 1 volt, 24 pulses per broad side, and 2 pulses per narrow side.

as the separation s is increased.

In Fig. 3.10, the transient responses with a complex dielectric constant e, = 10 + iO.1 are presented. The signal amplitudes are smaller than those in Fig. 3.6(b) due to the dielectric loss. In Fig. 3.11, the transient responses with copper as the conductor material are presented. The surface resistance is assumed to be 2.61 X 10-7 V1 ohms. The signal amplitudes are slightly different from those with a perfect conductor because the copper itself is good conductor.

112

3. Modelling of Lossy Microstrip Lines

Comparison of Conductor Loss Parameters.
w/h R(O/m) R(O/m) [12] aZoh/R. (dB) aZoh/R. (dB) [12]
0.1 0.07125 0.06605 23.713 21.981
0.2 0.04477 0.03993 14.899 13.289
0.3 0.03398 0.02986 11.309 9.937
0.4 0.02795 0.02400 9.303 8.120
0.6 0.02124 0.01848 7.069 6.150
1.0 0.01492 0.01315 4.967 4.376
1.2 0.01308 0.01173 4.353 3.904
1.4 0.01165 0.Dl063 3.876 3.538
2.0 0.00872 0.00843 2.901 2.807 tr - 00

Table 3.5. Comparison of the conductor loss parameters for a m.icro8trip line of finite thicknes8.

Conclusions

The spectral domain scalar Green's function in a lossy isotropic stratified medium is derived. A rigorous integral equation formulation for the charge distribution on the surfaces of the microstrip lines with finite thickness embedded in arbitrary layers of an isotropic stratified medium is derived. Using the spectral domain Green's function, a multiconductor transmission line analysis is formulated to investigate the propagation properties of coupled lossy microstrip lines. Both the frequency and the transient responses of coupled lines with different load conditions can be obtained. An efficient algorithm is devised based on this approach.

References

113

1.0
0.9
0.8
0.7
0.6
Vlll) 1.0,----,----r---,--.---,

VI (0)

1',(1)

0.5

i

j 0.0 ~-="------f--f:'__:~--t~-I

~

1',(1)

-0.5

10152025~

-1.0 '--_-'-_....J.,._---' __ -'-_-' o

101l

see IIl0II

'"" (10' rA.dianl/.ec.)

Ti.mt(ps)

Figure 3.6 Response of two symmetric micros trip lines, h = 0.2mm, w = 0.125mm, t = 5JLm, .! = 0.125mm, I = 5cm, fT = 10, all conductors are perfect, Zt = Z2 = Zs = Z4 = Z.o ::::; 72.2700. (a) Frequency response. (b) Transient response.

I .0 ,------,--_,.-----r----,----,

0.5
!
" 0.0
~
~
·0.5 Vl(l)

V1(O)

'0 c::::L'~ ~ .. ::C.=: ..

-1.0 '---__._ __ ...I.-_--'- __ ....I-_---J

20"

400

600

1000

Time (pI)

Figure 3.7 Transient response of two symmetric microstrip lines, h = O.2mm, w = O.125mm, t = 5JLm, .! = 0.125mm, I = 5cm, e; = 10, all conductors are perfect, Zl = Z2 = Zs = Z4 = ZoO ::::; 45.8620.

114

3. Modelling of Loss), Microstrip Lines

1.121 ~--..---..---..---.,-----,

111.5
!
.. 111.121
If
~
·111.5 ·1.121 '---__.,----'---_._---'------'

111

20111

401:1

60111

81211:1

100111

Time (pI)

Figure 3.8 Transient response of two symmetric microstrip lines h = 0.2mm, w = 0.125mm, t = 51-'m, _, = 0.25mm, I = 5cm, t'r = 10, all conductors are perfect, Zl = Z2 = Zs = Z4 = ZeO R:i 66.3450.

1.0
VI(O) VI (I)
".5
! VI(O)
.. 111.121
..
~ V2(O)
~ V2(1)
·0.5 (O~'=t

"I .. frfO .= ...
C7 - 00
.1.0
111 21211:1 41210 61211:1 8121111 lool:l
Tim. (pI) Figure 3.9 Transient response of two symmetric microstrip Ilaes, h = 0.2mm, w = 0.125mm, t = 51-'m, _, = 0.375mm, I = 5cm, e, = 10, all conductors are perfect, Zl = Z2 = Zs = Z4 = ZeO R:i 63.5670.

References

1.0 ,....-----r---,---...,.---,....---,

0.5
-
Q
-t.
j 0.0
~
'0.5 Vdl)

VI (0)

V2(1)

'1.0 '--_--I. __ -'- __ -'- __ ...___--'

o

200

800

400

600

115

1000

Time (pI)

Figure 1.10 Transient response of two symmetric microstrip lines, h = O.2mm, w = 0.125mm, t = 5Jl.m, ,= 0.125mm, 1= 5cm, e; = 10 + iO.1, all conductors are perfect, Zl = Z2 = Zs = Z4 = Z,O R! 72.2700.

1.0 r-----,---,---...,.---,....----,

k 0.0 i"'-~--~;-+--l..---.......e~~....;

~

V2(1)

'0.5

-1.0 '------'----'- __ I--_--1.._----l

o

200

400

8011l

1000

Time (PI)

Figure 1.11 Transient response of two symmetric mlcroshlp Ilnes, h = O.2mm, w = 0.125mm, t = 5Jl.m, , = 0.125mm, I = 5cm, e .. = 10, all conductors are copper, Zl = Z2 = Za = Z4 = ZeO R! 72.2700.

116

3. Modelling of Lossy Microstrip Lines

References

[1] 'Iripathi, V. K., "Asymmetric coupled transmission lines in an inhomogeneous medium," IEEE Trans. Microwave Theory Tech., MTT-23, No.9, 734-739, September 1975.

[2) Marx, K. D., "Propagation modes, equivalent circuits, and characteristic terminations for multiconductor transmission lines with inhomogeneous dielectrics," IEEE Trans. Microwave Theory Tech., MTT-21, 450-457, July 1973.

[3} Djordjevic, A. R., T. K. Sarkar, and R. F. Harrington, "Timedomain response of multiconductor transmission lines," Proc. IEEE, No.6, 743-764, June 1987.

[4] Seki, S. and H. Hasegawa, "Analysis of crosstalk in very high-speed LSI/VLSI's using a coupled multiconductor MIS microstrip line model," IEEE Trans. Microwave Theory Tech., MTT-32, No. 12, 1715-1720, December 1984.

[5} Chilo, J. and T. Arnaud, "Coupling effects in the time domain for an interconnecting bus in high-speed GaAs logic circuits," IEEE Trans. Electron Devices, ED-31, No.3, 347-352, March 1984.

[6] Yuan, H. T., Y. T. Lin, and S. Y. Chiang, "Properties of interconnection on silicon, sapphire, and semi-insulating gallium arsenide substrates," IEEE J. Solid-State Circuits, SC-17, No.2, 269-274, Apri11982.

[7] Medina, F. and M. Homo, "Determination of Green's function matrix for multiconductor and anisotropic multidielectric planar transmission lines : a variational approach," IEEE Trans. Microwave Theory Tech., MTT-33, No. 10, 933-940, October 1985.

[8} Koul, S. K. and B. Bhat, "Generalized analysis of microstrip-like transmission lines and coplanar strips with anisotropic substrates for MIC, electro-optic modulator, and SAW application," IEEE Trans. Microwave Theory Tech., MTT-31, No. 12, 1051-1058, December 1983.

[9J Wei, C., R. F. Harrington, J. R. Mautz, and T. K. Sarkar, "Multiconductor transmission lines in multilayered dielectric media," IEEE Trans. Microwave Theory Tech., MTT-32, No.4, 439-449, April 1984.

References

117

[10] Weeks, W. T., "Calculation of coefficients of capacitance of multiconductor transmission lines in the presence of a dielectric interface," IEEE Trans. Microwave Theory Tech., MTT-18, No.1, 35-43, January 1970.

[11] Pucel, R. A., D. J. Masse, and C. P. Hartwig, "Losses in microstrip," IEEE Trans. Microwave Theory Tech., MTT-16, No.6, 342-350, June 1968.

[12] Harrington, R. F. and C. Wei, "Losses on multiconductor transmission lines in multilayered dielectric media," IEEE Trans. Microwave Theory Tech., MTT-32, No.7, 705-710, July 1984.

[13] Syahkal, D. M. and J. B. Davies, "Accurate solution of micros trip and coplanar structures for dispersion and for dielectric and conductor losses," IEEE Trans. Microwave Theory Tech., MTT-27, No.7, 694-699, July 1979.

[14] Kong, J. A., Electromagnetic Waves Theory, Wiley-Interscience, New York, 1986.

[15] Ali, S. M., T. M. Habashy, and J. A. Kong, "Dyadic Green's function for anisotropic multilayered media," submitted for publication.

[16] Fukuoka. Y., Q. Zhang, D. P. Neikirk, and T. Itoh, "Analysis of multilayer interconnection lines for a high-speed digital integrated circuit," IEEE Trans. Microwave Theory Tech., MTT-33, No.6, 527-532, June 1985.