Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance...
Transcript of Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance...
![Page 1: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/1.jpg)
Incorporation of Conductor Losses into theStripline Simulator UA-FWLIS by Usinga Surface Impedance Formulation
Yi Cao,1 Xing Wang,2 Zhaohui Zhu,3 Steven L. Dvorak,1 John L. Prince1
1 Department of Electrical and Computer Engineering, The University of Arizona, Tucson, AZ 857212 NVIDIA Corp, 2701 San Tomas Expressway, MS: 12, Santa Clara, CA 950503 Intel Corp., 5000 W Chandler Blvd., MS: CH5-157, Chandler, AZ 85226
Received 2 November 2006; accepted 4 May 2007
ABSTRACT: We use a surface impedance formulation to enable the MoM-based full-wave
layered interconnect simulator, UA-FWLIS, to handle conductor losses for stripline inter-
connects. Because these approaches are fully compatible with the previously developed ana-
lytical calculations for the reaction matrix elements, the computational efficiency of UA-
FWLIS is not affected by including conductor losses. VVC 2008 Wiley Periodicals, Inc. Int J RF and
Microwave CAE 18: 187–194, 2008.
Keywords: method of moments; full-wave; surface impedance; conductor losses
I. INTRODUCTION
With increasing clock rates in high-performance
VLSI systems, it is becoming more and more impor-
tant to accurately model the performance of intercon-
nects, especially at high frequencies. The MoM-
based, full-wave layered interconnect simulator (UA-
FWLIS) is an efficient interconnect simulation tool
that has been developed for the purpose of modeling
stripline interconnects [1–6]. Because the closely
spaced ground planes cut off the propagation of all
the higher-order modes in the parallel-plate wave-
guide in stripline packaging structures, the reaction
matrix in UA-FWLIS is very sparse. Thus, by using
sparse matrix solution techniques and other accelerat-
ing techniques, UA-FWLIS can achieve much higher
computational efficiency when compared with Agi-
lent Momentum [7].
In the early versions of UA-FWLIS, all conduc-
tors were assumed to be perfect electric conductors
(PEC). However, conductor losses have to be
accounted for in high-frequency interconnect simu-
lations. Surface impedance formulations are the
simplest way to handle conductor losses in inter-
connect and packaging structures, and are widely
used [8–12]. In this article, we incorporate a sur-
face impedance formulation into UA-FWLIS. In
addition to reducing the number of unknowns, we
also show that the surface impedance formulation
avoids the branch-cut singularities that appear
when the ground planes are modeled as finite con-
ductivity regions. The absence of these branch-cut
singularities allows us to directly apply our previ-
ously developed techniques for obtaining closed-
form representations for the reaction elements. It
also allows us to apply the sparse matrix techni-
ques that are developed in [7].
In the following sections, we first validate the use
of surface impedance boundary conditions in lossy
stripline structures. We show that for practical inter-
connect applications, the electric field expressions
Correspondence to: S. L. Dvorak; e-mail: [email protected]
DOI 10.1002/mmce.20277Published online 19 February 2008 in Wiley InterScience (www.
interscience.wiley.com).
VVC 2008 Wiley Periodicals, Inc.
187
![Page 2: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/2.jpg)
that result from the surface impedance approach
accurately approximate the exact solution for the
ideal case where an analytical solution can be easily
obtained. We also discuss the branch-cut singularities
for these two cases. The details of the surface imped-
ance approach for modeling the losses on the traces
and ground planes are also given.
II. SURFACE IMPEDANCEFORMULATION FOR THE TRACES
Because UA-FWLIS employs a surface current for-
mulation, an effective surface current is defined first.
This allows the effects of conductor losses to be eas-
ily included into UA-FWLIS by adding an extra term
to every reaction element. According to Ampere’s
law, the magnetic field and the total current through
the cross section of a long conductor line satisfy the
following equation
ZC
H � dl ¼ZC
n̂ 3 Htan � ðn̂ 3 dlÞ
¼ZC
Jeffs � ðn̂ 3 dlÞ ¼ I; ð1Þ
where C is the perimeter of the cross section and the
unit vector n̂ is chosen normal to the perimeter. Here,
we can see that n̂ 3 Htan can be considered as an
effective surface current density that plays the same
role as the surface current on a PEC. We then use
this effective surface current as the unknown in the
problem and apply the general boundary condition
Zeffs n̂ 3 Htan ¼ Etan ¼ Zeff
s Jeffs ð2Þ
on the nonideal conductor surface, where Zeffs is a
properly defined surface impedance.
On the surfaces of the lossy traces, the integral
equation that is to be solved by the MoM becomes
Zeffs Jeff
s � Eitan ¼ Es
tan
��surface of non-PEC
¼ � �L � Jeffs :
ð3Þ
Therefore, the conductor losses can be included in
the reaction elements by writing
Zmn ¼ Tm; �L � Jeffns
� �þ Tm; Zeff
s Jeffns
� �: ð4Þ
Note that the reaction elements for the lossy traces
consist of a PEC term plus an additional term that
models the effects of the conductor losses on the
traces. Because the additional term is very easy to
compute when the surface impedance is a constant
along the surface of every cell, the reaction elements
can still be calculated efficiently by the previously
developed residue series representation [1].
In general, the surface impedance Zeffs varies
with the location on the perimeter of the cross sec-
tion and for different frequencies, thus, making it
difficult to find an analytical expression for Zeffs .
However, we found that for frequencies higher
than 1 GHz, the wave impedance of a conductive
half space, Z0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffijxl=r
p, is an appropriate approx-
imation for the surface impedance in most off-chip
cases. In most PCB and packaging structures, the
thickness of a typical wire is on the order of
20 lm, and the width is much larger than the thick-
ness. For copper conductors, the skin depth at
1 GHz is 2.1 lm, which is a small fraction of the
thickness of an off-chip line. Therefore, the dimen-
sions of the cross section for off-chip interconnects
are large enough to accurately apply Z0 as the sur-
face impedance at frequencies higher than 1 GHz.
If simulations at lower frequencies are needed,
where the current is distributed relatively uniformly
over the cross section, then a location-dependent
surface impedance [12], can be used. When Z0 is
applied as the surface impedance, we can see that
for a fixed frequency, Zeffs is a constant and the
term including the losses in expression (4) is just
the inner product of the expansion and testing func-
tions. For the case of rooftop expansion and testing
functions, this lossy term has a nonzero value only
for the cases of self-reactions and the reactions
between overlapping cells.
III. SURFACE IMPEDANCEFORMULATION FOR THE POWER/GROUND PLANES IN STRIPLINESTRUCTURES
In this section, we validate the use of a surface im-
pedance formulation for stripline configurations with
lossy ground planes. Here, we first study a simple
homogeneously filled stripline. Such structures can
be modeled as a three-layer dielectric structure as
shown in Figure 1. In this model, the lossy upper and
lower ground planes are considered as semiinfinite
regions with high conductivity, i.e., good conductors.
After investigating this general model we found that
the surface impedance approach provides a good ap-
proximate solution for the electric fields as compared
with the three-layer approach when the inverse Fou-
rier transform variable |kq| is not too large and r is
188 Cao et al.
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 3: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/3.jpg)
large enough. This condition is always met for the
packaging problems where UA-FWLIS is applicable.
After defining the following variables
k1 ¼ k2 ¼ k ¼ xffiffiffiffiffile
p; k0 ¼ k3 ¼ x
ffiffiffiffiffiffile0
p;
kz1 ¼ kz2 ¼ kz; kz0 ¼ kz3 ¼ k0z; e0 � �j
rx; ð5Þ
kq ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffik2 � k2
z
q¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffik02 � k02z
q¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffik2x þ k2
y
q;
we can obtain expressions for the tangential spectral
domain electric fields in regions 1 and 2 (for the pur-
pose of saving space, only the TE mode results are
given)
Figure 1. Three-layer dielectric media model for lossy
ground planes.
TE~EðiÞx ¼ �
~Jsk2k2
y ½ejkzznið1 þ k0z=kzÞ2 þ e�jkzznið1 � k02z =k
2z Þ�
2xek2qkz½ejkzdð1 þ k0z=kzÞ
2 � e�jkzdð1 � k0z=kzÞ2�ejkzzi
�~Jsk
2k2y ½e�jkzznið1 � k0z=kzÞ
2 þ ejkzznið1 � k02z =k2z Þ�
2xek2qkz½ejkzdð1 þ k0z=kzÞ
2 � e�jkzdð1 � k0z=kzÞ2�e�jkzzi
TE~EðiÞy ¼
~Jsk2kxky½ejkzznið1 þ k0z=kzÞ
2 þ e�jkzznið1 � k02z =k2z Þ�
2xek2qkz½ejkzdð1 þ k0z=kzÞ
2 � e�jkzdð1 � k0z=kzÞ2�
ejkzzi
þ~Jsk
2kxky½e�jkzznið1 � k0z=kzÞ2 þ ejkzznið1 � k02z =k
2z Þ�
2xek2qkz½ejkzdð1 þ k0z=kzÞ
2 � e�jkzdð1 � k0z=kzÞ2�
e�jkzzi
ð6Þ
where k02z 5 2jxlr 2 k2q and the variables zi and zni
are defined as
i ¼ 1 : z > zn; z1 ¼ d � z; zn1 ¼ zn
i ¼ 2 : z � zn; z2 ¼ z; zn2 ¼ d � znð7Þ
Next, we employ surface impedance boundary
conditions to model the lossy ground planes at z 5 0
and z 5 d.
The expressions for the electric field components
can now be written as follows:
TE~EðiÞx ¼ �
~Jsk2k2
y ½ejkzzniðZeffs þ xl=kzÞ2 þ e�jkzzniðZeff2
s � x2l2=k2z Þ�
2xek2qkz½ejkzdðZeff
s þ xl=kzÞ2 � e�jkzdðZeffs � xl=kzÞ2�
ejkzzi
�~Jsk
2k2y ½e�jkzzniðZeff
s � xl=kzÞ2 þ ejkzzniðZeff2s � x2l2=k2
z Þ�2xek2
qkz½ejkzdðZeffs þ xl=kzÞ2 � e�jkzdðZeff
s � xl=kzÞ2�e�jkzzi
TE~EðiÞy ¼
~Jsk2kxky½ejkzzniðZeff
s þ xl=kzÞ2 þ e�jkzzniðZeff2s � x2l2=k2
z Þ�2xek2
qkz½ejkzdðZeffs þ xl=kzÞ2 � e�jkzdðZeff
s � xl=kzÞ2�ejkzzi
þ~Jsk
2kxky½e�jkzzniðZeffs þ xl=kzÞ2 þ ejkzzniðZeff2
s � x2l2=k2z Þ�
2xek2qkz½ejkzdðZeff
s þ xl=kzÞ2 � e�jkzdðZeffs � xl=kzÞ2�
e�jkzzi
ð8Þ
where zi and zni are defined in eq. (7).
If we compare the general expressions in eq. (6)
with those in eqs. (8), we find that when kq 5 0 and
k02z 5 2jxlr, the expressions for the electric field for
the three-layer and the surface impedance models are
exactly the same provided we let Zeffs ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffijxl=r
p.
This is expected, since the special case of kq 5 0 is
associated with the normal incidence plane wave case
and Z0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffijxl=r
pis the wave impedance in the con-
ductive regions. We can also show that eqs. (8) are
Incorporation of Losses into UA-FWLIS 189
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 4: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/4.jpg)
good approximations to eq. (6) when |kq / k| is not too
large and r � xe0, i.e., it is a good conductor. In
Figure 2, we plot the relative error for the x-compo-
nent of the TE spectral domain electric field TE~Ex for
the surface impedance formulation, as compared with
the three-layer model formulation, as a function of
kq/k and r/xe, where we assume that the results for
the three-layer model are the exact results. From
Figure 2, we can see that for a good conductor (r �xe0) and for the range of kq that UA-FWLIS encoun-
ters, the use of the surface impedance as a lossy
boundary condition provides a very accurate model
for the lossy ground planes even when we chose the
simplest form for Zeffs . Note that the case |kq| � k is
associated with highly attenuated modes, which do
not contribute appreciably to the reaction elements.
This conclusion is also supported by the study on the
time-domain surface impedance of a homogeneous
lossy half-space [13], which indicated that when the
conductivity of the half-space s is large, the effect of
the different incident angles only appears at very
early-time of the reflected field. For the case of highly
conductive lossy ground planes, this means that only
frequencies that are well beyond hundreds of GHz will
see such effects and subsequently the surface imped-
ance approach would cause errors. In UA-FWLIS, the
value of kq depends on the pole locations and |kq| will
not be too large if the residue series for calculating of
the reaction elements converge quickly. This has been
found to be true for all the inspected cases and the
computational efficiency of UA-FWLIS also largely
depends on this important feature.
Another advantage of the expressions in eqs. (8)
when compared with eq. (6) involves the potential
branch-cut singularities. The branch-cut singularities
are associated with the multivalued square root func-
tions. Because it can be shown that eqs. (8) are even
functions of the variable kz, there are no branch-cut
singularities in these expressions. Therefore, we can
directly apply the techniques that were developed in
[1–6] to obtain closed-form expressions for the reac-
tion elements when the surface impedance formula-
tion is employed. In a similar manner, it can also be
shown that eq. (6) are even functions of kz, so they do
not contain branch-cut singularities associated with
this square root. However, eq. (6) do contain branch-
cut singularities associated with the square root
k0z ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffik02 � k2
q
q. When performing a contour integra-
tion analysis of the integrals, one would have to
deform the contour around this branch cut. Fortu-
nately, since k0 would have a large negative imagi-
nary part, this branch-cut contribution would be small
and could be ignored.
Recall that the expression for the reaction element
between two horizontal cells can be written as [14]
Zmn ¼j
4p2xe
Z 1
0
Z p
�p
~Jnsxe���ze¼zen
~Tmsxtð�kq cos bmn;
� kq sin bmnÞ���zt¼ztm
�GTM
���ze¼ztm
cos h cos bmn
þ GTE���ze¼ztm
sin h sinbmn�kqdhdkq; ð9Þ
for the case of PEC ground/power planes. In the case
of lossy ground/power planes, only the terms GTE and
GTM will be affected by the use of the surface imped-
ance boundary conditions. However, since the Gfunctions are not functions of q (the angular integra-
tion variable), the surface impedance formulation
will not have an effect on the analytical calculation
of the angular integral, and the efficient ILHI based
algorithm can still be used. To apply residue theory
to compute the semiinfinite integral over kq, we need
to find all the poles for the function in the reaction
integrand. Unfortunately, unlike the PEC case, the
TE and TM modes have different poles when there
are lossy conductor planes. In addition, these poles
are now the roots of transcendental equations. In con-
trast, the poles in the PEC ground plane case are ana-
lytically known, i.e., kz 5 np/d, n 5 1,2, . . . . We
employ a Newton iteration method for complex func-
tions [15] to find the poles for the lossy plane case.
The roots for the PEC case can be used as the initial
values for the Newton’s iteration method. In all the
cases we investigated, convergence results in three or
four iterations, since we are assuming good conduc-
tors. In fact, since Z0 is a relatively small value in the
case of good conductors, the roots kz, especially those
associated with the TM mode, are almost identical to
Figure 2. Relative error for the electric field from the
three-layer model and the surface impedance approach.
190 Cao et al.
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 5: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/5.jpg)
np/d. Therefore, if the ground planes have high con-
ductivity, then the effects of lossy ground planes
should be very small.
IV. SIMULATION RESULTS
To validate UA-FWLIS with the conductor loss
enhancement, we investigate several examples and
compare the S-parameters from UA-FWLIS with
those obtained by the commercial MoM-based simu-
lation tool, Agilent Momentum. Note that the S-pa-
rameters in this section are all calculated using the
characteristic impedance of the transmission line as
the reference impedance at each port.
A. Example 1—Multi-Line Filter
The first example we tested is the multiline, stripline
filter, which is shown in Figure 3.
The substrate is taken to be FR4, which has a rela-
tive permittivity of 4.4 and a loss tangent of 0.0001.
The spacing between the two ground planes is 600
lm, and the conductor traces are centered in the mid-
dle of the substrate. All traces are 30 mm long and
400 lm wide, and they are separated by 30 lm from
each other. The ground planes and traces are taken to
be copper, which has a conductivity of 5.8 3
107 S/m. In Figure 4, the S-parameters are plotted
over a frequency range from 5 to 12 GHz. For com-
parison purposes, the results for PEC lines, and lossy
lines with lossless ground planes, are also shown in
these figures. It can be seen that the results from both
simulators agree very well when losses are included.
Slight increases in the insertion loss and return loss
can be observed when the lossy results are compared
with the PEC line results. It should also be noticed
that the effect associated with lossy ground planes is
quite small compared with the effects caused by the
losses on the interconnect lines, as previously antici-
pated.
B. Example 2—Crossover Lines
The next example that we tested is a stripline cross-
over structure, which is shown in Figure 5.
Figure 4. S-parameter results for the filter. (a) S11 magnitude. (b) S21 magnitude.
Figure 3. Multiline filter structure.
Figure 5. Stripline crossover structure.
Incorporation of Losses into UA-FWLIS 191
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 6: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/6.jpg)
The substrate is the same as in Example 1. The
spacing between the two ground planes is 400 lm,
and each line is 22.5 mm long and 120 lm wide. The
ends of the lines without ports were left as open cir-
cuits. Layers 1 and 3 are both 160 lm thick, and the
thickness of the middle layer is 80 lm. The material
for ground planes and the traces is copper with a con-
ductivity of 5.8 3 107 S/m. In Figure 6, we plot S11
and S21 obtained by both agilent momentum and UA-
FWLIS over a frequency range from 4 to 20 GHz.
Once again. we see that the results from both simula-
tors agree very well, and the effect of finite-conduc-
tivity planes is small compared with the effects of the
lossy interconnect lines.
V. CONCLUSIONS
In this article, the capability of handling conductor
losses, from both traces and ground planes, was built
into the full-wave MoM-based stripline simulator,
UA-FWLIS. By introducing an effective surface cur-
rent and using surface impedance boundary condi-
tions, the losses on the conductor traces and ground
planes are easily incorporated into UA-FWLIS by
adding an extra term to the reaction elements and
modifying the electric fields, respectively. These
approaches did not affect the angular integral in the
expression of reaction elements, so the analytical cal-
culation of the reaction matrix elements, which is
based on the use of ILHI and residue theory, was
maintained after including the conductor losses, thus,
also maintaining the computational efficiency of UA-
FWLIS. A comparison between the S-parameter
results from UA-FWLIS with those from Agilent
Momentum validated the accuracy of the enhanced
version of UA-FWLIS. We also observed that the
effects of including finite-conductivity power/ground
planes is relatively small compared with those of the
lossy interconnect lines.
REFERENCES
1. Z. Zhu, Q. Li, X. Wang, S.L. Dvorak, and J.L. Prince,
Extension of an efficient moment-methods-based, full-
wave layered-interconnect simulator to finite-width
expansion functions, IEEE Trans Adv Packaging 30
(2007), 841–850.
2. D.L. Heckmann and S.L. Dvorak, Novel closed-form
expressions for MoM impedance matrix elements for
numerical modeling of shielded passive components,
IEEE Trans Magn 35 (1999), 1534–1537.
3. S. Kabir, S.L. Dvorak, and J.L. Prince, Reaction anal-
ysis in stripline circuits, IEEE Trans Adv Packaging
24 (2001), 347–356.
4. M.M. Mechaik and S.L. Dvorak, Series expansions
for the incomplete Lipschitz-Hankel integral Jeo(a,z),Radio Sci 30 (1995), 1393–1404.
5. M.M. Mechaik and S.L. Dvorak, Series expansions
for the incomplete Lipschitz-Hankel integral Yeo(a,z),Radio Sci 31 (1996), 409–422.
6. Z. Zhu, D.L. Heckmann, S.L. Dvorak, and J.L.
Prince, Numerical computation of incomplete Lip-
schitz-Hankel integrals of the Hankel type for com-
plex-valued arguments, Radio Sci 40 (2005), 1–17.
7. X. Wang, Y. Cao, S.L. Dvorak, J.L. Prince, and Z.
Zhu, Investigation of the sparse MoM reaction matri-
ces produced in stripline packaging problems, 56th
Electronics Components and Technology Conference,
San Diego, CA, June 2006, pp. 1256–1261.
8. T.E. van Deventer, P.B. Katehi, and A.C. Cangellaris,
High frequency conductor and dielectric losses in
Figure 6. S-parameter results for the crossover. (a) S11 magnitude. (b) S21 magnitude.
192 Cao et al.
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 7: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/7.jpg)
shielded microstrip, IEEE MTT-S Dig 3 (1989), 919–
922.
9. N.K. Das and D.M. Pozar, Full-wave spectral-domain
computation of material, radiation, and guided wave
losses in infinite multilayered printed transmission lines,
IEEE Trans Microwave Theory Tech 39 (1991), 54–63.
10. J. Aguilera, R. Marques, and M. Horno, Quasi-TEM
surface impedance approaches for the analysis of
MIC and MMIC transmission lines, including both
conductor and substrate losses, IEEE Trans Micro-
wave Theory Tech 43 (1995), 1553–1558.
11. I.P. Theron and J.H. Cloete, On the surface imped-
ance used to model the conductor losses f microstrip
structures, IEE Proc Microwave Antennas Propag 142
(1995), 35–40.
12. A. Rong, A.C. Cangellaris, and L. Dong, A novel
effective surface impedance formulation for efficient
broadband modeling of Lossy thick strip conductors,
IEEE MTT-S Dig 3 (2003), 1959–1962.
13. H.Y. Pao, Z. Zhu, and S.L. Dvorak, Exact, closed-
form representations for the time-domain surface
impedance of a homogeneous Lossy half-space,
IEEE Trans Antennas Propag 52 (2004), 2659–
2665.
14. R.F. Harrington, Time Harmonic Electromagnetic
Fields, IEEE Press, Wiley, 2001.
15. X. Wang, S. Kabir, J. Weber, S.L. Dvorak, and J.L.
Prince, A study of the fields associated with vertical
dipole sources in stripline circuits, IEEE Trans Adv
Packaging 25 (2002), 272–279.
BIOGRAPHIES
Yi Cao received the B.S.E.E., M.S., and
Ph.D. degree from Shanghai Jiao Tong Uni-
versity, Shanghai, China, in 1994, 1996,
and 1999. From 2004, he is a postdoctoral
researcher at the University of Arizona. His
main research area is the modeling for high-
speed interconnects and packaging, includ-
ing the numerical techniques for solving
field problems and circuit simulation.
Xing Wang received his B.S. degree in
1999 in the Dept. of Electronics in
Tsinghua University, Beijing, China. He
received both M.S. and Ph.D. degrees in
the Dept. of Electrical and Computer
Engineering at The University of Ari-
zona, USA, in 2002 and 2006, respec-
tively. His major is in the interconnect
packaging simulation/design area, espe-
cially in the techniques of full-wave simulation methods. He is
now working in the signal integrity research lab at nVIDIA
Corp., Santa Clara, USA. His interest of research is in packaging
design and simulation, high speed/frequency IC interconnect mod-
eling, simulation, and methodologies.
Zhaohui Zhu (M’06) received the B.S.
degree in electrical engineering from the
University of Science and Technology of
China in 1991, the M.S. degree in signal
processing from the Institute of Electron-
ics, Academia Sinica in 1994, and Ph.D
degree in the Department of Electrical
and Computer Engineering at the Univer-
sity of Arizona, Tucson in 2005. From
1994 to 2000, she was with Telecommunication Department of
China National Clearing Center, where she worked as a system
engineer. She is currently working as a packaging engineer for
Design Process Development in ATD, Intel Corporation. Her
research interests include signal integrity, electromagnetic model-
ing of high-speed circuits, electromagnetic transients, wave propa-
gation, and theoretical and computational electromagnetics.
Steven L. Dvorak received his B.S.
(1984) and Ph.D. (1989) degrees in Elec-
trical Engineering from the University of
Colorado, Boulder. Dr. Dvorak is cur-
rently a Professor in the Department of
Electrical and Computer Engineering at
the University of Arizona, Tucson, Ari-
zona. He served as an Assistant Professor
in this department from 1989 to 1996 and
an Associate Professor from 1996 to 2004. Dvorak previously
held a position with TRW Space and Technology Group from
1984 to 1989. His principal interests include electromagnetic mod-
eling of high-speed interconnects, electromagnetic transients, wave
propagation, theoretical and computational electromagnetics, optics,
geophysical applications of electromagnetics, applied mathematics,
and microwave measurements. Dr. Dvorak is an elected member of
the International Union of Radio Science Commissions B and F,
and a member of the IEEE and Tau Beta Pi. Dr. Dvorak received
the Antennas and Propagation Society S. A. Schelkunoff Prize Pa-
per Award in 1997 and the URSI Young Scientist Award in 1996.
He was also awarded the Department of Electrical and Computer
Engineering IEEE and HKN Outstanding Teaching Award and the
Andersen Consulting Teaching Award in 1994.
John L. Prince (S’65-M’68-SM’78-
F’90) received the BSEE degree from
Southern Methodist University, and as
an NSF Graduate Fellow received the
MSEE and Ph.D. degrees in electrical
engineering from North Carolina State
University. He was a Professor of Elec-
trical and Computer Engineering and
Director of the Center for Electronic
Incorporation of Losses into UA-FWLIS 193
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce
![Page 8: Incorporation of conductor losses into the stripline simulator UA-FWLIS by using a surface impedance formulation](https://reader031.fdocuments.net/reader031/viewer/2022020510/575005c61a28ab1148a64c37/html5/thumbnails/8.jpg)
Packaging Research at the University of Arizona. He came to
the University of Arizona in 1983. He was the Principal Investi-
gator of the Semiconductor Research Corporation (SRC) Pro-
gram in VLSI Packaging and Interconnection Research at the
university from 1984 until his death in December of 2005. In
1991–1992, he was Acting Director, Packaging Sciences at
SRC. He had extensive industrial experience. He was active in
consulting work in both the reliability and packaging areas. He
was co-author on two books in the field of electronic packag-
ing, Simultaneous Switching Noise of CMOS Devices and Sys-
tems, by Senthinathan and Prince, and Electronic Packaging:
Design, Materials, Processing and Reliability, by Lau, Wong,
Prince and Nakayama. He taught courses in electronic packag-
ing and the University of Arizona. His research interests cen-
tered on developing modeling and simulation techniques for
switching noise in packages and MCMs, on modeling and simu-
lation techniques for mixed-signal system packaging, and on
the development of high frequency measurements on packaging
structures. He authored or co-authored over 210 papers in the
field of electronic packaging and 35 papers in the fields of
semiconductor device physics, process development, and reli-
ability.
194 Cao et al.
International Journal of RF and Microwave Computer-Aided Engineering DOI 10.1002/mmce