This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES 1

Plane-Pair PEEC Model for Power Distribution

Networks With Sub-Meshing Techniques
Leihao Wei, Student Member, IEEE, Liang Li, Ketan Shringarpure, Member, IEEE, Albert E.
Ruehli, Life Fellow, IEEE, Edward Wheeler, Senior Member, IEEE, Jun Fan, Fellow, IEEE,
Bruce Archambeault, Fellow, IEEE, and James L. Drewniak, Fellow, IEEE

Abstract—We present an improved plane-pair partial element

equivalent circuit (PEEC) model for power distribution network
modeling based on the PEEC formulation. The model can include
via connections, decoupling capacitor macro-models, and discon-
tinuities such as holes in the plane-pairs. An efficient approximate
inductance sub-meshing model is described for large printed cir-
cuit plane-pairs with complex geometries and numerous vias. The
modified nodal analysis (MNA) used leads to a flexible circuit solu-
tion where we can compute inductances, resistances, impedances,
or other circuit models, including dc solutions. The MNA equa-
tions include effective optimizations such as the placement of ca-
pacitors. Today, a large class of methods are available based on
numerous formulations including finite-difference time-domain, fi-
nite-element method, integral equation model, and cavity models.
Each of the approaches has its own type of problems for which it Fig. 1. Plane-pair geometry example from [4].
is most suitable.
As illustrated in Fig. 1, the IC power pins are connected
Index Terms—Complex printed circuit boards (PCBs), effi- through vias that connect to the PCB’s power and reference
ciency, impedance computations, inductance, partial element plane-pair. The voltage between these planes is established
equivalent circuits (PEECs), partial inductance, sub-meshing,
voltage regulator module (VRM). by the voltage regulator module (VRM). Strategically placed
low-impedance decoupling capacitors are connected between
the power/reference planes. The aim of multiple capacitors
I. INTRODUCTION connected between the power/reference planes is to provide
a low impedance even at high frequencies for PCB power
integrity [2], [5].
An important aspect of the design of IC systems is the ability
P RINTED circuit board (PCB) power distribution systems
must be designed to limit the variation of voltage at all
integrated circuit (IC) chips. This represents a challenging task
to model the power distribution network (PDN) in all its com-
plexity. In this paper we consider modeling the PCB plane part
of the PDN. The efficiency of the plane-pair partial element
as switching speeds as well as the current supply requirements
equivalent circuit (PPP) approach is based on modeling plane-
increase with new designs. Currently, clock frequencies can be
pairs, as was done in [6].
several gigahertz with currents exceeding tens of amperes for
Here, we concentrate on the plane-pair, with one plane
high-end designs. The IC, the on-chip power distribution, must
typically the reference or ground plan and the other the power
provide a low impedance at the highest frequencies. The pre-
or voltage plane. Different approaches are used for plane-pair
vious modeling methods of PDN can be found in, for example,
modeling. One type of model is based on microwave cavity
[1]–[3].
modeling [7]. The original cavity model has evolved into
several different application specific approaches, e.g., [8]. The
Manuscript received May 22, 2015; revised September 13, 2015 and De-
long list of different techniques for plane-pair modeling also
cember 11, 2015; accepted December 19, 2015. This work was supported in
part by the National Science Foundation (NSF) under Grant IIP-1440110. includes lumped 2-D DE circuits or finite-difference time-do-
L. Wei is with the Department of Electrical Engineering, University of Cali- main (FDTD) models [9] and the transmission-line matrix
fornia at Los Angeles, Los Angeles, CA 90025 USA (e-mail: leihao@seas.ucla.
(TLM) approach [10], [11].
edu).
L. Li is with Cisco Systems, San Jose, CA 95134 USA. The conventional plane-pair partial element equivalent cir-
K. Shringarpure is with Apple Inc., San Jose, CA 95014 USA. cuit (PEEC) approach without acceleration techniques is time
A. E. Ruehli, J. Fan, B. Archambeault, and J. L. Drewniak are with the Elec-
and storage intensive due to couplings between all elements
tromagnetic Compatibility (EMC) Laboratory, Missouri University of Science
and Technology, Rolla, MO 65401 USA. [12]. This is a significant obstacle in full-wave analysis when
E. Wheeler is with the Department of Electrical and Computer Engineering, dealing with large PCBs. Fortunately, an improved differential
Rose-Hulman Institute of Technology, Terre Haute, IN 47803 USA.
coupling model [13] can be used to sparsify coupling.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. It has been noted that the inductive far coupling is stronger in
Digital Object Identifier 10.1109/TMTT.2016.2518164 conventional PEEC models than the capacitive coupling [14].

0018-9480 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES
A solution technique is based on the observation that the far
coupling is weaker for the inverse of the partial inductance ma-
trix . Approximations of the inverse were used to partially
avoid costly run times. An example for improvement of the ap-
proach is given in [15]. However, the approach presented in this
paper is much more efficient for parallel planes since the differ-
ence inductive coupling can be exploited for plane-pair models
[16]. This approach is fundamental for the speed improvement,
which we call the PPP approach.
We point out the advantages of the PPP technique in compar-
ison to the widely use conventional differential equation (DE)
models that have been used by many researchers, e.g., [17].
While the DE approach is fast, its accuracy suffers at disconti-
nuities and boundaries and very narrow plane-pairs, which is an
issue similar to short-length transmission-line models. Hence,
the PPP method better accounts for plane boundaries and holes
in planes, etc.
We use parallel-plate-type models for the inclusion of capac-
itances. This will be validated in the results section. Corrections
to the capacitance models can be made by the inclusion of fringe
capacitance approaches.
Speed-up techniques in the PPP approach are based on rectan-
gular Manhattan meshing such that partial inductance sub-ma-
trices for the - and -directions are decoupled. As done in pre-
vious work [16], [18], an approximation of the current distribu-
tion is used to accomplish a mesh reduction without the intro-
duction of – couplings between the and matrices.
For this reason, we devised the PPP approach to model the
system with fewer unknowns. A key advancement made in this
work is multi-level meshing. The opposite plane-pair currents
and the difference-nodes included in the approach implemented
in modified nodal analysis (MNA) [19] matrices are suitable
for the multi-level mesh reduction. The significant gain in the
solution speed realized by using sub-meshing will be presented
in the results section. This approach enables its practical use for
large and complex PCB problems.
The equivalent circuit for the PDN decoupling capacitors can
be changed depending on the application. Two models can be
used for an inductance-only model. In one model type, the ca-
pacitors are replaced by shorts, which essentially results in the
capacitor impedance only due to the inductance of the via con-
nection. Another inductance-only model can also include the in-
ductance part of the capacitor impedance (ESL). For both cases,
the solution vector is real, which results in a fast computing
time.
If we also add resistance of the PEEC cells and capacitors to
the solution, we solve the R–L problem, which leads to a com-
plex solution vector. Adding the capacitances with or without
the (ESC) capacitance of the decoupling capacitors results in a
complex impedance model. Both Joule and dielectric losses can
be readily added to the model with the result that the approach
presented in this work is both flexible and powerful.
In Sections II and III, the PPP model and its PPP implemen-
tation are detailed. Section II-B introduces the modeling of the
vertical interconnects or vias, capacitances, and loss models.
Acceleration with an approximate multi-level submeshing is
presented in Section IV. Finally, validation of simulations using
this approach are given in Section V.
Fig. 2. Plane subdivision to PEEC circuit (with inductances only), a hole is
modeled by deleting a node.
Fig. 3. Smallest example problem for two planes.
II. OUTLINE OF PPP MODEL IN FREQUENCY DOMAIN
A. Inductance Model for PPP
Partial inductance has been discussed in [20], allowing the
inductance to be split into constituent partial inductances. In
addition, the orthogonality of the currents in the vias and the
planes allows the modeling of vertical elements (vias) without
partial mutual coupling inductances in the horizontal direction
(plane).
Plane-pairs are subdivided using the orthogonal PEEC [12]
meshing shown in Fig. 2. A partial inductance is assigned to
each rectangular mesh cell indicated by dashed lines, in either
the or direction [16], where the partial inductances are con-
nected to the nodes. The grey -directed cells and half-cells are
pointing down in Fig. 2, while the orthogonal -cells and half
cells at the boundaries are pointing to the right.
Both planes require the same congruent meshing. Holes in
the plane-pairs can easily be made by eliminating the appro-
priate partial inductances in the models, as shown in Fig. 2.
The orthogonality between the currents prevents mutual cou-
plings between and cells, which further increases matrix
sparsity. Fig. 3 represents the simplest possible example for a
PEEC two-plane model to explain the basic formulation where
the cells are represented by lines [16]. Four nodes are used per
plane to represent the meshing with one power via and one ref-
erence via. A current is injected at nodes 4 and 4 in order to
find the inductance between the planes at port 4. Also, the load,
which is connected between the planes at nodes 1 and 1 , is
simply represented by a short-circuit connection, the simplest
model for a decoupling capacitor valid due to the large capac-
itance values. The plane-pair inductance—the inductance be-
tween the power and reference planes—is therefore separated
from via inductances by the plane inductance. In this example,
the inductance at N4 is given by .
The fundamental structure of the MNA matrix, shown in (20)
at the bottom of page 6, is the same for practical problems with
. In order to guarantee a good conditioning of the MNA
matrix we use mA, V, GHz, ns as normalized units and k , pF,
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WEI et al.: PPP MODEL FOR PDNs WITH SUB-MESHING TECHNIQUES 3

TABLE I
COMPARISON OF SECTION-TO-SECTION COUPLING INDUCTANCE DECAY

Fig. 4. Coupling between two cell pairs.

Fig. 6. Section-to-section approximation by one- and two-filament method.

the mutual coupling between cells can be entirely neglected at

Fig. 5. PEEC equivalent circuit of coupling between cell pairs. larger distances while still maintaining accuracy to the solution
with the inclusion of all mutual terms. More importantly, by set-
and H for the circuit. Note that circuit element stamps for the ting long-distance coupling to zero, it leads to a sparse matrix,
shorting and decoupling capacitors are added at the end of the
unknown vector so that changes can be made without touching (nH) (4)
the rest of the matrix. This formulation was presented in [16]
and [18]. The wider cells partial inductances can be approximated
In the difference inductance model [16], [21], the cell pairs by using a two-filament representation for each conductor.
and shown in Fig. 4, represented by the equivalent circuit For two filaments in each surface we have and
shown in Fig. 5 represent two sections in the plane, each con- representing conductor , and and representing
taining a power and a ground cell component. The voltage drop conductor . For a normalized cell width of 1, fila-
on the left cell pair induced by current on the right ments are placed at and by Gaussian
cell pair is given by (1). The mutual partial inductance quadrature with a weight of 0.5 each. We then have
between sections is given in (2). The mutual partial inductance for
and can be found by the formula for the mutual the difference cell section to cell section coupling.
inductance of rectangular sheets [22]. Of course, this formula Generally, for -filaments approximations
can also be used to find the self partial inductance of a section
( ) in (3), (5)
(1)
For the difference section pairs,
(2)
(3)

Dipole-like cancellation ( ) is observed in the mutual cou-

pling between cells due to the opposite currents of each element.
Therefore, mutual coupling reduces rapidly.
These difference segment couplings decay much faster than
the single-layer partial mutual inductances. Fast approximations
is applied at middle distances and the couplings are neglected (6)
for large distance compared to the plane separation .
For the purpose of accelerating MNA matrix element calcula- The comparison is given in Table I. In this example, the sec-
tion, the mutually inductance (4) between cells is accurately ap- tion length is 1 mm 1 mm and vertical cell spacing is 0.2 mm.
proximated with a one-filament model, where and In Fig. 6, we compare their convergences to zero thickness rect-
are section side lengths and [21]. On the other hand, angular sheets model versus the distance of two sections. The
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES

result shows that the two-filament approach of a partial induc-

tance is better than the one-filament approach since it provides
faster convergence to the zero thickness rectangular sheet model
(solid line). Essentially, we use the two filament model to re-
place the zero thickness model at middle distances in this paper.
In this work we use a multilevel sub-meshing strategy to
accurately capture the effects of current crowding in the vicinity
of the vias. This permits smaller mesh sizes to be used in the
vicinity of vias, where currents and their spatial variation are Fig. 7. Partial mutual inductance for two filament wires.
both significant. At the same time, using larger meshing away
from via regions thereby maintains accuracy while dramati-
cally reducing run times in comparison to solutions without
submeshing.

B. Approximate Models for the Via Connections

The required complexity of the via models depends on two
factors, one being the highest frequency for which the model is
to remain valid and the other the maximum tolerable compute
time. Fortunately, due to the mid frequency range in our appli-
cations, one filament model suffices to represent a via [4].
Rectangular approximations for vias are used in several pa-
pers [16], [21]. One can also approximate the number of con-
tacts to the via to represent a round via [4]. The round via has a
radius and length . We assume that is larger than the diam-
eter . With a high-frequency skin-effect approximation,
the partial self inductance of a round wire is approximated using
the partial self inductance of a zero thickness cylindrical tube in Fig. 8. Top view of partial capacitances, dashed lines enclose current cells.
(7), where [23]. Since this formula is inaccurate if
we also use the approximate round wire formula (8) [24].
A square model approximation is used in this work and the and
via connects to the plane with the via diameter equal to side
length,

C. Adding Capacitances to PPP Model

The plane capacitances lead to additional resonances at
higher frequencies. Very high-frequency PDN models require
the inclusion of capacitances as well as inductances. It has
(7)
already been shown that inductive coupling is relatively local
for plane-pairs. Since it is well known that capacitive coupling
for plane-pairs is more local than inductive, it is sufficient to
add a conventional DE model to the PPP capacitance model.
(8) Further, this enhances the overall simplicity of the model.
High-frequency solutions can be obtained without including
The coupling between vias is accurately represented by two retardation since in the difference models the couplings are
parallel filaments [13], [24], shown in Fig. 7. This approxima- very local.
tion holds provided that the via diameter is small compared to Using the parallel-plate capacitor model, the partial capaci-
via spacing. The analytical form of the partial inductance is tance is , where is the cell area, as shown in Fig. 8,
given for this case by (9), and is the plane separation. The capacitances are stamped into
the main diagonal of the upper left part of the MNA matrix (20).
Four cases for capacitance are shown in Fig. 8 and a capac-
itance plate is shown by the gray area where the capacitance is
proportional to with different ratios [4].
(9)
D. Plane Skin-Effect Model and Losses for Dielectric
with Losses can be included in the PEEC circuit models at dif-
ferent levels guided by the usual accuracy and complexity trade-
offs. Most of the PPP models are applied such that fast solutions
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WEI et al.: PPP MODEL FOR PDNs WITH SUB-MESHING TECHNIQUES 5

result. The most efficient model uses zero thickness partial in- TABLE II
ductances, and we assume that the frequency is sufficiently high EXAMPLE OF MNA MATRIX STAMP FOR VIAS
so that the conductor thickness is larger than several skin-depths
, where is the conductor thickness and is the conduc-
tivity of the planes,

(10)

We can use several different models for the plane resistance TABLE III
NUMERICAL INTEGRATION METHODS
for a cell. The resistance of the planes can easily be included
in the MNA equations as is done in (20) with added to the
cell coupling self inductance. The usual 1-D skin-effect model
could be used for finite thickness conductors, or

(11)
Similarly, at time , branch , node to , the Kirchoff's
where the factor of 2 accommodates the two plane resistance current law (KCL) equation can be written as (15),
and . To satisfy causality, the model must be Hilbert
consistent, and we also need to take the impact of the inductive
skin-effect part into account [25].
Including dielectric loss is not of key importance for mid-
frequencies in the lower GHz range. Debye models with few
elements can easily be added to the capacitor model to account
for dielectric loss where again the added accuracy will result in
increased compute time [26].
(15)
III. TIME-DOMAIN PPP
We have shown the overall PPP MNA formulation in the fre- Thus, the PPP MNA matrix in the time domain takes the fol-
quency domain. However, it is worth mentioning that this for- lowing form in (16):
mulation in the frequency domain can be easily converted to a
time-domain solution. Consider two standard DEs for a capac-
itor and inductor in (12), which describes time characteristics
for a capacitor from node to and an inductor at branch . An
example for the MNA inductance stamp for via inductances is (16)
given in Table II.
First- or second-order numerical integration techniques Note that if we take uniform time steps, , the time-domain
in (13) with coefficients listed in Table III, can be used for MNA matrix will not change, which leads to a fast solution.
time-domain PPP, where presents the th time step. For
the purpose of stability and accuracy, we use the second-order IV. ACCELERATION THROUGH THE USE OF
backward difference (BD2) integration method in this work, MULTILEVEL SUB-MESHING
which is given by the last row in Table III,
A. Sub-Dividing Mesh Cells and Weighted KCL
Here, an example of three levels of sub-meshing is given in
(12) Fig. 9 to illustrate the transitioning of meshing. Each sub-mesh
level halves the size of the mesh dimensions. To achieve a con-
(13) tinuous transitioning of current across the boundary, weighted
KCL equations can be used to make transition as continuous as
At time , node , the Kirchoff’s voltage law (KVL) equa-
possible, where the current in each partial inductance is assumed
tion can be written as (14) with appropriate derivative approxi-
to be uniformly distributed on cells.
mation,
Based on the ratio of the corresponding boundary of two sub-
meshing cross-section lengths, one can obtain weighted KCL
coefficients as discussed in [4].
Type I

Type II

(14)
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES
Fig. 9. Transitioning region of sub-meshing, contains Type I–V nodes from [4].
Shading highlights the inductive cells, and dash lines are boundaries of current
cells.
Type III
Type IV: G8
Type V: S1
B. Approximation of Mutual Inductance Coupling Using
Multi-Level Sub-Meshing
Thus far, we considered two speed-up techniques used, the
symmetry of the two planes, and the multi-level meshing. In
this section, we consider two other approaches. In the first ap-
proach, we use approximations for inductive coupling, while in
the second one we entirely neglect these small inductive cou-
plings. Both approaches are possible due to the rapid decay of
the difference inductive coupling. This results in a sparse matrix
whose solution is fast to compute.
The decay of the difference coupling and the approximate
coefficients for transmission-line-type problems was observed
in [13]. For plane-pair modeling a similar approach was applied
in [16], [18]. Here, we extend the approach to the multilevel
sub-meshing presented in Section IV-A. The rapid decrease in
the mutual inductance between cells versus distance between
cells is an important task addressed in this section.
First, consider the criterion used to switch from the cell-pair
to cell-pair inductance to the two-filament pair to two-filament
pair inductance. The relative error of calculating mutual induc-
tance by closed-form expressions compared to that from two-fil-
ament approximations is given by (18) with 3% error used here
as sufficient accuracy for most purposes,
(18)
An example is provided here of coupling between 0.125 mm
0.125 mm section size to 0.25 mm 0.25 mm section size
with 0.2-mm plane separation. The partial mutual inductance is
calculated by closed-form formula and the approximation for-
mula. From our study it is clear that when the section distance
is larger than approximately 0.74 mm, the 3% approximation
rule is met. With this method, the family of curves given in
Fig. 10 can be used to find required section distance. The re-
quired section distance is normalized to the larger value of sec-
tion sizes and plane separation. The index indicates the dif-
ference in level between two sections, e.g., means, two
commensurate sections. As shown in Fig. 10, less required nor-
malized section distance is necessary when the initial section
size is large.
Second, we consider the criterion used to neglect coupling
entirely according to (4), and one may define the normalized
inductance as the mutual partial inductance divided by the self
partial inductance in (19) [21]. An example of fast decay of
mutual normalized partial inductance is plotted in Fig. 11. At a
distance of 1.35 mm, the mutual partial term is already less than
10 of the self partial term. Hence, these mutual terms may
be neglected entirely for a sufficiently large section distance.
Fig. 12 shows a family of curve of required normalized section
distance versus section size,
(19)
(20)
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WEI et al.: PPP MODEL FOR PDNs WITH SUB-MESHING TECHNIQUES 7

Fig. 14. Simplified model.

Fig. 10. Required normalized section distance for the approximation formula.

Fig. 11. Normalized inductance for zero estimation. Fig. 15. Impedance frequency spectrum.

TABLE IV
RUN-TIME COMPARISON

Fig. 12. Required normalized section distance for zero estimation.
4) ; ; S m.
The frequency range is from 10 MHz to 1 GHz. Three levels
of sub-meshing are used with four contacts on each side of
the via, total unknowns on the object are 4708, resulting in a
calculation time of 272 s. The current code is implemented in
MATLAB script, although a C code would speed up calcula-
tion by a factor of 2 to 10. A CST model with 139 365 meshcells
takes 212-s simulation time. The result shows close agreement
with CST simulations, as shown in Fig. 15.

B. Run-Time Reduction
The multi-level sub-meshing method reduces the number of
unknowns, leading to significant reductions in the solve time,
allowing larger systems to be modeled than those possible prior
to implementing the sub-meshing strategy.
Fig. 13. Original model. We use a larger plane-pair with 500 mm 250 mm plane size
and 0.2-mm plane spacing as the test geometry. The center of
V. VALIDATION OF SIMULATIONS the short is located at (250 mm, 62.5 mm), and the source center
is located at (250 mm, 187.5 mm). The via size for both short
A. Impedance Sweep Validation and source is 0.25 mm 0.25 mm. We use two, three, and five
The part of the real PCB structure in Fig. 13 was simplified contacts of voltage nodes on each side of a via for different
as shown in Fig. 14. The test parameters are as follows. configurations.
1) Plane separation mm (10 mil). Comparison for the time reduction is shown in Table IV. We
2) Via diameter mm. are approximating the via with a more realistic round-shape ap-
3) Short is at (100 mm, 60 mm); source is at (25.2 mm, proximation in five contacts configuration. Further, it is inter-
2.8 mm). esting to consider a resistance-only solution. Its run-time results
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES
are also given for comparison. The MNA matrix is very sparse
for this case since resistance models are not mutually coupled.
We did not provide uniform mesh information because there
are too many unknowns to be incompatible with the modest
computer resources available to us.
VI. CONCLUSIONS
The PPP method with the use of multi-level sub-meshing
requires significantly shorter run times than uniform meshing
while maintaining high accuracy compared to the traditional
PPP approach. The approach results in refined meshing near
the vias and other connections, hence capturing the rapid spatial
variation of current. Regions far from the via and other connec-
tions are represented by more sparse meshes, which resulted in
a significant reduction in the number of unknowns. Run time is
further reduced through the rapid decay of the inductive cou-
plings, which leads to further spasification.
The PPP method shows its promise and flexibility in mod-
eling a plane-pair in a PDN. Although we recognize that mod-
eling a plane-pair is only part of a PDN model, it is feasible to
cascade layer pairs, as is conventionally done.
REFERENCES
[1] W. D. Becker et al., “Modeling, simulation and measurement of mid-
frequency simultaneous switching noise in computer systems,” IEEE
Trans. Adv. Packag., vol. 21, no. 2, pp. 157–163, May 1998.
[2] L. Smith, R. Anderson, D. Forehand, T. Pelc, and T. Roy, “Power distri-
bution system design methodology and capacitor selection for modern
CMOS technology,” IEEE Trans. Adv. Packag., vol. 22, no. 3, pp.
284–291, Aug. 1999.
[3] M. Swaminathan, K. Joungho, J. Novak, and J. P. Libous, “Power
distribution networks for system-on-package: Status and challenge,”
IEEE Trans. Adv. Packag., vol. 27, no. 2, pp. 286–300, May 2004.
[4] L. Wei, K. Shringarpure, A. Ruehli, E. Wheeler, and J. Drewniak,
“Plane-pair PEEC models for PDN using sub-meshing,” in IEEE
23rd Electr. Perform. Electron. Packag. Syst. Conf., Oct. 2014, pp.
159–162.
[5] M. Popovich and E. Friedman, “Decoupling capacitors for
multi-voltage power distribution systems,” IEEE Trans. Very Large
Scale Integr. (VLSI) Syst., vol. 14, no. 3, pp. 217–228, Mar. 2006.
[6] K. Jingook, K. Shringarpure, F. Jun, K. Joungho, and J. L. Drewniak,
“Equivalent circuit models for power bus design in multi-layer PCBs
with via arrays,” IEEE Microw. Wireless Compon. Lett., vol. 21, no. 2,
pp. 62–64, Feb. 2011.
[7] T. Okoshi and T. Miyoshi, “The planar circuit-an approach to mi-
crowave integrated circuitry,” IEEE Trans. Microw. Theory Techn.,
vol. MTT-20, no. 4, pp. 245–252, Apr. 1972.
[8] J. Kim, J. Fan, A. E. Ruehli, J. Kim, and J. L. Dewniak, “Inductance
calculations for plane-pair area fills with vias in a power distribution
network using a cavity model and partial inductances,” IEEE Trans.
Microw. Theory Techn., vol. 59, no. 8, pp. 1909–1924, Aug. 2011.
[9] H. H. Wu, J. W. Meyer, K. Lee, and A. Barber, “Accurate power supply
and ground plane models,” IEEE Trans. Antennas Propag., vol. 22, no.
3, pp. 259–266, Mar. 1999.
[10] P. B. Johns and R. L. Beurle, “Numerical solution of 2-dimensional
scattering problems using a transmission-line matrix,” Proc. IEEE, vol.
59, no. 9, pp. 1203–1208, Sep. 1971.
[11] S. Roy and A. Dounavis, “Macromodeling of multilayered power
distribution network based on mulitconductor transmission line ap-
proach,” IEEE Trans. Compon., Packag,. Manuf. Technol. B, vol. 3,
p. 1047, Jun. 2013.
[12] A. E. Ruehli, G. Antonini, J. Esch, J. Ekman, A. Mayo, and A. Orlandi,
“Non-orthoganal PEEC formulation for time and frequency domain
EM and circuit modeling,” in Proc. IEEE Int. Electromagn. Compat.
Symp., May 2003, vol. 45, pp. 167–176.
[13] A. E. Ruehli, C. Paul, and J. Garrett, “Inductance calculations using
partial inductances and macromodels,” in Proc. IEEE Int. Electromagn.
Compat. Symp., Aug. 1995, pp. 23–28.
[14] A. Devgan, H. Ji, and W. Dai, “How to efficiently capture on-chip in-
ductance effects: Introducing a new circuit element ,” in Proc. Int.
Comput.-Aided Design Conf., San Jose, CA, Nov. 2000, pp. 150–155.
[15] D. Daroui and J. Ekman, “Efficient PEEC-based simulation using re-
luctance method for power electronics applications,” Appl. Comput.
Electromagn. Soc. J., vol. 27, no. 10, pp. 830–841, Oct. 2012.
[16] F. Zhou, A. E. Ruehli, and J. Fan, “Efficient mid-frequency plane
inductance computations,” in Proc. IEEE Int. Electromagn. Compat.
Symp., Fort Lauderdale, FL, USA, Aug. 2010, pp. 831–836.
[17] M. Swaminathan and A. E. Engin, Power Integrity Modeling and De-
sign for Semiconductors and Systems. Upper Saddle River, NJ: Pren-
tice-Hall, 2007.
[18] E.-P. Li, Electrical Modeling and Design for 3D System Integration.
Hoboken, NJ, USA: Wiley, 2012.
[19] C. Ho, A. Ruehli, and P. Brennan, “Interactive circuit analysis and de-
sign using APL,” in Proc. IEEE Int. Circuits Syst. Symp., May 1975,
pp. 504–509.
[20] A. E. Ruehli, “Inductance calculations in a complex integrated circuit
environment,” IBM J. Res. Develop., vol. 16, no. 5, pp. 470–481, Sep.
1972.
[21] L. Li, A. E. Ruehli, and J. Fan, “Accurate and efficient computation
of power plane pair inductance,” in Proc. IEEE Int. Electromagn.
Compat. Symp., Philadelphia, PA, USA, Aug. 2012, pp. 167–170.
[22] P. K. Wolff and A. E. Ruehli, “Inductance computations for complex
three dimensional geometries,” in Proc. IEEE Int. Circuits Syst. Symp.,
1981, pp. 16–19.
[23] A. E. Ruehli, G. Antonini, and L. Jiang, “Skin-effect model for round
wires in PEEC,” in IEEE EMC Int. EMC Symp., Rome, Italy, Sep. 2012,
pp. 1–6.
[24] F. W. Grover, Inductance Calculations: Working Formulas and Ta-
bles. New York, NY, USA: Dover, 1962.
[25] A. E. Ruehli, G. Antonini, and L. Jiang, “Skin effect loss models for
time and frequency domain PEEC solver,” Proc. IEEE, vol. 101, no. 2,
pp. 451–472, Feb. 2013.
[26] G. Antonini, A. E. Ruehli, and A. Haridass, “Including dispersive di-
electrics in PEEC models,” in Electr. Perform. Electron. Packag. Dig.,
Princeton, NJ, USA, Oct. 2003, pp. 349–352.
Leihao Wei (S’12) received the B.S. and M.S. de-
grees in electrical engineering from the Rose-Hulman
Institute of Technology, Terre Haute, IN, USA,
in 2014 and 2015, respectively, and is currently
working toward the Ph.D. degree in electrical en-
gineering (with an emphasis on physical and wave
electronics) at the University of California at Los
Angeles, Los Angeles, CA, USA.
He is also a Member of the Terahertz Electronics
Laboratory, University of California at Los Angeles.
Liang Li, photograph and biography not available at the time of publication.
Ketan Shringarpure (GSM’10–M’13), photograph and biography not avail-
able at the time of publication.
Albert E. Ruehli (M’65–SM’74–F’84–LF’03)
received the Ph.D. degree in electrical engineering
from the University of Vermont, Burlington, VT,
USA, in 1972, and an Honorary doctorate from
Luleå University, Luleå, Sweden, in 2007.
He has been a member of various projects
with IBM, including mathematical analysis, semi-
conductor circuits and devices modeling, and as
Manager of a very large scale integration (VLSI)
design and computer-aided design (CAD) group.
Since 1972, he has been with the IBM T. J. Watson
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
WEI et al.: PPP MODEL FOR PDNs WITH SUB-MESHING TECHNIQUES
Research Center, Yorktown Heights, New York, USA, where he was a Research
Staff Member with the Electromagnetic Analysis Group. He is currently an
Emeritus of IBM Research and an Adjunct Professor in the electromagnetic
compatibility (EMC) area with the Missouri University of Science and Tech-
nology. He has authored or coauthored over 200 technical papers. He edited
Circuit Analysis, Simulation and Design (North Holland, 1986 and 1987).
Dr. Ruehli has served in numerous capacities for the IEEE. In 1984 and 1985,
he was the Technical and General Chairman, respectively, of the ICCD Inter-
national Conference. He has been a Member of the IEEE Administrative Com-
mittee (AdCom) for the Circuit and System Society and an Associate Editor
for the IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED
CIRCUITS AND SYSTEMS. He has given talks at universities including keynote
addresses and tutorials at conferences, and has organized many sessions. He is
a Member of SIAM. He was the recipient of the IBM Research Division or IBM
Outstanding Contribution Award (1975, 1978, 1982, 1995, and 2000). He was
the recipient of the Guillemin–Cauer Prize Award for his work on waveform
relaxation (1982) and a Golden Jubilee Medal (1999), both from the IEEE Cir-
cuits and Systems (CAS) Society. In 2001, he was the recipient of a Certificate
of Achievement from the IEEE EMC Society for inductance concepts and the
partial element equivalent circuit (PEEC) method. He was the recipient of the
2005 Richard R Stoddart Award, and the 2007 Honorary Life Member Award
of the IEEE Electromagnetic Compatibility Society for outstanding technical
performance. In 2010, he was the recipient of a Best Paper Award of the EPEPS
Conference for his work on optimized waveform relaxation.
Edward Wheeler (S’94–M’95–SM’97) was born in
Vincennes, IN, USA, in 1959. He received the Ph.D.
degree from the University of Missouri–Rolla, Rolla,
MO, USA, in 1996.
He holds the Lawrence J. Giacoletto Endowed
Chair with the Rose-Hulman Institute of Technology,
Terre Haute, IN, USA, where he is currently a Pro-
fessor with the Electrical and Computer Engineering
Department. He is a member of the Electromag-
netics and High-Speed Design Group, Rose-Hulman
Institute of Technology.
Dr. Wheeler was the recipient of the IEEE Third Millennium Medal in 2001.
Jun Fan (S’97–M’00–SM’06–F’16) received the
B.S. and M.S. degrees in electrical engineering from
Tsinghua University, Beijing, China, in 1994 and
1997, respectively, and the Ph.D. degree in electrical
engineering from the University of Missouri–Rolla,
Rolla, MO, USA, in 2000.
From 2000 to 2007, he was with the NCR Corpora-
tion, San Diego, CA, USA, as a Consultant Engineer.
In July 2007, he joined the Missouri University of
Science and Technology (formerly the University of
Missouri–Rolla), where he is currently an Associate
9
Professor. He is the Director of the Missouri S&T Electromagnetic Compati-
bility (EMC) Laboratory and the Director of the National Science Foundation
Industry/University Cooperative Research Center for EMC. His research inter-
ests include signal integrity and electromagnetic interference (EMI) designs in
high-speed digital systems, dc power-bus modeling, intra-system EMI and RF
interference, printed circuit board (PCB) noise reduction, differential signaling,
and cable/connector designs.
Dr. Fan served as the Chair of the IEEE EMC Society TC-9 Computational
Electromagnetics Committee (2006–2008). He was a Distinguished Lecturer of
the IEEE EMC Society (2007 and 2008). He currently serves as the Chair of the
Technical Advisory Committee, IEEE EMC Society, and is an Associate Editor
for the IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY and the
IEEE Electromagnetic Compatibility Magazine. He was the recipient of an IEEE
EMC Society Technical Achievement Award in 2009.
Bruce Archambeault (M’87–SM’02–F’06) re-
ceived the B.S.E.E degree from the University
of New Hampshire, Durham, NH, USA, in 1977,
the M.S.E.E degree from Northeastern University,
Boston, MA, USA, in 1981, and the Ph.D. degree
from the University of New Hampshire, in 1997. His
doctoral research was in the area of computational
electromagnetics applied to real-world electromag-
netic compatibility (EMC) problems.
He is an IBM Distinguished Engineer Emeritus
and an Adjunct Professor with the Missouri Univer-
sity of Science and Technology, Rolla, MO, USA. He has taught numerous
seminars on EMC and signal integrity across the USA and the world, including
the past 12 years with Oxford University. He has authored or coauthored
a number of papers in computational electromagnetics, mostly applied to
real-world EMC applications. He authored PCB Design for Real-World EMI
Control and lead authored the EMI/EMC Computational Modeling Handbook.
Dr. Archambeault is a Member of the Board of Directors, IEEE EMC Society,
and a past Board of Directors Member for the Applied Computational Electro-
magnetics Society (ACES). He currently serves as the Vice President for Confer-
ences of the EMC Society. He has served as a past IEEE/EMCS Distinguished
Lecturer, EMCS TAC Chair, and Associate Editor for the IEEE TRANSACTIONS
ON ELECTROMAGNETIC COMPATIBILITY.
James L. Drewniak (S’85–M’90–SM’01–F’07)
received the B.S., M.S., and Ph.D. degrees in elec-
trical engineering from the University of Illinois at
Urbana–Champaign, Urbana, IL, USA.
He is currently with the Electromagnetic Com-
patibility Laboratory, Department of Electrical
and Computer Engineering, Missouri University
of Science and Technology, Rolla, MO, USA. His
research is in electromagnetic compatibility, signal
and power integrity, and electronic packaging.