IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO.

1, FEBRUARY 2015

135

Efficient Analysis of Power/Ground Planes Loaded

With Dielectric Rods and Decoupling Capacitors
by Extended Generalized Multiple Scattering Method
Xinxin Tian, Yao-Jiang Zhang, Senior Member, IEEE, Dazhao Liu, Liangqi Gui, Qingxia Li, Member, IEEE,
and Jun Fan, Senior Member, IEEE

AbstractGeneralized multiple scattering (GMS) method, previously proposed for signal integrity analysis of vias, is now extended to power integrity analysis of power/ground planes loaded
with circular dielectric rods and decoupling capacitors. The transition matrices of the rods and decoupling capacitors are derived
from boundary value problems and equivalent circuits, respectively. The transition matrices are then regarded as loads to the
radial scattering matrix obtained by the GMS method. Therefore,
a parallel-plate impedance matrix, which characterizes the power
integrity performance, can be obtained. To understand physically
different noise suppression mechanisms in power/ground planes
with dielectric rods or photonic crystals, the field distributions in
power/ground planes are derived in more detail. It is found that
there are three kinds of resonances: one due to the cavity formed
by the plane pair, one due to the cavity formed by surrounding
dielectric rods, and one caused by the individual dielectric rod itself. The accuracy and efficiency of the extended GMS method are
verified by comparing with a commercial full-wave solver.
Index TermsElectromagnetic band gap (EBG), multiple
scattering method, photonic crystal, signal/power integrity, simultaneously switching noise (SSN).

I. INTRODUCTION
ELIABLE design of power delivery networks (PDNs) is
becoming more and more challenging with the higher
clock frequencies, faster edge rates, and lower voltage levels
found in dense high-speed printed circuit boards (PCBs) [1].
Hybrid field-circuit via models [2][5] have revealed that the
PDN in a multilayered structure of PCBs actually serves as the
return current path for signal link paths (SLPs) or signal channels
consisting of interconnects of traces, parallel plates, and vias.
High-speed signals along vias crossing a pair of power/ground

Manuscript received June 11, 2014; revised August 27, 2014 and September
24, 2014; accepted September 28, 2014. Date of publication November 20,
2014; date of current version February 13, 2015.
X. Tian, L. Gui, and Q. Li are with the Department of Electronics and
Information Engineering, Huazhong University of Science and Technology,
430074 Wuhan, China (e-mail: tianxx1988@gmail.com; guilq@hust.edu.cn;
qingxia_li@hust.edu.cn).
Y.-J. Zhang and J. Fan are with the Department of Electrical and Computer
Engineering, Electromagnetic Compatibility Laboratory, Missouri University
of Science and Technology (formerly University of Missouri-Rolla), Rolla, MO
65409 USA (e-mail: yaojiang.zhang@gmail.com; jfan@mst.edu).
D. Liu was with the Department of Electrical and Computer Engineering,
Electromagnetic Compatibility Laboratory, Missouri University of Science and
Technology (formerly University of Missouri-Rolla), Rolla, MO 65409 USA. He
is now with Apple Inc., Cupertino, CA 95014, USA (e-mail: dlhbf@mst.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TEMC.2014.2364269

planes may induce noise currents in nearby power/ground vias

and thus cause voltage fluctuations in the PDN. This kind of
simultaneously switching noise (SSN) or power/ground bounce
noise needs to be suppressed to maintain both signal and power
integrities in multilayered PCBs.
A typical way to suppress the SSN is to add decoupling
capacitors between power and ground planes [6]. Due to the
parasitic inductance dominating at higher frequencies, the decoupling capacitors are mainly effective in megahertz range [7].
RF isolation using power islands is another effective approach
to prevent noise from propagating from one area of the PCB
to others [8]. However, dc connections and high-speed traces
crossing the gap are challenges for the gap design. Electromagnetic band-gap (EBG) structures on the power/ground planes
have been extensively explored to prevent noise from propagating between power/ground planes in gigahertz frequencies
[9][11]. Well-designed stopband of periodic EBG structures
can effectively eliminate resonant frequencies of the cavity constructed by the power/ground planes. To fabricate an EBG PDN,
whole or partial power and/or ground planes are etched into a
two-dimensional (2-D) periodic structure. While EBG structures
may provide possible solutions to high-frequency PDN designs,
periodic patterned power/ground planes raise signal integrity
concerns due to the disturbed return current paths for signal
traces. To avoid this drawback of EBG structures deforming
power/ground planes, a photonic crystal power/ground layer
(PCPL) was proposed [12][15]. PCPL structures are formed
by periodically embedding dielectric rods into the original dielectric between power/ground planes. The periodicity of the
rod lattice and the dielectric contrast with the surrounding dielectric material can significantly suppress unwanted noise in
the expected stopbands while not deteriorating signal integrity
of traces.
To analyze the performance of a PCPL structure between
power/ground planes, full-wave solvers such as finite difference
time-domain (FDTD) and finite-element method (FEM) are usually adopted [12][15]. These solvers are flexible in dealing
with inhomogeneous materials and arbitrarily shaped geometries. However, these mesh-based numerical solvers are usually
time consuming. Recently, contour integral method (CIM) was
applied in analysis of PCPL EBG structures [16]. In this method,
both the boundary edges of power/ground planes and circular
circumferences of rods are discretized into many segments and
circular ports are defined along dielectric rods to consider both
the isotropic and anisotropic modes.

0018-9375 2014 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.

136

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO. 1, FEBRUARY 2015

In this paper, the generalized multiple scattering (GMS)

method proposed in [17] for via modeling is extended to power
integrity analysis of a power/ground plane pair loaded with
photonic crystals and decoupling capacitors. Different from
[16], a radial scattering matrix is first extracted without meshing circumferences of either source ports or dielectric rods.
The impedance matrix characterizing the noise coupling among
source ports is obtained by the transformation proposed in [18]
from the radial scattering matrix.
In Section II, the problem and notations are first introduced.
Generally, the parallel-plate impedance matrix, namely, Zpp
herein, is the main parameter used to characterize the power
integrity performance of power/ground planes. As indicated in
[5] and [18], Zpp reflects the self- and mutual coupling among
radial ports of vias, namely, via ports herein. In the extended
GMS method, however, a radial port is also assigned to the center of each circular dielectric rod, noted by rod ports (it will
be shown that decoupling capacitors can be treated similarly as
special rod ports). Both isotropic and anisotropic parallel-plate
modes are used to expand the field distributions near all the
radial via and rod ports. Then, the transition matrix is derived
to relate the expansion coefficients for each rod by enforcing
the boundary conditions along the circumference of each rod.
Furthermore, the transition matrix of a decoupling capacitor is
obtained from physics-based via circuit model [2][4] and the
radial port voltage/current definitions. On the other hand, an
overall radial scattering matrix among all radial via and rod
ports can be obtained through numerical addition theorems of
cylindrical harmonics by the GMS method. Finally, the transition matrices of rods and decoupling capacitors are loaded to the
radial scattering matrix. Through some matrix manipulations,
the radial scattering matrix of only via ports SR
pp can be derived.
Then by the transformation pair of radial scattering matrix SR
pp
and impedance matrix Zpp developed in [18], a Zpp matrix of
only via ports can be calculated.
In Section III, a formula of field distributions for a plate
pair loaded by photonic crystals is given to facilitate the discussions on different resonance mechanisms. Several examples
are provided in the Section IV to demonstrate the accuracy and
efficiency of the extended GMS method by comparing with a
full-wave solver.
II. EXTENDED GMS METHOD

Fig. 1. Via or source ports in an arbitrarily shaped plate pair loaded with
circular dielectric rods and decoupling capacitors: (a) top view and (b) side
view (not in scale).

matrix Zpp . In the algorithm proposed in this paper, the vias,

rods, and decoupling capacitors are all regarded as radial ports,
which have been described in [5], [18]. While the via radial ports
can be viewed as source ports, the radial ports associated with all
rods and decoupling capacitors will be treated as loaded ports.
The total number of loaded ports is Q = Qr + Qc . Therefore,
there are a total of P + Q radial ports in the plate pair. To facilitate the derivation of the algorithm, Ports 1, 2, . . . , P denote the
source ports of vias and Ports P + 1, P + 2, . . . , P + Q denote
the loaded ports of both rods and decoupling capacitors. The
locations and radii of the ports are represented by
i = (i , i )
and ri , i = 1, 2, . . . , P + Q, respectively. r i and r i are used
to represent the relative permittivity and permeability of the ith
rod.
A local polar coordinate system is set up at the center of each
circular radial port and the coordinates of a point are represented
st = (st , st ) is defined as
by
i = (i , i ). The symbol

A. Problem Description and Notations

Consider noise coupling among different ports in an arbitrarily shaped plate pair loaded with Qr circular dielectric rods and
Qc decoupling capacitors as shown in Fig. 1. Port voltages and
currents are defined between top and bottom perfect electric
conductor (PEC) plates. The edges of the plate pair are denoted
as , which can be approximated as either perfect magnetic
conductor (PMC) or perfect electric conductor (PEC) boundary conditions. The separation of top and bottom plates is h,
which is usually electrically small at the frequency of interest.
The relative permittivity and loss tangent of the dielectric layer
between two plates are r and tan , respectively.
Suppose there are P vias located in the plate pair, whose mutual coupling is characterized by a P P plate-pair impedance

s
t |
st = |

(1)

st = arg {
s
t }

(2)

t are the locations of point s and t in the global

where
s and
polar coordinate system, respectively. The vertical electric field
in the plate pair can be expressed in terms of cylindrical waves
in the local coordinate system of the ith port as [18]
Ez(i) =

Mi

m =M i

i
(i) (2)
i
jm
{a(i)
m Jm (k ) + bm Hm (k )}e

(3)

where k = k0 r and k0 is the wavenumber in free space. Jm ()

(2)
and Hm () are the mth-order Bessel and Hankel functions of
second kind, respectively. The truncation number Mi can be
different for different ports. If the ith port is a via or decoupling

TIAN et al.: EFFICIENT ANALYSIS OF POWER/GROUND PLANES LOADED WITH DIELECTRIC RODS AND DECOUPLING CAPACITORS

Fig. 3.
Fig. 2.

Equivalent circuit of a decoupling capacitor in a parallel-plate pair.

Circular dielectric rod in a plate pair.

(i)

H(2) =
capacitor, Mi should be zero as only the zero-order, isotropic
parallel-plate mode needs to be considered in Zpp . On the other
hand, for a dielectric rod, Mi may have to be larger than zero
to include higher order or anisotropic parallel-plate modes of
the field variation along the azimuthal direction of the rod. The
(i)
(i)
expansion coefficients am and bm are arranged into vectors to
facilitate later discussions as

(i)
(i)
a(M i )
b(M i )

(i)
(i)

a
b

(M
(M
+1)
+1)
i
i
(4)
a(i) =
and b(i) =
.

(i)
(i)
a(M i )
b(M i )
Note that the vectors a(i) and b(i) are not independent but
related each other due to reflections of dielectric rods and decoupling capacitors, and also wave transformations among different
coordinate systems in the plate pair.

For each circular dielectric rod and decoupling capacitor, the

expansion coefficient vector a(i) and b(i) related by a (2Mi +
1) (2Mi + 1) diagonal transition matrix T(i) is given as [19]
b(i) = T(i) a(i) .

(5)

For the rod ports, boundary conditions have to be enforced as

shown in Fig. 2. The electric and magnetic fields outside of the
rod i are written as
Mi


i
(i)
i
(i) (2)
i
a(i)
Ez (1) =
J
(k
)
+
b
H
(k
)
ejm
(6)
m
m
m
m
m =M i

H(1) =

Mi

m =M i

k (i) 
i
(2)
i
am Jm (ki ) + b(i)
H
(k
)
ejm .
m
m
jw
(7)

The bold font j denotes the imaginary unit in this paper, and
the electric and magnetic fields inside of the rod can be expressed
as
Mi


(i)
i
jm i
c(i)
(8)
Ez (2) =
m Jm (ki ) e
m =M i

Mi

m =M i

ki (i) 
i
cm Jm (ki i ) ejm
jw

(9)

where the primes on the Bessel and Hankel functions indicate

the differential to the corresponding variable; k = k0 r r and

ki = k0 r i r i are wavenumbers outside and inside of the rod.

The boundary conditions of the continuities of tangential fields
should be enforced at i = ri and lead to
(i)

(i)

(10)

(i)

(i)

(11)

Ez (1) = Ez (2)
H(1) = H(2) .
Substituting (6)(9) into (10) and (11) lead to
(i) (2)
(i)
a(i)
m Jm (kri ) + bm Hm (kri ) = cm Jm (ki ri )

k[a(i)
m Jm (kri )

(2) 
b(i)
m Hm (kri )]
(i)

(12)


ki [c(i)
m Jm (ki ri )].

(i)

(13)

(i)

Then, a transition coefficient Tm = bm /am can be derived

from (12) and (13) as
Tm(i) =

B. Transition Matrix of Dielectric Rods

and Decoupling Capacitors

(i)

137



(ki ri ) kJm
(kri )Jm (ki ri )
ki Jm (kri )Jm
(2)

(2)

 (k r )
kHm (kri )Jm (ki ri ) ki Hm (kri )Jm
i i

. (14)

A decoupling capacitor is usually modeled as a lumped circuit

used in circuit simulators. However, in order to incorporate a decoupling capacitor into the extended GMS method, the reflection
coefficient of radial cylindrical harmonics needs to be derived
first from the lumped circuit of a decoupling capacitor and the
via-plate capacitance in its physics-based via model [2][4]. As
shown in Fig. 3, a decoupling capacitor is loaded between two
plates through two vias. Instead of its own capacitance of C,
parasitic series resistance R, inductance L (including the loop
of interconnects), and the via-plate capacitance Cv extracted
analytically in [4] are also considered in the equivalent circuit
of the decoupling capacitor. Thus, the input impedance for the
equivalent circuit of decoupling capacitors can be expressed by
Zcap =

((1/jC) + jL + R) (1/jCv )
.
(1/jCv ) + ((1/jC) + jL + R)

(15)

On the other hand, the input impedance of decoupling capacitors is also defined as the ratio of radial port voltage and port
current as
Ez (1) h
Vi
Zcap =
=
(16)
Ii
2ri H(1)
where ri is the radius of the via barrel. The fields near the via
(left on in Fig. 3) are also expanded as incident and reflection

138

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO. 1, FEBRUARY 2015

cylindrical waves as (6) and (7). Substituting (6) and (7) into
(16), the input impedance becomes
Zcap =

Mi


(i)

m =M i

(2)

jh Jm (kri ) + Tm Hm (kri )
.

 (kr ) + T (i) H (2) (kr )
2kri Jm
m
m
i
i

(17)

As mentioned before, Mi is set to 0 as the via barrel radius

is electrically small for the decoupling capacitor, and this only
zero-order parallel-plate mode needs to be considered. So the
input impedance can be given in terms of cylindrical waves as
in [5, eq. (38)]
(i)

Zcap =

(2)

jh J0 (kri ) + T0 H0 (kri )
.
2kri J1 (kri ) + T (i) H (2) (kri )
0
1

(18)

Fig. 4. Illustration of ports and edge meshes of an arbitrarily shaped plate pair
for numerical addition theorems by BIE method [17].

Then the transition coefficient of decoupling capacitor can be

expressed as
(i)

T0

Zcap J1 (kri ) + (jh/2kri )J0 (kri )

(2)

(2)

C. Radial Scattering Matrix by GMS Method

. (19)

(jh/2kri )H0 (kri ) + Zcap H1 (kri )

Equation (19) can be validated in several special cases. If

(i)
(2)
Zcap in (19), T0 = J1 (kri )/H1 (kri ), which is the
reflection coefficient for a PMC cylinder, and if Zcap 0 in
(i)
(2)
(19), T0 = J0 (kri )/H0 (kri ), which is the reflection coefficient for a PEC cylinder. In these two cases, (19) is consistent
(i)
with [4, eq. (9)]. On the other hand, when T0 = 0 is used in
(18), the input impedance becomes
Zcap =

jh J0 (kri )
.
2kri J1 (kri )

(20)

Using small argument approximation of the Bessel functions J0 (x)  1, J1 (x)  x/2, when x 0, into (20), the input impedance is further derived as
Zcap =

1
jCp

(21)

where Cp = ri2 r 0 /h is the parallel-plate capacitance of a circular disk, which is reasonable and expected at low frequencies
for a finite circular plate pair. All of the aforementioned discussions for special cases demonstrate the correctness of (19),
which transforms the equivalent circuit of a decoupling capacitor to its transition coefficient of cylindrical harmonics.
Therefore, the diagonal transition matrix in (5) can then be
filled by (14) for dielectric rods and (19) for decoupling capacitors. Furthermore, the transition matrix of Q loaded ports of
both dielectric rods and decoupling capacitors can be obtained
as

Physically, the outward waves expressed by the Hankel functions from the jth port will illuminate the ith port, whose mathematical description is the addition theorem [17]


(2)

Hm  (kj )ejm =

SR (m, m )Jm (ki )ejm

(ij )

(24)

m =

where SR (m, m ) stands for the addition theorem coefficient

j
(2)
from the outward cylindrical harmonics Hm  (kj )ejm  to the
i jm i
i
, and ( , i ) and
regular cylindrical harmonics Jm (k )e
j
j
( , ) are the local polar coordinates of Port i and Port j as
illustrated in Fig. 4, respectively. The addition theorems link the
expansion coefficient vectors in (4) for Port j and Port i as
(ij )

(ij )

a(i) = SR b(j )

(25)

(ij )

where SR is a (2Mi + 1) (2Mj + 1) matrix to transform

the outward cylindrical harmonics of Port j to the regular cylindrical harmonics of Port i. For an infinitely large or finite large
circular plate pair, analytical addition theorems can be used to
(ij )
calculate SR [17], [20], [21]. For an irregular parallel-plate
pair, boundary integral equation (BIE) method was proposed to
obtain the addition theorem coefficient [17]. In the BIE method,
the edge with PMC boundary conditions is discretized into Ne
(ij )
segments as shown in Fig. 4, and SR can be obtained numerically as (detailed derivations are given in [17])
(ij )

SR

(ij )

= SF

+ M(i) (ZM )1 H(j )

(26)

(ij )

bl = Tl al
where Tl is also a diagonal matrix as
(1)
T
0

0
T(2)

Tl = .
..
..
..
.
.

(22)

0
0
..
.
T(Q)

where SF is a (2Mi + 1) (2Mj + 1) matrix, representing

direct illuminations from Port j to Port i, whose elements are
obtained analytically by


SF (m, m ) = Hm m  (kj i )ej (m m ) j i (1 ij )

(ij )

(23)

(2)

(27)

where ij is Kronecker delta symbol and the reflection waves

from the plate-pair edge is considered in the second term
on the right side of (26) where the Ne Ne matrix ZM is
the impedance matrix in the BIE method and its elements are

TIAN et al.: EFFICIENT ANALYSIS OF POWER/GROUND PLANES LOADED WITH DIELECTRIC RODS AND DECOUPLING CAPACITORS

139

transition matrix of dielectric rods and decoupling capacitors

given in (22). Using (22) yields
al = T1
l bl .
Fig. 5. S p p loaded by transition matrix T l of dielectric rods and decoupling
capacitors.

On the other hand, from (32)

al = Slv bv + Sll bl .

obtained by [17]
ZM (t, s) =

(33)

(34)

Then bl can be expressed by bv as

j0.25k ts

(2)
nt H1 (kts ),

0.5ws1 ,

t = s
t=s

bl = Tl (I Sll Tl )1 Slv bv .

(28)

t
s | and ts =
ts /ts . Here, s and t are the
where ts = |
segments of the plate-pair edge , s, t = 1, 2, . . . , Ne . The vecs are defined at the centers of the segments t and s
tors
t and
and ws is the length of segment s. The Ne (2Mj + 1) matrix
H(j ) considers the outward cylindrical harmonics of Port j illuminating the segments of the plate-pair edges and its elements
are obtained by

Substituting (35) into (32) leads to

av = [Sv v + Sv l Tl (I Sll Tl )1 Slv ]bv .

av = SR
pp bv .

t = 1, 2, . . . , Ne .
(29)
On the other hand, M(i) is a (2Mi + 1) Ne matrix, which
transforms the currents on the segments to the radiated fields on
all the ports. Its elements can be calculated by [17]
M(i) (m, s) =

(2)
 Hm
(ksi )
ejm s i
si ns
4j
si

jm (2)
ns
z
Hm (ksi ) .
si

(30)

For a plate pair with P via or source ports and Q loaded ports,
the combined matrix of the addition theorem coefficients can be
obtained from (26) as

(11)
{1N }
(12)
SR
SR
SR p

{2N }
(22)
S (21)
SR
SR p

(31)
Spp =

..
..
.
.
.
.

.
.
.
.

{N 1}
SR p

{(N 2}
SR p

{N N }
SR p p

where Np = P + Q is the total port number. By distinguishing

the via or source ports and the loaded ports, Spp can be expressed
into following submatrices as:
  
 
av
Sv v Sv l
bv
=
(32)
al
Slv Sll
bl
where av , bv and al , bl are the expansion coefficient vectors for
P via or source ports and Q loaded ports, respectively. Clearly,
(32) can be regarded as a linear response of the plate-pair edges
illuminated by the incident waves from all of ports.
D. Zpp of a Plate Pair Loaded by Dielectric Rods
Fig. 5 shows a microwave network of a plate pair with P +
Q radial ports described in (32), which is terminated by the

(36)

According to [18, eq. (23)], the relationship between av and

bv is actually characterized by a radial scattering matrix of the
P via or source ports in a plate pair loaded by Q dielectric rods
and decoupling capacitors as

H(j ) (t, m ) = {Hm  (ktj )ejm t j },

(2)

(35)

(37)

Then, by comparing (36) and (37), SR

pp can be expressed as
1
SR
pp = Sv v + Sv l Tl (I Sll Tl ) Slv .

(38)

Note that for all the vias and decoupling capacitors, the port
radii are via barrel radii, which are so electrically small that the
truncation number Mi is set to be zero. This means only the
zero-order, isotropic parallel-plate modes are considered. Then,
according to the transform pair developed in [18], the impedance
matrix Zpp can be calculated from the radial scattering matrix
SR
pp by
R 1
Zpp = [H0 + J0 SR
pp ][H1 + J1 Spp ]

(39)

where
H0 = diag{H0 (kri )h}

(2)

(40)

J0 = diag{J0 (kri )h}



2kri (2)
H1 (kri )
H1 = diag
j


2kri
J1 (kri ) .
J1 = diag
j

(41)
(42)
(43)

III. VERTICAL ELECTRIC FIELD DISTRIBUTIONS

To study different resonant mechanisms in a plate pair loaded
with dielectric rods and decoupling capacitors, the calculation
of vertical electric field distributions is derived through the GMS
method here. The plate pair area is discretized into Nin + Nout
points. All the points on the parallel plate can be classified
into two categories: Nin points inside the rods and Nout points
outside the rods.
For a point (i , i ) located inside the ith rod, the Ez component can be easily calculated by applying boundary condition

140

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO. 1, FEBRUARY 2015

on the edge of the rods using (6)(11) as

Mi


(i)

(i)

(2)

am Jm (kri ) + bm Hm (kri )
i
Jm (ki i )ejm .
Jm (ki ri )
m =M i
(44)
For all points located outside of any rod, the Ez fields are
the superposition of radiated fields from all of rods and source
ports. In this case, the GMS method still can be used to calculate
the Ez fields outside of rods by regarding all of the Nout points
outside of rods as radial ports with only zero-order parallel-plate
(k j )
modes [18]. Then, instead of (26), a new SR new needs to be
calculated as
Ez =

)
1 (j )
SR new = SF new + M(k
new (ZM ) H .
(k j )

(k j )

(45)

Here, k denote the kth points of the plate domain (outside the
(k j )
rods), k = 1, 2, . . . , Nout , where SF new is a 1 (2Mj + 1)
(k )
matrix whose elements are obtained by (27) and Mnew is a
1 Ne matrix whose elements are obtained by (30). It is worth
mentioning again that each field point here is viewed as a port
considering only zero-order cylindrical harmonics (Mi = 0).
The new SR new can be written in the following matrix:

(11)
{1N p }
(12)
SR new SR new
SR new

{2N p }
(22)
S (21)
SR new

R new SR new
.
Sppnew =

..
..
.
.
.
.

.
.
.
.

(N

1)

out
SR new

{N

2}

out
SR new

{N

N }

out
p
SR new

(46)
Note that Nout is the field point number and Np is the number
of total source, rod, and decoupling ports. From (22) and (32),
the outward wave coefficient vector bl can be expressed by bv
1
bl = (T1
l Sll ) Slv bv

(47)

where bv is the known vector of source ports. If the final port i

is excited, the coefficient b(i) is

0, m = 0
(i)
.
(48)
bm =
1, m = 0
The Ez component of all the points outside the rods and
decoupling capacitors, in [18, eq. (35)], can be calculated as
Eext
= Sppnew b
z

(49)

where bis the total vector of outward wave coefficients of all

the ports
 
bv
b=
.
(50)
bl
IV. NUMERICAL EXAMPLES AND DISCUSSIONS
Here, several examples are provided to validate the extended
GMS method. All the full-wave simulations for validations were
conducted on a machine with an Intel Xeon CPU E5620, four
Tesla C2070 GPUs and 64 GB RAM by a three-dimensional
(3-D) full-wave field simulator based on the finite integration

Fig. 6. Top view of a PCPL structure to suppress noise coupling among three
ports. The lossless dielectric rod has r = 102, and the lossless substrate has
r = 2.33.

technique (FIT) [22], in which simple lumped ports are defined

between top and bottom plates as shown in Fig. 1(b).
Fig. 6 shows a PCPL structure [14], which is formed by two
rectangular plates whose dimensions are 62.5 mm 100 mm
and a dielectric thickness h = 0.8 mm. Forty high dielectric
circular rods are embedded on the lossless dielectric substrate
whose dielectric constant is 2.33. The power/ground metal plates
are modeled as PEC with a thickness of 0.03556 mm. The rod
pitch and the rod radius are 12.5 and 2 mm, respectively. The
height of the rods is the same with the parallel-plate separation.
The relative dielectric constant of the lossless rods is 102. The
origin of the Cartesian coordinate system is located at the bottom
left corner of the parallel plates, and the three via or source ports
are located at P1 (12.5 mm, 25 mm), P2 (37.5 mm, 75 mm), and
P3 (25 mm, 46.5 mm), respectively.
The noise coupling among the source ports can be described
by S-parameters converted from the parallel-plate impedance
matrix of (39). Fig. 7(a) compares the magnitude of S21 by the
extended GMS method with different azimuthal mode number
for dielectric rods. It can be seen that at frequencies lower than
about 3 GHz, only zero-order or isotropic cylindrical harmonics
(M = 0) is good enough to describe the field distribution near
the dielectric rods. This means the dielectric rods can be modeled
as parallel-plate capacitors loaded to the plate pair as indicated
in (20) or (21).
With the increase of frequency, however, higher order or
anisotropic cylindrical harmonics (M 1) must be considered
to correctly model the fields outside and inside the dielectric
rods. In this example, converged |S21 | can be achieved when
the first-order cylindrical harmonics are considered. This means
only six unknowns are required in the extended GMS method
to expand the fields expressed in (3) for each dielectric rod. The
efficiency of the semi-analytical algorithm is expected to be very
high when compared with full-wave solvers. For this example, it
only took about 30 s by a MATLAB code of the extended GMS
method while FIT simulator spent 23447 s (more than 6 h).
Fig. 7(b) compares the converged |S21 | of the extended GMS
method with that by FIT simulator. Very good agreements are

TIAN et al.: EFFICIENT ANALYSIS OF POWER/GROUND PLANES LOADED WITH DIELECTRIC RODS AND DECOUPLING CAPACITORS

141

Fig. 7. (a) Convergence of |S 2 1 | with modes in the extended GMS method;

(b) comparison of |S 2 1 | by the extended GMS and FIT simulator.

observed between these two methods. The main discrepancies

are in the stopband from 3 to 6 GHz. This is because of the spurious rapid resonances in FIT simulations due to the drawback of
the time-domain solver for high-Q resonators formed by photonic crystals. It should be mentioned that these comparisons
could also be quantified using the feature selective validation
(FSV) technique specified in the IEEE standard established in
[23] and [24].
To manifest different stop band mechanisms in a plate pair
loaded with dielectric rods, vertical field distributions are shown
in Fig. 8 at frequencies of 1.0, 4.0, and 6.0 GHz, respectively. At
these three frequencies, the magnitude of S21 are all less than
40 dB. Here, Port 1 is the exciting port. Fig. 8(a) shows the
amplitude and phase distribution of Ez at 1 GHz. It can be seen
that the weak coupling between Port 1 and Port 2 is caused by
the whole plate-pair cavity resonance. The fields are not greatly
disturbed by the loaded dielectric rods. At this low frequency,
all of dielectric rods can be viewed as loaded capacitances to
the power/ground plates. On the other hand, Fig. 8(b) provides
the results at 4 GHz. Clearly, four dielectric rods near Port 1
construct a rectangular cavity. There are many these kinds of
dielectric-rod cavities in the photonic crystal structure to realize

Fig. 8. Electric field distribution at different resonant frequencies: (a) 1 GHz,

(b) 4 GHz, and (c) 6 GHz.

a frequency band gap to prevent noise waves propagating from

Port 1 to Port 2. Further observation on Fig. 8(c) indicates that
at 6 GHz, a single dielectric rod itself becomes a resonator,
thereby preventing noise coupling in the power/ground planes.
The resonant modes of a single dielectric rod have been studied
by mode-matching methods [25][27].
A second example is an irregularly shaped power/ground
plane pair with a lattice of dielectric rods and decoupling capacitors as shown in Fig. 9. The thickness of the dielectric between the two plates is 0.25 mm. The background substrate
has a relative dielectric constant of r = 4.3 and loss tangent of tan = 0.025. The decoupling capacitors are placed
at nearby locations of two ports on the board whose capacitance

142

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO. 1, FEBRUARY 2015

Fig. 9. Top view of a plate pair loaded with dielectric rods and decoupling
capacitors.

is C = 0.1 F, with a series resistance R = 1 m and a parasitic inductance L = 1 nH. The radii of via barrels and via
holes are 0.127 and 0.381 mm, respectively. The calculated viaplate capacitance Cv is 26 fF, which can be safely neglected at
frequencies less than 10 GHz. Ninety-eight dielectric rods are
place at the connection part of the board in order to prevent
the noise from propagating between the left and right sides of
the board. The lossless dielectric rods have a relative dielectric
constant of r = 100. The rod pitch and the rod radius are 7 and
1.4 mm, respectively. The power/ground metal plates are modeled as PEC with a thickness of 0.03556 mm. In the Cartesian
axis system, the origin was placed at bottom left corner of the
parallel plates, and the two lumped ports were located at Port 1
(15 mm, 75 mm) and Port 2 (160 mm, 80 mm).
Fig. 10 compares the mutual impedances between Port 1 and
Port 2 in the plate pair with and without dielectric rods and/or
decoupling capacitors. The black solid line is for the mutual
impedance of two ports in a solid plate pair; the dashed red line
is for the plate pair with dielectric rods but without decoupling
capacitors; and the blue dot line is for the plate pair with both
dielectric rods and decoupling capacitors.
From Fig. 10(a), it can be seen that below 4 GHz, the photonic crystals have no obvious impact on the noise coupling but
lead to a stopband from about 4 to 7 GHz and another from 8.4
to 9.8 GHz. In these two stopbands, the mutual coupling can
normally be reduced by more than 20 dB comparing with that
of the solid plates. On the other hand, the decoupling capacitors have almost no impact on noise suppression at frequencies
higher than 1 GHz.
Fig. 10(b) indicates the decoupling capacitors can effectively
reduce the mutual coupling at frequencies less than 1 GHz. Note
that in this frequency band, the dielectric rods behave like the
decoupling capacitors in that they lower the resonant frequencies by increasing the capacitance between two plates. Therefore, decoupling capacitors and embedded photonic crystals can
be complimentary to mitigate noise coupling in power/ground
planes of a PCB at low and high frequencies.
V. CONCLUSION
GMS method, a semi-analytical approach, is extended from
via modeling to power integrity analysis of power/ground planes

Fig. 10. Mutual impedance from Port 1 to Port 2 for cases of solid plate pair,
with only dielectric rods and with both decoupling capacitors and dielectric
rods. (a) High-frequency behavior. (b) Low-frequency behavior.

loaded with dielectric rods or photonic crystals and decoupling capacitors. The transition matrix of the dielectric rods
and decoupling capacitors are derived by considering the axially anisotropic modes and isotropic modes, respectively. In
addition, the electric field is calculated to distinguish various
resonance mechanisms in a plate pair with photonic crystals. It
is found that there are three kinds of resonant structures in a
plate pair with dielectric rods: plate-pair cavity, cavity formed
by surrounding dielectric rods, and cavity of dielectric rods.
The proposed method has been validated by comparing the results with full-wave simulations. It has been demonstrated that
the method is very efficient and flexible for modeling PCPL
structure and decoupling capacitors in design process of PDNs.
REFERENCES
[1] T.-L. Wu, H.-H. Chuang, and T.-K. Wang, Overview of power integrity solutions on package and PCB: Decoupling and EBG isolation,
IEEE Trans. on Electromagnet. Compat., vol. 52, no. 2, pp. 346356,
May 2010.
[2] R. Rimolo-Donadio, X. Gu, Y. H. Kwark, M. B. Ritter, B. Archambeault,
F. de Paulis, Y.-J. Zhang, J. Fan, H. Bruns, and C. Schuster, Physics
based via and trace models for efficient link simulation on multilayer
structure up to 40 GHz, IEEE Trans. Microw. Theory Techn., vol. 57, no. 8,
pp. 20722083, Aug. 2009.
[3] C. Schuster, Y. Kwark, G. Selli, and P. Muthana, Developing a physical
model for Vias, in Proc. IEC DesignCon Conf., Santa Clara, CA, USA,
Feb. 69, 2006, pp. 124.

TIAN et al.: EFFICIENT ANALYSIS OF POWER/GROUND PLANES LOADED WITH DIELECTRIC RODS AND DECOUPLING CAPACITORS

[4] Y. Zhang, J. Fan, G. Selli, M. Cocchini, and F. D. Paulis, Analytical evaluation of via-plate capacitance for multilayer printed circuit
boards and packages, IEEE Trans. Microw. Theory Tech., vol. 56, no. 9,
pp. 21182128, Sep. 2008.
[5] Y.-J. Zhang and J. Fan, An intrinsic circuit model for multiple vias in
an irregular plate pair through rigorous electromagnetic analysis, IEEE
Trans. Microw. Theory Tech., vol. 58, no. 8, pp. 22512265, Aug. 2010.
[6] M. Swaminathan, J. H. Kim, I. Novak, and J. P. Libous, Power distribution
networks for system-on-package: Status and challenges, IEEE Trans. Adv.
Packag., vol. 27, no. 2, pp. 286300, May 2004.
[7] J. Fan, J. L. Drewniak, J. L. Knighten, N. W. Smith, A. Orlandi, T. P. Van
Doren, T. H. Hubing, and R. E. DuBroff, Quantifying SMT decoupling
capacitor placement in DC power bus design for multi-layer PCBs, IEEE
Trans. Electromagn. Compat., vol. 43, no. 4, pp. 588599, Nov. 2001.
[8] J. Fan, Y. Ren, J. Chen, D. M. Hockanson, H. Shi, J. L. Drewniak,
T. H. Hubing, T. P. Van Doren, and R. E. DuBroff, RF isolation using power islands in DC power bus design, presented at the IEEE Int.
Symp. Electromagn. Compat., Seattle, WA, USA, Aug. 1999.
[9] T.-K. Wang, T.-W. Han, and T.-L. Wu, A novel Power/Ground layer using artificial substrate EBG for simultaneously switching noise suppression, IEEE Trans. Microw. Theory Tech., vol. 56, no. 5, pp. 11641171,
May 2008.
[10] P. H. Rao and M. Swaminathan, A novel compact electromagnetic
bandgap structure in power plane for wideband noise suppression and
low radiation, IEEE Trans. Electromagn. Compat., vol. 53, no. 4,
pp. 9961004, Nov. 2011.
[11] F. de Paulis, L. Raimondo, and A. Orlandi, Impact of shorting vias
placement on embedded planar electromagnetic bandgap structures within
multilayer printed circuit boards, IEEE Trans. Microw. Theory Tech.,
vol. 58, no. 7, pp. 18671876, Jul. 2010.
[12] A. Scogna, T.-L. Wu, and A. Orlandi, Noise coupling mitigation in
PWR/GND plane pair by means of photonic crystal fence: Sensitivity
analysis and design parameters extraction, IEEE Trans. Adv. Packag.,
vol. 33, no. 3, pp. 574581, Aug. 2010.
[13] G.-Z. Wu, Y.-C. Chen, and T.-L. Wu, Design and implementation of a
novel hybrid photonic crystal power/ground layer for broadband power
noise suppression, IEEE Trans. Adv. Packag., vol. 33, no. 1, pp. 206211,
Feb. 2010.
[14] A. Scogna, A. Orlandi, V. Ricchiuti, and T.-L. Wu, Impact of photonic
crystal power/ground layer density on power integrity performance of
high-speed power buses, in Proc. IEEE Int. Symp. Electromagn. Compat.,
2009, pp. 8085.
[15] T.-L. Wu and S.-T. Chen, A photonic crystal power/ground layer for
eliminating simultaneously switching noise in high-speed circuit, IEEE
Trans. Microw. Theory Tech., vol. 54, no. 8, pp. 33983406, Aug. 2006.
[16] J. B. Preibisch, X. Duan, and C. Schuster, An efficient analysis of
power/ground planes with inhomogeneous substrates using the contour
integral method, IEEE Trans. Electromagnet. Compat., vol. PP, no. 99,
pp. 110, 2013.
[17] Y.-J. Zhang and J. Fan, A generalized multiple scattering method for
dense vias with axially anisotropic modes in an arbitrarily shaped plate
pair, IEEE Trans. Microw. Theory Tech., vol. 60, no. 7, pp. 20352045,
Jul. 2012.
[18] Y.-J. Zhang, G. Feng, and J. Fan, A novel impedance definition of a
parallel plate pair for an intrinsic via circuit model, IEEE Trans. Microw.
Theory Tech., vol. 58, no. 12, pp. 37803789, Dec. 2010.
[19] W. C. Chew; L. Gurel; Y.-M. Wang; G. Otto, R. L. Wagner, and
Q. H. Liu, A generalized recursive algorithm for wave-scattering solutions in two dimensions, IEEE Trans. Microw. Theory Tech., vol. 40, no. 4,
pp. 716723, Apr. 1992.
[20] E. D. Caballero, H. Esteban, A. Belenguer, and V. Boria, Efficient analysis
of substrate integrated waveguide devices using hybrid mode matching
between cylindrical and guided modes, IEEE Trans. Microw. Theory
Tech., vol. 60, no. 2, pp. 232243, Feb. 2012.
[21] L. Tsang and D. Miller, Coupling of vias in electronic packaging and
printed circuit board structures with finite ground plane, IEEE Trans.
Adv. Packag., vol. 26, no. 4, pp. 375384, Nov. 2003.
[22] CST Studio Suite. (2013). [Online]. Available: https://www.cst.com/
[23] A. P. Duffy, A. J. M. Martin, A Orlandi, G Antonini, T. M. Benson,
and M. S. Woolfson, Feature selective validation (FSV) for validation of
computational electromagnetics (CEM). Part IThe FSV method, IEEE
Trans. Electromagn. Compat., vol. 48, no. 3, pp. 449459, Aug. 2006.
[24] A. Orlandi, A. P. Duffy, B. Archambeault, G. Antonini, D. E. Coleby,
and S. Connor, Feature selective validation (FSV) for validation of
computational electromagnetics (CEM)Part II: Assessment of FSV

143

performance, IEEE Trans. Electromagn. Compat., vol. 48, no. 3, pp.

460467, Aug. 2006.
[25] M. W. Pospieszalski, On the theory and application of the dielectric
post resonator, IEEE Trans. Microw. Theory Tech., vol. MTT-25, no. 3,
pp. 228231, Mar. 1977.
[26] Y. Kobayashi and M. Katoh, Microwave measurement of dielectric properties of low-loss materials by the dielectric rod resonator method, IEEE
Trans. Microw. Theory Tech., vol. MTT-33, no. 7, pp. 586592, Mar. 1977.
[27] B. W. Hakki and P. D. Coleman, A dielectric resonator method of measuring inductive capacities in the millimeter range, IRE Trans. Microw.
Theory Tech., vol. MTT-8, no. 4, pp. 402410, Jul. 1960.

Xinxin Tian is working toward the Ph.D. degree in

electrical engineering from Huazhong University of
Science and Technology (HUST), Wuhan, China.
She is currently a Visiting Student with the Electromagnetic Compatibility Laboratory, Missouri University of Science and Technology (Missouri S&T),
Rolla, USA. Her research interests include electromagnetic interference and compatibility and signal/power integrity issues in high-speed PCBs.
Dr. Tian is a recipient of Best Student Paper Award
in 2014 IEEE International Symposium on Electromagnetic Compatibility.

Yao-Jiang Zhang (S97M01SM11) received the B.E. and M.E. degrees

in electrical engineering from University of Science and Technology, China in
1991 and 1994, respectively. He received the Ph.D. degree in physical electronics from Beijing (Peking) University in 1999. He was with the Tsinghua
University as a post-doctoral research fellow from 1999 to 2001. He was a
senior research engineer and research scientist in Institute of High Performance
Computing (IHPC), Agency for Science, Technology and Research (A*STAR),
Singapore from Aug. 2001 to Aug. 2006 and from Sep. 2008 to Apr. 2010, respectively. He worked for the EMC laboratory, Missouri University of Science
and Technology (Missouri S&T, formerly University of Missouri-Rolla) from
Sep.2006 to Sep. 2008.
He is currently working as an associate research professor. His research interests include computational electromagnetics, electromagnetic interference and
compatibility, signal/power integrity issues in high-speed electronic packages
or PCBs, photonic crystal and nano-photonic devices.

Dazhao Liu received the B.S. degree in electrical engineering from Tsinghua
University, Beijing, China, in 2008, and the M.S. and the Ph.D. degrees in electrical engineering from Missouri University of Science and Technology (formerly
University of Missouri-Rolla), Rolla, USA, in 2010 and 2013, respectively.
He is currently with Apple Inc., Cupertino, CA, USA. His current research
interests include electrostatic discharge, advanced RF measurement, method of
moment, and signal integrity.

Liangqi Gui received the B.S. degree in electrical engineering from Wuhan University, Wuhan, China, in
1998, and the Ph.D. degree in electrical engineering
from Huazhong University of Science and Technology (HUST), Wuhan, China, in 2005.
He is currently an Associate Professor in the
Science and Technology on Multi-Spectral Information Processing Laboratory, Department of Electronics and Information Engineering, HUST. His current
research interests include wireless communication,
microwave remote sensing, EMC, and passive THz
imaging technology.

144

IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 57, NO. 1, FEBRUARY 2015

Qingxia Li (M08) received the B.S., M.S., and Ph.D.

degrees in electrical engineering from Huazhong University of Science and Technology (HUST), Wuhan,
China, in 1987, 1990, and 1999, respectively.
He is currently a Professor in Science and Technology on Multi-Spectral Information Processing Laboratory, Department of Electronics and Information
Engineering, HUST. His current research interests
include microwave remote sensing and deep space
exploration, electromagnetic theory and application,
antenna array, and signal processing.

Jun Fan (S97M00SM06) 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, USA, in 2000.
From 2000 to 2007, he was a Consultant Engineer with NCR Corporation,
San Diego, CA, USA. In July 2007, he joined the Missouri University of Science
and Technology (formerly University of Missouri-Rolla), where he is currently
an Assistant Professor with the Missouri S&T EMC Laboratory. His current
research interests include signal integrity and EMI designs in high-speed digital
systems, dc power-bus modeling, intra-system EMI and RF interference, PCB
noise reduction, differential signaling, and cable/connector designs.
Dr. Fan was the Chair of the IEEE EMC Society TC-9 Computational Electromagnetics Committee from 2006 to 2008, and was a Distinguished Lecturer
of the IEEE EMC Society in 2007 and 2008. He is currently the Vice Chair of the
Technical Advisory Committee of the IEEE EMC Society. He is the recipient
of the IEEE EMC Society Technical Achievement Award in August 2009.

Copyright of IEEE Transactions on Electromagnetic Compatibility is the property of IEEE

and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.