Received August 11, 2018, accepted October 2, 2018, date of publication November 9, 2018, date of current version December

3, 2018.
Digital Object Identifier 10.1109/ACCESS.2018.2876886

Modeling of Multilayered Power/Ground Planes

Based on Resonant Cavity Algorithm
JUN WANG , JIANMIN LU, XIUQIN CHU , YANG LIU, AND YUSHAN LI
Key Laboratory of High-Speed Circuit Design and EMC, Ministry of Education, Xidian University, Xi’an 710071, China
Corresponding author: Jianmin Lu ([email protected])
This work was supported in part by the National Natural Science Foundation of China under Grants 61871453 and 61501345 and in part by
the Open Funds of the Key Laboratory of High-Speed Circuit Design and EMC under Grants 2017KFKT/B10 and 2017KFKT/B03.

ABSTRACT Based on the resonant cavity algorithm, a method is presented to model the multilayered
power/ground planes with stitching vias at the high-speed packages and the printed circuit board.
A multilayered structure can be seen as a combination of two basic units: adjacent planes with a different
reference voltage and that with the same reference voltage. Then, the behaviors of the switching current
in and the physics-based models for the multilayered power/ground planes with different basic units are
presented. By analyzing the influence of stitching vias to the return path of the switching current in the
multilayered structure, it is obtained that the remainder plane pairs act as a multiport load to the plane pair
where the current load is located on. The algorithm for a plane pair with a multiport load is presented on the
basis of the resonant cavity algorithm to efficient mode the multilayered power/ground planes. At the same
time, this modeling method can be used in auto-decoupling design to consider the placements of decoupling
capacitors. The modeling method is corroborated by the comparison with a 3-D electromagnetic commercial
tool with wideband Z-parameter computation from dc to 10 GHz.

INDEX TERMS Multilayer, power distribution network, power integrity, resonant cavity model.

I. INTRODUCTION through-hole (PTH) vias. Due to the discontinuity of the

In modern high speed digital circuits, the design of power return path of signals’ return current, the return current will
distribution network (PDN) is becoming a challenging task to be coupled into the power/ground plane pairs, which may
supply of clean power to the switching circuits. The process cause a resonant phenomenon in the power/ground plane
dimension of integrated circuits (IC) has been aggressively pairs. In order to mitigate the noise coupled from the signal
reduced to yield smaller transistors, higher integration den- lines, the planes with the same reference voltage are stitched
sity, and faster operating speed. As a result, the simultane- by vias, which can provide a low-impedance path to the return
ous switching noise (SSN) induced by a large number of current [3].
simultaneously switching circuits and the parasitic induc- Another important way to mitigate SSN is adding capac-
tance of power/ground planes is becoming more and more itors to PDN [4]–[7]. Capacitors work as a local charge
serious [1]. It may affect sensitive circuits and its peak value library to provide charges to switching circuits in time.
may even cause logic failure. The power/ground planes play Hence a good decoupling design becomes very important to
an important role in PDN to mitigate the SSN [2]. In the ensure a robust power distribution network in a high speed
time domain, the power/ground planes not only perform as a package/PCB. In the decoupling design, the power distribu-
high-frequency capacitor to provide charges to current loads tion network is first modeled from the IC power-supplying
in time, but also provide a return path for the current of port to the voltage regulator modules (VRM) on the printed
signal lines. In the frequency domain, it is efficient to keep circuit board [5]. Then, according to the constraint of target
the impedance of PDN maintaining to a lower value over a impedance, appropriate capacitors are chosen to reduce the
wide frequency range. With the digital system’s complexity self-impedance of IC power-supplying port. But, the selec-
increasing, the PDN is designed as a multilayered one in tion of decoupling capacitor is very complex. As a result,
the printed circuit board (PCB) and packages. Transmission several auto-decoupling design methods were presented, such
lines residing in different layers are interconnected by plated as the big ‘V’ [4], decade methods [5] and swarm intelligence

2169-3536 2018 IEEE. Translations and content mining are permitted for academic research only.
67360 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 6, 2018
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
J. Wang et al.: Modeling of Multilayered Power/Ground Planes

algorithms (Genetic Algorithm [6], Particle Swarm optimiza- arrangements of the power/ground planes in a layer stack with
tion, Cuckoo Search Method, and Firefly Algorithm [7]). mixed reference planes are divided into two basic units: adja-
In these auto-decoupling design methods, the decoupling cent planes with different reference voltage and that with the
design is a continuous iterative process until the self- same reference voltage. Then, the return path of the switching
impedance of the power-supplying port less than the tar- current in the multilayered structure where some adjacent
get impedance in the frequency of interest. The model for planes with different reference voltage is analyzed. On the
the PDN not only should have a high accuracy, but also basis of the return current path, the physics-based model for
can be easily used to do numerical calculation. In previ- the multilayered structure is presented. Through analyzing
ous work, power/ground planes were modeled as lumping the physics-based model, It is obtained that the remainder
capacitive elements in the decoupling design [8]. However, plane pairs act as a multiport load to the plane pair where
as clock frequency increases to gigahertz, the behaviors of the current load is located on. Next, the modeling method for
the power/ground plane pairs that support the propagation the power/ground plane pair with multiport loads is presented
of radial waves become more complex [9]. The lumping to model the multilayered structure. In order to simplify
model becomes insufficient to characterize the features of the the derivation process, the multiport load is replaced by
power/ground planes at high frequency. In addition, for the a π -type circuit model to deduce the relationship between the
lumped PDN model, it is impossible to consider the influence voltages across the ports of the multiport load and the currents
of decoupling capacitors’ placements to the performance flowing into the load’s ports. Finally, several examples are
of PDN. Hence, a precise model of power/ground planes presented to validate the accuracy of this modeling method
that can exactly characterize its high-frequency performance through the feature selective validation (FSV) method. This
becomes more important. method not only has a high accuracy, but also had a high
Nowadays, many methods have been presented to model calculation speed. More importantly, this modeling method
the power/ground planes [10]–[13], such as partial element can be used as a basic algorithm in the auto-decoupling
equivalent circuit (PEEC), multilayer finite element method design. The advantage of this modeling method is that it can
(MFEM), Method of Moment (MOM) and Finite Differ- consider the influence of decoupling capacitor’s placements
ence Time Domain (FDTD). Although these methods can in the auto-decoupling design. However, due to the limit of
be applied exactly to model the power/ground planes, they the resonant cavity algorithm [9], the modeling method is
are too time-consuming. The theoretical expression of the incapable of dealing with the structure of completely arbitrary
resonant cavity model is one of the fasted methods obtained shape.
by solving the 2-D Helmholtz wave equation for a rect- This paper is organized as follows. In Section II, the return
angular plane pair [14]. The accuracy and the efficiency path of switching current in a multilayered power/ground
of this algorithm has been discussed by correlating with planes is analyzed and the physics-based circuit model for
S-parameter measurements [15], [16]. In addition, several multilayered power/ground planes is presented. In Section III,
works [17]–[19] has been made to improve the poor con- the modeling method for the power/ground planes with mul-
vergence of the resonant cavity algorithm due to its double- tiport loads is presented on the basis of the resonant cavity
infinite series. Another drawback of this algorithm is that it theory. In Section IV the accuracy of the modeling method
can only be applied to the regular power/ground planes with- is validated by several cases. In Section V, some specific
out load circuits. In [20], the closed theoretical expression for conclusions are summarized.
the power/ground planes with single-port load circuits was
deduced which extends the application scope of the resonant II. PHYSICS-BASED MODEL FOR MULTILAYERED
cavity algorithm. By being combined with the segmentation POWER/GROUND PLANE
method and the inverted composition method, the applica- According to different design requirements, there are
tion of the resonant cavity algorithm is extended to model different arrangements of power and ground planes in a
the irregular shaped power/ground planes with single-port multilayered stack. A multilayered structure can be seen
loads [20], [21]. But the modeling method in [20] is limited to as a combination of two basic units: adjacent planes with
the simple two-layered structure. In [22], a modeling method different reference voltage and that with the same reference
for the multilayered structure was presented. In this paper, voltage. The return current has different behaviors in the
the multilayered structure is decomposed into many single power/ground planes with different basic units.
plane pairs. The single plane pairs is modeled firstly and
then the equivalent model for the multilayered structure is A. CURRENT BEHAVIORS IN A MULTILAYERED
obtained by connecting all the circuit models of the single STRUCTURE WHERE ADJACENT PLANES WITH
plane pairs. By contrast, the modeling method presented in DIFFERENT REFERENCE VOLTAGE
this paper takes the multilayered power/ground planes as a A simple four-layer structure with stitching vias is shown
whole. in Fig. 1(a). The current drawn or released by the switching
In this paper, the modeling method for the multilay- circuits can be represented as a vertical current source. The
ered power/ground planes with stitching vias is presented power/ground plane pair can be considered as a dielectric
on the basis of the resonant cavity algorithm. Firstly, the resonant cavity with a PEC bottom wall and a PEC top wall.

VOLUME 6, 2018 67361

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 1. Multilayered power/ground planes (a) Multilayered

power/ground planes where adjacent planes with different reference
voltage (b) via capacitance and inductance definition for a via segment.

The propagation of the return current is confined in the region

between planes at frequency above 10MHz because the skin
depth is thinner than the PDN copper planes [9]. However,
in the structure, shown in Fig. 1(a), the return current in one
plane pair will propagate into another plane pair through the
anti-paid of stitching vias. FIGURE 2. Physics-based circuit model for the multilayered structure.
When the vertical current source releases current to
the PDN, the return current spreads out on the inner surfaces
of the first plane and a large part of current return through
the plane-to-plane distributed impedance and the ground via.
A small part of the current will propagate into the second cav-
ity through the power via. A large part of current on the above
surface of the third plane will return to the below surface
of the second plane through the plane-to-plane distributed
impedance, and then snake around and change surfaces on
the same plane. A small part of the current on the above
surface of the third plane will snake around and transmit to
the above surface of the fourth plane. Then, the current return
to the below surface of the second plane through ground
vias and return to the current source through the anti-pad
of power vias. The return path of return current is shown
in Fig. 1(a).
For the physics-based model of the four-layer structure,
as shown in Fig. 1(b), the via barrel and the anti-pad are
modeled by the inductor and capacitor, respectively, and the
plane pairs are modeled by the resonant cavity algorithm.
After connecting all the circuit models, the physics-based
model of the multilayer PDN with stitching vias is extracted
as shown in Fig. 2. The models of power vias and ground
vias are connected by the red solid-line and the black dot-line,
respectively. The ‘via1-4’and ‘ref1-4’ represent the locations
of the vias on the power and ground planes. The anti-pad
parasitic capacitors that are public elements between two
resonant cavities make the electrical characteristics of this
FIGURE 3. Simplified circuit model (a) Circuit model for the two-port
model more complex. network ‘‘Load1’’ (b) Circuit model for the two-port network ‘‘Load2’’
From the analysis of the return path of the switching (c) Simplified circuit model for the physics-based circuit model.
current, it is obtained that, the remainder plane pairs act as
a multiport load to the first plane pair. The third cavities,
ground stitching vias, and the via-stub of power stitching via The number of the load’s port depends on the number of
in Fig. 3(a) can be equivalent to a two-port load, marked as power stitching via. The equivalent circuit model for the
‘‘Load1’’, attached to the second cavity as shown in Fig. 3(b). multilayer PDN in Fig. 2 is shown in Fig. 3(c). The circuit
The number of the load’s port depends on the number of model for the two port network ‘‘Load2’’ is shown in Fig. 3(b)
ground stitching via. The circuit in Fig. 3(b) can also be which contains the second cavity, power stitching, the third
equivalent to a two-port network attached to the first cavity. cavity and ground stitching vias.

67362 VOLUME 6, 2018

J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 4. Multilayered power/ground planes where adjacent planes

with the same reference voltage.

FIGURE 5. Physics-based circuit model for the multilayered structure FIGURE 6. Simplified circuit model (a) Circuit model for the two-port
in Fig. 4. network ‘‘Load1’’ (b) Circuit model for the two-port network ‘‘Load2’’
(c) Simplified circuit model for the physics-based circuit model.

B. PHYSICS-BASED MODEL FOR A MULTILAYERED

STRUCTURE WITH ADJACENT PLANES WITH
THE SAME REFERENCE VOLTAGE in Fig. 6(a). It can be seen as a multiport load for the second
The multilayered PDN in which some adjacent planes with cavity and it is connected to the upper anti-pad capacitors as
the same reference voltage is shown in Fig. 4. The second and shown in Fig. 6(b), which can be simplified as a multiport cir-
the third plane are the adjacent planes with the same reference cuit with a multiport load. The equivalent circuit model for the
voltage. The physics model of the multilayer PDN is shown multilayered PDN in Fig. 4 is shown in Fig. 6(c). It is the same
in Fig. 5. The via barrel and the anti-pad are modeled by as the final equivalent circuit of the structure whose adjacent
the inductor and capacitor, respectively. The models of power planes are with different reference voltage. In a multilayered
vias and ground vias are connected by the red solid-line and structure, it is can be seen the combination of these two basic
the black dot-line, respectively. The plane pairs are modeled units. Through above analysis, it is obtained that the task of
by the resonant cavity method. Since three cavities are con- modeling a multilayered structure can be simplified as that
nected by the anti-pad capacitors of the power stitching via as modeling a power/ground pair with multiport loads.
a whole, the method of cascade network is used to analyze this
multilayered structure. The anti-pad capacitance are divided III. MODELING A POWER/GROUND PLANE WITH
into two parts and each of them is assigned to different plane MULTIPORT LOAD
pair. Take the anti-pad capacitors C3 as an example. These Based on the resonant cavity model, the modeling method
anti-pad capacitors connect the second and third cavity as a for the power/ground planes pair with multiport loads is
whole. The anti-pad capacitors C3 are divided into two parts presented.
C31 and C32. The capacitor C31 is assigned to the second
cavity and the capacitor C32 is assigned to the third cavity. A. RESONANT CAVITY EQUATION
The ratio of the upper capacitance C31 to the lower capaci- The propagation features of the power/ground planes are
tance C32 is equal to the ratio of the lower cavity’s dielectric usually described as the self or transfer impedance. The reso-
thickness to that of the upper cavity. The equivalent model for nant cavity equation can be used to efficiently compute the
the third cavity and the power/ground stitching vias is shown self and transfer impedance of a power/ground plane pair.

VOLUME 6, 2018 67363

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 7. Structure for a simple power/ground plane pair.

Consider the plane structure in Fig. 7, which contains

of two planes of dimensions a × b, separated by a FIGURE 8. Resonant cavity circuit model for power/ground plane pair.
dielectric of thickness d. The dielectric constant is ε.
This structure can be considered as a dielectric resonant
cavity with a perfect electrical conductor (PEC) bottom
wall and a PEC top wall. Since the plane spacing d
is far smaller than λ(the wavelength) and a, b  d,
the TM modes have 0 mode number along the z direc-
tion. The three main field components are Ez, Hx, and Hy.
‘‘E’’ is the electric field; ‘‘H’’ is the magnetic field; the
subscripts represent the field direction. Because the bound-
ary conditions along four sidewalls of the resonant cavity
FIGURE 9. Equivalent π -type circuit model for a n-port network.
are assumed to be perfect magnetic conductor boundaries,
the impedance matrix of a power/ground plane pair at two
arbitrary ports can be computed by the resonant cavity
algorithm: where
√
q
jωµd
∞ X
∞
εn2 εm
2 f (x , y , x , y ) fmn = (m/a)2 + (n/b)2 /2 εµ (7)
i i j j
X
Zij (ω) = 2 − k2
(1) mπ xi mπ txi nπyi nπ tyi
ab
n=0 m=0
kmn Nmni = εm εn cos sinc cos sinc (8)
a 2a b 2b
mπ xj mπ txj nπyj nπ tyj
where d is the thickness of the dielectric between the two Mmnj = εm εn cos sinc cos sinc (9)
planes; a and b represent the plane dimensions; m and a 2a b 2b
√n εab
are propagating modes; εn , εm = 1, for n, m = 0 and 2, Cmn = (10)
otherwise; d
d
Lmn = (11)
f (xi , yi , xj , yj ) εab(2π fmn )2
√
mπ xi mπ txi nπ yi nπ tyi 2π fCmn (1 + d tan δ π f µσ )
= cos( )sinc( ) cos( )sinc( ) Gmn = √ (12)
a 2a b 2b d π f µσ
mπ xj mπ txj nπyj nπ tyj The resonant cavity circuit model of the power/ground
× cos( )sinc( ) cos( )sinc( ) (2)
a 2a b 2b planes is shown in Fig. 8. Although it has a high efficiency,
 mπ 2  nπ 2 it can be seen from (6) that the resonant cavity equation has
kmn = + (3)
a b a poor convergence due to its double-infinite series. A fast
√ √ algorithm, double-frequency approximation method (DFA)
k = k − jk = ω εµ − j(ω εµ(tan δ + r/d)/2)
0 00
(4)
s in [19], was presented to improve its convergence. In this fast
2 algorithm, the propagating modes whose resonant frequency
r = (5)
ωµσ beyond the frequency range of interest are represented by an
LC parallel circuit. Then, the power/ground planes pair can be
δ is the dielectric loss angle; r is the skin depth; (txi , tyi ) and modeled by the resonant cavity model with a certain number
(txj , tyj ) are the coordinates of the ports; (txi , tyi ) and (txj , tyj ) of propagating modes. In Section II, it is obtained that the
are port dimensions. remainder plane pairs can be seen as a multiport load for the
√
Assume k 00  k 0 and kmn = 2π fmn εµ, Equation (1) can plane pair where the current load is located on. In order to
be rewritten as: simplify the derivation process, the multiport load is replaced
∞ X
X ∞
Nmni Mmnj by a π-type circuit model, as shown in Fig. 9. The expression
Zij (ω) = (6) of the elements in the equivalence π-type circuit model is
jωCmn + 1/jωLmn + Gmn
n=0 m=0 deduced as follows.
67364 VOLUME 6, 2018
J. Wang et al.: Modeling of Multilayered Power/Ground Planes

circuit model is shown as follows:

 
Z11 Z12 · · · Z1n
 Z21 Z22 · · · Z2n 
 .. .. .. .. 
 
 . . . . 
Zn1 Zn2 · · · Znn
 n −1
FIGURE 10. Diagram for a multiport circuit with a multiport load. X 1 1 1
− ··· −
 k=1 Z1,k Z1,2 Z1,n 


n
 
 − 1 1 1 
 X
B. EQUIVALEN π-TYPE CIRCUIT MODEL FOR ··· − 
Z Z Z2,n 
 
A MULIPORT LOAD =
 2,1
k=1
2,k (16)
. . .. ..

The equivalent π -type circuit model for a n-port load is shown 
 .. .. . .


in Fig. 9. The impedance matrix of the n-port network is 
n

 1 1 X 1 
shown as follows. 
− − ···

Zn,1 Zn,2 Zn,k
  k=1
V1
 V2   1 1 1
 The expression of the impedance of the elements in the
Zn−port =  .  ··· π -type circuit model is shown as following equation:
 
 ..  I1 I2 In  
Vn Z1,1 Z1,2 · · · Z1,n
 
 Z2,1 Z2,2 · · · Z2,n 
 
Z11 Z12 · · · Z1n
 . .. .. 
 
 Z21 Z22 · · · Z2n 
 .. ..
= .

.. .. .. 

(13) . . . 
 .. . . . 
 
Zn,1 Zn,2 · · · Zn,n
Zn1 Zn2 ··· Znn
 1 1 1 
where Vi and Ii (i = 1, 2, . . . , n) are the port voltages and n −
Y12
··· −
Y1n 
port currents of the π -type circuit. Through Kirchhoff’s
 P
 k=1 Y1k
 

current law, the relationship between the port currents and 
 1 1 1 

port voltages is obtained as follows:  − ··· − 
 Y21 n
P Y2n 
 Y2k 
 (17)
V1 V1 − V2 V1 − V3 V1 − Vn = k=1
I1 = + + + ··· + .. .. ..
 
 .. 
Z1,1 Z1,2 Z1,3 Z1,n 
 . . . . 

V2 V2 − V1 V2 − V3 V2 − Vn  
I2 = + + + ··· +  1 1 1 
Z2,2 Z1,2 Z1,3 Z1,n  − − ··· 
 Yn1 Yn2 n 
..
P
Ynk
. k=1
Vn Vn − V1 Vn − V2 Vn − Vn−1
In = + + + ··· + (14) where Yi,j (i, j = 1 to n) is the element of the conductance
Zn,n Z1,2 Z1,2 Z1,n−1 matrix of the n-port network.
where the impedance Zi,j is equal to Zi,j (i, j = 1 to n).
Equation (14) can be rewritten as follows after being C. ALGORITHM FOR THE MULTIPORT CIRCUIT
rearranged: WITH A MULTIPORT LOAD
For the multilayered power/ground planes where some adja-
n 
cent planes with the same reference voltage, the circuit model
X 1 1 1
− ··· − for the planes with the same reference voltage can be equiv-
 Z 1,k Z1,2 Z1,n 
    k=1
 
I1 n

V1 alent to a multiport circuit with a multiport load, as shown
 1 X 1 1  in Fig. 6 (b). To deduce the expression of the circuit model
 I2   − ··· −  V2 
 
  
 ..  =  Z2,1 Z2,k Z2,n  .  in Fig. 6 (b), a simplified circuit diagram is shown in Fig. 10.
 .. 
k=1
 .   .. .. .. ..

Assume the load is connected to the multiport circuit’s port k
.

In

 . . .  Vn

to port n. The currents flow into the ports k, k + 1, . . . , n − 1,
n
 1 1 X 1  and n are divided between the circuit model and the load. Let
− − ···
 
Zn,1 Zn,2 Zn,k the multiport circuit be presented as:
k=1
(15) V = ZM IM (18)
The relationship between the impedance matrix of the where the voltage matrix V is the voltage across every point
n-port load and the impedance of the elements in the π -type and V = [V1 , V2 , . . . , Vk , . . . , Vn ]T . The current matrix IM is

VOLUME 6, 2018 67365

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 11. Resonant cavity circuit model for the power/ground plane with single-port loads and a n-port
network.

the current flowing into each port of the n-port circuit model By substituting (20) and (21) into (18), V in (18) can be
and IM = [I1 , I2 , . . . , Ik , . . . , In ]T . The expression of the written as follows.
impedance matrix ZM is shown as follows.
V = ZM I − ZM ∗YL V (22)
 
M
Z11 M
Z12 ··· M
Z1k ··· M
Z1n
 0 0 ··· 0 ··· 0
 M 0 0 ··· 0 ··· 0 
M M M
 Z21 Z22 ··· Z2k ··· Z2n  .. .. . .. ..
 
.

. . . . .

  ···
 .. .. .. .. 

.. YL = 
 
.. (23)
 
 . . . . ··· .  .

   0 0 · · · Ykk Ykn 
ZM . .. .. .. .. 
 
= M .. (19)  ..
 
M M . M . ··· . . . 

Z
 k1 Zk2 ··· Zkk Zkn
0 0 ··· Ynk ··· Ynn

 .. .. .. .. .. 
 
 . . ··· . . . 
  The impedance matrix ZW
for the n-port circuit with a
 M M M M

Zn1 Zn2 ··· Znk ··· Znn multiport load can be written as follows by arranging (22).
ZW = (En×n +ZM ∗ YL )−1 ZM (24)
The relation between the current Ih1 (h = k, k +
1, . . . , n) flowing into the n-port circuit and that Ih2 (h = D. EQUATION FOR POWER/GROUND PLANES WITH
k, k + 1, . . . , n) flowing into the load is shown as A MULTIPORT LOAD
follows: The resonant cavity circuit model for a power/ground plans
pair with a π -type multiport load and single-port load is
Ik1 = Ik − Ik2 , Ik+11 = Ik − Ik+12 , · · · , In1 = In − In2 shown in Fig. 11. The number of the propagating modes of
(20) the resonant cavity circuit model is set as β according to the
DFA fast modeling method. Variables used to deduce the self-
The current flowing into the load can be obtained by the impedance of the circuit model are also shown in Fig. 11.
Y parameter matrix and the port voltages of load, shown as By the Kirchhoff’s current law, the expressions for the
follows. currents I1h (h = 1, 2, · · · , β), flowing into each propagating
mode, are obtained as follows.
    
Ik2 Y11L Y12 L · · · Y1k L  Vk I11 = I21 + I31
 Ik+12  
 
Y21L Y22 L · · · Y2k L  Vk+1 

   I12 = I22 + I32
 ..  =  .. .. ..   ..  (21)
  
.. ..

 .   . .  . 
  . . .

 
L L
Yk1 Yk2 · · · Ykk L
In2 Vn I1β = I2β + I3β (25)

67366 VOLUME 6, 2018

J. Wang et al.: Modeling of Multilayered Power/Ground Planes

According to the property of the ideal transformer, the cur- S

V2S = V12 S
+ V22 + . . . + Vβ2
S

rent I2h and I3h (h = 1, 2, . . . , β) can also be expressed as V1 M12 V2 M22 Vβ Mβ2
follows. = + + ... +
N1 N2 Nβ
Vh ..
I2h = (h = 1, 2, · · · , β) (26) .
Nh Zh
I3h = Mh1 I1S + Mh2 I2S + . . . + Mhα IαS (h = 1, 2, · · · , β) VαS = V1α
S S
+ V2α + . . . + Vβα
S

(27) V1 M1α V2 M2α Vβ Mβα

= + + ... + (33)
N1 N2 Nβ
Then, the expressions for the currents I1h (h = 1,
2, · · · , β) can be rewritten as following equation. The relationship between the current I and voltage Vh
(h = 1, 2, · · · , β) is obtained as follows by Substituting (29)
V1
I11 = + M11 I1S + M12 I2S + . . . + M1α IαS and (33) into (32).
N1 Z1 V 
V2 1
I12 = + M21 I1S + M22 I2S + . . . + M2α IαS   N1
N2 Z2 N1 I 
 V2


..  N2 I   
. N2
 
 ..  = Aβ×β  (34)
 
..

Vβ  .   
I1β = + Mβ1 I1S + Mβ2 I2S + . . . + Mβα IαS (28)  . 
Nβ Zβ Nβ I
 
 Vβ 
where β is the total number of the propagating modes of Nβ
the resonant cavity model and β = (n + 1)(m + 1); Zh , where the parameter matrix Aβ×β is expressed as (35), as
Nh and Mhg (h = 1, 2, · · · , β; g = 1, 2, · · · , α) are the shown at the top of the next page. Then, the expression of
mode impedance and port coefficients of the resonant cavity voltage Vh (h = 1, 2, · · · , β) can be presented as follows.
model. According to the property of the ideal transformer,  Xβ 
the current I1h (h = 1, 2, . . . , β) can also be expressed as N1 I Nj A−1
  1j
follows. V1  Xβj=1 
−1 
 
 V2   N2 I Nj A2j 
I11 = N1 I , I12 = N2 I , · · · , I1β = Nβ I (29)  ..  = 
   j=1 (36)
 .   .. 
.

where I is the current flowing into the port P1. The expression
 
Vβ  Xβ 
of the current IgS , flowing into the single-port loads is shown Nβ I Nj A−1
βj
j=1
as follows.
The self-impedance of port P1 is obtained as following
VgS
IgS = (g = 1, 2, · · · , α; g 6 = i, i + 1, . . . , i + n) (30) equation.
ZgS
β β
V1 + V2 + · · · + Vβ X X
where ZgS
and (g = 1, 2, · · · , α; g 6 = i, i + 1, . . . , i + n)
VgS Z11 = = Ni Nj A−1
ij (37)
I
are the impedance of and the voltage across the gth single- i=1 j=1
port load Sg , respectively. The expression of the current
where A−1ij is the element of the inverse matrix of the
flowing through the π -type multiport load is shown as
coefficient matrix Aβ×β . The trans-impedance between the
follows.
port P1 and the gth load port is expressed as follows.
n n S
VgS − Vi+j
0 VgS
β β
X X
IgS = IgS + S
Ig,i+j = + S
ZgS
X X
j=0 j=0,i6=g
Zg,i+j Z1g = Mig Nj A−1
ij (g = 1, 2, . . . , β) (38)
(g = i, i + 1, . . . , i + n) (31) i=1 j=1

S
where ZgS and Zg,i+j (g = i, i+1, . . . , i+n) are the impedance IV. EXAMPLE VALIDATION
of the elements of the π -type multiport load. Variables IgS , The utilization of the modeling method is discussed based on
0
IgS , and VgS (g = i, i + 1, . . . , i + n) are shown in Fig. 11. the multilayered power/ground planes with through or blind
Equation (28) can be rewritten as (32), as shown at the stitching vias.
top of the next page, by substituting (30) and (31) into it.
By Kirchhoff’s voltage law, the expressions for the voltages A. MULTILAYERED STRUCTURE WITH THROUGH VIAS
across the ports of each single-port load and the π -type n-port Consider a simple four-layer power distribution network with
network are obtained as follows. power and ground stitching vias as shown in Fig. 12. The
adjacent planes are with different reference voltage. It con-
V1S = V11
S S
+ V21 + . . . + Vβ1
S
tains of three plane pairs of dimensions 60 mm × 40mm
V1 M11 V2 M21 Vβ Mβ1 separated by a dielectric of permittivity 4.4. There are four
= + + ... +
N1 N2 Nβ stitching vias located at (30 mm, 25mm), (34 mm, 25mm),

VOLUME 6, 2018 67367

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

n ViS − Vi+j S n S − VS
 S
V1 M11 V1S M12 V2S ViS X Vi+n X Vi+n i+j
. . .

I11 =
 + S
+ S
+ + M 1i ( S
+ S
) + · · · + M 1i+n ( S
+ S
)
N1 Z1



 Z1 Z2 Zi,i j=0,j6=g Zi,i+j Zi+n,i+n j=0,j6=n Zi+n,i+j

M1β VαS


+... +



ZαS




S S S n ViS − Vi+j S S n S − VS
Vi+n

Vi+n
I12 = V2 + M21 V1 + M22 V2 + . . . + M2i ( Vi +

i+j
 X X
) + · · · + M ( + )

 S S S S 2i+n S S

 N Z
2 2 Z 1 Z2 Z i,i Z i,i+j Zi+n,i+n Zi+n,i+j

 j=0,j6=g j=0,j6=n
M2β VαS
 +... +

 ZαC
.


..






Mβ1 V1S Mβ2 V2S ViS
n ViS − Vi+j S S n S − VS
Vi+n

 Vβ X Vi+n X i+j
. . .

I = + + + + M ( + ) + · · · + M ( + )

1β βi βi+n



 Nβ Zβ Z1 S Z2 S S
Zi,i j=0,j6=g Zi,i+j S S S
Zi+n,i+n j=0,j6=n Zi+n,i+j



Mββ VαS


+... +



ZαS
(32)

 α M2 i+n n α i+n n α i+n n 

1 X 1g
X X M1k − M1i+j X M1g M2g X X M1k − M1i+j X M1g Mβg X X M1k − M1i+j
+ + M 1k S
+ M2k S
··· + Mβk S
 Z1 g=1 ZgS ZgS ZgS
 
k=i j=0
Zk,i+j g=1 k=i j=0
Zk,i+j g=1 k=i j=0
Zk,i+j 
 
 X α i+n n α M2 i+n n α i+n n 
 M2g M1g X X M2k − M2i+j 1 X 2g
X X M2k − M2i+j X M2g Mβg X X M2k − M2i+j 
 + M1k S
+ + M2k S
··· + Mβk S

 ZgS Zk,i+j Z2 S
Zg Zk,i+j ZgS Zk,i+j 
Aβ×β =  g=1
 k=i j=0 g=1 k=i j=0 g=1 k=i j=0 
. . .


. . .. .

. . . .
 
 
 α i+n n α i+n n α M2 i+n n

 XM M
βg 1g
X X Mβk − Mβi+j X Mβg M2g X X Mβk − Mβi+j 1 X βg
X X Mβk − Mβi+j 

S
+ M1k S S
+ M2k S
··· + S
+ Mβk S

Zg Zk,i+j Zg Zk,i+j Zβ Zg Zk,i+j
g=1 k=i j=0 g=1 k=i j=0 g=1 k=i j=0

(35)

(32 mm, 21 mm), and (36 mm, 21 mm), respectively. The commercial tool, High Frequency Structure Simulator
radii of the via and the anti-pad are 0.3 mm and 0.6 mm, (HFSS), are shown in Fig. 13.
respectively. Other parameters are shown in Fig. 12. The cur- It shows that the accuracy of (37) is improved as the num-
rent drawn by switching circuits from the power distribution ber of propagating modes used to model the power/ground
network is represented by a vertical current source that is plane pair increases. The method, feature selective validation
located at (23 mm, 21 mm). The current in the first cavity (FSV), is used to compare the self-impedance computed
flows into the second and the third resonant cavity through by (37) with that simulated by HFSS to see how similar
the anti-pad of the power and ground stitching vias. The the two data sets are [23]–[25]. The FSV method comprises
cylindrical wave excited by the current source will cause a three component measures: the amplitude difference mea-
resonance in the second and third cavities, which influences sure (ADM) which is to compare amplitude, the feature
the characteristics of the first cavity. difference measure (FDM) which is to compare the rapidly
The physics-based circuit model for the structure changing feature, and the global difference measure (GDM)
in Fig. 12 is the same as that shown in Fig. 2. The via-stubs which is combined by ADM and FDM. FSV interpretation
of power stitching vias, via1 and via 2, in the third cavity for point-by-point ADM/FDM/ GDM value and six-point
structure can be equivalent to single-port loads connected to scale is shown in Table 1. The trends of the point-by-point
the third cavity structure. The via-stubs of ground stitching ADM/FDM/GDM values with the number of the propagating
vias, via 3 and via 4, in the first cavity structure can also be mode are shown in Fig. 14. With the number of propagating
equivalent to single-port loads. The parasitic parameters of modes used in (37) increasing, it can be observed that the
stitching vias, such as anti-pad capacitance and via barrel accuracy of (37) is improved rapidly.
inductance, are extracted by ANSYS Q3D Extractor. The Good agreement can be seen for the self-HFSS to see how
resonant cavity algorithm is used to compute the impedance similar the two data sets are [23]–[25]. The FSV method com-
of the power/ground plane. The self-impedances of port P1 prises three component measures: the amplitude difference
in the frequency domain computed by (37) with differ- measure (ADM) which is to compare amplitude, the feature
ent number of propagating modes and that simulated by difference measure (FDM) which is to compare the rapidly

67368 VOLUME 6, 2018

J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 14. Grade-spread chart for the self-impedance of port P1.

FIGURE 12. Multilayered structure where adjacent planes with different

reference voltage (a)parameters of structure (b) structure in HFSS.

FIGURE 15. Multilayered Structure where some adjacent plane with the
same reference voltage.

mode are shown in Fig. 14. With the number of propagat-

ing modes used in (37) increasing, it can be observed that
the accuracy of (37) is improved rapidly. Good agreement
can be seen for the self-impedance both in magnitude and
phase. When the number of propagating modes is (60∗45),
the ADM value even reaches to the ‘‘EXCELLENT’’ level.
However, It is also can be seen that, from Fig. 13, the error
between the two data sets becomes lager at the frequency
range higher than 7GHz. It is the influence of the bandwidth
of the lumped LC equivalent circuit in the DFA algorithm
that results to the variation of the accuracy at high frequency.
FIGURE 13. Self-impedance of port P1 on the multilayered structure. The circuit structure where some adjacent planes are with the
same reference voltage is shown in Fig. 15. The second and
TABLE 1. FSV interpretation scales.
third plane are with the same reference voltage. Its physics-
based circuit model is the same as that shown in Fig. 5. The
structural parameters are the same as that in Fig. 12. The
impedance of the multiport load consisting of the stitching
vias and the second and third cavities is computed by (24).
The self-impedances of port P1 in frequency domain com-
puted by (37) with (60∗45) propagating modes and that sim-
ulated by HFSS are shown in Fig. 16. Good agreement can be
seen from the comparison of the self-impedances. Due to the
influence of the bandwidth of the LC equivalent circuit in the
DFA algorithm, the accuracy of the algorithm varies at high
changing feature, and the global difference measure (GDM) frequency.
which is combined by ADM and FDM. FSV interpretation
for point-by-point ADM/FDM/ GDM value and six-point B. MULTILAYERED STRUCTURE WITH BLIND VIAS
scale is shown in Table 1. The trends of the point-by-point Consider a simple five-layered power distribution net-
ADM/FDM/GDM values with the number of the propagating work shown in Fig. 17 where the adjacent planes with

VOLUME 6, 2018 67369

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

FIGURE 16. Self-impedance of port P1 on the multilayered structure FIGURE 18. Self-impedance of port P1 on the multilayered structure with
where some adjacent planes with the same voltage. blind stitching vias.

FIGURE 17. Multilayered Structure with blind power stitching vias.

different voltage. It contains of three ground planes and two

power planes of dimensions 60 mm × 40mm separated by
a dielectric of permittivity 4.4. There are four stitching vias
located at (30 mm, 25 mm), (34 mm, 25 mm), (38 mm, FIGURE 19. Locations of the Decoupling Capacitors Attached to the Plane
Pair.
25 mm), (32 mm, 21 mm), (36 mm, 21 mm), and (40 mm,
21 mm), respectively. The current drawn by the switching
circuits from the power distribution network is represented mitigate power noise efficiently through providing charges to
by a vertical current source and located at (23 mm, 21 mm). the current load. As clock frequency increases, it becomes
Other parameters are also shown in Fig. 17. more and more important to help designers to understand
For this five-layered power distribution network, it is the the effect of decoupling capacitor and do some decoupling
combination of these two basic units: adjacent planes with design efficiently. The purpose of this modeling method is to
different reference voltage and that with the same refer- provide a basic algorithm for the auto-decoupling design for
ence voltage. Since the power stitching vias are blind via, a multilayered PDN, which can consider the placements of
in the physics-based circuit model, there is only a multiport decoupling capacitors.
load connected to the first cavity. When computing the self- A test case is developed to check the accuracy of this
impedance of port P1, the current flowing into the single- method to model the multilayered power/ground planes
port load in (32) should be ignored. The self-impedance of with decoupling capacitors. Ten decoupling capacitors are
port P1 computed by (37) with (60∗45) propagating modes attached to the first cavity around the port 1. Each decoupling
and that simulated by HFSS are shown in Fig. 18. Good capacitor has a capacitance of 0.39 nF with 0.52 nH parasitic
agreement can be seen from the comparison of the self- inductance and 0.5 Ohm parasitic resistance. Their locations
impedances. However, due to the influence of the bandwidth on the first cavity are shown in Fig. 19. The impedance
of the LC equivalent circuit in the DFA algorithm, the accu- of port P1 is shown in Fig. 20. The comparison with the
racy of the algorithm varies at high frequency. result simulated by HFSS illustrates the modeling method’s
accuracy for the multilayered structure with capacitors.
C. APPLICATION OF THIS MODELING METHOD IN Compared with the self-impedance of the multilayered
DECOUPLING DESIGN structure without decoupling capacitors, suppression of the
In high-speed circuit, decoupling capacitors play an impor- resonant peaks in the multilayered structure with decoupling
tant role in the power distribution network, which can capacitors because of the capacitive nature of the capacitors

67370 VOLUME 6, 2018

J. Wang et al.: Modeling of Multilayered Power/Ground Planes

[4] S. Weir, ‘‘Bypass filter design considerations for modern digital systems, a
comparative evaluation of the big ‘V,’ multi-pole, and many pole bypass
strategies,’’ in Proc. Design Conf., Santa Clara, CA, USA, Feb. 2006,
pp. 1608–1630.
[5] L. D. Smith, R. E. Anderson, D. W. Forehand, T. J. Pelc, and T. Roy,
‘‘Power distribution system design methodology and capacitor selection
for modern CMOS technology,’’ IEEE Trans. Adv. Packag., vol. 22, no. 3,
pp. 284–291, Aug. 1999.
[6] K. Bharath, E. Engin, and M. Swaminathan, ‘‘Automatic package and
board decoupling capacitor placement using genetic algorithms and
M-FDM,’’ in Proc. 45th ACM/IEEE Design Autom. Conf. (DAC),
Anaheim, CA, USA, Jun. 2008, pp. 560–565.
[7] J. N. Tripathi, N. K. Chhabra, R. K. Nagpal, R. Malik, and J. Mukherjee,
‘‘Damping the cavity-mode anti-resonances’ peaks on a power plane by
swarm intelligence algorithms,’’ in Proc. IEEE Int. Symp. Circuits Syst.
(ISCAS), Seoul, South Korea, May 2012, pp. 361–364.
[8] J. Kim, K. Shringarpure, J. Fan, J. Kim, and J. L. Drewniak, ‘‘Equiv-
alent circuit model 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.
FIGURE 20. Self-impedance of port 1 on the multilayered structure with [9] N. Na, J. Choi, S. Chun, M. Swaminathan, and J. Srinivasan, ‘‘Modeling
ten decoupling capacitors. and transient simulation of planes in electronic packages,’’ IEEE Trans.
Adv. Packag., vol. 23, no. 3, pp. 340–352, Aug. 2000.
can be observed. The self-impedance of port P1 at low fre- [10] A. E. Ruehli, ‘‘Equivalent circuit models for three-dimensional multicon-
quency is influenced by these decoupling capacitors. How- ductor systems,’’ IEEE Trans. Microw. Theory Techn., vol. MTT-22, no. 3,
ever, at high frequency, the impedance of port P1 is almost pp. 216–221, Mar. 1974.
[11] J. Y. Choi and M. Swaminathan, ‘‘Decoupling capacitor placement in
unchanged. Since the series resonant frequency of the decou- power delivery networks using MFEM,’’ IEEE Trans. Compon., Packag.,
pling capacitor (0.353 GHz) is too low to influence the Manuf. Technol., vol. 1, no. 10, pp. 1651–1661, Oct. 2011.
impedance of port P1 at high frequency. [12] Y. Ji and T. H. Hubing, ‘‘On the modeling of a gapped power-bus structure
using a hybrid FEM/MoM approach,’’ IEEE Trans. Electromagn. Compat.,
vol. 44, no. 4, pp. 566–569, Nov. 2002.
V. CONCLUSION [13] J.-G. Yook, N. I. Dib, and L. P. B. Ratehi, ‘‘Characterization of high
Based on the resonant cavity algorithm, a modeling method frequency interconnects using finite difference time domain and finite
element methods,’’ IEEE Trans. Microw. Theory Techn., vol. 42, no. 9,
for the multilayered power/ground planes with stitching vias pp. 1727–1736, Sep. 1994.
has been discussed in this paper. Firstly, the influence of [14] T. Okoshi, Planar Circuits for Microwaves and Lightwaves. Munich,
the stitching vias to the return path of switching current is Germany: Springer-Verlag, 1984.
[15] G.-T. Lei, R. W. Techentin, P. R. Hayes, D. J. Schwab, and B. K. Gilbert,
analyzed. It is the anti-pad capacitance that transforms the ‘‘Wave model solution to the ground/power plane noise problem,’’ IEEE
switching current to other cavities and connect the cavities Trans. Instrum. Meas., vol. 44, no. 2, pp. 300–303, Apr. 1995.
as a whole. On the basis of the return current path, it is [16] N. Na and M. Swaminathan, ‘‘Modeling and transient simulation of planes
in electronic packages for GHz systems,’’ in Proc. IEEE 18th Topical
obtained that the remainder plane pairs act as a multiport Meeting Elect. Perform. Electron. Packag., Oct. 1999, pp. 149–152.
load to the plane pair where the current load is located in. [17] Z. L. Wang, O. Wada, Y. Toyota, and R. Koga, ‘‘Convergence acceleration
Then, this paper describes the physical principle, formulation, and accuracy improvement in power bus impedance calculation with a
fast algorithm using cavity modes,’’ IEEE Trans. Electromagn. Compat.,
and implementation of the modeling method for multilayered vol. 47, no. 1, pp. 2–9, Feb. 2005.
power/ground planes. According to the different structure [18] C. Wang et al., ‘‘An efficient approach for power delivery network
of multilayered planes and different types of stitching vias, design with closed-form expressions for parasitic interconnect induc-
tances,’’ IEEE Trans. Adv. Packag., vol. 29, no. 2, pp. 320–334,
the results computed by the modeling method presented in May 2006.
this paper are comprised with that simulated by HFSS. The [19] J.-M. Lu, Y.-S. Li, and M.-S. Zhang, ‘‘An efficient SPICE-compatible cav-
FSV results show the accuracy of the modeling method. ity resonant model for microstrip lines,’’ IEEE Trans. Compon., Packag.,
Manuf. Technol., vol. 1, no. 4, pp. 574–585, Apr. 2011.
It is also demonstrated that the modeling method can be [20] J. Wang, J. Lu, X. Chu, Y. Liu, and Y. Li, ‘‘Modeling and simulation
applied to solve the auto-decoupling design for a multilayered of planes with decoupling capacitors,’’ IEEE Trans. Compon., Packag.,
power/ground planes. The advantage of this modeling method Manuf. Technol., vol. 6, no. 7, pp. 1087–1098, Jul. 2016.
[21] J. Wang, J. Lu, and Y. Li, ‘‘Placement of decoupling capacitor
is that it can consider the influence of decoupling capacitor’s on packages based on effective decoupling radius,’’ in Proc. 18th
placements in the auto-decoupling design. IEEE Electron. Packag. Technol. Conf., Singapore, Nov./Dec. 2016,
pp. 195–198.
[22] R. Rimolo-Donadio et al., ‘‘Physics-based via and trace models for effi-
REFERENCES cient link simulation on multilayer structures up to 40 GHz,’’ IEEE Trans.
[1] P. Gupta and A. B. Kahng, ‘‘Efficient design and analysis of robust power Microw. Theory Techn., vol. 57, no. 8, pp. 2072–2083, Aug. 2009.
distribution meshes,’’ in Proc. 19th Int. Conf. Very Large Scale Integr. [23] A. Orlandi, G. Antonini, C. Ritota, and A. P. Duffy, ‘‘Enhancing fea-
Design (VLSID), Jan. 2006, pp. 337–342. ture selective validation (FSV) interpretation of EMC/SI results with
[2] L. D. Smith, ‘‘Simultaneous switch noise and power plane bounce for grade-spread,’’ in Proc. IEEE Int. Symp. Electromagn. Compat., vol. 2,
CMOS technology,’’ in Proc. IEEE 8th Top. Meeting Elect. Perform. Aug. 2006, pp. 362–367.
Electron. Packag., Oct. 1999, pp. 163–166. [24] IEEE Standard for Validation of Computational Electromagnetics Com-
[3] M.-S. Zhang, J.-F. Mao, and Y.-L. Long, ‘‘Power noise suppression using puter Modeling and Simulations, IEEE Standard 1597.1-2008, 2008.
power-and-ground via pairs in multilayered printed circuit boards,’’ IEEE [25] IEEE Recommended Practice for Validation of Computational Electro-
Trans. Compon., Packag., Manuf. Technol., vol. 1, no. 3, pp. 374–385, magnetics Computer Modeling and Simulations, IEEE Standard 1597.2-
Mar. 2011. 2010, 2011.

VOLUME 6, 2018 67371

 J. Wang et al.: Modeling of Multilayered Power/Ground Planes

JUN WANG received the B.S. degree in elec- YANG LIU received the M.Eng. and D.Eng.
tronic information science and technology from degrees from Xidian University, Xi’an, China,
the Langfang Teachers College, Langfang, China, in 2004 in 2007, respectively.
in 2008, and the D.Eng. degree from Xidian Uni- He is currently an Associate Professor with
versity, Xi’an, China, in 2017. the Electronic and Engineering Institute, Xidian
She is currently a Post-Doctoral Researcher University. His current research interests include
with the Electronic and Engineering Institute, power integrity, signal integrity, and image
Xidian University. Her current research interests processing.
focus on signal integrity, jitter, and power integrity.

JIANMIN LU received the B.S. degree in mate-

rial engineering from Zhengzhou University,
Zhengzhou, China, in 2005, and the Ph.D. degree
in electrical engineering from Xidian University,
Xi’an, China, in 2012.
He is currently with the Institute of Electronic
CAD affiliated with the Key Laboratory of High-
Speed Circuit Design and EMC, Ministry of Edu- YUSHAN LI received the B.S. degree in electronic
cation, Xidian University. His current research engineering from Harbin Engineering University,
interests focus on power noise suppression, time- Harbin, China, in 1968, and the M.S.E. degree
domain simulation, and high-speed interconnect modeling. in electronic engineering from Xidian University,
Xi’an, China, in 1981.
XIUQIN CHU received the M.Eng. and D.Eng. In 1981, he joined the School of Electronic
degrees from Xidian University, Xi’an, China, Engineering, Xidian University, as a Faculty Mem-
in 1997 and 2003, respectively. ber. From 1986 to 1987, he was a Visiting Scholar
She is currently an Associate Professor with and a Research Associate with the Department of
the Electronic and Engineering Institute, Xidian Electrical and Computer Engineering, University
University. Her main research interests are digital of Miami, Coral Gables, FL, USA. He is currently a Professor and a Ph.D.
image processing and signal integrity. Candidate Advisor with the School of Electronic Engineering, Xidian Uni-
versity. His current research interests include electronic design automation
and signal integrity analysis.

67372 VOLUME 6, 2018