Anda di halaman 1dari 5

2.

5D Methodologies for Electronic Package and PCB


Modeling: Review and Latest Development
En-Xiao Liu

Xing-Chang Wei, Er-Ping Li

Department of Electronics and Photonics, A*STAR Institute


of High Performance Computing (IHPC), Singapore
liuex@ihpc.a-star.edu.sg

College of Information Science & Electronic Engineering,


Zhejiang University, Hangzhou, Zhejiang, China

Abstract This paper first surveys 2.5D methodologies, which


are based on modal decomposition principle, for efficient
modeling of electronic packages and multilayer printed circuit
boards (PCB). It then reports the latest development of 2D
discontinuous Galerkin method (2D DGTD) based 2.5D
methodology for analysis of signal and power integrity in
multilayer PCBs. Moreover, the recent study of handling narrow
slots in the power-ground planes by a hybrid 1D- and 2D- DGTD
method is also presented. Numerical examples are given with
simulation results compared against measurement as well as fullwave simulation results.

II. OVERVIEW OF 2.5D METHOD

I. INTRODUCTION
The semiconductor technology keep evolving along the dual
paths of Moores law and more than Moore [1]. For a
successful design of high-speed electronic circuits and
systems, one needs to ensure signal and power integrity and
electromagnetic compatibility (EMC). Modeling as well as
simulation plays an ever-increasing role at the pre-layout
exploration and post-layout verification stage of a design.
Continuous increase in operating speed coupled with decrease
in supply voltage poses many challenges both to the design
and the modeling and simulation of high-speed systems.
Innovative methodologies for electrical modeling and design
are highly demanded to resolve both silicon complexity and
system complexity of sub-nanometer integrated circuits,
electronic packages and electronic systems.
Though high-speed electronic packages and multilayer
PCBs are usually far from being electrically large, they are
enormously complicated. Field or circuit method alone often
can neither deal with the humongous system-level simulation
properly, nor complete within a reasonable simulation time
and moderate memory usage. A plethora of computational
methodologies have been proposed to tackle the challenges
facing system-level modeling of packages and boards. Those
methods can be classified in many ways: hybrid, field, and
circuit approaches; numerical, semi-analytical, and analytical
approaches; frequency domain and time domain approaches;
and full-wave, quasi-static, and static methods. A review of
those different methods can be found in several reference,
such as [1].
Among all the modeling technologies, the 2.5
dimensional (2.5D) modeling methodology has evolved into

978-1-5090-1442-2/16/$31.00 2016 IEEE

an important category of methods, which strike a balance


between the accuracy of the full-wave methods and the speed
of circuit based methods. The current paper reports the
working principles and recent advancement of 2.5D
methodologies. It is a special domain decomposition approach
based on physics principles and natural physical structures, as
compared to other domain decomposition methods based on
rigorous mathematics, such as the finite element method with
domain-decomposition [2].

High-speed packages or printed circuit boards (see an


example in Fig. 1) are usually multilayer structures, which
consist of parallel conductive plates, multiple vias,
transmission lines such as microstrip lines and striplines, and
other discontinuities such as slots and apertures. Two
conductor plates neighboring with each other form a plate pair
(PP) which is essentially a parallel-plate waveguide or cavity.
The plate pairs naturally decompose the original multilayer
package or PCB into several largely decoupled domains. The
vertical vias connect neighboring domains through the
common via anti-pads. Each individual domain can be
represented by a microwave multi-port network. Those
networks are chained with one another in sequence by using
scattering, admittance or impedance (S/Y/Z) representations
into a complete electrical network of the entire structure such
as an electronic package or a PCB. Note that for the
convenience of presentation, we will exclude the modeling of
discontinuities other than vias, such as slots and apertures, and
leave it to the end of the paper for discussion.
A typical plate pair in Fig. 1, e.g., PP3, comprises two
conductive plates, striplines, vias, and a substrate. For
effective modeling of this typical domain, the modal
decomposition approach is employed to treat the two
decoupled modes separately, namely, the transmission line
mode (associated with stripline and microstrip line) and the
parallel-plate mode (associated with power-ground plane
pairs) [3-5]. The modal decomposition together with model
re-combination is elaborated in the next section.

801

III. DETAILS OF 2.5D METHOD


The section describes in detail how to model each
geometrical component (see Fig. 1(c)) and the corresponding
modes in a typical plate pair (see Fig. 1(b)).

shapes.
Over the years, researchers have developed a wide range of
2D methodologies including the 2D finite difference time
domain method (2D FDTD), 2D finite different frequency
domain method (2D FDFD), 2D finite element method (2D
FEM), contour integral (2D integral) method, and the recent
developed 2D discontinuous Galerkin method (2D DGTD).
The 2D DGTD method will be elaborated in the next Section.
The 2D FDTD method resolves the transverse magnetic
(TMz) modes in a plate pair by the following equations [8, 9]:

wEz
wt
wH x
wt

1 wH y wH x

 Jz

wy
H wx

1 wEz
,
P0 wy

wH y

1 wEz

wt

P0 wx

(1)

(2)

where H and P0 are the substrate permittivity and free-space


permeability, respectively. t represents time. J z is the zdirection excitation source. The 2D FDFD method has also
been used for this purpose [10].
The 2D FEM method [11] solve the following governing
equations for the plate pair:
xy u E z

 jZP H xy and xy u H z

( jZH  V d )E z (3)

where V d is the substrate conductivity.


2D IE method [3] solves the following equation for the
plate pair at its contour:

V r
Fig. 1
The 2.5D method based on modal decomposition domain
decomposition with is illustrated by using a multilayer PCB board as
an example [6]: (a) a representative multilayer PCB board, (b) a
typical plate pair comprise conductive plates, striplines and vias, (c)
parallel-plate modes, transmission-line modes, and cylindrical modes
corresponding with structures in (b).

k
2j

R n cH1 2 kR V r c

 jK dH 0 2 kR J n r c dl c
C
R

(4)
where H 0 is the second type of Hankel function of zero
2

orders, and H1 first order. r is the observation point, and r'


the source point. dl' follows the contour C. k is the substrate
wave number, and K the substrate wave-impedance.
2

A. Parallel Plate Pair

Each plate pair domains thickness is usually smaller than


one-tenth of the wavelength at the working frequency of
packages or PCB board. And the substrate material is usually
homogeneous for each domain. Thus the fields inside such a
parallel-plate pair is governed by a two-dimensional
Helmholtz equation with respect to the vertical electric
component ( Ez ), and the boundary of each domain can be
treated as a perfect magnetic conductor (PMC). The resonant
cavity model [7] was adapted for the modeling of plate pairs
with regular shapes. The 2D electromagnetic field solver,
which is one of the key components in a 2.5D method, are
widely used to modeling the special cavity formed by a
power-ground plane pair commonly found in printed circuit
boards (PCBs) and electronic packaging. Compared with the
resonant cavity method, 2D numerical methods are more
versatile in handling power-ground plane pairs in arbitrary

Fig. 2
(a) a stripline in cross-section view, and (b) it is split into
two transmission lines (microstrip lines) [1].

802

D. Cascading Individual Networks

The above networks for each domain or plate pair can be


connected together by using scattering (S) [13] or admittance
(Y) [12] parameters into the full network representing the
original full package or PCB.

Fig. 3
Via, stripline and plate pair models are linked together in a
five-port network.

Fig. 4
Cascading individual network for each parallel plate
domain into a full network representing the original package or PCB
board.

B. Trace

Traces, such as microstrip and striplines, can be described


by transmission lines with equivalent circuit or by network
parameters (S/Y/Z). Striplines require additional treatment in
order to be recombined with the parallel plate model and via
model. The reason is in traditional transmission line theory, a
stripline consists of the upper and lower ground planes that
have equal electrical potential. In the packages and PCBs,
however, those two reference ground planes for striplines
usually function as power-ground planes which have different
electrical potential. Therefore, in order to account for such a
potential difference, a stripline has to be split into two
microstrip lines [3] (see Fig. 2) to facilitate the model
recombination depicted in the next section. No special
treatment is needed to model microstrip lines including the
upper and lower striplines (see Fig. 3b) in the context of the
2.5D method. Depending on the structures (uniform or nonuniform) and operation frequencies of the microstrip lines, 2D
methods and/or 3D full-wave methods can be used to obtain
the equivalent circuits and/or the network parameters (S/Y/Z)
of the microstrip lines.

IV. LATEST DEVELOPMENT OF 2D DGTD METHOD


A new 2D field solver, based on 2D Time Domain
Discontinuous Galerkin (2D DGTD) method, is recently
developed for power-ground plane pair modeling without
apertures [14] and with apertures.
The DGTD method is superior to the FDTD method
because the former can model complex geometries and use
spatial discretization with high-order basis function for
improved accuracy. In contrast to the implicit TD-FEM
method, the DGTD method uses explicit time-stepping
scheme which reduces simulation time and memory usage.
The 2D DGTD method solves for the  mode in the powerground plates governed by 2D Maxwells equations. Similar
to the conventional FEM method, the Galerkins test
procedure is first used to result in weak-form equations. Now
numerical fluxes ( as well as  ) replace the
tangential magnetic and electric fields in the weak-form of
Galerkins equations. Such replacement differentiates the
DGTD from the conventional time-domain FEM and releases
tangential fields from being strictly continuous across the
shared edge of adjacent triangular mesh elements. Final
mathematical manipulation produces the strong-form
Galerkins equations as follows [14]:

C. Model Recombination for Plate-Trace-Via

The three-port network of vias in Fig. 3(a) is shown in Fig.


3(b) as a grey box. The stripline is split into two traces: one
trace is placed between Port 1 and 4, the other is between Port
2 and 5. The plate pair in the form of a S/Y/Z network is
connected to Port 3. The vias model, stripline model, and
plate-pair model are thus seamlessly merged into an integrated
network [12] which describes the initial plate pair (see Fig.
1(b)).

803

wE z ( t )
wt


D x H y (t )  D y H x (t )  V E z (t )  M 1 NF
E z ( t ), (5)

P0

P0

wH x ( t )
wt

wH y (t )
wt


 D y E z (t )  M 1 NF
H x ( t ),

(6)


D x Ez (t )  M 1 NF
H y (t ),

(7)

are the unknown nodal electric and


where and
magnetic field and both are vectors. M and D u are
mass and differentiation matrices, and N is

face matrix. is the number of nodes on element ks


edges ( ).
Its temporal discretization adopts a low-storage five-stage
fourth-order explicit Runge-Kutta method [15]. Assume that

, where
the ordinary differential equation is

 For time step , the Runge-Kutta method is


given below [14]:
p (0)
i >1, },5@ :

(i )
k

vn ,

ai k (i 1)  't L( p (i 1) , t n  ci 't )


p (i )
v n 1

Fig. 6 (a) A board with a 0.5 mm wide slot; (b) comparison of |S11|.

p (i 1)  bi k (i )

, (8)

p (5) ,

where is the time step size, and  are the


Runge-Kutta coefficients.
In order to handle power-ground planes with apertures, the
surface equivalence theorem is applied to represent the
aperture with its equivalent magnetic current governed by the
1D Telegraphers equations, which are solved by 1D DGTD
method. The power-ground plane model was solved by 2D
DGTD method. At each time step, 1D DGTD and 2D DGTD
are coupled to each other by updating the fields in the narrow
slots. The marching in time will stop when the
electromagnetic field results converge.

(a)

IV. RESULTS AND DISCUSSIONS


First, a pair of regular power-ground plane with
rectangular shape (see Fig. 5a) has the dimensions of 75 mm
long and 50 mm wide. Its substrate thickness is 1.1 mm with
 4.1, and 0.015. Fig. 5b shows that the S21
produced by the 2D DGTD using 6-order basis functions has
good correlation with the measurement data and CST
simulation results.

Fig. 5 (a) Test board; (b) comparison of |S21|.

Second, a board with a narrow aperture is shown in Fig.


6a and its geometrical parameters are indicated in Fig. 6b.
The accuracy of the coupled 1D DGTD and 2D DGTD
method for the modeling of power-ground plane with a
narrow aperture is again validated.

(b)

(c)

Fig. 7
A multilayer board consisting of eight conductive plates,
four striplines, and two through-hole vias: (a) top view, (b) crosssection view; and (c) 3D view. (Unit: mm)

Third, a multilayer test board as presented in [13] is used as


an example to compare different 2.5D methods based on
domain decomposition approach. Fig. 7 indicates all the
dimensions of the multilayer PCB board. The conductor
plates thickness is 0.01 mm. The substrates dielectric
constant is 3.84 and loss tangent is 0.033. The periphery of the
board is shorted with stitching vias running from top to
bottom conductor plates. The stitching vias have the same
diameters as the two through-hole vias. Fig. 8 shows relatively
good agreements between the results from the 2.5D method
(cavity), measurement [13], and full-wave HFSS simulation
with PCB vias (PEC-via) and PEC boundary conditions (PECB. C.). The full-wave PEC-via represents the case in Fig. 7,
where the periphery of the power-ground planes are shorted
by discrete conductor vias and simulated by the full-wave
solver. In contrast, the full-wave PEC-B.C. represents the case
(not shown in Fig. 7) where the periphery of the PCB are

804

assumed to be solid PEC. The results for those two cases have
some difference, because the array of via fences is
approaching the case of solid PEC walls, but not exactly the
same because of the little gaps between the neighboring PEC
vias. Fig. 8a seems to suggest that the S11 by the 2.5D
(Cavity) method are better than those by the full-wave
method. The actual reason is that the fabricated structure in
Fig. 7 has two rows of conductor via fences along certain
regions of the periphery of the PCB. Because of lacking
available details, the full-wave simulation is not able to
replicate exactly the shorted vias of the PCB as manufactured
in [13]. Note that the 2.5D method based on 2D-FEM
implemented in the SIwave 3.5 software (an early version of
SIwave, denoted as 2.5D (2D-FEM*) in Fig. 8, yields
inaccurate results for this study case, where the reason is
probably that the early versions of SIwave did not do the
plate-trace-via model properly or cannot handle the PEC via
boundary in this study case.

simulation speed and memory usage and are suitable for whatif studies of complex electronic packages and PCBs to achieve
signal and power integrity. Real-world high-speed PCB
boards and electronic packages are complicated and keep
evolving with technology advancement. Therefore, the 2.5D
methods need to be further developed to keep up with the
continuous advancement of high-speed electronic packages
and PCBs, such as new structures, new materials, new designs
and new technologies.
REFERENCES
[1]

[2]

[3]

[4]

[5]
[6]

(a)

[7]
[8]
[9]
[10]

[11]

[12]

(b)
Fig. 8
Scattering parameter results by domain decomposition
methods, measurement and the full-wave FEM method.

[13]

VII. SUMMARY
The underlying principle and details of 2.5D methods are
elaborated and the latest development of a 2D field solver
based on 2D DGTD method is also presented. The natural
domain decomposition reveals intuitively the physical
phenomena happening in multilayer packages and boards. The
2.5D methods win over full-wave methods in terms of

[14]

[15]

805

E.-P. Li, X. Wei, A. C. Cangellaris, E.-X. Liu, Y. Zhang, M. D'Amore,


J. Kim, and T. Sudo, "Progress Review of Electromagnetic
Compatibility Analysis Technologies for Packages, Printed Circuit
Boards, and Novel Interconnects," IEEE Trans. Electromagn. Compat.,
vol. 52, no. 2, pp. 248-265, May 2010 (invited paper).
Y. Shao, Z. Peng, and J.-F. Lee, "Full-Wave Real-Life 3-D Package
Signal Integrity Analysis Using Nonconformal Domain Decomposition
Method," IEEE Trans. Microwave Theory Tech., vol. 59, no. 2, pp.
230-241, Feb. 2011.
X. C. Wei, E.-P. Li, E.-X. Liu, and R. Vahldieck, "Efficient Simulation
of Power Distribution Network by Using Integral Equation and Modal
Decoupling Technology," IEEE Trans. Microwave Theory Tech., vol.
56, no. 10, pp. 2277-2285, Oct. 2008.
J. Fang, Y. Chen, Z. Wu, and D. Xue, "Model of interaction between
signal vias and metal planes in electronics packaging," in Proc. IEEE
3rd Topical Meeting on Electrical Performance of Electronic
Packaging, 1994, pp. 211-214.
A. E. Engin, W. John, G. Sommer, W. Mathis, and H. Reichl,
"Modeling of striplines between a power and a ground plane," IEEE
Trans. Adv. Packag., vol. 29, no. 3, pp. 415-426, 2006.
E.-P. Li and E.-X. Liu, "Domain and modal decomposition for efficient
signal and power integrity analysis of multilayer packages and PCBs,"
in IEEE MTT-S International Microwave Symposium, Montreal,
Canada, June 2012, pp. 1-3.
T. Okoshi, Planar Circuits for Microwaves and Lightwave. Munich,
Germany: Springer-Verlag, 1984.
W. K. Gwarek, "Analysis of arbitrarily shaped two-dimensional
microwave circuits by finite-difference time-domain method," IEEE
Trans. Microwave Theory Tech., vol. 36, no. 4, pp. 738-744, 1988.
E.-X. Liu, X. Wei, Z. Z. Oo, and E.-P. Li, "A 2D Time Domain Method
for Electrical Analysis of Electronic Packages," in Proc. 10th Electron.
Packag. Technol. Conf. , Singapore, Dec. 2008, pp. 254-259.
A. E. Engin, K. Bharath, and M. Swaminathan, "Multilayered FiniteDifference Method (MFDM) for Modeling of Package and Printed
Circuit Board Planes," IEEE Trans. Electromagn. Compat., vol. 49, no.
2, pp. 441-447, 2007.
J. E. Bracken, S. Polstyanko, I. Bardi, A. Mathis, and Z. J. Cendes,
"Analysis of system-level electromagnetic interference from electronic
packages and boards," in Proc. IEEE Conf. Electrical Performance of
Electronic Packaging, 24-26 Oct. 2005, pp. 183-186.
Z. Z. Oo, E.-X. Liu, X. C. Wei, Y. Zhang, and E.-P. Li, "Cascaded
Microwave Network Approach for Power and Signal Integrity Analysis
of Multilayer Electronic Packages," IEEE Trans. Compon. Packag.
Manuf. Technol., vol. 1, no. 9, pp. 1428-1437, Sept. 2011.
F. De Paulis, Y.-J. Zhang, and J. Fan, "Signal/Power Integrity Analysis
for Multilayer Printed Circuit Boards Using Cascaded S-Parameters,"
IEEE Trans. Electromagn. Compat., vol. 52, no. 4, pp. 1008-1018,
Nov. 2010.
Hui Min Lee, Siping Gao, En-Xao Liu, and et al., Two-Dimensional
Discontinuous Galerkin Time-Domain Method for Modeling of
Arbitrarily Shaped Power-Ground Planes, IEEE Trans. Electromagn.
Compat., vol. 57, no. 6, pp. 1744 - 1747, Dec. 2015.
J. S. Hesthaven, and T. Warburton, Nodal Discontinuous Galerkin
Method: Algorithms, Analysis and Applications. New York: Springer,
2008.

Anda mungkin juga menyukai