Computational Electromagnetics for Electromagnetic

Compatibility/ Signal Integrity Analysis

Li Er-Ping , PhD, IEEE Fellow

Advanced Electromagnetics and Electronic Systems Lab.
A*STAR , Institute of High Performance Computing (IHPC)
National University of Singapore
Erpingli@ieee.org

IEEE EMC DL Talk

Missouri Uni of ST
Aug. 15, 2008

Motivation of the Talk

Number of chapters wrote to me and asked to
talk on EMC Modeling, i.e.
Status of the numerical techniques
Applicability
Problems vs methods
Whether the simulation can solve 100% EMC
problems? If not why still develop and use it?

Outline
Overview of Computational Electromagnetic Modelling
Few Common Numerical Methods for EMC Modeling
MoM;
FDTD;
FEM.

Principle
Advantages/disadvantages
Typical Applications
Simulators

Modelling of Multilayered IC Packages

Motivation
Method Overview
Recent Development

Outlook and Summary

Simulation Challenges

Overview of Computational
Electromagnetic Modelling

The Needs for EMC Simulation

EMC is necessity

to guarantee no or least EM disturbance to the environment and to guarantee a correct work in environment EM
disturbance

to have a robust design in normal environment

The EMC becomes critical and more difficult

logic speed increase (frequency increase, transition time decrease) => high frequency emission increase

IC technologies evolution (size decrease, node capacitors decrease, digital level decrease, integration increase)
=> noise margin decreases and more sensitive to HF disturbances

power electronics evolution (digital control, switching frequency and power increase) makes harder standard
EMC emission compliance and design robustness

The EMC problems can be diversified

at system level: intersystems EMC and intrasystems EMC

at electronic board level

at chip/component level

EMC must be take into account at the beginning of the design

EMC modeling and simulation tools is required

to help system engineers in the architecture definition

to help electronic engineers in the product design with EMC consideration

to help industrial companies for reducing the time and cost of retrofit

Maxwell Equations

4
curl H =
c
1
curl E =
c
div D = 4

1 D
j +
c t
B
t

James Clerk Maxwell

(1831-1879)

div B = 0
... Light is an electromagnetic wave
governed by the interaction of electric
and magnetic fields.

Brief History of Electromagnetic Computation

1950s
Structural analysis
1965
A. M. Winslow (first EE application)
1966
Yee, Finite Differential Time Domain (FDTD)
1969
P. P. Silvester (waveguide analysis)
1970
Johns & Beurle (Transmission Line Method)
1974
K. K. Mei (unimoment method for
and antenna analysis)
1974
A. Ruehli( PEEC )
1980
J. C. Nedelec (vector elements)
1982
S. P. Marin (combined with boundary
equations for scattering analysis)
1983
P. P. Silvester & R. L. Ferrari, Finite
for Electrical Engineers, 1st ed.
1985Extensive developments for EM problems

J. M. Jin, FEM for EM, IEEE Press

scattering

integral
Elements

Recent Progress
Higher-order vector elements
Hybridization with boundary integral method
Hybridization with asymptotic methods
Time-domain finite element method
Fast multipole method
Multilevel Fast Multipole Method
Fast High Order Method

10

Computational Electromagnetic Methodologies

CEM
F.T.

Time Domain
PDE
FDTD

IE

TLM

IETD

FETD FVTD

Hybrid
Methods
J. M. Jin, FEM, IEEE Press, H.D. Bruns, et al ,

Frequency Domain

Low Frequency
PEEC

PDE

FEM FDM

High Frequency

IE

GO

PO

PEEC MOM

GTD

PTD

CGFFT

UTD

UAT

FMM
IEEE Trans on EMC, vol. 49, no. 2, 2007

AIM

SBR

Computational Electromagnetic for EMC

Analytical method:
only available for problems with a high degree of symmetry.

Numerical Methods (low frequency methods)

Integral equation based methods: Method of Moments (MoM), PEEC, Fast
Multipole Method (FMM), etc.
Differential equation based methods: finite element method (FEM),
finite difference method (FDTD), FIT, TLM, FVTD, etc.

Asymptotic Approaches (high frequency methods)

Geometric Optics (GO), geometric theory of diffraction(GTD)
Physical optics (PO), physical theory of diffraction (PTD)

Hybrid Methods
Numerical method cum numerical method: FEM-MoM
Numerical method cum asymptotic method: MoM-PO/GTD/PTD

H.D. Bruns, et al ,

IEEE Trans on EMC, vol. 49, no. 2, 2007, C. Paul, Intr of EMC, John Wiley

11

12

Review of Numerical
Methods
Method of Moments (MoM)

13

Method of Moments (MoM)

Method of moments (MoM)
transforms the governing integral equation of a given problem, by
weighted residual techniques, into a matrix equation to be solved
numerically on a computer

Illustration of the procedures of MoM:

>Consider the inhomogeneous equation (integral equation)<

L = f

Linear Operator (integral)

known function
(excitation)
Unknown function to
be determined

An example of the electric field integral equation (EFIE) for a perfectly

conducting (PEC) object illuminated by an incident field

n E i ( r ) = n j J s G ( r , r ) +
s J s ) G ds
(
j

S
Known
rS
Unknown
Greens function
(Incident field)

(surface current)

14

Method of Moments (MoM)

L = f

Illustration of the procedures of MoM (Contd):

Discretization

Meshing the structure into elements

Expanding the unknown function by using basis functions

n =1

vn

The original integral equation becomes:

n =1

< wn , Lvn > =< wn , f >

n =1

Testing (conversion)
Choose a set of testing functions
it into a matrix equation

Lvn = f

,w
take
the inner product, then convert
m

[ Z ]{ } = {b}

Solution & Post-Processing:

Solve the matrix equation for the unknown currents;
Calculate desired quantities.

Method of Moments (MoM)

15

Integral Equations for a given electromagnetic problem are formulated based

on the equivalence principle (alternatively on Greens identity)
Surface Integral Equation
Based on surface equivalence theorem: Fields outside an imaginary closed surface
can be determined by placing over the surface, suitable electric and magnetic currents
that satisfy the boundary conditions.
Suitable for impenetrable (PEC) body & homogeneous media

Volume Integral Equation

Based on volume equivalence theorem: Replace inhomegeneity of an object by
equivalent volume electric and magnetic currents that radiate in background medium.
Suitable for penetrable (inhomogeneous) media

Basis functions
Entire-domain basis function (regular domain)
Sub-domain basis function (complicated and arbitrary domain)
Examples --Pulse, roof-top, triangular, hexahedron and tetrahedron)

16

MOM for Simulation of EM Susceptibility

Incident Field
Coupling Algorithm

MOM simulation

L
W

Equivalent Circuit

EM
interference

Shielding cabinet

i
H

External cable

Circuit simulation
30

2k

Short-circuit current

180

-1k

Resistance
Reactance

Magnitude (mA)

Impedance (ohms)

25
1k

240

120
20

60

15

0
-60

10

Magnitude
-120
Phase

-180
-2k

200

400

600

800

Frequency (MHz)

Surface current

Equ Impedance

1000

0
200

400

600

Frequency (MHz)

Equ Current

Y W Liang and E P Li,A systematic coupled approach for EM susceptibility analysis of a shielded device with
multilayer circuits, IEEE Trans. On EMC, vol.47, no.4, 2005

800

-240
1000

Phase (degree)

4-layered PCB

17

MOM for Simulation of EM Susceptibility

Multi-layered PCB analysis with SPICE
Ambient EM interference: harmonic

plane wave

0.08
10 ohms
150 ohms

0.04

50 ohms
500 ohms

90 ohms

Out port 14

0.00

11

s
12

13

14

15

Voltage (V)

0.20

Out port 13

3=2.1

0.00

6
1

7
2

8
3

9
4

10
5

2=3.2

1=4.3

0.20
Out port 8

0.00
0.04

Out Port 3

0.00
0

200

400

600

800

Equivalent source

All ends of the traces are terminated

with the resistors.

1000

Frequency (MHz)
Y W Liang and E P Li,A systematic coupled approach for EM susceptibility analysis of a shielded device with
multilayer circuits, IEEE Trans. On EMC, vol.47, no.4, 2005

Method of Moments: Fast Algorithm

Fast Algorithm
The basic concept of fast algorithms is to decompose the MoM
matrix into near- and far-interaction components
To reduce the memory requirement for matrix storage and
accelerate matrix-vector multiplication
Typical fast algorithms:
(ML)FMM [(multi-level) fast multipole method],
CG-FFT (Conjugate gradient fast fourier transform)
AIM (adaptive integral method)
2-D representation of the procedures of
the AIM algorithm

AIM
(adaptive
integral
method)

(1)
(4)
(2)

(3)

(1)
(2)
(3)
(4)

Project panel current density onto grid

Compute potentials using FFT
Interpolate grid potentials onto panels
Compute near zone interactions directly

18

19

Method of Moments: Fast Algorithm

FMM & MLFMM (fast multipole)

One-level interaction (N2) among current elements

Down
(disaggregt.)

Up
(aggregt.)

A multilevel tree structure showing the interaction among the current

elements via the aggregation & disaggregation procedure

Connection between hubs

MoM : Z = b

O( N 2 )
FMM :
Nearinteraction

[Ref.] W. C. Chow, http://www.ccem.uiuc.edu/chew/aces2000_files/frame.htm

( Z NN +

Interaction from low-level

to upper level hub

V + T V ) = b O( N log N )
Interaction from upper-level
hub to low-level (elements)

Method of Moments: Fast Algorithm

Computation Cost (CPU time & memory requirements)
Conventional Method of Moments
O(N2)memory requirement for matrix storage & O(N3) operations for direct solution method
O(NiterN2) operations for iterative solver

Comparison of CPU time and memory usage between conventional MoM and Fast algrorithm

Significant reduction in CPU time and memory usage motivate the

development of fast algorithm
[Ref.] J. M. Jin, Finite Element Method in Electromagnetics, 2nd ed. Wiley Iterscience

20

Method of Moments (MoM)

MOM is strong in solving open domain problems involving
impenetrable (PEC) or homogeneous objects, and it has been
successfully applied to closed problems such as waveguides
and cavities as well
MoM is applicable to many EM-related application areas:
Electrostatic problems,
Wire antennas and scatterers,
Scattering and radiation from bodies of revolution or bodies of
arbitrary shape
Transmission lines
Aperture problems
Biomedical problems

21

Method of Moments (MoM)

Commercial Software
Numerical Electromagnetic Code (NEC)

Developed at the Lawrence Livermore National Laboratory

Frequency domain antenna modeling code for wire & surface structures

FEKO
EMC analysis, antenna design, microstrip antennas and circuits, dielectric media,
scattering analysis, etc.

IE3D
MMICs, RFICs, LTCC circuits, microwave/millimeter-wave circuits, IC interconnects
and packages, patch/wire antennas, and other RF/wireless antennas

22

23

Examples
Applications: Aviation industry
Method: MoM with FMM
Solver developed at IHPC, Singapore

GUI

RCS results

[Ref.] E. P. LI , et. al., IHPC Fast Algorithm Development -- Research Report, 2005.

Current distribution

24

Parallel Fast Integral Equation Simulation Method

Packaging Structure

V. Jandhyala,et al, IEEE EPEP, pp287-290, 2006

Simulation Time vs Number of Processors

25

26

Review of Numerical
Methods
Finite-Difference Time-Domain
Method (FDTD)

27

Finite-Difference Time-Domain Method (FDTD)

Finite Difference Time-Domain (FDTD) method,
first introduced y K.S. Yee in 1966, and later
developed by Taflove and others, is a direct solution
of Maxwells Time-dependent curl equations.
It is a robust, easy-to-understand , easy-toimplement techniques. It is one of the most popular
time-domain method for solving EM problems.

E
=

H = E + E

t
n+

1
2

df ( x0 )
= f ' ( x0 )
dx
f ( x0 + x 2) f ( x0 x 2)

x
1
2

H x | i , j ,k = H x | i , j , k
(

n
n
n
n

E
|
E
|
E
|
E
|

z ( i , j +1 2,k )
z ( i , j 1 2,k )
y ( i , j ,k +1/2 )
y ( i , j ,k 1/2 )
t

y
z

Finite-Difference Time-Domain Method (FDTD)

28

z Two interleaved grid points (E & H)

z E & H calculated alternatively at every half time step
z Time step is limited by the Courants condition
1
t
1 x 2 + 1 y 2 + 1 z 2

Yees cell in 3-D FDTD

simulation.

Strengths of FDTD:
Easy modeling of complex material
configuration
No matrix inversion involved
Easily adapted to parallel processing
Easy to generate broadband data
Ability to perform both transient and steady
state analysis

K S Yee, IEEE Trans A&P, vol.14 1966

Weaknesses of FDTD:
Mesh density is determined by fine geometric
details of the problem
staircase error for curve structure
Need to mesh the entire simulation domain
Need ABC (PML) to truncate unbounded
problem domain

Finite-Difference Time-Domain Method (FDTD)

29

Recent Development:
Domain decomposition; conformal FDTD; ADI-FDTD; Pseudo-spectral
FDTD
Other time domain methods
FIT (finite integration technique)
TLM (transmission line method)
FVTD (finite volume time domain)
Applications of FDTD
A variety of areas: Wave Propagation, Microwave/Antenna, highspeed electronics, photonics, biomedical problems

Finite-Difference Time-Domain Method (FDTD)

Commercial simulators
CST MicroWave STUDIO

Based on the Finite Integration Technique (FIT)

Full-wave electromagnetic field simulation software

Remcom XFDTD
Applications including microwave circuits, antennas, EMC, Scattering,
Photonics, Bio-EM, etc.

30

31

Examples for SI, EMI

Before Fix

After Fix

Courteous of CST, Hitachi Data storage

Examples

32

Crosstalk is concerned at dense path

Signal propagation along the lines and interference the
adjacent lines

Most coupling in hinge region

Imbalanced layout in this region

Read
Pair

Write
Pair

-Actuator
Pair

TFC
Pair
Write-to-Read
Crosstalk
Write-to-Read
Crosstlk
0

Signal Transfer
Voltage Transfer (dB)

Crosstalk (dB)

-10

-20

-30

-40

-50

-60
1

10 11 12 13 14 15 16 17 18 19 20

Frequency (GHz)

Courteous of CST, Hitachi Data storage

0
-3
-6
-9
-12
-15
10

100

1,000

Frequency (MHz)

10,000

37

Review of Numerical
Methods
Finite Element Method (FEM)

Finite Element Method (FEM)

Fundamentals of FEM:
FEM is a numerical technique to obtain the approximate solutions to boundary value
problems of the mathematical physics.
the equations in a Finite element (FEM) analysis can be formulated either by a
variational method (Ritz method) or a weighted residual method (Galerkins method)
[Also used by Method of Moment]
Variational method: Minimizing an energy functional
3D time harmonic EM problem

| H |2 | E |2 J E
+

F =
dV
V
2
2 j
2
energy stored in magnetic &
electric fields

energy dissipated/supplied by
conduction currents

Procedures of FEM Analysis:

Discretizing solution regions into finite number of subregions or elements
Deriving governing equations (elemental equation) for a typical element
Assembling of all elements in the solution region to form matrix equation
Solving the system of equations obtained.

38

Finite Element Method (FEM)

Features:
Remarkable advantage of FEM is the flexibility in terms of modeling any
complicated geometries, distribution of media.
Good handling of inhomogeneous medium (Each element can have different
material property)
Sparse matrix equation (each element only interacts with elements in its own
neighborhood)
Require to mesh the entire domain (object + background)
ABC, PML or FE-BI need to be used to truncate the mesh for unbounded
problems
Linear and nonlinear , 2-D/3D problems.
Widely used in frequency domain
Recent Development:
High-order ABC, domain decomposition, high-order elements
Commercial Software
Ansoft HFSS

39

EMI Simulation Examples

Antenna with Mould

Bushing
Monopole Model

Wire-frame Model of
PDA

60

Measured
E-field (dBuV/m)

50

Simulated

40

30

20

10

0
55.296

73.728

81

92.16

108

135

Frequency (MHz)

162

189

216

243

E-field on Casing

41

43

Review of Numerical
Methods
Comparison of Three CEM
Methods

44

Comparison of Three CEM Methods

FDTD

MOM

FEM

Principle

Direct solution of
Maxwells equations

Need Frequencydependent Greens

function

Variational principle
(minimizing energy
functional)

Equation

Differential equation

Integral equation

Differential equation

Methods

Transient or
steady state

Time-domain method;
Obtain responses over
a broad band
frequencies by Fourier
transform

Frequency domain
method -- response at
one frequency for one
solution of the matrix
equation

[Ref.] M. Sadiku and A. F. Peterson, IEEE Proc. 1990 Southeastcon, pp.42-47

Frequency domain method-response at one frequency

for one solution of the
matrix equation
Remedy: fast frequencysweeping approach to
obtain response over broad
band

Comparison of Three CEM Methods

45

Methods

FDTD

MOM

FEM

Geometry
materials

Inhomogeneity easy
Arbitrary shape
staircase error

Inhomogeneity difficult

Nonlinearity, ihhomogeneity
easy

Meshing

Entire domain
discretized

Normally only surfaces

discretized

Entire domain discretized

Matrix
equation

No matrix equation

Dense matrix

Sparse matrix

Boundary
treatment

Open boundary difficult

Absorbing boundary
needed

Open boundary easy

Open boundary difficult

Absorbing boundary
needed

Suitable
problems

Most developed time

domain method;
applicable to a variety of
electromagnetic
problems

More efficient to deal with

open domain problems
involving impenetrable
(PEC) or homogeneous
objects

More efficient for closed

region problems involving
complex geometries &
inhomogeneous media
objects;

46

Modelling of Multilayered IC
Packages
Background & Motivation

47

Packaging
Background

(Def.) Housing and interconnection of ICs to form product

To take up the slack --- difficulty of Moore's law scaling
Next-generation packaging --- 3D & more complicated
Emerging as

Motivation

Modeling and simulation

Enabler to reduce Cost & achieve Performance

- Present 2D

- Future 3D

Modeling and simulation techniques/tools for

Next-generation 3D packaging in high demand

Challenges

Objectives

[Short-term] To develop fast and accurate modeling techniques

for electrical & electromagnetic analysis of next generation

3D IC packaging and system integration
[Long-term] To explore a multi-physics platform: electrical-optical
-thermal-mechanical modeling

Multi-physics nature
[Mechanical, Thermal, Electrical et. al.]
Multi-scale nature
[nanometer to centimeter & DC to 10s/100s of GHz ]
Complexity
[Plenty of vias (intel--40k), signal traces & multiple P/G planes]

[Source: Georgia Tech]

[Source: IMEC, IZM]

49

Modelling of Multilayered IC
Packages
Method Overview

Approaches on IC Package Modeling

Circuit approach

Field approach

Coupled circuit-field approach

50

Wideband Modeling of Complete Signal Paths

51

in the Multi-layered Packages and Board by using the Multilumped Modeling Method
IC

IC

Complete signal path (chip to board)

Circuit Board
Modeling of high-speed system in package integration

Segmentation of a complete signal path

+
Chip

=
Substrate
Substrate

Flip chip
interconnect

strip package
Chip
trace with two
90 bends

Reference: I. Ndip, 2005 ECTC

+
=

Power/Ground Plane 1

+
Power/Ground Plane 2

Substrate

Through-hole via (strip to strip transition)

For illustrative reasons, dielectric layers
above plane 1, between both planes and
beneath plane 2, are removed

- strip package
trace with two
90 bends

Board

BGA ball

Equivalent Circuit Model of the Complete Signal Path

Port 1

Port 2

Flipchip
Interconnect

90 bend

Complete Signal Paths

Reference: I. Ndip, 2005 ECTC

90 bend

Via

90 bend

90 bend

BGA ball

52

53

Model Order Reduction Method/ MacroModel

R2=5

On-chip Interconnect

Vin
P1

P2

(2)
Fast 3D Field Solver
PSTD

C=10pf

(2)
(2)

FMM

Parameters Extraction
{S(s),Y(s) or parasitic parameters)

EMI Analysis
(Radiated
emission)

Model Order Reduction

(PRIMA)
Macromodel Synthesis
d
x(t ) = A x(t ) + B u (t )
dt
y (t ) = C x(t ) + Du (t )

Driver/Receiver
Circuit Model
E P Li, E X Liu, L W Li, IEEE Trav Adv. Packg. Vol 27, 2004

R2=50

Vout

Port Waveforms at
Interconnect Subnetwork
(2)
Signal Integrity Analysis
by SPICE Simulator
Electrical Performance
Analysis for Full Chip

Field Model-Full Wave Model

Parallel Computing & Domain Decomposition for IC Package
Modeling based on FDTD

Using Parallel FDTD

Iterative Bi-section approach for domain
decomposition & Load balancing
Ref: E P Li, et al, JEMAP, 2004,

EPEP, 2006

54

55

Results

CPU Time spent vs Processors

Magnetic field distribution at layer FC3

Ref: E P Li, et al, JEMAP, 2004,

Y.J. Zhang, et al, IEEE Sym. 2007

56

Modelling of Multilayered IC
Packages
Recent Development
Semi-analytical Method

57

New Algorithm Development

Domain-Decomposition Approach

Y2
YY22

Y3

Y4

YY44

YY1
1
YY3
3

Y1
[Phillips]

SIP (System in Package) on Silicon Carrier

Domain-Decomposition
Different parts are modeled by using different
optimized methodologies

58

Flowchart of Our Algorithm

Chip

Chip

Signal Layer

Layout of a Package
(Input)

GND
PWR

Domain
Decomposition

GND
PWR
Signal Layer

Chip

Top

Chip

Inner Domain

Top/bottom Domain

(P/G Planes, Vias,

Striplines etc.)

(ustrip lines, via bends, etc)

N-Body Scattering Theory

(NBST)

Parameter
Extraction

Sign
GND

GND

Inner

PWR
GND
PWR

Bottom

PWR
Signal Layer

Network Parameters [Y] or

[Z]

Equivalent Circuits
RLCG

New
NewFeatures
Featuresof
ofthe
theAlgorithm
AlgorithmCompared
Compared
to
toexisting
existingtechniques
techniques
2.5
2.5DDmethod
methodto
tosolve
solve3D
3Dproblem
problem
Semi-analytical
Semi-analyticalmethod
methodfast
fastbut
butaccurate
accurate
Efficiently
EfficientlyModeling
Modelinglarge
largenumber
numberof
ofvias
vias&&
Power/ground
Power/groundplanes
planes

System-level Electrical
modeling of entire
package

Modeling of Top/Bottom Domain

Equivalent Circuits

Integral Equation Method

Region 1
L1

L1
C
Region
2

Region
1

1a

Region
3

L1

Region
4

1b

Region 3
L3
C
3a

[Er-Ping LI, En-Xiao Liu et. al, IEEE EMC Symposium 2007]

L3
c

Region 2

C
3b

1c

2a

L1

L3
b

L3

L2

L2

L2

3c

4a

2b

Region 4
L4
a

L4
c

C
4b

2c

L2
d

L4
b

L4
d

C
4c

Modeling of Inner Domain

GND
PWR
GND
PWR

N-Body Scattering Theory

(NBST)
for analyzing coupling among multiple vias in
the presence of multilayered P/G planes

Superposition: Etotal = Einc + Escat(multiple)

Schematic cross sectional view of wave interactions among many cylinders inside an IC package;
The total electromagnetic wave is a superposition of the incident and scattered waves.

Inner Domain Modeling

Modal Expansion

TE & TM Modes
Parallel Plate Radial
Waveguide Mode (PPWG)
-- Adjacent P/G Planes

The governing Maxwells equations

have two independent solutions: the
TE (transverse electric) & TM
(transverse magnetic) modes. Each of
the two modes can be further
decomposed into a parallel plate radial
waveguide mode and a cylindrical
mode.

Cylindrical Mode (CYM)

-- Vias

E
E z = a mn
cos ( k m z ) Z n ( k ) e jn for TM
m

H
H z = a mn
cos ( k m z ) Z n ( k ) e jn for TE
m

[Z. Z. Oo, En-Xiao LIU et. al, ECTC 2007]

Multiple Coupling among Large Number of Vias

Scattering among N Vias
N via M q

( ) =

Nq

q =1 m = 0 n = N q

f qmn H n(2) ( km q ) e

jnq

cos( m z )

Scattered
fields

Coefficients of the incident wave

(excitation)

f q = Tq aq + qp f p
p =1; p q

N via

Unknown
coefficients
of the
scattered
wave

T (transition)
Matrix

Translation
Matrix

63

Experimental Validation

Test board 2:

Port 1

PCB 1
0

-5

-10

-10

-20

|S21| (dB)

|S11| (dB)

Port 2

two SMA ports (signal vias)

thickness 1 mm
via diameter 0.1 mm
substrate material 4.1; loss tan 0.02

-15
Measurement
Our method

-20

-40

-25
1

-30

Frequency (GHz)

Measurement
Our method

-50
1

Frequency (GHz)

Validation: L-shaped P/G Plane and Cut-out Structure

64

An irregular-shaped power/ground plane

substrate material 2.65; loss tan 0.003

two SMA ports at (61,5) and (10,42)
cut-out (15x15) in power/ground plane
60
SMM with FDCL
Measurement

50

|Z21| (dB)

40
30
20
10
0

Unit: mm

Measurement Ref.: Ping Liu et al, IEEE Trans EMC, vol. 47, no. 4, pp. 889-898.

Frequency (GHz)

Comparison against Ansoft HFSS

65

101 pins (vias) in three power/ground layers

0.25

smm
hfss

0.2

|Y_11|

0.15

0.1

0.05

0
1

Freq (GHz)

35

smm
hfss

30

|Z_11|

25
20
15
10
5
0
1

Freq (GHz)

101 Pins (PEC vias)

Three conductor layers (P/G/P) &

0.8-mm total thickness

20 x 20 mm dimension

66

Comparison against Ansoft HFSS

No. of
power/ground
layers

No. of
vias (pins)

Memory

CPU time

HFSS

Our
Approach

HFSS

Our Approach

20 MB

6 MB

10min 50
sec

43 sec

101

138 MB

43MB

77min 51sec

14min 57sec

101

420 MB

140MB

320min
29sec

20min 14sec

221

Out of memory

500MB

N.A

135min 53sec

50 MB

15 MB

13min 50sec

3min 21sec

4
Package Dimensions:
20 mm by 20 mm

Computing resource:
1.3 GHz CPU, 512 MB memory

Our approach:
About 1/3 of HFSSs memory usage
About 5 to 15 times faster than HFSS

Outlook and Summary

Simulation tools have been greatly improved in the last 10 years,
they are much faster and accurate now, it really could help the EMC
engineers to solve number of problems, but not 100%.
The present simulation tools can do
predict small scale and the regional EMI,
quickly predict and diagnose the regional EMI problems, i.e. Where
is the major source of the radiations,
optimize the performance for various designs at lower cost
compared to experiments ,
Be able quantitative assessments of insight EM performance, for
which experiment is unable to do ,
Overall shorten the design cycles.

68

Outlook and Summary

The present simulation tools cant
Still cant accurately simulate the entire system
level EMI, from components, board to system
levels,
Cant accurately simulate the large and complex
EMC problems,
No proper tools and accurate methods to simulate
the electromagnetic susceptibility, which is
nowadays critical for EMC design.

69

Outlook and Summary

This presentation gave an overview review of
three commonly used numerical Methods for
EMC;
it may help engineers to choose the right method/tool
for the right problem;
Integral methods vs. differential methods;
Surface methods vs. volume methods;
Time-domain methods vs. frequency methods.

70

Outlook and Summary

The future requirements
Simulation of Signal integrity and power integrity
simulation in real-world entire PCB and package system
is still challenge
Mixed thermal and electrical multiphysics simulation with
multi-scale natures
Comprehensive electromagnetic susceptibility simulation
is also required
A virtual EMC-test lab system should be developed to
model the entire system level EMC.

71

72

The Dream for EMC Modeling

Virtual EMC Lab

GUI

EMC/EMI
standards

EMC
modeling
techniques

Validation

Virtual EMC Lab

Conventional EMC testing
house

HPCV
computer
system

73