Simulating EMC/EMI Effects for High-Power Inverter Systems

Simulating EMC/EMI Effects for 

High Power Inverter Systems 

Emmanuel Batista 

Vincent Delafosse, 

Alstom Pearl 

Ryan Magargle Ansoft Corporation 

emmanuel.batista@transport.alstom.com 

vdelafosse@ansoft.com 

rmagargle@ansoft.com 

© 2008 Ansoft, LLC All rights reserved. Ansoft, LLC Proprietary

Acknowledgments 

This work has been based on the work of 

Emmanuel Batista, J.M. Dienot, M. Mermet-Guyennet 

Special Thanks: 

– 

– 

– 

P. Solomalala (Pearl/Alstom) 

O.Roll, X. Legoar, D. Prestaux, 

X. Wu, M. Rosu, S. Kher (Ansoft) 


Pearl: Power Electronics Associated Research Laboratory 

Research 

Viability 

and Validation of technologies Development 

and maintenance 

Passive Components 

Active Components 

Packaging 

and validation of prototypes 

Models-Simulation-Fabrication 

EMC 

Solve multi-domain/temps/structure 

Design of methods for conception 


Motivation 

• 

• 

• 

• 

High power IGBT based inverter systems have specific EMC/EMI 

requirements 

The prediction of EMC/EMI fields is very difficult . Physical prototyping 

can result in long design cycles 

Simulation tools can help with the use of several techniques 

The physical quantities in the inverter that need accurate simulation are: 

– 

– 

– 

– 

– 

Quantity of current going through the conductors 

Frequency dependent parasitics (RLC) between conductors 

IGBT characterization curves 

Power dissipation 

Emitted fields 


Overview 

• 

• 

• 

• 

• 

• 

Introduction to the power study 

Static electromagnetic field study 

Parasitics extraction 

IGBT characterization 

System simulation 



Introduction 

• 

These 

power converters 

Pantograph 

Traction Supply 

Traction 

Motor 

IGBT Module Top Row 

are used 

Inverter 

in high 

© 2008 Ansoft, LLC All rights reserved. Ansoft, LLC Proprietary 

AM 

3~ 

speed trains (TGV) 

Inverter Leg

Introduction 

Emitter 

Dielectric 

Gel 

Diode 

Chip 

IGBT 

Chips 

6.5kV IGBT Module Characteristics 

Baseplate 

Ceramic 

Substrate 

Collector 

Packaging 

6.5kV-600A 

6.5kV 600A 

Module 

24 IGBT and 

12 Diode Chips 


Introduction 

6.5kV IGBT Module Analysis 


• 

• 

• 

• 

Include package in IGBT 

performance 

Find DC current distribution 

Find switching currents for 

power dissipation 

Use power dissipation to 

determine environmental 

electromagnetic fields

Introduction 

IGBT Module 

Pack 3D 

accurate model 

Electromagnetic 

(EM) study 

Model design developed 

• 

Parameters 

Extraction 


at 

Alstom/Pearl 

Design and Couplings 

Model 

IGBT Model 

Tridimensional IGBT pack model and EM study 

• Parasitic model extraction 

• IGBT circuit model 

• Far Field Study for Electric Field EM 

Far Field Study

Introduction 

Different 

Finite Element Method 

Modeling 

Boundary Element 

Method 

• 

techniques will 


be 

seen 

Tridimensional IGBT pack model and EM study 

• Parasitic model extraction 

• IGBT circuit model 

System Simulation 

• Far Field Study for Electric Field EM 

Finite Element 

Method

Overview 

• 

• 

• 

• 

• 

• 


Static electromagnetic field study 






ElectroMagnetic Study 

• 

• 

Module layout 

verification 

The module contains 8 IGBTs in parallel: 

does each IGBT receive the same amount of 

current? 

– 

– 

If the current flows un-evenly, this will cause 

mechanical stress and reliability issues. 

Electromagnetic simulation is required. We 

use Maxwell3D. 



• 

• 

• 

• 

• 

The layout 

The DC solver 

in imported 

is 

The input current 

The sink 

used 

from 

(600 A) is 

(return current 

Outputs: conduction path 

path) is 

the CAD tool 

defined 

defined 

600 A Sink 

and current 

distribution 



The structure is meshed 

using automatic and 

adaptive meshing 

Current Distribution 

IGBTs on, Diodes off 



• 

• 

• 

The end IGBTs see less current 

than the center ones. 

This can cause reliability issues 

as the center IGBTs will be 

overloaded 

An optimization of the copper 

tracks can be made in order to 

equalize the currents. 

Igbt1a and Igbt4a have the 

highest quantity of current 


Overview 

• 

• 

• 

• 

• 

• 


Near field electromagnetic study 






Parasitics Extraction 

• 

• 

• 

Once the layout is optimized, the next step is to extract the resistance, 

inductance and capacitance (RLC) parameters of the package. 

For this 

Example 

we 

use the boundary 

for two conductors 

element 

method 

in Q3D 



• 

Frequency Dependent Effects 

• 

• 

• 

Integrated power-electronic modules exhibit frequency-dependent 

behavior due to eddy current and skin effects. 

In these cases, it may not be sufficient to rely on resistance and 

inductance extracted at a single operating frequency 

For example, coax conductors: 

Low Frequency High Frequency 

Same 

geometry 

Different 

frequency 


= 

Different Parasitics


• 

Gate 

Extracting parameters is straightforward as the nets are 

automatically assigned. 

net 

Emitter 

Collector net 


net


• 

• 

How do we set up the frequency sweep? 

– 

– 

We 

Through Nyquist sampling, we know that to capture a 

time step of Ts, we need to obtain frequency domain 

information up to: 

1 

Fmax 

= 

2× 

t 

For a time domain waveform with a risetime of 80 ns, in 

order to capture the ringing in the time domain, we would 

want to capture at least 4 samples during this risetime. 

This implies a sampling time of 20 ns 

need 

to solve 

s 

up to 50 MHz (= 1/20ns) 



• 

The simulation outputs consist of the RLC matrices for different frequencies 



• 

• 

How do we 

Basic methodology: 

• 

• 

• 

• 

use the parasitics in the circuit simulator? 

Compute N-port S parameters (frequency sweep) 

Convert this into information that circuit simulator understands 

Circuit simulator performs inverse FFT to find impulse response 

n 

y( 

nΔt) 

≅ ∑ s(( 

n − k) 

Δt) 

x( 

kΔt) 

k = n−( 

N −1) 

Convolution is used to produce time-domain results 


t 

∫−∞ Y ( jω 

) = S( 

jω) 

X ( jω) 

→ y( 

t) 

= s( 

t) 

⊗ x( 

t) 

= s( 

t −τ 

) x( 

τ ) dτ 

Copper shield 

Silicon 

Polyethylene 

Silicon 

Copper 

Voltage 

1.1 

1.1 

1.1 

1.0 

950.0m 

900.0m 

876.5m 

Voltage versus Time Using Different 2D Extractor Mode 

Damping 

VM11.V [V] 

VM_Linear_1Hz_Model.V [V] 

VM_Linear_1MHz_Model.V [V] 

VM_Frequency_Model.V [V] 

Phase 

17.55u 18.00u 18.50u 19.00u 19.50u 

20.00u 

Time (Seconds)

Overview 

• 

• 

• 

• 

• 

• 


Near field electromagnetic study 






SIMPLORER 8 

• 

• 

Simplorer 8 is a circuit simulation 

tool for solving multi-domain lumped 

circuit problems. 

Link projects together to achieve 

dynamic linking of multiple 

simulations on a single sheet. 


SIMPLORER 8 

• 

Modeling 

• 

• 

• 

• 

New Parametrization 

tool for IGBT 

Enhanced SMPS Library - Over 

450 New VHDL-AMS DC/DC 

Converter Models in SMPS Digital 

Co-simulation 

Spice–Pspiceintegration 

Enhancements to individual 

models … 


System Integration 

• 

How do we 

import the results 

from 

• 

• 

• 

• 

• 

Q3D?: Q3D dynamic link 

2 Types of links: Single Frequency or 

Frequency dependent 

No need 

to manually 

import output file 

Simplorer incorporates directly the Q3D 

project 

If some results are not available, Simplorer 

dynamically launches Q3D 

Parameters and variables can be passed 

between S8 and Q3D 



Wattmeter 

IGBT 

Vc 

Power Module from Q3D 

for board parasitics 


Vg

IGBT Characterization 

2 types of models 

• 

• 

• 

Basic Dynamic 

Accurate models of the semiconductors 

are needed to achieve a good circuit 

simulation 

Simplorer 8 offers a parameterization 

tool for IGBTs 

The user needs to import the data from 

the datasheet 

are available 

and Average Dynamic 

in Simplorer 8: 



Objective Average Basic 

Dynamic 

DC 

characteristics 

Electrical 

Dynamics 

Thermal 

Dynamics 

Capacitance 

Models 

- 

Transfer characteristic 

Advanced Dynamic 

Ic(Vge) accurate 

- Output characteristic Ic(Vce) accurate in 

the regions of voltage and current saturation 

- Intrinsic temperature dependency 

- 

Considered 

Partial Fractional or 

Continued Fractional 

- Default Full access to the 

C(V) C(V) characteristics 



Sub circuit of the basic dynamic IGBT model 



SheetScan 



• 

• 

Once all the curves 

The tool 

fits 

and data are entered, start 

the data to the internal 

Characterization tool 

extraction 

Simplorer model using 

Genetic 

Algorithm 

Component dialog 



Test Circuit 

rise time= 40 μs 

fall time = 50 μs 

Ansoft Corporation Simplorer1 

50.00 3000.00 

15.00 

R2.I [A] 

40.00 

30.00 

20.00 

10.00 

0.00 

-10.00 

U1.VCE 

2500.00 

2000.00 

1500.00 

1000.00 

500.00 

0.00 

499.90 499.95 500.00 500.05 500.10 

Time [us] 

500.15 500.20 500.25 500.30 

2000.00 

1500.00 

1000.00 

500.00 


switch_on 

TR 

TR 

TR 

Curve Info 

Ansoft Corporation Simplorer1 

50.00 2500.00 

15.00 

Curve Info 

R2.I [A] 

40.00 

30.00 

20.00 

10.00 

0.00 

-10.00 

U1.VCE 

Vce 

Switch on 

Switch off 

switch_off 

Ic 

Vce 

0.00 

-15.00 

999.00 999.50 1000.00 1000.50 1001.00 

Time [us] 

1001.50 1002.00 1002.50 1003.00 

Ic 

U1.VCE 

VM2.V 

R2.I 

TR 

5.00 

0.00 

-5.00 

10.00 

5.00 

0.00 

-10.00 

-5.00 

-10.00 

-15.00 

U1.VCE 

VM2.V 

TR 10.00 

TR 

R2.I 

VM2.V [V] 

VM2.V [V]


2500 Voltage 

Source 

Line Resistance 

and Line Inductance 

Vg: Gate Voltage 

(+/-15V) 



• 

• 

• 

Issue: 

• 

• 

Accurate simulation of the switching of the IGBTS requires 

very small time steps (hmin = 10ps) 

System simulation requires 

long time step 

(t = 5ms) 

Simplorer allows the user to dynamically change hmin and hmax 

using State Graphs. 

When 

the switching 

has occured, the time step 


can 

be 

increased. 

Switching Steady state Switching


-Vge 

Vce 

Vge 

Vce 

Ic 

Reduce Time Step HMin 

Ic 

Vge 

Vce 

Vce, Vge, Ic 

over time (Igbt3b) 


Vge 

Vce 

Ic 

Vge 

Ic 

g 

Vce 

c 

e 

Ic


Vg 

Vce 

Vge 

Power 

The power pulse duration is much smaller than the rise/fall time 


Ic 

of Ic 

and Vce


Instantaneous power level through Igbt3a 



Power levels of the full set of IGBT’s on switch on 



• 

• 

Igbt1a and Igbt4a receive the highest 

power levels. 

This is consistent with the DC 

Conduction Maxwell3D solution 

Igbt1a 

Igbt4a 



The fundamental frequencies of the power range 

between 16 and 54 MHz 

t @ 

Pmax 

(μs) 

t @ P 


FTT of the power through Igbt1a 

Most of the power level is below 110 MHz 


Emitted Fields 

• 

• 

There is very high power going through the IGBTs (almost 60 000 W in this 

study) during a very short period of time (60 ns). This switching can cause 

EMI issues in the inverter, but also in the surrounding equipment 

To be 

• 

• 

answered 

using 

the finite 

Will the module radiate? 

Are the field 

levels 

surrounding 

element 

method 

the module within 

in HFSS: 

mandated levels? 



• 

• 

• 

• 

• 

The power pulse in the IGBTs have most of the 

energy in the 16- 110 MHz range. 

The largest 

There is 

metallic 

piece 

is 

a chance of having 

λ 

< 4 * L = 600 mm. 

This is 

By itself, the module will 

150 mm in the module 

radiation if 

for a frequency 

not radiate. 

of 500 MHz. 

However, the power module in the train is 

surrounded by other metallic objects than can be 

fairly large. These objects can cause the radiation 

of electric fields during switching. 

Maxwell’s Equations 

div D = ρ 

curl E = -∂B/∂t 

div B = 0 

curl H = J + ∂D/∂t 



Exposure 

limits 

defined 

by European 

Community 


• 

• 

Regulators impose 

maximum levels of 

electric fields close to 

electric equipment. 

In the 10-110 MHz 

range: 

Emax=61V/m


• 

• 

Each 

IGBT pad is 

The port lies between 

excited 

using 

the collector 

lumped 

ports 

and emitter pads 



• 

• 

The structure is discretized with adaptive meshing. The meshing 

frequency is 100 MHz 

The frequency 

sweep 

ranges from 15MHz to 120 MHz 



Spectrum 

(MHz) 

For each 

Power 

(W) 

frequency, the power amplitude is 

E field at 1m for 1000w 

(V/m) 

E field at 1m 

(V/m) 

16.52892562 21439.97604 2.6312 56.41286497 

33.05785124 8635.09049 2.7994 24.17307232 

49.58677686 5579.619715 2.8731 16.0308054 

66.11570248 4131.16773 3.063 12.65376676 

82.6446281 3276.823585 3.4045 11.15594589 

99.17355372 2712.888158 3.8924 10.55964586 

115.7024793 2308.359536 4.4861 10.35553171 

132.231405 2022.75744 4.905 9.921625241 

Spectrum from 

Outputs from Simplorer 

Inputs for HFSS 

Simplorer 

Outputs From HFSS 

(normalized results) 

Fields Levels 

entered 



mag 

E @ 100 MHz, Power = 10 000W 


• 

• 

• 

The E field is very localized 

close to the module even at 

100 MHz 

However, the very high 

power can lead to large 

values of E field even far 

from the module 

This design is fine at 

110MHz. 

Power E field at 1m 

Spectrum (M Hz) (W) (V/ m) 

115.7024793 2308.359536 10.35553171

Conclusion 

• 

• 

• 

• 

We have seen that the combination of several simulation techniques can 

give a good approach to EMC/EMI issues, both in conduction and emission 

modes 

Accurate prediction requires the use of Finite Element Methods, Boundary 

Element Methods, System Simulation along with Accurate Component 

Characteristics 

Package traces need optimization to balance current distribution 

The simulated module does not radiate for the given harmonics, and is 

within regulated near field field limits.

Simulating EMC/EMI Effects for High-Power Inverter Systems

Create successful ePaper yourself

Delete template?

Save as template?