Content uploaded by Thomas Lagier
Author content
All content in this area was uploaded by Thomas Lagier on Dec 28, 2018
Content may be subject to copyright.
Design of a 1200 V, 100 kW Power
Converter: How Good are the Design
and Modelling Tools?
The Future of Simulation in Power Electronics
Packaging for Thermal and Stress Management
- ECPE Workshop
Thomas LAGIER, Cyril BUTTAY, Piotr
DWORAKOWSKI, Benhur ZOLETT
Introduction
During the design phases, the mistakes must be reduced
Accurate simulations can help the designers to reduce the risks
However, several limitations have been found in the tools
Purpose of the presentation
To identify the main limitations
To propose a enhanced simulation flow
21/11/2018Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 2
Agenda
Context
Simulation flow approach
Enhanced simulation flow
Conclusion & perspectives
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 3
Context
Motivations
Design flow validation
Performance estimation
1 year design & implementation
(2014-2015)
Design from a “blank page”
Specifications of our prototype
Input voltage : 900-1200 V
Output voltage : 450-600 V
Nominal power : 100 kW
Switching frequency : 20 kHz
Cooling : Forced air
21/11/2018Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 4
Limiting the design mistakes for high power converters
Context
Specifications of the power modules
Half bridge power module
5 MOSFET dies & 5 diodes dies per switch position
1,7 kV - 2 x 250 A
Design from die to converter level
SiC MOSFET & Diodes dies
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 5
Transformer
Control system
Design & simulation flows
System level studies
Use of Matlab calculation scripts
First estimation of the losses
Specification of the power modules
Power modules design
Number of dies in parallel
Use of the datasheet of the dies
Thermal characteristic of similar
power module
3DFE simulations
Determination of the parasitic
inductance and current distribution
System simulations
General approach
System level
calculation/simulations
Power module
specification & design
Design validation
Manufacturing
Datasheet of the dies
_Current, voltage & temperature levels
System specifications
Datasheet of the dies
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018
3D models
6
Simulation flow
Electromagnetic simulations
Reduced model
•Overvoltages
•Current reparation between the dies
•Temporal simulation domain
Electromagnetic simulations
(Ansys Q3D ®)
•Parasitic elements (inductances,
resistances, capacitances)
•Frequency domain simulation
(DC to 100 MHz)
•S parameters matrices
•RLGC matrices
•Equivalent circuit
Components considered
Power modules
Busbars
Gate driver
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 7
Circuit simulations
(Ansys Simplorer ®)
Simulation flow
Complexity of the model
80 MOSFET & diode dies to considered
High order (several hundreds) reduced models
Impossible to simulate the entire circuit
System time domain simulations
Input capacitors (RLC models) Input source
Busbar
(reduced model)
Power modules and
dies
Transformer (equivalent model)
Secondary inverter
Primary inverter
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 8
Simulation flow
Multi terminal systems : no transient solver
Use of “DC conduction solver”
Consideration of the RMS values of the current
No proximity/skin effects considered
Validity for thermal simulations ?
No electro-thermal transient simulations for multi terminal systems
Currents
Circuit simulations
(Ansys Simplorer ®)
Current densities
(Ansys Maxwell 3D ®) Thermal simulations
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018
Losses
9
Simulation flow
Accurate models are required to predict the losses
Hypothesis Error
First approximation
(Matlab script)
Polynomial approximation from datasheets
Spontaneous commutation lossless 25 %
Second approximation
(Matlab script)
Polynomial approximation from double pulse
Spontaneous commutation lossless 15 %
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 10
Next step : losses estimation with simulations ?
Simulation flow
Global system level simulation
Too complicated electro-magnetic reduced models
No Bbehavioral models for the power switches
No complete simulations of the system (control + power circuits)
No losses estimation
Limited ElectroMagnetic Disturbances consideration
No accurate thermal simulations
Complexity to consider both AC+DC excitations
No coupling between thermal and electromagnetic simulations
No coupling between thermal and mechanical simulations
No dielectric simulations
Issues during the tests
ElectroMagnetic Disturbances
Thermo-mechanical deformation on the power modules
Conclusion on the approach used
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 11
Enhanced simulation flow
Enhanced design flow
New approach proposed
Switches behavioral models
Thermal simulation
Dielectric simulations
Overvoltage
Losses estimation
Control/power interactions
EMI
…
Temperature
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018
Temperatures
Electro-mechanical stress
Voltage whistand
…
13
Electromagnetics reduced models
Open circuit
Capacitance @ 5 MHz
Measurement : 4,04 nF
Simulation : 3,95 nF
Short circuit
Inductance @ 5 MHz
Measurement : 49,7 nH
Simulation : 42 nH
Reduced models are satisfying but the order might be reduced
First order behavior
Use of simplified models in order to increase the simulation speed ?
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 14
Behavioral models for power switches
Several possibilities
Spice, VHDL-AMS
Models available in the software
Datasheets are not accurate enough to tune the models
Precise characterization are necessary
Behavioral models are the keys for simulation
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 15
Static characterizations
Dynamic
characterizations
3DFE models
(commutation cell)
Simulation model
Behavioral models power switches
Clamped inductive load test
1200 V –500 A
Power module based on CPM2-1700-0045B & CPW5-1700-Z050B dies
Behavioral models are good but might be improved
Vin Ids
L
Qlow
Qhigh
Vds
Switching losses comparison (IEC 60474)
Measure Simulation
Turn off (mJ)40 30 - 25 %
Turn on (mJ)35 45 + 30 %
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 16
Electro-thermal simulations
Proposed approach
Steady-State
Thermal Solver
Heat generation
(W/m3)
Temperature
Fields
(°C)
Mechanical
Structural
Solver
AC Conduction
Simulation
F1(Hz)
AC Conduction
Simulation
F2(Hz)
AC Conduction
Simulation
Fn(Hz)
DC Conduction
Simulation
Thermal
Displacement
Fields
(m)
N++
System simulations results
Currents (A)
Uni-directional coupling
Bi-directional coupling
Legend
Fluidic
simulation
Case
temperature
(°C)
Losses (W)
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 17
Electro-thermal simulations
Simulation results
98 % of the losses are in the dies
2 % of the losses in the connections
Experimental validation to be performed
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018
Temperature
0,7% 1,5% 0,2%
32,3%
65,3%
DC conection
20kHz conection
60kHz conection
Switching dies
Conduction dies
Losses in the connections
Losses in the power module
18
Operating point : 1,2 kV –125 A –20 kHz
Tc= 20 °C
Conclusion & perspective
3DFE simulation is necessary in the design of components
Global system level simulation helps to validate the design
Simplified models are the key for system level simulations
Behavioral models for power switches
Simplified 3DFE reduced models
The compromise between accuracy and simulation cost has to be found
Design of a 1200 V, 100 kW Power Converter: How Good are the Design and Modelling Tools? - Thomas LAGIER 21/11/2018 19
Thank you for your attention