Virtual prototyping for power diode and IGBT development

Virtual prototyping for power diode and IGBT development Maria Cotorogea Peter Türkes Andreas Groove 30 th Working Group Bipolar

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling for power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 2

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling of power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 3

Introduction IGBTs and power diodes are bipolar devices Losses are dominated by stored charges Development target: higher power density higher switching frequencies IGBT-modules: Chip development Package development Virtual Prototyping Group in Munich Why virtual prototyping in IGBT-module development? Reduce development costs and time by reducing learning cycles Target: accurately predict switching behavior Strategy: rollout on technology and package projects Model requirements: physics based models for IGBTs and diodes knowledge of parasitic elements and couplings fast model implementation and simulation 4

Introduction Not in focus Device triggered oscillation mechanisms Failure mechanisms Device reliability Outcome of VP activities Provide interfaces between different simulation levels (device simulation circuit simulation system simulation) Evaluate and enhance today s modeling precision of compact models to describe switching behavior of bipolar devices within SOA Consider electrical parasitics due to package design Assess current distribution in modules with several devices in parallel Investigate effect of manufacturing process tolerances in FE and BE Consider thermal couplings in modules and propose an electro-thermal co-simulation as full circuit-simulation approach 5

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling of power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 6

Virtual Prototyping Approach Virtual chip-module development flow Virtual module development CAD layout FEM Model FEM Simulation Parasitics compact model Application circuits Design / Process Circuit Module switching characteristics System-level models simulation Chip layout POR Process simulation Device model Device simulation Device compact model Virtual chip development Virtual data sheet 7

Virtual Prototyping Approach General Concept behind the Idea Switching behavior switching conditions power circuit & driver power devices & packages inherent electrical & thermal couplings Considerations short time scale: high du/dt & di/dt large time scale: self-heating Simulation needs appropriate electro-thermal models for IGBTs and diodes electrical and thermal parasitics models modeling techniques accounting for 2D or 3D geometrical design tools capable of facilitating a rapid iterative virtual design process 8

Automated model generation Manual model generation and calibration Minimum loss of precision! Virtual Prototyping Approach Physics-based device compact model 2D-TCAD Electrical parasitics compact model Circuit simulation 3D-FEM P DC P AC Thermal network as compact model 9

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling of power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 10

Compact modelling of power devices Description of charge dynamics in the drift region I tot = I p(0) 0 SCR I tot = I n(d) p+ n- n n+ Determined by implicit equation, derived from Poisson-Equation Charge dynamics in the base is described by the ambipolar diffusion equation (ADE): p t = p τ + D 2 p a x 2 with n p R Base, Q base Solution Methods for the local charge distribution: Fourier Transformation Kraus-Approximation using hyperbolic Functions Finite Element Method (FEM) Finite Difference Method (FDM) Advantages of FDM over Kraus-Ansatz FDM needs less approximations Example: p-i-n diode Approximation as a pure 1D problem! We need four boundary conditions Resistance modulation of small base-regions can be described Diode: no Fit-Parameter for ratio of plasma suspension points 11

Compact modelling of power devices Description of charge dynamics in the drift region I tot = I p(0) 0 SCR I tot = I n(d) p+ n- n n+ Determined by implicit equation, derived from Poisson-Equation Charge dynamics in the base is described by the ambipolar diffusion equation (ADE): p t = p τ + D 2 p a x 2 with n p R Base, Q base Solution Methods for the local charge distribution: Fourier Transformation Kraus-Approximation using hyperbolic Functions Finite Element Method (FEM) Finite Difference Method (FDM) Disadvantages of FDM over Kraus-Ansatz FDM has more unknowns and thus equations Example: p-i-n diode The SPICE code is significantly more complex! Approximation as a pure 1D problem! Convergence: Choose the correct boundary conditions We need four boundary conditions 12

Compact modelling of power devices Description of charge dynamics in the drift region Discretize base into equidistant points where the carrier concentration is calculated p i t = p i τ + D p i 1 2p i + p i+1 a x 2 Each point one equation Time integration directly calculated during transient analysis in SPICE Local integration for Q base (t) and R base (t) discrete sums with SPICE subcircuit Numerical Stable Boundary Conditions (BC) Note: Moving SCR boundary, thus x is changing with time High Injection at both junctions (anode & cathode): First 4 and last 4 points are used in Lagrange polynomials for local gradients at both junctions (anode & cathode): p 0 x = 1 6 x 11p 0 + 18p 1 9p 2 + 2p 3 =- μ n μ n +μ p 1 q A D I tot p n x = 1 6 x 2p n 3 + 9p n 2 18p n 1 + 11p n = μ p μ n +μ p 1 q A D I tot 13

out in OUT in3 in2 in1 Compact modelling of power devices Implementation of charge dynamics model in SPICE Vp P1 P2 P3 P4 P5 P6 P7 P8 P9 U8 relem2 dynamic charge and resistance 1 2 3 U9 diffusion 1 2 3 U10 diffusion 1 2 3 U1 diffusion 1 2 3 U2 diffusion 1 2 3 U3 diffusion 1 2 3 U4 diffusion 1 2 3 U5 diffusion 1 2 3 U6 diffusion 1 2 3 U7 diffusion Iin p1 p2 p8 p9 U11 dgl_randbed ADE Submodel pl pr 1K R1 V1 1 boundary-conditions ARB1 H1 U2 red @ time green @ time + t dp dt i D ( p 2 i 1 pi 1) pi (2 D x ) 2 x integrator 14

Compact modelling of power devices Implementation of charge dynamics model in SPICE Plasma concentration in a diode during turn-off Unique feature of a SIMetrix model 15

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling for power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 16

Assessment of modelling precision Work flow Calibration TCAD and compact chip models are calibrated on measurements of single chips mounted on DCBs Parasitics are extracted for the module and the test setup directly from the CAD layout (no C, R @ DC and L @ 100MHz) Parasitics of the driver are fitted from characterization measurements Precision assessment DoE for dynamic characterization in test circuit varying R g, V CC, I Load, T j and L s Measurements and simulations for at least 24 switching conditions Mounting as single-chip and in module package Evaluation of IGBT turn on and off + diode reverse recovery For each switching conditions extraction of up to 68 switching parameters from simulated and measured curves Qualitative assessment: overlay of simulated and measured switching transients for different switching conditions Quantitative assessment: simulation error (deviation) of extracted parameters for single-chip and module mounting 17

Assessment of modelling precision Switching curves Vce Turn on Turn off Vce Rg Ic TCAD Models (S-Device) Ic Rg Compact models (SIMetrix) 18

Assessment of modelling precision Extracted dynamic parameters [W] 19

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling for power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 20

Electro-thermal co-simulation in SPICE Motivation Currently used thermal models: Foster model (partial-fraction network, data sheet parameters) Cauer model (continued-fraction network, physically correct) Matlab model (direct combination of step responses from FEM analysis with power signal via convolution integral) Serious draw back: amount of RC elements when considering thermal couplings in complex modules Good to have: thermal model that can be directly integrated in circuit simulation Coupling with electrical model that considers loss distribution 21

Electro-thermal co-simulation in SPICE Thermal model generation with MOR tool for ANSYS Generates reduced thermal model without calculation of a temperature field. MOR = Model Order Reduction (procedure) ROM = Reduced Order Model (result) 3D-CAD material data relevant devices FEM network definition boundary conditions (convection etc.) load (power loss) 22

Electro-thermal co-simulation in SPICE Thermal model generation with MOR tool for ANSYS Generates reduced thermal model without calculation of a temperature field. MOR = Model Order Reduction (procedure) ROM = Reduced Order Model (result) 3D-CAD material data relevant devices FEM network definition boundary conditions (convection etc.) load (power loss) 62mm module Foster model with 624 Rs & Cs: Fitting time=15.6h Working time=23.6h Total time = 29.2h ROM for 180 dimensions with 6840 elements: Working time=2.0h Total time = 2.1h PrimePack TM 3 module Foster model with 9408 Rs & Cs: Fitting time=235h Working time=262h Total time = 410h ROM for 1000 dimensions with 146k elements: Working time=3h Total time = 12h 23

Electro-thermal co-simulation in SPICE Electro-thermal simulations with ROM for PP3 Electro-thermal model in SIMetrix results from merging the thermal ROM and the electrical compact model of the PP3 (semiconductor models were modified for power calculation and thermal handling) thermal network Example: Compact model of a PrimePack TM 3 with 48 semiconductors Only electrical network Output Transient power loss distribution in the module at fixed temperature Period: 35µs Convergence fail after 48 pulses! Estimated analysis time for 10ms: 37.8h Estimated analysis time for 100s: 43y! Transient multi-pulse simulation Electro-thermal network Output Transient temperature and power loss distribution in the module until thermal steady state reached Period: 35µs Convergence fail after 11 pulses! Estimated analysis time for 10ms: 140.3h Estimated analysis time for 100s: 160y 24

Electro-thermal co-simulation in SPICE Approach Transient multi-pulse electro-thermal simulation: not viable because of analysis time and convergence stability Next approach: perform separated, but coupled iterative simulations with pure electrical and pure thermal model simulate stationary temperature and power loss distribution in the module at thermal steady state in a DC/DC converter simulate transient temperature and power loss distribution in the module during thermal ramp up with sinusoidal current load (emulates inverter operation)

Total loss per chip Plus G1_11 G1_21 G1_31 G1_41 G1_51 G1_61 G2_11 G2_21 G2_31 G2_41 G2_51 G2_61 Diode2_11M Diode2_21M Diode2_31M Diode2_41M Diode2_51M Diode2_61M HE2 IGBT2_12M IGBT2_22M IGBT2_32M IGBT2_42M IGBT2_52M IGBT2_62M Diode1_12P Diode1_22P Diode1_32P Diode1_42P Diode1_52P Diode1_62P IGBT1_11P IGBT1_21P IGBT1_31P IGBT1_41P IGBT1_51P IGBT1_61P Diode1_11W Diode1_21W Diode1_31W Diode1_41W Diode1_51W Diode1_61W Diode2_11W Diode2_21W Diode2_31W Diode2_41W Diode2_51W Diode2_61W HE1 IGBT1_12W IGBT1_22W IGBT1_32W IGBT1_42W IGBT1_52W IGBT1_62W IGBT2_12W IGBT2_22W IGBT2_32W IGBT2_42W IGBT2_52W IGBT2_62W HG2 IP1 IP2 G1_11 G1_21 G1_31 G1_41 G1_51 G1_61 G2_11 G2_21 G2_31 G2_41 G2_51 G2_61 Diode2_11M Diode2_12M Diode2_21M Diode2_22M Diode2_31M Diode2_32M Diode2_41M Diode2_42M Diode2_51M Diode2_52M Diode2_61M Diode2_62M HE2 IGBT2_12M IGBT2_22M IGBT2_32M IGBT2_42M IGBT2_52M IGBT2_62M Diode1_12P Diode1_22P Diode1_32P Diode1_42P Diode1_52P Diode1_62P IGBT1_11P IGBT1_21P IGBT1_31P IGBT1_41P IGBT1_51P IGBT1_61P Diode1_11WDiode1_12W Diode1_21WDiode1_22W Diode1_31WDiode1_32W Diode1_41WDiode1_42W Diode1_51WDiode1_52W Diode1_61WDiode1_62W Diode2_11WDiode2_12W Diode2_21WDiode2_22W Diode2_31WDiode2_32W Diode2_41WDiode2_42W Diode2_51WDiode2_52W Diode2_61WDiode2_62W HE1 IGBT1_12W IGBT1_22W IGBT1_32W IGBT1_42W IGBT1_52W IGBT1_62W IGBT2_12W IGBT2_22W IGBT2_32W IGBT2_42W IGBT2_52W IGBT2_62W HG2 Plus G1_12 G1_22 G1_32 G1_42 G1_52 G1_62 G2_12 G2_22 G2_32 G2_42 G2_52 G2_62 IGBT2_11M IGBT2_21M IGBT2_31M IGBT2_41M IGBT2_51M IGBT2_61M Diode1_11P Diode1_21P Diode1_31P Diode1_41P Diode1_51P Diode1_61P HC1 IGBT1_12P IGBT1_22P IGBT1_32P IGBT1_42P IGBT1_52P IGBT1_62P IGBT1_11W IGBT1_21W IGBT1_31W IGBT1_41W IGBT1_51W IGBT1_61W IGBT2_11W IGBT2_21W IGBT2_31W IGBT2_41W IGBT2_51W IGBT2_61W HG1 Minus Wechsler FF1400R12IP4_RdcLac100MHz U1 PIL1 G1_12 G1_22 G1_32 G1_42 G1_52 G1_62 G2_12 G2_22 G2_32 G2_42 G2_52 G2_62 Diode2_12M Diode2_22M Diode2_32M Diode2_42M Diode2_52M Diode2_62M IGBT2_11M IGBT2_21M IGBT2_31M IGBT2_41M IGBT2_51M IGBT2_61M Diode1_11P Diode1_21P Diode1_31P Diode1_41P Diode1_51P Diode1_61P HC1 IGBT1_12P IGBT1_22P IGBT1_32P IGBT1_42P IGBT1_52P IGBT1_62P Diode1_12W Diode1_22W Diode1_32W Diode1_42W Diode1_52W Diode1_62W Diode2_12W Diode2_22W Diode2_32W Diode2_42W Diode2_52W Diode2_62W IGBT1_11W IGBT1_21W IGBT1_31W IGBT1_41W IGBT1_51W IGBT1_61W IGBT2_11W IGBT2_21W IGBT2_31W IGBT2_41W IGBT2_51W IGBT2_61W HG1 Minus Wechsler PIL3 3.379714324295 PIL5 249.779500651404 PIL7 3.092342728174 246.560629722351 PIL9 244.141734692882 PIL11 243.510914717039 PDL1 0.0457366750958548 PDL3 0.0583527217828008 PDL5 0.0639353704677629 PDL7 0.0619137081089526 PDL9 0.0573678645351192 PDL11 0.0442727672843596 PIH1 PIH3 0.0198797591381681 PIH5 0.0202047956111984 0.0204301527359044 PIH7 0.02071274836967 PIH9 0.0212157919284937 PIH11 0.0221505077710705 PDH1 110.070345323269 PDH3 111.551162487456 PDH5 PDH7 113.406335506563 PDH9 114.331689203209 PDH11 114.5789462689 109.482827935673 vjdl1 vjil1 vjdl5 vjil5 vjdl9 vjil9 DL1 P_IL DL9 IL9 R-L-Ersatzschaltung Aufbau Lastspule P_DL Plus 1.4k I1 Wechsler Diode2_51M Diode2_51W IGBT2_51M IGBT2_51W 500u R5 Diode2_11M Diode2_11W IGBT2_11M IL1 DL5 IGBT2_11W R4 Diode2_31M Diode2_31W IGBT2_31M IL5 IGBT2_31W R R17 C1 1n G2_11 G2_31 G2_51 IC vjil1 vjil3 vjil5 vjil7 vjil9 vjil11 vjdl1 vjdl3 vjdl5 vjdl7 vjdl9 vjdl11 vjih1 vjih3 vjih5 vjih7 vjih9 vjih11 vjdh1 vjdh3 vjdh5 vjdh7 vjdh9 vjdh11 IC I(in) OUT ARB1 I(in) vjdl2 vjil2 vjdl6 vjil6 vjdl10 vjil10 DL2 DL10 Diode2_52M Diode2_52W IL10 Plus IGBT2_52M IGBT2_52W Aufbau mit Quelle 36n L1 37.5m R1 Minus Diode2_12M Diode2_12W IGBT2_12M IL2 DL6 IGBT2_12W 600 V1 R6 Diode2_32M Diode2_32W IGBT2_32M IL6 IGBT2_32W R24 R16 Substrat 1 G2_12 Substrat 5 G2_52 G1_11 Substrat 3 G2_32 G1_31 G1_51 R8 R23 R15 170n L2 IGBT1_51P IH9 IGBT1_51W Diode1_51P Diode1_51W P_IH DH1 P_DH DH9 1.0 R2 Pulse(-15 15 10u 110n 110n 387u 669u) V2 10 R7-15 V3 1.3 R3 100n L5 IGBT1_11P IGBT1_11W Diode1_11P Diode1_11W Diode1_31P Ansteuerung IH1 IGBT1_31P IGBT1_31W Diode1_31W IH5 DH5 vjih1 vjdh1 vjih5 vjdh5 vjih9 vjdh9 G1_12 G1_32 G1_52 HG1 HE1 HG2 he g OUT ARB4 V(g)-V(he) HE2 R9 R22 R14 Vge IGBT1_12P IH2 IGBT1_12W Diode1_12P Diode1_12W IGBT1_32P IGBT1_32W Diode1_32P Diode1_32W IGBT1_52P IH10 IGBT1_52W Diode1_52P Diode1_52W DH2 IH6 DH6 DH10 Vge vjih2 vjdh2 vjih6 vjdh6 vjih10 vjdh10 U2 FF1400R12IP4_Therm_N100 vjil1 vjil3 vjil5 vjil7 vjil9 vjil11 vjdl1 vjdl3 vjdl5 vjdl7 vjdl9 vjdl11 vjih1 vjih3 vjih5 vjih7 vjih9 vjih11 vjdh1 vjdh3 vjdh5 vjdh7 vjdh9 vjdh11 gnd vjil2 vjil4 vjil6 vjil8 vjil10 vjil12 vjdl2 vjdl4 vjdl6 vjdl8 vjdl10 vjdl12 vjih2 vjih4 vjih6 vjih8 vjih10 vjih12 vjdh2 vjdh4 vjdh6 vjdh8 vjdh10 vjdh12 Vtambient IP1 IP2 HC1 HE1 HE2 HE1 I(in) OUT ARB2 I(in) k a OUT ARB3 V(A)-V(K) e c OUT ARB5 V(c)-V(e) VD ID Vce ID VD Vce vjdl3 vjil3 vjdl7 vjil7 vjdl11 vjil11 DL3 DL7 DL11 Diode2_21M Diode2_21W IGBT2_21M IL3 IGBT2_21W Diode2_41M Diode2_41W IGBT2_41M IL7 IGBT2_41W Diode2_61M Diode2_61W IL11 IGBT2_61M IGBT2_61W R13 R21 R29 G2_21 G2_41 G2_61 vjdl4 vjil4 vjdl8 vjil8 vjdl12 vjil12 DL4 DL8 DL12 Diode2_22M Diode2_22W IGBT2_22M IL4 IGBT2_22W Diode2_42M Diode2_42W IGBT2_42M IL8 IGBT2_42W Diode2_62M Diode2_62W IL12 vjil2 vjil4 vjil6 vjil8 vjil10 vjil12 vjdl2 vjdl4 vjdl6 vjdl8 vjdl10 vjdl12 vjih2 vjih4 vjih6 vjih8 vjih10 vjih12 vjdh2 vjdh4 vjdh6 vjdh8 vjdh12 vjdh10 IGBT2_62M IGBT2_62W R12 R20 R28 Substrat 2 G1_21 G2_22 Substrat 4 G1_41 G2_42 Substrat 6 G1_61 G2_62 R11 R19 R27 IGBT1_21P IH3 IGBT1_21W Diode1_21P Diode1_21W IGBT1_41P IH7 IGBT1_41W Diode1_41P Diode1_41W IGBT1_61P IH11 IGBT1_61W Diode1_61P Diode1_61W DH3 DH7 DH11 vjih3 vjdh3 vjih7 vjdh7 vjih11 vjdh11 G1_22 G1_42 G1_62 PIL2 R10 R18 R26 246.8240960714 PIL4 244.834384910374 PIL6 247.936990677803 PIL8 242.569039417704 PIL12 PIL10 240.406160611496 PDL12 239.887128801927 PDL10 0.0411749905897605 0.0578156757479723 PDL8 PDL6 0.0634768963893531 PDL4 0.0661033337570826 PDL2 0.0617732620971844 0.0500369742007769 PIH6 0.0204086514234834 PIH8 0.0206984417949577 PIH10 0.0212303454981039 PIH12 0.0222155205472372 PIH4 PIH2 0.02018857578181 PDH12 0.0198589220380567 PDH10 107.488108345139 112.929778003241 PDH2 108.866242343757 PDH4 110.4003961383 PDH6 111.946280909337 PDH8 112.7864118265 IGBT1_22P IH4 IGBT1_22W Diode1_22P Diode1_22W IGBT1_42P IH8 IGBT1_42W Diode1_42P Diode1_42W IGBT1_62P IH12 IGBT1_62W Diode1_62P Diode1_62W DH4 DH8 DH12 vjih4 vjdh4 vjih8 vjdh8 vjih12 vjdh12 Electro-thermal co-simulation in SPICE Approach Transient multi-pulse electro-thermal simulation: not viable because of analysis time and convergence stability Next approach: perform separated, but coupled iterative simulations with pure electrical and pure thermal model simulate stationary temperature and power loss distribution in the module at thermal steady state in a DC/DC converter simulate transient temperature and power loss distribution in the module during thermal ramp up with sinusoidal current load (emulates inverter operation) Electrical model: transient doublepulse simulation at given f and D with updated T script linkage Thermal model: DCOP with updated losses Mean temperature per chip 26

Electro-thermal co-simulation in SPICE Example: PrimePack TM 3 Stationary pseudo electro-thermal simulation Circuit: buck converter, low-side IGBT is switched Design: converter layout calculation for T jmax =150 C T initial = C Convergence criterion < 0.05 C 10 iteration needed Total analysis time: 3.3h High-side diode (DH) Low-side IGBT (IL) Result: mean temperature of each chip in good agreement with full FEM simulation 27

every 32 switching periods Total loss per chip Plus G1_11 G1_21 G1_31 G1_41 G1_51 G1_61 G2_11 G2_21 G2_31 G2_41 G2_51 G2_61 Diode2_11M Diode2_21M Diode2_31M Diode2_41M Diode2_51M Diode2_61M HE2 IGBT2_12M IGBT2_22M IGBT2_32M IGBT2_42M IGBT2_52M IGBT2_62M Diode1_12P Diode1_22P Diode1_32P Diode1_42P Diode1_52P Diode1_62P IGBT1_11P IGBT1_21P IGBT1_31P IGBT1_41P IGBT1_51P IGBT1_61P Diode1_11W Diode1_21W Diode1_31W Diode1_41W Diode1_51W Diode1_61W Diode2_11W Diode2_21W Diode2_31W Diode2_41W Diode2_51W Diode2_61W HE1 IGBT1_12W IGBT1_22W IGBT1_32W IGBT1_42W IGBT1_52W IGBT1_62W IGBT2_12W IGBT2_22W IGBT2_32W IGBT2_42W IGBT2_52W IGBT2_62W HG2 IP1 IP2 G1_11 G1_21 G1_31 G1_41 G1_51 G1_61 G2_11 G2_21 G2_31 G2_41 G2_51 G2_61 Diode2_11M Diode2_12M Diode2_21M Diode2_22M Diode2_31M Diode2_32M Diode2_41M Diode2_42M Diode2_51M Diode2_52M Diode2_61M Diode2_62M HE2 IGBT2_12M IGBT2_22M IGBT2_32M IGBT2_42M IGBT2_52M IGBT2_62M Diode1_12P Diode1_22P Diode1_32P Diode1_42P Diode1_52P Diode1_62P IGBT1_11P IGBT1_21P IGBT1_31P IGBT1_41P IGBT1_51P IGBT1_61P Diode1_11WDiode1_12W Diode1_21WDiode1_22W Diode1_31WDiode1_32W Diode1_41WDiode1_42W Diode1_51WDiode1_52W Diode1_61WDiode1_62W Diode2_11WDiode2_12W Diode2_21WDiode2_22W Diode2_31WDiode2_32W Diode2_41WDiode2_42W Diode2_51WDiode2_52W Diode2_61WDiode2_62W HE1 IGBT1_12W IGBT1_22W IGBT1_32W IGBT1_42W IGBT1_52W IGBT1_62W IGBT2_12W IGBT2_22W IGBT2_32W IGBT2_42W IGBT2_52W IGBT2_62W HG2 Plus G1_12 G1_22 G1_32 G1_42 G1_52 G1_62 G2_12 G2_22 G2_32 G2_42 G2_52 G2_62 IGBT2_11M IGBT2_21M IGBT2_31M IGBT2_41M IGBT2_51M IGBT2_61M Diode1_11P Diode1_21P Diode1_31P Diode1_41P Diode1_51P Diode1_61P HC1 IGBT1_12P IGBT1_22P IGBT1_32P IGBT1_42P IGBT1_52P IGBT1_62P IGBT1_11W IGBT1_21W IGBT1_31W IGBT1_41W IGBT1_51W IGBT1_61W IGBT2_11W IGBT2_21W IGBT2_31W IGBT2_41W IGBT2_51W IGBT2_61W HG1 Minus Wechsler FF1400R12IP4_RdcLac100MHz U1 PIL1 G1_12 G1_22 G1_32 G1_42 G1_52 G1_62 G2_12 G2_22 G2_32 G2_42 G2_52 G2_62 Diode2_12M Diode2_22M Diode2_32M Diode2_42M Diode2_52M Diode2_62M IGBT2_11M IGBT2_21M IGBT2_31M IGBT2_41M IGBT2_51M IGBT2_61M Diode1_11P Diode1_21P Diode1_31P Diode1_41P Diode1_51P Diode1_61P HC1 IGBT1_12P IGBT1_22P IGBT1_32P IGBT1_42P IGBT1_52P IGBT1_62P Diode1_12W Diode1_22W Diode1_32W Diode1_42W Diode1_52W Diode1_62W Diode2_12W Diode2_22W Diode2_32W Diode2_42W Diode2_52W Diode2_62W IGBT1_11W IGBT1_21W IGBT1_31W IGBT1_41W IGBT1_51W IGBT1_61W IGBT2_11W IGBT2_21W IGBT2_31W IGBT2_41W IGBT2_51W IGBT2_61W HG1 Minus Wechsler PIL3 3.379714324295 PIL5 249.779500651404 PIL7 3.092342728174 246.560629722351 PIL9 244.141734692882 PIL11 243.510914717039 PDL1 0.0457366750958548 PDL3 0.0583527217828008 PDL5 0.0639353704677629 PDL7 0.0619137081089526 PDL9 0.0573678645351192 PDL11 0.0442727672843596 PIH1 PIH3 0.0198797591381681 PIH5 0.0202047956111984 0.0204301527359044 PIH7 0.02071274836967 PIH9 0.0212157919284937 PIH11 0.0221505077710705 PDH1 110.070345323269 PDH3 111.551162487456 PDH5 PDH7 113.406335506563 PDH9 114.331689203209 PDH11 114.5789462689 109.482827935673 vjdl1 vjil1 vjdl5 vjil5 vjdl9 vjil9 DL1 P_IL DL9 IL9 R-L-Ersatzschaltung Aufbau Lastspule P_DL Plus Diode2_51M Diode2_51W IGBT2_51M IGBT2_51W 500u R5 1.4k I1 Wechsler Diode2_11M Diode2_11W IGBT2_11M IL1 DL5 IGBT2_11W R4 Diode2_31M Diode2_31W IGBT2_31M IL5 IGBT2_31W R R17 C1 1n G2_11 G2_31 G2_51 IC vjil1 vjil3 vjil5 vjil7 vjil9 vjil11 vjdl1 vjdl3 vjdl5 vjdl7 vjdl9 vjdl11 vjih1 vjih3 vjih5 vjih7 vjih9 vjih11 vjdh1 vjdh3 vjdh5 vjdh7 vjdh9 vjdh11 IC I(in) OUT ARB1 I(in) vjdl2 vjil2 vjdl6 vjil6 vjdl10 vjil10 DL2 DL10 Diode2_52M Diode2_52W IL10 Plus IGBT2_52M IGBT2_52W Aufbau mit Quelle 36n L1 37.5m R1 Minus Diode2_12M Diode2_12W IGBT2_12M IL2 DL6 IGBT2_12W 600 V1 R6 Diode2_32M Diode2_32W IGBT2_32M IL6 IGBT2_32W R24 R16 Substrat 1 G2_12 Substrat 5 G2_52 G1_11 Substrat 3 G2_32 G1_31 G1_51 R8 R23 R15 170n L2 IGBT1_51P IH9 IGBT1_51W Diode1_51P Diode1_51W P_IH DH1 P_DH DH9 1.0 R2 Pulse(-15 15 10u 110n 110n 387u 669u) V2 10 R7-15 V3 1.3 R3 100n L5 IGBT1_11P IGBT1_11W Diode1_11P Diode1_11W Diode1_31P Ansteuerung IH1 IGBT1_31P IGBT1_31W Diode1_31W IH5 DH5 vjih1 vjdh1 vjih5 vjdh5 vjih9 vjdh9 G1_12 G1_32 G1_52 HG1 HE1 HG2 he g OUT ARB4 V(g)-V(he) HE2 R9 R22 R14 Vge IGBT1_12P IH2 IGBT1_12W Diode1_12P Diode1_12W IGBT1_32P IGBT1_32W Diode1_32P Diode1_32W IGBT1_52P IH10 IGBT1_52W Diode1_52P Diode1_52W DH2 IH6 DH6 DH10 Vge vjih2 vjdh2 vjih6 vjdh6 vjih10 vjdh10 U2 FF1400R12IP4_Therm_N100 vjil1 vjil3 vjil5 vjil7 vjil9 vjil11 vjdl1 vjdl3 vjdl5 vjdl7 vjdl9 vjdl11 vjih1 vjih3 vjih5 vjih7 vjih9 vjih11 vjdh1 vjdh3 vjdh5 vjdh7 vjdh9 vjdh11 gnd vjil2 vjil4 vjil6 vjil8 vjil10 vjil12 vjdl2 vjdl4 vjdl6 vjdl8 vjdl10 vjdl12 vjih2 vjih4 vjih6 vjih8 vjih10 vjih12 vjdh2 vjdh4 vjdh6 vjdh8 vjdh10 vjdh12 Vtambient IP1 IP2 HC1 HE1 HE2 HE1 I(in) OUT ARB2 I(in) k a OUT ARB3 V(A)-V(K) e c OUT ARB5 V(c)-V(e) VD ID Vce ID VD Vce vjdl3 vjil3 vjdl7 vjil7 vjdl11 vjil11 DL3 DL7 DL11 Diode2_21M Diode2_21W IGBT2_21M IL3 IGBT2_21W Diode2_41M Diode2_41W IGBT2_41M IL7 IGBT2_41W Diode2_61M Diode2_61W IL11 IGBT2_61M IGBT2_61W R13 R21 R29 G2_21 G2_41 G2_61 vjdl4 vjil4 vjdl8 vjil8 vjdl12 vjil12 DL4 DL8 DL12 Diode2_22M Diode2_22W IGBT2_22M IL4 IGBT2_22W Diode2_42M Diode2_42W IGBT2_42M IL8 IGBT2_42W Diode2_62M Diode2_62W IL12 vjil2 vjil4 vjil6 vjil8 vjil10 vjil12 vjdl2 vjdl4 vjdl6 vjdl8 vjdl10 vjdl12 vjih2 vjih4 vjih6 vjih8 vjih10 vjih12 vjdh2 vjdh4 vjdh6 vjdh8 vjdh12 vjdh10 IGBT2_62M IGBT2_62W R12 R20 R28 Substrat 2 G1_21 G2_22 Substrat 4 G1_41 G2_42 Substrat 6 G1_61 G2_62 R11 R19 R27 IGBT1_21P IH3 IGBT1_21W Diode1_21P Diode1_21W IGBT1_41P IH7 IGBT1_41W Diode1_41P Diode1_41W IGBT1_61P IH11 IGBT1_61W Diode1_61P Diode1_61W DH3 DH7 DH11 vjih3 vjdh3 vjih7 vjdh7 vjih11 vjdh11 G1_22 G1_42 G1_62 PIL2 R10 R18 R26 246.8240960714 PIL4 244.834384910374 PIL6 247.936990677803 PIL8 242.569039417704 PIL12 PIL10 240.406160611496 PDL12 239.887128801927 PDL10 0.0411749905897605 0.0578156757479723 PDL8 PDL6 0.0634768963893531 PDL4 0.0661033337570826 PDL2 0.0617732620971844 0.0500369742007769 PIH6 0.0204086514234834 PIH8 0.0206984417949577 PIH10 0.0212303454981039 PIH12 0.0222155205472372 PIH4 PIH2 0.02018857578181 PDH12 0.0198589220380567 PDH10 107.488108345139 112.929778003241 PDH2 108.866242343757 PDH4 110.4003961383 PDH6 111.946280909337 PDH8 112.7864118265 IGBT1_22P IH4 IGBT1_22W Diode1_22P Diode1_22W IGBT1_42P IH8 IGBT1_42W Diode1_42P Diode1_42W IGBT1_62P IH12 IGBT1_62W Diode1_62P Diode1_62W DH4 DH8 DH12 vjih4 vjdh4 vjih8 vjdh8 vjih12 vjdh12 Electro-thermal co-simulation in SPICE Approach Transient multi-pulse electro-thermal simulation: not viable because of analysis time and convergence stability Next approach: perform separated, but coupled iterative simulations with pure electrical and pure thermal model simulate stationary temperature and power loss distribution in the module at thermal steady state in a DC/DC converter simulate transient temperature and power loss distribution in the module during thermal ramp up with sinusoidal current load (emulates inverter operation) Electrical model: 1 double-pulse simulation at given f sw and D with updated T and Iload script linkage Thermal model: quasi-continuous transient simulation with updated losses Mean temperature per chip every 32 switching periods 28

Electro-thermal co-simulation in SPICE Example: PrimePack TM 3 Results for transient pseudo electro-thermal simulation Switching: high-side IGBT, D=0.5, f sw =1500Hz Load: sinusoidal current source, f=1hz, I max =1240A T initial = T ambient = C Dt for thermal model update: 21.408ms Simulated time range: 9,6336s Total analysis time: 77h Mean T in IGBTs and diodes Chip-specific T in high-side IGBT Low-side diode (DH) High-side IGBT (IL) T IGBT T Diode 29

Outline 1 Introduction 2 Virtual Prototyping Approach 3 Compact modelling for power devices 4 Assessment of Model Precision 5 Electro-thermal co-simulation in SPICE 6 Summary and Outlook 30

Summary Virtual prototyping flow for device and package development Detailed evaluation of simulation precision Current distribution in modules for design optimization Investigation of impact of main FE and BE tolerances Simulation-based system-level models for thermal converter design Enhanced IGBT and diode compact models Electro-thermal co-simulation at circuit level 31

Outlook Compact chip models: Equivalent circuits Verilog-A Extraction of compact chip models from TCAD flow Automated generation of compact chip models Vision Establish MOR technique to predict temp. distribution Establish automated calibration routine Extend VP to discrete IGBTs in evaluation boards 32

33

34