See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/312155578 Reliability and Lifetime of Power Modules Tutorial: Industry Best Practices in Reliability Prediction and Assurance for Power Electronics Presentation · September 2016 DOI: 10.13140/RG.2.2.34359.32160 CITATION READS 1 4,214 1 author: M. Thoben University of Applied Science and Arts Dortmund 81 PUBLICATIONS 650 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Reliability of Power Modules View project All content following this page was uploaded by M. Thoben on 09 January 2017. The user has requested enhancement of the downloaded file. Reliability and Lifetime of Power Modules Tutorial: Industry Best Practices in Reliability Prediction and Assurance for Power Electronics M. Thoben Infineon Technologies, Max Planck Str. 1, D-59581 Warstein EPE´16 ECCE Europe 2016-09-05 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 2 Main inverter for hybrid and electric vehicles Full integration: radial distributed electronics source: BMW source: BMW Stand alone solution: radial distributed electronics Small mechanical dimensions Design flexibility for system integration Energy efficiency source: ZF Mechatronic solution: Electronics attached to transmission 2016-02-15 Lifetime System cost Copyright © Infineon Technologies AG 2016. All rights reserved. 3 Failure Rate Lifetime modeling / Mission profile is the basis for service life calculations Early „Infant Mortality“ Failure e.g. Assembly failures Wear Out Failures Random Failures Time 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 4 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 5 Reliability For „standard“ packages there are well defined „standard“automotive test qualification For power modules the LV 324 released in 2014 Qualification of Power Modules for the use in components of Electrical and Hybrid Vehicles General requirements, Test conditions and Tests Can be ordered at supplier portal of BMW, Daimler, VW e.g. (VW 82324 GS / BMW GS 95035) 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 6 Power Module Test matrix 8.2 QL-01 Power Cycling (PCsec) 8.2.1 Intention this is the basic test for ensuring the lifetime 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 7 PC conditions Tjmax Tj Tc Tcmax Internal heating and external cooling Conditions according : IEC 60749-34:2011 ∆Tj = 50~100K Tjmax = 150°C One test with IL>=0.85 ICN T1 T2 1 cycle Tamb Ic Icmax 0A PC,sec ton <= 5 s Focus: wire bond / die attach PC,min ton >= 15 s Focus: wire bond / die attach / substrate solder 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 8 Power cycling (sec.) Test procedure: TJ measurement by VCE (T) › TJ measurement indirectly by use of VCE (T) dependency @ 100mA current › VCE values directly measured after load current switch off @ 100mA and calculate TJ by use of VCE (T) correlation › Linear or polynomial fit Load current Measurement current TJ VCE @ 100mA Calculation via VCE (T) calibration VCE @ 100mA 2016-06-26 for internal use only Copyright © Infineon Technologies AG 2016. All rights reserved. 9 During the Validation Test example Power Cycling (sec.) 1 Online monitoring of VCE / Vf; TJ; TC; IC; RthJ-C data during the PC (sec) test – VCE increase due to bondwire lift-off / chip solder delamination – RthJ-C increase due to chip solder fatigue Assessment for eol: VCE/Vf limit: +5% of initial value RthJ-C limit: +20% of initial value 2016-06-26 for internal use only Copyright © Infineon Technologies AG 2016. All rights reserved. 10 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 11 Typical wear out mechanisms in power modules degradation of chip metallization wire bond lift-off Degradation of chip solder DBC solder joint cracking wire bond lift-off degradation of chip metallization solder joint cracking Degradation of chip solder Thermo mechanical stress leads to wear out of interconnects in power modules 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 12 Combination of PowerCycling and Thermal-Cycling with test condition: Tjmax=175 °C, Tcoolant,max=122 °C, Tcoolant, min=22 °C Twater_min 22°C Twater_max 122°C theating (passive) 10min tcooling (passive) 5min 200 180 Tjmax=175°C Tj 160 140 T [°C] 120 100 Tcase 80 60 40 20 0 0 100 200 300 400 500 t [s] 600 700 800 900 ILast 220A ton (active) 2s toff (active) 4s No of active cycles per passive cycle 100 Tjmax (standardmodule) 175°C Tjmax (improved module) 173°C 1000 Combination of PowerCycling and Thermal Cycling simulating Cold start condition for Power electronics at engine coolant loop 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 13 Degradation in substrate and chip solder after combined Thermal Cycling/PC test with Tjmax=175 °C and Tcoolant, max=122 °C Solder layer DCB-baseplate US-image in Solder Layer Chip-DCB After 240000 short / 2400 long cycles Delamination after 2400 long cycles Improved .XT System solder Degradation (according electrical measurement) No relevant Degradation >6500 long cycles possible Degradation of chip solder/wirebond and substrate solder occurs in parallel under combined test conditions Specific lifetime model for substrate solder necessary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 14 Influence of chip metallization and wire bond material on degradation mechanism 1,0E+08 Power Cycles 1,0E+07 1,0E+06 1,0E+05 Chip metallisation V2 + improved Al wire Combined test 1,0E+04 1,0E+03 20,0 Reference module Chip metallisation V1 + improved Al wire 40,0 60,0 80,0 100,0 120,0 140,0 160,0 ∆Tj Chip metallisation V1 Chipmetallisierung V2 Reference Power module Tjmax = 180°C Tjmax = 175..200°C Tjmax = 175..200°C More cycles are reached compared to combined test conditions Modification of chip metallisation and wire bond influences/improves reliability Lifetime model has to be adapted to specific failure mechanism 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 15 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 16 Lifetime models N f A T e a j N f K TJ 1 e ( Ea ) k B T j ,mean 2 TJ 273 ton 3 I 4 V 5 D 6 (Lesit project. 1997) (R. Bayerer et al. CIPS 2008) (U. Scheuermann et al. PCIM 2013) (Tj=Tlow in °C) Lebensdauerauswertung 1,0E+07 Comparison of Power Cycling Tests with CIPS 2008 model Anzahl der Zyklen 1,0E+06 1,0E+05 1,0E+04 40,00 50,00 60,00 70,00 80,00 90,00 100,00 110,00 120,00 ΔTj [K] Test 600V IGBT Test 1700V IGBT CIPS2008 1200V Standard BSM100GS120DLC Test 1200V IGBT CIPS2008 600V CIPS2008 1700V A. Hensler et al. : Proc. Braunschweiger Symposium Hybrid- und Elektrofahrzeuge (2011) Liftime is influenced by several factors: ton / wire thickness / current) Models have to reflect influencing parameters to be more precise set date Copyright © Infineon Technologies AG 2013. All rights reserved. Page 17 Reliability specification – HybridPACK Modules PC curves described for different Tvjmax Reliability specification below DT40K estimated 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 18 Reliability specification – PC ton time dependency If ton < 0.1s thes N=N(0.1s) If ton > 60s thes N=N(60s) ton influence taken into account for 0.1s<ton<60s 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 19 Reliability specification – thermal cycling 95% Curve Reliability specification for thermal cycling has to taken into account additionally to PC 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 20 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 21 Simulation Reliability PC,TC,TST (Tjmax = 125°C) 1,0E+11 systemsolderdegradation Copper Chipsoder delamination Power cycling 1,0E+10 dotted lines: estimated 1,0E+09 n (No. of Cycles) Wire bond lift-off Power Cycling: Low-, Medium & High Power Modules Solder 1,0E+08 1,0E+07 Traction Modules 1,0E+06 Standard Modules 1,0E+05 1,0E+04 10 100 Delta Tj in K Temperaturverlauf PC Chipsolde r System solder 160 Tmax 140 N= c1(wp)c2 Temp PLWK T w p d p ; 120 100 C PLWK Wire bond IGBT Module 80 da C C1 * w p 2 dN 60 ton 40 Simulationsmode ll 2016-08-28 toff 20 0 293 295 297 t [s] Copyright © Infineon Technologies AG 2016. All rights reserved. 299 301 22 Simulation of stress/strain in solder joint or wire bond Stress Temperature Viscoplastic strain - Hysterese wp strain energy density w p d p ; C da C C1 * w p 2 dN Strain energy density is calculated for solderjoint from simulation of PC 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 23 Usage of physics of failure simulation: Influence of on time on reliability of chip solder joint Variation of ton in Power Cycling Test (wton ) - c2 ; c2 2.1 - c2 N cyc (tref ) (wref ) N cyc (ton ) Simulation is used to investigate influencing parameter ton Exact prediction of crackpropagation not necessary Simulation results confirm empirical lifetime model for ton dependency 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 24 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 25 Use cases in a (x)eV › The system [electrical motor] has to support all these use cases Acceleration cruising hill hold / stall condition braking 2016-06-27 Copyright © Infineon Technologies AG 2016. All rights reserved. 26 Use-cases and mission profile equivalently important for proper inverter designs Use-cases describe individual situations Mission profile describe the actual usage 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 27 Calculation of life time from the given mission profile Electrical characteristics VCE sat, VF , Eon , Eoff , Erec Motor + drive control VDC, phase current, m, cos j, fs Mission profile Loss Calculation Loss profile Thermal simulation Thermal Parameters Zth jc, Zth CH , Zth H ambient , IGBT / diode Cycle numbers with different T Cooling Conditions Flowrate, Tcool Life time modeling Temperature Profile Tj max IGBT / diode / solder Climatic conditions Calculation of T occurrence Tambient 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. Life time Consumption per year 28 Calculation of lifetime based on mission profile Overview of calculation steps Mission Profile I_RMS V_DC F_S P_Diode COS_PHI M P_IGBT Power loss model Power loss profile Thermal model Temperature profile T Cycles @ T T, Tjmax,ton counting Lifetime model Power cycling / Thermal cycling Lifetime consumption 29 Copyright © Infineon 2016-06-27 Technologies AG 2016. All rights reserved. Dr. Krzysztof Mainka From Mission Profile to Power Loss Profile T 273 Eon ( I , V , T ) Eon ( I nom , Vnom,150) 150 273 E ( I ,V , 25) ln on nom nom Eon ( I nom,Vnom,150) 298 ln 423 T 273 Eoff ( I , V , T ) Eoff ( I nom, Vnom,150) 150 273 T 273 Erec ( I , V , T ) Erec ( I nom, Vnom,150) 150 273 Eoff ( I nom,Vnom, 25) ln Eoff ( I nom,Vnom,150) 298 ln 423 E ( I ,V , 25) ln rec nom nom Erec ( I nom,Vnom,150) 298 ln 423 I V I nom Vnom I V I nom Vnom I I nom Switching Losses 0.5 V Vnom Conduction losses Phase current, battery voltage, modulation index cos(ϕ)), switching frequency Mission profile parameters and the module’s characteristics are combined to a power loss profile. Power loss profile For a standard three phase inverter, well known mathematical formulas are used 2016-06-27 Copyright © Infineon Technologies AG 2016. All rights reserved. 30 Temperature profile computation RDs1 P_IGBT*xP TEMP_SOLDER + V RTs1 CDs1 P_Diode*xP CTs1 IGBT Passive IGBT TEMP_T IGBT Active P_IGBT*xP P_Diodepas*xP + RTp1 RT1 RT2 RT3 RT4 RT5 CT3 V Power Module CTp1 CT1 RDp1 RD1 CDp1 CD1 CT2 RD2 CT4 CT5 RD3 RD4 RD5 CD3 CD4 CD5 TEMP_D + V Diode P_IGBTpas*xP Diode Passive FEM Simulation for thermal model generation CD2 P_Diode*xP Diode Active Thermal Model of Power module for IGBT / Diode substrate solder considering cross coupling Thermal model for impedance and cross coupling of IGBT/diode Model for solder temperature approximating the step response Model for specific cooling system (e.g. heat transfer coefficient) 2016-06-27 Copyright © Infineon Technologies AG 2016. All rights reserved. 31 Temperature profile processing Active and passive ΔT cycles Cycles @ T Passive ΔT are defined as the difference between the maximum temperature reached during the active phase of the cycle and the ambient temperature (cold start). Passive cycles take into account the maximum temperature swing. Considering passive cycles ensures that calculations are conservative. Separating active/passive cycles simplifies life time calculations. 32 Copyright © Infineon 2016-06-27 Technologies AG 2016. All rights reserved. Dr. Krzysztof Mainka Temperature profile processing Edge-counting vs. Rainflow › The number, amplitude and resulting temperature of the ΔT are calculated from the temperature profile › Different counting algorithms interpret the same temperature profile differently T T T T ASTM E1049 or AFNOR A03-406 Dr. Krzysztof Mainka Copyright © Infineon Technologies 2011. All rights reserved. Page 33 Comparison of counting method Why Rainflow? Calculation error % K.Mainka, M. Thoben, O.Schilling: „Lifetime calculation for power modules, application and theory of models and counting methods“. EPE 2011 Dr. Krzysztof Mainka Copyright © Infineon Technologies 2011. All rights reserved. Page 34 Lifetime estimation Equivalent stress Active ∆T N N f(∆T, Tj, ton)This function is failure mechanism specific N Mission Profile Active ∆T ∆Tref ∆T Neq Passive ∆T N N Passive ∆T Equivalent Stress ∆Tref ∆T Optional ∆Tref f(∆T, Tj, ton) Mission Profile can be summarized to single equivalent stress value life time consumption from active and from passive cycles 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 35 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 36 Influence of module stack on maximum temperature during operation with same die size Increase of 40 to 50K for Power module without baseplate compared to direct cooled Module Significant impact of power module stack on lifetime 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 37 Power losses and temperature for different vehicle operations Power losses and temperature during recuperation with Power module 200A /direct cooled Power losses and temperature during motorstart with Power module 200A /direct cooled Increasing high losses on diode during recuperation (decreasing motor speed) 0.5s Boost operation with significant higher losses on the IGBT 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 38 Influence of chipsize and cooling method on temperature Required Operation Cycles and resulting temperature increase for different chip size and cooling in Power module with Copper baseplate Chipsize cooling direct 200A indirect 200A direct 400A indirect 400A required cycles [Mio/lifetime] boost coldstart motorstart recuperation 1 recuper. 2 IGBTDiode IGBTDiode IGBTDiode IGBT Diode IGBT Diode 51 44 44 49 77 33 41 84 23 48 73 51 50 50 81 36 63 103 33 57 27 21 24 20 33 19 28 41 16 24 46 40 28 24 34 20 48 61 26 34 0.15 0.15 0.12 0.12 1.5 1.5 0.8 0.8 2.4 2.4 Heat capacity of the baseplate is sufficient for 0.5s motorstart Direct cooling significantly reduces temperature ripple at recuperation 400A Chipsize necessary for required cycles although Tjmax not reached 2016-02-15 Copyright © Infineon Technologies AG 2016. All rights reserved. 39 Application parameters influencing lifetime regarding power cycling capability Small Changes in application parameters have high impact on lifetime - 10% Rth improvement=+39% life time -10% switching frequency =+17% life time 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 40 Cooling conditions: Rth definition for lifetime calculation Specification limit : 20% above initial Rth Typical value before stress test Rth for lifetime estimation should consider typical value(before stress test), Power Cycling test under LV324 condition already includes Rth degradation set date Copyright © Infineon Technologies AG 2013. All rights reserved. 41 Electrical Characteristics: Power Losses › Different definition for loss calculation are possible for Lifetime Simulation: – Switching losses: +10% to +20% Eon, Eoff, Erec from datasheet. – Conduction losses: Vcesat, Vf + 10% OR datasheet MAX values – Huge Impact on Lifetime: typ Eon, Eoff, Erec, Vcesat, Vf Lifetime => +10% Eon, Eoff, Erec, Vcesat, Vf => ≈ 50% Lifetime To prevent „overengineering“ loss data for Lifetime estimation should focus on typical Values before stress test, Value should address defined statistic range (e.g. 0.5*(MAX – TYP) + TYP) (includes systematic&statistic distribution but no EOL and measurement tolerance) 42 Electrical Characteristics: Same Rg and Vge but different switching speed Datasheet nomenclature Eoff from 10%Vce to 2%Ic Driver A Driver B Identical power module equal operating conditions! 100% Just using the same Rg and Vge is by far not enough to achieve equal switching performance! (losses may double with same Rg) Additional information is needed (e.g. dv/dt; di/dt) for correct loss data, measurement with final assembly/driver stage ensure precise loss data 43 Thermal cycling requirements for different cooling loops for 15 years lifetime N cyc: (TCASE ) C4 ( From reliability specification TCASE -4.5 ) 80K T_solder_max [°C] 85 85 85 85 85 85 85 85 85 85 85 85 T_water_min = Tc_min [°C] -25 -20 -15 -10 -5 0 5 10 15 20 25 30 delta Tc [K] 110 105 100 95 90 85 80 75 70 65 60 55 150 300 300 600 750 900 1350 1500 1500 1500 1050 1050 Cycles per lifetime equivalent number of T=80K for acceleration exponent :4,5 629 1020 819 1300 1274 1182 1350 1122 822 589 288 194 10590 A: coolant temperature is controlled to 70°C / temp. substrate solder: 85°C T_solder_max [°C] 55 60 65 70 75 80 85 90 95 100 105 110 T_water_min = Tc_min [°C] -25 -20 -15 -10 -5 0 5 10 15 20 25 30 delta Tc [K] 80 80 80 80 80 80 80 80 80 80 80 80 150 300 300 600 750 900 1350 1500 1500 1500 1050 1050 150 300 300 600 750 900 1350 1500 1500 1500 1050 1050 10950 Cycles per lifetime equivalent number of T=80K for acceleration exponent :4,5 Required Cycles B: coolant temperature has constant swing compared to ambient / temp. substrate solder: 55~110°C required test cycles for 15 years lifetime: 11000cycles, although solder joint maximum temperatures differs System level information needed for precise lifetime simulation 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 44 Contents 1 Motivation 2 LV324 Power Module Qualification 3 Power modules wear-out mechanisms 4 Lifetime models and Reliability Specifications 5 Physics of failure simulation 6 Mission profile simulation 7 Effect of parameters on life time expectations 8 Summary 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 45 Power Module with high reliabilty power cycling test showed a lifetime consumption of less than 10% today’s available power module technology ensures reliable power electronics over the vehicle lifetime when applying described methods 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 46 Summary › In the presentation methods are presented to ensure and determine lifetime based on mission profile › A detailed knowledge of degradation mechanisms of today’s power modules is available › If the failure mode is changing, e.g. crack in wire bond moves to crack in metallization layer, different solder materials or other assembly technologies are applied, the lifetime model has to be modified. › Physics of failure Finite Element simulations can help to describe and understand lifetime models › It is necessary to understand the reliability requirements from the applications in order to optimize the design of the module and all system parameters. › Today’s available power module technology ensures reliable power electronics over the vehicle lifetime when applying described methods 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 47 Time for Questions Thank you for your attention! Are there any questions? 2016-08-28 Copyright © Infineon Technologies AG 2016. All rights reserved. 48 Literature › [1] Thoben M., Mainka K., Bayerer R., Graf I., Münzer M.: From vehicle drive cycle to reliability testing of Power Modules for hybrid vehicle inverter. PCIM Europe 2008. › [2] Mainka K., Thoben M., Schilling O.: Lifetime calculation for power modules, application and theory of models and counting methods. EPE 2011 › [3] Schilling O., Schäfer M., Mainka K., Thoben M., Sauerland F. Power Cycling Testing and FE Modelling Focussed on Al Wire Bond Fatigue in High Power IGBT Modules. ESREF 2012 › [4] AN2010-02 Use of Power Cycling Curves for IGBT4 › [5] M. Thoben, F. Sauerland, K. Mainka, S. Edenharter, L. Beaurenaut: Lifetime Modeling and Simulation of Power Modules for Hybrid Electrical / Electrical Vehicles. ESREF 2014 View publication stats