Design tools and Simulations for Power Electronics Packaging Dr Hua Lu University of Greenwich Power Electronics Centre Annual Conference 5-6 July 2016 1 CMRG Group at Greenwich • CMRG: Computational Mechanics and Reliability Group • Working on power electronics: – Chris Bailey, Hua Lu, Pushparajah Rajaguru, Catherine Tonry, Mohammad Shahjalal • Ongoing power electronics research activities – UPE Components integration theme • WP CI4: Design Tools and Modelling – T4.1 Design for electro-thermal management – T4.2 Design for Reliability and Robustness – T4.3: Virtual Prototyping Framework for Power Electronic Systems – UPE cross cutting topics • Press-Pack, LED drivers, Multi-domain optimization Power Electronics Centre Annual Conference 5-6 July 2016 2 Physics of Failure Based Design and Reliability Analysis Tool Select Component FEA • Geometry • Material properties Electro-thermal Temperature Data acquisition • Clech • Analytical • Damage mech. • Surrogate model • Passives Power Electronics Centre Annual Conference 5-6 July 2016 System Reliability Risk Analysis Risk Analysis Damage & Lifetime 3 Failure Mechanisms To be Considered S. Yang, A. Bryant, P. Mawby, D. Xiang, L. Ran and P. Tavner, "An Industry-Based Survey of Reliability in Power Electronic Converters," in IEEE Transactions on Industry Applications, vol. 47, no. 3, pp. 14411451, May-June 2011. Types of failure mechanisms • Semiconductor device level • HCI, TDDB • Passive components • Changes in capacitance, ESR, leakage current • Inductor winding insulation break down • Packaging failure • Wirebond lifting/heel cracking • Solder joint degradation • Die cracking • Substrate delamination/cracking • First step is to focus on capacitor and power module failure Power Electronics Centre Annual Conference 5-6 July 2016 4 Reliability Prediction Methods • Primarily physics of failure based reliability prediction – Identify failure mechanisms • e.g. Crack in components and interconnect, degradation of dielectrics, etc. – Understand the physical processes • E.g. Electric currents and voltages, Fluid flow, heat transfer, diffusion, electromigration – Predict lifetime based on physical process analysis under given conditions • E.g. critical crack length, ESR etc. • Advantages – More accurate than traditional “handbook” method – Optimization at design stage • Disadvantages – Need lots of data and deep understanding of damaging processes • Structure, Failure and lifetime model, Environment, Material properties – Often complicated and time consuming • Expensive experiments and specialised software tools and training for numerical simulation Power Electronics Centre Annual Conference 5-6 July 2016 5 Capacitors Failure Mechanisms • Failures: high leakage, open, short, ESR/capacitance drift • Mechanisms: cracking, assembly and handling damage, electrolyte drying, electro-chemical migration, corrosion … Aluminium Electrolytic capacitors Failure rates: p b CV Q E Acceleration factors: capacitance, quality, environment Source: NAVSEA Power Electronics Centre Annual Conference 5-6 July 2016 6 Capacitor Lifetime Models Aluminuim electrolytic capacitor Vr Tm T10a T L L0 2 V0 n Ta, Tm are ambient and maximum temperature rating, ΔT is the temperature rise from ripple current. Vr and V0 are voltage rating and applied voltage. L0 is load life rating, n =1 and 7 for aluminium electrolytic and film capacitor respectively. Ceramic capacitors Vr L L0 V0 3 Tm Ta 8 Lifetime models may also be written as : • A primary reason for wear out in aluminium electrolytic capacitors is due to vaporization of electrolyte and degradation of electrolyte due to ion exchange during charging/discharging • In a physics based model, volume needs to be tracked and used in lifetime prediction: Power Electronics Centre Annual Conference 5-6 July 2016 7 Solder/Wirebond Lifetime Models • Coffin Manson Lifetime model • Energy based • Strain based Nf -Number of cycles N f AT b N f Af a T b expEa / RTmax N f aW pb N f a bp W p -Plastic work density/cycle -Plastic strain range/cycle p 0.9 W p ij d t ij p p t t They are correlated fatigue lifetime and are sometimes called “damage indicators” Plastic work density 0.8 2 d p : d p 3 W p 0.7 0.6 Wp(MJ/m3) t 0.5 0.4 0.3 0.2 0.1 0.0 0 2,000 4,000 6,000 8,000 10,000 12,000 t(s) Power Electronics Centre Annual Conference 5-6 July 2016 8 Single and Multiple Stress level Loading • Rainflow algorithm for cycle counts at different stress levels Miner’s Damage accumulation law if N fi Wi D 1 k n W i 1 i i This damage can be used to predict lifetime (based on the loading or given future loading) data Power Electronics Centre Annual Conference 5-6 July 2016 Rainflow cycle counting 9 Analysis: Surrogate Models • Identify design variables (DVs) • Use DOE to find points in the design space • carry out FEA simulation to obtain damage/lifetime for each design. • Surrogate models are obtained by fitting simple functions to the results W W X , X , X X X X Example with 3 DVs 3 1 2 3 i j 1 3 ij i j i 1 i i E p 0.0092 0.00416 x1 1.00 1006 x2 0.000445 x3 2.50 10 07 x1 * x2 0.00027 x1 * x3 5.00 1007 x2 * x3 0.001914 x1 * x1 1.99 10 05 x2 * x2 0.00015 x3 * x3 Power Electronics Centre Annual Conference 5-6 July 2016 10 Stress Analysis: 3D -> 2D • 2D FEA simulation is much faster than 3D • But – Not always a good approximation – Inaccurate results • Can 2D modelling be used for lifetime prediction and design optimization? What are the limitations? Power Electronics Centre Annual Conference 5-6 July 2016 11 3D, 2D Comparison 0.6 ΔW(MPa) 0.5 0.4 0.3 0.2 0.1 0 A VM stress and deformation show significant difference B C D E F A. B. C. D. Plane strain Axisymmetric 3D slice model with two sides fixed 3D slice model with symmetry boundaries on one side and the other is coupled E. 3D slice model with one symmetry plane and the other side is free F. 3D Model Power Electronics Centre Annual Conference 5-6 July 2016 12 2D FEA for Trend & Lifetime Prediction 0.9 ΔW3D(MPa 0.8 0.7 0.6 0.5 0.4 0.3 0.19 0.21 0.23 0.25 ΔW2D(MPa) 0.27 0.29 • ΔW values are different for 3D and 2D models • ΔW3D and ΔW2D is linearly related (Plane strain 2D): ΔW3D =5.25ΔW2D-0.812 Power Electronics Centre Annual Conference 5-6 July 2016 13 Stress Analysis: Clech Algorithm • • • • Stress states are described by τ and γ Time dependent temperature loading Temperature increases step by step At each time step stress and strain evolves along a stress reduction line 12 DT2 T0 1 / GT 1 1 / K 1 / GT 12 12 / K DT2 T0 2 12 crt 2 2 / K DT2 T0 Stress reduction line / K DT T0 Hall, P. M., 34th Electronic Components Conference, New Orleans, LA, May 14-16, 1984, pp. 107-116. Power Electronics Centre Annual Conference 5-6 July 2016 14 Clech Algorithm Results / K DT T0 • FEA is used to simulate the stress and strain • Extract the average stress and strain at the edge of solder joint model. • Fit the stress reduction line equation to the data. • Mean values of K and D at different temperature is used. 450 400 T(K) 350 300 250 200 0 5000 10000 Stress-strain for a range of solder thickness t(s) Regular temperature loading Power Electronics Centre Annual Conference 5-6 July 2016 15 Clech vs. FEA for Different dT Temperature loading: • • • Effect of temperature range on damage Trend quite similar Approximate linear relationship 0.35 0.18 0.30 0.16 ΔWClech(MPa) ΔW(MJ/m3) 0.20 0.15 Clech 0.10 Tmax(C) Tmin (K) Tmax(K) dT -40 125 233.15 398.15 165 -25 110 248.15 383.15 135 -10 95 263.15 368.15 105 5 80 278.15 353.15 75 y = 0.7033x - 0.0379 0.14 FEA 0.25 Tmin(C) 0.12 0.1 0.08 0.06 0.04 0.05 0.02 0 0.00 50 70 90 110 130 150 170 190 0.00 ΔT Power Electronics Centre Annual Conference 5-6 July 2016 0.05 0.10 0.15 0.20 0.25 0.30 0.35 ΔWFEA(MPa) 16 Global CTE Mismatch • Effect of baseplate CTE • Substrate CTE: 7.5x10-6/°C Substrate CTE=10ppm/K • As global CTE mismatch diminishes, Clech Algorithm seems to fail Power Electronics Centre Annual Conference 5-6 July 2016 17 New. Time Domain Damage Model Damage rate: • New features Visco-plastic strain and microstructure evolution: – Effect of plastic strain distribution – Effect of damage history – Effect of temperature Crack length vs. cycles Power Electronics Centre Annual Conference 5-6 July 2016 18 PowerLife GUI Power Electronics Centre Annual Conference 5-6 July 2016 19 Summary and Future Work • Fast physics of failure reliability analysis tool • Stress analysis and lifetime models implemented to predict wirebond/solder failures • Future work – Continued development of the software – Collection of data and models – New stress analysis functionalities for IGBT and other semiconductor devices – Capacitor analysis and physics of failure lifetime model – Adding design optimization functionality – System level optimization for reliability linking in with DTM cross cutting projects – More integration with external design and analysis software Power Electronics Centre Annual Conference 5-6 July 2016 20