reliability of power modules using sherlock

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com© 2004 – 2010

Reliability of Power Modules Using Sherlock

DfR WebinarJune 16, 2014

Greg Caswell and Craig Hillman

DfR Solutions, LLC

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com

o Introduction

o Power PCB Applications

o Common Issues

o Lifetime Expectations

o Failure Mechanisms

o Virtual Qualification Approach

o Sherlock Solution

Agenda


Power Modules Are Used in Several Market Segments

Thermoelectric Modules

Voltage Power Modules

Solar Power Modules

Automotive Power Modules

200W Power Amp

IGBT

Switching Power Supply


o High Temperature Environments

o Possible Vibration and Shock Environments

o Temperature and Power Cycling Environments

o Very High Current Flows and Thermal Transfer Requirements

o A variety of materials forming the product

o Substrate tiles bonded to copper baseplate

What Do They All have in Common?


o IGBT – Rail application – 30 years (Each module 100FIT)

o Power Module – Automotive Application – 20 years

o 10W/cm2

o DBC Substrate bonded to heatsink

o Vibration, shock, humidity, salt spray

o Cost

o Solar Power Inverters-25 years

Example Life Expectancies


Semicron Thermal Module


o Thermo-mechanical fatigue induced failureso CTE mismatcho Temperature swings

o Bond Wire Fatigueo Shear Stresses between bond pad and wireo Repeated flexure of the wireo Lift off (fast temperature cycling effect)o Heel Cracking

o Die Attach Fatigueo Solder Fatigue

o Voids

o Device Burn Outo Automotive- degradation of power

o Solder Fatigueo Bond wire failure (lift off due to fast temperature cycling)

o Structural Integrity – ceramic substrate to heat sink in thermal cyclingo IGBTs – solder joint fatigue, wirebond liftoff, substrate fracture, conductor

delamination

Failure Mechanisms


Bond Wire Fatigue Due to Thermal Effects

Bredtmann, et al, “Options for Electric Power

Steering Modules a Reliability Challenge.”

Automotive Power Electronics, September 2007


o After 100 cycles of -55 to 200C – DBC Delamination

o By 1000 cycles there were cracks in AlN substrate and extensive solder joint failures

Example of Substrate Delamination

Scofield, Richmond and Leslie, ”Performance and Reliability Characteristics of

1200V,100A 200C Half Bridge SiC MOSFET-JBS Diode Power Modules,”

IMAPS -International Conference on High Temperature Electronics

May 2010


Failure Modes- Solder and Silicon Cracking

Mitsubishi, “Power Module Reliability”

Cracks between DBC Substrate and also between silicon die and bond wire


o Examples of wire bond fatigue cracking and also wire bond lift off

Failure Modes - Wire Bond Cracking and Lift Off

Dynex – AN5945 – IGBT Module Reliability


o The stress conditions in the chart are for a railroad braking application

Typical Mission Profile


o Typically, extensive qualification testing is performed to ascertain the reliability of the power module as shown

IGBT Qualification Tests-Environmental


o Power module industry believes copper wire is more robust than aluminumo Changes being implemented for electric drivetrain

o Part of improvement is believed to be due to reduced temperature variation from improved thermal conductivity

o Part of improvement could be due to recrystallizationo Can result in self-healing

o Part of improvement could be more robust fatigue behavior

Copper Wire and Temperature Cycling

D. Siepe, CIPS 2010

N. Tanabe, Journal de Physique IV, 1995


o Copper clearly superior

Aluminum vs. Copper – Temperature Cycling

10

100

106 108107 109

N. Tanabe, Journal de Physique IV, 1995J. Bielen, EuroSime, 2006


Thermal Aging of Cu Wire Bonds vs. Gold

J. Onuki, M. Koizumi, I. Araki. IEEE Trans. On Comp. Hybrids & Manfg. Tech.

12 (1987) 550


o Cu is comparable in cost to aluminum but less proven –used on low cost products (not those where the cost of the IC is much greater than the package).

o Cu bonding is slower (5 wires/sec) so that adds process cost if high I/O

o Pd coating helps but adds cost

Points


o Palladium (Pd) coating creates galvanic couple with copper

o Studies have demonstrated thinning or loss of Pd coating during bonding

o Uncertain if JEDEC test with acceleration factor based on Peck’s equation (based on aluminum/gold galvanic couple) is still valid

o Push out of aluminum pad

o Could result in subsurface cracking (metal migration?)

o Uncertain if existing JEDECtemp cycling test is sufficient todrive crack growth

Major concerns identified by DfR


o Failure mechanisms

o CTE mismatch resulting in plastic strain

o Thermo-mechanical fatigue as a result of temperature cycling

o Coarsening

Die Attach Fatigue


Typical Thermal Stress Failures in a Die-Substrate Assembly

Die-Substrate Assembly

Chip, E1,α1

Substrate, E2,α2

Adhesive, E0, α0

Crack at the chip’s corner is due to the

interfacial stresses

Crack at the chip’s surface in its mid-portion

is due to the normal stresses in the chip

Crack/delamination at the

adherend/adhesive interface (adhesive

failure of the bonding material)

Is due to the interfacial stresses

Crack in the body of the adhesive (cohesive failure)

is due to the interfacial stresses


Typical Failure Modes in Die-Substrate and Similar Assemblies

� Typical failure modes in die-substrate assemblies are:

1) adherend (die or substrate) failure: a silicon die can fracture in itsmidportion or at its corner located at the interface;

2) cohesive failure of the bonding material (i.e., failure of the die-attachmaterial); and

3) adhesive failure of the bonding material (i.e., failure at theadherend/adhesive interface).

� An adhesive failure is not expected to occur in a properly fabricatedjoint. If such a failure takes place, it usually occurs at a very low loadlevel, at the product development stage, and should be regarded as amanufacturing or a quality control failure, rather than a material’s or astructural one.


Die Attach Solder Reliability

Marie Curie ECON2 2008


Sherlock

o User Friendly

o Quick

o Flexible

o Intuitive

o Reliable

o One of a Kind

o State of the Art


Why Sherlock

o Mil-HBK-217 actuarial in nature

o Physics based algorithms to time consuming

o Need to shorten NPI cycles and reduce costs

o Increased computing power

o Better way to communicate


PoF: The Complexity Roadblock

−

=

211

2

2

1 11exp

TTK

E

V

V

t

t

B

a

n

)%063.0exp(~51.0~

exp RHkT

eVT f −×

∝

( )

⋅

−++++⋅=⋅∆⋅−

aGGA

h

GA

h

AE

L

AE

LFLT

bcc

c

ss

s

9

2

221112

ναα


Traditional Iterative NPI Cycle


NPI Cycle Using PoF Modeling


Why DfA? Total Costs are Determined During Design

95% of the O&S Cost Drivers are Based on Decisions Made during Design.

Source: Architectural Design for Reliability, R. Cranwell and R. Hunter, Sandia Labs, 1997

29


Concurrent Engineering


o The foundation of a reliable product is a robust designo Provides margin

o Mitigates risk from defects

o Satisfies the customer

Introduction


o What is Sherlock?

o How will Sherlock help you?

o How is Sherlock unique?

o What can Sherlock do?

o How can Sherlock solve my design challenges?

o How does Sherlock work?

o Conclusion


What is Sherlock?

o Physics of Failure Design Reliability Analysis Tool

o Predicts product failure early in design process, quickly and accurately

o Electronics-focused –Used across all industries


How will Sherlock help me?

o Save time, money, resources

o Produce better, more reliable products

o Reduce warranty costs

o Accelerate product development

o Increase profitability

o Enhance customer satisfaction

Deeper Insight, Earlier in the Design Process

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com35

ICWearout

Thermal Cycling

Shock/Vibe

PTH Fatigue

Solder Fatigue

3D FEA

DFMEA

ICTMTB

F

MTB

F

Hi-Fidelity

ICWearout

Thermal Cycling

Shock/Vibe

PTH Fatigue

Solder Fatigue

3D FEA

DFMEA

ICTMTB

F

Hi-Fidelity

What Can Sherlock Do?


How Does Sherlock Work?

Phase 1: Data Input

o Parses standard EDA files (schematic, layout, parts list) automatically

o Uses embedded libraries (part, package, materials, solder, laminate)

o Can build box-level finite element analysis model in minutes

Phase 2: Sherlock Analysis

o Produces holistic analysis critical to develop reliable products

o Easy to assign and create standard structures and conditions

o Assessment options include: Thermal Cycling, Mechanical Shock, Natural Frequency, Harmonic Vibration, Random Vibration, Bending, Integrated Circuit Wearout, Thermal Derating, Failure Rate, Conductive Anodic Filament, High Fidelity PCB Model

Phase 3: Report & Recommend

o Presents results in multiple formats: Tabular / Histogram / Life curve / Overlay

36


Intuitive

• Easy-to-locate commands

• Industry terminology (parts list, stackup, pick & place, etc.)


Compatible with wide variety of reliability metrics

38


Handles very complex environments

39


Import Standard Design Output Files (Gerber/ODB)

40


Parts Listo Color coding of data origin

o Minimizes data entry through intelligent parsing and embedded package and material libraries

41


Stackupo Automatically generates stackup and copper percent (%)o Library with ~700 laminate materials with 48 different properties

42


Power Module materials Alumina, and Silicon Nitride in database

Die attach reliability is factored in for solder materials

Wire bonds are assessed using a separate calculator

43


Automated Mesh Generationo Identifies optimum mesh density based on board sizeo Expert user no longer required; model time reduced by 90%

44


Featureso Global Part Databaseo FEA 3D Modelo Sub-Assembly Analysis

o FEA 3D Viewero Result Managemento Component lead modeling

45


FEA 3D ModelingICT and Shock/Vibration Analysis

• Fully 3D elements for the PCB, components and mount points

• Increase simulation accuracy

• More reliable meshing algorithm

• Increased analysis flexibility

• Sub-assemblies, heat-sinks, chassis analysis

FEA Engine

• Multi-core and 64 bit support

• Faster analysis


Sub-Assembly Analysis

o Attach one or more CCA sub-assemblies to a primary CCAo Mezzanine cards supported by

standoffs

o Edge-connected cards

o Sherlock automatically analyzes the main CCA and all CCA sub-assemblies during a single ICT or Shock/Vibration analysis task

o Layer results and component results automatically generated for all circuit cards

47

DaughterCard

Mezzanine Card


o High Fidelity PCB

o Newest Feature

o Sherlock can identify and mesh every copper feature within a PCB or substrate

o Provides unrivaled insightinto risks due to warpage, thermal issues, mechanicalloads, etc.


Vibration/Shock/Bendingo Loading

o Mounting

o Direction

49


Thermal Cycling Fatigueo Cumulative Damage Index (CDI)

o Time to failure

o Thermal profile and Flowtherm results

50


Automated Thermal Deratingo Min / Max temperature assigned to each part (Operating and Storage)

o Sherlock automatically combines operating temperatures with thermal profiles

o Parts that exceed their ratings are flagged

51


DFMEAo The only Design Failure Mode Effects Analysis (DFMEA) software dedicated to electronics

52


Current DFMEA Tools (cont)

o Sherlock – DFMEA moduleo Module in Design Reliability Assessment tool

o Imports data from BOM or ODB++ design files (generated by PADS / Expedition)

o Later in design cycle than desired

o Earlier than typically being done today

o Facilitates integrating Reuse into DFMEA

o Highest level of automation – Standardized but customizable

o Common formatting – Can import / export to Excel

o Generates DFMEA by “Subcircuit” (Reuse Function)

o SEV / DET based on Standard Practices (Design Standards, etc.)

o OCC (Occurrence Rating)

o Based on component Reliability Predictions o Delphi Warranty Database

o Multiple Industry databases

o Future:

o Mission profile will drive reliability predictions / OCC


Baseline

Component Level DFMEA Generation

Schematic

Parts List

• Circuit Function / SubFunction / Sub-SubFunction

• Reference Designator

• Part Number

SherlockDesign Assistant

Delphi

Component

Database

(DMS)

Failure Rate ClassDelphi

DFMEA

Data Source

Failure Rate Class

• Failure Rate

• Failure Modes

• Fail Mode Distribution

Default:

• SEV, OCC, DET

• Failure Cause

• Failure Effects

• Preventions

• DetectionsBaseline DFMEA Assumes:

• Typical Application

• Not Safety Related

Description, properties


ICT Module (optional)

o Uses embedded FEA engine to compute board deflection and strain cause by ICT fixture


Results Summaryo Results tabulated into a scoreo Easy report generation

o PDF Formato Includes results and inputs

56


o Only software tool that provides a complete life curve

57

Constant Failure Rate Generic Actuarial MTBF Database

PTH Thermal Cycling Fatigue

Thermal Cycling Solder Fatigue

Vibration Fatigue

Over All Module

Combined Risk


o Fully Validated

58


Validation – Avionics Manufacturer

o “Following discussions with DfR and confirmation of the PCB material, from the supplier, I was able to refine the model for the QFN package and the PCB construction to predict the first failure of a QFN package at around 700 cycles.

It was interesting to note that the PCB material choice significantly altered Sherlock’s predicted solder joint life and choice of PCB material needs to be carefully considered from this perspective.

Subsequent real cycling of the test board has indeed produced a failure at around 770 temperature cycles and so appearing to add some (albeit limited) validation of the Sherlock prediction.

With the refined model failures of well soldered BGA joints were not predicted by Sherlock till around 3000 cycles. Our supplier has now finished the thermal cycles on the real boards seeing no BGA failures after about 1200 cycles.”

59


Validation – GPS Manufacturer

o With all material and design information received, Sherlock predicted failure of I200 (Actel Microcontroller) in just over 300 thermal cycles of -55C to 125C

60


Why Did I200 Fail Thermal Cycling?

o The Atmel device has two die

o This information is in the Atmel datasheet

o Results in a ball grid array (BGA) with a low coefficient of thermal expansion (CTE)

o The PCB was constructed with TU-622-5 laminate

o This material has a high coefficient of thermal expansion (CTE)

o This material is in the Sherlock laminate library

o Certain part packages (e.g., BGAs) are very sensitive to differences in CTE

61


Validation – Satellite Manufacturer

o DfR created an ad-hoc Sherlock project using only a picture and a component

o DfR predicted 1016 cycles to first failure and 1693 cycles to characteristic life

o Test data showed 750 cycles to first failure

o No stackup infoo Polyimide board

o Solder fatigue calculator was also used as a standalone tool


Summary - What is Physics of Failure (PoF)?

o Common Definition: o The process of using modeling and simulation based on the fundamentals of physical science (physics, chemistry, material science, mechanics, etc.) to predict reliability and prevent failures

o Mechanisms that can be modeled include fatigue, creep, diffusion, etc.

o The foundation of a reliable product is a robust designo Provides margino Mitigates risk from defectso Satisfies the customer

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com© 2004 – 2010

Thank You!Greg Caswell

Sr. Member of the Technical Staff

DfR Solutions

[email protected]