This document discusses physics of failure (PoF) reliability assurance software. It begins with an introduction to design for reliability (DfR) and the history and limitations of traditional DfR approaches. It then discusses how PoF is now important for modeling wearout failures. The document provides examples of modeling solder joint fatigue, vibration analysis, and shock environments. It also discusses validation of the software predictions against experimental data. In summary, the software allows modeling of reliability from multiple failure mechanisms to provide assurance that design goals are met.
Physics of Failure Electronics Reliability Assurance Software
1. Physics of Failure Electronics Reliability Assurance Software
Cheryl Tulkoff, Nate Blattau, & Randy Schueller
Senior Members of the Technical Staff
at DfR Solutions
IPC APEX EXPO 2010
2. Design for Reliability (DfR)
â˘DfR: A process for ensuring the reliability of a product or system during the design stage before physical prototype
â˘Reliability: The measure of a productâs ability to
ââŚperform the specified function
ââŚat the customer (with their use environment)
ââŚover the desired lifetime
3. History
â˘DfR has been a concept promoted by electronics community since the early 1950âs
â˘DARPA identified DfR as an âArea of Promiseâ to resolve issue with Defense Systems Reliability in 1958
Identification of Certain Current Defense Problems and Possible Means of Solution, INSTITUTE FOR DEFENSE ANALYSES, 1958
4. Why DfR?
Architectural Design for Reliability, R. Cranwell and R. Hunter, Sandia Labs, 1997
5. Limitations of Current DfR
â˘Too broad in focus (not electronics focused)
â˘Too much emphasis on techniques (e.g., FMEA and FTA) and not answers
âFMEA/FTA rarely identify DfR issues because of limited focus on the failure mechanism
â˘Overreliance on MTBF calculations and standardized product testing
â˘Incorporation of HALT and failure analysis (HALT is test, not DfR; failure analysis is too late)
âFrustration with âtest-in reliabilityâ, even HALT, has been part of the recent focus on DfR
6. DfR and Physics of Failure (PoF)
â˘Due to some of the limitations of classic DfR, there has been an increasing interest in PoF (aka, Reliability Physics)
â˘PoF Definition: The use of science (physics, chemistry, etc.) to capture an understanding of failure mechanisms and evaluate useful life under actual operating conditions
7. Why PoF is Now Important
Failure Rate
Time
Electronics: 1960s, 1970s, 1980s
No wearout!
Electronics: Today and the Future
Wearout!
8. Solder Joint (SJ) Wearout
â˘Elimination of leaded devices
âProvides lower RC and higher package densities
âReduces compliance
Cycles to failure -40 to 125C
QFP: >10,000
BGA: 3,000 to 8,000
QFN: 1,000 to 3,000
CSP / Flip Chip: <1,000
9. SJ Wearout (cont.)
⢠Design change: More silicon, less plastic
⢠Increases mismatch in coefficient of thermal
expansion (CTE)
BOARD LEVEL ASSEMBLY AND RELIABILITY
CONSIDERATIONS FOR QFN TYPE PACKAGES,
Ahmer Syed and WonJoon Kang, Amkor Technology.
10. Reliability Assurance -- Definition
â˘Reliability is the measure of a productâs ability to
ââŚperform the specified function
ââŚat the customer (independent of environment)
ââŚover the desired lifetime
â˘Assurance is âfreedom from doubtâ
âConfidence in your productâs capabilities
â˘Typical approaches to reliability assurance
ââGut feelâ
âEmpirical predictions (MIL-HDBK-217, TR-332)
âIndustry specifications
âTest-in reliability
â˘Sherlock is a reliability assurance software based upon physics of failure algorithms
11. Motivation
â˘Ensuring sufficient product reliability is critical
âMarkets lost and gained
âReputations can persist for years or decades
âHundreds of millions of dollars won and lost
â˘Designing in Reliability before prototype build & test
âSaves costs
âReduces development time
â˘Opportunities for improvement, automotive example:
âTotal warranty costs range from $75 to $700 per car
âFailure rates for E/E systems in vehicles range from 1 to 5% in first year of operation
â˘Hansen Report (April 2005)
âDifficult to introduce drive-by-wire, other system-critical components
â˘E/E issues will result in increase in âwalk homeâ events
12. Other Costs of Failure
Type of Business Lost Revenue per Hour
Retail Brokerages $6,450,000
Credit Card Sales Authorization $2,600,000
Home Shopping Channels $113,750
Catalog Sales Centers $90,000
Airline Reservation Centers $89,500
Cellular Service Activations $41,000
Package Shipping Services $28,250
Online Network Connect Fees $22,250
ATM Service Fees $14,500
Supermarkets $10,000
Does not include liability and loss of market share
13. Reliability and Design
â˘The foundation of a reliable product is a robust design
âProvides margin
âMitigates risk from defects
âSatisfies the customer
14. Currently Available DfR Tools
â˘FMEA â many limitations
â˘MIL-HNBK-217 MTTF Calculations â also many limitations (no solder joint considerations)
â˘FEA modeling â good but often expensive and limited to a few components.
â˘Sherlock â a new tool that models all the circuit cards assemblies and provides predicted life curves from many failure mechanisms.
15. Limitations of MTTF/MTBF
â˘MTBF/MTTF calculations tend to assume that failures are random in nature
âProvides no motivation for failure avoidance
â˘Easy to manipulate numbers
âTweaks are made to reach desired MTBF
âE.g., quality factors for each component are modified
â˘Often misinterpreted
â50K hour MTBF does not mean no failures in 50K hours
â˘Better fit towards logistics and procurement, not failure avoidance
16. Sherlock Coverage
â˘This software modeling tool predicts failures from
âSolder joint wear-out from thermal cycling (SAC305 or SnPb)
âConductive anodic filament formation
âPlated through hole fatigue
â217 MTBF calculations are also generated
â˘In addition the software uses FEA to determine
âBoard deflection from mechanical shock
âBoard deflection from vibration
âThe natural frequencies for the board based on the mount points.
17. Process Overview
â˘There are several high levels steps involved in running the software (named Sherlock). They are:
âCreate a Project
âDefine Reliability Goals
âDefine Environments
âAdd Circuit Cards
â˘Import Files
â˘Generate Inputs
âPerform Analysis
âInterpret Results
18. Inputs
â˘Gerber or ODB files for PCB and Pick & Place (w/ BOM)
â˘Thermal cycle conditions (Minerâs Rule is applied) â in the field or in test.
â˘Shock & Vibration conditions.
Layer Plot Examples
19. Identify Field Environment
â˘Approach 1: Use of industry/military specifications
âMIL-STD-810,
âMIL-HDBK-310,
âSAE J1211,
âIPC-SM-785,
âTelcordia GR3108,
âIEC 60721-3, etc.
â˘Advantages
âNo additional cost!
âSometimes very comprehensive
âAgreement throughout the industry
âMissing information? Consider standards from other industries
â˘Disadvantages
âMost more than 20 years old
âAlways less or greater than actual (by how much, unknown)
20. Field Environment (cont.)
â˘Approach 2: Based on actual measurements of similar products in similar environments
âDetermine average and realistic worst- case
âIdentify all failure-inducing loads
âInclude all environments
â˘Manufacturing
â˘Transportation
â˘Storage
â˘Field
21. Field Environment (example)
â˘For automotive electronics outside the engine compartment with minimal power dissipation, the diurnal (daily) temperature cycle provides the primary degradation-inducing load
â˘Absolute worst-case: Max. 58ÂşC, Min. -70ÂşC
â˘Realistic worst-case: Phoenix, AZ (USA)
âAdd +10ÂşC due to direct exposure to the sun
Month Cycles/Year Ramp Dwell Max. Temp (oC) Min. Temp. (oC) Jan.+Feb.+Dec. 90 6 hrs 6 hrs 20 5 March+November 60 6 hrs 6 hrs 25 10 April+October 60 6 hrs 6 hrs 30 15 May+September 60 6 hrs 6 hrs 35 20 June+July+August 90 6 hrs 6 hrs 40 25
23. Solder Joint Fatigue
â˘Two most common solder types are available.
âEutectic tin-lead (SnPb)
âLead-free SAC 305 (Sn-3.0%Ag-0.5%Cu)
âAdditional solders may be added in the future
âSpecified at the board or component level
25. Validation
Example
QFN
Solder Material Cycles to Failure (calc) Cycles to Failure (exprm) Stress Strain Energy Name
Tin-Lead 496 631 2.28E+01 3.326 QFN-52 Tee, Ng, Yap, Zhong
Lead-Free 7938 7800 3.639 6.63E-02 HVQFN-24 de Vries, Jansen, van Driel
Lead-Free 9079 5250 2.828 5.80E-02 HVQFN-48 de Vries, Jansen, van Driel
Lead-Free 3366 4500 5.528 0.4021 HVQFN-72 de Vries, Jansen, van Driel
Tin-Lead 2463 1635 8.932 0.67 QFN-44 Tee, Ng, Yap, Zhong
Tin-Lead 976 2015 17.76 1.702 QFN-36 Tee, Ng, Yap, Zhong
Tin-Lead 956 2165 19.36 1.725 QFN-28 Tee, Ng, Yap, Zhong
Tin-Lead 3542 2928 10.23 0.4658 QFN-20 Tee, Ng, Yap, Zhong
Lead-Free 1437 1280 10.04 0.3663 QFN-40 Mukadam, Meilunas, et al
Lead-Free 1448 2063 10.92 0.3635 QFN-42 Mukadam, Meilunas, et al
Lead-Free 3651 803 5.565 0.1442 QFN-44 Mukadam, Meilunas, et al
Tin-Lead 760 947 16.77 2.17 QFN-20 Zhang and Lee & Kim, Han, et al
Solder Properties Package Properties
Author(s)
QFN Sherlock Validation Profile
100
1000
10000
100000
100 1000 10000 100000
Predicted
Experimental
26. Validation
BGA
BGA Sherlock Validation Graph
100
1000
10000
100000
100 1000 10000 100000
Predicted
Experimental
Large scatter in data is typical of
experimental results for BGAs
27. Assessment of IPC-TR-579
â˘Based on round-robin testing of 200,000 PTHs
âPerformed between 1986 to 1988
âHole diameters (250 Îźm to 500 Îźm)
âBoard thicknesses (0.75 mm to 2.25 mm)
âWall thickness (20 Îźm and 32 Îźm)
â˘Advantages
âAnalytical (calculation straightforward)
âValidated through testing
â˘Disadvantages
âNo ownership
âValidation data is ~18 years old
âUnable to assess complex geometries (PTH spacing, PTH pads)
â˘Complex geometries tend to extend lifetime
âDifficult to assess effect of multiple temperature cycles
â˘Can be performed using Minerâs Rule
â˘Software conducts calculations for all plated through holes and thermal cycles (combined using Minerâs Rule)
28. Vibration Environment
Number of natural frequencies to look for within the desired frequency range
Single point or frequency sweep loading Techniques are available for equivalence random vibration to harmonic vibration
29. Vibration (cont.)
â˘Vibration loads can be very complex
âSinusoidal (g as function of frequency)
âRandom (g2/Hz as a function of frequency)
âSine over/on random
â˘Vibration loads can be multi-axis
â˘Vibration can be damped or amplified depending upon chassis/housing
âTransmissibility
â˘Response of the electronics will be dependent upon attachments and stiffeners
â˘Peak loads can occur over a range of frequencies
âStandard range: 20 to 2000 Hz
âUltrasonic cleaning: 15 to 400 kHz
30. Vibration (cont.)
â˘Failures primarily occur when peak loads occur at similar frequencies as the natural frequency of the product / design
â˘Natural frequencies
âLarger boards, simply supported: 60 â 150 Hz
âSmaller boards, wedge locked: 200 â 500 Hz
âGold wire bonds: 2k â 4kHz
âAluminum wire bonds: >10kHz
31. Mechanical Loads (Vibration)
â˘Exposure to vibration loads can result in highly variable results
âVibration loads can vary by orders of magnitude (e.g., 0.001 g2/Hz to 1 g2/Hz)
âTime to failure is very sensitive to vibration loads (tf ďľ W4)
â˘Very broad range of vibration environments
âMIL-STD-810 lists 3 manufacturing categories, 8 transportation categories, 12 operational categories, and 2 supplemental categories
32. Interpretation (Vibration)
â˘SAC is âstifferâ than SnPb
âFor a given force / load, it will respond with a lower displacement / strain (elastic and plastic)
â˘Low-cycle fatigue (plasticity driven)
âUnder displacement-driven mechanical cycling, SnPb will tend to out-perform SAC (e.g., chip scale packages [CSP])
âUnder load-driven mechanical cycling, SAC will tend to out-perform SnPb (e.g., leads of thin scale outline packages [TSOP])
â˘High-cycle fatigue (elasticity driven)
âStiffer solder (i.e., SAC), lower strain range
33. Vibration Software
Implementation
c L
c
ďş
ďĽ ď˝
⢠The software uses the finite element results for board
level strain in a modified Steinberg like formula that
substitutes the board level strain for deflection and
computes cycles to failure
⢠Critical strain for the component
Îś is analogous to 0.00022B but modified for strain
c is a component packaging constant, 1 to 2.25
L is component length
38. Environments (Mechanical Shock)
â˘Initially driven by experiences during shipping and transportation
â˘Increasing importance with use of portable electronic devices
âA surprising concern for portable medical devices
âFloor transitions (1 to 5 inch âdropâ)
â˘Environmental definitions
âHeight or G levels
âSurface (e.g., concrete)
âOrientation (corner or face; all orientations or worst- case)
âNumber of drops
39. Software Shock
â˘Implements Shock based upon a critical board level strain
â˘Will not predict how many drops to failure
â˘Either the design is robust with regards to the expected shock environment or it is not
â˘Additional work being initiated to investigate corner staking patterns and material influences
45. Possible Actions
â˘Based on the reliability assessment one may decide to increase reliability by:
âChanging package types
âChanging location of components
âChanging the mount point locations
âIncreasing Cu thickness in PTHs
âEtc.
â˘Trial and error can be used on the virtual board
â˘The software can also be used to determine the TC test conditions that best simulate the field use conditions.
46. Reliability Assurance Tool
â˘This powerful software tool uses the principles of PoF to predict the life of CCAs prior to prototypes being built.
â˘Optimization of the design layout can now take place early in the design cycle which greatly improves the chances of designing it right the first time.