Guaranteeing Reliability with Vibratin Simulatin and Testing Dr. Nathan Blattau
. Nathan Blattau, Ph.D. - Senir Vice President Has been invlved in the packaging and reliability f electrnic equipment fr mre than ten years. His specialties include best practices in design fr reliability, rbustness f Pb-free, failure analysis, accelerated test plan develpment, finite element analysis, slder jint reliability, fracture, and fatigue mechanics f materials. 2 9000 Virginia Manr Rd. Ste. 290, Beltsville MD 20705 301-474-0607 www.dfrslutins.cm
Vibratin Fatigue High cycle fatigue (HCF) due t mechanical stress induced by vibratin Millins f cycles t failure Small changes in stress have large impacts n time t failure Accrding t U.S Air-Frce statistics 20 percent f all failures bserved in electrnic equipment are due t vibratin prblems Steinberg D.S. Vibratin analysis fr electrnic equipment. Jhn Wiley & Sns, 2000. 3
Designing in Reliability, Earlier is Cheaper Reduce Csts by Imprving Reliability Upfrnt 4 4
Vibratin Fatigue Lifetime under mechanical cycling is divided int tw regimes Lw cycle fatigue (LCF) High cycle fatigue (HCF) LCF is driven by inelastic strain (Cffin-Mansn) 2N p f c -0.5 < c < -0.7; 1.4 < - 1 /c > 2 HCF is driven by elastic strain (Basquin) f b e N f E 2-0.05 < b < -0.12; 8 > - 1 /b > 20 f 5 5
Vibratin Harmnic and Randm Single frequency Randm vibratin is a cntinuus spectrum f frequencies MIL-STD-810G AN INTRODUCTION TO RANDOM VIBRATION Tm Irvine 6
Mechanical Lads (Vibratin) Expsure t vibratin lads can result in highly variable results Vibratin lads can vary by rders f magnitude (e.g., 0.001 g 2 /Hz t 1 g 2 /Hz) Time t failure is very sensitive t vibratin lads (t f W 4 ) Very brad range f vibratin envirnments MIL-STD-810 lists 3 manufacturing categries, 8 transprtatin categries, 12 peratinal categries, and 2 supplemental categries 7
Typical Vibratin Levels Harmnic Steinberg D.S. Vibratin analysis fr electrnic equipment. Jhn Wiley & Sns, 2000. Randm MIL-STD-810G Figure 514.6C-1 US Highway truck vibratin expsure 1 hur is equivalent t 1000 miles 8
Excessive Vibratin Randm Vibratin 9.8 t 28 Grms 0.07 t 0.5 G 2 /Hz Natural Frequency 72 Hz Results With BGA s, SnPb slder always utperfrmed leadfree Less cnclusive fr leadless/leaded parts All BGA-225 Wdrw, IPC/APEX 2006 9
Fatigue Expnent High Cycle Fatigue? N N field test test field c Assume the slder strain is directly prprtinal t the bard level strain Cffin Mansn Fit t data prvides c = 1.5 Expnent value is t lw; representative f lw-cycle fatigue High-cycle fatigue expnent is typically 4 t 6 r higher MIL-STD-810, Steinberg Lw cycle fatigue behavir can nt be extraplated t HCF behavir 10
Thrugh Hle Slder Intercnnect Vibratin Fatigue HCF failures typically ccur in the lead r slder jint Cmpnent Mtin Bard bending 11
Surface Munt Vibratin Fatigue Bard Bending SnPb failure @ 1200 µε SnAgCu failure @ 1200 µε 12
SnPb Fatigue Crack Very fine cracks Secndary micrcracks Evidence f phase carsening 13
SnAgCu Fatigue Crack Shrinkage Crack Well defined crack path Shrinkage crack prvided crack initiatin site 14
Using Simulatin
Sherlck Design Analysis Sftware 16
Parts List 17
Bundary Cnditins (Munt Pints) 18
Bard Prperties (Stackup) 19
Test Vehicles and Mdeling Finite Element Analysis in Sherlck 20
Sherlck - Mdal Analysis 1st Natural Frequency: 159.45 Hz 21
Mdal Analysis One accelermeter lcated at the center misses 3 frequency mdes 22
Bard Strain Analysis Bard level strain (at cmpnents) Green: predicted (peak values) Purple: experimental (averaged ver strain gage regin) 23
Bard r Cmpnent Mtin - FEA 24
Simulatin During Test Plan Develpment Simulatin Identify critical cmpnents prir during plan develpment Identify critical frequencies r bard respnses Identify lcatins fr accelermeter placement Munting cnsideratin Bundary cnditins Munting cnfiguratins 25
HALT Example - Vibratin HALT chambers utilize repetitive shck (RS) t generate vibratin input Type f randm vibratin Pneumatic hammers strike the chamber table a generate the vibratin HALT vibratin is nt designed t replicate field envirnments Designed t expse weak links Rapid assessment Input/Output is typically displayed as a Grms value What des this mean? Is it suitable fr simulatin, can it be used in Sherlck? 26
HALT Chamber Settings HALT vibratin is specified by a Grms level What des this represent The rt mean square acceleratin (Grms) is the square rt f the area under the ASD (acceleratin spectral density) curve in the frequency dmain The Grms value is typically used t express the verall energy f a particular randm vibratin event Can it be used fr simulatin? N A straight Grms value can represent an infinite number f acceleratin and frequency cmbinatins The time histry f the table excitatin must be captured and prcessed int a usable frmat fr the simulatin 27
HALT - Repetitive Shck Induced Vibratin 28 Input HALT level set t 33 Grms
PSD (G 2 /Hz) Vibratin Data Pst Prcessing FFT perfrmed n the time histry data t generate a PSD (pwer spectral density) 1 0.1 24 Grms 0.01 0.001 0.0001 0.00001 Qualmark Chart/Hanse Respnse belw 5KHz 0.000001 0.0000001 5.7 Grms 0 20000 40000 60000 80000 100000 120000 Frequency Even thugh bth chambers are prgrammed t utput 16 Grms the PSD prfile is higher acrss the full frequency range fr the Qualmark, the cntent belw 5KHz used as Sherlck inputs Knw yur chamber! HASS r HALT prfiles can be very different 29
FFT f the Time Histry Different FFT parameters 32.8 Grms 32.7 Grms Frequency 1.20E+01 1.00E+02 7.84E+02 1.57E+03 2.35E+03 3.14E+03 3.92E+03 4.71E+03 5.49E+03 6.27E+03 7.06E+03 7.84E+03 8.63E+03 9.41E+03 1.02E+04 1.10E+04 PSD 7.50E-05 2.00E-04 4.81E-02 5.33E-02 1.54E-01 2.52E-01 2.78E-01 1.39E-01 9.59E-02 6.18E-02 2.48E-02 3.70E-02 4.55E-02 6.39E-02 9.85E-02 1.06E-01 30
HALT Vibratin HALT vibratin can cause high cycle fatigue f cmpnents Simulatin can identify critical lcatins fr accelermeter placement Simulatin can als identify critical cmpnents befre HALT testing Simulatin prir t HALT requires the develpment f a suitable vibratin prfile Leverage prir HALT vibratin data Prduct shuld have similar mass and munting Need the time histry data (time verses acceleratin) Data shuld include table utputs fr multiple Grms levels 31
Bundary Cnditins (Munting) 32
Pre-HALT Example X,Y,Z RS vibratin HALT prfile cnverted t PSD prfile 33
HALT Vibratin - Mdal Analysis Identify lcatins fr accelermeters Identify critical cmpnents 216.14 Hz 263.15 Hz 34
Accelermeter Lcatins Sherlck supprts Virtual Accelermeters The mdal analysis identifies the regins f greatest bard respnse which can be used t place the HALT accelermeters Thrugh hle crystal scillatr may need t be staked prir t HALT Nf = 263.2 Hz 35
Prduct Respnse t Multiply Axis Vibratin 36
Accelermeter Respnse 37
Critical Cmpnents Excessive lead stains (1.4%) will lead t a rapid HALT failure X1 needs t be staked 38
During the HALT Planning Stage Sherlck identifies critical cmpnents and they are mitigated befre HALT is perfrmed Different munting cnfiguratins are investigated in Sherlck and mdified t prevent unrealistic mvement during the HALT test The hard questins Hw much f a margin d I need What is a gd vibratin level, depends n the prduct and use 39
High Cycle Vibratin Fatigue Testing - Example Test time fr HCF usually take days t weeks Type f vibratin applied depends n the gals Randm vibratin prduct qualificatin testing Harmnic vibratin fundamental understanding f fatigue behavir (SN curves) Example - Harmnic vibratin f 208 I/O BGAs Resnant frequency: ~160 Hz Testing at ~153Hz SAC 305, SnPb 80 mils, 90 mils, 95 mils, 105 mils 6.8 Gs, 7.8 Gs, 8.3 Gs, 9.8 Gs 40
High Cycle Fatigue Testing High cycle fatigue testing can take weeks n a electrdynamic shaker 41
Analysis One Accelermeter lcated at the center f the PCB, Frequency sweep 20 500 Hz 42
Experimental - Cycles t Failure Stepwise failure behavir characteristic f differing stress levels alng bard length 43
Predictin (Steinberg Displacement Based) Step 1: Calculatin f maximum deflectin (Z 0 ) Z 0 9.83 PSD 2 f 2 n f n Q Randm PSD is the pwer spectral density (g 2 /Hz), f n is the natural frequency f the CCA, G in is the acceleratin in g Q is transmissibility (assumed t be square rt f natural frequency) Harmnic Z 0 9.8G f in 2 n Q 44
Predictin (Steinberg) Step 2: Calculate critical displacement B is length f PCB parallel t cmpnent c is a cmpnent packaging cnstant 1 t 2.25 h is PCB thickness r is a relative psitin factr 1.0 when cmpnent at center f PCB L is cmpnent length 0. 00022B Z c chr L At critical displacement, cmpnent can survive a minimum 20 millin cycles under randm vibratin 10 millin cycles fr harmnic 45
Predictin (Steinberg) Step 3: Life calculatin N c is 10 r 20 millin cycles Several assumptins CCA is simply supprted n all fur edges N N Z c 0 c Z0 Mre realistic supprt cnditins, such as standffs r wedge lcks, can result in a lwer r higher displacements Chassis natural frequency differs frm the CCA natural frequency by at least factr f tw (ctave) Prevents cupling Vibratin ccurs at rm temperature Depending upn the cnfiguratin and lading, vibratin at lwer r higher temperatures can increase/decrease lifetime Des nt cnsider the influence f in-plane displacement (i.e., tall cmpnents) r cmpnents lcated at areas f high curvature 6.4 46
Example Predictin (Steinberg) 208 I/O BGA, 1.6 mm thick PCB, 159 Hz Distance between standff Y = 76.2 mm X = 152.4 mm = 6 inches Package dimensins 14.4 x 14.4 mm Center cmpnent (Critical deflectin) Z c = 0.00022 * 6 / (1.75 * 0.062 *1*SQRT(0.567)) = 0.0162 inches peak N c = 10 millin cycles t failure Bard deflectin G in = 9.8 g s Z = 9.8*9.8*SQRT(159)/(159^2) = 0.0479 inches 95.8 mils peak t peak, clse t measured value f 105 mils Cycles t Failure 0.0162 N 10,000,000 0.0479 6.4 0 9700 0. 00022B Z c chr L Z 0 9.8G f N N in 2 n Z Q c 0 c Z0 6.4 47
High Cycle Fatigue (Predictin) During vibratin, bard-level strain is prprtinal t slder r lead strains and therefre can be used t make time-t-failure predictins Requires cnverting cycles-t-failure displacement equatins (Steinberg) t use strain The critical strain fr the package types is a functin f package style, size, lead gemetry N 0 c c N c L ζ is analgus t 0.00022B but mdified fr strain c is a cmpnent packaging functin L is cmpnent length c 0 n 48
FEA Failure Predictin During vibratin the bard strain is prprtinal t the slder r lead strains and therefre can be used t make time t failure predictins This requires cnverting the cycles t failure displacement equatins (Steinberg) t use strain The strain fr the cmpnents is nw pulled frm the FEA results The critical strain fr the package types is a functin f package style, size, lead gemetry ζ is analgus t 0.00022B but mdified fr strain c is a cmpnent packaging functin L is cmpnent length n N 0 N c c c 0 c L 49
Sherlck Vibratin Fatigue Predictin 105 mils Deflectin Steinberg using measured deflectin Steinberg using cmputed deflectin 50
Sherlck Vibratin Fatigue Predictin 90 mils Deflectin Steinberg using measured deflectin 51
Cnclusins Finite element based simulatins fr vibratin can capture: Cmplex bundary cnditins and whether the PCB is adequately supprted Cmplex mde shapes, mdal analysis where d yu put the accelermeters during test Facilitate strain r curvature based predictins Displacement based methds (Steinberg) High displacement des nt always indicate high stress Only based n the natural frequency f the bard Utilizes Miles equatin t equivalence randm vibratin t harmnic Can prvide very cnservative results Experimental measurements can be used t validate simulatins and verify respnse, nt typically dne t failure but fr qualificatin purpses Cmbining simulatins and fatigue predictins with qualificatin test results prvides the mst value Is the vibratin respnse as expected Cnfidence in the reliability predictins 52
Any Questins 53