Life Prediction in Leadfree Solder Joint and PCB Metallization Under Shock and Vibration by Arjun Angral A thesis submitted to the Graduate Faculty of Auburn University in partial fulfillment of the requirements for the Degree of Master of Science Auburn, Alabama August 6, 2011 Keywords: Solder Joint Reliability, Life Prediction for Electronics, Electronic Packaging, Fatigue, Lead-Free Solder, Cu-Traces Approved by Pradeep Lall, Chair, Thomas Walter Professor of Mechanical Engineering George T. Flowers, Professor of Mechanical Engineering Barton C. Prorok, Associate Professor of Materials Engineering ii Abstract In the age of portable communication devices, with electronics playing a vital role in all aspects of our daily lives, their reliability is of great concern to the industry. In the highly competitive portable electronics market, the product life-cycles are constantly shrinking, as new technologies emerge, making older ones obsolete at unprecedented rates. As a result, the OEMs face the challenge of developing novel products within shorter development cycles. Furthermore, with increased consumer expectations of superior performance and high quality, the reliability of the products is no more an afterthought in the product design process. The consumer electronics industry is largely driven by tends towards miniaturization and high functionality. Due to their smaller and lighter construction, paired with careless handling at the hands of the consumer, portable electronics are under constant risk of mechanical abuse in the form of accidental drops. This work serves to address the need for reliability models, which predict the useful life of electronic components without carrying out exhaustive testing procedures. The study explores the feasibility of employing the Finite Element Method to simulate mechanical phenomenon, as an alternative to conducting experimental tests. The study demonstrates the use of validated Finite Element simulations based results to develop life prediction models for electronic components subjected to drop and shock loadings. iii In this work, the drop and shock reliability of electronic components has been extensively researched. The study investigates the reliability aspects associated with contemporary packaging architectures, such as Ball Grid Arrays (BGAs) and Package- On-Package (PoP) structures; with special focus on second level interconnects, fabricated with novel Pb-free solder alloys. The Finite Element Method has been employed to predict stresses and strains in solder interconnect, during transient-dynamic shock events. Digital Image Correlation based strain measurements on the test vehicle have been used to validate the FE models. The Levenberg-Marquardt Algorithm, a non linear least squares minimizing method, has been used to evaluate constants which relate the interconnect life-in; terms of shock-events-to-failure; to the stain levels it experiences during each event. The study also looks at the reliability issues associated with Copper Traces on the PCB surface, investigating the effects of variation in their geometries and orientation on their reliability. Furthermore, a fatigue life model, derived from the experimental data, is presented for enabling life prediction of Cu-traces in drop/shock. The reliability studies presented in this work, offer an insight into BGA and PoP solder interconnect and PCB metallization failure mechanisms. The developed life prediction models enable easy yet effective assessment of electronic component reliability, eliminating the need for exhaustive testing procedures, thereby shortening the product development cycle. iv Acknowledgments Pride Praise and Perfection belong to the almighty alone. So, first of all I would like to offer my heartfelt salutation to the Supreme Being for the physical and mental strength bestowed upon me and in whose faith, I was able to complete this task. With an overwhelming sense of legitimate pride and genuine obligation, which gives me exuberant privilege, I reiterate my indebtedness to my research advisor Dr. Pradeep Lall, for his incessant guidance, innovative ideas, prompt help and keen interest during the study. I humbly express my sincere thanks to Dr. George T. Flowers and Dr. Barton C. Prorok for agreeing to serve on my committee. Their candid discussions, honest appraisal and constructive criticism, helped me a great deal in completing this work. Every effort is motivated by an ambition and all ambitions have an inspiration behind. I owe this pride place to my Pa- Ashok Angral and Ma - Nirmal Angral. Their unwavering, ecstatic love, dedication and encouragement has always rejuvenated my strength and rekindled my sense of duty. Fortunately the spirit has been sustained by unsustained, unflinching support, which has been forthcoming perpetually from my friends, Muralidharan, Mahendra, Sonali, Vijay, Ananthakrishnan and Jaspreet who are always with me for every reason. With heartfelt obligation and love, I thank Mistu, her moral support, encouragement and advice helped me throughout my graduate student life. v It is a rare privilege to get to work in a place where every single colleague is a great friend. I thank all my colleagues: Ryan, Prashant, Mandar, Rahul, Vikrant, Sandeep and Dinesh for all the help. It was a pleasure knowing and working with all of you. Lastly, I would like to acknowledge the financial support that the Center for Advanced Vehicle and Extreme Environment Electronics provided me. I thoroughly enjoyed each of the CAVE Review meetings and firmly believe that the exposure and knowledge the center provided me with is a valuable addition to my intellectual assets. vi Table of Contents Abstract ............................................................................................................................... ii Acknowledgments ............................................................................................................. iv List of Figures .................................................................................................................... ix List of Tables .....................................................................................................................xv 1. Introduction ......................................................................................................................1 1.1 Electronic Packaging Overview .........................................................................1 1.2 Reliability Concerns...........................................................................................7 1.3 Vibration and Shock Reliability.......................................................................10 1.4 Finite Element Modeling for Transient Dynamic Simulations ........................13 1.5 Digital Image Correlation ................................................................................15 1.6 Thesis layout ....................................................................................................18 2. Literature Review...........................................................................................................21 2.1 Solder Joint Reliability ....................................................................................21 2.1.1 Lead-Free (Pb-Free) Solders .............................................................23 2.2 Drop and Shock Testing...................................................................................25 2.3 Finite Element Modeling for Transient Dynamic Simulations ........................28 2.4 Digital Image Correlation ................................................................................30 vii 3.Fatigue Constants For Life Prediction Of Second Level Solder Interconnects Subjected To Drop & Shock ......................................................................................33 3.1 Introduction ......................................................................................................33 3.2 Experimental Test Boards, Setup and Procedure .............................................36 3.3 Evaluation of Fatigue Constants for Lead-Free Solder Interconnects .............40 3.4 Life Prediction of Lead-Free Alloy Based Solder Interconnects ....................69 3.5 Failure Analysis and Effect of Thermal Aging on Drop/Shock Reliability.....70 4. Board Trace Fatigue Models And Design Guidelines For Electronics Under Shock- Impact .........................................................................................................................72 4.1 Introduction ......................................................................................................72 4.2 Test Vehicle .....................................................................................................75 4.3 Accelerated Testing .........................................................................................80 4.4 Finite Element Modeling .................................................................................82 4.5 Trace Design Assessment ................................................................................91 4.6 Copper Trace Design Guidelines .....................................................................98 5. Reliability Studies On Package-On-Package Components In Drop And Shock Environments ............................................................................................................101 5.1 Introduction ....................................................................................................101 5.2 Package-on-Package (PoP) Architecture ......................................................102 5.3 PoP Test Vehicle ............................................................................................104 5.4 PoP: Surface Mount Assembly ......................................................................110 5.5 Reflow Parameters .........................................................................................114 5.6 PoP: Drop Testing ..........................................................................................120 5.7 Finite Element Models For Reliability Evaluation ........................................124 5.8 PoP Vs BGA: Susceptibility to failure...........................................................135 viii 5.9 Failure Analysis .............................................................................................136 6. Summary and Conclusions .........................................................................................140 Bibliography ...................................................................................................................144 ix List of Figures Figure 1.1: Electronic packaging industry: trends .............................................................. 3 Figure 1.2: Different types of Electronic Packages, classified according to their mounting type and lead structures. [Source: Renesas Electronics Corporation]..........4 Figure 1.3: Increasing complexity and performance capabilities of modern chip- carriers. [Source: Corporate Package Development, STMicroelectronics] ................ 5 Figure 1.4: Hierarchy of electronic packaging. .................................................................. 6 Figure 1.5 : Typical ball grid array (bga) electronic package architecture ......................... 8 Figure 1.6: Vibration testing profiles for survivability assessment of electronic components [Ref: MIL-STD-810F] .......................................................................... 11 Figure 1.7: JEDEC standard shock pulse [Ref: JEDEC, JESD22-B111] ......................... 12 Figure 1.8: The principle of DIC [Source: Dantec Dynamics A/S] .................................. 17 Figure 2.1: Typical drop test apparatus and mounting scheme for PCB assembly .......... 28 Figure 3.1: Test vehicle used for the study and the dropping orientations. ...................... 37 Figure 3.2: Experimental set-up and schematic for controlled drop test with high speed cameras for DIC based strain measurements. ................................................. 38 Figure 3.3: Package Strain and corresponding continuity history, with peaks in continuity indicating failure. ..................................................................................... 39 Figure 3.4: Shematic describing methodology adopted for development of life prediction models for leadfree solder interconnects.. ................................................ 39 x Figure 3.5: Schematic explaining basic principle of Rainflow Counting Algorithm. ...... 42 Figure 3.6: SAC105 Pristine- Interconnect strain history and corresponding strain amplitude histogram .................................................................................................. 43 Figure 3.7: SAC105 Thermally Aged- Interconnect strain history and corresponding strain amplitude histogram ........................................................................................ 44 Figure 3.8: Sn3.5Ag Pristine- Interconnect strain history and corresponding strain amplitude histogram .................................................................................................. 45 Figure 3.9: Sn3.5Ag Thermally Aged- Interconnect strain history and corresponding strain amplitude histogram ........................................................................................ 46 Figure 3.10: SAC305 Pristine-Interconnect strain history and corresponding strain amplitude histogram .................................................................................................. 47 Figure 3.11: SAC305 Thermally Aged- Interconnect strain history and corresponding strain amplitude histogram ........................................................................................ 48 Figure 3.12: 3d Error Space - variation of error with ?A? and ?n? .................................... 50 Figure 3.13: Depiction of convergence of global minima in hybrid methods. ................. 54 Figure 3.14: Variation in initial guesses to the ?A? and ?n? values given as an iNput to the algorithm. ............................................................................................................. 56 Figure 3.15: Variation of ?A? and ?n? values across bounded error space and error minimization. ............................................................................................................. 57 Figure 3.16: Fatigue constants computation for Pristine SAC105 Interconnects based on solder interconnect strain history using LM Algorithm based error minimization. ............................................................................................................. 59 Figure 3.17: Fatigue constants computation for thermally aged SAC105 alloys based on solder interconnect strain history using LM algorithm based error minimization. ............................................................................................................. 61 Figure 3.18: Fatigue constants computation for Pristine Sn3.5Ag Interconnects based on solder interconnect strain history using LM algorithm based error minimization. ............................................................................................................. 63 Figure 3.19: Fatigue constants computation for thermally aged Sn3.5Ag interconnects based on solder interconnect strain history using LM algorithm based error minimization. ............................................................................................................. 65 xi Figure 3.20: Fatigue constants computation for pristine SAC305 interconnects based on solder interconnect strain history using LM algorithm based error minimization. ............................................................................................................. 67 Figure 3.21: Fatigue constants computation for thermally aged SAC305 alloys based on solder interconnect strain history using LM algorithm based error minimization. ............................................................................................................. 69 Figure 4.1 : Schematic showing methodology adopted for development of fatigue constants and evaluation of copper trace reliability .................................................. 74 Figure 4.2: Test vehicle schematic and photograph showing the location and layout of traces on both sides of the board. .............................................................................. 76 Figure 4.3: 3D-Digital Image Correlation ? principle ...................................................... 79 Figure 4.4: Schematic showing nature and direction of loading for accelerated testing .. 81 Figure 4.5: Contour plot (global model) showing variation of longitudinal tensile stresses on the PCB ................................................................................................... 83 Figure 4.6 Schematic describing the finite element modeling methodology used, with implementation of global-local submodeling ............................................................ 84 Figure 4.7: Global local submodeling: Using outputs from driving nodes of the validated global model to drive corresponding driven nodes of the submodel ......... 85 Figure 4.8: Submodels for different straight pad-trace junction geometries .................... 86 Figure 4.9: Sub models for different angled pad/trace junction geometries ..................... 86 Figure 4.10: Multidirectional Trace Sub-models. ............................................................. 87 Figure 4.11: Comparison of strains contours as obtained by DIC and as predicted by Finite Element Analysis. ........................................................................................... 88 Figure 4.12 : Variation of longitudinal stresses along the length for different trace geometries ................................................................................................................. 89 Figure 4.13: Failure locations, as observed from FE simulations and accelerated tests for straight traces ....................................................................................................... 89 Figure 4.14: Model predictions of stress concentrations and observed ............................ 90 Figure 4.15: Comparison of peak strains in different trace/pad geometries. .................... 91 xii Figure 4.16: Increasing stress concentration with increasing difference in trace width and pad diameter. ...................................................................................................... 92 Figure 4.17: Variation in maximum principal stress with angle made with longitudinal direction of the board ................................................................................................ 93 Figure 4.18: Effect of trace orientation on max principal strain ....................................... 94 Figure 4.19: A comparison of stress concentrations developed at the trace-pad junctions for angled traces. ........................................................................................ 95 Figure 4.20: Stress concentrations at sharp angles between traces. .................................. 96 Figure 4.21 : Figure showing optimum trace orientation for maximum reliability .......... 97 Figure 4.22: Fatigue curve describing the low cycle fatigue behavior of copper traces .. 98 Figure 5.1: A schematic of package-on-package stack showing both the constituent packages viz. the bottom (PSvfBGA) and the top (CSP) packages. ....................... 103 Figure 5.2: A schematic of package-on-package test vehicle developed, showing the three different daisy chain channels and other board attributes. ............................. 105 Figure 5.3: Schematic describing test vehicle channels and other attributes .................. 106 Figure 5.4: Single component JEDEC standard Package-on-Package test vehicle, fabricated and used for the study. ............................................................................ 106 Figure 5.5: Schematic showing matching daisy chain net-lists on the top of the bottom package (PSvfBGA) and the bottom of top package (CSP) .................................... 107 Figure 5.6 : Schematic showing matching daisy chain net-lists on the bottom of the bottom package (PsvfBGA) and the PCB ............................................................... 108 Figure 5.7 : Schematics showing the classification of the two tiers of Solder Interconnects into three daisy chain net-lists for the test component used in the study (Source: Amkor Technology) ........................................................................ 109 Figure 5.8: Schematic of PoP stack with relevant dimensions. ...................................... 110 Figure 5.9: A comparison between regular and PoP-specific solder paste, showing a difference in amount of transcription achieved by dipping. .................................... 112 Figure 5.10: Process flow adopted for assembling package-on-package components on the test vehicle .................................................................................................... 113 xiii Figure 5.11: Mechanism of warpage due to CTE mismatch. ........................................ 114 Figure 5.12: Solder reflow defect: Excessive warpage and tombstoning resulting in open connections ..................................................................................................... 116 Figure 5.13: Solder reflow defect: Improper reflow due to low peak temperature and less TAL resulting in un-collapsed solder joints ..................................................... 116 Figure 5.14: Solder reflow defect: Bridging of interconnections leading to short- circuits, resulting from high peak temperature and/or high TAL............................ 117 Figure 5.15: Solder reflow defect : Head-in-Pillow defect resulting from premature activation of flux due to high ramp rate and failure in removal of oxide film. ....... 117 Figure 5.16: Solder reflow defect : Sliding of the package owing to high warpage due to thermal shock ...................................................................................................... 118 Figure 5.17: Good solder interconnections obtained by optimized reflow profile. ....... 118 Figure 5.18: Different reflow profiles adopted with different parameters. The optimum profile (Profile-2) is highlighted in red. ................................................... 119 Figure 5.19: Effect of reflow parameters on PoP assembly and induced defects. ......... 120 Figure 5.20: Schematic of experimental set-up for controlled drop test with DIC and DAQ systems ........................................................................................................... 121 Figure 5.21: Experimental set-up for controlled JEDEC drop test using LANSMONT? drop test tower. ............................................................................. 122 Figure 5.22: 3D-Digital Image Correlation- basic principle [Panchagade, 2006] ........ 123 Figure 5.23: Strain repeatability through different shock events .................................. 124 Figure 5.24: Global-local Submodeling: outputs from defeatured global model used to drive submodel preserving geometrical and material detail of the area of interest. .................................................................................................................... 125 Figure 5.25:Detailed submodel preserving the geometrical detail and various material layers of the PoP stack with the corner solder balls modeled in detail while all other I/Os are modeled as Timoshenko beam elements. ......................................... 126 Figure 5.26: PoP assembly submodel- critical corner interconnect .............................. 127 Figure 5.27: Figure showing correlation between DIC and FE simulation based strain contours on the test vehicle surface- incurred during drop event. ........................... 128 xiv Figure 5.28: A comparison of out of plane displacements as evaluated by DIC and predicted by FE simulations .................................................................................... 129 Figure 5.29: Strain histories extracted at the four corner interconnects in the bottom tier of solder balls. ................................................................................................... 131 Figure 5.30: Strain histories extracted at the four corner interconnects in the top tier of solder balls. ......................................................................................................... 131 Figure 5.31: A comparison of strain cycles experienced by solder interconnects in the top and bottom tiers located at corner-1 as shown in schematic ............................. 132 Figure 5.32: A comparison of strain cycles experienced by solder interconnects in the top and bottom tiers located at the corner-2 as shown in the schematic ................. 132 Figure 5.33: A comparison of maximum and minimum out of plane strains experienced by corner solder interconnects in the bottom tier. ............................... 133 Figure 5.34: A comparison of maximum and minimum out of plane strains experienced by corner solder interconnects in the top tier. ..................................... 133 Figure 5.35: Stress distribution in the critical corner interconnect at different time steps during shock event. ......................................................................................... 134 Figure 5.36: A comparison of stresses induced in a corner interconnect for .................. 135 Figure 5.37: A comparison of out-of-plane stresses induced in corner PCB side copper pads interconnect for PoP and BGA packages ............................................ 136 Figure 5.38: Figures showing failure locations as observed by cross sectioning and as predicted by FE simulations .................................................................................... 137 xv List of Tables Table 1.1 Thermal and mechanical properties of various packaging materials .................. 8 Table 3.1: Package architecture of 100 I/O CABGA on test board .................................. 38 Table 3.2: Fatigue constants and life predictions.............................................................. 70 Table 4.1: Different pad/trace geometric combinations incorporated .............................. 77 Table 5.1: PoP submodel: elements and material properties used .................................. 127 1 CHAPTER 1 INTRODUCTION 1.1 Electronic Packaging Overview The field of electronic packaging has been constantly evolving and adapting to meet the ever accelerating demands of the consumer electronics market. Electronic Packaging is traditionally defined as the back-end-of-the-line process that transforms silicon based integrated circuits (IC) into functional assemblies. Packaging brings together the fundamentals of Mechanical Engineering, Physics and Electronics towards the development of highly functional computing devices. Development and characterization of reliable materials for packaging and establishment of interconnections between different layers of an electronic component has remained the cornerstone of leading research in electronic packaging for the past few decades. Leadfree and halogen- free solder alloy materials in particular have attracted great research attention. Their compliance with recently established Restriction of Hazardous Substances directive (RoHS) has accounted for their proliferation throughout the packaging industry. Furthermore, their high degree of comparability with conventional tin-lead based solder alloys with regard to manufacturability and production has made them quite acceptable as primary soldering materials. But, much waits to be explored as the reliability concerns associated with these novels materials have started to surface. 2 Gordon Moore, co-founder of Intel, referring to the state-of-art in the semiconductor industry, famously stated in 1965 ?The amount of transistors which can be inexpensively placed on an Integrated Circuit device doubles every 18 months". The statement, initially made in the form of an observation and forecast, has since been widely accepted and become the trend driving the semiconductor industry [Moore et al, 1998]. The electronic packaging industry is not oblivious to the market trends that drive the semiconductor industry, its similarities with which tread a very fine line. With decrease in IC feature size and consequential reduction in cost per IC, there is an evolving need for cost-effective and reliable packaging solutions also. While the future IC chips are predicted to be larger in size and to have more I/Os, the advancements in electronic packaging are consequentially going to be driven by the market requirements for cost-effectiveness, miniaturization and high functionality. In compliance with the consumer electronics market trends, as the products continue to become smaller and more functionally packed, exhaustion of horizontal packaging space has further led to development of 3D stacking solutions. 3D package or die stacking provides more functionality in smaller dimensions through expanding packages in the vertical dimension. System-In-Package (SiP) architectures are being widely accepted as a solution to system integration. System-in-Package incorporates electronic and non- electronic devices such as optical (cameras, proximity sensors), and Micro-Electro- Mechanical-Systems(MEMS) devices (such as accelerometers and gyroscopes), and their interconnections in a single package, to form smart structures. With these fast paced developments in the Electronic packaging industry, the reliability aspects of novel 3 systems need to be thoroughly addressed before they can be introduces in commercial products. Figure 1.1: Electronic packaging industry: Trends [Adapted from JISSO Japan Technology Roadmap, 2001 Edition and S.L. Buedo (Universidad Autonoma de Madrid)] Electronic Packaging Systems can be broadly classified either on the basis of their mounting type as 1) Surface Mount Components and 2) Through Hole Components; or on the basis of their packaging type as 1) Hermetically Packaged or 2) Plastic Encapsulated packages. The former classification of electronics packaging is the most widely accepted. T h r u H ol e ? D I P ? P G A S u r f a ce Mou n t ? Q FP ? T S O P ? S O J ? B GA C h i p S ca l e ? C S P ? Fl i p C h i p ? W L P 3 D P a c k a g i n g ? ? P o P ? S t a ck e d D i e 4 Figure 1.2: Different types of electronic packages, classified according to their mounting type and lead structures. [Source: Renesas Electronics Corporation] S ur f a c e M oun t T y pe A r e a A r r a y P a c k a g e G ul l W i ng L e a d P a c k a g e N on - L e a d P a c k a g e J - L e a d P a c k a g e P a s s i v e D e v i c e s B G A , F B G A , H F B G A , H B G A T H B G A , L B G A , T F B G A , V F B G A T F L G A Q F P , H Q F P , L Q F P , H L Q F P , T Q F P , H T Q F P S O P , S S O P , H S O P , H S S O P , L S S O P , T S S O P V Q F N , H V Q F N , W Q F N , H W Q F N V S O N , W S O N , H W S O N S O J , Q F J SC - 6 2 , S C - 7 0 , S C - 7 4 , S C - 8 8 , T S S O P , H W S O N , L D P A K , D P A K , D 2 P A K , U F P , U R P D I P , S D I P , S I P , Z I P TO - 3 P , T O - 2 2 0 , D P A K ( L ) , L D P A K ( L ) L e a d t y pe P a c k a g e P a s s i v e s P i n ? I ns e r t i on T y pe 5 Figure 1.3: Increasing complexity and performance capabilities of modern chip- carriers. [Source: Corporate Package Development, STMicroelectronics] For facilitating ease of identification electronics are further classified within a system. The electronic packaging hierarchy classifies an electronic system in four levels. The zeroth level constitutes non-repairable parts, such as integrated circuits or passive devices. The first level identifies small components such as BGAs and QFPs which house the zeroth level devices and are mounted on PCBs daughter-cards. The second level packaging constitutes the PCBs that house different second level packages together into one functional unit. Lastly, the third and highest level of packaging identifies motherboards and pullout chassis on electrical cabinets which house several PCBs together to form a standalone electrical unit. The interconnects between the zeroth (chip) and the first (package) level of packaging are the wire-bonds. These are referred to as first level interconnects while the solder balls or leads, which connect the packages to their landing pads on the PCB are called second level interconnects. 6 Figure 1.4: Hierarchy of electronic packaging. [Adapted from ?Mechanical Design of Electronic Systems? by Dally, Lall and Suhling, 2008] S i l i co n W a f e r I C C hi p F i r s t Le v e l P a ck a ge ( B GA) S e cond Le v e l P a ck a ge ( P C B or C a r d) T hi r d Le v e l P a ck a ge ( M ot he r boa r d) F i r s t L ev el In t er c o n n ec t ( W i r eb o n d s ) S ec o n d L ev el In t er c o n n ec t ( S o l d er B a l l s ) 7 1.2 Reliability Concerns. Driven by the trends towards miniaturization and increased functionality, the electronic components are being made smaller and more complex. With advent of novel packaging technologies such as System-on-Chip (SoC), System-in-Package(SiP), Multi- Chip-Modules (MCM) and 3D packaging technologies such as PoP and Stacked-Die components, the complexity of packaging architectures has exponentially increased. As the interconnect counts increase and the pitch between them decreases, their susceptibility to failure increases. In the highly competitive electronics market, it is of outmost importance to the manufacturers to ensure highest levels of quality and reliability in usage conditions representative of the real world. A typical electronic package comprises of various different types of materials (Figure 1.5) with vastly varying material properties. For accurate reliability quantification, it is very import to have a clear understanding of how all these constituent materials of a component interact with each other during the products operational life. As an example, automotive electronic components such as Engine Control Unit (ECU) housed near the engine experience harsh thermal cycling during their operational life. These extreme thermal conditions lead to severe expansion and contraction of the components. A high degree of mismatch between different constituent layers of the package can often lead to delamination between layers and shearing of the interconnects. As another example, electronic components housed in portable electronic devices often experience shock pulses when the product is dropped. A careless design which does not allow for any compliance to deformation under shock-loading may result in drastic failure of the component and as a result the whole device. 8 Figure 1.5 : Typical Ball Grid Array (BGA) electronic package architecture Table 1.1 Thermal and mechanical properties of various packaging materials In the light of growing concern for environmental preservation, various directives; such as the Restriction of Hazardous Substances (RoHS) and the Environmental Preservation Agency studies on life-cycle assessment (LCA) of the environmental impact of lead-free and tin-lead solder, as used in electronic products; have required the electronics manufacturers to comply to strict regulation on use of hazardous substances in their products. The use of Lead, a primary constituent of the solder alloys compositions, previously dominant in the industry, consequentially, had to be stringently regulated. Even though the total lead used in electronics makes up only a small fraction (?2%) of world lead consumption, while most of lead is used for batteries, the electronic packaging industry had to venture into alternate materials. Novel Tin-Silver based solder alloy Mol d C o m po und P r i nt e d C i r c ui t B o a r d S i l i c o n Di e Di e A t t a c h m e n t A d h e s i v e C o pp e r P a ds S o l de r I nt e r c o nn e c ts B T S ubs tr at e E lastic M od u lu s (E) (G P A) Mol d Com pound 23.5 15 S il icon D ie 162.7 2.5 Die Attac h Adhe sive 6.8 52 B T Sub stra te 18 12.4 S older B a ll s 30.5 24.5 P rinte d Ci rc uit B oa rd 17 14.5 M at e r ial Coe f f icie n t of T h e r m al E xp an sion ( CTE ) ( 10 - 6 1/K ) 9 compositions were developed to address the issue. However, their adverse effects on product reliability and high cost of compliance were widely cited as criticisms of these alternative materials. In the recent past, the high volume of research on the behavior of leadfree solders and their reliability has helped in development of a better understanding of their performance. However the reliability data available on these materials is still very limited and does not suffice to establish a comprehensive comparison with Lead-based alloys. In today's world, electronic systems are omnipresent and in many cases, they enable core functionality of the devices that encompass them. With electronic components residing at the heart of most of the mission-critical systems of the modern age- such as Air Traffic Control , Personal and Mass Communication Systems and Transportation, it is of outmost importance that the reliability concerns related with them be thoroughly addressed. Unlike mechanical constituents, of any system, electrical components generally do not wear out. Their failure is often undetectable before it occurs. The main reliability concerns associated with electronic components are associated with the following causes: 1. Thermally Induced stresses due to exposure to high operating temperatures 2. Mechanically Induced stresses due to careless handling-drop; and operation in vibrating environment. 3. Residual stresses from manufacturing processes 4. Electro-migration, Transient static discharge and other electrically induced failures. 10 With regard to the mechanical failures, especially those caused due to thermally or mechanically induced stresses, the interconnects between various levels of a component are rendered most vulnerable to failure. Furthermore, since they form a very significant part of a component, their failure can result in drastic failure of the whole component. This thesis encompasses reliability studies carried out on electronic component interconnects with special focus on second level interconnections. While the state-of the-art lies in diagnostic methods which reside in post-failure space. This work addresses the growing need for prognostic and life-prediction methodologies which enable failure isolation before occurrence. 1.3 Vibration and Shock Reliability Portable computing and communication devices, omnipresent in today?s world, constitute a major application of electronic components. Furthermore, emerging trends for smaller and lighter form factors have driven the electronic component industry towards robust designing or withstanding high shock and vibration environments. With a high probability of careless handling at the hand of the consumer, the electronics industry cannot undermine the importance of mechanical robustness of their products. The reliability issues associate with the interconnections between various electronic components of a product, especially, are of most importance since they are most susceptible to failure under harsh environments. For an assessment of the reliability of electronic components under drop and shock conditions representative of the real-world usage conditions, various testing procedures have been standardized. For evaluation of their reliability in vibration 11 environment, various different vibration profiles have been established to serve as representation of actual environment in which a component is intended to be deployed. Test methods used for Drop/Shock reliability quantification can broadly be classified as board-level and product-level tests, under constrained and unconstrained (free drop) conditions. The most widely accepted board level constrained drop test, is the one developed by the Joint Electron Device Engineering Council (JEDEC). The JEDEC test standard ?Board Level Drop Test Method of Components for Handheld Electronic Products?, JESD22-B111, is used by electronics manufacturers to evaluate the performance of surface mount components in drop/shock. Figure 1.6: Vibration testing profiles for survivability assessment of electronic components [Ref: MIL-STD-810F] 12 Figure 1.7: JEDEC standard shock pulse [Ref: JEDEC, JESD22-B111] The standards established for reliability quantification however, adopt a highly conservative approach. The tests, like the one outlined in JEDEC standard JESD22-B111, subject the test vehicle to a very harsh shock pulse with constraint conditions resulting in high level of deformations. Consequentially, mapping of the standard test results to product level performance is very challenging. While a component may perform relatively worse in a standard test, it might turn out to survive through its service life when deployed in the field, since in an actual product, the component failure mode and time of occurrence is dominated by its housing design and the level of shock its subjected to; which may vary with ?say, different drop orientations and heights. Additionally, factors such as the product mass, structural compliance etc also play a vital role in affecting the shock pulse that the component finally experiences. Despite the availability of standard testing procedures, the component deigns process is still very challenging, owing to the possibility of a multitude of failure inducing environments that the product much can be subjected to-each of which the product cannot be tested for. It is very difficult to analyze various design variations, of a product by subjecting each variation to exhaustive testing procedures. An underlying theme of this thesis is development of life- 13 prediction models and design guidelines which eliminate the need of such exhaustive testing procedures and aid in speeding up the product development cycle. 1.4 Finite Element Modeling for Transient Dynamic Simulations In the highly competitive electronics market, the product development cycles are continuously shrinking. New technologies are being rapidly developed and introduced into the products at unprecedented rates. Conception-to-Production times have substantially shrunk as novel technologies continue to make older ones obsolete within the span of a few months. With tight production schedules and small product development cycles, exhaustive testing of all the design iterations of a product has become unfeasible. Owing to the complexities involved in in-situ measurement of Solder joint response to the transient dynamic shock phenomenon, analytical methods have established their superiority over experimental methods. Finite element methods in particular have gained popularity in the electronics industry, in the development of new packaging materials, designs and assembly processes. The analytical method involves discretization of a continuous structure into ?finite elements? governed by specific material models. When carefully devised and validated with experimental observations, the technique is capable of accurately simulating material responses to static as well as transient dynamic events. Various Finite Element modeling approaches have been used to address reliability issues specific to electronic packaging [Dally 2008]: 1. Equivalent layer models to represent the solder joints and simulate their behavior under drop impact. 14 2. Solid-to-solid sub-modeling techniques to analyze BGA reliability for drop impact using half the PCB board. 3. Shell-to-solid sub-modeling using beam-shell-based quarter symmetry model to reduce the computational time. 4. Symmetry of load and boundary conditions is used to obtain computational efficiency and decrease the model size. The FE method based simulations make it easier to account for scale differences between the dimensions of the individual layers of an electronic assembly, such as the solder interconnects, copper pad and chip interconnects. [Dally, 2008]. I the past, several techniques to simulate electronic packaging subjected to shock pulses such as those induced by drop testing test have been developed. To this effect, in addition to employing implicit FE method by translating the input acceleration pulses into effective support excitation loads on the test vehicle; [Yeh 2004], researchers have also used the explicit finite element solver to simulate transient dynamic event [Lall 2004, 2005, Xie 2002, 2003, Wu 1998, 2000]. Explicit time integration serves as a better approach to transient dynamic deformation of electronic assemblies since it approximates the phenomenon as a wave propagation problem. The governing equation for a dynamic system can be expressed as [Cook 1989]: ? ?? ? ? ?? ? ? ? ? ?nextnnn RRDCDM ??? i n t??? (1.1) For a linear problem, {Rint}n=[K]{D}n, where [M], [C] and [K] are the mass, damping and stiffness matrices respectively and {D}n is the nodal displacement vector as each time step. Methods of explicit direct integration calculate the dynamic response at 15 the time step (n+1) from the equation of motion, using the central difference formulation and known conditions at one or more preceding time steps as shown below [Cook 1989]: ? ? ? ? ? ? ? ?? ? ? ? 122in t12 2 1122 11 ?? ?????? ?????????????? ??? nnnnextn DCtMtDMtRRDCtMt (1.2) Other FE techniques used in electronic packaging reliability studies include implicit global models[Irving 2004, Pitaressi 2004, Syed 2005], global-local submodeling [Wang 2004, Tee 2003, Wong 2003, Zhu 2001, 2003, 2004], smeared property models [Jie 2004, Lall 2006] and explicit-implicit sequential model [Zhu 2005]. The global-local submodeling is a very powerful approach which significantly reduces the computational time by dividing a model into two parts. In the first run, a defeatured ?global model? is created and used to simulate the phenomenon using boundary conditions representative of the actual test. Subsequently, detailed ?submodels? of the area of interest are developed. These submodels preserve the geometric and material intricacies of the area of interest in complete detail. The outputs from the global model simulation are extracted at key locations and provided as driving inputs to the submodel. The Finite Element method has been extensively used for simulating transient dynamic phenomenon in electronic assemblies, in the studies presented in this thesis. 1.5 Digital Image Correlation With the ease in availability of low-cost and easy-to-use hardware and software facilities for digital image acquisition, digital image analysis, nowadays, is becoming extremely popular in experimental mechanics. It is widely used in strain field 16 measurement as a powerful tool to post-process fringe images generated by moir? effect or other interferometry methods. High speed photography has been used to measure deformations in sheet metal forming, Automotive crash testing, rail vehicle safety[Kirpatrick, 2001], air-plane safety[Marzougui, 1999], modal analysis of turbine blades, high strain rate Split- Hopkinson bar tests, dynamic fracture phenomenon, and package hermiticity (MIL-STD- 883) tests. High-speed cameras measure impact speed, force, and deformation due to shock, and thermal loading. Previously, the measurement of derivatives of field quantities, such as strains, was limited to a specific physical locations or discrete target points in an electronic structure. It was not feasible to extract data at a very large number of locations by using discrete targets because of the time consuming nature of the process. Digital Image Correlation (DIC) is a state-of-art technique which overcomes such limitations of other optical methods. A major advantage of the technique is that the sample preparation is simple and quick. To prepare the specimens for testing, they are to be speckle painted, which can be easily done using spray paint. Transient deformation data can then be recorded with the help of high-speed cameras, which is then processed to evaluate full-field strains. In electronic packaging science, digital image correlation has been used to study various mechanical phenomenon such as stresses in BGA package solder interconnects under thermal loading[Zhou 2001; Rajendra 2002; Zhang 2004 2005; Xu 2006; Lall 2009, 2010], material characterization at high strain rates [Tiwari 2005, Lall 2010], crack propagation in solder interconnects in drop/shock [Lall, 2009, 2010] and strains in flip-chip dies under thermal loading [Kehoe 2006]. In the recent past, 17 it has also been used to measure deformation kinematics with the help of ultra high-speed data acquisition and video systems. DIC is also used to acquire Experimental data which is correlated to the finite element models for validation. Figure 1.8: The principle of DIC [Source: Dantec Dynamics A/S] Essentially, Digital image correlation is a digital speckle correlation based method to measure full field deformation on the surface of an object based on tracking a geometric point before and after deformation and using it to calculate the displacement field. It has s widely been used for measuring shock deformation in electronic assemblies. 18 The technique works by comparing a sub image around a reference pixel in the original/reference image and then comparing it with the sub images corresponding to different pixels in the deformed image using a predefined correlation function to describe the difference of the two digital sub images. A full field displacement contour is obtained by shifting the reference pixel in the original image and applying this method to all the other pixels of the images. It is therefore obvious that the resolution/accuracy of the DIC procedure relies on the pixel size of the camera image. The technique is described in further detail in Chapter 3 of the thesis. 1.6 Thesis Layout This thesis presents reliability studies performed on various electronic component structures. The aim of this work was to develop Finite Element models to simulate transient dynamic phenomenon in electronic assemblies with special focus on solder interconnects. Using FE simulations, validated with experimental measurements, the work aims at development of life prediction models and design guidelines which eliminate the need for exhaustive testing during the design iterations and aid in shortening the product development cycle. Chapter 2 presents literature survey on solder joint reliability. Experimental methods used in electronic packaging reliability studies and various life prediction models used in electronics. A major portion of this work deals with reliability o second level interconnects. The chapter therefore outlines previous research carried out in this field in order to differentiate the work presented in this thesis. Board level reliability studies carried out on BGA packages in particular have been emphasized. Previous work 19 pertaining to the drop/shock and vibration performance of the novel lTin-Silver based leadfree solder alloys has also been presented. The import of the work presented herewith, in context of its applicability to and usefulness in product development cycles, the current state-of-art in electronic packaging reliability studies has been summarized. The following Chapter 3 presents a study carried out a BGA test vehice with three different lead free alloys system used for solder interconnects ?with an aim to develop a life prediction model for these interconnects in drop/shock environments. The chapter describes the tests carries out on pristine and thermally aged test vehicles and compares their performance in drop tests. The methodology adopted to evaluate a set of fatigue constants; which govern the mathematical life prediction model; including a non linear least squares minimizing algorithm, the Levenberg-Marquardt (LM) algorithm, has been described in detail. Finally, the chapter validates the accuracy of the life prediction model developed and advocates its applicability to product development cycles. In the following section, Chapter 4, reliability studies carried out on an indigenous test vehicle ? developed with an intention to study PCB surface Copper Traces- have been presented. The chapter details the test vehicle attributes, the experiment designed to study the effect of variations in copper trace attributes on their reliability in drop and shock and the strain-life based power law developed to describe copper trace fatigue. Conclusively the study along with presenting a fatigue life prediction model, summarizes the effects of PCB surface copper trace orientations and dimensions on their reliability. Design guidelines for laying out copper traces on the PCB surface, ensuring maximum reliability have been suggested for use as empirical rules. 20 Chapter 5 encompasses studies carries out on novel Package-on-Package (PoP) components. While the current state-of-art resides in manufacturability issues associated with PoP structures, the chapter addresses the need for a reliability quantification of PoP structures. The study describes in articulate detail the methodology adopted to fabricate the PoP test vehicle as per JEDEC standards. While presenting an analysis of the e hurdles faced during the test vehicle fabrication, in the form of reflow-related defects, the study also serves to standardize an optimum reflow profile for fabrication of PoP assemblies. Later part of the chapter deals with development of detailed global-local FE models for simulating the PoP structures? response to standard drop tests and provides an insight into the failure mechanisms of these novel structures. In conclusion, the study compares the reliability issues associated with PoP assembly and reliability in drop and shock scenarios and compares them with an equivalent BGA. Finally, the concluding Chapter 6 summarizes the purpose and findings of the thesis and briefly discusses the scope for future work. 21 CHAPTER 2 LITERATURE REVIEW In the age of communication technology, electronics are omnipresent. With the ever increasing significance of electronics, the import of their reliability is also increasing. Lately, the prime focus of packaging research has shifted from manufacturing processes to reliability issues involved with electronics. This chapter entails a background of the research done on electronic packaging reliability; with a special focus on BGA reliability in drop/shock environments; and summarizes recent developments in predictive failure modeling for the same. 2.1 Solder Joint Reliability Since the Interconnects perform a critical function of communicating between functional parts of a package, they form a vital part of the electronic assembly. Additionally, the fact that they the most mechanically complex part of the assembly and their miniature form, render them most vulnerable to failure. The solder interconnects between the package and the PCB, experience the highest susceptibility to failure under harsh loading conditions and upon failure, render the device irreversibly useless. The quality and reliability of solder interconnects are often considered an index of the reliability of the package on the whole. [Woodgate, 1987] described a perfect joint as 22 one showing complete wetting of solder to its corresponding pad- outlining this as a universal requirement that applies to both through-hole and surface mount assemblies. The pursuit of component miniaturization with increased functionality had led the industry towards fine pitch components. With the need for increasing interconnections between packages and PCB, and owing to the limitations of manufacturing processes, when the pitches between the legs of perimeter array surface mount components could not be reduced any further, area array assemblies were conceived. Over the years, area array surface mount components have evolved from primitive Pin-Grid-Arrays to novel Ball-Grid-Array packages. BGAs, which have now proliferated through the packaging industry, cater to the need of high I/Os and reliable interconnects between the packages and the PCB, by accommodating them in a very small form. Furthermore, their ease of fabrication and comparable robustness has established them as suitable replacements for area array surface mount components. In addition to providing electrical connectivity, the solder joints of a BGA component also preserve the mechanical integrity of the component by serving as compliant structures between the component and the PCB. As a result, the solder interconnects consume a large portion of the mechanical damage caused to the assembly in a drop/shock event. Rapid shearing caused due to CTE mismatch between the component and the PCB in thermal loading scenarios and cyclic fatigue at high strain rates during shock events are attributed as major causes of failure in solder joints. In terms of their thermal performance, [Chiang 2000] suggested two reflow passes as a means to achieve better reliability characteristic of BGA type packages. A twofold increase in their characteristic life when subjected to thermal cycling due to 23 significant reduction in equivalent plastic strain, the energy density and the Von-Mises stress of the solder joint was shown. Different package architectures variables and their effects on BGA solder interconnect reliability have been articulated [Syed, 1996]. The improved solder joint reliability of perimeter array packages over area array packages has also been established [Syed, 1996]. Syed, also reported on the direct impact the BT substrate thickness had on solder joint reliability- thicker BT substrate, he reported, enhanced the solder joint reliability. Previously, it has been shown that large fillets and low standoff heights of the BGA solder joints leads to an increase in their fatigue life under power cycling [Charles 1990]. It has been shown that large fillet angles serve to reduces high magnitudes of stress concentrations as they result in an increased the net cross-sectional area within the joint. While the corner solder balls, have been shown to be most vulnerable to failure in drop/shock scenarios, it has been shown that the layout of solder balls is very critical since it affects the load distribution on critical solder ball [Tee 2004].The effects of solder ball metallurgy and the inter-metallic compound (IMC) thickness at the solder-copper interface have been looked into [Zhong, 1999]. The effect of the pad surface finish has alse been researched, [Bradley 1996 and Suhling 2002] showed superior HASL pad finish thermal reliability performance over immersion nickel and palladium based pad finishes for BGA packages. 2.1.1 Lead-Free (Pb-Free) Solders In light of growing concern for the environmental effects of hazardous materials, directives such as the Restriction of Hazardous Substances (RoHS) and Waste from 24 Electrical and Electronic Equipment (WEEE) have restricted the use of lead-based components in electronic manufacturing industry. The advent of Pb-Free alloys in the packages industry while inevitable at first, has now been widely accepted. A vast multitude of lead-free alloy compositions have been studied for their applicability in electronics. The Tin-Silver-Copper (Sn/Ag/Cu) often referred to as SAC family of alloys have been established as one of the most suitable replacements for their leaded counterparts. Previously, researchers have outlined the variation in SAC alloy performance with different packaging architecture, in addition to underlining their strong temperature dependence [Syed 2001, Zhang 2003 and Vandevelde 2004]. Lowering the silver content in the SAC compositions has been shown to result in a significant improvement in their board level reliability in drop/shock testing. In contrast, researchers [Zhang 2008 and 2009] have shown that higher silver content SAC alloys have enhanced thermal reliability. The SAC alloy performance has been compared with conventional leaded solder compositions [Schubert 2003 and Clech 2005]. Experimental studies conducted have shown that a significant enhancement in SAC based solder-joint reliability can be achieved by using compliant plastic substrates. With regard to their effect on drop/shock performance, researchers have investigated the effects of variation in lead-free solder alloy composition on their reliability. SAC alloy compositions with low silver content have been shown to be resistant to high strain rates under mechanical shock and to have improved drop reliability [Zhu 2008, Che 2008, Lall 2008a, Pandher 2008, 2007, Kim 2007]. Additionally, it has been shown that by lowering the silver content of SAC alloys, an 25 increase in their creep rate, can be achieved [Zhang 2008]. Researchers, [Song 2008, Huang 2007] have suggested adding Ni to SAC alloys to improve thier drop reliability. It has also been shown that an improvement in their drop performance can be achieved by using SAC interconnects on Ni/Cu/Au surface finishes [Kawashiro 2008]. Solder joint reliability has been shown to enhance with UBM/penetration layer/SnAg lead free solder bump structures [Choi 2007]. While a lot of research has been conducted on the effects of Pb-free alloy compositions on their drop/shock performance, the life prediction models for Pb-free alloys presented in this study, are largely beyond the state of art. 2.2 Drop and Shock Testing Owing to their small form factor and light weight, portable electronics are highly susceptible to mechanical shocks due to accidental drop. Solder interconnects between different components in a handheld electronic product are the most susceptible to failure in such scenarios. Mechanical characterization of the interconnects and an assessment of their reliability in shock environments is critical to their long term survivability in the real world. Since the electronic components housed in modern electronics are packed with a high level of functionality, the operating temperature in these devices; due to power dissipation; often rise to harmful levels in the vicinity of the packages. Presenting an insight into the effects of overlapping thermal and mechanical stresses on their reliability, Mattila et. al. [Mattila 2007], have investigated the effects of temperature on the drop tests reliability of BGA components. Lall et. al. [Lall 2006, 2009] have also investigated the effects of overlapping stress due to thermal and mechanical loading on packaging by 26 subjecting them first to thermal loading followed by drop tests. While a most realistic approximation of their performance in real world can only be obtained by conducting product level tests, these tests are often very expensive and accounting for small design variations by conducting exhaustive testing is very time consuming. Package-level tests, which act as simplified real life impact scenarios are therefore performed to serve as an acceptable replacement of board-level drop tests that save cost and time [Yeh 2005, 2006, Ong 2003]. For their reliability assessment, electronic components are subjected to a vaiety of board level tests such as bending impact [Kim 2006], ball impact [Lai 2006], pendulum impact or Charpy [Ratchev 2007], high speed bend [Seah 2006], microimpact [Ou 2005], cold ball pull and high speed shear tests [Johnson 2007]. These tests have been performed to serve as tools to study impact toughness, fracture toughness and other contributing factors that directly impact the shock reliability of electronic components. Amongst the mechanical tests mentioned, the drop test is the most prominent. JEDEC has standardized the drop test [JEDEC 2003] by specifying the drop impact pulse and test board configuration. The standardized test serves as a common ground for assortment of semiconductor component manufacturers to compare the solder joint reliability under impact. The JEDEC standard drop test however has its limitations, in that it has too many redundant loading conditions, which result in reduced sample sizes of each loading condition for statistical analyses. Zhao et. al. [Zhao 2007] addressed this limitation of the testing standard by proposing an alternative board design with only one loading condition and a sufficiently large sample size, while applications of the response spectra to a JEDEC standard drop test board subjected to different JEDEC drop test conditions Tsai et. al. [Tsai 2007] have also been demonstrated. 27 The JEDEC standard specifies that the test vehicle should be mounted with the package side facing downwards to create a more critical loading condition [Yeh 2004]. As per the test, the drop orientation of the test vehicle is required to be horizontal or at zero degrees during the drop test. However, the drop orientation standardized by JEDEC may not be the only orientation in which the test board may fall in a real world scenario. Additionally, the supports and clamps to the board as outlined in the standard are too generic to be representative of the actual clamping conditions in a product housing. Experiments conducted on cell phones by Liu et. al. [Liu 2005], Seah et. al. [Seah 2002] and Ong et. al. [Ong 2003] have underlined the sensitivity of impact reliability to the impact angle of the product. Chong et. al. Despite its limitations, the JEDEC standard continues to be the most widely accepted drop test standard. The following figure shows the schematic of a JEDEC standard drop test of a printed circuit assembly as outlined in JEDEC JESD-B111. 28 Figure 2.1: Typical drop test apparatus and mounting scheme for PCB assembly (Source: JEDEC) 2.3 Finite Element Modeling for Transient Dynamic Simulations The Finite Element Method, owing to its versatility and accuracy in simulating complex mechanical phenomenon, has been exploited for reliability prediction in Electronic Packaging. The FE method had been used to make solder joint life predictions under different loading scenarios such as thermal cycling, drop and shock, vibration and bending. A vast majority of the life prediction models based on the finite element 29 method, account for the reliability of the components based on accumulated field quantities(deformations) and their derivatives such as like creep strain, inelastic strain energy density and nonlinear plastic work. Previous researchers have employed several modeling methods to simulate transient dynamic events. Researchers have used smeared property models [Lall 2004, 2005], Timoshenko beam models with conventional and continuum shell [Lall 2006, 2007, 2008], global-local submodels [Tee 2003, Wong 2003, Zhu 2003, 2004] and Implicit transient analysis with Input-G Method [Luan 2004] for simulating electronic packages subjected to mechanical loadings. [Wang 2001] applied [Anand 1985] unified creep model to represent viscoplastic deformation behavior of solders. He concluded that inelastic deformation behavior calculated by the model can be applied for solder joint reliability predictions. [Zhang 2000] proposed a novel life prediction model that takes into account the damage evolution to accurately predict solder joint reliability. Tee et.al. [Tee 2004] used the maximum normal peel stress of the critical solder joint extracted from finite element simulation as a failure criteria and presented a life prediction model to quantify package reliability by estimating the number of drops to failure for a package subjected to board level drop tests. Since bending is one of the prime causes of failure in electronic components mounted on PBC surface, Shetty et.al. [Shetty 2003] used three-point and four-point bending tests for evaluating the reliability of chip scale packages under curvature loads. In the study, deformation energy based empirical reliability model was used to calculate mean life time to failure, under mechanical bending. The study further established a 30 relationship between average cycles to failure and average strain energy density by volume averaging the strain energy density over the top layer of the solder joint using the volume average technique proposed by [Zhan 20003, Darveaux 1992 and 2000]. In this study, the FE method has been extensively used to simulate transient dynamic as well as static mechanical phenomenon and to make reliability predictions for solder interconnects in drop and shock. 2.4 Digital Image Correlation Owing to the complexity of electronic assemblies, it is not feasible to place deformation measuring devices such as strain gages on them. Additionally, conventional deformation measuring techniques only evaluate localized values. With a multitude of components on a single test assembly, it is beneficial to have an estimate of full field deformations for subsequent evaluation of individual component response. Digital Image Correlation (DIC) is a novel technique which works by tracking a geometric point on the specimen before and after deformation, thereby enabling full field strain and displacement measurement during a transient dynamic event. Geometric points on the test specimen are distinguished by speckle coating the surface of interest. Previous studies have established the effect of size, consistency and density of the speckle pattern on the accuracy of the method [Zhou 2001, Amodio 2003, Srinivasan 2005]. The DIC method has its merits over strain measurement by using strain gage, which measure strain only at localized points on the specimen. Furthermore, the technique does not involve contact with the specimen during the monitored event and specimen preparation is very quick and easy. 31 The Digital Image Correlation technique has been extensively used in this work to measure field quantities (deformation) and for subsequent calculation of their derivatives(strain). The technique has been employed in this work, in conjunction with high speed imaging systems to evaluate transient dynamic board response to JEDEC standard shock events as well as static bend tests. High speed photography has previously been used to measure deformations in sheet metal forming, Automotive crash testing, rail vehicle safety[Kirpatrick, 2001], air- plane safety[Marzougui, 1999], modal analysis of turbine blades, high strain rate Split- Hopkinson bar tests, dynamic fracture phenomenon, and package hermiticity (MIL-STD- 883) tests. High-speed cameras measure impact speed, force, and deformation due to shock, and thermal loading. Previously, the measurement of derivatives of field quantities, such as strains, was limited to a specific physical locations or discrete target points in an electronic structure. In the field of electronic packaging, DIC has been used to calculate full field deformations and deformation gradient in electronics [Lall 2007c, 2008b-d, 2009, Miller 2007, Park 2007a,b, 2008]. Previously, the DIC based strain measurements technique has been demonstrated to be useful for transient strain measurement in electronic assemblies, in the presence of rigid body motion Lall 2007 2008]. Digital image correlation also been employed to study deformations in printed circuit assemblies for mobile devices [Lall 2007, Miller 2007, Park 2007], material characterization [Jin 2007, Park 2007, Thompson 2007], for evaluation of stresses and strain in flip-chip dies under thermal loading [Kehoe 2006] and for calculating stresses in solder interconnects of BGA packages under thermal loading conditions [Zhou 2001, Yogel 2001, Zhang 2004, Zhang 2005, Sun 2006]. DIC 32 has also been used for evaluating elastic modulus of underfill materials at elevated temperatures during four-point bending tests [Park 2007a, Shi 2007]. In conjunction with high resolution SEM, the method has also been used to study the stresses released at the component surface before and after ion milling [Vogel 2007]. Although the technique has been widely used in thermal analysis of electronic packaging, it has recently gained a lot of attention in the field of transient dynamics as well. DIC algorithms in conjunction with high speed imaging systems are now widely used as tools for acquisition and analysis of deformation images. The technique has been extensively used in this study for evaluation of test vehicle deformation during transient dynamic as well as static mechanical events. 33 CHAPTER 3 FATIGUE CONSTANTS FOR LIFE PREDICTION OF SECOND LEVEL SOLDER INTERCONNECTS SUBJECTED TO DROP & SHOCK 3.1 Introduction Handheld electronic products such as cell phones, cameras, calculators etc are highly susceptible to shock and drop during their service, owing to their compact form factors. Electronic packages which constitute the most functional parts of these products also end up being the most vulnerable to failure as a result of shock. Second level interconnect failure has been established as the most common failure mode under application of mechanical shock. Joint Electron Devices Engineering Council (JEDEC) standardized drop test is one of the most common experimental methods used to quantify shock and vibration reliability. The test involves subjecting a test board of prescribed design, to a specified shock pulse. Owing to their small size and inaccessible feature, Dit is difficult to measure deformations and their derivatives across interconnects using strain gage and other conventional experimental techniques. In the light of increasing concern for environmental preservation, lead free solder alloys have proliferated through the electronics industry. Leadfree solder alloy systems are being increasingly used as solder interconnects in electronic packages. It is important to determine the life of these lead free alloy-systems in order to place them as a worthy 34 replacement for the lead based solder alloys. Previously, in order to evaluate the reliability of lead free interconnects in different environments, representative of their service life, they have been subjected to various mechanical tests such as drop and shock, thermal ageing and thermal cycling. Previously, many researchers have studied the effects of variation in solder alloy composition on their reliability. Low silver content SAC alloys are seen to be resistant to high strain rate under drop shock and improve drop reliability [Zhu 2008, Che 2008, Lall 2008a, Pandher 2008, 2007, Kim 2007]. Researchers have established increase in the creep rate, achieved by lowering silver content of SAC alloys [Zhang 2008]; improved drop reliability, by adding Ni in SAC alloys [Song 2008, Huang 2007]; improvement in the drop performance of SAC BGAs on Ni/Cu/Au surface finishes [Kawashiro 2008]. Solder joint reliability is seen to enhance with UBM/penetration layer/SnAgCu lead free solder bump structures [Choi 2007]. SnAgCu and SnPb based solder alloys on OSP are seen to have similar fracture energy at lower shear speed. As shear speed increases however; beyond 100mm/sec fracture energy of SAC is seen drops to zero [Sweatman 2008]. In this chapter, the concept of relative damage-index based on the lead-free interconnect transient strain history; evaluated using Digital Image Correlation and Explicit Finite Element Analyses; has been established for life-prediction of leadfree solder interconnects . Solder alloy systems studied include Sn1Ag0.5Cu (SAC105), Sn3Ag0.5Cu (SAC305) and 96.5Sn3.5Ag. Transient full field strains on the test board surface, during the shock-impact have been measured using Digital Image Correlation in conjunction with high-speed cameras operating at very high frame rates (?50000 fps). In addition the effect of sequential stresses due to thermal aging and shock-impact on the 35 failure mechanisms of the interconnects has also been studied. Previous researchers have attempted evaluation of transient dynamics at the board and component-levels using various simulation techniques including, explicit finite-elements [Lall 2004, Xie 2002, 2003, Wu 1998, 2000] implicit global models [Irving 2004, Pitaressi 2004], and global- local sub-models [Tee 2003, Wong 2003, Zhu 2001, 2003, 2004]. Life prediction of new lead-free alloy-systems under shock and vibration however, is largely beyond the state of art. In this study, explicit sub-modeling and DIC based transient strain histories of the solder interconnects have been used to develop life prediction models for solder interconnects in drop and shock environments. Three different lead-free solder alloys have been used in pristine and thermally aged states. The concept of relative damage index based on the solder interconnects strain history has been developed to quantify the damage accrue after each shock event, with an assumption of linear superposition of damage. An attempt is made to predict number of drops to failure for each package subjected to the prescribed drop tests. DIC is used to measure displacement, velocity and strain values across entire speckle coated region of a test board. Field quantities and their derivatives obtained from DIC are thereby used as boundary conditions for explicit sub- models using node-based sub-modeling technique. Strain obtained at discrete locations on the test vehicle using DIC is used to validate the accuracy of the finite element simulations of the shock events. Levenberg Marquardt algorithm ? a non liner least squares minimizing algorithm, has been used to derive key constant values (called Fatigue Constants) which define the mathematical life prediction model used to access interconnect reliability. 36 3.2 Experimental Test Boards, Setup and Procedure The test vehicles used for experimentation as a part of this study, were set to the prescribed JEDEC standard for drop testing of electronics (JESD22-B111) with overall board dimensions being 132mm? 77mm? 1mm. The test vehicles had 15 CABGA packages populated on one side; in a 3 row by 5 column format; each having an I/O count of 100 and pitch of 0.8mm). The test board was made up of FR4 substrate. Three different solder alloys compositions (SAC105, SAC 305 and Sn3.5Ag) were used as the interconnect materials. The surface finish of test board on the populated side was Ni-Au, while it was Cu-OSP on the other side. The pre-test criteria mentioned in JEDEC standard No. 22-B111 were met. The drop height was adjusted to achieve specified G level and pulse duration (1500 G, 0.5 ms, half-sine pulse). Throughout the testing process, DIC based strain values and continuity data were continuously acquired using high speed data acquisition systems. The drop- event was simultaneously monitored with ultra high-speed video cameras operating at 50,000 frames per second. 37 Figure 3.1: Test vehicle used for the study and the dropping orientations. Pa c k a g e U p si d e 0 ? d r o p o r i e n t a t i o n JE D E C 0 ? d r o p o r i e n t a t i o n I n t e r c o n n e c t a r r a y c o n f i gu r a t i o n 1 0 m m x 1 0 m m 38 Table 3.1: Package architecture of 100 I/O CABGA on test board Figure 3.2: Experimental set-up and schematic for controlled drop test with high speed cameras for DIC based strain measurements. B all C o u n t 100 B all p itch ( m m ) 0 . 8 Die Size( m m ) 5 . 5 5 Su b s tr ate T h ick n ess ( m m ) 0 . 2 3 2 Su b s tr ate P ad T y p e NSMD B all Diam eter ( m m ) 0 . 4 8 1 0 X 1 0 m m , 1 0 0 I /O, C A B GA PC CCD CCD T es t V ehi cl e D r o p T a b l e R i g i d B a s e G u i d i n g R od ? ? 2 5 ? 39 Figure 3.3: Package strain and corresponding continuity history, with peaks in continuity indicating failure. Figure 3.4: Schematic describing methodology adopted for development of life prediction models for leadfree solder interconnects.. T es t B o a r d J E D E C Sta n d a r d D r o p T es t w i th Hi g h Sp ee d Im a g i n g - 3 0 0 0 - 2 0 0 0 - 1 0 0 0 0 1000 2000 3000 4000 0 0 . 0 0 2 0 . 0 0 4 0 . 0 0 6 0 . 0 0 8 0 . 0 1 0 . 0 1 2 0 . 0 1 4 0 . 0 1 6 St r a i n D a ta ( T i m e D o m a i n ) R a i n f l ow C y cle C ou n t i n g 0 5000 10000 15000 20000 25000 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 475 500 525 550 575 600 St r a i n D a ta ( Fr eq u en c y D o m a i n ) N on L i n e a r C u r v e Fit t i n g U s i n g L M A l g or i t h m F A T I G U E C O N ST A N T S (A & n ) ? ? ? N k n k k A M 1 ) 2 ( ? =1 40 3.3 Evaluation of Fatigue Constants for Lead-Free Solder Interconnects This study introduces the concept of Relative Damage Index for quantification of damage accrues in the solder interconnects after each drop.RDI assumes linear superposition of damage after each drop and is defined such that it achieves magnitude of the damage equal to ?1? at failure. Mathematically, if damage after kth drop is Dk and total damage at failure is D, then assuming linear superposition of damage, RDI is defined as: 11 ???Nk kDD (3.28) Re-writing Miner rule and Coffin-mansion relationship based on the assumed logarithmic relationship between strain and number of cycles, the expression for RDI can be modified as: 1 )2(1 ??? ? N k nk k A M ? (3.29) where, ?k? is the bin-index for the histogram, ??/2 is the solder interconnect strain amplitude, M is the total number of bins in the histogram, N is the number of cycles subjected on the sample in the kth histogram bin during all the drops until-failure of the device, and D is the damage index. This study uses the strain histories of the critical interconnects, obtained from validated FE models and failure data obtained from experimental testing to supply Equation 3.29 with the number of bins M, strain amplitudes ??/2 and the number of cycles N. Since the strain data obtained from simulations in in amplitude-history form, it gives no information about the constituent cycles of varying both in amplitudes and 41 frequency. In order to convert the amplitude history of strain, into a more meaningful form; wherein the data was decomposed as counts of cycles of constant amplitude; a cycle counting algorithm known as the Rainflow Counting Algorithm as used. The algorithm converts the transient amplitude history of strain into histograms of constant cycle amplitudes and number of cycles. The transient strain signal analyzed, was seen to have a large number of very small strain amplitude cycles and very few large strain amplitude cycles. The Rainflow cycle counting algorithm decomposes a non uniform sequence of peaks and valleys into an equivalent set of loading blocks. A flow of rain is begun at each strain reversal in the history and is allowed to continue to flow unless, (a) the rain began at a local maximum point (peak) and falls opposite a local maximum point greater than that from which it came. (b) The rain began at a local minimum point (valley) and falls opposite a local minimum point greater (in magnitude) than that from which it came. (c) it encounters a previous rainflow. Detailed rules for cycle counting are described in [ASME 1997, Bannantine 1990, and Downing 1982]. 42 Figure 3.5: Schematic diagram explaining basic principle of Rainflow Cycle Counting Algorithm. Solder interconnect strain data for three different packages for each test board, obtained from validated submodels was used for decomposition by rainflow analysis and subsequently for computation of the fatigue constants which establish the life prediction model. From simulation, strain histories of all the solder interconnects for each package were extracted. The interconnect which experienced relatively maximum strain amplitudes was chosen for rain flow analysis. The simulated transient-dynamic solder interconnect strain data in time-domain was thereby transformed into histograms of load T i m e Str a i n R ai nf low C oun ti ng Sa m p l e Str a i n A m p l i tu d e Hi s to r y T i m e Str a i n T i m e Str a i n C y c l e s S tr a i n A m p l i tu d e Ra i n f l o w A n a l y s i s f o r C o m p r es s i v e P ea k s Ra i n f l o w An a l y s i s f o r T en s i l e P ea ks C y c l e Am p l i tu d e Hi s to g r a m 43 cycle amplitude and number of cycles. Rainflow analysis was been carried out for both pristine and thermally aged test boards for all the three solder alloy compositions. Following figures show interconnect strain history and corresponding strain amplitude histogram for all three the solder alloys, for both pristine and thermally aged states. Interconnect Strain histories are clearly seen to have a large number of very small strain amplitude cycles and very few large strain amplitude cycles. Figure 3.6: SAC105 pristine - interconnect strain history and corresponding strain amplitude histogram S A C10 5 P r is t in e P a c k a g e 3 - 3 0 0 0 - 2 0 0 0 - 1 0 0 0 0 1000 2000 3000 4000 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 T im e ( s e c ) M ic r o S t r a in ( E x x ) S A C10 5 P r is t in e P a c k a g e 3 0 5000 10000 15000 20000 25000 25 100 175 250 325 400 475 550 S t r a in A m p lit u d e ( M ic r o S t r a in ) Cyc le s S A C 1 0 5 P ri s t i n e P a c k a ge 8 - 1 5 0 0 - 1 0 0 0 - 5 0 0 0 500 1000 1500 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e ( s e c ) M i c ro S t ra i n ( E x x ) S A C 1 0 5 P ri s t i n e P a c k a ge 8 0 10000 20000 30000 40000 50000 25 50 75 100 125 150 175 200 S t ra i n A m pl i t u de ( M i c ro S t ra i n ) C y c l e s SA C 1 0 5 Pr i s ti n e Pa c k a g e 1 3 -4 0 0 0 -2 0 0 0 0 2000 4000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (s e c ) Mi c r o Str a i n (Ex x ) S A C10 5 P r is it n e P a c k a g e 1 3 0 10000 20000 30000 40000 25 75 125 175 225 275 325 375 425 475 525 S t r a i n A m p l i t u d e ( M i c r o S t r a i n ) C y c l e s 44 Figure 3.7: SAC105 Thermally aged - interconnect strain history and corresponding strain amplitude histogram SA C 1 0 5 T h e r m a l l y a g e d Pa c k a g e 2 - 1 0 0 0 0 0 10000 20000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e ( S e c ) M i c r o S t r a i n ( E x x ) SA C 1 0 5 T h e r m a l l y a g e d Pa c k a g e 2 0 500 1000 1500 2000 2500 25 2 0 0 3 7 5 5 5 0 7 2 5 9 0 0 1 0 7 5 1 2 5 0 1 4 2 5 1 6 0 0 1 7 7 5 1 9 5 0 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s SA C 1 0 5 T h e r m a l l y a g e d Pa c k a g e 1 3 -4 0 0 0 -2 0 0 0 0 2000 4000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (Se c ) Mi c r o Str a i n (Ex x ) SA C 1 0 5 T h e r m a l l y a g e d Pr i s i ti n e Pa c k a g e 1 3 0 5000 10000 25 125 225 325 425 525 625 725 825 925 1025 1125 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s SA C 1 0 5 T h e r m a l l y a g e d Pa c k a g e 1 4 -1 0 0 0 0 0 10000 20000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (Se c ) Mi c r o Str a i n (Ex x ) SA C 1 0 5 T h e r m a l l y a g e d Pa c k a g e 1 4 0 500 1000 1500 2000 2500 25 2 0 0 3 7 5 5 5 0 7 2 5 9 0 0 1 0 7 5 1 2 5 0 1 4 2 5 1 6 0 0 1 7 7 5 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s 45 Figure 3.8: Sn3.5Ag pristine - interconnect strain history and corresponding strain amplitude histogram S n 3 .5 A g P r is t in e P a c k a g e 1 - 2 0 0 0 0 - 1 0 0 0 0 0 10000 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 T im e ( s e c ) M ic r o S t r a in ( E x x ) S n 3 .5 A g P r is t in e P a c k a g e 1 0 1000 2000 3000 25 425 825 1225 1625 2025 2425 2825 3225 S t r a i n A m p l i t u d e ( M i c r o S t r a i n ) Cyc le s S n 3 .5 A g P r is t in e P a c k a g e 2 - 2 0 0 0 0 0 20000 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 T im e ( s e c ) M ic r o S t r a in ( E x x ) S n 3 .5 A g P r is t in e P a c k a g e 2 0 1000 2000 3000 25 375 725 1075 1425 1775 2125 2475 2825 T im e ( s e c ) Cyc le s S n 3 .5 A g P r is t in e P a c k a g e 1 2 - 1 5 0 0 0 - 5 0 0 0 5000 15000 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2 T im e ( S e c ) M ic r o S t r a in ( E x x ) S n 3 .5 A g P r is t in e P a c k a g e 1 2 0 1000 2000 3000 25 225 425 625 825 1025 1225 1425 1625 1825 S t r a in A m p lit u d e ( M ic r o S t r a in ) Cyc le s 46 Figure 3.9: Sn3.5Ag thermally aged - interconnect strain history and corresponding strain amplitude histogram S n 3 . 5 A g T h e r m a l l y a g e d P a c k a g e 2 - 2 0 0 0 0 - 1 0 0 0 0 0 10000 20000 0 50 100 150 200 T i m e ( s e c ) M i c r o S t r a i n ( E x x ) S n 3 . 5 A g T h e r m a l l y a g e d P a c k a g e 2 0 1000 2000 25 300 575 850 1125 1400 1675 1950 2225 2500 2775 3050 3325 S t r a i n A m p l i t u d e ( M i c r o S t r a i n ) C y c l e s S n 3 . 5 A g T h e r m a l l y a g e d P a c k a g e 4 - 2 0 0 0 0 - 1 0 0 0 0 0 10000 20000 0 50 100 150 200 T i m e ( S e c ) M i c r o S t r a i n ( E x x ) Sn 3 . 5 A g T h e r m a l l y a g e d Pa c k a g e 4 -9 0 0 100 1100 25 350 675 1000 1325 1650 1975 2300 2625 2950 3275 St r a i n A m p l i tu d e (M i c r o St r a i n ) C y c l e s S n3 . 5 A g Th e r m a l l y a ge d P a c k a ge 1 3 -5 0 0 0 -2 5 0 0 0 2500 5000 0 50 100 150 200 T i m e (S e c ) M i c r o S tr a i n (E x x ) S n 3 . 5 A g T h e r m a l l y a g e d P a c k a g e 1 3 0 5000 25 75 125 175 225 275 325 375 425 475 525 575 S t r a i n Am p l i t u d e ( M i c r o S t r a i n ) Cy c l e s 47 Figure 3.10: SAC305 pristine - interconnect strain history and corresponding strain amplitude histogram S A C 3 0 5 P r i s t i n e P a c k a g e 4 -2 0 0 0 -1 0 0 0 0 1000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (s e c ) Mi c r o Str a i n (Ex x ) SA C 3 0 5 Pr s i ti n e Pa c k a g e 4 0 5000 10000 15000 20000 10 40 70 1 0 0 1 3 0 1 6 0 1 9 0 2 2 0 2 5 0 2 8 0 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s S A C 3 0 5 P r i s t i n e P a c k a g e 8 -4 0 0 0 -2 0 0 0 0 2000 4000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (s e c ) Mi c r o Str a i n (Ex x ) SA C 3 0 5 Pr i s i tn e Pa c k a g e 8 0 2000 4000 6000 8000 10 110 210 310 410 510 610 710 810 910 1010 1110 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s S A C 3 0 5 P r i s t i n e P a c k a g e 1 4 -1 0 0 0 0 1000 2000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (Se c ) Mi c r o Str a i n (Ex x ) SA C 3 0 5 Pr i s i tn e Pa c k a g e 1 4 0 5000 10000 15000 10 40 70 1 0 0 1 3 0 1 6 0 1 9 0 2 2 0 2 5 0 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s 48 Figure 3.11: SAC305 thermally aged interconnect strain history and corresponding strain amplitude histogram Damage Superposition and Life Prediction based on Relative Damage Index In this study, the test vehicles were subjected to JEDEC standard shock pulses. Since the drops were highly repeatable, the corresponding solder interconnect strain histories were also seen to be repeatable and hence damage was extrapolated to the point of failure assuming linear superposition. The relationship between cumulative cycle S A C 3 0 5 T h e r m a l l y a g e d P a c k a g e 2 - 1 5 0 0 0 - 1 0 0 0 0 - 5 0 0 0 0 5000 10000 15000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e ( S e c ) M i c r o S t r a i n ( E x x ) SA C 3 0 5 T h e r m a l l y a g e d Pa c k a g e 2 0 1000 2000 3000 25 525 1025 1525 2025 2525 3025 3525 4025 4525 5025 5525 Str a i n A m p l i tu d e (M i c r o Str a i n ) C y c l e s S A C 3 0 5 T h e r m a l l y a g e d P a c k a g e 3 -1 0 0 0 0 1000 2000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e (Se c ) M i c r o S t r a i n ( E x x ) S A C 3 0 5 T h e r m a l l y a g e d P a c k a g e 3 0 5000 10000 15000 20000 25 75 125 175 225 275 325 375 425 475 525 S t r a i n A m p l i t u d e ( M i c r o S t r a i n ) C y c l e s S A C 3 0 5 T h e r m a l l y a g e d P a c k a g e 7 - 2 0 0 0 0 0 20000 0 0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2 T i m e ( S e c ) M i c r o S t r a i n ( E x x ) S A C 3 0 5 T h e r m a l l y a g e d P a c k a g e 7 0 5000 25 225 425 625 825 1025 1225 1425 1625 1825 2025 2225 S t r a i n A m p l i t u d e ( M i c r o S t r a i n ) C y c l e s 49 counts till present drop and the damage accrued yet, as established previously in this chapter is: 1 )2(1 ??? ? N k nk k A M ? (3.30) It should be noted that the if the cycles to failure (N) is known from dropping the test vehicle to failure, the cumulative cycles to failure Mk can be related to the cumulative solder interconnect strain amplitude ??/2 by the coefficient ?A? and the exponent ?n?. This part of the study involves evaluation of these constants, hereby referred to as Fatigue Constants. For determination of these constants, a non linear least square minimizing algorithm known as the Levenberg-Marquardt Algorithm was used. Using the interconnect strain data in the form of constant amplitude strain histograms from the simulations and the failure data from the experimentation, initial guesses to the values of ?A? and ?n? within a bounded space were given as inputs to the LM algorithm. The algorithm then explored the bounded error space for global minima. The values of the fatigue constants corresponding to the global minima were then used for the life prediction model. A typical output from the LM algorithm showing the variation of error in the multidimensional error space is shown in Figure 3.14. 50 Figure 3.12: 3d Error space - variation of error with ?A? and ?n? Levenberg-Marquardt Algorithm In this study, the values of the constants ?A? and ?n? have been obtained which in turn are used to predict the number of strain cycles and thereby the drops to failure. The relation between the number of strain cycles to failure and the strain amplitudes is non linear owing to the presence of the exponent ?n?. The optimization of the expression for Relative Damage Index for computing the values of ?A? and ?n? is complex and involves use of non-linear optimization techniques. The Levenberg Marquardt Algorithm has been employed for the same. Equation 3.30 can be expressed as f(p)=1 where p is the parameter vector comprising of ?A? and ?n? pertaining to a particular interconnect type. For obtaining the values of ?A? and ?n?, the problem is formulated as a non linear least squares problem whereby an initial parameter estimate p0 is provided and the squared error between the Pa c k a g e 2 E r r o r Sp a c e G l o b a l M i n i m a ?A ? V a l u e c o r r es p o n d i n g to m i n i m u m er r o r ?n ? V a l u e c o r r es p o n d i n g to m i n i m u m er r o r 51 desired and the obtained measurement vector is minimized. If ? =1-f(p) is the error and g(p)= ?T? is the squared error, the least square problem can be formulated as [Minimize g(p) such that : p ? R ] Here g(p) can be approximated by its Taylor series expansion as: g(p+?p)= g(p) +? g(p)T ?p (3.31) Where ? is the gradient operator. Now for the above approximation to be good, the increment ?p is to be chosen such as to minimize the error function ? =1-f(p).Then, the minimizer parameter vector p, for the error function can be represented as )()(21))((21)( 2 pgpgpgpF Ti ?? ? (3.32) Where F(p) represents the objective function for minimizing the residuals gi .For optimizing the objective function, the an initial parameter estimate p0 is provided and it is desired to find the vector p? that best satisfies the relation f(p)=1 while minimizing the squared distance g(p).An iterative procedure for finding the global minima is adopted .The iteration involves finding a descent direction ?h? and a step length ??? which gives a good decrease in the F value. The objective function for a small step length can therefore be approximated by its Taylor expansion as: F(p+?h)=F(p)+VhTF?(p) (3.33) If the function F(p+ ?h) is seen to decrease, the function is said to be proceeding in the descend direction ?h?. This implies that comparing F(p) with F(p+ ?h) should give hTF(p)<0.If no such h exists, then gradient F?(p) is zero, showing that the function is stationary. The step length is obtained by a procedure called line search whereby the distance from the current parameter vector p in the descent direction that brings about a 52 decrease in the objective function is quantified. The direction of steepest descent is known to be in the direction of the negative gradient of error i.e along h= ?F?(p).The method in which the optimization is carried in this direction h=-F?(p) is called the Steepest Descent Method. This method performs well in steep regions where slow convergence is advisable. For shallow regions, where fast conversion is advisable, the Newton-Gauss method works well. This method is based on the idea that non linear functions can be approximated as linear functions based through Taylor expansion, as the error in the system reaches a minimum. For a small ||?p||, a Taylor series expansion leads to the approximation hJ (p )g(p )h)g(p ??? (3.34) where J is the Jacobian Matrix. Substituting for g(p), in equation for F gives, )()(21))((21)( 1 2 pgpgpgpF Tmi i ?? ?? (3.35) )h()h(21)h(L)hp(F T ????? (3.36) hJJh21gJh)p(F)h(L hJJh21gJhgg21)h(L TTTT TTTTT ??? ??? (3.37) The gradient L is represented by, hJJgJhL TT ??)(' (3.38) When the gradient of a function is zero, it is known to be stationary. Therefore the descent direction towards the region in error space where the error function becomes stationary is given by 53 gJh)JJ( TgnT ?? (3.39) In each step, in the Gauss Newton method ? = 1 found by line search principle is used. Levenberg Marquardt alternates between the steepest descent method and the Gauss- Newton method depending on the proximity from the minimum in the error space. In this method a positive constant referred to as damping is added to the diagonal of JTJ to facilitate control over the convergence of the error function. The damping determines the rapidness of the convergence, large damping implying slow convergence and vice versa. A gain ratio parameter? , is used for updating the iterations carried out by the LM algorithm. Mathematically, ? is defined as )gJh(h21 )px(F)x(F )h(L)0(L )hx(F)x(F TT ?? ????? ? ???? (3.40) A large gain ratio indicates that )p(L? is a good approximation to )pp(F ?? , and there is a need for decrease in ? so that next LM step is closer to Gauss-Newton step. If ? is small then, )p(L? is a poor approximation and we should increase ? with the twofold aim of getting closer to the steepest descent direction and reducing the step size. A schematic diagram explaining the LM Algorithm is shown below. 54 Figure 3.13: Depiction of convergence of global minima in hybrid methods. Evaluation of Fatigue Constants Form the definition of cumulative damage index (equation 3.30) it can be seen that for any particular interconnect type, if the fatigue constants ?A? and ?n? are known, the number of cycles and thereby drops to failure can be predicted. The relationship between these constants and the strain cycle amplitude and frequency is non-linear. Due to this non-linearity, we have used the Levenberg-Marquardt Algorithm to evaluate ?A? and ?n?. The algorithm has been modified to incorporate the equations relating the fatigue constants and the strain parameters. E r r or D e scent D i r e ct i on E r r o r f r o m I n i ti a l G u e s s G a us s N e w t on M e t hod S t e e pes t G r a di e nt D e s c e nt M e t hod 55 The constants ?A? and ?n? stand in the expression for Cumulative Damage Index as: 1 )2(1 ??? ? N k nk k A M ? Or f(A,n)=1 (3.41) From the strain data pertaining to a particular interconnect in a package under drop loading, the strain is converted from time domain to frequency domain using the Rainflow Algorithm as explained above and a histogram is plotted. This data in frequency domain is plugged in to equation 53 such that ?k? is the bin-index for the histogram, ? e/2 is the interconnect strain amplitude while N is the total number of bins in the histogram and M is the total number of cycles in the kth bin during all the drops until failure. The two unknowns in the equation viz. the coefficient A and the exponent n are computed by providing an initial guess to the values of A and n and carrying out iterations as per the LM algorithm to explore the error space for a global minima. When iterating close to the global minima, the method does a linear approximation through Taylor expansion to the objective function in the neighborhood of the parameter to be evaluated using. This requires the calculation of the Jacobian of the function with respect to each unknown. ?? ??? ??? ? ? ??????? ?? ? ? N k kk n k A M n f 1 2lo g2 ?? (3.42) ? ? ?? ?? ? ? ? ?? ?? ? ? ? ??? ? ??? ? ??? ? ? N k k n k A M A f 1 2 2 ? (3.43) 56 Initial guess values to the parameter vector [A,n] were provided as input to the Levenberg-Marquardt algorithm. The values were varied one at a time over a range surrounding the expected values. The algorithm explores the error space by computing minimized error for each value of the parameter vector. A schematic diagram explaining the variation of A and n values is shown below. Figure 3.14: Variation in initial guesses to the ?A? and ?n? values given as an input to the algorithm. The minimized errors are plotted against the corresponding ?A? and ?n? values which gives 3-d plot of the error space as shown in the following figures. 57 Figure 3.15: Variation of ?A? and ?n? values across bounded error space and error minimization. Minimized Error 58 P a c k a ge 3 P a c k a ge 8 59 Figure 3.16: Fatigue constants computation for pristine SAC105 interconnects based on solder interconnect strain history using LM algorithm based error minimization. P a c k a ge 1 3 60 Pa c k a g e 2 Pa c k a g e 1 3 61 Figure 3.17: Fatigue constants computation for thermally aged SAC105 alloys based on solder interconnect strain history using LM algorithm based error minimization. Pa c k a g e 1 4 62 Pa c k a g e 1 Pa c k a g e 2 63 Figure 3.18: Fatigue Constants computation for pristine Sn3.5Ag interconnects based on solder interconnect strain history using LM Algorithm based error minimization. Pa c k a g e 1 2 64 Pa c k a g e 2 Pa c k a g e 4 65 Figure 3.19: Fatigue constants computation for thermally aged Sn3.5Ag interconnects based on solder interconnect strain history using LM algorithm based error minimization. Pa c k a g e 1 3 66 .. Pa c k a g e 4 Pa c k a g e 8 67 Figure 3.20: Fatigue constants computation for pristine SAC305 interconnects based on solder interconnect strain history using LM algorithm based error minimization. Pa c k a g e 1 4 68 Pa c k a g e 2 Pa c k a g e 3 69 Figure 3.21: Fatigue constants computation for thermally aged SAC305 alloys based on solder interconnect strain history using LM algorithm based error minimization. 3.4 Life Prediction of Lead-Free Alloy Based Solder Interconnects The relative damage index established in this study has been used to predict the number of drops to failure for both pristine and thermally aged solder interconnect alloys subjected to drop test. Exposure to thermal aging reduces the number of drops to failure for both experimental data and simulation. The proposed methodology enables evaluation of the damage equivalency in the application of interest and for the packaging interconnect query. Damage during the life of the product should not exceed ?1? for the design to have good survivability in drop and shock applications. The constants used for damage progression are specific to solder alloy system. Table 3.2 shows the Fatigue constants ?A? and ?n? developed using two methods. Model predictions and correlation Pa c k a g e 7 70 with experimental data for SAC105, SAC305 and Sn3.5Ag Pristine and thermally aged condition are also shown. Table 3.2: Fatigue constants and life predictions 3.5 Failure Analysis and Effect of Thermal Aging on Drop/Shock Reliability All the failed packages were cross-sectioned to investigate the failure modes. Observed failure modes include, failure at the solder copper interface board side as well as on component side and copper trace crack. The observed failures were predominant across the interface of solder interconnect and copper pad on board side for thermally aged samples of both the alloy system and in the case of pristine samples combination of two or more failure modes were observed. The thermally aged solder interconnects were observed to have lower drop and shock survivability as compared to pristine interconnects. For both solder alloy system Weibull distribution exhibit different slopes between pristine and thermally aged samples indicating the change in the failure modes. Additionally the fatigue constants evaluated for all the cases correspond to the decreased reliability of thermally aged interconnects. The ?A? value was seen to decrease from A ctual Pre di cte d Fail ure Dr op Fail ure Dr op SAC 1 05 Pristine 3.0 0 74700 -0.0 3 0.8 9 1065 1199 -12 .63 8.0 0 78799 -0.0 2 78965 -0.0 4 0.8 4 1579 1872 -18 .55 13 .00 83395 -0.0 6 1.3 1 1622 1237 23 .74 SAC 1 05 T he rmal ly Ag ed 2.0 0 75200 -0.3 8 0.9 7 138 143 -3.3 2 13 .00 74102 -0.3 3 1.1 8 226 191 15 .33 14 .00 74198 -0.4 0 0.8 4 120 143 -18 .81 Sn 3.5 Ag Pristin e 1.0 0 75197 -0.2 8 0.9 4 172 183 -6.2 8 2.0 0 80793 -0.3 0 0.8 7 162 186 -14 .73 12 .00 83699 -0.2 5 1.2 1 233 192 17 .44 Sn 3.5 Ag T he rmal ly Ag ed 2.0 0 72704 -0.4 0 1.3 6 114 84 26 .29 4.0 0 79207 -0.5 0 0.8 6 66 77 -16 .20 13 .00 77202 -0.5 0 0.8 7 97 111 -14 .78 SAC 3 05 Pristine 4.0 0 79899 -0.1 8 1.0 9 501 460 8.1 6 8.0 0 76701 -0.2 0 0.9 6 423 438 -3.6 4 14 .00 81905 -0.2 3 0.9 5 165 174 -5.2 3 SAC 3 05 T he rmal ly Ag ed 2.0 0 77095 -0.3 3 0.9 8 161 165 -2.2 9 3.0 0 83704 -0.4 3 0.7 4 163 220 -35 .17 7.0 0 75694 -0.2 3 1.1 6 215 185 13 .77 79502 -0.2 0 78831 -0.3 3 -0.3 7 76371 -0.4 7 74500 79896 -0.2 8 A v g n Computed RD I Err or(% ) A lloy Pa ck age A n A v g A 71 pristine to thermally aged conditions for all the three solder alloys compositions which is in agreement with the failure data. Additionally, the exponent ?n? was seen to increase in the negative domain from pristine to thermally aged cases which correspond to a decreased reliability after thermally aging as observed in the drop testing. This chapter entails a life prediction model which facilitates reliability prediction of solder interconnects without exhaustive testing. In this study Digital image correlation (DIC) technique in conjunction with high speed imaging was been used to acquire full filed strain, displacement and velocity components for lead-free board assemblies subjected to impact. Obtained velocity components from experimental technique (DIC) were transferred into finite element analysis as boundary conditions using node-based submodeling approach to predict solder level strain history. Failure modes for the both pristine and thermally aged lead-free alloys have been reported. Damage superposition and life prediction model of the solder interconnect based on relative damage index has been developed. Levenberg Marquardt Algorithm has been used to evaluate the fatigue constants ?A? and ?n? which govern the life prediction model. Using these constants and the strain data from a validated simulation, the life of the solder interconnects of the assembly subjected to shock tests, can be predicted in terms of the number of shock events to failure. The proposed model eliminated any need for exhaustive testing procedures, significantly improving the product development cycle. 72 CHAPTER 4 BOARD TRACE FATIGUE MODELS AND DESIGN GUIDELINES FOR ELECTRONICS UNDER SHOCK-IMPACT 4.1 Introduction Driven by the trends towards miniaturization and increased functionality, modern electronic systems are being built into more intricate and smaller packages. The mechanical robustness of these smaller and more complex packages is of great concern to the electronics industry. With advances in packaging technology, more reliable interconnects are being designed as a result of which the accountability for failure shifts to copper traces which form the primary failure mode. Previous researchers have addressed copper trace fatigue reliability [Farley 2009] and existence of copper-trace failures in drop-shock [Tee 2009]. This paper addresses the need for life prediction models for PWB copper traces in shock and vibration environments. The study focuses on low cycle fatigue of copper traces which is simulated by subjecting the PWB to cyclic three point bending. Owing to the complexity of the test vehicle, global-local finite element models were developed for simulating the board response. The study addresses the need for empirical rules for trace layout on the PWB which ensure maximum reliability. The effect of trace orientation, trace-pad fillet angle and trace width on its 73 reliability has been investigated. Using Digital Image Correlation based strain responses and Finite Element Model based stress responses, mathematical models for damage accumulation and life prediction of copper traces have been formulated and validated with experimental failure statistics. In applications ranging from handheld consumer products to military grade mission critical equipment, reliability prediction of smaller and more intricate electronic packaging configurations has been a challenge to researchers in the semiconductor industry. Solder interconnects in CSPs and BGAs have been found by the researchers to be the weakest link in the electronic assemblies, most susceptible to damage in harsh environments. Previous researchers have studied second-level interconnects with focus on design enhancement, fatigue modeling [Mattila 2004, Chai 2005, Lall 2006, tee 2004] and life prediction [Lee 2000, Lall 2009, Luan 2005]. In presence of bending stresses due to drop, shock and vibration, the copper traces on PCB surface experience high tensile stress magnitudes. These stresses are often cyclic and result in failure due to fatigue in terms of trace crack. Researchers [Tee 2008] have shown that improvements in solder interconnect design has led to a shift in mode of failure from interconnect fracture to PCB copper trace cracks. Previously, the effect of trace orientation [Tee 2009] and fatigue behavior of copper traces in cyclic mechanical loading has been studied [Farley 2009]. However, life prediction of copper traces in shock and vibration is beyond the state of art. This study focuses on detailed failure analysis of board traces subjected to high magnitudes of cyclic mechanical loads with an objective of developing board design guidelines and models for reliability prediction of copper traces subjected to mechanical loads. Emphasis is laid on analysis of copper board trace failure analysis and 74 development of a fatigue model for their life prediction. Effects of trace orientation and trace to pad dimensional ratio on its susceptibility to failure has been studied. Analysis of board traces failure is done by integrating material characterization, experimental, modeling techniques. Since the junction of traces and solder pads in electronic assemblies are most likely to fail owing to the stress concentrations in these regions, this study focuses on modeling and failure analysis of these regions of the assembly. Effect of various pads to trace fillet angles as a function of their dimensional ratio has also been studied. The effect of trace orientation on reliability has also been studied. Printed circuit board assemblies without components have been tested to isolate the copper trace failure and develop a mathematical model for life prediction of copper traces. Figure 4.1 : Schematic diagram showing methodology adopted for development of fatigue constants and evaluation of copper trace reliability T e s t Boar d Th r e e P oi n t Be n d i n g S tr ai n C or r e l ati on Wi th G l ob al F E M od e l s S t r a i n D a t a T r ac e F ati gu e b y th r e e p oi n t b e n d i n g C y c l e s t o F a i l u r e F E S i mu l ati on s S t r a i n s a t f a i l u r e l o c a t i o n s St ra i n L i f e C u rve ? ? ? ? c ff b f f N2N2 E2 ?? ? ? ?? C O PPE R T R A C E F A T I G U E MO D EL 75 4.2 Test Vehicle The test vehicle used for the study comprised of a JEDEC standard board without any components on it. The 132mm X 77mm board has 9 traces running along its length on either side. To study the effect of pad to trace dimensional ratio on its reliability, the 9 traces on either side were laid as combinations of 3 different pad and trace sizes. The absence of components on the board helps in isolating copper trace cracks from other failure modes that occur in interconnects and other parts of the component assembly. Absence of second-level interconnects eliminates the possibility of the failure mode involving copper trace delamination. The inclusion of traces laid out in a ?zig-zag? pattern enables an assessment of the effect of the various angles at which the traces merge with a solder pad. Table 1 shows the pad and copper trace combination on the test vehicle. Pad trace combinations have been studied to understand the effect of various trace geometries. 76 Figure 4.2: Test vehicle schematic showing the location and layout of traces on both sides of the board. T op S i d eBo tto m S i d e 77 Table 4.1: Different pad/trace geometric combinations incorporated in the test vehicle Measurement of Circuit Board Deformation In this study, the Digital Image Correlation technique was used to measure full field strains on the surface of the Test Vehicle. The strain field measured using DIC was to be subsequently used for validation of the FE Models developed as a part of the study. Digital image correlation (DIC) is an optical method, used to measure full field deformation and their derivatives on the surface of a loaded structure. DIC technique involves application of speckle pattern on the surface of the structure with alternate mists of black and white paints and tracking a geometric point on the speckle patterned surface 78 before and after loading and using it to compute both in-plane as well as out-of-plane deformations in the structures. Previously, Digital Image Correlation (DIC) has been used in the electronic industry for various applications. DIC is used to measure full field displacement and deformation gradient in electronic assemblies subjected to drop and shock [Lall 2007b, 2008a,b, Miller 2007, Park 2007, 2008], damping ratio on the surface of the board [Peterson 2008] examination of velocity, rotation, bending on portable products subjected to impact test [Scheijgrond 2005], stresses in solder interconnects of BGA packages under thermal loading [Zhou 2001, Yogel 2001, Rajendra 2002, Shi 2004, Zhang 2004, Zhang 2005, Xu 2006, Bieler 2006, Sun 2006], stresses and strain in flip-chip die under thermal loading [Kehoe 2006] and warpage measurement on dual chip module [Ouimet 2008]. However, the application of DIC technique to study solder level deformation and strain for transient event such as shock and drop of the PCB assembly has not been completely explored. Behavior of the solder interconnects in printed circuit board subjected to shock and drop is a major reliability issue in portable electronics. In the Digital Image Correlation technique displacement field quantities are obtained by tracking a geometric point before and after deformation [Zhou 2001, Amodio 2003, Srinivasan 2005, Kehoe 2006 and Lall 2007b, 2008a]. The tracking is achieved using digital image processing of speckle pattern on the specimen surface. Figure 4.3 describes the basic principle of DIC. 79 Figure 4.3: 3D-Digital Image Correlation (DIC) principle The sub image before impact is referred to as I1 (r) and the one after impact as I2 (r) respectively. The two are related as follows: (3.1) (3.2) Where U(r) is the displacement vector at pixel r = (x, y, z)T . The difference in the positions of the current pixel and the reference pixel gives the in-plane as well as out- of-plane displacement U(r) of this reference pixel. Full-field displacement can thus be found out by changing the reference pixel location on the speckle patterned surface and following the same procedure described as above. A subimage around a reference pixel ?r? in the undeformed image is then compared with the subimages corresponding to different pixels in the deformed image using a predefined correlation function. Difference in the intensity of the two subimages is used to evaluate local translation of speckles. Three typical correlation functions are defined as follows: [Zhou 2001] )]r(Ur[I)r(I 12 ?? )]r(Ur[I)r(I 21 ?? 80 Absolute difference: ???? ?? ???? dr)r( dr)r()'rr(1)'r( I IIC 1 12A (3.3) Least square: ???? ?? ???? dr)r( dr)]r()'rr([1)'r( 21 212 L I IIC (3.4) and Cross-Correlation: ? ? 2/12221 21C dr)'rr(dr)r( dr)'rr(1)'r( II I)r(IC ?? ????? ?? ???? (3.5) where ? (M x N) is the area of the sub-image around reference pixel r, r` is the current pixel, CA(r`) is the current absolute correlation function, CL(r`) is the current least square correlation function, and CC(r`) is the current cross-correlation function. The cross- correlation functions provide the correspondence between matching subsets in images of the un-deformed and deformed states. It is an iterative spatial-domain cross-correlation method. This method maximizes the cross-correlation coefficient between a subset in the reference image I1 and the deformed image I2. 4.3 Accelerated Testing Stresses experienced by board traces caused due to drop and shock in actual operating environment were simulated by subjecting the PCB to cyclic repetitive loading. The experimental procedure involved determining overstress destruct limits of the PCB and then performing accelerated tests at magnitudes well below the overstress limits. The experimental setup involved a 3-point bending fixture used in conjunction with a universal testing machine (Figure 4.4). The board was supported in two places, 81 symmetric about its center and subjected to prescribed out-of-plane deflection in the center. Owing to the nature and symmetry of loading, it was expected that failure in the traces would occur in tension, and therefore the traces on either side were subjected to cyclic tensile stresses until failure. The test board was allowed to ?roll? at the supports to avoid membrane stress in the board. Strain data was acquired by placing strain gages at six strategic locations on the board. The prescribed out of plane deflection was chosen such that strains indicated by indicated by strategically placed strain gages were linear up till their highest limits. Figure 4.4: Schematic showing nature and direction of loading for accelerated testing The test vehicle was tested at different loading rates to investigate the effect of rate of loading on modes of failure. The test vehicles were failed by subjecting them to two different maximum cyclic stress levels to obtain different cyclic-life corresponding to different stress levels in order to populate a ?S-N curve describing the low cycle fatigue behavior of the traces. This curve was further used in development of a power law describing the fatigue life of the traces. Continuity across the traces was monitored using 82 a high speed data logger and failure data was obtained by recording the cycle counts at which traces failed. Failed traces were then observed under a microscope for failure analysis. It was observed that in all the cases the failure in traces occurred only at the pad/trace joint. This is attributed to the stress concentrations at these locations due to the discontinuities in geometry. The failure location on each trace was noted for comparison with subsequent finite element analysis. 4.4 Finite Element Modeling Owing to the complexity of the printed circuit board and in order to preserve the geometrical details of the traces, a Global-Local sub modeling approach was used. Since the failures observed during accelerated testing were localized in the trace/pad joints, sub models preserving acute geometric details of these regions of the PCB were developed. A defeatured global model of the PCB was created without copper trace architectural details (Figure 4.5).Outputs from the global model were used to drive the local sub-models of all the trace pad junctions on each trace. 83 Figure 4.5: Contour plot (global model) showing variation of longitudinal tensile stresses on the PCB 84 Figure 4.6 Schematic describing the finite element modeling methodology used, with implementation of global-local submodeling D IC Ba s e d F ul l F i e l d S t ra i n G l oba l M ode l S i m ul a t i on S ubm ode l D ri ve n by G l oba l M ode l Re s ul t s S ubm ode l S i m ul a t i on us e d for F a i l ure P re di c t i on 85 Figure 4.7: Global-local submodeling: Using outputs from driving nodes of the validated global model to drive corresponding driven nodes of the submodel Trace and Pad Sub-Models Detailed Submodels were developed for all of the trace-pad junctions. All the pad/trace junctions on each trace were driven as a local sub model deriving its boundary conditions from the output of the global model. V a l i da t e d G l oba l M ode l D e t a i l e d S ubm ode l D ri vi ng N ode s D ri ve n N ode s 86 Figure 4.8: Submodels for different straight pad-trace junction geometries Figure 4.9: Sub models for different angled pad/trace junction geometries Three categories of Submodels were made for each trace with a different trace/pad dimensional ratio. 1) Nine submodels detailing the nine straight trace/pad junctions with different trace/pad ratios (Figure 4.8); 2) Nine Submodels detailing the nine angled pad/trace junctions with different trace/pad ratios (Figure 4.9); 3) Trac e - 1 Trac e - 2 Trac e - 3 Trac e - 4 Trac e - 5 T race - 6 Trac e - 8Trac e - 7 87 Multidirectional trace Submodels with traces routed from 0? to 315? at intervals of 45? (Figure 4.10). Figure 4.10: Multidirectional trace sub-models. Model Validation The finite element models developed for the trace-pad geometries were subjected to boundary conditions representing a three point bending test with loading magnitudes equal to the accelerated test parameters. The full field strain contours obtained from DIC based measurements were compared with FE simulation prediction to validate the FE model. A good correlation was obtained between the two. 88 Figure 4.11: Comparison of strains contours as obtained by DIC and as predicted by FEA. F E S i mu l ati on P r e d i c te d S tr ai n s 89 Additionally, a good correlation was achieved between areas of high stresses as shown by the simulations and the actual failure locations (Figure 4.13 & Figure 4.14). Figure 4.14 shows the correlation of model predictions with the failure locations. Figure 4.12 : Variation of Longitudinal stresses along the length for different trace geometries Figure 4.13: Failure locations, as observed from FE simulations and accelerated tests for straight traces V a r i a t i o n o f S t r e s s e s A l o n g l e n g t h o f B o a r d 0 . 0 0 E + 0 0 1 . 0 0 E + 0 8 2 . 0 0 E + 0 8 3 . 0 0 E + 0 8 4 . 0 0 E + 0 8 5 . 0 0 E + 0 8 6 . 0 0 E + 0 8 7 . 0 0 E + 0 8 8 . 0 0 E + 0 8 0 1 2 3 4 5 6 T r a c e L o c a t i o n s S t r e s s e s ( P a ) T r a c e - 1 T r a c e - 2 T r a c e - 3 T r a c e - 4 T r a c e - 5 T r a c e - 6 T r a c e - 7 T r a c e - 8 T race Nu mbe rs A1 A2 A3 B1 B2 B3 C1 C2 C3Model P re dic ti on s O bse rved F ail ur e 90 Figure 4.14: Model predictions of stress concentrations and observed locations of failure. 91 4.5 Trace Design Assessment Effect of Trace/Pad ratio Detailed failure analysis coupled with finite element simulations of the failed regions showed a trend in the failure pattern which could be related to the geometries of the trace/pad junctions. Stress levels in the trace-pad junctions were seen to decrease with a decreased difference in trace width and pad diameter. Figure 4.15: Comparison of peak strains in different trace/pad geometries. 92 Figure 4.16: Increasing stress concentration with increasing difference in trace width and pad diameter. For similar trace/pad ratio traces with bigger pads were seen to be more reliable than those with smaller pads( Figure 4.15). The traces furthest away from fixed supports I n c r e a s i n g D i f f e r e n c e i n T r a c e W i d t h a n d P a d D i a D e c r e a s i n g L o n g i t u d i n a l S t r a i n L e v e l s 93 were seen to fail first. The trace geometries correspond to the three trace groups shown in Figure 9 and Table 1. A similar trend of increased strain with increase in (pad diameter, d)/(trace width, w) is observed for all pad diameters. Effect of Trace Orientation From an assessment of the trace geometries tested, the traces with least trace/pad ratio were seen to be most reliable. Multi directional trace models were developed for these traces with traces routed from 0? to 315? at intervals of 45? (Figure 12). Figure 4.17: Variation in maximum principal stress with angle made with longitudinal direction of the board 94 Figure 4.18: Effect of trace orientation on max principal strain The finite element simulations revealed a trend in the stress levels in traces oriented at different angles. The traces which were laid at an angle closer to 0 and 180 degrees were seen to experience highest levels of stresses (Figure 13). The maximum stresses in traces were seen to reduce as the angle approached 90?. Effect of ?Turning Angle? From finite element simulations of the angled trace geometries, the effect of ?turning angle and its best possible orientation on a PCB was analyzed. It was observed that irrespective of the trace width and the trace/pad ratio, maximum stresses occurred at the fillet between the two traces, at the point where they met a pad. 0 0 . 0 0 1 0 . 0 0 2 0 . 0 0 3 0 45 90 135 180 225 270 315 M a x i m u m Pr i n c i p a l Str a i n A n g l e w i th l e n g th o f th e b o a r d E f f e c t o f t r a c e A n g l e o n M a x i m u m P r i n c i p a l S t r a i n 95 Figure 4.19: A comparison of stress concentrations developed at the trace-pad junctions for angled traces. The longitudinal stresses in these traces produced a tearing effect at these locations resulting in crack initiation at the fillets (Figure 4.19). Model predictions indicate that sharp angles increased the strain concentration and obtuse turning angles sustained lower strain concentrations. 96 Effect of Trace Routing Directions In terms of the trace routing angle orientation, It was observed that if the traces were routed in such at way that the trace ?bends? opened up and closed under flexural bending of the board, the stress concentrations produced were very high and resulted in crack initiation at inner edges of the bend. It is suggested that the trace routing angles should be made as obtuse as possible to avoid stress concentrations. Additionally, bends in traces, should be placed such that they do not open up when the board is flexed as indicated in the figure that follows. Figure 4.20: Stress concentrations at sharp angles between traces. Fail u re Location 97 Figure 4.21 : Figure showing optimum trace orientation for maximum reliability Copper Trace Fatigue Failure Model Combining the life distributions from experimental testing and stress magnitudes at failure locations from the Finite element simulations, a relationship describing the fatigue behavior of thin film copper traces was developed. A generalized strain life power law is used to describe the fatigue failure of thin film copper traces. The strain data used is derived from the full-field measurements and are subject to its accuracy. The durability of the copper in PCB traces as observed from the study, can be described using a generalized strain life power law given as: ? ? ? ?cffbff N2N2E2 ?????? (4.6) Where ?? is the total strain range at failure location in a cycle, f? and f? are the failure stress and strains while Nf is the cycle count to failure. Owing to the nature of the accelerated tests conducted, the failures incurred in the traces were due to low cycle fatigue and therefore this equation can be reduced to: Bf AN )( ??? (4.7) L ongi t udi na l D i re c t i on O f Boa rd Re d u c e d S t r a i n l e v e l s due to p ro p e r t ra c e o ri e n t a t i o n H i g h S t ra i n l e v e l s ?t e a ri n g ? e ffe c t p ro d u c e d b y b o a rd fl e x u re . 98 The fit values from experimental data are: A = 8.63 x 10-27 and B = -12.195. The relation between the strain life in terms of cycles to failure and the trace strain values can be approximated by a log-log function. For the dataset obtained from the accelerated tests and the finite element simulations, the best fit to the dataset was obtained by the function: Figure 4.22: Fatigue curve describing the low cycle fatigue behavior of copper traces 4.6 Copper Trace Design Guidelines Accelerated stress tests have been performed on a test vehicle having only copper traces in order to isolate the copper trace failure modes. The effect of trace dimension, trace/pad ratio and trace orientation on their reliability have been studied. The following observations have been made 1. Failures in copper traces subjected to tensile stresses due to PCB bending in drop and shock, are localized at pad/trace junctions. This can be attributed to the stress concentrations at these locations sue to geometric discontinuities. C u Str a i n V s C y c l e s to F a i l u r e y = - 0 . 0 0 0 4 L n ( x ) + 0 . 0 0 7 1 3 . 0 0 E - 0 3 3 . 5 0 E - 0 3 4 . 0 0 E - 0 3 4 . 5 0 E - 0 3 5 . 0 0 E - 0 3 100 1000 C y c l e s t o F a i l u r e ( N f ) C o p p e r T r a c e S t r a i n C y c l e s t o F a i l u r e ( N f ) 99 2. The traces with least difference in trace and pad dimensions are more reliable. For traces with similar trace/pad dimensional ratio, wider traces and bigger pads are seen to be more reliable. 3. The traces laid out on the PCB, at orientations parallel to the longitudinal direction of the board, are least reliable when the PCB is subjected to bending due to drop or vibration. The stress levels in the traces reduce and their reliability increases as the angle of their orientation becomes closer to the perpendicular to the length of the board. 4. The traces, when turned to conform to the circuit layout, should be turned at as obtuse an angle as possible to prevent high stress concentrations. Also, the turns should be oriented such that the flexing of the circuit board under action of any shock/vibration does not cause the bend to open up or close. Copper Trace Design guidelines 1. The traces should be made wider at the trace pad interfaces to minimize the difference in trace/pad dimensional ratio. If the width if the traces are a concern with regard to material consumption and trace density, the trace width can be reduced after a certain distance from the trace/pad interface. 2. Since the trace failures were seen to occur mostly in exposed (regions not covered by solder mask) regions of the trace, any discontinuities in the trace dimensions in these regions should be avoided. 3. The traces should be laid out on the PCB at angles as close to 90? (w.r.t the length of the board) as possible. If the layout of components on the PCB does 100 not support this, the traces should at least be routed from the pads at angles close to 90? and oriented to lie in other directions, by turning the trace at shallow angles. In order to develop a mathematical model describing the fatigue behavior of the traces, a Strain vs. cycles to failure curve has been populated from the results obtained from accelerated stress tests. Fatigue Constants governing the linear-log function based mathematical model characterizing the fatigue behavior of copper traces on PCBs subjected to drop/shock or vibration have been developed. The model which can be driven by strains predicted by finite element simulations can be used for assessment of low cycle fatigue behavior of copper traces and their life prediction eliminating any need for exhaustive testing procedures. 101 CHAPTER 5 RELIABILITY STUDIES ON PACKAGE-ON-PACKAGE COMPONENTS IN DROP AND SHOCK ENVIRONMENTS 5.1 Introduction The consumer electronics industry stands at a critical juncture where manufacturers strive to incorporate more functionality in smaller packages. In the highly competitive consumer electronics market, a continued demand for products with smallest possible form-factor yet high functionality has led to the proliferation of 3D packaging technologies. Package-on-Package (PoP) architectures, in particular have attracted a lot of interest, especially in portable electronics industry. The advantages of these stacked 3D architectures include simplified and compact design, savings of board space allowing for more package landings, reduced pin counts and optimized production costs. While a lot of recent research, in the field of PoP architectures has been focused on development of optimum process flows and warpage control during reflow, the effects of reflow parameters on the quality of PoP build and the associated reflow defects(other than warpage) have not been extensively researched. Additionally, the reliability issues associated with PoP assemblies in drop and shock environments have largely been neglected. Since PoP architectures find their applications mainly in portable electronics, which are susceptible to frequent drops and careless handling at the hand of the consumer, the reliability of PoP architectures in environments representative of the real world is critical to their success in the industry. In this study, Single component PoP test vehicles were fabricated as per JEDEC standards for quantifying the reliability of PoP packages in drop and shock. Daisy chained double-stack PoP components were used to identify failure for subsequent drop/shock performance analysis. Experimental strain data acquired using Digital Image Correlation and high speed continuity data- for identifying failure was used in conjunction with validated FE simulations of drop test events; for assessment of reliability of PoP architectures in drop/shock environments. Validated node based global- local FE simulations were used to predict strains in critical solder balls in both layers of the PoP stack. The drop/shock reliability studies and life prediction models presented in this chapter, present an insight into PoP failures and eliminate the need for exhaustive testing procedures. 5.2 Package-on-Package (PoP) Architecture Driven by the trends towards miniaturization and increased functionality, the electronics components used in consumer and industrial products are being made smaller and more intricate. In order to achieve high functionality, more components are being placed within smaller housings, which lead to exhaustion of XY space, leaving scope for further population only in out-of-plane directions. The concept of stacking packages, exploits this availability of space in z-direction to pack more functionality in smaller form factors. In recent years, Package-on-Package assemblies, more commonly known as PoP assemblies have gained wide acceptance, especially in the portable electronics 103 market. PoP assemblies cater to the demand of high memory density and powerful processing all, built into one package stack. PoP is an integrated circuit technology that vertically combines discrete memory and logic devices into a single module. A typical PoP consist of a high performance logic device as the bottom package , while high capacity or combination memory device is fabricated as mating stacked die BGA to serve as the top package [Dreiza, 2005]. Figure 5.1: A schematic of Package-On-Package stack showing both the constituent packages viz. the bottom (PSvfBGA) and the top (CSP) packages. PoP assemblies have always been compared to Stacked Die CSPs (SCSP). With proliferations in advanced wafer thinning technology, SCSPs facilitate low package profiles while being compliant with regular SMT assembly procedures and with relatively low packaging cost. PoPs however have their own advantages including flexibility in memory and processor combinations, and scope for testing at individual package level instead of the memory-logic device system as a whole. 104 While PoP architecture is gaining acceptance in electronics industry, low assembly yield and reliability issues associated its their drop/shock and thermal performances are of great concern to the manufacturers. In order for PoP assemblies to proliferate throughout the packaging industry, there is an urging need for development of standardized assembly processes which give good yields and reliable builds. Additionally, with the development of technology, packages will be manufactured with increased I/O counts and reduced pitch which would result in lower standoff height between top and bottom packages- which further highlights the need for standardized assembly processes and established reliability quantifications. This study addresses the need for a standard assembly process and an insight into board level reliability of PoP assemblies. This chapter validates a previously suggested assembly process and suggests an optimum reflow profile for assembling PoP modules while minimizing reflow defects. Furthermore, this part of the study aims at quantification of board level reliability of PoP assemblies by performing JEDEC standard tests followed by detailed failure analysis to isolate failures in mode and occurrence. 5.3 PoP Test Vehicle For this study, a JEDEC standard test board was fabricated at NSF-CAVE3. The test board made up of FR-4 had a single PoP assembly, populated on one side only. The PoP stack fabricated, consisted of a 12mm X 12mm Package-Stackable-Very-Fine-Pitch BGA (PSvfBGA) as the bottom package with another 12mm X 12mm CSP stacked on its top. The bottom package (PvfBGA) had the solder balls arranged in a perimeter-array configuration with an I/O count of 305, while the top package (CSP), also a perimeter 105 array package, had an I/O count of 128. The pitch between the solder balls was 0.5mm for the bottom tier and 0.65mm for the top tier. Different solder alloys, representative of the actual combinations used in the industry, were used in both top and bottom packages for achieving extended reliability. The total height of the package stack was 1.2mm. Figure 5.2: A schematic of Package-On-Package test vehicle developed, showing the three different daisy chain channels and other board attributes. 106 Figure 5.3: Schematic describing test vehicle channels and other attributes Figure 5.4: Single component JEDEC standard Package-on-Package test vehicle, fabricated and used for the study. A+, A - : D a i sy C h a i n o f P S vf B G A , B o t t o m p a cka g e b a l l s B+, B - : D a i sy ch a i n o f t o p j u n ct i o n ( T o p F B G A t o P S vf B G A 1 1 6 p i n m e m o r y i n t e r f a ce ) C +, C - : D a i sy ch a i n o f t o p j u n ct i o n ( T o p F B G A t o P S vf B G A 1 2 co r n e r b a l l s) ? D ai s y ch ai n s fo r t o p j u n ct i o n (t o p FB G A t o PS v fBG A ) ro u t ed fro m ?i n s id e? p ack ag e fo o tp ri n t v ia th ro u g h - h o l es 107 Having a single component on the test vehicle, with the interconnects which are being monitored for continuity, divided into several different daisy chains, facilitates easier fault isolation for subsequent failure analysis. For this purpose, the solder balls in the two tiers of the PoP stack were divided into three different daisy chains. The top tier of solder balls between the bottom package (PSvfBGA) and the top package (CSP) was divided into two different daisy chains- one connecting twelve corner solder balls, three in each corner and the second connecting all the other solder balls. These two daisy chains, connected to the copper pads on the PCB through two pairs of solder balls from the bottom tier. The bottom tier of solder balls, between the package stack and the PCB was included in a different daisy chain. Figure 5.5: Schematic showing matching daisy chain net-lists on the top of the bottom package (PSvfBGA) and the bottom of top package (CSP) (Source: Amkor Technology) 108 Figure 5.6 : Schematic showing matching daisy chain net-lists on the bottom of the bottom package (PsvfBGA) and the PCB (Source: Amkor Technology) 109 Figure 5.7 : Schematics showing the classification of the two tiers of Solder Interconnects into three daisy chain net-lists for the test component used in the study (Source: Amkor Technology) D a i s y Ch a i n n e t l i s t o f P S v fBG A , Bo t t o m p a c k a g e b a l l s D a i s y c h a i n n e t l i s t o f t o p j u n c t i o n (T o p F BG A t o P S v fBG A 1 1 6 p i n m e m o ry i n t e rfa c e ) D a i s y Ch a i n n e t l i s t o f t o p s i d e (T o p F BG A t o P S v fBG A 1 2 c o rn e r b a l l s ) 110 Figure 5.8: Schematic of PoP stack with relevant dimensions. Since SAC105 for Ni/Au finish and SAC125Ni for Cu OSP finish have been highly recommended for applications requiring improved board level drop [Syed 2006]; SAC105 and SAC125Ni were used as solder for the balls on the top and bottom tiers of the PoP stack, respectively. Additionally, SAC105 and SAC125Ni have been shown to have better performance in board level drop tests as compared to SAC305 and SAC405 solder alloys and hence were preferred in the test vehicle. 5.4 PoP: Surface Mount Assembly A regular BGA surface mount assembly process involves three basic steps- 1) Stencil printing the solder paste on the PWB copper pads 2) placing the package on the PWB while matching package footprint to the landing pads and 3) Reflow. While the process has been well established and optimized for BGA assembly, it cannot be utilized 111 for Package on Package stacking without significant alterations. The factors which render the regular BGA assembly process unsuitable for PoP assembly are as follows: Firstly, the bottom package in a PoP assembly is generally of a very thin profile which tends to induce a lot of warpage during reflow owing to CTE mismatch between the laminate and the mold cap. Regular solder paste printing cannot efficiently address the issue of warpage in the bottom package. In a regular screen printing process, the thickness of the metal mask is of the order of 120um which owing to half of it being flux, generates a solder paste deposition of around 60um thickness. Compared to a typical bottom package warpage, which at its peak can be as much as about 100um, this thickness of solder paste does not suffice to yield a good connection for all the solder balls. On the other hand, if the thickness of the solder paste deposition is increased, it has been shown[Yoshida,2006] that the surface tension from the large amount of the transcribed solder results in pushing the substrate of the bottom package up during reflow. Secondly, to enable lower stack profile, the bottom package is designed to have a mold cap surrounded by copper pads on the substrate for receiving the solder balls from the top package (CSP). This results in a non-level surface on the top of the bottom package which renders it unsuitable for screen printing solder on the pads. Additionally, it?s necessary to supply adequate mount of flux and/or solder paste between the solder balls from the top CSP and the copper pads on the bottom package to remove the oxide film and achieve good connections. Regular solder pastes, used for stencil printing applications have very high viscosity which results in a lower transcription by dipping. For PoP assembly, novel solder pastes have been developed which facilitate transfer of solder paste on to the 112 solder balls by dipping the packed in the solder paste. These novel PoP pastes, have lower viscosity and smaller grain size [Takagi, 2005] which results in better ?pick-up? of paste after dipping. Figure 5.9: A comparison between regular and PoP-specific solder paste, showing a difference in amount of transcription achieved by dipping. The solder dipping process, necessary for transcribing solder paste to the top tier of solder balls, can also be extended for solder paste supply to the bottom tier of solder balls. It has been shown that using solder paste dipping instead of screen printing for the bottom package also, results in a better yield. For the PoP assembly used in this study, the process flow was as follows: 1. Dipping the bottom package (PSvfBGA) into low-viscosity PoP solder paste 2. Placement of the bottom package on the PCB 3. Dipping the top package (CSP) into PoP solder paste 4. Placement of the top package (CSP) on the bottom package (PSvfBGA) 5. Reflow R e g u l a r S o l d e r P a s t e S o l d e r P a s t e D e s i g n e d F o r P o P A s s e m b l y 113 Figure 5.10: Process flow adopted for assembling Package-on-Package components on the test vehicle It has been shown that the paste dipping depths that enable solder paste coverage of up to 50-65% of solder ball height yield best results. Accordingly, two different dipping channels of 0.12mm and 0.25mm depths were used for dipping the bottom and the top packages owing to differences in their solder ball dimensions. The process involved doctor-blading the solder paste, to conform to the channel depth followed by dipping the package in the paste, with controlled force. The Solder balls were optically examined for good deposition before placement and subsequent reflow. D i p P S vf B G A i n P a s t e P l a c e P S vf B G A on B oa r d D i p C S P i n P a s t e P l a c e C S P on P S vf B G A R e f l ow 114 5.5 Reflow Parameters Reflow Profile has a very important effect on the final yield of PoP assembly process. In this study, various experiments were conducted to study the effect of reflow profile parameters on PoP assembly. The defects that occurred as a result of experimenting with the reflow profiles were analyzed for optimization of the reflow profile. A major factor influencing the quality of the PoP build was warpage of the package. The warpage during reflow at high temperatures, especially in the bottom package, has been reported to be as large as 100um. A mismatch in the CTE of the various components of the package viz. the die, the mold cap and the substrate, results in a negative (concave) warpage of the bottom package and positive (convex) warpage of the top package [Yim, 2010]. Additionally, the use of different solder alloys for assembly, all of them Lead-free Compositions make it even more challenging to accommodate reflow parameters that minimize warpage while staying within the narrow reflow process window that is required to achieve optimum yield with lead-free alloys. Figure 5.11: Mechanism of warpage due to CTE mismatch. While the solder dipping procedure accommodates warpage due to reflow to an extent, by supplying an adequate amount of solder paste/flux, large values of warpage, still result in open connections. Owing to the geometry of the component, the warpage is 115 seen to be more severe in the bottom package, especially near the corner of the package. As a result a lot of open connections were observed in the corner solder balls of the bottom tier of I/Os. Previously, researchers have outlined PoP stacking process flow for high yields [Yoshida 2006, Dreiza 2006]; investigated the effects of material properties on PoP package warpage behaviors [Yim, 2010] and developed models for warpage prediction in PoP stacks during reflow. [Amagai, 2010]. However reflow defects, other than associated with PoP assembly have largely been neglected. Variations to the reflow profile parameters, suggested to account for warpage in the PoP constitutive packages during reflow, often lead to other defects. Bound by the propensity of these defects, the optimum profile window becomes very small. This study, aims at optimization of the PoP reflow profile accounting for all the major reflow defects. As a part of the study, different reflow profiles with varying Ramp Rates, Peak Temperatures and Time-Above-Liquidus were used to reflow the assembly. Visual inspection of the assembly following the reflow was performed to analyze the effects of change in these reflow parameters on the yield of the process. The most common types of defects observed during reflow are as follows: 1. Tombstoning: wherein one side of the package gets lifted up and the I/Os on the other side solidify first, resulting in an improper package landing. 2. Open Joints, resulting from excessive warpage in the top and bottom package. 3. Bridging or Sliding of I/Os due to high peak temperatures. 4. Improper Reflow: Un-collapsed solder balls due to low peak temperatures of less TAL. 116 5. Solder-Balling i.e. accumulation of tiny solder ball from the solder paste, non the peripheral edge of the solder ball after reflow. Figure 5.12: Solder reflow defect: Excessive warpage and tombstoning resulting in open connections Figure 5.13: Solder reflow defect: Improper reflow due to low peak temperature and less TAL resulting in un-collapsed solder joints 117 Figure 5.14: Solder reflow defect: Bridging of interconnections leading to short- circuits, resulting from high peak temperature and/or high TAL Figure 5.15: Solder reflow defect : Head-in-Pillow defect resulting from premature activation of flux due to high ramp rate and failure in removal of oxide film. 118 Figure 5.16: Solder reflow defect : Sliding of the package owing to high warpage due to thermal shock Figure 5.17: Good solder interconnections obtained by optimized reflow profile. An optimized reflow profile was developed after a detailed analysis of the effects of the aforementioned parameters on the quality of the build. The optimized reflow profile had the following attributes: 119 1. A gentle Ramp-rate of about 1.5?C /min which was neither too high to result in bridging, tombstoning or excessive package warpage nor too low to fail in activating the flux in the solder paste and result in Solder-Balling or Head-in- pillow type defects due to presence of oxide film 2. Ramp-to-Spike profile which unlike Ramp-Soak-spike profile does not prematurely break down the flux activators in the solder paste, instead preserves then throughout the preheat stage promoting better wetting. 3. Optimum Peak temperature of 255?C which is well above the liquidus of the solder alloys used (? 217?C) but not too high to cause bridging or tombstoning 4. Optimum TAL of about 90 sec which is long enough for proper formation of solder joint while not being too long to cause de-wetting and induce head-in- pillow type defects. Figure 5.18: Different reflow profiles adopted with different parameters. The optimum profile (Profile-2) is highlighted in red. 120 Figure 5.19: Effect of reflow parameters on PoP assembly and induced defects. 5.6 PoP: Drop Testing For evaluation of the Package-on-Package assembly reliability in shock, the test vehicles were subjected to JEDEC standard drop, inducing a shock pulse of 1500G for 0.5ms. The continuity of all the three daisy chains were continuously monitored throughout the drop event using high speed DAQ systems. The test boards were speckle coated and the drop event recorded using High speed cameras for subsequent evaluation of in-situ full field strain using Digital image Correlation technique. 121 Figure 5.20: Schematic of experimental set-up for controlled drop test with DIC and DAQ systems H i g h S pe e d C a m e r a s ( 1 0 , 0 0 0 f ps) T e st V e h i c l e Pu l se S h a p e r D r o p T a b l e D r o p T o w e r D IC S e t u p Hi - S pe e d D a t a A c qu i si t i o n f o r C o nt i nu i t y M o ni t o r i ng 122 Figure 5.21: Experimental set-up for controlled JEDEC drop test using LANSMONT? drop test tower. Digital image correlation (DIC) is an optical method to measure full field deformation and their derivatives i.e strain on the surface of a loaded structure. The Digital Image Correlation technique involves application of speckle pattern on the surface of the structure with alternate mists of black and white paints which are then tracked as geometrical points during the monitored event, before and after loading and using it to compute both in-plane as well as out-of-plane deformations in the structures[Zhou 2001, Lall2007]. 123 Figure 5.22: 3D-Digital Image Correlation- Basic principle [Panchagade, 2006] High speed Data Acquisition systems were used to monitor the continuity of the three daisy chains through the drop event, for detection of failure. Failure in the assembly was defined as a peak in the resistance in any of the daisy chains which occurs as crack propagates through an I/O resulting in an open connection. After performing drop tests on the test-vehicle, until all the three daisy chains were rendered open the failed assemblies were preserved for subsequent failure analysis failure analysis. Previously it has been shown that small variation in drop orientation can produce significantly varying transient-dynamic board responses. To avoid variations in drop orientation, significant effort was made to establish repeatability of the drop event. 124 Figure 5.23: Strain repeatability through different shock events To ensure repeatability of the drop event, strain histories, monitored using strain gages mounted at key location on the test board and shock pulse measured using an accelerometer were monitored for consistency. 5.7 Finite Element Models For Reliability Evaluation Finite elements models were developed for simulating the drop/shock events and evaluating the peak strain and stress values and the specific locations where they occur. Owing to the complexity of the package on Package assembly and in order to preserve the geometrical intricacies of the package stack in complete detail, a node based Global- Local sub modeling approach was used. Since the failures observed during accelerated drop testing were primarily seen to be localized in the corner solder balls, these I/Os were modeled in complete detail while all the other Interconnects were modeled as Timoshenko beam elements. A defeatured global model of the test vehicle was created while using smeared properties for the PoP package stack. The outputs from the global 125 model were then used to drive the local sub-models which preserved the geometrical details of the PoP stack. Figure 5.24: Global-local submodeling: outputs from defeatured global model used to drive submodel preserving geometrical and material detail of the area of interest. PoP Stack Sub-Model Previously explicit sub-models have been used for evaluation of solder joint reliability in drop and shock[Lall 2006]. In this study detailed sub models preserving the geometrical intricacies of the PoP stack were developed. The sub-model accurately captured all the different material layers of the stack. In the first run, submodels with all interconnects modeled as Timoshenko beam elements with section and material properties, representative of the actual solder balls were developed. These models predicted highest strains in corner solder interconnects in both tiers and established these as the most vulnerable to failure. In a subsequent model developed, these solder balls were modeled in complete detail while all other interconnects in both the tiers were G l o b a l M o d e l w i t h p a c k a g e fa c i n g d o w n D e t a i l e d S u b m o d e l d ri v e n b y o u t p u t s fro m g l o b a l m o d e l 126 modeled as Timoshenko beams. Owing the as mismatch in the solder ball dimensions and pitch between interconnects in both the layers, the modeling of the PoP stack was found to be very challenging. Figure 5.25:Detailed submodel preserving the geometrical detail and various material layers of the PoP stack with the corner solder balls modeled in detail while all other I/Os are modeled as Timoshenko beam elements. P o P A s s e m b l y S c h e m a t i c 127 Figure 5.26: PoP assembly submodel- critical corner interconnect Table 5.1:PoP submodel: Elements and material properties used Model Validation The full-field filed quantity values on the surface of the PoP test vehicle from DIC based measurements were correlated with those obtained by Finite Element Method based simulations, to validate the accuracy of the model in predicting the stress and strain T o p P a c k a g e M o l d En c a p s u l a n t T o p S o l d e r Ba l l Bo tto m P a c k a g e S u b s tr a te Bo tto m S o l d e r b a l l P C BC o m p o n e n t E l e m e n t T yp e s M a t e r i a l P r o p e r t i e s De n s i t y E l a s t i c M o d u l u s P o i s s o n ' s R a t i o P C B S4 R , C 3 D8 R 1 8 1 9 . 5 6 9 4 1 . 6 0 E + 1 0 0 . 3 3 Su b s t r a t e C 3 D8 R 1400 2 . 4 1 E + 1 0 0 . 3 M o l d C o m p o u n d C 3 D8 R 1970 1 . 5 5 E + 1 0 0 . 2 5 C o r n e r So l d e r I n t e r co n n e ct s C 3 D8 R 7346 4 . 7 0 E + 1 0 0 . 3 5R e m a i n i n g So l d e r I n t e r co n n e ct s B 3 1 128 values in the PoP assembly. Figure 22 shows a comparison of z-displacement as measured experimentally and as predicted by FE simulations on the surface of the PCB. Figure 5.27: Figure showing correlation between DIC and FE simulation based strain contours on the test vehicle surface- incurred during drop event. 129 Figure 5.28: A comparison of out of plane displacements as evaluated by DIC and predicted by FE simulations T =0 .7 5 m s a f ter i m p a c t T =3 .5 m s a f ter i m p a c t T =5 .2 5 m s a f ter i m p a c t T =1 3 m s a f ter i m p a c t 130 Additionally, strain gages were mounted at critical locations on the test board. Strain histories from the strain gage were used to validate the global-model by comparing them to strain histories predicted by FE simulation at the same location. A good correlation between filed quantity (displacement) and derivatives of field quantity (strain) values from experimental observations and FE simulations was obtained which validates the accuracy of the model (Figure 22). The model was then be used drive the detailed global local submodels and predict stress/strain peaks and the locations where they occur in high magnitude or concentration. Model Predictions The validated FE model simulations were used to predict failure modes and locations. A preliminary simulation with all the interconnects modeled as Timoshenko beam elements with representative section properties showed that the corner interconnects in both the tiers of I/Os to incurred highest levels of stresses, hence these interconnects were modeled in complete detail in subsequent modeling iteration. Strain histories at the critical corner interconnects were extracted for both the tiers of solder balls. A comparison of strain histories at the four corners of the PoP stack showed a diagonal-symmetry in the strain responses. The interconnects in the top left and bottom right corner were seen to experience similar strain cycles as were the interconnects in the top right and bottom left corners of the PoP stack. 131 Figure 5.29: Strain histories extracted at the four corner interconnects in the bottom tier of solder balls. Figure 5.30: Strain histories extracted at the four corner interconnects in the top tier of solder balls. The FE simulations also showed that the strains induced were substantially higher in bottom tier of solder balls as compared to the top tier. As a result, the corner I/Os in 12 3 4 D R O P - 0 . 0 0 1 - 0 . 0 0 0 8 - 0 . 0 0 0 6 - 0 . 0 0 0 4 - 0 . 0 0 0 2 0 0 . 0 0 0 2 0 . 0 0 0 4 0 . 0 0 0 6 0 . 0 0 0 8 0 0 . 0 0 1 0 . 0 0 2 0 . 0 0 3 0 . 0 0 4 0 . 0 0 5 0 . 0 0 6 St r a in T i m e ( S e c ) S t r a i n H i s t or i e s a t Fou r C or n e r I n t e r con n e ct s ( B ot t om T i e r ) C o r ne r - 1 C o r ne r - 2 C o r ne r - 3 C o r ne r - 4 - 0 . 0 0 0 6 - 0 . 0 0 0 4 - 0 . 0 0 0 2 0 0 . 0 0 0 2 0 . 0 0 0 4 0 . 0 0 0 6 0 0 . 0 0 1 0 . 0 0 2 0 . 0 0 3 0 . 0 0 4 0 . 0 0 5 0 . 0 0 6 St r a in T i m e ( Se c) S t r a i n H i s t or i e s a t Fou r C or n e r I n t e r conn e ct s ( T op T i e r ) C or ne r - 1 C or ne r - 2 C or ne r - 3 C or ne r - 4 12 3 4 D R O P 132 the bottom tier, i.e. the interconnects between the bottom package(PSvfBGA) and the PCB were expected to fail first. Figure 5.31: A comparison of strain cycles experienced by solder interconnects in the top and bottom tiers located at corner-1 as shown in schematic Figure 5.32: A comparison of strain cycles experienced by solder interconnects in the top and bottom tiers located at the corner-2 as shown in the schematic - 0 . 0 0 1 - 0 . 0 0 0 8 - 0 . 0 0 0 6 - 0 . 0 0 0 4 - 0 . 0 0 0 2 0 0 . 0 0 0 2 0 . 0 0 0 4 0 . 0 0 0 6 0 . 0 0 0 8 0 0 . 0 0 1 0 . 0 0 2 0 . 0 0 3 0 . 0 0 4 0 . 0 0 5 0 . 0 0 6St r a in T i m e ( Se c) St r a i n Hi s t o r i e s fo r I n t e r c o n n e c t s a t Co r n e r - 1 B ot t om I nt e r co nne ct T op I nt e r conn e ct 12 3 4 D R O P - 0 . 0 0 1 - 0 . 0 0 0 8 - 0 . 0 0 0 6 - 0 . 0 0 0 4 - 0 . 0 0 0 2 0 0 . 0 0 0 2 0 . 0 0 0 4 0 . 0 0 0 6 0 . 0 0 0 8 - 0 . 0 0 1 6E - 18 0 . 0 0 1 0 . 0 0 2 0 . 0 0 3 0 . 0 0 4 0 . 0 0 5 0 . 0 0 6St r a in T i m e (Se c) St r a i n Hi s t o r i e s fo r I n t e r c o n n e c t s a t Co r n e r - 2 B ot t om I nt e r conne ct T op I nt e r conn e ct 12 3 4 D R O P 133 Figure 5.33: A comparison of maximum and minimum out of plane strains experienced by corner solder interconnects in the bottom tier. Figure 5.34: A comparison of maximum and minimum out of plane strains experienced by corner solder interconnects in the top tier. - 1000 - 800 - 600 - 400 - 200 0 200 400 600 800 C or ne r - 1 C or ne r - 2 C or ne r - 4 C or ne r - 3St r a in ( M icr o s t r a in s ) M a x a n d M i n Z - St r a i n s i n F o u r c o r n e r I n t e r c o n n e c t s - B o t t o m T i e r M a x S t r a i n M i n S t r a i n 12 3 4 D R O P - 500 - 400 - 300 - 200 - 100 0 100 200 300 400 C or ne r - 1 C or ne r - 2 C or ne r - 4 C or ne r - 3St r a in ( M ic r o s t r a in s ) M a x a n d M i n Z - St r a i n s i n F o u r c o r n e r I n t e r c o n n e c t s - T o p T i e r M a x S t r a i n M i n S t r a i n 12 3 4 D R O P 134 Additionally, in both the tiers of solder balls, the stresses were seen to be higher towards the bottom (solder ball to substrate/PCB) side. This indicates a high probability of crack development in the interface between solder ball and copper pad at the PCB side. Figure 5.35: Stress Distribution in the critical corner interconnect at different time steps during shock event. T i m e = 0.00 06 S e c T i m e = 0.00 30 S e c T i m e = 0.00 36 Se c T i m e = 0.0024 S e cT i m e = 0.0018 S e c T i m e = 0.0012 S e c 135 5.8 PoP Vs BGA: Susceptibility to failure In order to compare the reliability of PoP Stacks with BGAs of similar dimensions, a separate submodel representing a BGA with same dimensions and weight as the PoP stack was developed. The stresses in the bottom tier of interconnects from the PoP stack were compared to those induced in a regular BGA interconnects. The FE simulations of both the package types subjected to a similar shock pulse predicted that the stresses in the bottom tier of interconnects of the PoP stack were higher in magnitude as compared to those induced in the BGA interconnects. Figure 31 shows a comparison of the stress in the dropping direction, in the corner interconnects, which were modeled in complete detail in both the submodels. Figure 5.36: A comparison of stresses induced in a corner interconnect for Package-on-Package and BGA packages P o P BG A 136 In order to gauge their susceptibly to cratering/copper pad delamination, the out- of-plane stresses in the PCB side copper pads were compared between the PoP stack and a regular BGA. Figure 5.37: A comparison of out-of-plane stresses induced in corner PCB side copper pads interconnect for PoP and BGA packages The copper pads for the PoP stacks were seen to experience higher stresses than those corresponding to a regular BGA. However with the alloy composition used in this study, no cratering or copper pad delamination type of failures were observed. 5.9 Failure Analysis After failures was observed as peaks in resistance across a daisy chain in the assembly, the package stack was potted in epoxy and polished for identification of crack location, by visual inspection under a Microscope. P o P BG A 137 In accordance with the FE model predictions, a majority of the failures were seen to occur in the bottom tier of solder balls. The three daisy chains that separated the top tier of interconnect from the bottom tier, facilitated easier isolation of failure. In agreement with the FE model predictions, the bottom tier of solder interconnect were seen to fail first. Since the daisy chain connecting the top tier of solder interconnects communicated with the pads on the PCB through two pairs of I/Os in the bottom tier, these I/Os were inspected for failure to make sure the open event was not due to failures in these connecting interconnects. Figure 5.38: Figures showing failure locations as observed by cross sectioning and as predicted by FE simulations 138 Additionally, cracks observed in the corner solder balls established them as the most susceptible to failure. The observed Failure locations were compared with locations where the FE simulations predicted occurrence of high stresses. A good correlation was observed between the experimentally observed and predicted failure locations. Conclusively, in this study a JEDEC standard single component test vehicle was fabricated for evaluation of Package on Package reliability in drop/Shock environments. It was observed that selection of Solder Paste and reflow profile parameters play a very crucial role in the yield of PoP assembly process. The process flow suggested by previous researchers, wherein solder dipping for both the tiers of solder balls, is advocated has been further validates in this study. Additionally, an optimum reflow profile obtained by detailed analysis of the effects of profile parameters on induced reflow defects has been suggested. The suggested reflow profile has the following attributes: 1. Ramp-to-Spike profile, instead of Ramp-Soak-Spike profile to avoid premature flux activation 2. Gentle ramp rate to allow for timely flux activation and adequate wetting and dissolution of oxide film preventing proper solder connections and avoid thermal shock to the components. 3. A for allowing moderate peak temperature well above the liquidus of the solder used, but low enough to avoid excessive warpage 4. An optimum TAL ob about 60-90 seconds to facilitate proper reflow without defects. Additionally, validated finite element simulations were used to predict failure locations in PoP assemblies. A detailed analysis of the strains experienced by the two 139 tiers of interconnects was performed. A comparison of the strain cycles experienced by the corner interconnects in the two tiers, established that the corner interconnects in the bottom tier of I/Os was the most susceptible to failure. Cross sections of failed PoP stacked validated the FE model based predictions, in that, no failures were observed in the top tier of interconnects and the cracks observed in the bottom tier were localized to the corner interconnects. In order to compare their reliability with BGAs, , a comparison of the stresses incurred in the bottom tier of interconnects of a the PoP stack with those induced in the interconnects of a BGA of similar attributes was made. The PoP assembly processes and reflow parameters suggested in this study can be used for achieving a high-yield of PoP assembly. Furthermore, the validated FE models developed can be used to evaluate effects of design iterations in PoP assemblies eliminating the need for exhaustive testing. 140 CHAPTER 6 SUMMARY AND CONCLUSIONS The motivation behind this study is derived from the need for life prediction models for electronics in drop and shock. With packaging technology evolving at unprecedented rates, the product development cycles are shortened to meet the demands of the highly competitive market. This work, addresses the issue by demonstrating the feasibility of FE modeling for performing efficient reliability studies. Additionally, presented in this work, are life prediction models for leadfree BGA solder interconnects and PCB surface copper traces. These models, developed from validated FE model outputs and actual testing data, eliminate the need for exhaustive testing procedures. Using these models, the effects of various design iterations on the life of interconnects and copper traces in assemblies subjected to drop/shock can be analyzed without testing each design variation. The technique heavily relies on the accuracy of the Finite Element models developed to represent the assembly subjected to mechanical loading. It is therefore of outmost important that the techniques presented in this study be used with caution and a thorough understanding of the design variables. Chapter 3 presents a mathematical model which relates the life of BGA interconnects in terms of shock-events-to-failure, to the strain levels experienced during a single drop event; through a set of constants referred to as fatigue constants. The values 141 for these constants were seen to vary between solder alloy compositions and also with thermal aging of the components. The variation in the constant values between thermally aged and pristine interconnects, correlated well with the degrading effect thermal aging had on the reliability of the interconnect. It should be noted that the values of fatigue constants presented in this work are for Pristine and a single thermal aging condition only. The values governing the model cannot be used for assemblies thermally aged at different temperatures and for different durations; without substantial error in the predicted life. For pristine assemblies however, the constants developed in this work can be used with the error in predicted life well within experimental bounds. The life prediction model developed in this work, relies heavily on the accuracy of the FE model that the strain history for the interconnect is extracted from. It is therefore important to make sure that the FE simulation of the shock event is correlated with experimental results. With a validated FE model, the strain histories extracted at the critical solder interconnects can be used as an input to the life prediction model. Minor design variations can be incorporated in the FE model and the output from the simulation can again be used in conjunction with the Life Prediction model to see its effect on the life of the interconnects. The life prediction model and methodology presented in this chapter, eliminate the need for exhaustive testing of each design variation and equip the component designer with a powerful tool to evaluate components designs in terms of reliability. While this study presents life prediction models for sequential thermal aging and drop, future studies could address assemblies which undergo simultaneous thermally aging and shock induced mechanical loading. 142 The study on cupper traces aimed at evaluating the reliability of PCB surface copper traces. The study set forth a set of empirical guidelines which summarize the effects of trace geometries and orientation on their drop/shock reliability. The design guidelines provide a PCB designer with a set of rules that he can follow to ensure highly reliable copper traces-without a detailed knowledge of their failure mechanisms. Additionally, the study presents a mathematical model, which can be used to evaluate the fatigue life of copper traces on the PCB surface. It would be interesting to perform similar studies on copper traces embedded between different layers of a composite PCB. While the fatigue model presented in this study is valid for low cycle fatigue, a similar model which covers high cycle fatigue can also be developed. In conjunction, the two models could be used for life prediction of cupper traces in subjected to high g-level shock pulses, be decomposing the strains experienced by the traces into constituent cycles of constant algorithm, as was done for the BGA interconnects in Chapter 3. While the study addresses the need for empirical design guidelines for ensuring Cu-trace reliability, a life prediction model for the same would also aid the cupper trace design process. The study on novel Package-on-Package architecture addresses the issues involved with manufacturing these components. IN the study, a JEDEC standard single component test vehicle was fabricated for evaluation of Package on Package reliability in drop/Shock environments. The effects of Solder Paste attributes and reflow profile parameters on the yield of PoP assembly process were studies. Additionally, an optimum reflow profile obtained by detailed analysis of the effects of profile parameters on induced reflow defects was suggested. The study further used , validated finite element 143 simulations to predict failure locations in PoP assemblies by performing a detailed analysis of the strains experienced by the two tiers of interconnects. A comparison of the strain cycles experienced by the corner interconnects in the two tiers, established that the corner interconnects in the bottom tier of I/Os was the most susceptible to failure. Cross sections of failed PoP stacked validated the FE model based predictions, in that, no failures were observed in the top tier of interconnects and the cracks observed in the bottom tier were localized to the corner interconnects. In conclusion, a comparison of the stresses incurred in the bottom tier of interconnects of a PoP stack with those induced in the interconnects of a BGA of similar attributes was made with an intention of comparing the two in terms of their drop/shock reliability. In conclusion, this study presents life prediction models, suggests optimum manufacturing processes and design guidelines which aid in design and fabrication of reliable electronic components, within the constraints of short product development cycles. 144 BIBLIOGRAPHY Abaqus 6.9-1 Documentation 2009 Abdulhamid, M., Basaran, C., ?Influence of thermomigration on lead-free solder joint mechanical properties?, Journal of Electronic Packaging 131:011002, 2009. Abdul-Baqi, A., Schreurs, P. J. G., Geers, M. G. D., ?Fatigue damage modeling in solder interconnects using a cohesive zone approach?, International Journal of Solids and Structures, Volume 42, pp. 927-942, 2005. Amagai, M., Suzuki, Y., A study of package warpage for package on package (PoP), Electronic Components and Technology Conference (ECTC), 2010 Proceedings 60th , vol., no., pp.226-233, 1-4 June 2010 American Society of Mechanical Engineers (ASME), ?Standard Practices for Cycle Counting in Fatigue Analysis?, ASTM E1049-85, 1997 Amodio, D., Broggiato, G., Campana, F., Newaz, G., Digital Speckle Correlation for Strain Measurement by Image Analysis, Experimental Mechanics, Vol. 43, No.4, pp. 396-402, 2003. Bannantine, J. A., Comer, J.J., Handrock, J.L., ?Fundamentals of Metal Fatigue Analysis?, Prentice Hall, 1990. Bhate, D., Chan, D., Subbarayan, L., Nguyen, L., Fatigue Crack Growth and Life Descriptions of Sn3.8Ag0.7Cu Solder Joints: A Computational and Experimental Study, Proceedings of the 57th Electronic Components and Technology Conference, pp. 558- 565, 2007. Bieler, T., Jiang, H., Influence of Sn Grain Size and Orientation on the Thermomechanical Response and Reliability of Pb-free Solder Joints, Proceedings of the 56th ECTC, pp. 1462- 1471, May 2006. Chai, T.C., Quek, S., Hnin, W.Y., Wong, E.H., Chia, J., Wang, Y.Y., Tan, Y.M., Lim, C.T., Board Level Drop Test Reliability of IC Packages, Electronic Components and Technology Conference, pp. 630-636, June, 2005 145 Che, F.X., Luan, J.E., Baraton, X., Effect of silver Content and Nickel Dopant on Mechanical Properties of Sn-Ag-based Solders, Proceedings of the 58th ECTC, Orlando, Florida, pp.485-490, May27th-May30th, 2008. Chen, K. M., Ho1, K. K., Jiang, D. S., Impact Of Lead Free Solder Materials On Board Level Reliability For Low-K Wlcsp, International Microsystems, Packaging, Assembly and Circuits Technology, pp. 243-245, 2007. Choi, J.Y., Kang, I.S., Park, B.J., Park, Y. M., Jung, G.J., Study on Inter-Metallic Compound (IMC) at Interface of Pb-Free Solder and Nickel UBM for Mobile Application, Proceedings of the 57th ECTC, Reno, Nevada, pp.1802-1808, May29th- June 1st, 2007. Chong, D.Y.R., Toh, H.J., Lim, B.K., Low, P. T.H., Drop Reliability Performance Assessment for PCB Assemblies of Chip Scale Packages (CSP), Electronic Packaging Technology Conference, pp. 262-269, June, 2005. Cook, R.D., ?Concepts and Applications of Finite Element Analysis? John Wiley & Sons, NY, 1981. Cote, K.J.; Dadkhah, M.S., Whole field displacement measurement technique using speckle interferometry, Electronic, Proceedings of the 51st Electronic Component & Technology Conference, pp.: 80-84, 29 May-1 June 2001. Davis, C., Freeman, D., Statistics of subpixel registration algorithms based on spatiotemporal gradients or block matching, Optical Engineering, Vol. 37, pp. 1290- 1298, 1998. Downing, S. D., and Socie, D. F., ?Simple Rainflow Counting Algorithms, International Journal of Fatigue?, Vol. 4, No. 1, pp. 31-40, 1982. Dreiza, M., et al., Stacked Package-on-Package Design Guidelines , International Wafer Level Packaging conference, 2005. Dreiza, M., L. Smith, G. Dunn, N. Vijayaragavan, J. Werner, Package on Package (PoP) Stacking and Board Level Reliability: Results of Joint Industry Study, Proceedings of IMAPS 2006. Eslampour, H., Lee, S., Park, S., Lee, T., Yoon, I., Kim, Y., Comparison of Advanced PoP Package Configurations, 60th Electronic Components and Technology Conference (ECTC), Las Vegas, Nevada, pp. 1946 ? 1950, June 1-4, 2010 European Directive 2002/95/EC, Restriction on the use of hazardous substances in electrical and electronical equipment. 146 Gu, J., Cooreman, S., et al., Full-Field Optical Measurement For Material Parameter Identification With Inverse Methods, WIT Transactions on The Built Environment, Vol. 85, 2006. Gu, J., Lim, C., Tay, A., Equivalent solder joint and equivalent layer models for the simulation of solder column failure under drop impact, Proceedings of 6th Electronics Packaging Technology Conference, pp. 547?552, 2004. Gu, J., Lim, C., Tay, A., Simulation of mechanical response of solder joints under drop impact using equivalent layer models,? in Proceeding of 55th Electronic Component Technology Conference, Orlando Florida, pp. 522?529, May31 -June 3, 2005. Helm, J., McNeill, S., Sutton, M., Improved Three- Dimensional Image Correlation For Surface Displacement Measurement, Society of Photo-Optical Instrumentation Engineers, Vol. 35, No. 7, pp. 1911-1920, July 1996. Ho, H.M., Oh, K.E., Wong, E.H., Lim, T.B., Application of moire interferometry in electronics packaging, Proceedings of the Electronic Packaging Technology Conference, pp. 277 ? 282, 1997. Huang, B., Hwang, H., Lee, N., A Compliant and Creep resistant SAC-Al(Ni) Alloy, Proceedings of the 57th ECTC, Reno, Nevada, pp. 184- 191, May 29-June 1, 2007 Irving, S., Liu, Y., Free Drop Test Simulation for Portable IC Package by Implicit Transient Dynamics FEM, 54th Electronic Components and Technology Conference, pp. 1062 ? 1066, 2004. Iyer, G., Ouyang, E., Kittidacha, W., Tantideeravit, S., Suresh, L.K., Pb-free Solder: SAC105 vs SAC305 Drop-Test Reliability Data Comparison, International Electronics Manufacturing Technology Symposium, pp. 251-255, 2007. Jang. J., De Silva, A.P.,Drye, J.E., Post, S. L., Owens, N.L., Lin, J., Frear, D.R., Failure Morphology After Drop Impact Test of Ball Grid Array (BGA) Package With Lead-Free Sn?3.8Ag?0.7Cu and Eutectic SnPb Solders, IEEE Transactions On Electronics Packaging Manufacturing, Volume:30, Issue 1, pp. 49 ? 53, January, 2007. JEDEC Standard JESD22-B111, Board Level Drop Test Method of Components for Handheld Electronic Products, 2003. Jie, G., Lim, C.T., Tay, A.A.O., Equivalent Solder Joint and Equivalent Layer Models for the Simulation of Solder Column Failure under Drop Impact, Electronic Packaging Technology Conference, pp. 547-552, June, 2004. Johnson, M., E., Lu, H., Y., Lawhead, D., Tessier, T., Scott, D., Curtis, A., Using High 147 Speed Shear and Cold Ball Pull to Characterize Lead Free Solder Alloys and Predict Board Level Drop Test Performance, Electronic Components and Technology Conference, pp. 533-542, 2007. Kawashiro, F., Ujiie, M., Shibuya, K., Kurita, Y., Soejima, K., Kawano, M., Reliability Studies of Sn-Ag-Cu BGA Solder Joints on Ni/Cu/Au Surface Finish for SMAFTI Packaging Technology, Proceedings of the 58th ECTC, Orlando, Florida, pp.283 -289, May27th-May30th, 2008. Kehoe, L., Lynch, P., Gu?nebaut, V., Measurement of Deformation and Strain in First Level C4 Interconnect and Stacked Die using Optical Digital Image Correlation, 56th ECTC, pp. 1874-1881, May 2006. Kim, H., Zhang, M., Kumar, C., Liu, p., kim, D., Xie, m., Wang, Z., Improved Drop Reliability Performance with Lead Free Solders of Low Ag Content and Their FailureModes, Proceedings of 57th ECTC, Reno, Nevada, pp 962- 967, May 29-June 1, 2007. Kim, P., Kim, B., Ahn, E., Chung, T., Improvement of Drop Reliability in OSP/Cu Pad Finished Packages, Electronic Packaging Technology Conference, pp. 168-173, December, 2006. Kirkpatrick, S., Schroeder, M., Simons, J., Evaluation of Passenger Rail Vehicle Crashworthiness, International Journal of Crashworthiness, Vol. 6, No. 1, pp. 95-106, 2001. Lai, Y-S., Chang, H-C., Chiu, Y-T., Yeh, C-L., Song, J-M., Ball Impact Responses of Sn- Ag-Cu Solder Joints Doped with Ni or Ge, International Microsystems, Packaging, Assembly and Circuits Technology, pp. 197 ? 201, October, 2007. Lall, P., Choudhary, P., Gupte, S., Suhling, J., Health Monitoring for Damage Initiation and Progression during Mechanical Shock in Electronic Assemblies, IEEE Transactions on Components and Packaging Technologies, Vol. 31, No. 1, pp. 173-183, March 2008. Lall, P., Choudhary, P., Gupte, S., Suhling, J., Hofmeister, J., Statistical Pattern Recognition and Built-In Reliability Test for Feature Extraction and Health Monitoring of Electronics under Shock Loads, 57th Electronics Components and Technology Conference, Reno, Nevada, pp. 1161-1178, May 30-June 1, 2007. Lall, P., Gupte, S., Choudhary, P., Suhling, J., Solder-Joint Reliability in Electronics Under Shock and Vibration using Explicit Finite Element Sub-modeling, IEEE Transactions on Electronic Packaging Manufacturing, Volume 30, No. 1, pp. 74-83, January 2007. 148 Lall, P., Gupte, S., Choudhary, P., Suhling, J., Darveaux, R., Cohesive-Zone Explicit Sub-Modeling for Shock Life-Prediction in Electronics, 57th ECTC, pp. 515-527, Reno, NV, May 29- Jun 1, 2007. Lall, P., Gupte, S., Choudhary, P., Suhling, J., Solder-Joint Reliability in Electronics Under Shock and Vibration using Explicit Finite-Element Sub-modeling, 56th ECTC, pp. 428 ? 435, 2006. Lall, P., Hande, M., Bhat, C., Islam, N., Suhling, J., Lee, J., Feature Extraction and Damage-Precursors for Prognostication of Lead-Free Electronics, Microelectronics Reliability, Vol. 47, pp. 1907?1920, December 2007. Lall, P., Iyengar, D., Shantaram, S., Pandher, R., Panchagade, D., Suhling, J., Design Envelopes and Optical Feature Extraction Techniques for Survivability of SnAg Leadfree Packaging Architectures under Shock and Vibration, 58th ECTC, Orlando, FL, pp. 1036- 1047, May 27-30, 2008. Lall, P., Panchagade, D., Choudhary, P., Suhling, J., Gupte, S., Failure Envelope Approach to Modeling Shock and Vibration Survivability of Electronic and MEMS Packaging, 55th ECTC, pp. 480-490, 2005. Lall, P., Panchagade, D., Iyengar, D., Shantaram, S., Suhling, J., Schrier, H., High Speed Digital Image Correlation for Transient-Shock Reliability of Electronics, 57th ECTC, Reno, Nevada, pp. 924-939, May 29 ? June 1, 2007. Lall, P., Panchagade, D., Liu, Y., Johnson, R. W., and Suhling, J. C., Smeared-Property Models for Shock-Impact Reliability of Area-Array Packages, Journal of Electronic Packaging, Vol. 129(4), pp. 373-381, 2007. Lall, P., Panchagade, D., Liu, Y., Johnson, W., Suhling, J., Models for Reliability Prediction of Fine-Pitch BGAs and CSPs in Shock and Drop-Impact, 54th ECTC, pp. 1296-1303, 2004. Lall, P., Shantaram, S., Angral, A., Kulkarni, M., Explicit Submodeling and Digital Image Correlation Based Life-Prediction of Leadfree Electronics under Shock-Impact, 59th ECTC, San Diego, CA, pp. 542-555, May 25-29, 2009. Lee, Joon-Yeob, Hwang, Tae-Kyung, Kim, Jin-Young, Min Yoo, Sohn, Eun-Sook, Chung, Ji-Young, Dreiza, M., Study on the Board Level Reliability Test of Package on Package (PoP) with 2nd Level Underfill, Electronic Components and Technology Conference,. ECTC '07. Proceedings. 57th, vol., no., pp.1905-1910, 29, May- 1, Jun. 2007. Lee W.W., Nguyen L.T., Selvaduray G.S. ?Solder joint fatigue models: review and applicability to chip scale packages? Microelectronics Reliability,2000,Vol. 40, PP 231- 244 149 Lim, C.T. and Low, Y.J., Investigating the Drop Impact of Portable Electronic Products, 52nd Electronic Components and Technology Conference, pp. 1270-1274, 2002. Lin, W., Yoshida, A., Dreiza, M., Control of the Warpage For Package-on-Package (PoP) Design, SMTA International Conference, pp.320-326, 2006. Liu, S., Wang, X., Ma, B., Gan, Z., Zhang, H., Drop Test and Simulation of Portable Electronic Devices, International Conference on Electronic Packaging Technology, pp. 701-704, June, 2005. Lu, H., Applications of digital speckle correlation to microscopic strain measurement and materials' property characterization, Journal of Electronic Packaging, Volume 120, Issue 3, 275-279, September 1998. Lu, H., Yeh, C., Wyatt, K., Experimental evaluation of solder joint thermal strain in a CSP using digital speckle correlation, 1998 InterSociety Conference on Thermal Phenomena, pp. 241?245, 1998. Luan, J.E., Tee, T.Y., Goh, K.Y., Ng, H.S., Baraton,X., Bronner, R., Sorrieul, M., Hussa, E., Reinikainen,T., and Kujala, A., ?Drop Impact Life PredictionModel for Lead-free BGA Packages and Modules?,EuroSime Conference Proc., Germany, 2005, pp. 559-565 Masicat, J., Kumar, C., Nuda, M., An Efficient Certification Approach For New Sn-Ag- Cu Solder Alloy, Annual International Reliability 564 Physics Symposium, pp. 564-565, Phoenix, 2007. Mattila T .T., Kivilahti J.K.,? Failure mechanisms of lead-free chip scale package interconnections under fast mechanical loading?, Journal of Electronic Materials,2005,Vol. 34, pp 969-976 Mattila, T. T., Kaloinen, E., Syed, A., Kivilahti, J. K., Reliability of SnAgCu Interconnections with Minor Additions of Ni or Bi under Mechanical Shock Loading at Different Temperatures, Electronic Components and Technology Conference, pp. 331- 390, 2007. Miller, T., Schreier, H., Reu, P, High-speed DIC Data Analysis from a Shaking Camera System, Proceedings of the SEM Conference, Springfield, MA, June 4-6, 2007. Moore, G.E., et al., Cramming more components onto integrated circuits," Proceedings of the IEEE, vol. 86, no. 1, pp. 82-85, 1998. Ong,Y.C., Shim,V.P.W., Chai, T.C., LimC.T., Comparison of Mechanical Response of PCBs Subjected to Product-Level and Board-Level Drop Impact Tests, Electronic Packaging Technology Conference, pp. 223-227, June, 2003. 150 Ou, S., Xu, Y., Tu, K. N., Micro-Impact Test on Lead-Free BGA Balls on Au/Electrolytic Ni/Cu Bond Pad, Electronic Components and Technology Conference, pp. 467-471, June, 2005. Ouimet, S., Casey, Jon., Marston, K., Muncy, J., Corbin, J., Jadhav, V., Wassick, T., Depatie, I., Development of a 50mm Dual Flip Chip Plastic Land Grid Array Package for Server Applications, Proceedings of the 57th IEEE Electronic Components and Technology Conference, Reno, Nevada, pp.1900-1906, May 29-June 1, 2008. Pandher, R.S., Lewis, B.G., Vangaveti, R., Singh, B., Drop Shock Reliability of Lead- Free Alloys - Effect of Micro-Additives, Electronic Components and Technology Conference, pp. 669-676, March, 2007. Pandher, R., Healey, R., Reliability of Pb-Free Solder Alloys in Demanding BGA and CSP Applications, Electronic Systems, Proceedings of the 58th Electronic Component & Technology Conference, Orlando, Florida, pp. 2018-2023, May27th-May30th, 2008. Park, S., Al-Yafawi, A., Yu, D., Kwak, J., Lee, J., Goo, N., Influence of Fastening Methods on the Dynamic Response and Reliability Assessment of PCBS in Cellular Phones Under Free Drop, Proceedings of the ITherm, Intersociety Conference on Thermal and Thermo-mechanical Phenomena, Orlando, Florida, pp.876-882, May 28- May31, 2008 Park, S., Shah, C., Kwak, J., Jang, C., Pitarresi, J., Transient Dynamic Simulation and Full-Field Test Validation for A Slim-PCB of Mobile Phone under Drop / Impact, Proceedings of the 57th Electronic Component & Technology Conference, Reno, Nevada, pp. 914-923, May 29 ? June 1, 2007c. Peterson, D., Cheng, C., Karulkar, P.C., Characterization of Drop Impact Survivability of a 3D CSP Stack Module, Proceedings of the 58th Electronic Component & Technology Conference, Orlando, Florida, pp. 1648-1653, May27th-May30th, 2008. Pitaressi, J., Roggeman, B., Chaparala, S., Mechanical Shock Testing and Modeling of PC Motherboards, Electronics Components and Technology Conference, Las Vegas, Nevada, pp. 1047 ? 1054, June 1-4, 2004. Rajendra, Pendse, D., Zhou, P., Methodology For Predicting Solder Joint Reliability In Semiconductor Packages, Microelectronics Reliability, Vol. 42, pp. 301-305, 2002. Ren, W., Wang, J., Reinikainen, T., Application of ABAQUS/Explicit submodeling technique in drop simulation of system assembly, Proceeding of Electronic Packaging Technology Conference, pp. 541?546, 2004. Ren, W., Wang, J., Shell-based simpli?ed electronic package model development and its application for reliability analysis, Proceeding of Electronic Packaging Technology Conference, pp. 217?222, 2003. 151 Reu, P., Miller, T., High-speed Multi-camera DIC for Finite Element Model Validation, Part 1, SEM Annual Conference and Exposition on Experimental and Applied Mechanics, June 4-7, 2006. Scheijgrond,P.L.W., Shi, D.X.Q., Driel, W.D.V., Zhang,G.Q., Nijmeijer, H., Digital Image Correlation for Analyzing Poratable Electronic Products during Drop Impact Tests, 6th International conference on Electronic Packaging Technology, 2005. Seah, S.K.W., Lim, C.T.,Wong, E.H., Tan, V.B.C., Shim, V.P.W., Mechanical Response of PCBs in Portable Electronic Products During Drop Test, Electronic Packaging Technology Conference, pp. 120-125, May, 2002. Seah, S.K.W., Wong, E.H., Mai, Y.W., Rajoo, R., Lim, C.T., High-Speed Bend Test Method and Failure Prediction for Drop Impact Reliability, Electronic Components and Technology Conference, pp. 1003-1008, May,2006. Shah, K., Mello, M., Ball Grid Array Solder Joint Failure Envelope Development for Dynamic Loading, Electronic Components and technology Conference, Las Vegas, Nevada, pp. 1067-1074, June 1-4, 2004. Shi, X., Pang, H.L.J., Zhang, X.R., Liu, Q.J., Ying, M., In-situ Micro-Digital Image Speckle Correlation Technique for Characterization of Materials Properties and Verification of Numerical Models, IEEE Transaction on Component and Packaging Technologies ,pp. 659-667, 2004. Sjoberg, J.; Geiger, D.A., Shangguan, D., Process development and reliability evaluation for inline Package-on-Package (pop) assembly, Electronic Components and Technology Conference. ECTC 2008. 58th, vol., no., pp.2005-2010, 27-30 May 2008. Song, F., Lo, C.C.J., Lam, J., Jiang, T., Lee, S.W., A Comprehensive Paralell Study on the Board Level Reliability of SAC, SACX and SACN Solders, Proceedings of the 58th ECTC, Orlando, Florida, pp.146-154, May27th-May30th, 2008. Srinivasan, V., Radhakrishnan, S., Zhang, X., Subbarayan, G., Baughn, T., Nguyen, L., High Resolution Characterization of Materials Used In Packages Through Digital Image Correlation, InterPACK Conference Proceedings, July 17-22, 2005. Sun, Y., Pang, J., Shi, X., Tew, J., Thermal Deformation Measurement by Digital Image Correlation Method, Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronics Systems, pp. 921-927, May 2006. Sweatman, K., Suenaga, S., Nishimura, T., Strength of Lead-free BGA Sphere in High Speed Loading, Surface Mount Technology Association Pan Pacific, Kauai, Hawaii, pp.127-133, Jan22-Jan24, 2008 152 Syed, A., Kim, S. M., Lin, W., Khim, J.Y., Song, E.K., Shin, J.H., Panczak, T., An Methodology for Drop Performance Prediction and Application for Design Optimization of Chip Scale Packages, Electronic Components and technology Conference, Orlando, Florida, pp. 472-479, May 31-June 3, 2005. Syed, A., Kim, T.S., Cha, S.W., Scanlon, J., Ryu, C.G., Effect of Pb free Alloy Composition on Drop/Impact Reliability of 0.4, 0.5 & 0.8mm Pitch Chip Scale Packages with NiAu Pad Finish, Electronic Components and Technology Conference, pp. 951-956, 2007. Syed, A., Kim, T.S., Cho, Y.M., Kim, C.W., Yoo, M., Alloying Effect of Ni, Co, and Sb in SAC solder for Improved Drop Performance of Chip Scale Packages with Cu OSP Pad Finish, Electronic Components and Technology Conference, pp. 404-411, December, 2006. Takagi,Y., T.Notomi, and A. Fujita, Charactreristics of New Solder Paste for 3-D Package, MAP&RTS (2005) Tan, V. B. C., Tong, M. X., Lim, K.M., Lim, C.T., Finite Element Modeling of Electronic Packages Subjected to Drop Impact, IEEE Transaction on Component and Packaging Technology, Volume:28, Issue 3, pp. 555-560, September, 2005. Tee, T. Y. , Hun Shen Ng, Chwee Teck Lim, Eric Pek , Zhaowei Zhong , Board Level Drop Test and Simulation of TFBGA Packages for Telecommunication Applications, Proceedings of the 53rd ECTC, pp. 121-129, 2003 Tee, T. Y., Luan, J., Pek, E., Lim, C. T., Zhing, Z., Advanced Experimental and Simulation Techniques for Analysis of Dynamic Responses during Drop Impact, Electronic Components and Technology Conference, Las Vegas, Nevada, pp. 1088-1094, June 1-4, 2004. Tee, T. Y., Ng, H-S., Lim, C. T., Pek, E., Zhong, Z., Board Level Drop Test and Simulation of TFBGA Packages for Telecommunication Applications, 53rd Electronic Components and Technology Conference, pp. 121-129, 2003. Tee, T.Y., Ng, H.S., Syed, A., Anderson, R., Khoo, C.P., ogers, B.,? Design for Board Trace Reliability of WLCSP under Drop Test?, EuroSimE Conference Proc., Netherlands,2009 Tee, T.Y., Tan, L.B., Anderson, R., Ng, H.S., Low, J. H., Khoo, C.P., Moody, R., and Rogers, B., ?Advanced Analysis of WLCSP Copper Interconnect Reliability under Board Level Drop Test,? 10th EPTC Conference Proc., Singapore, 2008, pp. 1086-1095. Tiwari, V., Williams, S., Sutton, M., McNeill, S., Application of Digital Image Correlation in Impact Testing, Proceedings of the 2005 SEM Annual Conference and Exposition on Experimental and Applied Mechanics, June 7-9, 2005. 153 Towashiraporn, P., Crosbie, P., Brown, M., Cavallaro, A., Dynamic Crack Propagation Prediction of Solder Joints under Impact Conditions, Electronic Components and Technology Conference, pp. 689-694, 2007 Towashiraporn, P., Subbarayan, G., Desai, C. S., A hybrid model for computationally efficient fatigue fracture simulations at microelectronic assembly interfaces, International Journal of Solids and Structures, Volume 42, pp. 4468-4483, 2005 Towashiraporn, P., Xie, C., Cohesive Modeling of Solder Interconnect Failure in Board Level Drop Test,Proceedings of ITHERM, pp. 817-825, 2006 Tsai, K.T., Liu, F.L., Wong, E.H., Rajoo, R., High Strain Rate Testing of Solder Interconnections, Conference on High Density Microsystem Design and Packaging and Component Failure Analysis, pp. 1-6, June, 2005. Van Der Sluis, O., Timmermans,P.,H.,M., Van Silfhout, R., B., R., Van Driel., W., D., Zhang, G., Q., Delamination modeling of three-dimensional microelectronic system, Proceedings of the 58th Electronic Component & Technology Conference, Orlando, Florida, pp. 1593-1600, May27th-May30th, 2008. Vijayaragavan, N., Carson, F., Mistri, A.,Package on Package Warpage ? Impact on Surface Mount Yields and Board Level Reliability, Proc of 58th IEEE ECTC, Lake Buena Vista, FL, May 2008, pp 389-396 Vogel, D., Gollhardt, A., Sabate, N., Keller, J., Micheal, B., Reichl, H., Localized Stress Measurements ? A New Approach Covering Needs for Advanced Micro and Nanoscale System Development, , Proceedings of the 57th Electronic Component & Technology Conference, Reno, Nevada, pp.1490 -1497, May 29 ? June 1, 2007 Wang, V., Maslyk, D., Analysis of the Reliability of Package-on-Package Devices Manufactured Using Various Underfill Methods, International Conference on Electronic Packaging Technology & High Density Packaging (ICEPT-HDP), Shanghai, China, pp. 1 ? 3, July 28-31, 2008. Wang, Y.Y., Wang, F., Chai, T.C., Finite Element Modeling of CSP Package Subjected to Board Level Drop Test, Electronic Packaging Technology Conference, pp. 684-688, June, 2004. Wen, T., Ku, S., Efficient Evaluation of Substrate Warpage by Finite Element Method and Factorial Design Analysis, Proceedings of the 57th Electronic Component & Technology Conference, Reno, Nevada, pp. 1754-1759, May 29 ? June 1, 2007. Woodgate Ralph. ?Solder Joint Inspection Criteria?, Printed Circuit Assembly, February 1987, pp, 6-10. Wong, E. H., Lim, C. T., Field, J. E., Tan, V. B. C., Shim, V. P. M., Lim, K. T., Seah, S. 154 K. W., Tackling the Drop Impact Reliability of Electronic Packaging, ASME International Electronic Packaging Technical Conference and Exhibition, Maui, pp. 1 ? 9, July 2003. Wong, E.H., Lim, K.M., Lee, N., Seah, S., Hoe, C., Wang, J., Drop Impact Test - Mechanics and Physics of Failure, Electronic Packaging Technology Conference, pp. 327-333, May, 2002. Wu, J., Song, G., Yeh, C-P., Wyatt, K., Drop/Impact Simulation and Test Validation of Telecommunication Products, Intersociety Conference on Thermal Phenomena, pp. 330- 336, 1998. Wu J., Global and Local Coupling Analysis for Small Components in Drop Simulation, 6th International LSDYNA Users Conference, pp. 11:17 - 11:26, 2000. Xie, D., Arra, M., Shangkai, D., Phan, H., Geiger, D., Yi, S., Life Prediction of Lead free Solder Joints for Handheld Products, Telecom Hardware Solutions Conference, Plano, Texas, May 15-16, 2002. Xie, D., Arra, M., Yi, S., Rooney, D., Solder Joint Behavior of Area Array Packages in Board-Level Drop for Handheld Devices, 53rd Electronic Components and Technology Conference, pp. 130 ? 135, 2003. Xu, L., Pang, H., Combined Thermal and Electromigration Exposure Effect on SnAgCu BGA Solder Joint Reliability, Proceedings of the 56th ECTC, pp. 1154-1159, May 2006. Yeh. C., Lai, Y., Chang, H., Chen, T., Correlation Between Package-level Ball Impact Test and Board-level Drop Test, Electronic Components and Technology Conference, pp. 270-275, June, 2005. Yim, M.J., Strode, R., Adimula, R., Yoo, C., Effects of Material Properties on PoP Top Package Warpage Behaviors, 60th Electronic Components and Technology Conference (ECTC), Las Vegas, Nevada, pp. 1071 - 1076, June 1-4, 2010. Yogel, D., Grosser, V., Schubert, A., Michel, B., MicroDAC Strain Measurement for Electronics Packaging Structures, Optics and Lasers in Engineering, Vol. 36, pp. 195- 211, 2001. Yoshida, A., Taniguchi, J., Murata, K., Kada, M., Yamamoto, Y., Takagi, Y., Notomi, T., Fujita, A., A Study on Package Stacking Process for Package-on-Package (PoP), 56th Electronic Components and Technology Conference (ECTC) , San Diego, California, pp. 6, May 30 - Jun 02, 2006. Zhang, F., Li, M., Xiong, C., Fang, F., Yi, S., Thermal Deformation Analysis of BGA Package by Digital Image Correlation Technique, Microelectronics International, Vol. 22, No. 1, pp. 34-42, 2005. 155 Zhang, Y., Shi, X., Zhou, W., Effect of Hygrothermal Aging on Interfacial Reliability of Flip Chip on Board (FCOB) Assembly, Electronic Packaging Technology Conference, pp. 404-409, 2004. Zhao, J., Liu, F., Zhou, X., Zhou, H., Jing, J., Zhao, M., Improvement of JEDEC Drop Test in SJR Qualification through Alternative Test Board Design, Electronic Components and Technology Conference, pp. 946-950, 2007. Zhao, J, Luo, Y., Huang, Z., Ma, R., Effects of package design on top PoP package warpage, Electronic Components and Technology Conference, ECTC 2008. 58th, vol., no., pp.1082-1088, 27-30 May 2008. Zhang, Y., Zijie, C., Suhling, J., Lall, P., Bozack, M.J., The Effects of Aging Temperature on SAC Solder Joint Material Behavior and Reliability, Proceedings of the 58th Electronic Component & Technology Conference, Orlando, Florida, pp.99 -112, May27th-May30th, 2008 Zhou, P., Goodson, K. E., Sub-pixel Displacement and Deformation Gradient Measurement Using Digital Image- Speckle Correlation (DISC), Optical Engineering, Vol. 40, No. 8, pp 1613-1620, August 2001. Zhu, F., Two-dimensional Recursive Fast Fourier Transform Image Processing Applications, IEE Proceedings Radar and Signal Processing, Vol. 137, Issue 4, pp:262 266, 1990. Zhu, L., Marccinkiewicz, W., Drop Impact Reliability Analysis of CSP Packages at Board and Product System Levels Through Modeling Approaches, Inter Society Conference on Thermal and Thermomechanical Phenomena, pp. 296 ? 303, 2004. Zhu, L., Modeling Technique for Reliability Assessment of Portable Electronic Product Subjected to Drop Impact Loads, Proceedings of the 53rd ECTC, pp. 100-104, 2003. Zhu, L., Sub-modeling Technique for BGA Reliability Analysis of CSP Packaging Subjected to an Impact Loading, InterPACK Conference Proceedings, 2001. Zhu, W. H., Xu, L., Pang, J., Zhang, X., Poh, E., Sun, Y., Sun, A., Wang, C. K., Tan, H.B., Drop Reliability Study of PBGA Assemblies with SAC305, SAC105 and SAC105- Ni Solder Ball on Cu-OSP and ENIG Surface Finish, Proceedings of the 58th ECTC, Orlando, Florida, pp.1667-1672, May27th-May30th, 2008.