Esa Hussa

Absolute and relative fatigue life prediction methodology for virtual qualification and design enhancement of lead-free BGA

The semiconductor industry is driving toward lead-free solder due to environmental concern and legislation requirement. The industry has also concluded that SnAgCu solder alloy so far is the best lead-free alternative to SnPb solder. Therefore, most existing and new packages have to be tested and qualified using lead-free solder. One of the critical concerns is board level solder joint reliability during thermal cycling test. In this paper, the methodology for an absolute life prediction is described for virtual qualification of packages. A good absolute fatigue life prediction requires an appropriate solder creep model and actual test data on packages. Two new sets of lead-free Anands constants for SnAgCu solder are introduced for creep models. These Anands creep models are compared with other lead-free and eutectic solder model and the relative design trend is similar. A fatigue corrective factor is introduced to integrate the different solder models together for convenient relative design enhancement with acceptable range of absolute life prediction. These fatigue corrective factors can also be used to compare different finite element modeling assumptions such as element size and solution time step. Subsequently, design analysis is performed to study the effects of 11 key package dimensions and material properties. It is found that the relative design trend for packages with lead-free and eutectic solder is similar. Therefore, the design guidelines established for the previous eutectic solder is still valid for lead-free solder.

Evaluation of the drop response of handheld electronic products

The aim of the product level drop response evaluation presented in this paper is to provide goals and guidelines for the development of a board-level drop test methodology that would better reproduce the field use loading conditions of modern portable electronic devices. Eight commercially available smart phones from different manufacturers were evaluated for their free-fall drop response. For this purpose, miniature accelerometer and strain gauges were attached to various locations on the component board inside the product covers. The maximum strain, average rate to maximum strain, frequency of the effective mode shapes, and maximum deceleration were determined from the measured strain and deceleration histories. The determined values showed significant variation from drop to another and device to another, but it was noteworthy that the extreme magnitude of the strain, average rate to maximum strain, and deceleration can be very high: values as high as 10,000 μ (“micro-strain” = [10−6 m/m]), 26 s−1, and 10 kG were measured, respectively. Post analyses of the strain histories revealed that the shock impact response of the devices can be conceptually divided into two consecutive periods: (i) forced high amplitude bending/twisting of the component board at the moment of impact, and (ii) subsequent lower amplitude (resonance) vibration of the component board while the device bounces back from the site of impact. Maximum train values reached during Period (i) were typically much higher than the typical strain peaks during Period (ii). However, during Period (ii) sharp strain peaks were often identified whose maximum value occasionally went well above the maximum value during Period (i). Furthermore, any resonance vibrations initiated by the impact forces were dampened efficiently in all device models. In order to form a better understanding of what is causing the very high strains, the drop response of one of the devices was simulated by employing the Finite Element Method (FEM). The FEM results showed that the regions of high strains are highly localized. During Period (i) they are caused by the forced bending of the board by the surrounding mechanical structures, and during Period (ii) by internal collisions between the vibrating component board and the surrounding mechanical structures. On the basis of the characterization of the commercial portable devices, the following goals were set for the development of a board-level drop test methodology: a test board that simulates the response of portable electronic products to a free-fall drop impact should be able to produce: (1) board strain well above 3500 μ and (2) average strain rate as close as possible to 7 s−1. The experimental characterization of the mobile devices was carried out by Aalto University, while the device-level drop impact simulations were performed by Nokia.

Drop impact life prediction models with solder joint failure modes and mechanisms

Drop impact performance of solder joints of IC packages becomes a great concern for handheld products, such as mobile phones and PDA. Failure modes of solder joints under drop impact depend on solder alloys, interfacial strength, intermetallic formulation, and etc. Submodeling technique is applied to model detailed structure of critical solder joint. The stress and strain concentration at different locations of solder joint correlate well with failure modes observed during testing. Eutectic solder joint is more susceptible to bulk solder failure while Sn-4Ag-0.5Cu is more susceptible to intermetallic compound (IMC) layer failure. Softness of Sn-37Pb reduces the stress in IMC while increases the plastic strains in bulk solder. Life prediction model is determined by solder joint failure mode and mechanism. Stress criteria is suitable for IMC interfacial brittle crack while plastic strain criteria should be applied for life prediction of bulk solder ductile failure

An approach to board-level drop reliability evaluation with improved correlation with use conditions

The work presented in this paper has been carried out in order to find means to improve the existing methods of board-level drop reliability assessment to better represent the use environment loading conditions of modern portable electronic devices. To provide goals and guidance for the development work, eight commercially available smart phones from different manufacturers were evaluated for their free-fall drop response. The results show that the drop response can be divided conceptually into two parts: (i) forced bending/twisting of the component board at the moment of the impact, and (ii) subsequent (resonance) vibration of the component board. The strain magnitudes caused by the forced bending and twisting were much higher than those by the post impact vibrations. Furthermore, sharp strain peaks were often identified within a few milliseconds from the impacts. The results of the Finite Element simulations show that the distribution of strains on the component boards is highly nonuniform; The regions of high strain are localized because they are caused either by the forced bending (by surrounding covers and frames in part i) or by internal collisions between the component board and the surrounding mechanical structures (sharp strain peaks in part ii). The devices were characterized for maximum strain, average rate to maximum strain, vibration frequencies and maximum value of deceleration. The following goals were set for the development of a board-level drop test methodology: a test board that simulates the response of modern portable products should be able to produce board strain well above 3 500 μ (“micro-strain” = [10-6 m/m]) and mean rate to maximum strain close to 7 s-1. Reaching these objectives requires significant changes to the existing JESD22-B111 approach. An investigation on the effects of the shape of deceleration input pulse was carried out first, followed by modifications for the board support to further increase the maximum strain and the average strain-rate of the JESD22-B111 compliant printed wiring board. The support was modified by replacing the four point supports by line supports at both ends of the rectangular shaped board. This approach can achieve the maximum strains of 5 000 μ when coupled with the optimized deceleration input. In addition to the increased strain and strain-rate, dampening of the board vibration became more effective, which is also in a better agreement with the typical drop response of portable electronic products.

Effect of strain rate on adhesion strength of Anisotropic Conductive Film (ACF) joints

Advantages of Anisotropic Conductive Films (ACF) technology, especially the low bonding temperature, fine pitch capability, elimination of lead and low cost, have resulted in ACF interconnects being widely used in contemporary portable electronic devices. Such products are exposed to various types of mechanical loadings, such as shock and impact during handling and accidental drops, throughout their life-cycle. Such impacts involve a risk of a mechanical damage to the electronic components and their interconnections to the PWAs inside the product housing. Evidence indicates that damage to the ACF matrix can directly result in instability of the contact resistance of the bonding. This study evaluates the ACF adhesive strength under mechanical loading conditions. ACF adhesive strength with three selected width and pitch, two kinds of surface finishes: ENIG and OSP surface finish is studied at different dynamic load velocities for tensile and shear tests. ACF/ENIG interface is found to exhibit higher adhesion strength than ACF/OSP interface. Furthermore, some rate dependency in adhesive strength is observed. ACF adhesion strength is captured in a rate-dependent traction-separation constitutive response in Finite Element (FE) Cohesive Zone Models (CZM), to model the damage initiation and evolution in the ACF during tensile and shear tests.

Hybrid Simulation Method for PWB Level Drop Tests

PWB level drop tests are widely used as a standard test method to evaluate the reliability of PWB and packages under drop conditions (JEDEC Standard JESD22-B104-A). The drop height and test setup need be adjusted in order to achieve the requirements of a peak shock of 1500g and an impulse duration of 0.5 ms. Generally, the ground need be covered with a thin layer of rubber pad to absorb some of the impact energy. However, this rubber pad will bring challenges for modelling due to large deformation, nonlinear hyperelasticity, and contact. And sometimes, it may also cause the convergence problem. Therefore, a hybrid drop simulation method was developed. This hybrid method can not only circumvent the difficulties mentioned, but also increase the efficiency and reduce the CPU time of PWB drop simulation. When simulating a PWB board level drop test, generally, not only the PWB and the components assembled on it need be modelled, but also the drop vehicle, rubber pad, and ground should be included in the model. For the hybrid drop simulation, however, only part of drop vehicle need be modeled and there is no need to model the ground and the contact between the ground and the drop vehicle. Then an acceleration time curve measured from drop test is applied to the hybrid model so that the responses of the model will mimic the real drop situation. In this way, not only the simulation time is reduced due to smaller model sizes, but also can some difficulties related to large deformation, contact, and nonlinear material properties be avoided. Finally, a comparison of a bare PWB and a populated PWB drop cases was used to validate this hybrid drop simulation method. A reasonable correlation was achieved.Copyright