| الوصف: |
Solid-state lighting (SSL), which is based 0n semiconductor Lighting Emitting Diode (LED), is the most promising and reliable energy saving solution for future lighting applications. Since a bare LED die can hardly survive without a package, one of the most import function of the LED package is to facilitate the electrical connection and heat dissipation of the light engine mounted on a printed circuit board (PCB) by Land Grid Array (LGA) solder joints. These joints have been proven to be one of the most vulnerable link in the system. Therefore, evaluating the reliability of solder joints becomes vital to ensure the long term reliability of this new lighting product. A combined theoretical and experimental approach based on Finite Element (FE) calculations is a fast and commonly adopted way of estimating the solder reliability in microelectronics industry. However, there is no fatigue model that is particularly suitable for Land Grid Array (LGA) solder joints, which are mostly employed in board level LED packages. . Additionally, it would be desirable to have cruder engineering guidelines to quickly estimate the effect of the various geometries and dimensions for new trial LGA configurations. On the other hand, relatively accurate lifetime predictions still required Accelerated Life Tests (ALT). However, current reliability testing methods based on the detection of final catastrophic failure generally require very long test time (up to 9000 hours or 55 weeks), which does not meet the industrial target to limit the testing time to 6-12 weeks. A test method that is capable of carefully separating and in-situ monitoring the fatigue damages may be able to resolve the issue by offering the possibility of early termination of the test once enough information is collected to predict final and catastrophic failure. In addition, this capability can be attractive to make more accurate Remaining Useful Lifetime prognostic of solder joint in many critical electronics in other systems. In Chapter 2, it was demonstrated that the widely accepted fatigue model for predicting the lifetime of BGA solder joint no longer valid for assessing reliability of solder joint in a LGA assembly due to its inappropriate critical elements selection method. Therefore, a more suitable critical element selection method for LGA assemblies is proposed based on statistical analysis of creep energy density distribution for elements in the vicinity of the likely crack initiation point. By adopting this new critical element selection method, a new energy based fatigue model for predicting LGA solder life times has been established by combining lifetime measurements with corresponding Finite Element Method (FEM) simulations for different material combinations and different LED package configurations. The model has a much better predictive power than the model based on the BGA approach. In chapter 3, a phenomenological response surface model is derived for a fast qualification of the reliability of LED packages with different designs by conducting a series of FEM simulations. It seems that a smaller carrier size and larger solder standoff height in general will result in better solder reliability. This trend is more pronounced in solder with smaller carrier size and larger stand-off height. Additionally, solder reliability decreases rapidly with solder coverage decreasing. The ratio between thermal solder area and electrical solder area AR has a significant influence on solder joint reliability in a SSL system. The optimal AR value is also observed to relate to carrier size and solder coverage. In general, the optimal AR value increases with increasing carrier size, and this phenomenon is more pronounced in packages with smaller solder coverage. When the chip carrier size is relative small, it is advisable to make the thermal pad and electrical pad comparable. The proposed methodology in this work, a combination of energy based fatigue modeling and FEM modeling proves to be very valuable for solder reliability optimization for LED packages. This methodology can also be applied to optimize the package configuration in terms of thermal performance, electrical performance, cost and combinations thereof. The predictive power and the limitations of the approach are also discussed in Chapter 3. The advantages of in-situ high precision damage monitor during ALT are demonstrated, which offers great potential to save testing time. In Chapter 4, precise fatigue damage monitoring is proven to be achievable by conducting in-situ high precision (apx. ±80 n?) electrical resistance measurement of each individual solder joint during testing. The method was shown to be capable of capturing viscoplastic deformation accumulation, resulted crack initiation and sequenced crack propagation in an isothermal fatigue test. Moreover, in Chapter 5, it is demonstrated that using dedicated electrode configurations, different types of fatigue damages can be separated from each other or highlighted depending on the testing purpose. The effect of viscoplastic deformation can best be separated from those due to crack formation in one electrode configuration whereby progressive viscoplastic deformation induces resistance decay, whereas crack initiation and propagation provokes resistance increasing. It allows a rather precise yet conservative estimation of the crack initiation point. The other two configurations studied demonstrated superior sensitivity to crack opening and closure during fatigue cycling, which gives saw-teeth shaped signal patterns, in which case the measured resistance increases with both progressive viscoplastic deformation and propagating crack. The experimental findings agree with results from FEA simulation and periodic tomographic analyses. However, even for very high resolution of the electrical resistance measurements as in the present work, identification of the very early stages of crack initiation (less than 5% of the joint contact zone) remains not really feasible. In Chapter 6, the knowledge gained for monitoring failure under isothermal conditions is transferred to the testing of solder joints subjected to thermo-mechanical loading, employing a different crack initiation identification method. Accelerated temperature cycling test was done to four LED like board level ceramic packages with in-situ periodic high precision electrical resistance monitoring for each solder joint inside. Not the actual electrical resistance values but the change in Temperature Coefficient of Resistance (TCRc) of the solder joint was correlated to fatigue damage evolution in the joint. Both viscoplastic deformation accumulation and crack propagation increase the TCR. A method to identify the crack initiation point based on the noise analysis of in-situ TCRc monitoring signal of individual solder joint is presented. It is observed that once a major crack is present in the joint the noise level increases significantly, and the rate of TCR increase increases with further crack growth. The crack initiation point as determined from the electrical resistance data using a built-in variance analysis tool in Matlab, was verified by micro-tomographic results and FEM simulation predictions. The new method as developed can be a very attractive technique for both solder accelerated test and RUL prognostics. |
