Electronic packages fatigue life predicion under random vibration loading conditions

In this dissertation both finite element modeling and experimental techniques are employed to develop an assessment methodology to predict the fatigue life of electronic components under random vibration loading. The fatigue life estimation procedure, that would help analyst to make relatively accurate prediction of induced fatigue life, has presented and each step of the procedure has been explained.;To help develop and verify the fatigue life procedure, a specially designed PCB with Ball Grid Array (BGA) packages attached was tested with different vibration excitations profiles, both random and sinusoidal and the results were used in conjunction with fatigue models to predict failures. Experimental modal analysis has been used throughout this dissertation to build a well correlated finite element model and to set the tests parameters.;The first step of the prediction procedure aimed at characterizing fatigue properties of the Pb-free solder joint (SAC305/SAC405) by generating their own S-N (stress-life) curve. A sinusoidal vibration over a limited frequency band centered at the test vehicle's 1st natural frequency was applied and the time to failure was recorded. The resulting stress was obtained from the FE model through harmonic analysis in ANSYSRTM. Spectrum analysis specified for random vibration, as the second step, was performed numerically in ANSYSRTM to obtain the response Power Spectral Density (PSD) of the critical solder joint. The volume averaged von Mises stress PSD was calculated out of the FEA results and then was transformed into time history data through inverse Fourier transform. Rainflow cycle counting was then used to count the damaging cycles in the time record to be used along Miner's rule to estimate cumulative damage of the critical solder joint.;Result obtained show that the calculated fatigue life based on the Rainflow cycle counting results, the S-N curve, and the modified Miner's rule agreed with actual testing results, especially for SAC 405 solder. The higher percent error seen in SAC 305 predictions is attributed to the poor fit of the S-N curve obtained for the material which can be corrected by running more tests. A comparison between predictions obtained through well known Steinberg's methods and current procedure showed that more accurate prediction is obtained following the time history analysis adopted by the procedure.;More testing focused on verifying the applicability of the procedure to different loading and boundary conditions and on analyzing stress state at the critical solder balls during excitation. Here it was shown that the procedure can be used to predict the fatigue life fairly accurately regardless of the profile or the boundary conditions, and the results also helped identifying worst conditions that could lead to some catastrophic failures by studying the stress state and identifying the most damaging modes. In general it was found that the first mode is the most damaging in any excitation, however second or higher modes might contribute significantly to the damage as in the 3 standoffs case.;Finally the effect of temperature was studied by assessing the fatigue life at elevated temperature (100°C), where it was found that temperature impacts both the system response and the fatigue life. More work is recommended here to accurately model the temperature effect into the prediction procedure.