Proportional Hazard Model of doped low creep lead free solder paste under thermal shock

Abstract

In this experiment, the thermal performance of various lead-free doped solder paste alloys on leaded and leadless packages on laminate substrates were investigated. The Primary Goal of this test is to find a manufactureable solder paste that will mitigate the effects of aging on lead free solder joints. The results discuss the effects of different paste materials, the effects of doping level, and reflow profile on the resulting reliability of a variety of surface mount components. The test vehicle consists of 35 mm (PBGA 1156), 31mm (SBGA 304), 15 mm (CABGA 208) and 6mm (CABGA 36) ball grid array packages with 1.0 mm, 1.27mm and 0.8mm pitch, respectively. Additionally, a leadless 5mm QFN package (MLF 20) with 0.65mm pitch and 2512 SMRs are included in order to understand the effect of doped solder paste on conventional packages. The test boards are built with (a) three different reflow profiles on 6 mil stencil thickness, and (b) one reflow profile on a 4 mil stencil thickness, to study the differences in doping volume effects of the new doped alloys. The test vehicles were subjected to high temperature accelerated life test (HALT) in liquid shock (thermal shock) testing. Each test vehicle underwent 3000 thermal cycles with peak temperatures of -40°C to +125°C on a 15-minute thermal profile (5 minutes dwell time and 2.5 minutes transition time). Reliability of the test packages were determined from the ability of the components and solder interconnects to withstand the thermal stresses induced by alternating high and low temperature extremes. The time to failure were right censored after 3000 cycles. The experimental variables include paste materials, solder paste manufacturer, Reflow profile, stencil thickness, packages and component solder sphere. The response variable used in this test is Time to Failure (cycles). The effect of experimental variables on Time to Failure were assessed at 5% level of significance using Proportional Hazard Model. It is found that all the experimental variables except reflow profile have significant impact on the Time to Failure. Since this experimental data consists of heavy censoring, Censored Quantile Regression model is developed and it is compared with Proportional Hazard Model.