Investigation on the failure criterion of reliability testing for Pb-free BGA packages

Abstract

Daisy-chained test vehicles are commonly used in board level reliability testing. By continuously monitoring the in-situ daisy chain resistance change over time, a failure could be captured during cycling and eventually the failure data could be used to establish the solder joints failure distribution under different testing conditions. One of the most debatable matters is that when should one to determine a failure to occur. Per IPC 9701A [1] a failure is defined as 10 1000-ohm events in 1 micro-second duration for event detector or 20% increase over the baseline resistance for data logger. Other threshold values such as 100, 300, or 500 ohms are also commonly used by packaging reliability community. Such a wide range of failure threshold values may introduce significant delta in terms of cycle numbers for Pb-free solder joints if different criteria would be used as reported by Henshall, etc [2]. Therefore a systematic study of the impact of using such diversified resistance values on the final failure distribution is necessary and important such that no big difference among reliability results from different sources. The purpose of this study is to investigate the impact of different failure thresholds on Pb-free solder joint failure distribution for most commonly used packages. The test vehicle, designed on an 8″×15″ double-sided printed circuit board (PCB) with multiple test sites, was populated on both sides with daisy-chained components. To reflect the real situation, the components were selected to include different package types (FCBGA, PBGA, CSP, QFN, etc), different pitches (0.4–1.0 mm), and different package size (6–50mm). The assembled test vehicles then went through 0C–100C thermal cycling, the cycle numbers corresponding to different resistance thresholds were recorded and compared. The test results showed that the failure threshold has significant impact on Pb-free solder joint failure distribution, thus it is important to unify the failure criterion such that the reliability results from different sources could be compared side by side. For some packages especially small wire-bond packages that have relatively low baseline resistance, the 20% failure criterion may be too sensitive to the resistivity changes caused not by solder joint failure but other events such as connection cable resistivity change over time or temperature.