Automatic Solder Joint Failure Mode Analysis Based on Dye & Pry Image Processing

Ball Grid Array (BGA) interconnects have been widely used for many different components, including NAND, ASIC and DDR. They are among the most vulnerable parts of a microelectronic device under environmental loading conditions such as thermal cycling and/or mechanical shock/vibration during production, transportation and service. Damages/cracks are most likely to initiate and propagate in BGA solder joints, which can lead to the failure of devices. Dye and pry (D&P) method is commonly used to analyze crack location and area. However, this method is labor intensive if manually done and can generate inconsistent results depending on operators' experience and skill. In the present work, we have developed an automated D&P analysis method based on computer image processing and physical understanding. First, individual ball images are automatically extracted from whole component D&P image. Then, K-mean clustering is applied to sort all pixels of each solder ball image into three different groups based on their red-green-blue values. Next, two failure modes, namely, crack propagating through a solder ball (Mode-C) and crack penetrating into substrate/PCB laminate (Mode-A/E), can be distinguished according to the brightness of the individual image corresponding to metal or plastic. At last, the dye percentage of either mode can be calculated and reported with different colors. We repeat this procedure and generate the dye percentage for all solder balls automatically. The result matches well with human measurement on various components. The present method can not only shorten the D&P inspection time but also provide an accurate and consistent result on dye area percentage of different failure modes.