A Statistical Approach to Three-Dimensional Shape Inspection of Micro-Solder Balls

Introduction

The recent trend in semiconductor manufacturing has been to seek smaller parts with high integrating capabilities; the manufacturing processes have become more complex with the advent of the ball-grid-array (BGA)-based packaging processes. This has increased the prominence of BGA quality control and the resulting two-dimensional (2D) and three-dimensional (3D) BGA inspection research (Wu & Chen, 2004)Moreover, BGA ball diameters are now as small as 60 µm, thereby requiring more precise and effective defect inspections with the aim of ensuring quality and productivity. BGA inspections vary in type, depending on the stage of the packaging process, and they are typically non-contact-type inspections (Park &Kim, 2007). For example, special forms of radiation, such as X-rays, are often used where visual inspections of post-bonding cracking or breakage are infeasible (Feng et al., 2008). To inspect the shape of BGA balls before attaching them to a board or film, the typical methods employed are optical approaches that use visible light (Chen, 2007; Saito et al., 2003; Kim, 2012). BGA balls, having specular reflection characteristics similar to those of metal surfaces (Dash et al., 2018), require an effective inspection technique. For most specular reflection surfaces, certain patterns such as optical interference are projected onto the surface for image distortion analysis, or a ring light is used for geometrical optical analysis. In previous studies, 127-point light sources and 5-step ring illumination (Kim & Nguyen, 2013) were used as shape recognition methods for the three-dimensional inspection of BGA balls. To reduce inspection times, several studies have also been conducted on the simultaneous inspection of multiple balls of BGAs. For example, the images of BGA packages arranged on a printed circuit board (PCB) were simultaneously captured and were used to inspect the presence or absence of BGA balls and the distance between BGA balls. However, most of the research has focused on the inspection of BGAs after the ball placement stage, that is, on the array of solder balls positioned on the BGA substrate. For a solder ball as a material itself, often called “solder sphere” by material manufactures and suppliers, there are several aspects to be considered. Manufacturers need to ensure consistent supply of high-quality material and automated inspection technologies to produce precisely shaped balls (Perabo, 2014). Furthermore, the smaller the BGA, the more important it is to consider productivity during the inspection process, and the smaller the solder ball, the more difficult it is to handle. Miniaturization leads to productivity-compromising problems such as (a) delays in the inspection procedures due to the requirement of individual inspections for a considerable number of small balls and (b) the requirement of precision equipment for handling smaller balls in the pre- and post-inspection stages. Accordingly, more effective solutions for inspection are necessary.

Methods for position determination or the positioning of balls for BGA inspection have proposed (Lin & Wang, 2014); however, they are limited to the stage after ball placement. To eliminate the complexities involved with sophisticated equipment and improve productivity, our proposed method aims to simultaneously capture images of multiple solder balls arranged arbitrarily on a tray and accomplish inspections using statistical characteristics. The method involves a vision system comprising a camera and lighting system with spatially arranged light-emitting diodes (LEDs). The images of the LEDs, reflected on the surfaces of the BGA balls, are acquired, and the positions of the LEDs are determined through image processing. These data are then used to extract 3D-shape image information and applied to inspect the ball shapes.