Research On Machine Vision Alignment Technology Of Area Array Packaging

With the development of micro-electronic technology, the technology on area array packaging (such as BGA, Flip Chip), which provides more I/Os, smaller size and better performances of the chip than many standard packages,is developed rapidly. Meanwhile, the packaging equipments are more complex than ever. Vision alignment technology is one of the key techniques for packaging equipments. The structure of alignment system based on machine vision, the design of components, and the algorithm of image processing and alignment are systematically investigated in this dissertation.A frame of vision alignment system is firstly proposed. Then its functional modules are analyzed and the principia how to choose its components are given. Optical system that is the key module of the alignment system is studied, which concerns adoption of the LED illumination with coaxial beam and improvement on flying capture of the chip's and substrate's images by CCD.In the dissertation, the process and geometric features of area array chip are briefly presented. Furthermore, two methods of alignment, such as based-on the chip surface pins and based-on the fiducial marks, are induced. Comparing both of them, the method of based-on the fiducial marks is used due to its high precision and rate. Moreover, several common fiducial marks are analyzed and alignments by the marks are given. The result of experiments shows the circle mark is better than other marks for locating the chip of area array packaging.A new algorithm of sub-pixel edge detection for circle marks is systematically proposed. Firstly, real images of the marks are preprocessed by means of the methods such as image filtering, threshold segmentation, and edge extraction. The processing result and its analysis are given. In order to realize sub-pixel location using circle marks, some algorithms of sub-pixel edge extraction based on pixel rank images are discussed, included interpolation, fitting methods, and moment-based edge operator. Comparing these algorithms, a new algorithm of circle mark detection and its procedure are given, which is based on moment-based edge operator for sub-pixel edge extraction and least squire fitting for the position of circle marks. The experimental results clearly show that the algorithm can meet the requirement of the area array packaging technology. A software system is developed for image processing and analyzing in vision alignment of area array packaging. The software system realizes the algorithms mentioned above.