In-Line Inspection And Measurement Theory And Technology Of IC Packaging

IC(Integrated circuit) is widely used in all kinds of electronic products, profoundly affected the modern people’s life. And the requirements of the quality of electronic products become higher. Thereinto, one of the key factors affecting the quality of electronic products is the quality of IC packaging. Thus, effective detection measurements should be implemented to guarantee the quality. Traditional artificial visual and sampling methods have a large labor intensity, low detecting precision and poor reliability of faults, thus these classical methods are unable to meet the requirements of modern mass productions. With the requirements of high speed, high precision and low cost, the high-performance machine vision system should be studied.In this research, a high-performance machine vision system is developed, which can inspect the product quality of double-side IC with SOP(Small Outline Packaging). By using the proposed system, IC pins can be located with high-speed, the coplanarity of IC leads can be measured in high accuracy, and the mirco defects can be detected correctly in real-time. As a result, the IC packaging can be inspected automatically in high performance. Main research contents are as follows:(1) A novel BLOB(Binary Large OBject) analysis algorithm using RLE(Run Length Encoding) is proposed. Based on this fast BLOB analysis approach, IC pins can be located correctly. First, the new data structures for RLE and BLOB linked lists is designed, to accelerate the access and modifications of data. Second, to avoid the traditional labeling conflicts and simplify the comparisons of connectivity branches, an efficient algorithm for BLOB analysis is presented. Furthermore, the area feature of objects can be extracted once the BLOB linked list is dynamically created or modified. Finally, to evaluate the performance of the proposed method, the ICs with various types of SOP packages were located by the proposed algorithm. The experimental results not only demonstrate that the IC pins could be effectively and robustly located with the proposed algorithm, but also show that the presented BLOB analysis algorithm runs faster than other traditional methods.(2) A sub-pixel edge positioning method based on the 1D(one-dimensional) gray moment is developed. The position of IC lead tips in sub-pixel level can be obtained. The proposed sub-pixel positioning method has four steps, including estimation, correction, precision positioning and verification procedures. The proposed method improves the fast 1D gray moment method and enhances the noise immunity, not only improving the measuring accuracy, but also ensuring the measuring speed. The experimental results show that the proposed method can achieve equivalent computation costs as classic methods and higher accuracy. Therefore, the proposed four steps sub-pixel positioning method is more suitable for the coplanarity measurement of IC leads than others.(3) A new single camera vision system for online coplanarity measurement of IC leads with SOP package is presented. With the proposed system, the measuring accuracy and speed can be improved. Firstly, a novel imaging optical structure with multiple reflection mirrors and a total reflection rectangular prism is developed. With this structure, the two side views of IC leads are captured at the same time in a small FOV(Field Of View). To improve the assembly quality, the relationship between the assembling deviation of optical components and measurement error is investigated. Secondly, a fast algorithm for the coplanarity measurement of the IC leads is also proposed. This algorithm mainly consists of the IC lead tips searching, the local sub-pixel positioning and the local image correction. Finally, high precision measurement instrument is used to verify the accuracy of the proposed approach. The experimental results show that the proposed approach can measure the coplanarity of IC leads in real-time and the coplanarity measuring error is about 5 μm. Furthermore, the proposed method can result in better performances in measuring speed and accuracy than the traditional methods.(4) A new method based on the statistical learning theory is presented for defect inspection of molding surfaces in IC packages. By using the good samples, the gray value of each pixel in the same position of each region of interest(ROI) is counted, then the confidence interval of the normal distribution is formed independently for each pixel of each ROI. The scratches, pits and cracks can be detected correctly in real-time. The proposed method mainly includes: lens distortion correction, image tilt correction, circle detection, ROI location, statistical learning and inspection. The experimental results have proved that the proposed method can detect the defects correctly and completely.Finally, the future research direction of IC packaging inspection is proposed based on the summarization of the whole paper.