Study On Offline Bonding Quality Detection System

With the substantial increase of the integration of microelectronic circuits, a higher demand of the miniaturization of components and packaging technologies is put forward. Inside the semiconductor packing, the connection between the semiconductor and external pin and the connection between chips have been the key factor to ensure the chip and the external electrical connections. Wire bonding is the most important chip packaging technology. The quality of bonding is directly related to device performance and reliability of the entire package.After wire bonding, the quality inspection must be made to the wire bonding products. The main inspection items are visual inspection and pull thrust test. Pull thrust test is a destructive bond pull test, so only selected groups of solder joints are measured and can not cover all of the solder joints. Therefore, visual inspection has been the most effective qualitative method to evaluate the quality of the bonding. The traditional visual inspection requires inspectors to observe each bonding point and every bonding wires to determine whether they are in accordance with the technical requirements. The inspector make the initially determine the quality of the bonding products; remove the substandard products so as to prevent defective products go to the next process.In this paper, the image processing techniques has been used to realize diode automatic identification, positioning as well as lead width measurements. The main contents of this article are as follows:First, the paper introduces several assessment methods of inspecting wire bonding quality, such as visual inspection, destructive welding tension detection, ball bond shear detection, electrical testing, ray detection, ultrasonic testing, optical detection and these methods were compared with each other.Secondly, this paper makes comparison and study on algorithms of image segmentation, contour extraction and edge detection in image processing technology. Summarizes up the advantages and disadvantages of various algorithms and derive the most specific algorithm for this system. This paper firstly make binarization of bonding product image taken by the television microscope through the fixed threshold method, then make the binary image contour extraction and after that use the projection method to complete the positioning of the diode.Finally, make edge detection for the image and apply image thinning algorithm to get the single-pixel edge of the diode and gold thread. Calculate the proportion of size so that can achieve accurate measurement of the width of gold wire. The bonding quality off-line image detection system described in this paper can improve the quality of bonding quality inspection which was rely mainly on the level of knowledge and experience level of the inspectors, with no uniform quantitative criteria, easily lead to some of the human deviation. The system can also largely reduce the amount of visual fatigue. So, the research of this paper has a certain theoretical significance and practical value.