Mineo Nomoto

Automated solder joint inspection system using optical 3-D image detection

An automated system has been developed for visually inspecting the solder joints of SMDs (Surface Mounted Devices). The system is capable of inspecting fine pitch components down to 0.3 mm pitch QFPs (Quad Flat Packages). A unique image detection method was also developed to obtain precise 3-D images of solder joints. The principle of a confocal microscope is employed but plural sensors are used to detect reflected light at different focusing positions simultaneously. The system is unaffected by secondary reflection and dead angles. The warp in a PC (Printed Circuit) board surface is calculated in real time using the detected 3-D images, and board height to be detected in successive areas is predicted based on this calculation. Real-time automatic focusing control is then performed using newly developed defect detection algorithms, the system can recognize leads, pads and solder fillets from the detected images. Because 3-D shape features are extracted and used for defect judgment, user-defined parameters have been made easy to understand and/or to modify. Operational evaluation of the system confirms a 100% defect detection rate and a very low false alarm rate (0.16%).

Automatic screen-printed circuit pattern inspection using connectivity preserving image reduction and connectivity comparison

Describes a new automatic pattern inspection method which can reliably detect fatal circuit defects without giving false alarms due to circuit patterns with slightly jagged edges. Also described is a high-speed automatic inspection system for screen-printed circuit patterns of an unbaked ceramic layer, known as a green sheet. For fatal defect detection, the connectivity comparison method was adopted. In order to speed up this method and to further improve it, connectivity preserving image size reduction, a new concept, was developed.<<ETX>>

Automated Pattern Inspection System for Circuit Pattern of Multi-Layer Thin-Film Substrate using Optical Discrimination Method.

This paper describes an automated pattern inspection system for circuit patterns of multilayer thin-film on a ceramic substrate. Thin-film layers, usually made of resin and metal circuit patterns, are alternately formed on a ceramic substrate. To inspect all those thin-film layers inline visual inspection is used. This inspection involves detecting defects existing in a surface layer just after the layer is formed. Surface circuit patterns must be optically differentiated from patterns on inner layers. For this purpose, bright and dark-field combined illumination is used and the wavelengh of reflected light to be detected is selected in a specific range. Illumination is achieved by a 4-fold ring-light-guide structure and bright-field illumination. The “Local Pattern Comparison Method” is used for defect detection in which a detected binary image is locally compared with a reference image. This method can be implemented by means of an image processing hardware system using a pipelined architecture for real-time operation. An automated pattern inspection system was constructed with the above mentioned pattern detection and defect detection method. According to the evaluation results, the inspection system can reliably detect defects as small as 10 μm.

Automatic 2 1/2 D shape inspection system for via-hole fillings of green sheets by shadow image analysis

The authors describe an automatic 2 1/2 D shape-inspection system for via-hole fillings of green sheets, which are unsintered layers of multilayer ceramic substrates. Two shadow images re taken with two oblique illuminations from different directions. To determine whether the shadow results from a convex structure or a concave structure, the shadow regions are labelled by analyzing the relation between their edge positions. After labelling, concave and convex defects as small as 30 mu m can be detected independently by measuring the length of the shadow and comparing it to the predetermined criteria. In addition to concave and convex defects, lack-of-filling defects can also be detected with very high accuracy by counting the area of the filling pattern detected from an oblique direction.<<ETX>>