Automated inspection of printed circuit board patterns referenced to CAD data

A state-of-the-art system for automated detection of defects in printed circuit board patterns for mainframe computers is described. This system compares detected patterns with corresponding patterns, which are electronically generated from the CAD data. Several techniques are developed: modulation compensation, thresholding that focuses on detecting narrow gaps and thin shorts, alignment technique that allow inevitable distortions in a PCB, defect fatality recognition, and so on. As a result, highly reliable inspection system, which can detect short circuits as small as 7 mu m wide within 4.5 min. in a 600x800mm printed circuit board, is accomplished.