Real-time automatic inspection of macro defects in in-line TFT fabrication process

Abstract

In this article, an automatic detection and judgement method for macro defects in thin film transistor (TFT) fabrication process is proposed using a high-resolution line charge-coupled device camera as a preliminary inspection method for TFT liquid crystal display (TFT-LCD) panel. Macro defects are classified primarily into four types. Relevant detection and judgement methods are then applied according to the type of macro defect using different criteria such as diffraction pattern shifts, just noticeable differences, and the gradient of inspection images. The proposed method is verified with industrial experiments. In the experiment, 559 TFT glasses are used, which are sampled among the glasses and were judged as PASS (i.e. non-killer-defected glass) in the TFT fabrication process by a conventional human inspection method, but were judged as REJECT (i.e. killer-defected glass) in the module process. All macro defects in the sampled glasses were detected and rejected in the TFT fabrication process using the proposed method.