Defects detection on TFT lines of flat panels using a feed forward neural network

With the remarkable growth of demand for flat panel displays in recent years, having a wide verity of applications such as televisions, computers, cameras, hand phones, medical equipments, toys and etc, there is a stiff competition among manufacturers for high throughput product lines and low priced manufacturing. The demand for larger sizes of mother glass (e.g. Generation 10, 11 & 12) and high density thin film transistor (TFT) pitch patterning of FPDs have also been increasing. However when the pitch pattern density of panels is increasing the tendency of defects such as inter layer short circuits between TFT lines is also increasing. Therefore recently, manufactures have been forced to keep significant importance on detection of defects in early stages of production process of FPDs and repair them early. As a result, the speed and the precision of defects detection have been major issues for the manufactures of FPDs and researchers as well.

Automatic optical inspection (AOI) methods, which are based on still or video images, have been mainly used in the past for detection of defects during intermediate processes of fabrication lines of FPDs [5, 6]. Though the AOI methods, being based on images, are fast, non-contact and no damage occurs to glass substrates like pin probe methods, non-electrical defects such as particles on panel surface (micro dusts) and even slight color changes on TFT wirings can also be falsely detected as defects. A major drawback in AOI methods is that it is extremely difficult to distinguish all those non-electrical defects from electrical defects (open NG and short NG) that need to be repaired and restored. Furthermore, some common electrical defects, which occur with the increased pitch patterning, such as cracks on wirings and short circuits in between layers under wire crossings cannot be detected properly with AOI methods. In general, workers operating such repair systems are required to visually check all such defects detected by the inspection system and to judge whether a repair is needed or not which is tedious and cumbersome. Therefore to avoid operators being involved in such unnecessary work, it is necessary to improve the efficiency and accuracy of detecting electrical defects.

However, if it is only for detecting electrical defects then there is the commonly used pin probed inspection method of which electrode pins make direct contacts on each and every wiring on the panel surface and measure the current flown after applying some voltage. Though this method has the advantage of detecting nothing but electrical defects, it also has disadvantages such as very low-speed of inspection, poor adjustability for the changes of TFT circuit design and line pitch and the necessity of frequent replacing of pin probing fixtures of the inspection machines which is an expensive process.

For full text: click here

(Author: H.A. Abeysundara, Hiroshi Hamori, Takeshi Matsui, Masatoshi Sakawa

Published by Sciedu Press)