A Wearout Early Predication Method by Using MACD Technique

• Main Authors: Kan-Chun Hsieh, 謝侃君
• Other Authors: 黃德成
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/96452167784628562714

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 99 === In recent years, due to the prosperous development of the semiconductor industry, the reliability of CMOS has became an important research issue. The size of the integrated circuit is made smaller, so the Gate Oxide of CMOS also needs to be diminished and became thinner. Therefore, it is easy to lead to increase leakage current. If leakage current is too high, the product will turn to the period of wearout failure. The lifetime of the product can be separated as three periods, which is the early failure, random failure and wearout failure. The early failure can be measured in the process of testing by using Burn-In. When the user takes the product, it has already entered into the period of random failure. Then it will get into the wearout in the end. However, the duration before the product entering the wearout is hard to predict. If there is no proper precautionary method, it could cause serious losses. Therefore, we propose a statistical method by using MACD technique to find the time point when the product in the Bathtub Curve before entering the wearout. In the experiment of this research, we present the method of Single-point predication and Multi-point weighting predication for different situations. Also, we add the factors of temperature and voltage to moderate the Weighting coefficient, making the time point of predicting wearout can be more precise.