在本篇論文中，我們利用計算NBD及NCD之龍捲風模型晶圓圖分析法，針對隨機瑕疵（Random Defect）、製程偏移（Process Variation）及兩者結合所產生之晶圓圖模擬分析，結果發現產生故障晶粒的方法雖不同，但在B-C圖上以NBD為橫軸，NCD為縱軸之龍捲風瑕疵圖樣分佈卻趨於一致。再將七種常見系統性錯誤的晶圓瑕疵圖樣放至晶圓圖分析模型，最後在B-C圖上得到不同的瑕疵圖樣分佈，可見本研究所建立之晶圓圖分析模型，確實可作為一個晶圓瑕疵判讀的檢驗機，並進而達到降低成本、提高效能及良率的目的。 In this thesis, we use the tornado model, which counts the total number of bad dice (NBD) and changed directions (NCD) on a wafer amp, to simulate and analyze wafer maps with random defects, process variation and both. It is shown that different sets of wafer maps generated by different types of sources are almost with the same tornado trend in the B-C diagram, where NBD is as x-axis and NCD is as y-axis. Then we analyze seven systematic errors defect pattern in wafer by wafer map analyzing model and get different distribution of defect pattern in B-C diagram. The wafer map analyzing model we set up in this research can indeed be a checker machine which can judge the source of defect in wafer and achieve the goals of cost down and improving the yield rate.
