本篇論文是以過去提出的迴力棒特徵圖(Boomerang-Chart)為基準，藉由分析迴力棒的趨勢且產生迴力棒的上下限，使未來對晶圓的探討能更有效且省時。 首先，我們先對五種不同尺寸的晶圓所產生的迴力棒特徵圖(Boomerang-Chart)來做出五條迴力棒的中線，再來透過各個晶圓圖產生NBD及NCD兩個參數，把這兩個參數除上晶粒做正規化產生NNBD及NNCD兩個參數。接下來透過每個瑕疵數灑20點來紀錄當NNBD等於0.1、0.2、0.3、0.4、0.5所相對應的NNCD標準差。最後以中線加上1.96倍NNCD的標準差成為上限; 中線減去1.96倍NNCD的標準差成為下限,我們可以得到95%信賴區間的上下限網子來判別分佈的瑕疵數是否在上下限之間，進而得到提高測試效率及降低成本的目的。 ;In this paper, we use Boomerang Chart that we published in the past to analyze the wafer maps by generating upper bound and lower bound. Most importantly, we can reduce time and analyze efficiently. At first, we choose five kinds of size of wafers that we would like to analyze, and simulate basic centerline according to these five kinds of size. Second, we will create parameters NBD and NCD from every wafer in every data, and normalize the two parameters to create two new parameters, NNBD and NNCD. And then, we will create twenty points from every defect. When NNBD equals to 0.1, 0.2, 0.3, 0.4 and 0.5, we save the NNCD standard deviation. Last, we use the centerline to plus one point nine six times NNCD standard deviation and to subtract one point nine six times NNCD standard deviation. We can obtain the upper bound and lower bound in 95 percent confidence interval. According to the upper bound and lower bound, we understand whether the defects are inside or not. So, we set to get the achievement of increasing yield、testing efficiency and reduce the production cost.
