中文摘要 本篇論文針對過去實驗室研究的已分類瑕疵晶圓圖，以不同的方向重新計算晶圓圖的特徵參數，分析新參數和原晶圓參數產生的差異，並且對於晶圓圖的隨機性判斷又產生何種影響? 過去我們利用晶圓圖故障晶粒數( NBD )和故障晶粒連續線( NCL )產生迴力棒圖，並且有初步提出B-score來作為判斷隨機瑕疵晶圓圖的標準。當初，我們在計算故障晶粒連續線( NCL )時總是固定同一方向，本篇論文藉由旋轉和翻轉晶圓圖以產生七個不同方向的晶圓圖，在本質上該七片晶圓仍和原晶圓是同一片，但是由於NCL的計算方式，可能會造成數值變動，而B-score也會因此受到影響，最終導致隨機性檢定也可能出現偏差。 本篇論文便是要探討旋轉和翻轉的七個分身晶圓圖，在隨機性檢定中與原晶圓所判斷的隨機性是否一致，並觀察一致性的偏差，以此確保我們的檢測維持可靠度。 ;ABSTRACT In this paper, we recalculate the characteristic parameters of wafer maps in different orientations for the classified wafer maps in previous research. We analyze the difference between new parameters and original wafer parameters, and what effect does it have on the randomness test of wafer maps? In the past, we used Number of Bad Die (NBD) and Number of Contiguous Line (NCL) to generate Boomerang. We propose B-score as a standard for judging wafer maps. At that time, we always fixed the same orientation of wafer maps when calculating NCL. Now, we generate wafer maps in seven different orientations by rotating and flipping the wafer map. In essence, the seven wafers are still the same as the original wafer. However, due to the way NCL is calculated, it may cause numerical changes. Eventually, the randomness test may also be different. In this paper, we explore whether the seven clones and the original wafer is consistent in the randomness test to ensure that our test is reliable.
