建構半導體晶圓針測圖樣分類辨識模型

Development Pattern Recognition Model for Classification of Circuit Probe Wafer Maps on Semiconductor

呂建鋒(Chien-Feng Lu)

指導教授 ： 洪炯宗

國立中央大學/資訊電機學院/通訊工程研究所碩士在職專班/碩士(2012年)

摘要

在半導體業界中，晶圓針測 (Circuit Probe) 屬於生產過程中的後段測試，其測試結果包含了晶圓良率的好壞及追溯異常製程與設備所需之重要資訊，並且產生圖形化的故障圖樣分佈於晶圓上，此資訊為製程錯誤診斷及機台故障檢視提供很多有用的線索。因此，了解晶圓針測圖樣所代表的意義及造成故障圖樣的原因，在工程資料關聯性分析中是件非常重要的工作。為了減少以往人為檢視方式所消耗的時間，一個準確的晶圓自動化分類系統在工程資料分析中是不可或缺的一套工具。本論文使用影像處理中偵測直線之霍夫轉換(Hough transform)及偵測圓形之圓形霍夫轉換(Circular Hough transform)演算法擷取直線及圓形分佈故障圖樣之特徵，並且使用了數種常見的特徵擷取技巧來取得其它不同的故障圖樣。研究中提出一套結構化系統，將晶圓特徵擷取出後透過數種資料探勘分類演算法建立不同的分類器(classifier)，並且驗證所提出之數種特徵可有效消除雜訊對晶圓分類準確度的影響。最後以實際半導體資料及模擬資料評估不同資料勘分類演算法對於本研究所提出的特徵擷取方法有最佳的準確度。

關鍵字

半導體晶圓分類 ； 晶圓針測 ； 霍夫轉換 ； 特徵擷取 ； 資料探勘

並列摘要

Abstract－ Circuit probe test is an end of line testing that the individual die has been measured at wafer level in modern semiconductor manufacturing. The test results are visualized as a spatial distribution of the failures on the wafer which can provide some valuable information for the production of failures. In order to reduce time consumption by human operation, a great accuracy of automatic classification system is clear needed for engineering analysis. In this paper, we demonstrate how a robust feature extraction procedure using by classical Hough transform (HT) and circular Hough transform (CHT) can be adapted to detect lines and rounds spatial patterns on circuit probe wafer map. In addition, we also used several technique to detect others spatial patterns. These features which are effectively eliminate the influence of noise to perform pattern classification. The presented methodology is validated with real fabrication data and several data mining classification algorithms are presented to evaluate the advantage of this methodology.

並列關鍵字

semiconductor wafer classification ； circuit probe test ； data mining ； Hough transform ； features extraction

參考文獻

1.Palma, F.D., et al., Process diagnosis via electrical-wafer-sorting maps classification, IEEE International Conference on Data Mining, 2005.

2.Chen, F.-L. and S.-F. Liu, A neural-network approach to recognize defect spatial pattern in semiconductor fabrication, IEEE Transactions on Semiconductor Manufacturing, 2000. Vol. 13, No. 3.

3.Shino, W., et al., Effective excursion detection by defect type grouping in in-line inspection and classification, IEEE Trans. Semiconduct. Manufact. 1999. Vol. 12, p. 3-9.

4.Palma, F.D., et al., Unsupervised algorithms for the automatic classification of ews maps: a comparison, ISSM International Simposium on Semiconductor Manufactoring, 2005.

5.White, K.P., et al., Classification of defect clusters on semiconductor wafer via the Hough transformation, IEEE Transactions on Semiconductor Manufacturing, 2008. Vol. 23, No. 2.