針對改善圖樣品質之平行自動化N次測試圖樣產生方法

Abstract

由於IC的設計複雜度增加以及特徵尺寸的縮小，簡單的錯誤模型 (fault model)已經無法偵測到足夠的缺陷 (defect)。 為了偵測更多的缺陷同時避免使用複雜的錯誤模型，偵測同個錯誤多次是一個可行的辦法，這是因為有些缺陷會在多次的錯誤偵測中恰巧被偵測到。Mentor graphics 提出了兩項計量 (metric) (BCE 和 FOC) 來衡量多次偵測圖樣對於現實電路缺陷的次是品質。 本篇論文中提出了一個平行化，搜尋空間分割 (search space partitioning) 的 N 次測試圖樣產生技術，考量了上述提到的兩項計量並且優化它們。動態的錯誤選擇 (Dynamic fault selection) 被用來提升 BCE 計量；在圖樣產生途中的測試選擇 (test selection) 則用來提升 FOC 計量。 本論文提出的方法在ISCAS89 和 ITC99 測試電路上進行驗證，實驗結果顯示，和Mentor Graphics 所提出的 N次測試自動化圖樣產生流程相比，本論文提出的方法可以達到更好的測試品質。

Due to the increasing design complexity and the shrinking feature size of the modern IC’s, the single-stuck-at fault model is no longer adequate of covering sufficient amount of defects. To increase the defect coverage while avoiding using complex fault models, one solution is to detect each fault multiple times; by coincidental detections, the defect coverage is increased. Two metrics (BCE and FOC) are proposed by Mentor Graphics to estimate how well the multiple detect patterns can cover the real defects. In this thesis, a parallel, search-space partitioning N-detect ATPG methodology is proposed. The ATPG take the two multiple detect metric into consideration and aiming at improving the pattern quality according to them. Dynamic fault selection is used to increase the BCE metric and the test selection during ATPG is performed in favor of the FOC metric. The proposed techniques are validated using ISCAS89 and ITC99 benchmark cir-cuits. The experimental results show that the proposed techniques can achieve better pattern quality in terms of the two metrics comparing to the multiple detect ATPG flow proposed by Mentor Graphics.