路徑延遲錯誤(Path Delay Faults, PDFs)的測試樣本通常在每個邏輯閘的邊界輸入使用靜態(static)和強健(robust)敏化規則,因此延遲缺陷影響的路徑延遲錯誤將會被這些測試樣本無條件的捕捉到。然而在功能敏化(Functional Sensitized, FS)的情形下,如果不被功能敏化的邊界輸入主控值(controlling value)所覆蓋,依然有一個種類的缺陷可以被無條件的測試。 在這論文的第一部份,我們提供了一個新的樣本產生方法產生功能敏化路徑延遲錯誤來改善缺陷的偵測率。我們使用衝突分析來建立功能敏化決定樹。然後收尋整個樹來分析測試樣本的品質。採用這個方法得到的實驗結果表示,我們可以測到比強健或非強健(non-robust)樣本更多關鍵路徑(critical paths)的區段(有一個測試電路增加了24.03%) 為了測試小延遲缺陷(small delay defects),轉變錯誤(transition fault)樣本產生器應該對每個錯誤用長路徑來偵測。為了達到這個目的,論文的第二部分,我們把路徑延遲測試的觀念套用在轉變錯誤樣本產生器。而這個樣本產生器會使用最長的路徑來測試小延遲錯誤。我們把論文第一部份做出的功能敏化測試的方法套用到這部份的轉變錯誤樣本產生器,去尋找最長的功能敏化路徑來測試每個轉變錯誤。實驗結果,我們比較這方法產生的功能敏化測試樣本和強健或非強健測試樣本。結果顯示,功能敏化測試樣本存在著最長的路徑來測試每個轉變錯誤。在使用功能敏化的方法下,整體平均有5.34%的轉變錯誤可以被較長的路徑測試。
Test patterns of path delay faults (PDFs) are usually generated with static or robust sensitizing criteria for side inputs of gates, because defects affecting the delays of the PDFs will be captured by these patterns unconditionally. However, under functional sensitization (FS), there exist a class of defects that can be tested unconditionally, if they are not masked by the off-input controlling values. In the first part of this thesis, we propose a new pattern generation method for functionally sensitizable PDFs to improve the detectability of defects. We use conflict analysis to build a FS decision tree. Then traverse the tree for analyzing the quality of test patterns. It is shown in the experiments that with the proposed method, extra segments (up to 24.02% for one benchmark) of critical paths become testable compared with only robust/non-robust patterns. In order to test small delay defects, the transition fault pattern generation should test each defect with the long path. For this purpose, we apply the concept of PDFs for transition fault pattern generation in the second part of this thesis. This pattern generation searches the longest path for testing the small delay defects. We apply the FS ATPG in the first part into this transition fault pattern generation for testing each transition fault with longest FS path. In experimental results, we compare FS test patterns with robust/non-robust test patterns. It shows that FS test pattern exist the longest path for test each transition fault. In average, there are 5.34% transition faults detected with the longer path if FS is considered.