適用於群聚多重轉態延遲錯誤診斷技術

Abstract

在先進的製程中，晶片上的缺陷容易群聚在小區域。另外，局部電壓壓降也會造成部分區域的延遲錯誤。然而傳統診斷技術無法有效解決群聚延遲錯誤。這篇論文針對實體群聚多重轉態延遲錯誤提出一個診斷方法。我們使用一種效果原因的追蹤技術來尋找可能的延遲錯誤候選人。 此外，我們提出了一個近似覆蓋經驗法則來尋找一群延遲錯誤候選人。這個經驗法則允許試驗故障與模擬故障有些不同，如此一來錯誤遮蔽與增強效應可能被忍受。在ISCAS'89和ITC'99的電路上，實體群聚多重轉態延遲故障實驗中顯示出我們診斷技術的能力。當插入十個實體群聚多重轉態延遲錯誤時，我們的診斷技術正確率(0.55)比商業軟體的正確率(0.24)還要高。

For advanced technologies, defects on a wafer tend to cluster in a small area. Also, local IR drop can cause delay faults in a small region. However, traditional diagnosis technique cannot handle clustered delay faults very well. This thesis presents a multiple fault diagnosis technique for physically-clustered transition delay faults. We use an effect-cause tracing to identify possible transition delay fault candidates. Then, we propose an approximate covering heuristic to find a group of candidate faults. This heuristic allows some difference between test failure and simulation failure so that masking effect and reinforcement effect are likely to be tolerated. Simulation on ISCAS’89 and ITC’99 benchmark circuits with physically-clustered multiple transition delay faults demonstrate the effectiveness of our diagnosis technique. The accuracy of our technique with 10 physically-clustered transition delay faults (0.55) is much higher than that of a commercial tool (0.24).