BCH碼之測試結果壓縮診斷技術

Abstract

本論文提出一個以BCH碼為基礎之測試結果壓縮器並具有診斷功能之技術。一般而言，以BCH碼為基礎之壓縮器對於實作來說有些缺點。以時間壓縮為主的方法，其所需的額外面積是非常大的，以空間壓縮為主的方法，因為隨著可修正的個數增加，其壓縮率也降低。而我們提出使用，同時將時間與空間壓縮的方法，以達到減少資料量與面積。此外，我們使用虛擬布林函數最佳化相關軟體，來輔助診斷發生錯誤的掃描鍊與位置。最後，使用離散機率等方法，來計算分析診斷錯誤之機率，可以快速求得混疊機率大小。對於五十萬個掃描單元的設計中，當t=10的實驗結果顯示，其壓縮率大於2600倍，所需的面積比以時間壓縮為主之壓縮器少60%。

This thesis presents a test response compaction and diagnosis technique based on BCH error correction code. Traditional time domain BCH compaction is not very useful in practice because the area overhead is too large. On the other hand, space domain BCH compaction does not have sufficient compression ratio to support multiple error diagnosis. This technique shares the polynomials among scan chains to reduce the area overhead. A Boolean satisfiability optimizer is employed to diagnose the failing flop-flops. In addition, a mathematical model is proposed to analytically estimate the aliasing probability and to aid the design of compaction circuitry. Simulation results show that, for designs of 500K flip-flops, our compression ratio is more than 2,600, which is much higher than that of space domain compaction. The area overhead is less than 60% of that of time domain BCH compaction.