本論文提出一新型類比式停止疊代正規化對數和積LDPC解碼晶片設計。在低密度同位元校驗碼中以和積演算法、對數和積演算法及最小和演算法最為常見,在考量解碼效能及硬體複雜度後,本論文採用對數和積演算法來設計電路並以電流模式來實現,同時提出了一新型正規化類比解碼器,藉由正規化的調整能有效提高對數和積演算法解碼效能;在解碼速度上,亦增加了停止疊代判斷機制來提高其吞吐量。 最後本論文所提之新型類比式正規化對數和積解碼晶片,以TSMC 0.18μm 1P6M CMOS製程加以驗證。第一顆晶片包含了2436顆電晶體,晶片面積不含I/O PAD為0.055mm2,當資料吞吐量為11.85Mbps時,消耗功率為0.78mW。第二顆晶片新增停止疊代架構,共包含了2694顆電晶體,晶片面積不含I/O PAD為0.078mm2,當資料吞吐量為62Mbps時,消耗功率為0.85mW。實驗結果顯示,本論文所設計之解碼架構,不僅具有低複雜度,更有助於提升電路解碼速度,同時也提高了其解碼效能,可供未來通訊系統及SoC之運用。
In this thesis, we propose a new analog normalized log sum-product LDPC decoder with the stopping iteration method. The common LDPC decoding algorithms are the sum-product, log sum-product algorithm and the min-sum algorithm. For decoding performance and circuit complexity consideration, the log sum-product algorithm with current-mode is used in our design. Moreover, a new normalization method is proposed to improve log sum-product algorithm error-correcting performance, and a stopping iteration circuit to increase the throughput. Finally, the proposed LDPC Log Sum-Product decoders are implemented with TSMC 0.18μm 1P6M CMOS technology. The first decoder core size is 0.055mm2 including 2436 transistor counts, and the throughput is 11.85Mb/s while power consuming 0.78mW. The core size of second decoder with stopping iteration is 0.078mm2 including 2694 transistor counts, and the throughput is 62Mb/s while power consuming 0.85mW. Experimental results show that the analog decoders have smaller area and higher throughput, which can be applied to future communication systems and SoC applications.