SISO/MIMO BICM系統之LLR非線性量化

Abstract

眾所周知，軟式解碼(soft decoding)比硬式解碼(hard decoding)能夠得到更好的效能，然而軟式解碼的複雜度相對的也高很多，在一般通訊系統□軟位元(soft bit)通常是以對數近似比(log likelihood ratio; LLR)來代表，在進入解碼器之前LLR必須被量化，量化的位元數直接影響解碼器的複雜度。又在混合自動重傳的機制□(Hybrid Automatic Retransmission Request; HARQ)，舊的LLR值必須儲存以便與未來新的LLR作合併，LLR的位元數直接的影響到所需記憶體的容量。因此如何用最少的位元數達到效能的要求是一重要的研究議題。本論文採用最大互消息量(mutual information; MI)的原則對LLR做最佳的量化，主要的貢獻有三(1)考慮高QAM調變之LLR量化，(2)考慮多輸入多輸出(multiple input multiple output; MIMO)的系統之LLR量化，(3)考慮LLR量化在HARQ上之應用。模擬結果顯示，最大互消息量量化可以有效的提升系統效能，或降低所需之LLR位元數。

As well known, soft decoding can have better performance than hard decoding. However, the computational complexity of soft decoding can be much higher. In communication systems, soft-bit information is usually represented by the log likelihood ratio (LLR). Before entering the decoder, the LLR values must be quantized. The numbers of quantized bits directly influence the decoder’s complexity. Also, in the hybrid automatic retransmission request (HARQ), the old LLR values must be stored to be combined with the new LLR values. The numbers of quantized bits directly impact the memory size. Therefore, how to reduce the number of bits in LLR values is an important research topic. In this thesis, we use the principle of mutual information maximization to obtain the optimum quantization of the LLR. We have three contributions. First, we consider the LLR quantization in high QAM systems. Second, we consider the LLR quantization in multiple-input-multiple-output systems. Third, we consider the LLR quantization in HARQ systems. Simulations show that the LLR quantization with the quantization method can significantly enhance the system performance, or reduce the number of required bits.