運用改良式預先排序搜尋達成快速判斷正交分頻多工系統之最佳字元長度

Abstract

本篇論文提出兩種改良式預先排序搜尋法，藉以快速判斷OFDM系統之定點數字元長度，並提出預先排定搜尋序列的新法則以減少模擬次數及硬體複雜度。其一稱為「Type-1 Modified Preplanned Search」（T1-MPS），在forward iteration過程中，每次取序列中前兩個變數同時加1位元，讓系統的BER快速地達到指定的位元錯誤率；並且在backward iteration過程中減少字元長度以減少硬體複雜度。另一個方法稱為「Type-2 Modified Preplanned Search」（T2-MPS），此方法從預先排定的序列中取出前三個變數，個別加1位元形成三個新的字元長度組合，並選取輸出位元錯誤率最小的組合作為下一輪的基點。實驗平台是以SystemC函式庫設計正交分頻多工系統之接收器，實驗結果顯示，本論文所提出T1-MPS搜尋方法之模擬次數比其他搜尋方法之模擬次數少，與傳統預先搜尋方法相比，模擬次數減少約35%，硬體複雜度平均減少約1.9%；而T2-MPS搜尋方法與CDM搜尋方法相比，T2-MPS搜尋方法之模擬次數可減少60%以上，硬體複雜度相較於CDM搜尋方法則增加約0.2%。

Two modified preplanned search methods to determine the fixed-point wordlength for orthogonal frequency division multiplexing (OFDM) system are presented. A new search sequence determination criterion is utilized to trade-off the number of simulations and hardware complexity. The proposed Type-1 Modified Preplanned Search (T1-MPS) adds 1 bit to first two elements simultaneously in forward iteration to meet the bit error rate (BER) constraint, and decreases the wordlength in the backward iteration to reduce the hardware complexity. The other one is Type-2 Modified Preplanned Search (T2-MPS) that adds 1 bit respectively to the first three variables in the searching sequence to form three new wordlength combinations and then chooses one combination with the smallest BER for next iteration. Applying the proposed methods to the OFDM receiver, the experimental results show that the proposed T1-MPS method has the lowest number of simulations among the existing methods and the T2-MPS method can keep the lower hardware complexity. Compared with preplanned search method, the T1-MPS method reduces 35% trials and the hardware complexity is reduced about by 1.9%. Compared with CDM search method, the T2-MPS method can reduce more than 60% trials with the penalty of increased hardware complexity about 0.2%.