應用於正交分頻多工系統之框碼同步方法及其效能評估

Abstract

在現今高速傳輸的通訊系統當中，正交分頻多工 (OFDM) 技術因其有效使用頻寬以及抵抗多重路徑干擾的特性而受到重視。 OFDM系統的主要問題在於時間與頻率的非同步效應會嚴重地降低系統的效能，所以能夠估測正確的時間與頻率偏移變得相當的重要，而本篇論文著重在探討框碼時間同步 (frame timing synchronization) 的問題。 我們研究了數種基於循環前置 (cyclic prefix) 的框碼同步技術，並提出了兩個以整體搜尋法 (global search algorithm) 為基礎所改進的演算法。 在局部搜尋法 (local search algorithm) 中，我們只搜尋前一個框碼所估測出之時間點附近的取樣點，如此可在高訊雜比的情況下增加框碼時序的同步準確率。 另外一個整體搜尋改進法 (modified global search algorithm) 是一個較低複雜度，但可產生與其他演算法相似效能的演算法。 本論文所提出的方法皆有自我學習的功能，也就是說，可以在緩慢變化的通道環境下自動追蹤時序。 經由模擬的結果，我們所提出的演算法將與其他基於循環前置的框碼同步方法在運算複雜度以及同步效能上的表現逐一比較。

Bandwidth efficiency and multipath immunity make orthogonal frequency division multiplexing (OFDM) an attractive modulation technique for modern high data rate communication systems. A major concern of OFDM systems is that symbol timing and frequency synchronization errors can seriously degrade the system performance. Therefore, it is important to obtain accurate estimate of time and frequency offsets. In this thesis, we focus on frame timing synchronization. Several cyclic prefix based techniques are investigated and two modified algorithms based on global search algorithm are proposed. In the local search algorithm, we only search the sample points near the estimated point of previous frame; this can increase the accuracy of estimated timing at high SNR. Another algorithm, called modified global search algorithm, yields a comparable performance and lower complexity compared with other techniques. Our proposed algorithms are all self-learning in the sense that they can automatically track slowly time varying channel conditions. The performance and complexity of our proposed algorithms will be compared with other cyclic prefix based algorithms from simulation results.