IEEE 802.16e OFDM上行及OFDMA下行同步技術與數位訊號處理器實現之研究

Abstract

本篇論文介紹IEEE 802.16e正交分頻多工(OFDM)和正交分頻多工存取(OFDMA)的同步。我們討論了他們同步的問題、演算法、以及實做方面的議題。 在正交分頻多工系統，我們首先設計一套同步系統用來解決符碼時間偏移和小數部分載波偏移的問題，並用浮點數運算來實做。符碼時間是利用固定不變的同步碼(preamble)來同步，而小數部分的載波偏移是計算靠循環字首(cyclic prefix)的相關性(correlation)來估計其值。我們同時在可加性白色高斯雜訊通道(AWGN)以及多路徑Rayleigh衰減通道下做模擬，模擬速度高達60 km/h，並觀察其結果。 其次，我們把這些方法修改成定點運算的版本，並在數位訊號處理平台上，最佳化我們的程式的速度。雖然修改成定點運算會使效能衰減，但其接果依然可以接受。最後我們系統的每一塊功能區塊(function block)都能達到即時處理(real time)的要求。 在正交分頻多工存取系統中，除了符碼時間偏移和小數部分載波偏移之外，尚有整數部分載波偏移、取樣頻率偏移、和同步碼索引(preamble index)等需要同步。因為使用者端事先無法知道確切的同步碼，所以我們把同步碼視為一般符碼，利用計算靠循環字首的相關性來估計時間偏移和小數載波偏移。我們同樣利用同步碼的特性來同步整數載波偏移和同步碼索引。小數載波偏移需要靠平均每個循環字首相關性的結果來得到較準的結果，而取樣頻率偏移可以和載波偏移一起同步，因為他們有一樣的錯誤比例。我們也提供比較準確的時間同步以改善同步碼索引同步的錯誤率。 就像正交分頻多工系統一樣，我們用浮點運算來實做這些方法，並同時在可加性白色高斯雜訊通道以及多路徑Rayleigh衰減通道下做模擬，不過這邊的模擬速度高達300 km/h。

This thesis introduces the synchronization of IEEE 802.16e OFDM and OFDMA system. We discuss their synchronization problems, algorithms, and implementation issues. In the OFDM system, we first design a synchronization system to overcome the problems of symbol timing offset and fractional carrier frequency offset (CFO), and implement them with floating-point. The symbol timing is synchronized by invariant preamble, and the fractional CFO is estimated by cyclic prefix (CP) correlation. We simulate our system in both AWGN and multipath Rayleigh fading channel, which the speed is as high as 60 km/h, and see the performance. Next, we modified these methods into fixed-point version, and then optimize the speed of our programs on the digital signal processor (DSP) platform. Although the performance is degraded because of fixed-point modification, the results still can be accepted. Finally the function blocks of our system can all reach the requirement of real time. In the OFDMA system, in addition to symbol timing offset and fractional CFO, it still has integer CFO, sampling frequency offset (SFO), and preamble index need to be synchronized. Since the SS does not know the preamble in advance, we view preamble as a regular symbol and estimate the symbol timing and fractional CFO by CP correlation. We also use the feature of preamble to estimate the integer CFO and identify the preamble index. The fractional CFO needs be estimated by averaging every CP correlation result for a more accurate result, and the SFO can be synchronized with fractional CFO synchronization because they have the same error ratio. We also afford a fine timing estimation to improve the error rate of the preamble index identification. Like in the OFDM system, we implement these methods in floating-point version, and simulate them in both AWGN and multipath Rayleigh fading channel, but the speed is as high as 300 km/h here.