多輸入多輸出正交分頻多工與單載波分頻多工存取系統之同步與通道響應之同估測

Abstract

本論文首先針對長程演進技術(Long Term Evolution, LTE)的特點和所採用的技術作深入淺出的介紹。為了提高頻譜使用效率、提高系統效能表現和降低系統成本，長程演進技術在上行採用了單載波分頻多工存取技術，於下行採用了正交分頻多工技術。根據長程演進技術和ITU-R所制定的標準，我們考慮在LTE系統中的通道估測和三項前端電路不理想效應，他們分別是載波頻率偏移、取樣時脈偏移、實虛部不平衡和通道響應。以系統設計的觀點出發，我們完整地設計一套同步和通道響應同估測的程序。根據這套程序，我們分別在上行和下行提出了三套有效率的同步和通道同估測演算法。 我們提出的第一個演算法為針對多輸入多輸出下行通道中，基於時域的延展性卡門濾波器(EKF)所發展出的載波頻率偏移、取樣時脈偏移和通道響應同估測之演算法。有別於在以往的同估測作法是在頻域進行估測且殘餘頻率偏移所導致的載波間干擾(ICI)被忽略，我們所提出的方法是基於時域且不忽略載波間干擾。根據系統模擬結果，我們所提出的時域延展性卡門濾波器估測演算法在中度至高度之信噪比情況之下，其表現明顯優於以往在頻域採用遞迴性最小方差(RLS)的演算法。我們所提出的演算法明顯的消除了由於殘餘頻率偏移所造成在均方誤差(MSE)表現的錯誤地板(error floor)現象。 接下來，我們提出的第二個演算法為針對上行通道中，低複雜度的載波頻率偏移和通道響應同估測演算法。傳統的做法最大的缺點在於複雜度太高，而我們所提出的演算法是基於循環轉移Zadoff-Chu序列在時域的良好性質和改良之時域平行干擾消除(PIC)所發展出來的簡單且低複雜度的載波頻率和通道響應估測演算法。藉由反轉矩陣計算的避免，在通道估測所需要的複雜度大大地降低。此外，我們根據所提出的演算法而推導出了在多用戶干擾之下的通道估測的均方誤差下界。根據模擬結果，我們所提出的演算法比傳統方法的收斂速度要快，且收斂結果十分逼近我們所推導出來的通道估測之均方誤差下界。 最後，根據我們之前所提出的方法，在上行將實虛部非平衡的效應列入考慮。我們提出了針對單載波分頻多工存取系統的載波頻率偏移、實虛部非平衡和通道響應同估測演算法。在多用戶干擾方面我們採用了基於時域的改良式平行干擾消除方法並且在通道估測迴路中引入了簡化且低複雜度的卡門濾波器(KF)估測演算法。從模擬結果得知，我們所提出的演算法比前人所提出的演算法更能對抗高度頻率選擇性(frequency selective)通道。且通道之估測均方誤差能在在少數遞迴次數中收斂至推導之通道均方誤差下界。

In this thesis, the features and techniques of LTE system are first illustrated. In order to achieve high frequency efficiency, high system performance and lower cost, LTE adopt SC-FDMA as its uplink transmission technique and OFDMA as its downlink transmission scheme. Based on the specification of LTE and channel model specified by ITU-R, we consider three front-end non-idealities including CFO, SCO and IQ imbalance as well as channel estimation in LTE system. We propose an efficient synchronization and channel estimation procedure in the system design point of view. Based on the proposed procedure, CFO, SCO, IQ imbalance and channel impulse response are scheduled to be jointly and efficiently estimated by our proposed estimation method both in downlink and uplink, respectively. First we propose a time domain extended Kalman filtering based joint carrier frequency offset, sampling clock offset and channel estimation algorithm in MIMO OFDMA downlink. Unlike previous works, where the frequency induced ICI term is neglect and the algorithm is developed in the frequency domain, our method is based on time domain approach without ignore the ICI term. From the simulation result, our proposed time domain EKF approach is robust in medium to high SNR values compared with conventional frequency domain RLS method, where the error floor occurs due to residual frequency offsets and ICI effect. Second, we propose a low complexity joint estimator of carrier frequency offset and channel impulse response for SC-FDMA uplink. Unlike conventional method, which is computational exhausting, our proposed scheme is based on cyclic shifted Zadoff-Chu sequences and modified time domain parallel interference cancellation scheme, which provide simple and low-complexity solution for both CFO and channel estimation. The computational complexity of channel estimation is greatly reduced since no matrix inversion is involved. Furthermore, a multiuser interfered channel estimation MSE lower bound is derived. Simulation results show that our proposed algorithm is more robust when CFO is large, has faster convergence rate and achieves the derived lower bound in acceptable iteration numbers. Finally, based on the proposed algorithm, we take transmitter IQ imbalance into consideration and propose a joint estimator of CFO, IQ imbalance and channel impulse response for SC-FDMA in uplink. We make use of the modified time domain parallel interference cancellation for MUI cancellation and of the simplified low complexity Kalman filtering for channel estimation. Simulation results show that our proposed algorithm is very robust in highly frequency selective channel over previous work. The channel MSE is close to the derived lower bound in a few iterations.