LTE 及LTE-Advanced 上行通道估計之研究

Abstract

無線通訊在未來通訊的發展佔了很重要的地位，其中更以WiMAX(IEEE)跟 LTE(3GPP)這兩個分支為基礎個別發展；雖然大致上這兩個系統十分相似，但最 主要的差異就在於LTE 在上傳採用的是單載波分頻多工存取(SC-FDMA)的技術。 有別於WiMAX 所使用的正交分頻多工存取(OFDMA)，其優點在於能降低峰均 值比(PAPR)而使手機端(MS)的電池壽命更久、更省電，因而延長使用時間。本 篇論文介紹LTE 和LTE-A 單載波分頻多工存取(SC-FDMA)中，多通道傳輸通道 估計的問題、演算法、分析及模擬等議題，並延伸多輸入多輸出(MIMO)通道。 在通道估測中，我們分為單輸入單輸出(SISO)和多輸入多輸出的討論。首先 在單輸入單輸出通道中，我們先採用最小平方差(least square)的估計器，然後再 利用兩種方法在頻域去估測出載波的頻率響應並比較其效能。第一種方法是線性 最小均方根誤差(LMMSE)，利用最小平方差估計器所求得的參考訊號(reference signal)上載波頻率響應，乘上線性最小均方根誤差矩陣來使參考訊號更我平滑， 而線性最小均方跟誤差矩陣基於通道相關矩陣。第二種方法則是利用高斯分布當 作權重來使參考訊號平滑，我們稱為高斯窗口(GWD)。接下來兩種方式皆是用線 性內差的方式來得到資料載波上的頻域響應。在多輸入多輸出的系統中，我們利 用參考訊號的循環位移(cyclic shift)，在頻域和時域中分開多根天線所產生的資料 重疊。分開通道後，分別對各個通道做通道估計。 在模擬當中，我們先在加成性白高斯雜訊(AWGN)通道上驗證我們的模擬模 型，然後在放置於多重路徑的通道上模擬，首先結果發現在低訊號雜訊比(SNR) 的情況下，線性最小均方根誤差估計器的效能與高斯窗口的效能差不多，但高斯 窗口的複雜度會比線性最小均方根誤差來的低。再來我們觀察到在頻域做通道分 開比在時域做通道分開來的精準。 ii 我們研究的貢獻主要是降低單輸入單輸出的通道估計的複雜度，還有增加再 多輸入多輸出通道分開的準確度。

Wireless communication will play an important role in the evolution of communication in the future, especially basing on WiMAX (IEEE) and LTE (3GPP) to develop individually. Although the two systems are very similar, the biggest difference of them is that LTE using single carrier frequency division multiple access (SC-FDMA) technique in uplink transmission. The advantage of LTE is to decrease peak-to-average power ratio (PAPR) which saves more power of the batteries of user equipments (UEs) and efficiently extends the using time, rather than using OFDMA (orthogonal frequency division multiple access) which is adopted by WiMAX. This thesis will introduce the subjects of channel estimation problems, algorithms, analysis of multi-path transmission in SC-FDMA, and extend to the multi-input multi-output (MIMO) channel. In channel estimation, we divided into the discussion of the single-input single-output (SISO) and MIMO. In single-input single-output system, first we use least square estimator, then use two different methods to estimate the channel response in frequency domain and compare the performance between them. The first method is linear minimum-mean square error (LMMSE). We use reference signal channel response which estimates by least square estimator multiplied the linear minimum-mean square error matrix to make more smooth, and the linear minimum-mean square error matrix base on the channel correlation matrix. The second method is using Gaussian distribution window (GWD) to make reference signal more smooth. And then the two methods all use linear interpolation to get the channel response of data carriers. In multiple-input multiple-output system, we use cyclic shift of reference signal to separate channel form different transmit antenna in time domain and frequency domain. After channel separation, channel estimation for each channel. In simulation, we test and verify the simulate model which we proposed, in additive white Gaussian noise. Then we simulate in multi-path channel. First we find that LMMSE has similar performance as Gaussian distribution window in symbol error rate and mean square error, but Gaussian distribution window has lower complexity than LMMSE. Second we observe that frequency domain channel separation is better than time domain channel separation. Our contribution to the research is mainly to reduce the complexity of the SISO channel estimation, as well as to increase the accuracy of the MIMO channel separation.