基於離散傅立葉轉換技術之LTE上行通道估計

Abstract

隨著日新月異的高科技產品陸續發明，如智慧型手機和平板電腦等。提供使用者們更多的資料處理量也變得更加重要。而其中一個由3GPP發展的通訊協定LTE逐漸地成為主流。在LTE系統下，為了更能延長手機電量使用時間，一個藉由降低峰均值比(PAPR)的傳送技術單載波分頻多工存取(SC-FDMA)被應用在上行通道中。而本篇論文即在討論在單載波分頻多工存取下對通道估測的一些問題，演算法，和其分析及模擬。 在通道估測中，我們首先使用最小平方差(least square)的估計器，然後接著使用多種基於離散傅立葉轉換技術(DFT-based)的估計器來估測。基於離散傅立葉轉換技術的估計器是藉由轉換通道頻率響應(CFR)至一變換域，切除掉雜訊成分較多的區塊後再將這在變換域的類似通道脈衝響應給轉換回頻域。這些我們使用的基於離散傅立葉轉換技術的估計器有基於離散傅立葉轉換傳統型估計器、基於離散傅立葉轉換局部型估計器、基於離散餘弦轉換傳統型估計器、基於離散餘弦轉換局部型估計器，和基於離散傅立葉轉換對稱延展型估計器。其中傳統型和局部型的差別分別是其轉換是在一包含頻域防護帶(guard band)的全頻域，或是只在一由傳送訊號頻帶組成的局部頻域。而基於離散傅立葉和離散餘弦轉換分別是指用不同的方式轉換至變換域。基於離散傅立葉轉換對稱延展型估計器則是在轉化至變換域前用一鏡射複製的方式改變訊號特性以達到更加準確地估計通道。最後，我們增加了一個在變換域動態截切的方法當在消除雜訊區塊時以更加改進通道估計的效能。 在模擬中，我們先在AWGN通道上驗證模擬模型，後藉由在單根瑞利(Rayleigh)分布通道及多重路徑通道的比較和分析得知基於離散傅立葉轉換對稱延展型估計器和基於離散餘弦轉換局部型估計器有較好的效能。而動態截切的方法也額外提升了在MSE和SER上的效能。

With the invention of the smartphone, tablet PC and some other high-technique products, it becomes significant to offer users more data throughputs in wireless communication system. One of the main protocols, long term evolution (LTE), developed by 3GPP become more and more dominated nowadays. And to extend the capacity of the cell phone in LTE system, a new transmission technique single-carrier frequency-division multiple access (SC-FDMA) is adopted for the uplink (UL) by lowing the peak-to-average power ratio (PAPR) in transmission power. Our thesis mainly focuses on discussing the channel estimation in SC-FDMA including the subjects of problems, algorithms, analyses, and simulations. In channel estimation, we first use the least square (LS) estimator, and then use several DFT-based channel estimators in transformed domain. We do the DFT-based channel estimations by transforming the channel frequency response (CFR) into transformed domain, cutting off the region contained more noise components, and transforming this equivalent channel impulse response (CIR) back into the frequency domain. And these DFT-based methods we use are conventional DFT-based, partial DFT-based, conventional DCT-based, partial DCT-based, and symmetric extension DFT-based methods. The conventional ones mean that the transformations are done in the whole frequency domain contained the guard band rather than the partial frequency domain only composed of the signal bandwidth like the partial ones. And the DFT or DCT-based methods mean that we do the transformation by using DFT or DCT. The symmetric extension DFT-based one means that before transmitting the CFR into transformed domain, we mirror duplicate it first, then due to the property of signal we will get a more accurate estimation. Finally, we add the dynamic truncated method in transformed domain when we remove the noise part by symmetric DFT-based and DCT-based methods to get a further improvement. In simulation, we verify the simulated models in AWGN. Then through the comparison in single-path Rayleigh and multipath channel conditions, we can get the results that symmetric extension DFT-based and DCT-based have better performance than other methods in general, and with the dynamic truncated methods the symmetric extension DFT-based has an extra performance gain both in the MSE or SER.