無線通訊之訊號設計與接收技術

Abstract

在本論文中，我們研究高傳輸速率無線通訊的訊號設計及接收技術，期使能增進傳輸效能。我們提出數個改進的技術，其中二個是有關訊號設計方面，另外三個是有關接收技術。 首先，我們提出一種稱為最小範數訓練數列(min-norm training sequence)的數列給最少平方通道估計(least-square channel estimation)之用。使用這種訓練數列在最少平方通道估計中，當干擾為相加性的白色雜訊(AWGN)時，可以使通道估計誤差的均方值(mean-square error)達到最小。我們也證明了使用此種訓練數列所能達到的最小通道估計誤差的極限。但因為目前還沒有一個有效的方法來產生這種訓練數列，我們提出一種快速的搜尋方法來幫助尋找這一類的訓練數列。另外因為不同長度的訓練數列，會有不同的最小通道估計誤差極限，我們針對一些特殊長度的數列，證明該長度的訓練數列無法達到最低的通道估計誤差極限。 接著，我們提出一個縮短長度的通道估計方法。主要是因為無線通訊環境的通道長度會變化相當大，我們發現使用適當的通道長度來做通道估計，可以降低通道估計誤差。我們所提的縮短長度的通道估計方法，特別在低訊雜比(SNR)的環境下，可以有效降低通道估計誤差。同時我們也證明，若使用最小範數訓練數列，則使用與不使用縮短長度的最少平方通道估計結果是一樣的。 另外我們研究將訓練訊號與資料訊號的功率設為不同時的傳輸效能，特別是將前者的功率設為稍大，將後者的功率設為稍小，並使總傳送能量固定。我們推導一個二者最佳功率比的公式。一般在通訊系統中，訓練訊號與資料訊號的功率都是一樣的，但經由使用不同的功率，從電腦模擬的結果，我們發現可以達到較佳的傳輸效能，當然是在總傳送能量固定之下。也就是我們可以將傳送的能量做更有效的運用。 還有，根據輸入訊號的訊雜比，我們提出一個方法，用來選擇維特比等化器(Viterbi equalizer)所用的通道長度。同樣因為無線通訊環境的通道長度會變化相當大，以及維特比等化器的計算複雜度會隨著所用通道長度而呈指數性的增加，這種簡單的選擇方法，可以有效的選擇維特比等化器所用的通道長度。由電腦模擬的結果發現，在相同的傳輸位元錯誤率之下，大約可節省25%至10%的計算量。 最後，我們提出一個新的時空域信號處理的架構，用來消除同頻干擾(CCI)及相加性的白色雜訊。近來時空域信號處理常被研究用來消除同頻干擾，但是我們發現傳統的時空域信號處理架構在同頻干擾與相加性的白色雜訊同時存在時，無法提供很強健(robust)的訊號處理能力，特別是在同頻干擾的功率不會比相加性的白色雜訊大很多時。我們所提出的方法，在不同的同頻干擾與白色雜訊功率比之下，有很強健的表現。 我們使用蒙地卡羅(Monte Carlo)模擬來評估所提出方法的效能增益。另外我們也對適用於時空域訊號處理的通道模型做一簡短的介紹。我們也發現有二個原先並不是設計作時空域訊號處理的通道模型，經由稍微變化使用方法後，也可以當時空域訊號處理的通道模型。而且經由使用適當的參數，這二個通道模型可以產生相似統計特性的通道脈衝響應。

We consider signaling and receiving techniques for improved transmission performance at high baudrates for wireless communication. Several techniques are proposed, of which two concern signal design and three concern receiver design. First, a special kind of sequences called the min-norm training sequences are proposed for least-square channel estimation. By using one such sequence as training sequence, the mean-square channel estimation error can be minimized when least-square channel estimation are used in AWGN environments. We also prove the lower bound of the mean-square channel estimation error that can be achieved by applying this kind of sequences. Since no simple constructive methods exist for obtaining the min-norm training sequences, we propose a search approach to find sequences having nearly the same properties as the min-norm sequences. Because the lower bounds of different length min-norm sequences are not the same, we also prove the existence properties of some special length of sequence. Secondly, a channel estimation method using reduced channel length is proposed. Because the channel length varies greatly in wireless communication environments, we find that estimation error can be significantly reduced if proper channel length is used in channel estimation. The method that we propose can reduce the channel estimation error, especially in low SNR environments. Also, we prove that when the training sequence is min-norm sequence with unity normalized error norm, the results of the reduced-length channel estimation is the same with that of non-reduced length channel estimation. Thirdly, we consider using unequal power levels for the training signal and the data signal, with a higher power level for the former. We derive a mathematical expression for the optimal power ratio. In general transmission systems, the power level for the training signal and the data signal are the same. By using different power levels for training and data signals, we can achieve a better performance with total transmitted power being constant. That is, we can achieve better transmission power efficiency. Fourthly, a channel taps number selection method for Viterbi equalizer based on known input SNR is proposed. Since the channel length varies greatly in wireless communication environments and the computation complexity of Viterbi equalizer is exponentially increased as the taps number is increased, this method can provide a simple and efficient selection for the Viterbi equalizer. The resulted performance is the same with the fixed taps number method by using this selection method, but this method achieves 25% to 10% computation power reduction. Fifthly and lastly, a new space-time signal processing structure is proposed for reducing both the CCI and AWGN effect. Space-time signal processing has received much interest in wireless communication for its superiority on canceling the co-channel interference (CCI). But we find that the traditional combining structure does not perform robustly when both AWGN and CCI exist, especially when the interference-to-noise (INR) is not very high. The proposed space-time signal processing structure performs more robustly than the traditional structure for different INR environments. The above performance evaluation has been effected by small-scale Monte Carlo simulations. We also give a brief review of the wideband spatial channel models and extend two non-spatial channel models to spatial usage. By suitable choice of parameters, the two channel models can generate channel impulse responses of similar statistics.