多輸入多輸出無線通訊系統中等量增益波束成形之設計

Abstract

多輸入多輸出(Multiple-Input Multiple-Output, MIMO)技術可以獲得較高的頻譜效益以及改善對抗通道衰落的強固性，有效提升系統之傳輸鏈結品質。因此，在更高維度及更精緻的信號調變技術之外，多輸入多輸出 (MIMO)技術逐漸被越來越多無線通信系統所採用，以能夠更有效地使用珍貴的頻譜資源。在多輸入多輸出的技術中，相較於開迴路(open-loop)多輸入多輸出技術，閉迴路(closed-loop)技術由於可以在傳送端獲得通道狀態資訊(channel state information at the transmitter, CSIT)，因此可以提供額外的前置編碼增益。然而，要將完整的通道狀態資訊回饋至傳送端，其龐大的回饋資料在大部份的無線系統中，其頻寬有限的回饋通道顯然難以負荷，因此顯得這樣的方式不符實際需要。基於這樣的缺點，啟發了有限回饋前編碼技術的研究，其中量化的前編碼矩陣由接收機傳送之傳送機。 在本論文中，即針對有限回饋前編碼技術，提出了以下之設計: 1. 首先，我們利用純量量化設計了等量增益前置編碼器，此方法不需預先定義編碼簿。在此編碼器中，位元配置可被用來量化前編碼矩陣，以進一步改善系統效能。在先前技術中，有兩種傳統的位元配置機制:一種是均勻的(uniform)位元配置機制，另一種是最佳的(optimal)位元配置機制藉由竭盡的搜尋而得。均勻位元配置法具有簡單的特性，但是會導致明顯的效能衰落。另一方面，竭盡搜尋的方法雖然可以的到最佳的效能，但是極度高的運算複雜度導致在實際系統實現的不可行性。在此設計中，我們採用純量量化的等量增益前編碼器提出兩種位元配置機制，此兩種機制皆具有低複雜度，相較於使用竭盡搜尋法的效能最佳機制，能有效降低複雜度由回饋數值B的指數次方降低至線性次方。此外，提出的機制效能相當接近於效能最佳機制。在4傳輸1接收的通道環境中，假設回饋數值B為4，和效能最佳機制的位元錯誤率效能差距最小可達0.2dB。此外，隨著B上升，此差距會進一步縮小。最後，此兩種機制提供了在效能以及複雜度之間的良好權衡性。 2. 再來，針對純量量化等量增益傳輸前置編碼器，我們進一步探索通道的時間相關性，希望能藉由其中設計進一步增強量化的精準度。等量增益的優勢在於對於功率放大器的需求較低，可以降低系統的成本。此外，純量量化的優點在於可以快速編碼回饋資訊、不需預先定義編碼簿以節省儲存空間。於此設計中，我們提出了兩個基於純量量化的差分回饋機制。首先，此兩種機制皆具有低複雜度，相較於先前由英特爾提出利用旋轉進行差分回饋的標竿機制，他們節省了在接收端及發射端的旋轉矩陣建構及旋轉運算，這些皆牽涉到複雜且大量的複數矩陣運算。此外，由蒙地卡羅模擬顯示出於時間相關通道下:1)相較於傳統的一次性(one-shot) 波束成形向量機制，吾人所提出的低複雜度機制仍然能夠提供明顯的效能增強; 2)相較於無量化損失的等量增益傳輸向量，吾人所提出的機制的提供了可接受的性能劣化。在4傳輸2接收的通道環境中，他們的位元錯誤率效能和Grassmannian波束成形向量機制比較可以達到0.5 dB的增進，而和沒有經過量化的等量增益傳輸最佳效能比較，可以達到僅約0.15 dB的性能衰減；和英特爾提出的非等量增益傳輸差分回饋機制，可以達到僅約0.35 dB的性能衰減。

Multiple-input multiple-output (MIMO) techniques enable higher spectral efficiency and improve robustness against channel fading. Therefore, in addition to higher order and much sophisticated signal modulation techniques, MIMO technology has been adopted by more and more wireless systems to utilize the precious spectrum resource much effectively. Among MIMO techniques, the closed-loop MIMO can provide extra gain compared to the open-loop MIMO thanks to the provision of the channel state information at the transmitter (CSIT). However, the tremendous feedback amount of full CSIT is somewhat impractical in most wireless systems. This drawback motivates the research of limited feedback precoding techniques, where only quantized precoding vectors from the receiver is conveyed back to the transmitter. In this dissertation, for the limited feedback precoding techniques, we proposed the following designs: 1) We design equal gain precoders with scalar quantization which do not require the predefined codebook. In such precoders, bit allocation can be used to quantize the precoding vector/matrix to further improve the system performance. There are two conventional bit allocation schemes. One is the uniform bit allocation and the other is the optimal bit allocation obtained by using exhaustive search. The uniform bit allocation is simple but turns out to have obvious performance degradation. On the other hand, the exhaustive search method leads to the optimal performance. However, its computational complexity is extremely high and may be somewhat impractical to be realized. In this work, we propose two bit allocation schemes for the MIMO equal gain precoder with scalar quantization. The two proposed methods are both with low complexity and can effectively reduce the complexity to the linear order from the exponential order of feedback value B compared with the exhaustive search (optimal) method. Also, their performance is close to that of the exhaustive search (optimal) method. In a 4T1R (4-transmit and 1-receive) channel environment and feedback value B= 4 assumed, the least bit error probability (BEP) performance gap can achieve 0.2 dB. Moreover, as B becomes larger, the gap becomes smaller. Consequently, the proposed bit allocations provide good trade-off between performance and complexity. 2) We design the scalar quantized equal gain transmission (EGT) vectors adapted in the temporally correlated channels. The superiority of equal gain property lies in its lower requirement to the power amplifier (PA) and contributes to a low-cost system. Furthermore, the advantages of the scalar quantization for limited feedback systems include the faster encoding of feedback information in MISO channels and no need for predefined codebook with storage occupied. In this work, two differential feedback schemes based on the scalar quantization technique are proposed. At first, the two proposed schemes are both with low complexity, because they economize the use of rotation matrix constructions and rotation operations for quantization at the receiver and reconstruction at the transmitter compared with the prior benchmark work proposed by Intel using the rotation-based differential feedback scheme, which are involved in complicated and considerable complex matrix operations. Furthermore, Monte Carlo simulations are given to show that the proposed low-complexity schemes can still offer significant performance improvement compared with conventional one-shot beamforming schemes in the temporally correlated channels as well as acceptable performance degradation from the unquantized EGT vector. In a 4T2R (4-transmit and 2-receive) channel environment, their bit error probability (BEP) performance can achieve 0.5 dB improvement compared to Grassmannian (GS) beamforming scheme and around 0.15 dB degradation from that of the unquantized EGT vector and around 0.35 dB degradation from that of Intel’s scheme (non-EGT scheme).