臺灣大學: 電子工程學研究所吳安宇陳彥良Chen, Yen-LiangYen-LiangChen2013-04-102018-07-102013-04-102018-07-102011http://ntur.lib.ntu.edu.tw//handle/246246/256784從消息理論已知當傳送端獲得部分通道資訊，可利用預前編碼（Precoding）技術讓整體系統效能得到大幅提升，其好處包括可以有效對抗通道不良效應、降低接收端複雜度、提升系統吞吐量以及有效率分配多個使用者的通信資源…等，因此，預前編碼技術在很多前瞻通訊系統中扮演相當重要的角色，如現今的IEEE 802.3an、802.11n、802.16e/m以及3GPP-LTE都已經採用此類的設計，並將之納入其標準之中。針對不同的通訊系統，由於通道環境的不同，其預前編碼技術的需求也有所不同，因此本論文針對目前前瞻的有線及無線通訊系統，各自發展出所需要的預前編碼演算法及硬體架構設計。以下針對三種通訊環境作探討： 第一種是有線通訊環境，以10GBASE-T（IEEE 802.3an）乙太網路為主，所採用的預前編碼技術為湯林森-何洛緒瑪預前編碼（Tomlinson-Harashima Precoding），由於10GBASE-T所要求的百萬位元傳輸量，因此我們需要設計出高速的湯林森-何洛緒瑪預前編碼器，然而預前編碼器包含了非線性的迴饋電路，因此限制了它們在高速應用上的發展。所以本論文第一部分提出一個高速化方法，並且發展出泛用性的預前編碼器架構，在給定一個已知的設計規格下，所提出的方法可以在硬體複雜度和輸出的動態範圍之間做一個取捨，從而找出一個符合設計標準的近似最佳解。因此，此方法提供了更高的自由度，可在高速的預前編碼器作設計上的取捨。 第二種是室內無線通訊環境，以IEEE 802.11n為主，所採用的預前編碼技術為奇異值分解（Singular Value Decomposition），在接收端做完通道估測後，對所得到的通道矩陣進行奇異值分解，將一部分的分解結果回饋至傳送端，使傳送端可以使用最佳的預前編碼器來傳送資料，而另一部分的分解結果直接用於接收端做為解碼器，主要挑戰在於傳統的奇異值分解方法在運算複雜度及運算速度上都有很大的限制，因此本論文第二部分提出一套完整的可適性奇異值分解演算法及硬體架構設計，其所具備主要特色包含快速分解、有效提升硬體利用率、以及支援所有11n定義的傳收天線配對情形。 第三種是室外無線通訊環境，以802.16e/m及3GPP-LTE為主，所採用的預前編碼技術為碼簿搜尋（codebook searching），在行動通訊系統中，通道會快速變化，若是太慢回饋通道資訊，此資訊便不適用於當時的通道環境，而碼簿機制雖然可以節省回饋的資訊量，但是我們需要搜尋碼簿裡最佳的預前編碼器作為回饋的指標，此動作會大大的增加接收端的運算複雜度，因此本論文第三部分提出一系統化方法，可以在不失系統效能的情況下，有效減少碼簿搜尋範圍，並且大幅降低接收端的運算複雜度。 因此在本論文中，針對三種不同的通訊環境，我們提出相對應的高效能預前編碼技術，包含高速湯林森-何洛緒瑪預前編碼器、可重組態奇異值分解引擎、以及低複雜度碼簿搜尋機制，並且期望所提出的預前編碼技術也能夠應用在未來前瞻通訊系統之中。When the transmitter knows the channel information, precoding is an effective way to overcome channel effect and achieve high throughput transmission. With this desirable feature, precoding techniques have been adopted by several advanced communication standards, such as IEEE 802.3an, IEEE 802.11n, IEEE 802.16e/m, and 3GPP-LTE. Depending on the type of channel information and how fast the channel changes, the required precoding techniques are different for different communication standards. The precoding techniques are Tomlinson- Harashima precoding, singular value decomposition, and codebook for wireline systems, WLAN systems, and WMAN/WWAN systems, respectively. The design challenges of each precoding technique are described as follows. In the application of the IEEE 802.3an systems, the Tomlinson-Harashima (TH) precoders need to operate at a speed of 800MHz. However, the speed requirement is hard to meet since the TH precoders contain modulo devices and feedback loops. Therefore, we propose a two-time pipelining scheme to pipeline the TH precoders, which enables us to develop a generalized TH precoder architecture. The proposed scheme provides more degrees of freedom in designing high-speed pipelining TH precoders with build-in arbitrary speedup factors. In the IEEE 802.11n systems, the use of the singular value composition (SVD) technique can greatly enhance the system throughput. However, the high computational complexity and high decomposing latency are the important issues in applying the SVD to the real-time applications. Hence, we propose a complete adaptive SVD algorithm, as well as a reconfigurable architecture design for high-throughput wireless systems. The main design features include low decomposing latency and supporting all antenna configurations in a multi-input multi-output (MIMO) system. The proposed design is implemented in 90nm technology for the application of IEEE 802.11n systems. The chip result shows that for an 802.11n system, the average latency of our SVD engine is only 0.33% of the WLAN coherence time. Therefore, the proposed SVD engine is very suitable for the high-throughput WLAN applications. Due to the feedback delay constraint and limited feedback bandwidth, codebook-based precoding is a promising practical method in the WMAN/WWAN applications. According to the current channel condition, the receiver selects the optimal precoder from a codebook which consists of a finite set of precoders and sends the index of the chosen precoder over a limited-feedback channel. Conventional precoder selection criteria require high computational complexity. Besides, if the codebook is square, some precoder selection criteria are not feasible. Therefore, we propose a low-complexity precoder selection criterion which is applicable to any existing codebook. Compared with direct implementation, the proposed scheme has significant computational complexity reduction without performance loss. In summary, we propose three high-performance precoding techniques with their own design considerations for the advanced communication applications in this thesis. We expect that the proposed precoding techniques can also be applied to future advanced communication systems.1840733 bytesapplication/pdfen-US預前編碼湯林森-何洛緒瑪預前編碼奇異值分解碼簿多重輸出輸入正交分頻多工百億位元乙太網路precodingtomlinson-harashima precodingsingular value decompositioncodebookMIMOOFDM10GBASE-Tthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/256784/1/ntu-100-F94943010-1.pdf