考量數位家庭之服務模式具使用者多工(multiple access)，並兼顧各使用者所需之高品質影音服務的資料量，本論文利用回傳的通道資訊來研究多使用者多輸入輸出前編碼技術，目的是減少接收端的複雜度，盡可能將複雜度移到傳送端，讓接收端有低成本、小面積、低功耗的好處。且因802.11ac、WiMax和3GPP LTE-Advanced都支援8×8MIMO技術，在硬體實現上以8根天線為考量。 本論文中演算法部分，我們提出了低複雜度之多使用者多輸入輸出前編碼演算法(MU- MIMO precoding)，此演算法在支援之使用者數目愈多時，錯誤率之效能有愈好的表現，而此前編碼機制是利用QR分解和THP演算法為基礎，來消除使用者間與天線間之干擾源，再搭配Sorted QR分解和所提出的區塊能量分配(Proposed Bolck based PAL)機制，更進一步改善錯誤率之效能，藉由此能量分配機制，運算複雜度可大幅降低，而對系統性能卻只有些微影響。 在硬體實作方面則針對關鍵性模組—高吞吐量之8×8 Sorted QR分解—進行設計，最大特色為同時運算3個4×4 Sorted QR分解，並提供不同層數的排序功能。設計程序上，首先先評估QR分解演算法的運算量，且為了達到高傳輸率，在硬體實現上，選擇了Givens Rotation演算法，因其可藉由管線級CORDIC實現，同時，採用Systolic Array的硬體架構，來達到高吞吐量之目標，相較於先前的文獻，我們呈現了具有極佳之面積時間乘積之硬體效能之設計結果。In future services of digital home, transmission of high-quality audio and video data to multiple users is a necessity. Multi-user MIMO precoding schemes that utilize the feedback of the channel state information to reduce either the receiver complexity or to enhance the system performance become attractive. Thus, consumers can have the benefits of low cost, small area, low power but high quality in the receiver. Due to the fact that the next-generation wireless systems such as 802.11ac, WiMax, and 3GPP LTE-Advanced all support 8×8 MIMO configuration, we aim to offer the design, in algorithm and architecture, for 8 antennas. In this thesis, we propose a reduced-complexity multi-user MIMO precoding algorithm including sorted-QR decomposition, block-based power allocation, and THP algorithm. The proposed scheme outperforms other MU-MIMO precoding schemes when the system supposts more users. The sorted QRD and proposed power allocation are effective to improve the BER performance. In addition, our proposed power allocation further reduces complexity and has only slight BER degradation. Finally, the THP algorithm is adopted to cancel the interference in advance at transmitter. As to the hardware implementation, we propose a scalable design of high throughput 8×8 Sorted QRD. It can compute three sets of 4×4 Sorted QRD simultaneously. Various numbers of sorting layers are supported. The Givens Rotation algorithm of the QRD is adopted for its merit in pipelining with the CORDIC operation. Finally, a hardware-efficient design with good AT product is shown.
