在下一代第五代通訊系統中，巨量多輸入多輸出系統(Massive MIMO system)被認為是候選技術之一。隨著下一代通訊系統中基地台與使用者接收裝置能使用的天線數量大幅提升，巨量多輸入多輸出系統相對於傳統多輸入多輸出系統必須承受更高的運算複雜度。而為因應巨量多輸入多輸出系統增加的運算複雜度，多種前編碼(precoding)技術相應而生。論文中主要在探討巨量多輸入多輸出前編碼/波束成形系統，提出用於追蹤時變通道的快速收斂奇異值(singular value decomposition, SVD)分解法。由於在這些系統中只選擇了具較強空間增益的子通道進行傳輸，所以我們的SVD演算法利用了部分分解和時間相關的特性。此外在本論文演算法中所提出的自我調整逆冪次演算法(Self-Adjusting Inverse Power Method, SA-IPM)可以透過在每次迭代期間根據中間結果調整來實現快速收斂，而平行處理可以提升高產出率(throughput)實現。與具有超線性收斂的自冪次(Self-Power Method, SPM)相比，本論文所提出之自我調整逆冪次演算法(SA-IPM)具有更好的收斂性和更低的複雜度。而良好的通道追蹤能力也被證明。 在硬體設計方面，則以支援到10×10的矩陣做QR分解為考量，並可以支援2×2~10×10的矩陣維度，資料流格式則採用外部浮點數內部定點數形式來表示，以函蓋10×10通道矩陣的分部範圍，並使用座標軸旋轉數位計數器(Coordinate Rotation Digital Computer, CORDIC)來實現Givens rotation的運算，由五個CORDIC組成的脈動陣列(systolic array, SA)SA1和七個CORDIC組成的脈動陣列SA2來完成QR分解，而SA1和SA2內部都有管線化設計。完成10×10QR分解中的上三角矩陣R需要141個時脈數，單一矩陣Q需要154個時脈數，透過TSMC 40製程，最高時脈操作頻率來到110MHz以上。 ;Massive MIMO (multiple-input multiple-output) technique is considered to be one of the promising solution in the 5th generation wireless communication system. With the increase in the number of antennas that can be used by base stations and user devices in next-generation communication systems, massive MIMO systems have higher complexity than conventional MIMO systems. To reduce the increased complexity of the massive MIMO system, a variety of precoding techniques are developed. In this thesis, a fast- fast-convergence singular value decomposition (SVD) algorithm is developed for tracking time-varying channels in massive MIMO precoding/beamforming systems. Since only strong channel gain (singular value) are selected for data transmission in these systems, our SVD algorithm exploits the properties of partial decomposition and temporal correlation. Besides, the proposed self-adjusting inverse power method can achieve fast convergence by modifying the shift according to the intermediate result during each iteration. Thus, parallel processing is possible to facilitate high-throughput implementation. Compared to the self-power method with super linear convergence, the self-adjusting inverse power method has better convergence and lower complexity. Good channel tracking capability is also demonstrated. In QR decomposition hardware design which is supported for a matrix of 10×10, and it can support a matrix size form 2×2 to 10×10. The data stream format is expressed in the form of an external floating-point internal fixed-point number. The Givens rotation is realized by Coordinate Rotation Digital Computer (CORDIC). The QR decomposition is accomplished by systolic array (SA) one and systolic array two which are consisted by five CORDIC and seven CORDIC. The upper triangular matrix R in the 10×10 QR decomposition needs 141 clocks, and the unitary matrix Q needs 154 clocks. Through the TSMC 40nm process, the highest clock operating frequency reaches over 110 MHz.
