Summary: | 碩士 === 國立清華大學 === 電機工程學系 === 97 === Due to the growing demands of high-dimension multiple-input multiple-output (MIMO)
systems, the preprocessing of sorted-QR decomposition becomes one of the computational
bottlenecks in the QR-based MIMO detection, and the power consumption and
computation complexity of the MIMO detector are growing dramatically. This thesis
proposes an 8 × 8 MIMO detector which can be divided into two parts. First, the
proposed Givens-Rotation-based algorithm aims to improve the throughput and the
hardware utilization efficiency by stopping the sorting operations earlier or relaxing the
sorting condition for the parallel sorted-QR decomposition. The detection performance
degradation is negligible and better hardware efficiency can be obtained for the larger
number of the MIMO antennas. Then, we propose an 8 × 8 mixed K-best/QR-SIC detector
with the proposed modified sorted-QR decomposition as the preprocessing. The
modified sorted-QR decomposition is used to relax the sorting condition to improve the
latency and the hardware utilization rate. Compared to the K-best detector, the proposed
mixed K-best/QR-SIC detector can achieve lower computation complexity with
almost no performance loss in the high-dimension MIMO detection.
|