| Summary: | 碩士 === 國立成功大學 === 電機工程學系 === 102 === The multiple-input multiple-output (MIMO) technique has been widely adopted in modern wireless communication systems to enhance the throughput rate. Among various decoding techniques, the turbo-MIMO systems are introduced to further improve the decoding efficiency between the detector and decoder. For MIMO detector design, the fixed-complexity sphere decoding (FSD) algorithm features low computational complexity with acceptable performance and is suitable for highly parallel architecture development. However, unreliable soft information provided from the output of FSD detector usually leads to performance degradation. Although the turbo-MIMO system can update this soft information with multiple iterations, the overall throughput rate will be reduced accordingly. Design of low-complexity detectors with reliable soft information output is thus essential in efficient turbo-MIMO systems.
In this thesis, we presented an efficient turbo-MIMO system to update the soft information between the detector and decoder. An extended tree search technique is developed to improve the initial soft information of the soft-MIMO detector. Furthermore, a candidate node selection scheme is proposed to reduce the memory requirement of log-likelihood ratio (LLR) computation. A strategy of iteration profile is also provided to enhance the throughput. Compared to the LSD algorithm, the proposed scheme can reduce the complexity of LLR computation and the number of overall searched nodes by about 94.68% and 61.57% respectively, in 4×4 turbo-MIMO systems with 16-QAM modulation.
|
|---|
