Low Complexity Implementation Of MIMO Sphere Decoding Algorithm For LTE System

The development of wireless communication systems is promoted by future wireless data service. Owing to Multiple-input multiple-output (MIMO) has met the growing requirements of wireless data service, it is a key technology for future wireless communication systems. Especially for spatial multiplexing system, it is possible to achieve the good spectrum resource utilization without being able to increase the bandwidth. However, since the receiver needs to separate the spatially multiplexed data streams, the higher spectrum resource utilization comes at the cost of increased computational complexity at the receiver. This requires MIMO detector must have lower complexity with good performance.This thesis has analyzed recognized MIMO detection algorithms for spatial multiplexing system. This theis has chosen depth-first sphere decoding algorithm as a starting point, then focused on the low complexity implementation of soft-output sphere decoding in LTE system.This thesis has analyzed some basic concepts, such as partial Euclidean distance, Schnorr-Euchner enumeration, radius constraints, admissible sets, single tree search, log likelihood ratio, sorted QR decomposition for soft-output sphere decoding. This thesis has proposed some methods to improve system performance. Since sphere decoding has variable instantaneous throughput, this thesis has introduced a constraints on the maximum complexity of the decoder. In order to get maximum complexity, this thesis has proposed statistical method to set run-time constraints. For the performance loss of the run-time constraints case, this thesis has proposed a series of methods such as run-time constraints with weighting, run-time constraints with adjustable weighting, run-time constraints with LTE block. For 4×464-QAM system, compared to the case without weighting, the performance improvement is 2db@1e-3,2.2db@1e-3, 2.5db@1e-3, respectively. The results demonstrate the proposed methods are helpful for BER performance.This thesis has firstly proposed architecture for single tree search soft-output sphere decoding, then has finished the fixed-point realization. Finally, this thesis has presented some low complexity methods for timing, search complexity, power. According to search complexity, this thesis has proposed RIP-SE enumeration and LSR radius setting which can effectively reduce the invalid point. The throughput is increased by 2-fold with little performance degradation. According to timing, firstly, this thesis has optimized the fixed-point parameters in the case of run-time constraints, secondly, this thesis has proposed a parallel radius constraints implementation. The area is decreased by 50%. According to power, this thesis has imposed some low power methods such as Gray code, pipelined architecture, clock gating. The power is decreased by 65%. The results demonstrate the proposed methods are helpful for low complexity realization.