Efficient Architecture for Soft-Input Soft-Output Sphere Detection With Perfect Node Enumeration

The application of the turbo principle allows to exploit the full potential of multiple-input multiple-output (MIMO) communications at the cost of increasing the computational effort at the receiver. In the context of soft-input soft-output tree search detection, the computation of metric values and the optimal node order represents two of the most computationally demanding operations. Heuristic approaches may be applied to reduce the complexity, but their accuracy is compromised by the effect that the input soft information has on the node ordering. The design of adaptive, good-performing, and cost-effective tree search detectors for iterative receivers represents, hence, a challenging task. To alleviate these complexity and performance loss drawbacks, an efficient MIMO sphere detector realization is proposed in this paper. A novel smart-sorting enumeration approach offers a significant gain in terms of throughput (from 40% up to a factor 5) and energy efficiency (up to 80% energy saving in the low SNR regime) with regard to preceding implementations. Owing to the additional low delay and area cost reported, the proposed design represents a very promising candidate toward a fast, accurate, and efficient MIMO detector.