Fast and Area-Efficient Sphere Decoding Using Look-Ahead Search

Sphere decoding enables maximum likelihood (ML) detection with fairly low complexity in the MIMO wireless systems, but it takes hundreds cycles at low SNR environment. This paper proposes a fast decoding algorithm to reduce the decoding cycles using look-ahead search. Since the proposed decoding algorithm utilizes hardware resources to add other sub-trees into the search space successively, it helps not to go down into the sub-tree that has a small value at the root node but has large values at the child nodes. Scaling and enumeration techniques are also presented, which are effective in implementing the proposed sphere decoder. As a result, the proposed decoder saves about 30% decoding cycles at the cost of small hardware overhead compared to the conventional decoder.