System architecture and implementation of MIMO sphere decoders on FPGA
IEEE Transactions on Very Large Scale Integration (VLSI) Systems
50 years of CORDIC: algorithms, architectures, and applications
IEEE Transactions on Circuits and Systems Part I: Regular Papers
Dynamic Nulling-and-Canceling for Efficient Near-ML Decoding of MIMO Systems
IEEE Transactions on Signal Processing
A universal lattice code decoder for fading channels
IEEE Transactions on Information Theory
Closest point search in lattices
IEEE Transactions on Information Theory
On maximum-likelihood detection and the search for the closest lattice point
IEEE Transactions on Information Theory
A unified framework for tree search decoding: rediscovering the sequential decoder
IEEE Transactions on Information Theory
Algorithm and implementation of the K-best sphere decoding for MIMO detection
IEEE Journal on Selected Areas in Communications
A new efficient ordering scheme for sphere detection
Journal of Electrical and Computer Engineering
Hi-index | 0.00 |
The order of the channel matrix columns is an important factor that affects the computational complexity, memory requirement, and detection error rate of multi-input multi-output (MIMO) detectors. Novel efficient ordering schemes proposed in our previous work lead to advantages in the computational complexity and memory requirement of various maximum likelihood (ML) MIMO detectors. In this paper, we apply these ordering schemes to the K-Best detector--a near-ML MIMO detector suitable for high-throughput hardware implementations--and show that our novel ordering schemes improve the performance of the K-Best detector, especially when K is small. With a given detection error rate, our ordering schemes either lead to signal-to-noise ratio (SNR) gains, or allow even smaller K. Two of these ordering schemes, which can be easily embedded into the QR decomposition, are implemented in hardware. We adopt Givens based ordering schemes, due to their numerical stability when fixed point representations are used. Our hardware implementation results show that our novel ordering schemes incur negligible overheads and are particularly suitable for high-throughput implementations.