A multistage sphere decoder for frequency-selective MIMO channels

A new multistage sphere decoder (MSD) for the detection of multiple input-multiple output (MIMO) systems in frequency-selective channels is presented in this paper. The algorithm is based on the Schnorr-Euchner (SE) version of the sphere decoder (SD) but reduces its complexity by dividing the tree search of the SD in a number of non-overlapping stages that directly determine its performance and complexity. The MSD performs a reduced-size tree search in each one of the stages, applying interference cancellation between them to subtract the effect of previously searched stages. By selecting the number of stages of the MSD, different levels of performance can be attained, ranging from the maximum likelihood (ML) performance of the SD and the performance of the generalized decision feedback equalizer (GDFE). Simulations show that the MSD can reduce the complexity of the SD while approaching its ML performance. In particular, by reducing the worst-case complexity of the SD, the MSD is more suitable for hardware implementation even though it still presents a variable complexity.