Radius Choice Algorithm for Complexity Reduction of Probabilistic Tree Pruning Sphere Decoder for MIMO Systems

In digital communication systems, Maximum likelihood (ML) detection is the optimum method to decode a received signal vector. In wireless communication systems, the complexity increases with the constellation size and the number of antennas. It becomes infeasible to apply ML decoding to the practical systems as it searches through all the lattice points in the constellation. Sphere decoding is a popular method that provides near ML performance with reduced complexity. It provides optimal or suboptimal results with reduced complexity, as it searches for points which are within the specified radius of the sphere. The complexity of the sphere decoding is dependent on the two factors. The first one is the initial radius of the sphere, basically to begin the search process. The second factor is to update the radius when no points are found in the specified radius. In this paper, a combination of radius selection using Radius Choice Algorithm and radius updating using Probabilistic Tree Pruning-SD are proposed. First, the initial radii for spheres is obtained, in which expected number of points is some predefined values using radius choice algorithm and the values are stored in a look up table Using the radius from the LUT the search begins and if the points are not found inside the specified radius is updated using PTP-SD. The simulations are carried out for constellation size of 4-QAM and antenna size of 4 X 4 and 8 X 8 MIMO. The complexity is measured by the number of floating point operations (FLOPS) needed. It is seen that, complexity is reduced by an amount of 30 % at higher SNR, without degrading the performance. The proposed technique reduces the number of nodes visited during the search process.