The VK-402 searches only within a fixed number of lattice points, where the number of considered lattice points decreases as we descend from level N. To achieve a close to ML BER performance, channel column ordering is required.
Architecture
The VK-402 is FPGA prototyped for 4×4 MIMO system with 16-QAM modulation. Fig. 2 shows the architecture of the VK-402. The columns of the channel matrix H are assumed to be perfectly known at the receiver. They are first ordered according to number of lattice points considered at each level in the tree. Then, they are fed, along with the received vector, to the QR Decomposition block, which is based on CORDIC given rotation algorithm. The upper triangular matrix R and the rotated data vector y are both applied to the sphere decoder. The VK-402 is carefully and optimally designed to deliver a fixed throughput by using semi-parallel architecture for processing the tree nodes. Based on this implementation,the VK-402 provides an improved hardware implementation in terms of throughput and area, which compares favorably to other implementations.
The hardware implementation can be customized to accommodate other MIMO sizes, as well as different constellation sizes.
