A VLSI architecture of a K-best lattice decoding algorithm for MIMO channels

Lattice decoding algorithms have been proposed for implementing the maximum likelihood detector (MLD), which is the optimal receiver for multiple-input multiple-output (MIMO) channels. However the computational complexity of direct implementation of the lattice decoding algorithm is high and the throughput is variable. In this work, a K-best algorithm is proposed to implement the lattice decoding. It is computational inexpensive and has fixed throughput. It can be easily implemented in a pipelined fashion and has similar performance as the optimal lattice decoding algorithm if high value of K is used. In this paper, we describe a pipelined VLSI architecture for the implementation of the K-best algorithm. The architecture was designed and synthesized using a 0.35 /spl mu/m library. For a 4-transmit and 4-receive antennas system using 16-QAM, a decoding throughput of 10 Mbit/s can be achieved.

I. Introduction

For Space Division Multiplexing, where information is transmitted and received over several transmit and receive antennas in parallel, the optimal receiver is the maximum likelihood detector (MLD) [1]. Basically MLD tries to maximize the conditional probability , where is the received signal and is a possible transmitted signal. One way to implement this receiver is to exhaustively for all possible transmitted symbol vectors. This requires Q Euclidean distances calculation, where Q is the constellation size of each element in the transmitted vector, and M ‘ is the number of transmit antennas. The complexity is huge and increases exponentially with the number of transmit antennas. Recently, the lattice decoding algorithms [2]–[5] have been proposed to reduce its computation complexity significantly. The lattice decoder can be interpreted as a depth-first tree search approach with pruning. It has the disadvantage that the computation requirement varies with the input signal and hence the decoding throughput is also varying, which is not desirable for real time signal detection.