Spatial Multiplexing Architectures with Jointly Designed Rate-Tailoring and Ordered BLAST Decoding Part-II: A Practical Method for Rate and Power Allocation

The study of the class of new spatial multiplexing architectures (SMAs) is continued. As introduced in Part I of this paper, the SMAs consist of joint design of rate and power allocation to the spatially-multiplexed substreams at the transmitter and ordered nulling/canceling detection/decoding at the receiver. This was studied in Part I under the diversitymultiplexing tradeoff framework. Here the more detailed and practical problem of allocating rates and powers across the transmit antennas is investigated to minimize the overall system (uncoded or outage) error probability. Since the layer gains are unavailable to the transmitter, the rates and powers must be allocated based on the statistics of the layer gains. However, the channel dependent ordering rules make the precise distributions of the layer gains complicated or intractable. To solve this problem, a simple, yet effective, four-parameter hyperbola model is proposed to closely approximate the error probability of each layer. With this approximation, the computation of rate and power allocation according to the criterion of minimizing the maximal error probability of all the substreams is given. Simulation results validate the superior performance of the proposed SMAs, especially that of the Greedy ordering Rate Tailored SMA (GRT-SMA). Although the rate and power allocation method is obtained under the assumption of iid Rayleigh fading channel, the proposed SMAs also work well in other types of fading channels (such as in correlated and Rician channels).