Optimal Training-Based Scheme for MIMO Full-Duplex Wireless Links

Abstract

In a multiple-input-multiple-output (MIMO) full-duplex (FD) wireless link, training-based schemes are the most effective with digital self-interference (SI) cancellation methods after antenna and analog SI cancellations, and they help receiver to know responses of both residual SI channel for digital cancellation and transmission channel for demodulation. In this situation, the open issue is that how much training is needed in MIMO FD wireless links-too little training will cause the imperfect digital SI cancellations and the transmission channels are also improperly learned, too much training and the links have no time left for the data transmission before the fading channel changes. We compute a capacity lower bound of a MIMO FD channel that is learned at the receiver by training without any feedbacks, and maximize the bound as a function of the received signal-to-noise ratio $\rho $ , fading coherence time $T$ , and numbers of link antennas $M$ , $N$ . When training power is equal to data power, numerical results show that the optimal number of training symbols $T_{o}$ is $T_{o}\geq 2\mathrm {max}(M,N)$ if $T \gg \mathrm {max}(M,N)$ under appropriate $\rho $ . Conversely, if $T\approx 2\max (M,N)$ , $T_{o} .