Achievable Throughput in In-Band Full-D...

In-band full-duplex (IBFD) draws significant attention in many communication systems since it can improve spectral efficiency and data rates. Recent research activities in broadband power line communication (BB-PLC) also considered IBFD. However, implementation of IBFD for BB-PLC is highly challenging due to the harsh nature of PLC channels. Besides the remote link estimation, the estimation and cancellation of self*interference (SI) signals are necessary. In many research works, the feasibility of IFBD was analyzed by comparing signal-to-interference plus noise ratio (SINR) in the half-duplex and the full-duplex modes. However, to enable sufficient SI cancellation, very accurate SI channel estimation is needed, which may require the transmission of additional training symbols. In such a case, there is a trade-off between time spent on training and time used for data transmission. In the IBFD case, it is unclear what the optimal number of training symbols should be and what the effective throughput is. This paper investigates this problem and estimates the achievable throughput for IBFD in SISO, SIMO, and MIMO scenarios. Firstly, we perform Monte Carlo simulations over a database of measured channels to analyze bit allocation and its dependence on the number of training symbols. Secondly, we look for the optimal proportion between the training period and payload transmission to achieve the best throughput. Finally, we compute the average bit error rates, throughput, and bidirectional throughput gain.