Cooperative Bit Allocation in In-Band Full-Duplex Power Line Communication

Abstract

In-band full-duplex (IBFD) is an attractive technology in broadband power line communication (BB-PLC) because it helps to improve spectral efficiency. However, IBFD is challenging since it requires additional hardware and advanced signal processing to mitigate self-interference (SI) signals. SI cancelation architectures and channel estimation techniques determine the overall IBFD performance. Accurate SI channel estimation is required since imperfect SI cancelation reduces signal-to-interference-plus-noise ratio (SINR), causing an increase in data errors and a decrease in data rates. Although channel estimation can be improved by sending additional training symbols, increasing the training duration will lower data throughput. Thus, the training symbol number is an essential trade-off for IBFD performance in BB-PLC. In this paper, we investigate IBFD performance in single-input single-output (SISO), single-input multiple-output (SIMO), and multiple-input multiple-output (MIMO) communication scenarios, including the influence of the training period. By analyzing error vectors on a constellation diagram, we obtain the closed-form expressions for the symbol error probability (SEP) affected by IBFD and the training duration. Based on the obtained expressions, we propose a bit allocation algorithm to determine bit loading to ensure reliable IBFD communication. Furthermore, we suggest a procedure to compute the optimal training symbol number that maximizes throughput in IBFD. Using the proposed bit allocation strategy and a database of measured channels, we estimated the achievable bidirectional throughput and the throughput gain in IBFD compared to time division duplexing (TDD).