A Novel RTRLNN Model for Passive Intermodulation Cancellation in Satellite Communications

• Published in: 2018 14th International Wireless Communications & Mobile Computing Conference (IWCMC) pp. 18 - 23
• Main Authors: Liang, Bizheng, Bu, Xiangyuan, Li, Mucheng, Guo, Pengfei, Liu, Celun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2018
• Subjects: Adaptation models; Analytical models; Feature extraction; Integrated circuit modeling; Interference cancellation; Passive Intermodulation; RTRLNN; Satellite communication
• ISSN: 2376-6506
• DOI: 10.1109/IWCMC.2018.8450316

Passive intermodulation (PIM) often limits the performance of satellite communication systems with multi-carriers. The PIM interference has a peculiarity of time-varying and non-Markov. While the existing digital signal processing methods have limited effects, we propose a novel adaptive real-time recursion learning neural network (RTRLNN) which is suitable for dynamic PIM in satellite communications. The proposed novel RTRLNN method has fast convergence with its adaptive learning rate and accurate approximation ability. In this paper, a cancellation system with RTRLNN algorithm is designed for the PIM interference in satellite communications. The system has two processing sections which are pilot slot and data transmission slot. The system extracts the feature of PIM interference and trains itself in the pilot slot while achieves cancellation of PIM interference in the data transmission slot. Simulation results are presented on time and frequency domain illustrating the effectiveness of PIM interference cancellation. Compared with the modified least mean square(LMS) method, the proposed novel RTRLNN method significantly decreases the bit error rate(BER) of PIM interference. Furthermore, the proposed novel RTRLNN method shows an enhancement of 10\mathrm {d}\mathrm {B} in SIR gain while the \mathrm {E}\mathrm {b}/\mathrm {N}0 fixed 10\mathrm {d}\mathrm {B}.