Wiener-Hammerstein model and its learning for nonlinear digital pre-distortion of optical transmitters

Abstract

We present a simple nonlinear digital pre-distortion (DPD) of optical transmitter components, which consists of concatenated blocks of a finite impulse response (FIR) filter, a memoryless nonlinear function and another FIR filter. The model is a Wiener-Hammerstein (WH) model and has essentially the same structure as neural networks or multilayer perceptions. This awareness enables one to achieve complexity-efficient DPD owing to the model-aware structure and exploit the well-developed optimization scheme in the machine learning field. The effectiveness of the method is assessed by electrical and optical back-to-back (B2B) experiments, and the results show that the WH DPD offers a 0.52-dB gain in signal-to-noise ratio (SNR) and 6.0-dB gain in optical modulator output power at a fixed SNR over linear-only DPD.

arXiv commentAccepted version for Optics Express

References in corpus (5)

• Adam: A Method for Stochastic Optimization
• Identification of block-oriented nonlinear systems starting from linear approximations: A survey
• Laser Phase Noise Tolerance of Uniform and Probabilistically-shaped QAM Signals for High Spectral Efficiency Systems
• Fast Identification of Wiener-Hammerstein Systems using Discrete Optimization
• Nonlinear Interference Mitigation via Deep Neural Networks

Cited by in corpus (3)

• Artificial Neural Networks for Photonic Applications: From Algorithms to Implementation
• An Interpretable Mapping from a Communication System to a Neural Network for Optimal Transceiver-Joint Equalization
• Physics-oriented learning of nonlinear Schrödinger equation: optical fiber loss and dispersion profile identification