Low-complexity neural network equalization for long-haul coherent transmission with cascaded semiconductor optical amplifiers

Abstract

In this letter, we numerically investigate a long-haul coherent data transmission system with a cascade of semiconductor optical amplifiers (SOAs). We exploit low-complexity neural networks that can be implemented in real time to compensate for the accumulated distortions induced by a cascade of SOAs. This equalization provides an order-of-magnitude reduction in bit error rate at low dispersion (in the O-band), whereas higher dispersion degrades performance.

Figures (4)

• Figure 1: The scheme of the simulated data transmission system. At the transmitter: bit sequence generator, modulator, and RRC filter; in line: 16 spans each consisting of 73 km of standard fibre followed by an SOA and a Gaussian filter; at the receiver: linear dispersion compensation (LDC), matched RRC filter, NN equalizer, and demodulator.
• Figure 2: Comparison of the SOA model with the experimental data from Thorlabs' official webpage for SOA1013S thorlabs_SOA.
• Figure 3: BER as a function of launch power for the system with SOA and $\mathrm{LinAmp_{8dB}}$ (linear amplifier with $\mathrm{NF=8\, dB}$). (a): The scenarios are SOA (red), SOA with no fiber (blue), $\mathrm{LinAmp_{8dB}}$ (green). (b): SOA (red, from (a) for reference), SOA with $\beta_2 \rightarrow 0$ (purple), and $\mathrm{LinAmp_{8dB}}$ with $\beta_2 \rightarrow 0$ (green). In all cases, solid lines correspond to non-equalized systems, dashed lines to equalized systems. The forward error correction (FEC) threshold is $3.8 \times 10^{-3}$ (7% overhead).
• Figure 4: The dependence of BER on $\beta_2$ for system with SOA and $\mathrm{LinAmp_{8dB}}$ when NN is applied or not. Two constellations correspond to the wavelength at which NN equalization provides the maximum BER reduction (circled on the graph).