Experimental Characterization of Raman Amplifier Optimization Through Inverse System Design

Abstract

Optical communication systems are always evolving to support the need for ever-increasing transmission rates. This demand is supported by the growth in complexity of communication systems which are moving towards ultra-wideband transmission and space-division multiplexing. Both directions will challenge the design, modeling, and optimization of devices, subsystems, and full systems. Amplification is a key functionality to support this growth and in this context, we recently demonstrated a versatile machine learning framework for designing and modeling Raman amplifiers with arbitrary gains. In this paper, we perform a thorough experimental characterization of such machine learning framework. The applicability of the proposed approach, as well as its ability to accurately provide flat and tilted gain-profiles, are tested on several practical fiber types, showing errors below 0.5 dB. Moreover, as channel power optimization is heavily employed to further enhance the transmission rate, the tolerance of the framework to variations in the input signal spectral profile is investigated. Results show that the inverse design can provide highly accurate gain-profile adjustments for different input signal power profiles even not considering this information during the training phase.

Fiber-Agnostic Machine Learning-Based Raman Amplifier Models

“…Therefore, previous knowledge of the experimental system is not critical. Although not as accurate as numerical models, NN-based RA models have proven to be very efficient in predicting Raman gain profiles [16], [17], [19], [27], amplified spontaneous emission (ASE) noise [18], [34] and noise figure (NF) [20]. In these models, X incorporates just the Raman pump powers and has, therefore, dimension N p (number of pumps).…”

“…This is because the Raman gain strength and spectral shape depend on the material composition, the length, the attenuation, and the effective area of the optical fiber [26]. Relying on these fiber-specific models requires countless models to be available, one for each considered fiber span [27], and a massive amount of data to train each of them. Therefore, generalizable RA models that can predict the RA performance for fiber types unseen during training are critically needed.…”

Design of ultra-long-distance optical transport networks based on high-order remotely pumped amplifier

“…Distributed Raman amplification (DRA) and ROPA are very beneficial technologies to enlarge the transmission distance in long-haul WDM transmission systems thanks to their excellent low noise characteristics [2,[6][7][8][10][11][12][13][14][15]. Especially, an optimized combination of a DRA and a ROPA, was recently proposed to demonstrate an enhancement of optical signal to noise ratio (OSNR) and a long-distance terrestrial field-trial [7,8].…”

Online versus Offline Optimization Methods for Raman Amplifier Optimization

“…(2) Supervised training, consisting in training NN inv and NN using . Details about how the NNs architecture, training and performance in a wide range of gain profiles (flat, tilted, and arbitrary) can be found in [7]. The first step is the most time-consuming.…”

“…Recently, machine learning (ML) has been applied to learn the complex pump-signal relations in RAs [6]. It has been shown to reach highly accurate gain profile optimization for C-band [7], S+C and S+C+L-bands [2]- [4] transmissions. Based on neural networks (NN) that learns from data, such a data-driven approach performs ultra-fast offline optimization, and normally dedicated NN models are required for each specific experimental scenario [8].…”