NIST logo

EEEL and USC Collaboration Applies Artificial Neural Networks to Optical Performance Monitoring

For Immediate Release: September 1, 2009

*

Contact: Jeffrey Jargon
303-497-4961

High-performance optical networks are susceptible to degrading effects that can change over time. Knowledge of the degradation can be used to diagnose the network, repair the damage, drive a compensator/equalizer, and/or reroute traffic around a nonoptimal link. Therefore, it is valuable to monitor the network performance for many types of impairments, such as optical signal-to-noise ratio (OSNR), chromatic dispersion (CD), polarizationmode-dispersion (PMD), and fiber nonlinearity; any of these parameters can change with temperature, plant maintenance, and path reconfiguration. Key features of any optical performance monitor are simplicity in implementation and the ability to accommodate different modulation formats and impairments.

Researchers from EEEL and the University of Southern California (USC) have demonstrated a novel artificial neural network (ANN) model for optical performance monitoring. ANNs can be used to model arbitrary relationships between inputs and outputs. Therefore, they have the potential to be a powerful tool for performance monitoring in optical fiber communication systems as well as other aspects of complex optical system and device design. The NIST-USC model is capable of simultaneous identification of OSNR, CD, and PMD from common eye-diagram parameters for both 40 Gb/s return-to-zero on-off keying (RZ-OOK) and differential phase-shift-keying (RZ-DPSK) systems. The researchers experimentally verified their model using two cascading Mach-Zehnder modulators to generate 40 Gb/s RZ-OOK and RZ-DPSK signals. They then extended this technique to simultaneously identify accumulated fiber nonlinearity, OSNR, CD, and PMD from eye-diagram and eye-histogram parameters in a three-channel 40 Gb/s DPSK wavelength-division multiplexing (WDM) system. Simultaneous identification of accumulated fiber nonlinearity, OSNR, CD, and PMD causing changes from a baseline by use of the eye-diagram and eye-histogram parameters were obtained, and high correlation coefficients (> 0.95) were achieved with various baselines. Finally, the ANN model was also capable of simultaneous identification of in-phase/quadrature data misalignment and data/carver misalignment in return-to-zero differential quadrature phase shift keying transmitters.

X. Wu, J.A. Jargon, R.A. Skoog, L. Paraschis, and A.E. Willner, "Applications of Artificial Neural Networks in Optical Performance Monitoring," J. Lightwave Technol. 27 (2009) 3580-3589.