DWDM Data Model for Accurate Gain and Noise Figure

The patent applies preliminary action by pre-establishing data models for network elements before actual OSNR calculation is needed. These models are built using training data that captures the relationship between input/output optical parameters and noise coefficients/gains. When OSNR calculation is required, the pre-built models are directly applied, eliminating the need for real-time complex measurements while maintaining high accuracy.

The patent introduces data models as intermediaries between the physical network elements and the OSNR calculation process. Instead of directly measuring or presetting gain and NF values, the system uses data models that process input optical power, output optical power, and optical signal-to-noise ratio measurements to derive accurate noise coefficients and gains. This intermediary layer resolves the contradiction by enabling accurate calculations without requiring complex real-time operations.

The patent applies segmentation by dividing the complex task of OSNR calculation into separate data models for different network element types (e.g., optical amplifiers, ROADMs). Each network element has its own dedicated data model that is trained and optimized independently. This segmentation reduces overall system complexity by allowing each model to be managed separately while collectively providing accurate OSNR calculations across the entire network.

The patent utilizes parameter changes by training data models with multiple sets of training data that represent different operating conditions, wavelength combinations, and network configurations. The models learn to adapt their parameters (noise coefficients, gains) based on the specific input parameters (input optical power, output optical power, optical signal-to-noise ratio) provided during calculation, enabling accurate results across varying conditions without increasing structural complexity.