Blind Volterra Equalizer Learning for Mission-Mode NL Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Volterra series-based equalizers rely on reference signals or estimated signals derived from digital signal processing, which is impractical in mission mode due to significant intermediate DSP processing and challenges in recovering the input signal from the ADC output, especially in coherent optical receivers where distortion is severe.
Innovation Solution
A blind learning algorithm that calculates NL distortion correction parameters based on the ADC output signal, employing empirical calculation of the 4th moment and high-order autocorrelation functions to identify model parameters without requiring a reference input, enabling NL distortion compensation during system operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reference signal-based Volterra equalizer calibration is performed, then NL distortion correction accuracy is improved, but system complexity and processing requirements increase significantly
Solution Approach 1:
The system performs self-calibration by using the received signal itself as the reference, eliminating the need for external calibration equipment or complex test setups. The blind learning algorithm automatically identifies Volterra kernel parameters by processing the received signal through multiple stages including coarse calibration and fine calibration phases
Solution Approach 2:
The patent extracts only the essential components needed for calibration by separating the calibration process into distinct phases (coarse and fine calibration) and extracting only the necessary Volterra kernel parameters rather than attempting to characterize the entire system response
2Loss of time
If blind learning algorithm is used for NL distortion correction, then processing time and resources are conserved, but calculation complexity of high-order moments increases
Solution Approach 1:
The blind learning algorithm is divided into two distinct phases: coarse calibration that computes lower-order moments (up to 4th order) to establish initial Volterra kernel parameters, and fine calibration that computes higher-order moments (up to 6th order) to refine the parameters. This segmentation reduces overall computational complexity compared to computing all high-order moments simultaneously
Solution Approach 2:
The coarse calibration phase performs preliminary estimation of the Volterra kernel parameters using lower-order moments before the fine calibration phase refines these parameters using higher-order moments. This preliminary action reduces the computational burden of the final calibration by starting with reasonable initial estimates
Data Source
AI summary
Aspects of the subject disclosure may include, for example, conducting blind learning of correction parameters for non-linear (NL) distortion associated with one or more components of a system, resulting in learned correction parameters, wherein the conducting is performed without a need to identify or estimate a reference input associated with the one or more components, and causing the learned correction parameters to be applied to an output signal associated with the one or more components to compensate for the NL distortion. Other embodiments are disclosed.