Adaptive Averaging for Nonlinear Noise in Frequency Domain Equalization

The patent segments the frequency domain processing into two distinct stages: first performing standard frequency domain equalization to correct linear intersymbol interference, then performing a second frequency domain transformation and equalization pass to correct nonlinear distortions. This segmentation allows each stage to address specific types of channel impairments independently, resolving the contradiction between achieving convergence and correcting nonlinear distortions.

The patent introduces an intermediary frequency domain transformation step that converts the time-domain equalized signal back to frequency domain for a second equalization pass. This intermediary transformation acts as a bridge between linear and nonlinear equalization processes, enabling the system to first correct linear effects and then address nonlinear distortions that were not fully eliminated in the first pass.

The patent implements a feedback mechanism where the output of the first frequency domain equalization pass is fed back through another frequency domain transformation and equalization stage. This feedback loop allows the system to iteratively refine channel estimation by using the equalized signal from the first pass as input to the second pass, thereby improving channel estimation accuracy while maintaining computational efficiency through reuse of the same processing structure.

The patent maintains continuous useful action by performing equalization in a continuous two-stage process rather than discrete separate operations. The first frequency domain equalization pass continuously processes the signal, and its output continuously feeds into the second pass, ensuring that channel estimation and distortion correction occur in an ongoing manner that improves accuracy without interrupting the data flow or requiring additional computational overhead.