Frequency-Domain Equalization Circuit for Rational Oversampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing adaptive equalization filters in optical fiber communication require high computational complexity due to large time spreads, especially when using frequency domain filters with oversampling rates that are not integer multiples of the symbol rate, leading to increased calculation amounts and complex circuit designs.
Innovation Solution
An equalization signal processing circuit that divides input signals into M/L oversampled signals, performs frequency domain conversions and arithmetic operations using filter coefficients, and updates coefficients through an error backpropagation method, allowing for efficient circuit design by utilizing FFT and IFFT sizes that are powers of two.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a frequency domain filter is used to compensate for large time spreads, then the calculation amount is reduced, but the circuit design becomes complex when using non-integer multiple oversampling rates
Solution Approach 1:
The patent segments the frequency domain filter into multiple sub-filters, each handling a specific frequency band. This segmentation allows the use of simpler integer-based oversampling rates for each sub-filter while maintaining the overall effectiveness for large time spread compensation, thereby reducing individual circuit complexities.
Solution Approach 2:
The patent changes the oversampling rate parameter from non-integer to integer multiples, specifically designing the system to operate at 2x or 4x oversampling rates. This parameter change simplifies the FFT and IFFT operations, making the circuit design more manageable while still achieving effective equalization through the segmented filter approach.
2Reliability
If a large filter is used to compensate for effects with large time spread, then the compensation effectiveness is improved, but the calculation amount increases
Solution Approach 1:
The patent transitions from time-domain filtering to frequency-domain filtering, changing the dimension in which the equalization operation is performed. In the frequency domain, the filter operation becomes simple multiplication instead of convolution, dramatically reducing the calculation amount while maintaining the ability to handle large time spreads through the use of FFT and IFFT.
Solution Approach 2:
The patent changes the mathematical operation parameter from convolution in time domain to multiplication in frequency domain. This parameter change fundamentally reduces the computational complexity from O(N^2) to O(N log N) through the use of fast Fourier transform, enabling efficient processing of large time spread compensation.
3Reliability
If two-times oversampling is used in linear adaptive filter, then the Nyquist criterion is satisfied, but the calculation amount increases for non-integer multiple oversampling rates
Solution Approach 1:
The patent segments the oversampling operation into integer-based stages (2x or 4x), avoiding the need for complex non-integer oversampling. Each segment uses standard integer multiplication factors that are computationally efficient, while the segmentation allows the system to still satisfy the Nyquist criterion through the combined effect of the segmented processing stages.
Data Source
AI summary
An equalization signal processing circuit includes: a signal division unit that divides an input signal of oversampling of a rational number M/L multiple into M signals; a first frequency domain filter that performs an arithmetic operation of a first filter coefficient on M signals in a frequency domain; a second frequency domain filter that performs an arithmetic operation of a second filter coefficient on, for each L group, M signals on which the arithmetic operation of the first filter coefficient is performed; a time domain conversion unit that converts a signal added for each group into a signal in a time domain; a switch circuit that sequentially selects a signal converted into a signal in the time domain for each group; and a coefficient updating unit that updates the first filter coefficient and the second filter coefficient.