Time-Domain Digital Back Propagation: Algorithm and Finite-Precision Implementation Aspects

Abstract

We propose a nonlinear mitigation algorithm designed from an ASIC perspective, and analyze implementation aspects. Given 9 signal and 11 coefficient bits, reach is increased by 105% compared to linear compensation in single-channel 16-QAM transmission.

ASIC Implementation of Time-Domain Digital Backpropagation with Deep-Learned Chromatic Dispersion Filters

“…We have studied TD-DBP based on deep-learned CD filters from an ASIC perspective. It was shown that the obtained filters have similar signal resolution requirements compared to our previous work 5,6 , and significantly reduced coefficient resolution requirements. Moreover, reduced filter lengths directly translate into lower power dissipation and chip area.…”

“…A major issue with DBP based on the splitstep Fourier method (SSFM) is the large complexity caused by the fast Fourier transforms (FFTs). Time-domain DBP (TD-DBP) with finite impulse response (FIR) filters may be competitive [5][6][7][9][10][11][12] , assuming that the chromatic dispersion (CD) steps are sufficiently short. Design methods for the FIR filters include least squares 5,6,13 or wavelets 10 , but accumulating truncation errors due to repeated filter use can lead to severe performance degradations.…”

Wideband Time-Domain Digital Backpropagation via Subband Processing and Deep Learning

“…Real-time digital backpropagation (DBP) based on the split-step Fourier method (SSFM) is widely considered to be impractical due to the complexity of the chromatic dispersion (CD) steps. To address this problem, finite impulse response (FIR) filters may be used instead of fast Fourier transforms (FFTs) to perform time-domain CD filtering [1][2][3][4][5][6][7] . Indeed, the FIR filters can be as short as 3 taps per SSFM step, provided that the step size is sufficiently small (i.e., many steps are used) and the filters in all steps are jointly optimized 6 .…”

“…The complexity of time-domain DBP (TD-DBP) is dominated by the total number of CD filter taps in all steps. Recent work has focused on relatively narrowband signals (e.g., 10 Gbaud in 6 and 20 Gbaud in [3][4][5] ) for which the overall CD memory is low. Since the memory increases quadratically with bandwidth, it is not clear if TD-DBP can be scaled gracefully also to more wideband signals.…”

Model-Based Machine Learning for Joint Digital Backpropagation and PMD Compensation

“…As an application, we consider joint digital backpropagation (DBP) and polarization-mode dispersion (PMD) compensation, similar to [12][13][14][15]. Compared to previous work, the employed model uses hardware-friendly time-domain implementations [16,17] and does not assume any knowledge about the particular PMD realizations along the link (i.e., differential group delays (DGDs) and polarization states), nor any knowledge about the total accumulated PMD. We show that our model converges reliably to a solution with performance close to the case where PMD is absent from the link.…”