Comparison of linear and nonlinear equalization for ultra-high capacity spectral superchannels

Abstract

In ultra-high-speed (>400Gb/s per wavelength), high-spectral efficiency coherent optical communication systems using multi-carrier spectral superchannels, the maximum reach is severely limited due to linear and, foremost, nonlinear impairments. Hence, the implementation of advanced digital signal processing (DSP) techniques in optical transceivers is crucial for alleviating the impact of such impairments. However, the DSP performance improvement comes at the expense of increased cost and power consumption. Given that the computational complexity of the applied linear and nonlinear equalizers is the factor that determines the trade-off between the performance improvement and cost, in this study we provide an extended analysis on the computational complexity of various linear and nonlinear equalization approaches. First, we draw a complexity comparison between a conventional OFDM coherent receiver versus a filter-bank based OFDM receiver and it is shown that the latter provides significant complexity savings. Second, we present a comparison between the digital back-propagation split-step Fourier (DBP-SSF) method and the inverse Volterra series transfer function nonlinear equalizer (IVSTF-NLE) in terms of performance and computational complexity for a 32 Gbaud polarization multiplexed (PM)-16 quadrature amplitude modulation (QAM) OFDM superchannel. Keywords: Multi-band OFDM superchannels, filter-bank modulation, nonlinear compensation, computational complexity, power consumption. System vendors are interested in moving the pluggable transceivers for 100G coherent applications into smaller transceivers, such as the C form factor pluggable 2 (CFP2) in order to increase the bandwidth density. Although the CFP2 appears as a very promising candidate, it is challenging to include all the necessary elements of a 100G coherent transceiver (i.e. either in terms of the footprint or the electrical power budget). Another solution is the CFP2-ACO (analog coherent optics) [2]. In this design, the DSP ASIC is placed on the line card rather than on the module, reducing the footprint and the power consumption compared to other alternatives, such as the CFP or the OIF MSA. The main drawback for CFP2-ACO is that the flexibility is reduced since it can only be plugged into line card slots specifically designed for this transceiver technology. Under this restriction, the focus has now switched to the CFP2-digital coherent optics (DCO) which incorporates the DSP chip (anticipated to be released in 2017). However, the major challenge is to integrate the DSP chip in a limited package space while reducing the power consumption further in order to meet the stringent target specifications of CFP-DCO [3].According to OIF-Tech-Options-400G-01.0, the most viable solution to reduce the computational complexity of the DSP chip, and consequently, the power consumption in superchannel transmission systems (carrying bit rates greater than 100 Gb/s) is to divide the signal into many sub-channels operating at lower baud rate whilst maintaining the same...

Técnicas avançadas de projeto e processamento digital de sinais para sistemas ópticos sem repetição de alta capacidade

“…Atualmente, esforços são realizados para mitigar efeitos não lineares, para os quais o algoritmo de retropropagação digital (digital back-propagation -DBP) e a estimação de sequência de máxima verossimilhança (maximum likelihood sequence estimation -MLSE) se mostram como possíveis soluções [37]. Diversos trabalhos envolvendo o DBP e MLSE aplicados a sistemas ópticos com repetição já foram demonstrados na literatura [38,39,40,41,42,43,44,45].…”