[Paper Review] Joint Superchannel Digital Signal Processing for Ultimate Bandwidth Utilization

Mikael Mazur, Jochen Schröder+2 more|arXiv (Cornell University)|Nov 6, 2019

Optical Network Technologies48 references4 citations

TL;DR

This paper proposes joint superchannel digital signal processing using an optical frequency comb to enable super-Nyquist transmission with zero guard bands, achieving higher spectral efficiency by exploiting comb stability and aliasing in a two-sample-per-symbol receiver. The method reduces optimal guard-band from 1 GHz to 500 MHz and increases roll-off from 1% to 10%, enabling higher throughput and improved robustness without upsampling or additional equalization stages.

ABSTRACT

Modern optical communication systems transmit multiple frequency channels, each operating very close to its theoretical limit. The total bandwidth can reach 10THz limited by the optical amplifiers. Maximizing spectral efficiency, the throughput per bandwidth is thus crucial. Replacing independent lasers with an optical frequency comb can enable very dense packing by overcoming relative drifts. However, to date, interference from non-ideal spectral shaping prevents exploiting the full potential of frequency combs. Here, we demonstrate comb-enabled multi-channel digital signal processing, which overcomes these limitations. Each channel is detected using an independent coherent receiver and processed at two samples-per-symbol. By accounting for the unique comb stability and exploiting aliasing in the design of the dynamic equalizer, we show that the optimal spectral shape changes, resulting in a higher signal to noise ratio that pushes the optimal symbol rate towards and even \emph{above} the channel spacing, resulting in the first example of frequency-domain super-Nyquist transmission with multi-channel detection for optical systems. The scheme is verified both in back-to-back configuration and in single span transmission of a 21 channel superchannel originating from a 25GHz-spaced frequency comb. By jointly processing 3 wavelength channels at a time, we achieve spectral efficiency beyond what is possible with independent channels. At the same time, one significantly relaxes the hardware requirements on digital-to-analog resolution and bandwidth, and well as filter tap numbers. Our results show that comb-enabled multi-channel processing can overcome the limitations of classical dense wavelength division multiplexing systems by enabling tighter spacing to reach the ultimate spectral efficiency in optical communications.

Motivation & Objective

• To overcome spectral efficiency limits in dense WDM systems caused by guard bands and inter-channel interference.
• To eliminate the need for large guard bands in superchannels by leveraging the frequency-locked stability of optical frequency combs.
• To enable super-Nyquist transmission (symbol rate > channel spacing) through joint multi-channel digital signal processing.
• To reduce hardware complexity by avoiding upsampling and minimizing filter tap counts through optimized spectral shaping.
• To demonstrate that joint processing can improve signal-to-noise ratio and system tolerance without increasing system latency or computational overhead.

Proposed method

• The system uses a 25 GHz-spaced optical frequency comb as a stable, coherent transmitter source to eliminate relative laser frequency drift.
• Each of the 21 channels is coherently detected at two samples per symbol, enabling baseband processing without upsampling.
• A dynamic MIMO equalizer is extended to jointly mitigate inter-channel crosstalk by exploiting the comb’s inherent frequency locking.
• Aliasing is intentionally used in the equalizer design to replace the need for signal upconversion, simplifying hardware requirements.
• The spectral shape is optimized to increase roll-off from 1% to 10%, reducing equalizer memory and improving tolerance to transceiver impairments.
• Joint dispersion compensation is applied per channel, maintaining smooth filter response while preserving individual channel independence.

Experimental results

Research questions

• RQ1 Can joint multi-channel digital signal processing overcome the spectral efficiency limitations of traditional dense WDM systems with guard bands?
• RQ2 How does exploiting the frequency-locked nature of optical frequency combs enable interference cancellation beyond what is possible with free-running lasers?
• RQ3 To what extent can aliasing in a two-sample-per-symbol system replace upsampling for super-Nyquist transmission?
• RQ4 What is the optimal spectral roll-off factor when inter-channel interference is jointly mitigated via dynamic equalization?
• RQ5 Can joint processing reduce hardware complexity (e.g., DAC resolution, filter taps) while improving signal-to-noise ratio and system reach?

Key findings

• The effective guard-band was reduced from 1 GHz to 500 MHz in both back-to-back and 80 km transmission, demonstrating significant spectral efficiency gain.
• The optimal spectral roll-off increased from 1% to 10%, enabling better tolerance to transceiver impairments and reducing equalizer memory.
• The system achieved super-Nyquist transmission (symbol rate > channel spacing) through frequency-domain interference cancellation without upsampling.
• The joint processing scheme improved the effective signal-to-noise ratio, enabling higher throughput or extended transmission reach.
• The method reduced hardware complexity by eliminating the need for upsampling and minimizing filter tap counts, while maintaining high performance.
• The results confirm that comb-based superchannels with joint DSP can achieve ultimate spectral efficiency by leveraging intrinsic comb stability and aliasing in the equalizer design.