Reduced complexity equalization for coherent long-reach passive optical networks [invited]

TLDR: It is found that a cascade of linear filters can be combined into a single, truncated, linear, adaptive filter with negligible impact on receiver sensitivity, and by utilizing a multiplier-free tap weight update algorithm, the overall complexity of a digital coherent receiver can be significantly reduced, making it attractive for use in an ONU.

Abstract: Coherent receivers offer a potential solution for implementing a high-capacity, long-reach (up to 100 km) passive optical network (LR-PON), due mainly to their high sensitivity, frequency selectivity, and bandwidth efficiency. When using coherent receivers, received signals can be post-processed digitally to mitigate the specific impairments found in access networks and, additionally, relax the optical complexity requirements of the coherent receiver. However, the digital signal processing must itself be low complexity in order to minimize the overall complexity and power consumption of the optical network unit (ONU). This paper focuses on the impact of reduced complexity equalization algorithms on receiver sensitivity in a LR-PON. It is found that a cascade of linear filters can be combined into a single, truncated, linear, adaptive filter with negligible impact on receiver sensitivity. Additionally, by utilizing a multiplier-free tap weight update algorithm, the overall complexity of a digital coherent receiver can be significantly reduced, making it attractive for use in an ONU. Matched filtering, chromatic dispersion compensation, and polarization tracking are all performed by the adaptive equalizer. The performance of this low-complexity, multiplier-free equalizer is experimentally verified for 3 GBd polarization division multiplexed quadrature phase shift keying (12 Gbit/s) in both a back-to-back configuration and transmission over 100 km standard single-mode fiber.