100G backers divided on best approach

This article was originally published on fibresystems.org.

This summer the Optical Internetworking Forum (OIF) decided to forge ahead with the standardization of optical modules for 100 Gbit/s networking based on a modulation format called dual-polarization quadrature phase-shift keying (DP-QPSK).

DP-QPSK cuts the symbol rate on the fibre by a factor of four, by transmitting two bits of data per clock cycle on each of two polarizations. Thus a 100 Gbit/s data rate is reduced to a symbol rate of just 25 Gbaud on the fibre, which mitigates the impact of chromatic dispersion — an effect that causes high-speed pulses to spread out as they travel down the fibre.

“We have selected an implementation approach supported by a critical mass of photonic component vendors and users,” claimed David Stauffer of IBM, and the OIF’s PLL Working Group chair.

However, it seems that no sooner did the OIF give its blessing to DP-QPSK, than the dissenters started to crawl out of the woodwork.

One of the first vendors to actively back other options was ADVA Optical Networking, which is taking part in a research initiative called 100 Gigabit Ethernet Transport (100GET). The partners of the 100GET-METRO sub-project, which include optical software company VPIsystems, will use numerical simulations to investigate the robustness of different modulation formats to fibre impairments including higher-order effects like second-order polarization mode dispersion (PMD) and chromatic dispersion slope. Their aim is to find the best approach for 100 Gbit/s transport in metro networks, which may well have different requirements from long-haul networks.

Similarly, DWDM vendor Ericsson, which recently started talking about its 100 Gbit/s ambitions, is taking part in a different sub-project under the 100GET umbrella. The project participants are keeping a totally open mind on which modulation format would be best for DWDM transport: orthogonal frequency division multiplexing (OFDM), sub-carrier modulation (SCM), and multi-level phase shift keying (M-PSK) are all under investigation.

“100G transmission presents considerable challenges,” said Rodolfo Di Muro, optical product marketing manager for Ericsson’s business unit networks. “It is anticipated that 100 Gbit/s will coexist with 40 Gbit/s or 10 Gbit/s in the same network with minimum architectural change. Technology is not mature yet, and we can anticipate that one size does not fit all networks.”

Sensitive issues
Even Len Bosack, founder and owner of optical firm XKL, is ready to stick his oar in. He believes that DP-QPSK isn’t going to be up to the job at 100 Gbit/s because the optical signal will be too sensitive to chromatic dispersion and PMD.

“I keep suggesting to people that what you really want is polarisation-diverse 8-QAM [quadrature amplitude modulation, a name for advanced phase-shift keying formats], and that will get you a dandy 100G transmission system, which will go quite considerable distances,” he told fibresystems.org. 8-QAM is a coding scheme that transmits three bits of data per symbol. Combined with polarization multiplexing, it has a symbol rate on the fibre of 15 Gbaud.

“Yes you’ve given up some receiver sensitivity to get there, and the transport is a bit more complicated, but in my view it’s a tradeoff that is both feasible and desirable,” he adds.

Ironically Nortel, which is an ardent supporter of DP-QPSK at 40 Gbit/s, says it won’t be using the same modulation format at 100 Gbit/s. Nortel wants its 100 Gbit/s solution to be an overlay solution for 10 Gbit/s-engineered networks, which means slashing the symbol rate on the fibre from 100 Gbit/s right down to 10 Gbit/s.

In fact, so far fibresystems.org has only found one vendor that’s prepared to back DP-QPSK at 100 Gbit/s. Alcatel-Lucent doesn’t favour DP-QPSK for 40 Gbit/s transmission, due to the possibility of optical crosstalk with 10 Gbit/s signals running on the same fibre, which would have the same baud rate as a 40 Gbit/s DP-QPSK signal. However, the giant vendor thinks that DP-QPSK has potential at 100 Gbit/s.

“At 100 Gbit/s DP-QPSK suffers much less impact from nonlinear effects, so a modulation format that experiences issues at 40G becomes recommended at 100G,” said Paolo Ottolenghi, senior product manager at Alcatel-Lucent.

Reproduced with permission. © Institute of Physics and IOP Publishing Ltd.