Don't count 40G out!
The telecommunications market is going through an unprecedented economic dive and 40G — having to overcome all of these economic issues as well as the technical challenges of dispersion and optical signal-to-noise ratio — is being hit the hardest. Do you or your customers really need the extra capacity it offers? Why not just use four 10G ports? What about all those dispersion issues? Is it not time to sing the funeral march for 40G; 40G rest in peace — 40G RIP.
But while the pundits have formed a consensus view that 40G is an esoteric long-haul technology fraught with enormous economic and technical hurdles, the relentless march of Moore's law has produced breakthroughs making 40G not only feasible, but economical as well. Yields are up on 40G semiconductors, electro-absorption modulators (EAMs) at 40Gbit/s are available, low-phase noise clocks are solving jitter issues, and small, cost-effective chromatic and polarisation dispersion compensators are proving to be a reality. Add to this list framer and forward erro correction chips operating at 43Gbit/s, while link lengths are pushed past 1000km before 3R regeneration. I know: Agilent builds the equipment that tests these products.
Even so, it can be argued that 40G will be delayed due to another market reality. Long-haul transport has been the application that has launched each successive breakthrough in speed. Today, there is over-capacity in long haul and so many systems with few wavelengths lit that it would be better to light up the dark wavelengths, when more capacity is needed, than lay new fibre.
The adoption rate of 40G will be attenuated in the long-haul space because of this reality. But the assumption that long-haul transport is the only catalyst for 40G must be questioned. Economic breakthroughs are being made at the opposite end of the spectrum, on the very short reach (VSR) links that interconnect equipment within a Tier 1 central office or Point of Presence.
For VSR applications, all below 2km in reach, no dispersion compensation is needed and thus 40G works. A single 40G EAM can now do the standard NRZ modulation format. All these technical hurdles are eliminated alongside the associated cost issues. The one thing 40G modules must offer is lower cost than four 10G modules; that crossover point is now — Moore's law is relentless.
Coupled with these advances is a standard 40G electrical interface called SFI-5, the standard electrical interface for 40G transponder modules and OC-768/STM-256 framer chips. SFI-5 has allowed the first Multi-Source Agreement for 40G transponder modules, allowing economies of scale. These SFI-5-based 40G transponder modules will allow the lowest cost-per-bit connections; hence this will see these emerging products being almost completely focused on VSR applications. This is the application that will be the catalyst for 40G because it offers a clear economic benefit.
Since the VSR application mentioned is a core application, it would seem logical that core long-haul transport would follow. Once again, history may be a poor judge. Metro, specifically the metro core, could very well be the next application. The economics of going from 2km to 40, 80, or 100km links are not that different; a single amplifier with a dispersion compensator at the end of the fibre is required, but the fundamental line systems are not significantly different from those required for 2.5Gbit/s or 10Gbit/s DWDM systems. In fact, 40G may allow carriers to avoid DWDM altogether, using a single ring or stacked rings, and avoiding all the optical muxing and demuxing at each terminal. Now that would mean real cost savings.
This is the age of the pragmatists; whatever is most cost effective overall will be adopted. The deployment of 40G technology will be different from that of 10G and 2.5G before it, as it will not flow through the same applications in the same order. You can listen to the pundits and ignore 40G. But ignore it at your own risk.
Larry Desjardin is Agilent Technologies' High Bandwidth Program Manager