The German vendor has been invited to present the latest results from the EU-funded project 100GET-METRO, which it has been working on since September 2007 (see EU projects explore 100G alternatives).
Naturally, as a vendor of metro optical equipment, ADVA's interest in 100 Gbit/s is for metro deployments, which seem to have been largely neglected by current 100 Gbit/s development efforts.
The Institute of Electrical & Electronics Engineers (IEEE) is defining physical implementations for 100 Gigabit Ethernet over distances up to 40 km, while the Optical Internetworking Forum (OIF) is aiming to standardize 100 Gbit/s components and modules for long-haul networks, which target distances in the 1000 km range.
"From our perspective that leaves a window from 40 to 600 km," says ADVA's marketing director Stephan Rettenberger. These kind of distances are the sweet spot in Europe.
"Of course anything that can go 1000 km can go 600 km, but maybe that's not the ideal solution," Rettenberger suggests. The metro market is more cost-sensitive than the long haul, with transponders being the most expensive part of a metro optical platform.
The right price point will only be achieved if complicated transmitter or coherent receiver structures can be avoided. Not only are these costly, but the digital signal processing chips would generate more heat than a metro box can handle.
With this in mind, ADVA embarked upon the project with a couple of additional requirements: to transmit 100 Gbit/s on a single wavelength, and to do this using off-the-shelf components. "Anything that can use existing technology will not only result in a better price point, but a better time to market as well," Rettenberger points out.
The solution the vendor settled upon is a new modulation format that combines differential phase-shift keying (DPSK) with 3-level amplitude-shift keying (3ASK). In this scheme the symbol can take one of six possible values (see figure); two consecutive symbols are combined to transmit five bits of data.
This creates a scheme with an efficiency of 2.5 bits per symbol, and a modulation rate of 40 Gbaud. Lowering the symbol rate on the fibre is essential to make the system more tolerant of dispersion and other fibre impairments.
The beauty of the scheme is that it has relatively few components. On the transmit side, the signal can be generated with a single Mach-Zehnder modulator. (A second modulator, or pulse carver, could be added to shorten the pulses and boost performance.) On the receive side the phase difference between two symbols is detected in a delay-line interferometer with balanced detectors, while a separate photodetector is used to detect amplitude.
"40 Gbit/s DPSK technology is there; you can buy modules that do this, and all you need to add in is the intensity detection so you can distinguish between the three amplitude levels," explains Jörg-Peter Elbers, ADVA's vice-president of technology.
The invited paper OTuN6, which will be presented at 6pm on Tuesday March 24, describes simulation results on the tolerance of the DPSK-3ASK modulation scheme to transmission impairments and filter cascades.
ADVA has built a bench-top version of the transmitter and receiver and is working on the next step — "hard results" in the form of bit-error rate measurements and eye diagrams to show that the simulated performance can be achieved with real components.
Over the past few weeks, ADVA's engineers have been working flat out to get good results, and although they are getting really close, unfortunately they couldn't get the results they wanted in time to meet the March 11 deadline for submitting an OFC Postdeadline Paper.
Ultimately the vendor hopes to turn its bench-top set-up into a compact transceiver. "Ideally we expect it will have not more than twice the size of a 40 Gbit/s solution, even though it will have more than twice the capacity," says Rettenberger, adding that component integration could reduce the size even further.
Still no consensus
While everyone rushes to adds a new acronym to the ever-expanding array of options for 100 Gbit/s transmission, it raises an important question: does it really matter if the industry can't come to a consensus? (see 100G backers divided on best approach).
Rettenberger thinks it doesn't, at least not in this case. The client-side interfaces must meet the relevant standards, while the line side is likely to be proprietary. "That's not a problem because most systems are deployed in a bookend fashion, where the transmitter and receiver pair come from the same vendor," he says.
Standardization is generally a good idea because it helps component vendors to increase volumes, which brings down prices. But since ADVA's 100 Gbit/s solution is constructed from off-the-shelf components designed for 40 Gbit/s, the issue is neatly avoided.