OFC ‘09, ROADM technology and the future network

This week’s entry comes from Ian Clarke. Ian is one of our key technical guys in our Sydney facility, where Finisar develops and manufactures our WSS (Wavelength Selective Switch) modules.

******************************

OFC 2009 felt like a giant planning meeting for 100 Gb/s systems. Everywhere could be heard debate and proposals for components, protocols and modulation methods. The most difficult component is clearly the transponder. In particular, the ASICs for the receivers look very challenging to build. How do you build an analog–to-digital converter that runs at 60+ Gb/s; and when you do, how do you process the data fast enough?

However, my real interest was the requirements for wavelength selective switches (WSS), which also received plenty of attention. The most interesting requirement for an engineering manager like me is to know when we need to deliver these parts.

On Sunday, Glenn Wellbrock from Verizon described a series of trials of 100 Gb/s Ethernet signals. On Monday Jim King from AT&T described their initial build out of their 40 Gb/s system between “NFL cities” (the 25 biggest cities in the US). He stated that he could use 100 Gb/s channels now if he could get them, but in reality he didn’t think he would be deploying them seriously until 2011. Similarly, a Nokia–Siemens paper suggested that 2012 would be the critical year for 100 Gb/s systems. However, Glen Wellbrock’s message was that they wanted to be building 100 Gb/s ready systems that ran 40 Gb/s until the transponders were ready. This suggests that we are likely to see requirements for ROADMs, amplifiers and other components to be “100 Gb/s ready” quite soon.

This raises the question: what sort of WSS should we be building for 100 Gb/s systems? The required bandwidth depends on the modulation format that finally dominates. Last year, Chandresekhar’s paper from ALU Bell Labs used PM-RZ DQPSK (polarization multiplexed return-to-zero differential quadrature phase shift keying) needing a huge 44.5 GHz bandwidth (try to fit that in a 50 GHz window!). Renaudier’s paper (also from ALU, this time from France and also published last year, but mentioned in this year’s workshop) needed only 35 GHz to achieve zero penalty on a PM-RZ-QPSK with coherent detection. 16-QAM (quadrature-amplitude modulation) systems are more efficient, but are so complex to build that I feel would be unlikely to be the first system used. Other contenders are 8-PSK (phase shift keyed) and OFDM (orthogonal frequency division multiplexing), but these have their own issues. Coherent detection systems have advantages in CD chromatic dispersion) and PMD (polarization mode dispersion) tolerance. If we assume that the initial 100 Gb/s systems will use a coherent PM-RZ-QPSK modulation format, which is a reasonable guess at the moment, then targeting a ±17.5 GHz real, concatenated bandwidth for 2010 would be reasonable.

Ian Clarke is an Engineering Manager for Finisar Australia (Ian.Clarke@finisar.com).