Coherent Detection

A lot of money has been invested in 40G and 100G transceiver technology, but 10G's relative price performance advantage continues to improve. There is a simple explanation - building high speed transceivers isn't simple.

The current generation of 40G transceivers have economics that is appropriate only for deployment by large tier-1 operators. The technology is too complicated and expensive for high volume deployment.

The new generation of transceiver technology that is targeting 100G uses a radically different approach. The difference is the use of coherent detection. These transceivers rely more on digital electronics, than on analog photonics.

Digital electronics has the potential to improve the manufacturability and robustness of these transceivers. Coherent transceivers use adaptive digital signal processing to take on tasks that formerly required precision photonics. This radical change was necessary because traditional photonic methods just don't work at 100G. Adaptive electronics also eliminates the need for external dispersion compensation and improves link budgets. More important, adaptive digital processing takes on complex configuration and operations functions. With coherent detection transceivers automatically adapt to transmission conditions.

The challenge is that the electronic technology required to make coherent detection work is at the very limit of what is possible. While operation has been demonstrated using racks of FPGAs, a lot of work is still required to squeeze all of the required functionality into a few chips. The amount of logic required, and the extreme clock speeds, demand massive chips built using 45 nm or smaller technology. By any measure this is a high stakes high risk business.

We are told that the A/D, DSP and Gearbox chips will be in photonic component vendor labs sometime next year. Large TEMs that target tier-1 carriers may get samples of early generation transceivers in 2011 too. Ironing out the bugs will take time, so deployment should be expected to ramp slowly. A deployment power curve isn't likely until there is supply side price competition, and that is unlikely before we are well down the learning curve. Perhaps sooner, but likely by 2015, this technology will be ready for showtime.

The coherent detection approach may be attractive at 40G too. The electronics will be smaller and simpler than required for 100G. This may be a lower cost, lower risk path toward 100G. We need to watch to see if coherent 40G transceivers become available. This scenario will require R&D dollar strapped transceiver vendors to make some difficult choices.

The economics of 10G transceivers will continue to improve. Even with the most optimistic projections, 40G transceivers will require more than 4x the parts of 10G transceivers, so the cost per bit/second at 10G will be lower than the cost at 40G. Synchronous detection means that 100G transceivers will ultimately have better economics than 40G transceivers. The question is when will 100G transceivers be cost effective?

The Opex at higher transceiver speeds has the potential to be significantly better, just because there are fewer wavelengths, fibers and transceivers to manage. These Opex saving will be more important to large network operators, who will drive this market. Smaller datacenters and Tier-2 and tier-3 network operators are likely to see less relative benefit, and might be more tentative when deploying more costly 40G and 100G transceivers. Volume deployment in smaller networks is likely to lag the big networks by a few years.