The latest Xyratex patent applications enable embedded waveguide technology to be used competitively inside an optical backplane design. They come from a StoreLite research project.
"One of the main issues preventing the wider deployment of this technology has been the inability to adequately and cost effectively align the photonic signal path through a connector or other component to a waveguide buried within the layers of an optical backplane or plug-in card," said Ian Johnson, Xyratex' Chief Scientist. "Even very small misalignments will result in substantial signal losses thereby reducing the performance of the ultra high speed interconnect channel. Further, conventional micro-manipulator alignment techniques are perceived to be far too expensive to be practical."
The IT and communication industry is continuously deploying ever increasing numbers of high-speed devices, densely packaged within high-availability products. For example, the next generation Fibre Channel standard is researching into an 8 Gbit/s data rate, the next generation SAS/SATA standards are looking towards 6 Gbit/s and the Infiniband standard has already started deploying 10 (4x) and 30 (12x) Gigabaud connections. In addition, the emerging PCI Express standard can be configured in multiples of 2.5 Gigabaud channels, and 10 Gigabit Ethernet is beginning to be deployed more widely.
Johnson said: "These higher data rates make the provision of high speed interfaces and the subsequent design of systems within specified cost targets much more difficult due to the challenge of managing conflicting requirements such as crosstalk and RFI emissions."
Most of these high performance interconnects make up their high speed channels with multiple 2.5 Gigabaud links today, but Double Data Rate (DDR), i.e. 5 Gigabaud individual links, are already emerging and Quad Data Rate (QDR), 10 Gigabaud individual links are also being discussed in some of the standards. Further, the use of copper as the source material to carry these high speed electrical signals becomes problematic at speeds higher than 3 or 4 Gigabaud.
More costly dielectric materials are usually required above these speeds to ensure signal integrity over useful distances. This is to sustain the low bit error rate requirements of these high speed digital communication channels. Additionally, as the individual electrical links approach or exceed 10 Gbit/s, any purely electrical implementation becomes very difficult and the costs begin to significantly rise. A recent research study has suggested that the optimum cross over to optical-based solutions could be as low 6.25 Gigabaud per link.
Historically photonic solutions have been considered too expensive, but now optical component costs are falling and densely packaged VCSEL lasers can be purchased at prices approaching $1 per Gb/s in volume. Below this cost point, the issue begins to move toward the system integration implementation problems of high density and low cost optical paths.
To address this, Xyratex has researched a completely new method of repeatable line card insertion and cost-effective connection to an embedded optical backplane. It has successfully created a practical design demonstration with respect to high-speed data transfer across multimode polymer waveguides.
Johnson said about this: "We have successfully focused our attention on the development of a low cost optical connector with a particularly cost-effective waveguide alignment and a highly practical daughtercard accommodation methodology that can be widely used in many storage, networking and telecommunications applications."
The Xyratex design satisfies the following requirements:
* High bandwidth connection with low latency optical interconnect
* Reliable high precision connection with immunity to movements between line card and backplane (vibrations, air flow and PCB deformation due to thermal and mechanical stresses)
* High connection repeatability > 100 cycles
* Scalable to accommodate more channels and more connectors per board
* Low cost in mass production
Steve Thompson, Xyratex CTO, said: "The three patent applications filed and three being prepared ... will enable opportunities for innovative system architectures within our strategic roadmap."
He anticipates that the market will start to emerge in 2007 with volume product becoming commercially available in 2008. There may also be opportunities for earlier deployment in certain niche applications where device density or RFI issues are particularly difficult.
The Xyratex Storlite research project was a 30-month optical research and development program based in the UK and supported by the Department of Trade and Industry. It involved the collaboration of University College London, Edinburgh University and Xyratex industrial research partner, Glasgow-based Exxelis.