Brian Bailey - December 12, 2012
Silicon photonics is most certain a hot topic area these days as a number of companies work out how to overcome the challenges of putting optical components onto silicon using standard production techniques. The reward is data transfer rates of 25 gigabits per second and up which will provide a significant increase for communications, Ethernet LANs, high performance computing and many other applications.
The problem has almost been created by Moore’s law which has resulted in long interconnect path lengths, increased wire delays and power densities approaching those of a nuclear reactor. Clock frequencies had stopped rising and voltages are flattening out. Increases in compute capabilities have come from deploying multiple processors but these need to communicate and that is creating a bottleneck. Also, the communications hierarchy now spans from inter-chip to the cloud and it would be convenient to have a single technology that could address all of those needs. As an added advantage, silicon is an excellent optical material because:
- there is a very large refractive index difference between Si and SiO2. This means that waveguides can be constructed and that tight bends can be created making them suitable for integration on-chip with other circuitry.
- Transparent near IR
- Can be built on a conventional CMOS process
IBM has said that it has created test chips using standard silicon fabrication processes at 90nm. Kotura, a small startup serving this area, is claiming a 100 gigabit transceiver and hopes to increase that to deliver a terabit per second between servers. I published an article by Kotura over in the EDA Designline this week that also provides a lot more information about some of the components necessary to create these systems, such as lasers, wavelength division multiplexors, modulators and detectors.
It is clear that this is seen as an area for potential regional excellence with governments in Europe, Canada and Asia setting money aside to help research in this area and to help foster cooperation between companies that have pieces of the technology necessary for complete solutions. Several companies in the US are working in this area including IBM, Intel, Luxtera, Kotura as well as DARPA funded programs.
There are many hurdles that have to be overcome in order for this to become mainstream technology, and at this time everything is very much custom design. There are no PDKs or design rule decks, nor are there tools that will help with the design and verification of these systems – perhaps another area in which the EDA companies can carve out a niche.
If this is an area that is of interest to you, then there is a course available online, put together by the European Helios project. While it covers a lot of ground it almost requires some knowledge of the subject matter already as speaker notes do not accompany the slides. This makes it somewhat difficult to get the point made on some of the slides. It is available here.
Brian Bailey – keeping you covered
If you liked this feature, and would like to see a weekly or bi-weekly collection of related features delivered directly to your inbox, sign up for the IC Design newsletter.