Miniaturized monolithic transceivers promise to boost Ethernet capacity

Internet usage is changing. The number of individuals who are going online is on the rise and they are using this technology for an increasing range of activities, such as watching videos and gaming via broadband access.

These changes in behavior are driving a rocketing rise in data transfer that places greater strain on internet infrastructure. To reduce this burden, developers of technologies for fiber-optical communications are continuing to search for new ways to increase transmission capacity. This quest involves all parts of the fiber-optical communication chain, from backbone networks that provide access for individual users, to local connections between computers in the same building.

Linking computers together in a business environment is increasingly undertaken with Ethernet technologies. In the 1990s this type of infrastructure provided data transfer at 10, 100 and finally 1000 Mbit/s, and Ethernet speeds are now up to 10 Gbit/s (10 GbE).

European efforts
The next step will be 100 GbE. This race is already underway, and organizations such as the Institute for Electrical and Electronic Engineers (IEEE) and the International Telecommunication Union’s telecommunication standardization sector (ITU-T) are currently developing standards for fiber-optical transmission equipment that will enable those higher- capacity networks. Standards, along with multi-source agreements, are essential for driving the development of faster Ethernet networks, because they enable the builders of networks to source components from a variety of manufacturers.

Fabricating components for 100 GbE is very challenging because the route to higher speeds involves shrinking the component size. Scaling can cut capacitances, inductances, resistances and transit times, but success demands increased control of the dimensions, which in turn pushes up the complexity of fabrication.

To spur faster transmitters and receiver development, the European Union started to fund a handful of projects from 2005 to develop suitable technologies. These efforts, which marry the talents of universities with the expertise of companies involved in the telecommunications industry, include: a 75.7 million ($7.7 million) effort entitled “integrated photonic mm-wave functions for broadcasting connectivity (IPHOBAC)”, which combines radio and optical technologies; a project called “optoelectronic integration for 100 Gbit/s Ethernet optical networks (GIBON)”, which is led by the Alcatel-Thales III-V lab and backed by 71.9 million of funding; and a 76.6 million project led by Alcatel-Lucent, “100 Gbit/s carrier-grade Ethernet transport technologies (100GET)”.

We are also developing components for 100 GbE through a 73.8 million, three-year project called HECTO (high-speed electro-optical components for integrated transmitter and receiver in optical communication). This effort, which kicked off in November 2006 and is partly funded by the European Union, draws on the expertise of industrial and academic researchers from nine institutions (see box “The roles of the HECTO partners”). Its focus is the scaling of commercial component architectures employed for 10 GbE, so that they can offer a costcompetitive option for the deployment of 100 GbE.

The HECTO project has four primary objectives. The first is the full development of packaged transmitters and receivers that are suitable for optical systems based on serial 100 GbE signals. We are focusing on a single data stream at 100 Gbit/s, which needs a total transmission of about 110 Gbit/s to accommodate error-correction coding. Another target is to determine the specifications for all interfaces of the photonic components. This has to take into account emerging relevant standards, the identification of application areas for the components and the impact that this has on their specifications. The other two goals are to test the fully packaged transmitters and receivers in laboratory system test-beds and perform field trial tests; and to exploit the results of the project, through component vendors Syntune and U2T Photonics, which are consortium partners.

A two-pronged attack
During the HECTO project we have intentionally developed two distinct generations of hardware. As the project has progressed, we have modified the exact specifications for the transmitter and receiver modules. These adjustments are partly driven by a greater awareness of how we can optimize our technology, but they also allow us to accommodate any changes in standardization.

We evaluated the performance of our first generation of semiconductor devices last year. Although these chips could be viewed as just a test run that provides useful input to a subsequent set of devices, the results produced by these components have been published and they have attracted significant international interest. Tests are now underway to evaluate the performance of all parts of the transmitters and receivers in optical networks.

One of the initial highlights of our project is a transmitter that was produced by KTH and Syntune, which combines a light source and a lightintensity modulator on one chip with dimensions of just 1 mm × 0.5 mm. This features the first modulator design capable of 100 GHz that can be monolithically integrated with a distributed feedback (DFB) laser.

The DFB laser design employed in this transmitter spans the 1530–1550 nm range. This allows the source to be used in wavelength-division multiplexed fiber-optical communication systems that simultaneously transmit the light pulses from several lasers down a single fiber.

The modulator features a quantum-well stack and changes to the voltage applied across the device produce strong variations in absorption at the laser’s wavelength. A distributed microwave design is used in the travelling-wave electroabsorption modulator (TW-EAM), with the electrical signal following the light path and interacting with it over a length of about 0.2 mm. The short interaction length cuts electrical power consumption and increases the maximum bit rate.

We are now well into the final year of the HECTO project, and one of the outstanding tasks is to take our second-generation components and evaluate them in trials at Acreo’s test-bed in Kista. It is likely that the modules that were built in this project will eventually be deployed in more general measurement equipment for high-speed components for fiber-optical communications.

100 Gbit/s and beyond
The optical components that are built in the HECTO project use the same form of optical signals used today in 10 GbE. With this approach, the light intensity is switched between a high and a low power level to encode every binary bit in the data stream. This approach, known as on-off keying, is an economically viable solution at lower bit rates. However, at data rates of 100 Gbit/s and beyond the construction of the electrical circuit and electrical interconnects is very challenging, because they have to handle a frequency range that spans 0 Hz to around 100 GHz. This is 10 times as wide as that needed for 10 GbE.

To span this wider range we have miniaturized some parts of the electrical circuitry because this reduces loss and microwave reflections. For example, the diameter of the coaxial cable is just 1 mm. However, making this cable smaller pays the penalty of increased fragility and leads to greater electrical loss due to the finite conductivity of gold.

Increasing bit rates beyond 100 Gbit/s with the same type of solution would require further refinement to the electrical parts of the transmitters and receivers, and this would be very challenging. So we believe that it is more likely that other approaches will be used to move to faster speeds. One further downside of on-off keying is that it needs a relatively large optical bandwidth, and this is a weakness in wavelength-division multiplexing systems that use a narrow channel spacing to maximize capacity.

Several partners in the HECTO program are also participating in the 100GET project, which is investigating the benefits of other forms of optical signal. Options for improving the spectral efficiency of the light signal include schemes that modulate the intensity of the light and its phase. This effort is still at an early stage, but there are no obvious reasons why it will not be possible to slash the size of the relatively bulky components currently used in the lab to those realized in the HECTO project. Massive year-on-year increases in internet traffic are sure to continue, and these types of components will be needed to deliver the required expansion of internet capacity.