As the density of devices in enterprise networks and data centers increases, there is an emerging need for low-cost, 10G bit/sec Ethernet over twisted-pair cabling. A 10GBase-T standard in the works will support 10G Ethernet over new wiring at distances of up to 330 feet and existing wiring at distances of up to 182 feet.
An IEEE task force first met in March to begin hammering out the new 10GBase-T, or 802.3an, standard. The first draft is scheduled to be ready this year, and the group expects to finalize the standard in 2006.
To run multi-gigabit data rates on four-pair copper cabling, sophisticated digital signal processing is used to eliminate the effects of cross-talk between pairs of cable and to remove the effects of signal reflections. But it's difficult to eliminate noise that is external to the cable, such as electro-magnetic interference from outside sources or adjacent cables. Cable-to-cable noise, or alien cross-talk, prevents wiring from reliably operating under worst-case 330-foot conditions.
To address this problem, a Telecommunications Industry Association (TIA) subcommittee for cabling specifications is working to provide additional requirements for cabling installations up to 182 feet.
To support distances of 330 feet, as required in the U.S. market, substantial augmentation to the existing cabling specifications are required. This will result in a separate cabling system designed for 10GBase-T. The TIA cabling subcommittee currently is addressing this task.
The subcommittee must develop a new Physical Layer entity (PHY), which interfaces with the existing 10G media access control and Gigabit Media Independent Interface in the IEEE model. The PHY contains the functions that transmit, receive and manage encoded signals that are recovered from cabling systems.
The PHY will be based on pulse amplitude modulation (PAM). Using PAM, information is encoded by a stream of pulses with discrete amplitudes. This is the same modulation technique currently used in 100Base-T and 1000Base-T, but the symbol rates and digital signal processing techniques will be enhanced.
The question is, what type of PAM encoding will be used? There are two proposals on the table, PAM12 and PAM8. PAM12 operates at 825M symbols per second. It has the advantage of using lower baud rate than PAM8, which operates at 1000M symbols per second. Presentation materials show that PAM8 might be able to tolerate 20 percent more noise than PAM12. Because PAM8 separates sampled signal values into only eight distinct voltage levels, the difference between each level is greater than it would be for PAM12, which makes it easier to determine the level of each sample.
Also being considered is a technique known as Tomlinson-Harashima Precoding (THP), which takes the known signal values and equalizes them before transmission, rather than best-detected values. Low-density parity check, a common coding approach that decreases the signal-to-noise ratio needed to achieve a required bit error rate, works well with THP to reduce the amount of required transmit power.
There are many building blocks that make up the 10GBase-T system architecture. The IEEE task group has made a great deal of progress toward its goals and is continuing to flesh out technical proposals that will make 10GBase-T a reality by the summer of 2006.
Alexander is a senior project engineer for Nexans SA, and Debiec is a technical marketing specialist for Berk-Tek, a Nexans company. They can be reached at firstname.lastname@example.org and email@example.com, respectively.