Connections +
Feature

10GBASE-T and the IEEE

With the formation of a new task force, a 10 Gigabit Ethernet standard will be possible over twisted pair copper cabling. The plan is to bring a whole new world of speed and throughput to all areas of the network. Luc Adriaenssens explains.


July 1, 2004  


Print this page

There seems to be a never-ending need for increased speed and bandwidth and as companies try to keep up, many are realizing that their existing infrastructures are not adequate to support growth in the months and years to come.

As it becomes more popular, Ethernet

continues its relentless pace toward tenfold data rate increases. Just as Gigabit Ethernet deployments on UTP start to pick up increased momentum and 1000BASE-T interfaces become commonplace in switches, servers and even laptop PCs, a 10 Gigabit Ethernet standard for twisted pair cabling (named 10GBASE-T) is in development already.

With previous Ethernet data rate leaps, there was uncertainty regarding the media requirements and cabling categories that will be involved in 10GBASE-T standards, and that holds true this time around, as well. These elements have become the subject of industry debate, and, if the industry isn’t careful to explain standards, the market could become confused.

As has been the case in the past, a task force is helping to establish parameters. The IEEE 802.3an 10GBASE-T Task Force devoted to this project is focused on making the widest possible use of UTP cabling, both with current and enhanced technology.

The end result should be a standard that makes possible cost-effective, high-performance Ethernet solutions.

With the formation of the IEEE Task Force, a 10 Gigabit Ethernet standard will be possible over twisted pair copper cabling. This will bring a whole new world of speed, throughput and opportunity to all areas of the network.

To achieve “Broad Market Potential” (a key aim of the IEEE Task Force), the main challenge for the task force is to set the requirements for the complete solution, both electronics and cabling, that ensure 10 Gb/s capability.

While it is anticipated that Category 6 cabling will support 10GBASE-T (albeit to a limited 55 meter distance), the task Force has set a challenge to cabling vendors to develop solutions capable of supporting 10GBASE-T over 100 meters, with support for 4-connector channels and guaranteed performance out to 625 MHz.

If vendors are able to meet these challenges, the costs and efficiencies involved in the use of 10GBASE-T could be even more beneficial than initially believed.

The key targets for this new cabling channel specification include:

– IEEE Model No. 1 for Insertion Loss and ANEXT for 100-meter 4-connector channel performance:

– Extrapolation of all Category 6/Class E channel performance requirements to 625 MHz;

– Complete 100-meter 4-connector channel solution tested with sophisticated system-level tools that allow the modeling of worst-case configurations for all parameters, Evolutionary improvements in component design and performance, and

– Complete cabling system design and installation instructions

The March 2004 formation of the task force marked the culmination of a one-year IEEE Study Group period to determine the objectives for 10GBASE-T. While an original objective of the Study Group was to support 10 Gb/s transmission over Category 5e cabling, it became clear early in the process that a minimum of Category 6/Class E performance would be required.

However, the maximum distance over which 10 Gb/s transmission can be supported over Category 6/Class E cabling is still uncertain (with a minimum objective of 55 meters). It also is clear to the IEEE Task Force that there is a strong global preference for UTP cabling systems that avoid the shielding and grounding complexities of screened alternatives.

At the formation, the target performance objectives for 100-meter 4-connector channels were based on Category 6/Class E channel performance parameters extrapolated to 625 MHz, with the addition of alien crosstalk (the noise from adjacent channels) specifications.

Over the last few months, there has been much progress toward “locking-in” cabling requirements (in ISO/IEC and EIA/TIA) so that the work on transceiver designs can be based on viable cabling performance models.

Transmitting 2.5 Gb/s on each of the four pairs is no easy task. It requires multi-level coding that transmits multiple bits per Hertz of bandwidth.

Even with the planned sophisticated coding schemes, the minimum bandwidth exceeds 250 MHz and likely will approach 625 MHz. Sophisticated Digital Signal Processing (DSP) techniques also will be required to reduce the effects of within-cable impairments such as return loss and crosstalk (NEXT and FEXT).

The effects of alien crosstalk cannot be reduced effectively by the electronics and must be reduced to tolerable levels within the cabling.

The task force has been exploring these issues and interfacing with the ISO and TIA cabling standards to converge on its cabling channel requirements. It has progressed steadily on improving the specificity of the cabling channel requirements.

At the January meeting, the IEEE committee unanimously approved minimum channel requirements based on Category 6 extrapolated to 625 MHz. At the March IEEE meeting, the channel requirements were reaffirmed and appended with a set of alien NEXT requirements.

The powersum Alien NEXT requirements depend on the insertion loss of the channel as shown in the set of models above.

Agreement on some initial models is a significant milestone. There is general industry agreement regarding the application of Model No. 1 to the development of new UTP cabling for 10GBASE-T (sometimes called “augmented” Category 6/Class E).

Model numbers 2, 3 and 4 are based on existing Category 6/Class E channel performance requirements extrapolated to 625 MHz, with ANEXT requirements based on the insertion loss at specific or calculated lengths.

Model numbers 2 and 4 are designed to accommodate as much of the embedded base as practical. The current belief is that embedded Category 6/Class E cabling will support 10GBASE-T transmission for at least 55 meters.

Longer distances should be achievable on existing cabling exhibiting higher performance. Simple techniques such as cord upgrades also should extend distances. Model No. 3 is an alternative to Model No. 1 but may not be achievable on Category 6/Class E UTP, even with additional mitigation.

IEEE liaison letters have been sent to both TIA and ISO cabling committees with the request to further develop and standardize these requirements. Both committees have indicated that they are eager to complete this work and plan to provide more detailed specifications, measurement procedures, field-testing guidelines, etc.

Much progress has been made toward defining the cabling channel requirements to support 10GBASE-T over new as well as existing cabling.

Although clarity has improved with the agreed-upon models above, many remaining questions must be answered before “guarantees” can be made with confidence and credibility.

Some companies have developed solution prototypes. However, testing has detected unexpected phenomena in the channel due to complex interactions among components. This insight highlights the challenge of achieving robust 10 Gb/s performance using conventional design techniques.

In addition, solving component challenges in isolation does not translate to a guarantee of end-to-end channel performance. Refinements and system tuning optimizations will continue, and companies will eventually deliver a UTP solution that meets all 10GBASE-T requirements. When selecting such a solution, end users should assess claims of 10 Gb/s support critically.

Ultimately, those able to take advantage of a 10GBASE-T solution will enjoy installation and operations savings over comparable fiber solutions. In addition, 10Gbps switches cost less than SONET equipment. Expected early adopters of 10GBASE-T solutions include data centers, educational institutions, storage area network (SAN) providers and financial institutions.

IEEE 10GBASE-T Model Channel Insertion Loss ANEXT @ 100 MHz
No. 1 100 m Class F Insertion Loss 60 dB
No. 2 55 m Class E Insertion Loss 47 dB
N
o. 3
100 m Class E Insertion Loss 62 dB
No. 4 55-100 m Class E Insertion Loss Calculated based on IL (47-62 dB)

Luc Adriaenssens is vice president of Systimax Solutions R&D and head of Systimax Labs. He holds 27 U.S. patents, and he has an additional five pending patents in the fields of cable design, connector design and EMC.


Print this page

Related