The road to 10GBE transmission has been a long and treacherous one, writes George Georgevits. So how did we get here?

In Part 1 of this series we reviewed the history of data cabling, from the days of the earliest mainframe-based computer networks with their equipment dependent proprietary cabling, through the early days of Local Area Networks (LANs) using Category 3 and 4 unshielded twisted pair (UTP) media to today’s 100Mbps (100Base-T) and Gigabit Ethernet (GBE or 1000Base-T) LANs using Cat 5e media.

Gigabit Ethernet (GBE) Networks and Beyond

The Cat 5e standard (ANSI/TIA/EIA-568-B.2) was specifically written to assure support for GBE networks over permanent links and channels for runs of up to 100m. It was ratified in April 2001.

GBE networks achieve a 1,000Mbps throughput by simultaneously transmitting and receiving at 250Mbps per pair on all four pairs. This is known as simultaneous, bi-directional, full duplex transmission.

The technology associated with running GBE (and as we shall see later, 10G as well) over twisted pair further than 100m presents some unusual problems that are not found on lower speed networks.

The terminating electronics at each end of a GBE link rely on a device known as a ‘hybrid’ to separate the send and receive signals, exactly as is done in a telephone handset. Remember, a telephone only requires a single pair of wires to operate and allows you to speak and still hear the other party at the same time (i.e. it simultaneously sends and receives over the same pair).

Your telephone handset contains a voice frequency hybrid circuit.

But new problems arise with this arrangement when operated at high frequencies. Recall that the GBE signal contains frequencies up to 100MHz. At such high frequencies, the UTP pairs behave like transmission lines because the physical circuit length is commensurate with the signal wavelength.

With transmission lines, reflections are generated whenever the circuit impedance changes. This problem does not arise at voice frequencies because the wavelengths involved are very much longer than the actual physical length of the circuit. In a cable, the wavelength at 1KHz is some 200km whereas at 100MHz, it is just 2m.

Unfortunately, the hybrid in GBE terminating electronics cannot differentiate between a wanted signal sent from the distant end and an unwanted signal caused by a reflected signal. It treats both as valid received signals. Thus GBE circuits cannot function correctly when large reflections are present, such as those caused by a large discontinuity or, for that matter, a number of equally spaced small kinks in the cable.

The entire link (or channel) needs to be made from cable and components that are reasonably well matched to the nominal line impedance for the entire run.

These problems didn’t arise on lower speed Ethernet systems (100Mbps and below) because the send and receive circuits used different pairs. This means that when a Cat 5e installation is tested in the field, a ‘PASS’ result on the field tester only implies GBE support on that cabling if the constituent components are component level-compliant with the Standard.

This is because if large reflections are generated somewhere in the middle or towards the distant end of the tested link, they are attenuated by the intervening cable insertion loss so by the time they get to the field tester, it cannot see them. Hence a ‘PASS’ is reported.

However, a GBE link receiver will see them and become confused because the same cable loss applies to the wanted signal. So if you are planning to run GBE, beware of using components that are specified as link or channel level-compliant! These are meaningless terms and such components can play havoc on GBE circuits.

Category 6

The Cat 6 standard (ANSI/TIA/EIA-568-B.2-1) was ratified in June 2002. It was written to provide the level of transmission performance thought at the time to be adequate for supporting the next generation of high speed Ethernet networks over copper – 10 Gigabit Ethernet (10Gbase-T or 10G for short).

Unfortunately, much of the development work associated with running 10G over copper had not been completed at the time the Cat 6 Standard was ratified.

In addition, manufacturers had already been selling so-called ‘Cat 6’ product – product that had been designed to meet the performance requirements of the various drafts of the Cat 6 Standard that had been around for several years prior to the Standard finally being ratified.

As it turned out, the saga of Cat 6 was a repeat of what happened with the old Cat 5. There emerged a number of unforeseen transmission performance issues associated with supporting 10G over copper, for example performance needed to be specified to 500MHz, not 250MHz, and Alien Crosstalk (AXT), which was not specified for Cat 6 at all.

These technical requirements were not evident at the time of writing, and as a result the Cat 6 standard could not and did not address them.

The end result was that the Cat 6 Standard specified product that wound up being what could be considered of as just a premium version of Cat 5e product. At the end of the day, it would still only guarantee support for GBE to 100m.

Running 10G on Cat 6 cabling

If you really want to squeeze a little more out of your Cat 6 installation, there is a Technical Systems Bulletin (TSB-155) published by the TIA in 2007 called Guidelines for the Assessment and Mitigation of Installed Category 6 Cabling to Support 10GBASE-T. It provides recommendations for deploying 10GBase-T over installed compliant Cat 6 cabling systems.

It is worth noting, however, that this document is just a set of recommendations and not a standard.

Running 10G on Cat 6 is possible because the 10G terminating electronics are quite smart. They take care of most of the problems caused by cable impairments. What limits the circuit performance is the loss of signal due to cable insertion loss and the noise introduced by alien crosstalk (i.e. interference radiating into the cabling from adjacent circuits on other nearby parallel cables).

In brief, TSB-155 recommends that Cat 6 links and channels be characterised up to 500MHz by retesting for both the usual link/channel cabling parameters (i.e. return loss, pair-to-pair crosstalk, etc) and alien crosstalk (both ANEXT and AFEXT).

If the test results prove satisfactory support for 10G over shorter runs, of around 50m, may be possible over the installed Cat 6 system.

Should the alien crosstalk test results not meet the channel or permanent link minimum requirements of the Cat 6A standard, there may still be sufficient signal-to-noise margin to support running at 10G.

TSB-155 provides the means to make the necessary calculations to see if this is the case.

If it isn’t, TSB-155 provides guidelines for what to do to overcome this issue. Options include:

• Using only non-adjacent patch panel positions;
• Reducing alien crosstalk by unbundling cable runs;
• Separating patch cords and fly leads;
• Replacing Cat 6 patch cords and fly leads with Cat 6A-specified or shielded Cat 6 leads in places where they are bunched together (e.g. at patch panels);
• Reconfiguring any cross connects as interconnects;
• Replacing Cat 6 jacks with Cat 6A jacks; and,
• Replacing some critical horizontal cable runs with Cat 6A cable.

Category 6A

In February 2008 the TIA finally ratified the long awaited Cat 6A Standard. (I say ‘long awaited’ because there were at least 10 different drafts to my knowledge, spanning a period of over three years.)

In 2009, the TIA revamped its set of communications cabling standards and released the ANSI/TIA-568-C series. This included 568-C.2 Balanced Twisted-Pair Telecommunications Cabling and Components Standards – the current Cat 6A Standard.

This Standard also contains revised test methods that apply retrospectively to Cat 6 and Cat 5e.

The Cat 6A Standard was written specifically to support 10 Gigabit Ethernet (10Gbase-T or 10G for short) over unshielded twisted pair copper across distances of 100m.

Like GBE, it uses simultaneous, bi-directional, full duplex transmission.

From a transmission performance perspective, the key differentiating features are that circuits are qualified to 500MHz and, in addition to the usual tests, alien crosstalk testing is mandatory.

For field testing, a new generation of field tester was developed to cope with these more demanding requirements to a satisfactory level of accuracy.

At the component level, alien crosstalk performance requirements were specified for cable and connecting hardware, as well as channels. Having tested a lot of product in the lab, I can state unequivocally that these requirements are onerous and not easy to meet.

For unshielded cable, alien crosstalk is tested by setting up a 6-around-1 bundle of cable and requires almost 400 measurements, swept from 1 to 500MHz.

Some manufacturers have opted for a shielded cable (FTP) solution to overcome the difficulties presented by alien crosstalk in UTP cables. The shield does solve the problem of cable to cable induced interference, but introduced other issues which make standards compliance just as difficult (more about this in the next issue).

The alien crosstalk requirements for connecting hardware are also quite onerous. Alien crosstalk in connecting hardware is caused by the electromagnetic field from one connector being electromagnetically coupled into an adjacent connector, usually at the patch panel.

Some manufacturers have chosen to shield their jacks to minimise such unwanted stray pickup. Others use staggered port positions or every second port position in the patch panel order to reduce the level of pickup. However, all of these approaches have their drawbacks.

In summary, achieving category 6A compliance at the component level can be done, but it is not easy. Testing for component level compliance is also difficult and requires a well-equipped, standards-compliant lab setup.

Field testing for alien crosstalk is likewise complex and time consuming, but more about this in the next issue.

In the final part of this series we shall review the history of shielded solutions, the ISO 11801 standard and what the future holds for copper based data cabling.