Bandwidth expectations for future twisted-pair networks
The past is prologue as UTP technology forges ahead.
The past is prologue as UTP technology forges ahead.
Throughout history, magicians and fortunetellers have used many different methods for attempting to see into the future. The nature of enterprise networks is ever changing and, therefore, seeing into the future of our industry may also seem like a mystical exercise. Certain engineering models and technology developments (see sidebar, “From kilobits to gigabit,” page 15), however, can help make more accurate predictions.
Making 10-GbE over copper a reality
In late 2002, more than 100 subject matter experts attending the IEEE 802.3 [Ethernet] meeting evaluated the feasibility of forming a working group to standardize 10-Gigabit Ethernet transmission over 100 meters of copper twisted-pair cabling. Key advances in digital signal processing (DSP) made some representatives think this was feasible. For starters, the fiber standard for 10 Gigabit, 802.3ae, was complete. And a short distance (<15 meter) copper project-802.3ak 10GBaseCX4-was in progress.
Those present decided to form a study group chartered to present a formal proposal to IEEE. This process was successful and Working Group 802.3an is now developing the technology to make 10-Gbits/sec over twisted pair a reality for a 100-meter channel.
Increasing transmission rates by another factor of 10 will require significantly higher bandwidth than existing systems can provide. The theoretical throughput of the channel, called the Shannon Capacity, needs to be in excess of 18 Gigabits to allow for design penalties incurred when building active components. This is accomplished by increasing the bandwidth to 500 MHz, reducing transmission impairments, and using a more complex encoding scheme-PAM-12.
PAM refers to Pulse Amplitude Modulation, technology that allows multiple bits per symbol to be transmitted, resulting in data rates significantly higher than the maximum specified frequency of the cabling.
Noise cancellation technology makes 10GBase-T possible, with significant levels of cancellation for NEXT, ELFEXT and RL noise now a reality. This raises the signal-to-noise ratio for a given cable. Alien crosstalk (AXT), both near end and far end, is noise from outside the cable. AXT was not a significant noise source in earlier applications but now is considered a dominant noise source for 10GBase-T.
The largest AXT influence is from other surrounding cables. Since AXT tends to be random and unpredictable, it is very difficult for DSP to cancel. Initial modeling showed that a 15-dB improvement in AXT from typical Category 6 levels is needed to support 100 meters of 10G-BaseT data transmission, which means that Category 5e cabling is inadequate for further consideration.
This timeline shows the tenfold increases in speed for copper-based Ethernet standards, and the corresponding TIA documents specifying cable performance to support those speeds.
Based on this information, IEEE sent liaison letters to TIA/EIA TR-42 and IEC/ISO 11801 asking for cabling, component, and channel testing requirements to be developed. IEEE provided four channel models that differed in their AXT and insertion loss assumptions. Two channel models were related to new cabling to support 100-meter channel lengths and the other two relate to legacy Category 6 channels of up to 55 meters and 55 to 100 meters respectively. All four models required transmission parameters to be characterized up to 500 MHz.
Since IEC/ISO 11801 has Class F STP cable listed as an option, Class F insertion loss is an option for UTP designs to meet. The stronger signal levels allow a 2-dB higher AXT threshold while still maintaining the required signal-to-noise ratio. A new designation, Cat 6a (augmented) was created to describe these new high-performance UTP cabling systems.
At the time of this article, test methods for measuring AXT on Category 6a cabling, components, permanent links, and channels are still undefined and under study. The standards groups are working feverishly to complete this work so they won’t delay the projected 2006 publication date for the 802.3an 10GBase-T standard. The method currently favored by most participants is a tight bundle of six disturbing cables wrapped around a single victim cable.
AXT is measured as a power sum for all 24 disturbing pairs on each of the four victim pairs. This configuration can be used for cable-only, or for channel measurements that include connecting hardware.
Several vendors have released cabling products touting 10GBase-T support. Customers evaluating these products for purchase should make sure that supporting information showing AXT test data and test methods are clearly defined. Existing Category 6 products are likely to support 10 GBase-T for limited distances with 55 meters as a goal. Additional modeling is needed to precisely set the distance limit for legacy Category 6 cabling.
TSB-155 is being developed by TIA/EIA TR-42.7 to set cabling requirements for legacy Category 6 products. The first draft is now out for industry ballot. The standard for Category 6a cabling components is also out for first industry ballot. If approved, it will be published as TIA-568-B.2-10.
Copper forever and ever?
Data storage is getting cheaper and processor speeds are getting faster on a continuous basis. Moore’s Law predicted a doubling of microprocessor speeds every 18 months, and it appears to be continuing on track. The area required to store a terabyte of data in 1967 was more than 1 million square feet-the size of a large factory. Today, that amount of information can be stored on an area less than a square foot. The ability to move and store data quickly and at low cost will continue to drive the use of more content-rich applications.
Alien crosstalk is a concern in 10GBase-T systems, not just for cables, but for connecting hardware as well. This illustration animates the effects of adjacent connectors on each other in a patching field.
So, does it ever stop? Will copper media become obsolete? DSP technology will continue to improve, and it will be possible to provide higher levels of cancellation for impairments such as NEXT, ELFEXT, RL and perhaps AXT. Cabling technology improvements will also allow the use of higher frequencies. Larger copper diameters will further reduce insertion loss; however, preliminary Cat 6a cable designs use 22-AWG conductors that approach the practical limit for size.
Tight twist lays improve noise immunity, which also improves theoretical capacity. Category 6a UTP products also push the limit on practical twist lays, approaching three per inch. Cabling developments following Cat 6a will use higher bandwidth, improved noise cancellation and potentially increased use of shielding. But twisting tighter or making the cable bigger than the current 0.315-inch maximum allowed is not desirable.
I believe that doubling LAN throughput on UTP media to 20 Gbits/sec for a 100-meter distance is likely within 10 years. This prediction assumes additional cancellation of noise and some improvement of cabling performance. STP or coaxial cabling will be able to expand copper throughput further.
An insatiable appetite
It’s difficult to imagine using 10-GbE to the desktop on a broad scale within the next five years; however, within the useful life of a network installed today, it is very possible we will see wide scale deployment of 10-GbE in commercial buildings. Data centers could use 10-Gbit transmission today, and will be the early adopters of 10GBase-T equipment.
The truth is, as long as we need quick access to information-whether at work, at home, or anywhere in between-the need for faster data networks will be a requirement for the foreseeable future. The reality is that we have become accustomed to finding any piece of information essentially “on demand,” and that demand will only continue to grow.
Whether you service the information needs in the financial community, medical field, business world, transportation industry, government agencies or any other area, those who depend on you will continue to demand more information faster. To illustrate this further, think about the kinds of information being transmitted today:
• Medical imaging, which was once film-based, is now transmitted real-time to doctors and patients awaiting results.
• Computer-aided design (CAD) images, which were once simple line-drawings, are now 3D, full-color, high-resolution images sent electronically to engineers, contractors, builders, and consumers.
• Call centers must have access to millions of names in multiple databases, which grow larger while on-hold wait times are driven down.
• Homeland security initiatives require the capability to include alarm systems and video surveillance on the same network used by an information-technology system.
To be sure, the demands will be great, and, despite the fact that in 10 years we’ll be transmitting more information at a faster rate than any of us can imagine today, I’m fully confident that we have the foresight and technology to anticipate and meet the demand.
From kilobits to gigabit
In 1969, a general standard was developed for data transmission devices. This standard, known as RS-232, allowed for reliable connection between various types of communication equipment without compatibility problems. RS-232 facilitated data rates up to 38,400 bits per second. In 1973, development began on a proprietary bus topology LAN, which became a de facto standard for 10-Mbits/sec Ethernet. This became the basis for the Institute of Electrical and Electronic Engineers (IEEE; www.ieee.org) 802.3 standards for 10Base5 in 1983, 10Base2 in 1986, 10Base-T in 1991, 100Base-T in 1995, and 1000Base-T in 1999.
The IEEE serves as one of the key standards bodies that drive the development of network cabling. IEEE 802.3 Ethernet and IEEE 802.5 Token Ring recognized that standard telephone wire, designed to transmit kilobits, would not support the 10- to 16-Mbits/sec data rates desired. Early Ethernet networks specified bus architectures that used ThickNet and ThinNet 50-Ω coaxial cabling rated for 10-Mbits/sec data transmission.
IBM’s proprietary shielded twisted pair (STP) cables, Type 1, Type 9 and Type 2 composite, transmitted 16-Mbits/sec Token Ring signals. While these early networks served their purpose, improved microchip technology made it possible to look seriously at using UTP cabling for these applications. A star-wired UTP network is easier to install and troubleshoot relative to the coaxial bus and STP networks that were in use at that time. The culmination of this work was the publishing in 1990 of IEEE 802.3 10Base-T that specified the use of 4-pair UTP cabling, later designated as Category 3 UTP (certified up to a frequency of 16 MHz).
Category 4 cabling soon followed the introduction of Category 3 cabling and was certified to 20 MHz to focus on supporting 16-Mbits/sec Token Ring. Shortly after the introduction of Category 4, it was apparent that LAN technology would allow a step up to 100-Mbits/sec Ethernet. Category 5 cabling, rated up to 100 MHz, supplanted Category 4 as the highest-speed cabling system. The factor-of-10 increase in speed continued with Ethernet’s next-generation solution, 1000Base-T. In fact, it has been an objective of the IEEE Working Groups to achieve tenfold increases in data speed with no more than a tripling of the cost of the active equipment.
Category 5 requirements needed to be upgraded to Category 5e to support Gigabit networks that, while still rated at 100 MHz, have added the additional parameters of return loss (RL), equal-level far-end crosstalk (ELFEXT), and power-sum crosstalk requirements. Near-end crosstalk specifications (both pair-to-pair and power sum) were also improved by 3 dB. Field-test criteria were then established to certify legacy Category 5 and new 5e channels for Gigabit operation.
Immediately following the standardization of Category 5e, cabling standards groups, such as TIA/EIA TR-42 and ISO/IEC 11801, went to work on a UTP infrastructure specifically designed for twice the bandwidth of a 5e channel. This fresh design would be ideal for Gigabit networking and have capacity to support higher speed networks in the future. The resulting Category 6 channel is certified to a 200-MHz frequency, with cables and components testing out to 250 MHz. Work on Category 6 was completed in 2002, and the standard was published.