A Series of Extensive Experiments at a Berk-Tek Test Facility Offer Evidence That 1000BASE-T Ethernet Networks are Best Served by Using Category 6 Cabling Systems.
January 1, 2003
Since the ratification of the TIA/EIA Category 6 standard in June of this year, many users and specifiers of Category 5e cabling systems have been considering upgrading to Category 6.
Some others have been reluctant to make this move and feel that they have “all the cabling system they need.” This latter group points to the fact that IEEE 802.3, the group that writes the Ethernet networking standards, built the 1000Base-T Gigabit Ethernet standard to work over the installed base of category 5/5e.
The belief of the committee doing this work was that the installed base for LAN cabling is Category 5/5e, thus the solution for Gigabit Ethernet over 100 meters of copper must be able to work over Category 5/5e.
The purpose of this article is to show how “real world” environmental stresses and noise sources can adversely affect 1000Base-T when operating over Category 5e cabling systems. It will also be shown that 1000Base-T can operate with better performance and reliability when operating over Category 6 cabling systems.
To accomplish the Herculean task of engineering Gigabit Ethernet to work over Category 5/5e, chip designers were forced to build in a significant amount of Digital Signal Processing (DSP).
Algorithms to cancel the effects of NEXT (Near-end Cross talk), transmission echo (transmitted signal reflected back to the near-end receiver), as well as other parameters had to be included in the chip technology in order to make 1000Base-T work over Category 5/5e cabling systems.
Because the cable plant is usually Unshielded Twisted Pair (UTP), additional processing power was also needed to scramble the signal, which enabled EMI (Electromagnetic Interference) emissions to remain within FCC limits.
After all this DSP work, Category 5e is able to support 1000Base-T in a controlled, laboratory environment using well-behaved electronics and cabling.
After having been involved in many cabling installations where the surrounding environment has had an effect on network performance, researchers at the Nexans New Holland Competence Center (NHCC), located within Berk-Tek’s headquarters facility in New Holland, Pa., decided to initiate a series of experiments to stress this 1000Base-T technology under “real world” conditions.
The results of these experiments are revealing and put into question the practicality of using Category 5e to support Gigabit Ethernet.
IMPACT IN THE REAL WORLD
The NHCC has been operational for one year and during this period a number of projects have been completed that cast doubt on the reliability of standard Category 5e cabling systems when used to support 1000Base-T networks when subjected to non-optimal, “real-world” conditions.
A series of experiments offer evidence that 1000Base-T Ethernet networks are best served by using Category 6 cabling systems.
In any network, the performance of the Network Interface Cards (NIC’s), LAN on motherboards (LOMs) and hub/switch ports can vary widely.
The transceivers employed are designed to meet IEEE standards, but as in any manufacturing process, the normal “bell distribution curve” suggests that some will be stronger performers than others. Some devices might have a weak output signal, be more susceptible to noise, or be affected by temperature and humidity. In any of these cases, the user will have a workstation that could experience slow performance or excessive downtime, dropped sessions, and other network problems.
At the NHCC, laboratory-grade, 1000Base-T transceivers were tested and categorized based upon network performance. Transceivers that were classified as weak and strong were then put into service and connected to 100 meter, 4-connector channels.
The test results revealed the additional performance advantage a Category 6 channel exhibits versus a Category 5e channel.
As an example, a 13 time difference in bit-error-rate could mean the difference between a workstation working or crashing. Also, the products used were “laboratory-grade”, costing thousands of dollars.
A user, buying NICs, hubs, and switches at significantly lower cost, may experience a wider range in performance with the weaker units possibly performing even worse than those tested here.
In addition to weak transceivers, attention must be paid to the environment the cabling system is installed in. In our second test experiment we explored the effect elevated temperatures can have on network performance.
Most users believe that when cable is installed in a building with central heating and ventilating systems, the temperature throughout the cable pathway is constant.
However, in many commercial structures the air return plenum space can become quite hot if it is exposed to the outside walls or the roof.
There have been cases in the summer months where air return spaces can reach 54 degrees Celsius or higher. In this environment, the electrical properties of the cabling can change and potentially affect network performance.
Figure One shows a summary of error accumulations for a 1000Base-T network over a 24-hour period when Category 5e, Category 6 and Category 6+ cabling systems were exposed to elevated temperatures.
These results show that as the temperature of the cabling environment increased, 1000Base-T exhibits far fewer errors operating over a Category 6+ cabling system than over a Category 5e cabling system.
In our next experiment, we introduced the first of two noise sources, EMI (Electromagnetic Interference). Outside noise can be the “Achilles Heel” for 1000Base-T and it is clear from the test results that a Category 6 cabling system can protect a network better than a Category 5e cabling system in the presence of EMI.
There are instances when a commercial building may be near a radio or cellular transmitting tower. In urban areas this is especially of concern. Electrical noise can take many forms. The DSP technology that is built into 1000Base-T chips is not designed to deal with noise generated by external sources.
A third test involved analyzing Gigabit Ethernet performance when 100-meter, 4-connector cabling channels were subjected to various frequencies of EMI in a specialized EMC chamber that isolated the cabling system from any other noise sources. This chamber is used to determine the fitness for use of numerous electrical and electronic devices in the presence of EMI.
The results were revealing in that they pointed to a cabling property that up until now has not been a major consideration in UTP performance testing, cable balance. Longitudinal Conversion Loss (LCL), one element of cable balance, is now a recommended component performance property for cable in the TIA/EIA 568-B.2-1 Category 6 standard.
WHAT IS CABLE BALANCE?
Without delving into transmission line theory, cable balance is a measurement of how consistent a twisted pair cable is made throughout the manufacturing process. Specifically, positioning each of the eight wires in the center of the insulating plastic, consistently twisting and cabling the pairs along their length, and controlling tensions throughout the manufacturing process, including packaging, are just some of the steps that must be taken to build a cable with better balance.
Transmission over UTP cabling systems is done using differential signaling. Differential systems employ alternating current and voltages, thus at any given instant in time one wire in a pair will have a positive voltage and the other wire will have a negative voltage. As disturbances from outside the cable appear, noise voltage should be coupled onto each conductor in a pair “equally.” If noise is injected into a twisted pair equally then these unwanted artifacts can be subtracted out at the receiver. Noise voltage will only be coupled equally if the twisted pair is “perfectly balanced.”
In practice, it is not possible to create a “perfectly” balanced pair or twisted-pair cable. However, tightly controlled manufacturing processes can produce extremely well balanced pairs and UTP cables, which in turn are far less susceptibl
e to outside noise sources.
In our final test scenario, the benefits of good cable balance were examined when the cabling system was in the presence of Electrical Fast Transients (EFT) interference.
Electrical Fast Transient noise is present in all building power systems. These high frequency voltage “surges” can be detrimental to network performance if they are coupled onto network cabling systems.
Noise due to fast transients can come from electrical devices cycling or turning on and off such as lighting ballasts, compressors, copy machines, fax machines, pencil sharpeners, pumps, etc.
As wire and cable pathways become more and more congested and users opt for condensed raceways, the possibility exists for power wire to get close enough to network cabling such that the effects of EFT are significant.
In this test, the power wire and the network cable were run in commercially available plastic raceway that has compartments for each. The proximity of the two separate cabling systems was within two inches.
The result was that the EFT noise coupled to the Category 5e system significantly, whereas the Category 6 systems were not affected until the transients were at a voltage level beyond that which would be expected in a normal commercial building.
These results reinforce the point that cables with better balance, such as Category 6, are more immune to outside noise sources and do a better job of protecting network systems.
In each of these test cases, it is clear that the increased signal to noise margin of Category 6 cabling systems provides a measurable difference in network performance and reliability when compared to Category 5e cabling systems. It is also clear that using a Category 6 cabling system, that is better balanced than a Category 5e cabling system, will add performance margin.
This additional margin could provide the difference between poor and excellent network operation when cabling systems are exposed to stressed environments and unwanted noise sources.
Dan Kennefick is the Copper Products Business Manager for Berk-Tek, a division of the Nexans Company, responsible for copper products sold in the Americas region. A member of both BICSI and the NFPA Electrical Section, he has worked in the wire and cable industry since 1986.