Migrating from 10Base-T or 100Base-T to faster signaling creates unexpected problems for many network owners. These problems result from the added vulnerability of networks to failure caused by lost operating margin. This article describes four steps contractors and installers can take to surmount these problems.

The high-speed data communication revolution continues. Driven by an exponential increase in data processing capabilities, this revolution is turning data streams into torrents. Evolving technologies require even faster network protocols to handle the flow

The most common computer networking protocols now in use are 10Base-T and 100Base-T Ethernet, operating at the once-impressive rate of 10 Mbps (10 million data bits per second) and 100 Mbps respectively. Thanks to the foresight of telecommunication industry leaders, Ethernet transmission at these speeds requires only two of the four pairs of wires found in the widely installed UTP data cable. Not merely spare wires, the extra pairs were designed into the network scheme, anticipating the day when these protocols would be replaced by something even faster.

That day has come. 1000Base-T, or Gigabit Ethernet, is now being used in many networks. Existing network installations and almost all new ones are upgrading to Gigabit Ethernet. This protocol operates at a blistering 1 billion bits per second by using all 4-pairs and signaling bi-directionally through each pair!

As computer networks switch to 1000Base-T, network owners will experience more efficient network processing. Higher bit rates equal more bandwidth which, in turn, equals more efficient throughputs and ultimately, happiness for everyone. But this huge increase in network speed is not gained without a significant cost. Beside the obvious dollar costs, there is a significant hidden cost to both existing and new networks.

The hidden cost is the added vulnerability of any network to failure caused by lost operating margin. No one involved in the development of higher speed installations is immune. The impact is felt by LAN system designers, component manufacturers, distributors, installation contractors, field tester manufacturers and ultimately--and most significantly--the end user or client.

There is an old joke in the industry that states unless something extremely bad was done to a premise network, Ethernet will run on wet string. That's an exaggeration of course, but the point is valid: a typical 10Base-T network has a huge safety margin. Components, connections, cabling and installation can each be off spec and the network will still work. Some have reported a cable run with a splice in the middle, and despite the flaw, the link passed testing and 10Base-T functioned just fine. This convenient fact changes as network speeds increase.

Many network managers switching to faster Ethernet are seeing 10-15 percent of their Category 5e-rated nodes for which they paid large sums of money, fail to operate at the required speeds. Worse, some nodes don't work at all. What has gone wrong?

At higher speeds performance margins begin to shrink dramatically. The reasons are highly technical but they can be stated simply. Bi-directional signaling with four pairs adds new network complexity. These higher speed signals have less power and the noise accompanying them is relatively strong.

The signal to noise ratio is so low that very exotic and expensive processing techniques are designed into the network interface cards and other network gear. Chip and board designers are solving these technical problems, but chip design doesn't directly guarantee premise network margin. How can we be assured that the networks have more than enough safety margin to survive the transition to Gigabit and keep the customer happy?

First, contractors and installers must be trained to provide a first class installation. Installing cable and hardware for higher speed networks is a critical skill. Pull tensions, bend diameters, fill ratios, separation from power circuits, grounding systems, termination techniques, and many other factors must be considered, practiced, and mastered.

Installers can get good instruction by attending industry or BICSI endorsed training classes or by aligning with a premise hardware manufacturer. Most of these manufacturers sell their equipment as a system with a warranty backed by trained installers and field-testing.

Second, each installation will have more margin if the very best hardware, connectors, and cabling are installed. Systems are only as good as their weakest link. If you have a great installation but the parts are deficient, the network will have problems at higher speeds.

The TIA has developed component requirements for enhanced Category 5e, Category 6, and is working on those for augmented Cat 6A. If the performance values of your components meet these newest standards, by definition, the network will have more safety margin.

A third way to overcome the hidden cost of higher speed networks is to talk to manufacturers of hand-held testers. The newest testers are truly amazing. These devices will do tests such as return loss (RL) and time-domain analysis, that were previously limited to $40,000 network analyzers, but at a fraction of the cost. This new data will tell you volumes about the quality and margin of a network link or channel.

Last and fourth, it is vital to understand the data, especially return loss. RL is a summation of all the reflected signal energy coming backward toward the end where the signal originated. It is like an echo and should not be confused with crosstalk. RL limits have been defined for both Cat 5e and Cat 6 channels. When checking your network, analyze actual channel RL performance across the frequency range and note the margin that separates it from the limit lines. RL wasn't previously specified in network testing because it has no effect on slower Ethernet signaling. For the new, higher speed protocols, it is a critical measurement because it is a strong indicator of an installation's performance margin.

As an example, Figure 1 shows the RL measured for a 100-meter channel. The hardware and patch cords are rated Cat 5 and the horizontal cable meets the requirements for Cat 6. The horizontal/diagonal line starting at -17 db and ending at -6 db is the RL limit for Category 5e. Although all channel components interact, the horizontal cable dominates the lower frequencies (below 10 MHz). The high frequencies (above 100 MHz) are dominated by the connectors. The mid range (under the shaded area in Figure 1) is driven by patch cable quality. Although this channel meets the Cat 5e return loss requirement, the upward "bump" in the data in the critical middle frequencies can be improved.

Figure 1.

The shaded area in Figure 2 shows the improvement gained by simply changing to better patch cables. All other components remain the same. Notice that the channel RL is improved by 4-6 decibels in the frequency range that transmits most signaling energy. More margin in this range provides more network stability, fewer bit errors and stronger signals. This finding is often confirmed by field experience. Nodes that fail when a network is upgraded to faster Ethernet often "heal" with the substitution of better patch cords. We have heard many end users comment that "I don't know why, but our network seems to run better with these cords." The new generation of hand-held testers gives you the ability to measure this RL improvement on your network.

Figure 2.

The premise network industry is facing new challenges, and everyone is affected. Network professionals must protect themselves and their customers from potential problems generated by the migration to next-generation protocols. Those who succeed will have both the upgraded skills and an understanding of the hidden costs of higher speed networks so that they can effectively provide their customers with the performance they paid for.

For information on Quabbin Wire & Cable's state-of-the-art DataMax®6 patch cords and authorized assemblers, click here.

To comment on this paper e-mail us at engineering@quabbin.com.