Dynamic Performance

by Eric Lawrence Berk-Tek Technical Marketing Director RCDD

When 10BASE-T was a cutting-edge LAN application, the signal-to-noise ratio at the receiver was estimated using the attenuation-to-cross-talk ratio of the cabling. Actually, the attenuation-to-cross-talk of the cable itself was used since this pre-dated the link and channel specifications found in today’s TIA standards. This was sufficient because there was so much crosstalk present in early twisted-pair cables that the noise was dominated by the cross-talk from the near-end transmitter. Also, the interactions between the cable and the connecting hardware were so minimal that the relationship between the cabling and the components was insignificant.

With the commercialization of Category 5, and eventually Category 5e cabling systems, near-end cross-talk and other electrical performance parameters were improved to the point where the relationship between the cable and connecting hardware became significant. Link and channel specifications were created in an attempt to quantify this interaction in order to provide channel performance specifications to network elec­tronics designers. Although cross-talk was still considered to be the dominant noise source, the cross-talk in attenuation-to-cross-talk ratio became the near-end cross-talk of the channel.

With the emergence of digital technology, today’s situation is much different. Digital signal processing (DSP) has become so inexpensive that it is employed in virtually every modern LAN transceiver. Using DSP, near-end cross-talk can be cancelled to a certain extent. Thus, using the attenuation-to-cross-talk ratio as an indi­cation of signal-to-noise ratio is no longer accurate. Today’s twisted pair cables and channels also perform much better with only a small fraction of the cross-talk energy that they once had. Because of this, other noise sources, such as return loss, impedance mismatches, far-end cross-talk from multiple far-end transmit­ters, near-end cross-talk from multiple near-end transmitters, alien cross-talk, ambient electrical and electro­magnetic noise have become much more significant. Also, interaction between the passive components in the channel and the transceivers has become important.

The channel forms only one part of the communi­cation system. The signal-to-­noise ratio is critical at the point where the decision is made in the receiver as to what bit was intended. All system compo­nents play a role in determining the received S/N including noise from sources external to the communication system.

To design a cabling system that will maximize the signal-to-noise at the receiver becomes quite a challenge, considering the complexity of the noise environment and the sophistication of today’s digital signal processing.

It is a confusing task to sort through all the “marketing hype” and determine what will improve the network’s reliability. Improved network reliability is the goal because it will improve the bottom line through a lowered cost of ownership.

To help sort through the confusion, we need to return to the basics. In 1948, Claude Shannon published his now-famous commu­nication theory, which was renamed from “A mathematical theory of communication” on its first reprint. Shannon’s Law relates channel capacity to signal-to­noise across the usable bandwidth of the communication channel.

What Channel Capacity Means

Channel capacity is very important because it tells us the maximum data rate (bits of information per second) that can be transmitted from the source to the destination without error. This remains true even when the transmission channel is noisy, provided the bandwidth and received signal-to-noise requirements are met.

This does not mean that there will never be a bit error nor does it mean that the bit error rate is zero. There will always be a finite probability that an error will occur, and therefore, the bit error rate will always be greater than zero. What it does mean is that codes exist to define the communication scheme, such that the average probability of a bit error is as small as we want it to be. This is true for all symmetric, binary systems.

It also means that the message will be communicated from the source to the destination without error. In a digital communication system, there may be bit errors, and subsequent retransmissions, but the message will get through.

When a digital communication system operates at a data rate that is greater than the channel capacity, there will be a high probability of error. Thus, it is very important to have a sufficient signal-to-noise ratio for the application under actual, real-world operating conditions to ensure reliable and trouble-free operation. It is also very important to have sufficient bandwidth to ensure that the network application you are intending to run operates within the channel’s capacity.

Bit Errors

One way to express the probability of errors occurring at the receiver output is by the bit error rate (BER). The bit error rate can be estimated by counting the number of observed errors at the receiver as a ratio to the number of bits received. In the packet-oriented networks prevalent in data communications, this can be diffi­cult. This is especially true since many systems, like Ethernet, will discard the entire packet even if one bit is erroneous. To estimate the bit error rate in a packet based system, it is recommended that the packet error rate is measured instead. By assuming that if a packet contains an error, all bits in the packet are in error, the packet error rate equals bit error rate. This also provides a conservative estimation of bit error rate, ensuring adequate reliability. In reading marketing claims regarding bit error rate, with packet based appli­cations, be sure to understand the assumptions being made. They are not universal.

Bit error rate is a quantitative measurement of the reliability of a digital communication system. The lower the bit error rate, the more reliable the communication system is. Market research (Sage Research, December 1998) has shown that improved network reliability is the most important driver of network upgrades for more than 80% of the IT managers that were surveyed. BER is an excellent metric for network reliability.

Limitations on Channel Capacity

Channel capacity, as calculated using Shannon’s Law, can be misleading if not carefully used and interpret­ed. As previously stated, it represents the maximum data rate in which digital com­munication can reliably take place. It does not, however, have any provision for the practicality of the digital commu­nication system. Just because codes exist to obtain the data rate does not mean that it would be practical to manufacture a transceiver to send and receive them. Nor does it have any provision to ensure that the resulting equipment would have a palatable price tag.

With that in mind, it is important that constraints are placed on the assump­tions made when calculating the channel capacity of a digital communication sys­

tem to ensure that there can be practical implementations that would be able to offer robust operation under a variety of real-world operating conditions. By carefully choosing the proper noise sources and levels, this can be accomplished, ensuring that we are designing communications systems for real-world networks.

Although today’s integrated circuits are extremely powerful, they do have limits on complexity and speed. Fortunately, semiconductor and active equipment providers are constantly pushing this envelope. As the can­cellation, error-correction, wave-shaping and other DSP techniques improve, the maximum, practical channel capacity will naturally improve, always bounded by the ultimate theoretical capacity. It is important to keep in mind that channel capacity calculations are always dependent on the assumptions made of the noise sources. It is very easy to inadvertently make an apple-to-oranges comparison when looking at the channel capacities of two different systems if the assumptions are not fully understood.

Conclusions

Modern communication systems are no longer dominated by near-end cross-talk. Therefore, attenuation-to­cross-talk ratio (ACR) is no longer an accurate method to gauge signal-to-noise. To accurately characterize signal-to-noise, all dominant noise sources at the receiver output must be considered. Communication sys­tems are comprised of active and passive devices. The signal and noise are affected by all components of the system including interactions. For example, cross-talk noise created in the passive cabling system will be reduced both by the attenuation of the cabling system and the DSP cancellation in the receiver.

Bit error rate is an important metric of network reliability and ensuring adequate signal-to-noise under real-world conditions. Real-world conditions will affect signal-to-noise and the resultant bit error rate. It is impor­tant to have reliable network operations with the conditions under which the network will actually be operating.

Channel capacity tells us the growth potential of our communication channel to support higher data rates. Users need to exercise caution when comparing channel capacities of two different channels to make sure an apples-to-apples comparison is accomplished. Differences in assumptions will have a significant effect on channel capacity.

For more information, call 1-877-96CLEAR or visit our web sites. NetClear: http://www.netclear-channel.com