Ensure measurement accuracy for LAN testing by avoiding
the negative effects of EMI and improper grounding
The identification and elimination of noise has always been a major concern in the installation, testing, and certification of LAN networks, as well as a key part of ongoing network maintenance and troubleshooting. Contract cable installers and internal network maintenance staff must understand the various sources of noise that can interfere with legitimate signals and degrade data transmission over the network. Noise can come from external sources, such as electromagnetic interference (EMI) from nearby equipment and transmission lines, and from internally generated noise on the cabling due to poor workmanship issues.
As network performance levels and bandwidths continue to escalate, managing noise becomes an increasingly important issue. Higher speed means packing more data into shorter time cycles, and shorter signal pulses mean greater susceptibility to interference. Unfortunately, the negative effects of noise can sometimes take the form of subtle performance degradation rather than a “hard failure” of the network. Data that isn't properly received due to noise interference must be retransmitted, thus slowing down the completion of that transmission while also contributing to overall network congestion.
LAN test equipment plays a critical role in the management of noise in today's network environments. Without the ability to precisely measure, diagnose, and eliminate noise, network installations that are specified as high speed can actually be operating at much lower performance levels. On the other hand, testers that aren't able to clearly distinguish between different types and sources of noise can cause technicians to spend excessive amounts of time chasing down extraneous noise that may or may not have a significant effect on real-world network performance.
Types and sources of noise. Noise in the network can consist of a variety of types, including common mode and normal mode. Noise at its basic level represents a disturbance to the waveform that bears no relationship to the fundamental frequency and which therefore can interfere with the waveform's signal-carrying capabilities. Common mode noise is particularly important in twisted-pair networks because signal integrity is dependent upon maintaining the differential relationship or “common mode” between the wiring pairs.
Common mode noise is created from the difference in potential between two physically remote grounds. A poor ground system or ungrounded shielded cable can act as an antenna, which gathers the induced voltage and applies it to the input. This type of noise can be hard to eliminate and becomes increasingly more problematic as the frequency of the noise increases, especially for today's high frequency networks.
As shown in Fig. 1 above, when a data signal is attenuated through natural capacitance, severely differentiated, or contains a lot of noise, the individual signal pulses become much less distinct and the data less likely to be correctly received at the far end of the transmission.
With network speeds constantly increasing and circuit logic voltage levels simultaneously decreasing, the ability to maintain precise waveforms is becoming even more difficult. When transmitting signals using 3V logic at 350 MHz rates, the presence of less than 1V noise levels can have a major effect on the link's data-carrying capabilities.
Workmanship is another key factor that can affect noise levels. Over the past few years, great strides have been made in the quality of cabling, connectors, and other materials that make up the network, thereby ensuring a highly consistent level of impedance matching within any particular length of cabling or specific connector component; but the real challenges come into play when they're installed within real-world environments. Even using the best materials, critical workmanship issues like improper grounding, untwisting too much wiring at the termination point, or impedance mismatches between cabling and connectors can easily create unacceptable noise levels.
When two adjacent lines in a twisted-pair circuit carry a signal, an intentional capacitive coupling exists between the lines. To the extent that the signal is distributed as two equal and opposite phases, any transitions along the cabling length will disturb the differential phases by equal amounts and leave the difference intact. Good layout practice is vital to maintain proper coupling between the differential traces, thus ensuring that any noise introduced into the application environment is seen as common. Proper grounding is also a critical issue to keep the circuit from inappropriately collecting and accumulating radiated energy from surrounding sources.
External noise, such as EMI sources, can also be a major contributing factor in LAN link test failures. EMI can be radiated from a variety of devices that emit unintentional RF signals, such as computers, factory floor production equipment, television and stereo sets, fluorescent lights, power tools, power lines, and office equipment. Patch panels and wiring closets can present particularly difficult environments, with many different signals trying to find routes to ground by cross-coupling across nearby cable links.
If the surrounding building has poor grounding, the prevailing neutral-to-ground voltage conditions will be high and the radiated effects can impose themselves on any data cable in the area. One side issue to consider here is the use of shielded cabling. While shielded cabling can sometimes be helpful in reducing external radiated energy, it's important to remember that the shielding is typically tied into the building ground. If the building has poor grounding, the cable shielding can actually become a contributor to noise on the cable rather than a benefit.
Noise detection and analysis. Distinguishing between different types and sources of noise is a critical first step in troubleshooting noise-related failures. Unfortunately, different types of noise can “look” the same to some test instruments. Radiated energy between pairs in the form of crosstalk can look a lot like radiated noise from external sources to some testers. For instance, close proximity to external noise radiation sources can result in “false failure” modes that may consume significant amounts of a technician's time, as he or she tries to chase down non-existent problems within the cabling system.
Measurements such as NEXT (near end crosstalk) and ELFEXT (equal level far end crosstalk) can be useful tools for identifying noise within twisted-pair copper networks. Both of these techniques operate by putting signals on a wiring pair and then measuring the coupled energy of the field effect on adjacent pairs.
However, NEXT measurements typically aren't as useful for distinguishing noise sources because externally generated noise may be arithmetically summed into the total noise equation, thereby indicating a failure but not the reasons for the failure. In contrast, PowerSum ELFEXT is particularly useful in noise analysis. By selectively testing signals on the far end of the cable link, this test can provide a more detailed picture that helps sort out crosstalk from external noise. If failure indications are only on outside pairs, such as Pairs 7 and 8 or 1 and 2, it's a good indication of external energy.
The image shown in Fig. 2 demonstrates a specific situation in which significant spikes were observed between ground and neutral on a single-phase circuit. The spikes were first seen as an ELFEXT failure, which led to a search for external sources for the high-energy spikes that were being induced on to the cable. Further investigation revealed that the data cable passed closely by an AC power line that supplied power to a copy machine. The spikes occurred whenever the heater in the copy machine was energized.
Differences in LAN test equipment. Tester architectures and design approaches can play a critical role in the ability to detect and analyze differences in noise sources. Among the key issues that must be considered are the fundamental differences between digital domain testing and frequency domain (analog) testing.
All signal waveforms and noise patterns are inherently analog in nature. This may cause problems for some DSP-based testers, which acquire information digitally, to precisely distinguish between types and sources of noise. Depending upon the digital sampling methods and frequencies you're working with, DSP-based testers may unintentionally mask out subtle differences between noise types.
To overcome this issue, some digital testers are equipped with a “noise check” feature that detects voltage on the line prior to conducting any other testing. The initial “noise check” acts as a preliminary screening mechanism to characterize and filter out the external noise from subsequent testing. Unfortunately, because the external noise is part of the surrounding environment, it may be a critical issue that affects the actual operation of the network and therefore shouldn't be arbitrarily filtered out of the test and certification processes.
In comparison, high-performance analog testers can simultaneously “see” and display all of the relevant signal and noise waveforms on the line. This allows you to clearly distinguish between different energy types, identify various noise sources, and to determine which ones represent potential real-world problems.
In the final analysis, it's critical that the field test instrumentation you're using not only be able to conduct all of the tests specified by standards committees like TIA and ISO, but also be able to precisely identify, measure, isolate, and analyze noise throughout the entire network environment.
Pivonka is senior application engineer at Ideal Industries, Inc., San Diego, and Mazzuchelli is the president of EML Associates in Stoughton, Mich.