Robert A. Conte and Masood A. Shariff
AT&T Bell Laboratories
Those who install Category 5 unshielded twisted-pair cabling systems are often called upon to verify link performance before they hook up the active equipment in a local area network. This verification, presently based on the advisory test parameters listed in Annex E of the recently released Telecommunications Industry Association-568A standard, is a challenging technical problem involving precise electrical measurements between 1 and 100 megahertz.
Sellers of portable test equipment may claim that these measurements are routine. "Simply connect the tester to the link and let the device tell you whether the link passes or fails," they say. The reality of the situation is very different. Although portable testers may one day play an important role in link verification, the makers of these devices must first improve their high-frequency performance.
To deal with current inadequacies of portable Category 5 LAN testers, advocates offer a strategy of "buy a unit now, and software upgrades can be downloaded in the future to keep it up-to-date." This is an oversimplification. Although it is an easy procedure to download software that processes measured data in different ways, it is difficult, if not impossible, to modify the basic electrical performance of a piece of equipment with downloadable software. If you start with inaccurate data, you are likely to reach erroneous conclusions, no matter what processing is performed in software.
You must be prepared, then, to ask questions about a testing device and the data it produces before you judge the compliance of installed links. Before purchasing equipment, you should develop a set of requirements against which different units can be compared. These requirements should come from an understanding of what needs to be accomplished and how test equipment can help you reach your goals.
Three-step procedure
One way to view the field test procedure is as a three-step process. The first step is a sanity check.
Is there a simple check that ensures the unit is properly calibrated and producing reliable data? To ascertain if the unit has been damaged, you can use this check prior to testing. To accomplish this, perform a test or measurement in which the answer is known beforehand. If the new results do not agree with those previously obtained, there is a problem.
The second step is data collection and processing. Near-end crosstalk and attenuation are the most important parameters to be measured, but other parameters may also be useful.
The difficult part of this step is to determine what constitutes compliance with a standard. Does the geometry of the link you are measuring agree with that of the tables of numerical values stored in the measurement equipment? Should the numbers internal to the device be modified in some way before a comparison is made?
Problem resolution is the final step. Once accurate data is collected and processed, how does the test equipment help you resolve problems? If, for example, the device provides a pass or fail message but does not indicate where the problem lies, then it may be of little value in evaluating a building site.
This three-step procedure--sanity check, data collection and processing, and problem resolution--will naturally lead to a set of technical requirements and desirable features that can be used to compare and select equipment for purchase.
Measurement sanity checks
A sanity check can help identify such problems as faulty or crossed connections, as well as loose wires. The first such check is to measure the crosstalk of a link. Suppose the measurement is made by transmitting on pair X and receiving, or measuring, near-end crosstalk on pair Y. Once the data is collected, the same near-end crosstalk measurement can be made by transmitting on pair Y and receiving on pair X.
This reversal of transmit and receive pairs should be made in the tester by pressing buttons on the front panel. No components of the link under test should be touched. The two near-end crosstalk measurements should be equal, and the degree to which they differ is a measure of the accuracy and quality of the device. Agreement should be within the accuracy statements of the tester. If this is not the case, there is a problem with the unit.
Another useful sanity check is interchanging tip and ring in a near-end crosstalk test. This is more difficult because it requires building a test fixture that allows tip and ring to be interchanged easily. Unfortunately, a test instrument that can perform this test at the touch of a button does not exist. The most practical approach is to use an external test fixture (see "How to gain confidence in your UTP cable tester," March 1994, page 39).
The near-end crosstalk readings obtained before and after interchanging tip and ring should be in close agreement, within the accuracy of the measurement device. A noticeable disagreement indicates a problem with the unit.
The last sanity check is measuring attenuation. Measure link attenuation in both directions. Locate the transmitter of the test equipment at one end of the link, and take a measurement. Then, connect the transmitter to the opposite end of the link, and take another attenuation measurement. The measurements should be in close agreement; a major discrepancy indicates a problem with the unit.
No allowance is made for signal degradation caused by the connectors on the test equipment in the Annex E parameters SP-2840. This is a consideration when you are measuring near-end crosstalk. Crosstalk resulting from the connection between the link and test equipment is quantified as residual near-end crosstalk (see "Balance and residual crosstalk for field test instruments," September 1994, page 51). It is a measure of the near-end crosstalk reported by test equipment terminated by a connector and 100-ohm resistors. Ideally, you should measure no near-end crosstalk from this source; the amount measured is an indication of the quality of the test instrument.
Residual near-end crosstalk exceeding 50 decibels at 100 MHz is desirable, while 40 dB at 100 MHz is borderline unacceptable. Before purchasing equipment, you should know the value of this parameter.
Another point to consider is that the Annex E link consists of a fixed length of cable. The attenuation numbers reported by TIA are for the entire link of 100 meters plus connecting hardware and are not attenuation per unit length. It is possible, therefore, for a short Category 3 link to meet the attenuation and near-end crosstalk numbers for a 100-meter Category 5 link listed in Annex E.
Does this mean that this Category 3 link can be used to support Category 5 applications? No. A Category 3 link is susceptible to electromagnetic interference at higher frequencies.
Remember that a Category 5 link is defined as one that contains only properly installed Category 5 components. Category 5 cables must meet a specific attenuation-per-unit-length requirement. Testing based solely on Annex E cannot provide any information on the category of a link. The category must be known before testing begins, or you will not be able to determine performance.
Another critical parameter affecting accurate data collection is balance. This parameter is also important in measuring near-end crosstalk. The near-end crosstalk numbers in Annex E are for balanced signals that have no common-mode component. This means that the test equipment is required to emit a negligible common-mode component in the signals used in testing a link. The ratio of differential to common-mode components is often referred to as balance and is usually expressed in decibels.
Balance is affected by the quality of the magnetics used in the test equipment. This is a critical parameter, and no equipment should be purchased without understanding its value. A balance of 50 dB or greater at 100 MHz is considered excellent.
The last important electrical parameter is return loss. This is a measurement of how closely the source and load impedances of the test equipment match the characteristic impedance of the twisted-pair transmission line. Unless the transmission line is properly terminated at both ends by the measurement equipment, reflections that can corrupt measured data will occur, leading to inaccurate results and inappropriate performance determinations. A return loss of 20 dB at 100 MHz is considered good.
Interpreting test data
Once you have collected accurate data, you must process and interpret it. The desired result is that all measurements will indicate enough margin so that the user will be confident the link can support any application. But what do you do when your data indicates a problem?
The most difficult problem to resolve is near-end crosstalk, and its most likely source is excessive untwisting of wire pairs during termination on connecting hardware. This problem can be identified from measured data if the dynamic range of the test instrument is large enough.
Near-end crosstalk coupling should be uniformly distributed over the length of the cable run. In this case, the curve of near-end crosstalk loss versus frequency displayed on the screen of a test device will show alternating peaks and valleys at regular intervals as frequency increases. The spacing of the peaks and valleys is a function of the cable length; the spacing narrows as the length of the cable increases. This is an indication of a healthy link.
On the other hand, if the near-end crosstalk coupling is dominated by a single point and is not distributed over the length of the cable, then loss as a function of frequency will not exhibit any peaks and valleys; it will be a smooth function of frequency, which will appear as a linear curve on a logarithmic frequency scale. Point coupling can result from improperly terminated connecting hardware.
When the tester shows a problem with the link, examine the near-end crosstalk versus frequency dependence of the data to see if it has peaks and valleys or is a smooth curve. If you do not see peaks and valleys, then check the termination to your connecting hardware and you will probably find the problem.
The next requirement for the test instrument, then, is that it should have enough dynamic range that you can distinguish the peaks and valleys of a healthy link. A dynamic range of 60 dB for near-end crosstalk and an accuracy of a few decibels ensure peaks and valleys in the crosstalk data will be resolved adequately.
Advocates of portable test equipment argue that specification of the unit`s accuracy is necessary only when the measurement is close to the Annex E limit. According to this argument, only the worst-case margin to Annex E is important, and the rest of the near-end crosstalk data is not really useful. This approach is inadequate, however, if you want to solve problems and find problem links. If the test equipment prints out a number, you should know the accuracy of that number.
What to look for in Category 5 Testers
Here are the requirements for an ideal portable test device:
- It should be able to interchange transmit and receive during a near-end crosstalk test by pressing a few keys on the front panel. The interchange should be completely internal to the unit, and the near-end crosstalk measured before and after the reversal should agree within a few decibels at all frequencies. Discrepancies may indicate a potential problem with the internal switching matrix used in the device.
- A tip/ring reversal during a near-end crosstalk test should indicate agreement within a few decibels at all frequencies. This test is accomplished external to the equipment, and poor agreement before and after tip/ring reversal could indicate a problem with magnetics, balance, residual near-end crosstalk or all three.
- Attenuation measured in each direction on the link should be in agreement within 1 dB. Anything worse could indicate a problem with a passive loopback strategy used by the device.
- The residual near-end crosstalk should exceed 50 dB at 100 MHz.
- The balance should exceed 50 dB at 100 MHz.
- The return loss of the unit with respect to a 100-ohm reference should exceed 20 dB at 100 MHz.
- The dynamic range of near-end crosstalk measurements should extend to 60 dB, and accuracy statements should accompany the complete dynamic range of all measurements, not just those near the Annex E limits.
Robert A. Conte and Masood A. Shariff are members of the technical staff at AT&T Bell Laboratories, Middletown, NJ.