Field Measurements of Return Loss White Paper By Mark Johnston and Jim Tonti Microtest October 21, 1998 Page 1 12/15/99
Overview Return loss is an important new measurement used to qualify the performance of Category 5, 5E, and 6 cabling. It has just begun to be measured in the field, and this has raised a number of questions. Some of these questions can be resolved with education and training, while others may require action from equipment suppliers and the TIA. Other test parameters, such as NEXT, are now well understood. In the event of a NEXT failure on the link, for example, you can be certain that some part of the link is not in compliance with link NEXT requirements. With return loss, on the other hand, a failure, especially if it is close to the limit line, could be considered cautionary, rather than as a hard failure. Another important consideration is that the return loss requirements for Category 5, 5E, and 6 are all being developed and thus are subject to change. This White Paper defines and discusses some of these return loss issues, and offers suggestions to minimize unexpected results in field testing applications. What is return loss? Return loss is a measure of the ratio of signal transmitted into a system to the amount of that signal reflected (i.e. returned ) from that system back to the source. If the system were ideally matched to the source (i.e. its input impedance were exactly the complex conjugate impedance of the source), no signal would be returned. Any variation in the system impedance from the source impedance (typically 100 ohms) results in some returned signal. Real-life cabling systems do not have perfect input impedance, and therefore have a measurable return loss. Changes in twist, distance between conductors, varying copper diameter, dielectric composition and thickness variations, and other factors all contribute to slight variations in cable impedance. In addition, not all connecting hardware components in a link may have equal impedance. At every connection point there is the potential for a change in impedance. Each change in the impedance of the link causes part of the signal to be reflected back to the source. Return loss is a measure of all the reflected energy caused by variations in impedance of a link relative to a source impedance of 100 ohms. Each impedance change contributes to signal loss (attenuation) and directly causes return loss. There is a direct mathematical relationship between input impedance and return loss. Return loss is an important parameter for simultaneous bi-directional transmission systems (e.g. Full- Duplex 100BaseT, 1000BaseT, etc.), and is now a requirement for qualifying Category 5/Class D, Category 5E, and Category 6/Class E links. How is it measured? Because return loss measurements involve measuring signals going in to and coming out of a link simultaneously on the same pair, special measurement techniques and equipment are required. The standard approach is to use an RF Network analyzer, a directional coupler, and a balun to which the pair under test is connected. The directional coupler separates some of the transmitted and reflected signals passing through it to two separate ports, which are connected back to the analyzer. The transmitted and reflected signal levels are measured, and their ratio may be interpreted as return loss. However, directional couplers are not perfect (25 to 30 db of directivity is common), and baluns do not have good return loss characteristics themselves (20 to 25 db typically). A fair amount of correction is required to mathematically remove the non-ideal effects of the coupler and balun to make reasonably accurate measurements. For 100 ohm cabling systems, calibration is done with Open, Shorted, and 100 Ohm reference loads at the instrument s reference plane. All three calibration data sets are taken across the frequency range, with the magnitude and phase (i.e. the vector quantity) at each frequency being stored in the measurement instrument. When the cabling system is measured, these three complex data sets are applied to the measured data to determine the system s return loss. High dynamic range is required of a system to make accurate measurements. The calculation involving the Open, Shorted, and Terminated calibrations and the raw measured data is essentially a subtraction of the transmitted signal from the sum of the transmitted and reflected signals. Minutely small variations in any of Page 2 12/15/99
the calibration data or the measured data (e.g. noise caused by insufficient dynamic range) can cause many db of error in the final return loss calculation. A combination of very accurate factory calibration, ultrastable measurement hardware, and real-time field correction for gain and phase shifts is required to preserve return loss field test measurement accuracy. What limits are required? Return loss requirements have been proposed for Category 5/Class D, Category 5E, and Category 6/Class E basic/permanent links and channels. The table below summarizes the return loss limits in effect in the applicable draft standards. Frequency 1 f < 20 MHz 20 f < 100 MHz 100 f 200 MHz Category 5 Basic Link 1 15 15-7log(f/20) Category 5 Channel 1 15 15-10log(f/20) Category 5 Permanent Link 2 15 15-7log(f/20) Category 5E Basic Link 3 17 17-7log(f/20) Category 5E Channel 3 17 17-10log(f/20) Category 6 Basic Link 4 19 19-7log(f/20) 19-7log(f/20) Category 6 Channel 2,4 19 19-10log(f/20) 19-10log(f/20) Category 6 Permanent Link 3 19 19-7log(f/20) 19-7log(f/20) 1 TSB-95 ballot October 1, 1998 2 ISO/IEC JTC 1/SC 25/WG 3 N535: Updated draft specifications for Classes C, D, E, F. August 25, 1998 3 TIA 568A Addendum A5 (Category 5E) ballot October 1, 1998 4 TIA Category 6 Draft standard Revision 3, August 1998 The following graph demonstrates the proposed pass/fail limits for Category 5, 5E, and 6 channels, along with the actual performance of an installed Category 5E channel. This channel meets Category 5E return loss requirements. An important point to note is that the return loss of installed links is very close to required return loss. There is very little margin available to cover component variation or installation practices. 0-5 RL (db) -10-15 -20 Link Measurement Category 5 Channel Category 5E Channel Category 6 Channel -25 0 50 100 150 200 250 300 Page 3 12/15/99 Frequency (MHz)
How does the Microtest OMNIScanner measure return loss? The Microtest OMNIScanner measures return loss the same way a laboratory network analyzer does. It uses a full magnitude and phase (vector) measurement on each pair under test. At each frequency, a realtime measurement of the transmitted signal (loop-back) is made to correct for operating conditions, and Open, Shorted, and Terminated factory calibration data are applied to the vector measurements. The result is an extremely fast, accurate, and repeatable measurement of a link s return loss. How do measurements with the Network Analyzer compare with the OMNIScanner? Whenever comparing two test instruments, it is critical to ensure that the test environment is stable. Microtest has developed hardware tools that permit experts to directly connect cable or connectors to the OMNIScanner in the exact same manner as they connect them to the network analyzer. As well, we have software that interfaces between the OMNIScanner and the network analyzer that permits all key measurements to be performed and directly correlated on the PC for all pairs. If you are using a network analyzer and would like to perform this type of comparison, contact one of the authors for further information. When such a comparison is made, a very high correlation is seen. In NEXT comparisons, excellent agreement can be shown down to signal levels at 90 db out to 300 MHz. For return loss, excellent agreement is also shown across the entire range. In fact, this correlation is no different than would be obtained between two different network analyzers with two sets of baluns. An example of typical return loss agreement is shown in the figure below. NA / OMNIScanner RL Comparison on 92 meter Cat 5E Basic Link 0 1 mt 90 mt 1 mt -5-10 RL (db) -15-20 -25-30 -35 Network Analyzer RL OMNIScanner RL -40 0 50 100 150 200 250 300 Frequency (MHz) Page 4 12/15/99
This indicates that measurements made at the input port of the OMNIScanner will be the same as those made by a network analyzer. So if the OMNIScanner is accurate, and components are supposedly Category 5E compliant, why would links fail in the field? If I have Cat 5E cable and Cat 5E components, why do I sometimes fail return loss? There are several factors that contribute to return loss failures in field measurements. 1. Component interoperability Manufacturers have designed their plugs and jacks for maximum performance when they mate together. When you take a plug from supplier A and mate it with a jack from supplier B, the performance is generally no better than the poorer performing component. Exhaustive research has not been performed in this area, but tests suggest this effect can contribute up to 0.5 db variation in results. 2. Patch cord impedance variation Stranded cable is tested in long (often 1000 foot) lengths. Over long lengths, return loss of well-designed and constructed links is often excellent. However, the impedance structure of such links can vary significantly over short distances. Measurements on consecutive short (1 meter) lengths of stranded cable cut from the same spool have shown up to 5 db variations in return loss! The complex impedance of such short segments of cable is varying significantly from 100 ohms, especially at frequencies above 25 MHz. What does this mean? First, it can be a large and unaccounted-for source of unexpected return loss in the measurement. It also means that in the absence of manufacturer return loss-verified short segments of cable, test tool manufacturers need to individually qualify cordage to ensure it meets strict return loss consistency requirements. Microtest does this in its quality process for OMNIScanner test cords. Today, there is no standard for qualifying the return loss of patch cords. There is a draft standard for qualifying NEXT of patch cords, but it does not address return loss. For Category 5E and 6 links, Microtest strongly recommends users request return loss qualifying information from prospective suppliers. We have found some are doing an excellent job of qualifying cable, but many are not. As a result, there is a significant risk that patch cords can introduce enough return loss variation to fail a channel, where the installed basic or permanent link has passed. One way to improve the return loss performance of test cords or patch cables is to use longer lengths. Longer lengths may exhibit more consistent structure. This provides two benefits. First, the return loss performance of the cord is improved, and second, results are more consistent from cord to cord. 3. Installation practices TIA 568A mandates that all terminations should be twisted to within 0.5 inches of each connection. In case of Category 5 or 5E NEXT, there is enough margin in today s designs that installers may be a bit relaxed with this requirement and still easily meet NEXT requirements. However, the installation process significantly impacts return loss. Measurements show that relaxing the twist from 0.5 to 1 inch can increase return loss on some pairs by up to 2 db. This difference is more than enough to fail many links. 4. Lack of link margin Unlike NEXT at Category 5 levels, where a comfortable margin exists, actual return loss margin relative to the pass/fail line is very narrow. There is very little room for any issue that adds return loss. In fact, there is some concern that if a link is constructed with marginally passing components and cabling, and if a normal patch or test cord is used, the link may fail a return loss limit it has been designed to meet. Page 5 12/15/99
5. Short link In many cases the link is long enough so that its attenuation effectively hides the effects of far end connector, patch cord and termination impedance. But for short links, connector and patch cord return loss performance at the far end becomes a factor in a near end measurement. Not a great deal of study has been done in this area, and effects depend upon the length of the link, and relative magnitudes of the return loss of components at both the near and far ends. However, it is clear that marginal components will cause more return loss failures on short links (under 25 m) than on longer links. What can I do if a link fails return loss? 1. Observe which end of the link is failing. Inspect the terminations closely. Wire twists should be tight all the way to the connection points on the affected pair(s). Even ½ of untwist may make up to 2 db of difference in a link s return loss. 2. Try swapping the patch cord (if channel) or swapping ends with the test cord (if basic or permanent link). If the problem moves with the cord, consider the following: The equipment cord for channel/permanent link tests is constructed of components that meet or exceed the specified performance requirements for the cable and connectors. It is possible that the cable and components of the link under test are just barely compliant with TIA specifications, and the particular combination of the link under test and the equipment patch cord reveals this deficiency. (Return Loss specifications may have negative margin for worst case components and termination practices.) 3. Try using an equipment cord constructed of the same materials as that of the link under test. Microtest has available a custom patch cord kit, which allows users to construct equipment test cords from their own components. Matching a particular supplier s cordage and connectors may improve the impedance matching and thus the return loss of the link. 4. Try using an equipment or patch cord that is longer or shorter than the cord currently in use. This may shift the peak of the problem to a place where the return loss spec is easier to meet, or may actually result in some cancellation of return loss. In many cases, a 2-meter test cord (the maximum length currently allowed by the TIA) may give improved return loss results. 5. Verify that the test equipment used is capable of accurate return loss measurement performance. While the Microtest OMNIScanner exceeds these requirements, other products that can measure return loss do not. (This is also true for the ELFEXT measurement). What independent data does the test equipment supplier have that confirms return loss measurement accuracy? Conclusion Return loss is a difficult measurement to make, there is very little margin available in proposed standards, and these standards are still subject to change. A return loss failure, especially if all other parameters pass, is not necessarily an indication of a problem with the link. Given the number of issues yet to be resolved, marginal return loss results could be viewed as informative rather than as hard failures. We hope that this overview has helped you to understand some of the issues related to making field return loss measurements. By understanding and applying the techniques presented here, you will minimize any future return loss measurement difficulties, and reduce the instances of failed or marginal test results. Page 6 12/15/99