Return loss - reflections on the data link

Return loss with metallic conductors

One high-frequency parameter that often fails during acceptance measurements of data links is the so-called return loss (RL). This article will take a closer look at how the measurement of this parameter can go wrong despite the conscientious selection of high-quality components and careful installation of the cables on the distributors and data sockets. But before we go into the individual errors, we must first deal with the theory behind this measured value in order to be able to draw conclusions about possible causes of errors.

The return loss, also called backscatter loss, is a measure of the reflections that can occur on a data link. The return loss exists for metallic conductors and for optical fibres. It is the ratio of emitted power to reflected power, given as a logarithmic measure in decibels (dB). With metallic conductors, which are the topic here, such reflections occur at inhomogeneities of the wave resistance (impedance) within the cable run.

Stable impedance as a quality feature

The impedance of a data link should nominally be 100 ohms. It is made up of the individual variables resistance, leakage, inductance and capacitance. The value is also frequency-dependent to a small extent. These basic variables of the cable theory are essentially defined by the mechanical construction of the cable. Any mechanical overloading of the cable that affects the construction will automatically change the impedance and lead to deterioration of the reflection behaviour. It is important that all parts of the track, cable and connection components have the same impedance of 100 ohms, otherwise unwanted reflections will occur at the transition points and the transmission performance of the track will suffer as a result.

Possible sources of error can therefore be all the individual components of a data link, in addition to technical inadequacies when laying or terminating the links. Even the connection cords used for the active components can turn a good installation link into a problem if the impedances do not harmonise. In the further course of this article, we will take a detailed look at which source of error has what effect, using typical error patterns on the measuring device.

Measurement

The return loss is determined in the field using a cabling certifier such as the WireXpert from Softing IT Networks GmbH. Both LOCAL and REMOTE devices are connected to the data link and a single or auto test is started. The devices successively send a swept test signal in the frequency range of the selected test standard to each wire pair from each side over the route. The respective remote device provides the corresponding nominal termination. Each transmission channel now determines the specific return loss value via the integrated reflection measuring bridge and all the individual values obtained are processed and displayed in a clear collective graphic including a limit value curve. In simple terms, the higher the measured value achieved and consequently the greater the distance to the limit value curve, the better the reflection and transmission behaviour. However, it can also happen that individual measured values sporadically exceed the limit value curve during measurements in the lower frequency range despite the use of good material and appropriate processing. In this case, the measurement would actually have to be evaluated as faulty, but there are certain exception rules in the standards for the evaluation of some high-frequency parameters. Since special effects when measuring or properties of paths would evaluate a measurement as faulty, although the transmission properties are sufficient for error-free data transmission, some correction formulas have been integrated into the standards, which of course must also be implemented in the measuring devices.

For the reflection behaviour, for example, the following applies: If the insertion loss of the measured path is less than 3 dB, the return loss is not evaluated, but only indicated for information. The legitimisation behind this is that this can only happen in the lowest frequency range where there is no heavy data transmission and therefore the signals are not relevantly affected.

The effect of exception rules can be recognised in the WireXpert measuring device by a two-colour limit line. The first part of the limit curve is shown in green, here an exception rule takes effect and the measured values are independent of their position to the 
limit curve are "informative only" and do not contribute to an evaluation. Subsequently, the curve turns red and from here on the good/bad evaluation applies.

Locator return loss

If the return loss measurement for the route fails, then the WireXpert offers the return loss locator function for quick fault location, which provides a detailed profile of the signal reflections over the entire length of the route. The graphic shows the reflection behaviour as amplitude deflection at each position along the cabling route. A freely movable cursor can either be moved manually to a reflection event or successively set to the individual interference peaks by means of a peak detector. Below the collective graphic, the unit displays an overview table that shows good/bad indicators in pairs and pages on the one hand and controls the display content of the graphic on the other hand. In this way, the affected wire pair and the affected location can be clearly localised.

Causal research

We want to look at the possible sources of error in theory in order to find them again later in real measurement examples. Let's start with the installed components, which in the simplest normal case include the installation cables, the distribution panel and the data socket. Let's assume that the appropriate power categories were used here to achieve the desired power class, and let's assume a professional installation.

Now how can such inhomogeneities of impedance occur on a data cable so that the reflection values crash? This happens when the production processes have been disturbed and possibly one or more pairs of cores are not twisted correctly over a longer distance, or the shielding foil around a pair has possibly flipped over. Fortunately, such production faults are usually already discovered in the manufacturer's quality control and rarely reach the market. The same applies to the termination components. Here, too, it can sporadically happen that insufficient quality was produced and that errors now show up in the high-frequency parameters when the track is calibrated. But here, too, quality assurance puts a stop to this, at least with the well-known manufacturers.

Cable

Probably the most common return loss problems of a system are buried in the condition of the installed cable. As mentioned above, the stable characteristic impedance is probably the most important parameter of the cable to pay attention to. Any mechanical change will immediately lead to a change in the pad parameters and a shift in the impedance value. As a result, reflections will occur that will affect data transmission.

The worst stress you can put on a data cable is probably pulling too hard, possibly over too small a bend radius when installing the cable. Depending on the manufacturer, a modern double-shielded Cat 7 cable has a maximum tensile load of approx. 100 to 350 N, i.e. approx. 10 to 35 kg. This is not much, considering the often (too) small dimensioned cable runs that resist the cables. Such an overload of the cable will result in a strong change of the geometrical structure inside, which in turn will result in a change of the covering values. This causes a change in impedance, which leads to mismatches and consequently to reflections that weaken the transmission properties. It is not by chance that the professional speaks of "insertion" when inserting the cables and shudders at the word "pulling in". Such a cable can be recognised in the return loss locator by a measurement curve that shows interference amplitudes over the entire course. Unfortunately, (too) many installers still believe that a pull-in is no longer "visible" once all covers are closed and the cable is no longer visible. A certifier reveals such "mistreatment" of a cable with only one measurement. In this context, compliance with the permissible bending radii should also be pointed out. The values permitted here differ for the installation case and later operation. Please refer to the respective manufacturer's data sheets.

Defects from the third parties

However, there are also a number of damages to a cable that cannot be attributed to an installer and are only noticed after installation during calibration. A classic example of this is the forklift truck that has driven unnoticed over the laid-out cables and has destroyed the internal structure through its weight, but has not left any serious external marks on the sheath. If you then fail the RL measurement and display the location graph, you will find two conspicuous interference amplitudes on the cable run with a distance of about 2 metres, the same as the track width of the forklift.

The only thing that can help here is to replace the cable. Also nasty are faults caused by other trades, e.g. the drywaller who accidentally cut a data cable and then amateurishly repaired it with a luster terminal. A large spike will appear on the meter's display at the site of the operation. If you also look at the location graph for the crosstalk at this point, you will also see a large spurious amplitude here. If both important high-frequency parameters fail, one can assume an insufficient connection. Both the shielding mechanisms and the geometric structure of the cable have been destroyed. In contrast to our forklift from above, where mainly the geometry has suffered and the shielding mechanisms are still intact and therefore the crosstalk can definitely still be assessed as OK.

Connection components and connection technology

The next source of reflection interference is to be found in the connection components or rather in their contact with the cable. In the foregoing, we have already assumed the use of suitable components of the respective power category, so the cause of error of a mismatch between possible and required power is already excluded. However, there is still the factor of the technician and his connection skills. To be fair, one has to admit that these problems have become rarer nowadays due to the use of modern individual modules. With older compact boxes with LSA strips, it often happened that the high-frequency properties suffered due to massive changes in the cable geometry in the box, e.g. twisting the wire pairs too far open. Often, crosstalk and reflection errors occurred during remeasurement. If this is the case, the locator function of the WireXpert can be used to immediately determine the location of the fault at the beginning or end of the section. If the component is not internally defective, a clean reconnection usually helps to regain the high-frequency characteristics.

Error patterns

Here are some comments on the error patterns shown. The WireXpert is able to display the high-frequency characteristics both via frequency and location-dependent. The combination of both modes allows an exact and fast fault analysis.

Figure 1 shows an inconspicuous reflection curve over the entire frequency range of the selected 500 MHz standard as a reference. The spatial resolution also shows no anomalies and all combinations in the table are rated as good. This is how it should always be.

In picture 2 (Figure 2) you can see a section that was strongly overstretched during the insertion, the impedance was changed here and thus mismatches create reflections that cause the measurement to fail. In the locator it can be seen very nicely how the curve shows deflections over almost the entire length of the track.

A classic is measurement 3 (Figure 3). The return loss drops out and on closer inspection, two interference peaks of approximately the same height can be seen on all wire pairs. The distance between the peaks is about 2 metres, roughly the track width of a vehicle, possibly a forklift truck, which has driven over the laid-out cables, but which was not noticed when the cables were installed. The crosstalk behaviour over the track is also shown here. Since a good signal still comes out despite the "RL damage", the shielding mechanisms for the crosstalk (twisting, shielding foil) seem to be intact, i.e. the cable is not torn open, the wires are not exposed and untwisted. But the cable must come out!

In the fourth measurement (Figure 4), a return loss error has occurred again. This time the RL locator shows a large interference peak at a single point, but also again on all wire pairs. As a cross-check, the course of the crosstalk is again observed and this time a clear spike can also be seen here at the same location. This means that both the mutual shielding mechanisms of the wire pairs and the geometry of the cable were destroyed. What was the cause? The cable was accidentally cut at this point and "patched" again very unprofessionally with a simple luster terminal. But unfortunately, the high-frequency properties of luster terminals are not sufficient for this purpose. What are the remedies in such a case? You have three options: either replace the cable completely, shut down the track or actually reconnect it, but not with a luster terminal, but with special cable connectors, which are now available from some well-known manufacturers. But don't overdo it, only a maximum of ONE such connector per track is permitted.

Figure 5 shows the interference pattern when an inadequate component, such as a compact socket with too low a category, has been installed at one end of the line, or the high-frequency values have been caused by improper connection, e.g. with cores twisted on too far. In most cases, it helps to use suitable components or to reconnect in compliance with all current regulations, e.g. do not twist the wires more than 13 mm and do not invent your own connection diagrams.

The last measurement (Figure 6) shows the 3 dB rule in use. The limit curve is exceeded by the measured values in the lower (green) part, but the result is still positive. The RL locator display is equally positive. If even one value in the red part of the limit value curve were to exceed the line, the measurement would be assessed as "failed". No intervention in the system is necessary at this point.

Conclusion

The general rule for installing data links is that the cables used must not be "pulled in", but only "inserted". Care must also be taken when handling the data cables. Any mechanical stress must be avoided. When connecting, no experiments should be made, neither should one's own connection scheme be realised nor should one roughly tinker with the structure and the twisting system of the cable. Also when repairing lines, bear in mind that we are trying to transmit high-frequency signals here and not 50 Hz alternating current.

AUTHOR

Alfred Huber
Head of Technology,
Softing IT Networks GmbH