Network Cable Testing What Causes Data Loss?
An outline of the causes of network cable data loss and why gigabit cabling is more of a challenge than cables running at lower speeds.
You might be tempted to think that a cable is a cable is a cable, but then you look around and notice that there are lots of different types on sale, and that data rates seem to be much greater than they used to be. Why only a few years ago everyone had 10M Ethernet and now that’s old hat and everybody has at least 100M and increasingly 1000Base T (Gigabit) Ethernet. Something must be different.
Well the reality is that all cables are not equal, and increasing the data rate brings a lot of new factors into play. A cable that operates happily at 10M is a different beast at 1000M and as you start to use all four pairs in the cabling (as you have to do to get the faster rates) you introduce a whole raft of new problems that need to be considered when testing the cabling.
What Slows the Data Down?
An electrical signal is composed of electrons moving along the wire. As the frequency increases (and hence the potential amount of data you can send) a number of new phenomena appear, and these need to be taken into account when installing and testing the cabling. As a consequence the standards have had to change to take these effects into account.
There are three main classes of problem that cause data to be weakened, lost, or corrupted beyond acceptable limits.
- Attenuation (signal loss/degradation)
Attenuation (Signal Loss/Degradation)
We are all used to the fact that in nature as a signal travels further from its source it weakens. The same is true of electrical signals in wires. Beyond a limit (different for each type of cable) the signal is too weak or distorted to be recognisable. Cable length is the major factor for attenuation measurement.
Attenuation is also dependent on frequency, becoming greater as frequency increases.
Higher Temperatures increase attenuation too, about 0.4% per degree Celsius for Cat5e cabling.
Attenuation is measured in decibels (dB). Confusingly signal loss is measured in negative numbers, e.g. -3dB and the negative sign is usually ignored, so the example would read as 3db of loss. Consequently lower number are better. So 2dB is better than 4dB.
Just about all the electrical properties of the cabling have an effect on attenuation too.
Resistance is a function of the cross sectional area of the conductor. Resistance in the wire limits the signal and dissipates the energy as (a small amount of) increased heat. The longer or thinner the wires the greater the resistance.
The insulation covering the individual wires in the cable inevitably absorbs some of the signal. Since many wires are placed very close together they store this energy, acting in electrical terms, like a capacitor. High Density Polyethylene (HDPE) is commonly used because its electrical properties at high frequencies helps to minimise the losses. Cables that are designed for lower frequency applications may perform poorly at higher frequencies.
Electrically impedance is a combination of resistance, capacitance, and inductance expressed in Ohms. Typical cables are rated at around 100 Ohms. A so called Return Loss occurs when a signal hits a high impedance, for example an incorrect connector or a cable fault, and is bounced back. Potential bounced signals can cause problems on high speed networks, the higher the network speed the more pronounced the problem. Poorly fitted or wrongly specified connectors are a major cause.
Infinite impedance indicates a cut in the cable.
Zero impedance indicates a short circuit.
Any electrical signal on the wire not part of the sender’s original signal is classed as noise. Noise is generated both internally and externally.
Twisted pair cables produce no interference, the twists cancel each other out, in theory that is. In real life any variation in the thickness of the wire, in the cable insulation, and in the capacitance of wires or insulation will cause a mismatch and consequently noise. Good quality cables minimise the noise but cannot remove it altogether.
Electrical interference can come from many sources. Cables should always be installed in separate conduits away from mains cables. In industrial applications electric motors (in lifts/elevators) fluorescent lights and air conditioners, are major sources of interference.
In areas of electrical noise it is common to shield cables or to use other technologies, such as optical fibre to avoid interference.
Crosstalk is likely to be much greater than any other noise effect. When a signal travels down a conductor, an electric field is created, which interferes with any wires close by. This is Crosstalk and gets larger at higher frequencies and the more parallel the wires. The twists in the pairs should (in theory) cancel this effect. For good signal cancellation it’s important that the twists are symmetrical and that adjacent pairs have different twists.
Crosstalk is measured in decibels (dB).
Near End Crosstalk (NEXT)
Crosstalk at the end of the cable where the signal originates is called Near End Crosstalk (NEXT). The near end is up to about 30 metres from the source.
10M systems suffer more from crosstalk than 1000Base T systems. This is because more sophisticated cancelling techniques have to be employed on faster systems that use all 4 pairs of wires in the cable.
Crosstalk can be a problem on cable connectors or patch panels, where the wires have to be untwisted to make the connections. Always keep the section of untwisted cable to a minimum, even a very short piece of untwisted cable can introduce a large amount of crosstalk.
Far End Crosstalk (FEXT)
If you are following this then you won’t be surprised to discover that Far End Crosstalk (FEXT) is similar to NEXT, but at the far end of the cable. Signals will have been attenuated by the time they reach the far end of the cable and so are weaker. Not surprisingly FEXT is greater on short cables than on long cables. FEXT is a useful measure since it is involved in calculating the value of ELFTEXT, see the next section.
Equal Level Far End Crosstalk (ELFEXT)
This is a calculated value of the crosstalk between pairs measured at the far end of the cable. It takes into account the amount of signal loss. ELFEXT is calculated for each pair of cables and will be slightly different for each pair. Very high values indicate excessive attenuation or high far end crosstalk.
Pair to Pair Crosstalk
Since it is clear that a signal on one pair will have an effect on all the other pairs in the cable a good method of measuring the crosstalk is to place a signal on one pair and measure the disturbance on the others. This has to be done for each pair and at the near and far ends, resulting in 12 sets of measurements.
Measure Pair 1 to 2, 1 to 3, 1 to 4. And Pair 2 to 3, 2 to 4, and 3 to 4. Then go to the other end and repeat.
The worst value is the crosstalk for the cable.
Power Sum Crosstalk
Pair to pair interference is even more important in technologies that use all four cable pairs, since every pair affects every other.
In this test a signal is placed on all the pairs except one and a measurement taken. This needs to be done for 4 times, once for each pair, and again 4 times at the opposite end of the cable, resulting in 8 measurements. Again the worst result is power sum crosstalk for the cable.
Signal to Noise Ratio (SNR)
As the name implies this compares the signal strength to the amount of noise on the pair. A more accurate term for a similar thing is Attenuation to Crosstalk Ratio (ACR). Both terms are also slightly misleading. ACR isn’t a ratio, it’s just the difference between signal strength and NEXT, and SNR includes internal as well as external noise (which for most practical purposes makes no difference). You will hear both terms used to mean the same thing. Only in areas of extreme electrical interference will ACR and SNR be significantly different.
Obviously the difference between the level of noise and the level of signal is important. You are aiming to get a good signal much stronger than any noise. Attenuation greater as frequency increases and NEXT gets lower. The difference for any cable is the ACR. Theoretical bandwidth limits are always higher than the rates used in practical cabling.
Electrical signals travel very fast, but not infinitely fast. Typical twisted pair cables run at 60% to 90% of the velocity of light. The time taken for a signal to travel down the pair is the Propagation Delay. The propagation delay in itself is not normally an issue, the recommended cable lengths will already have taken this into account. However Delay Skew, described below is a significant factor.
Think back to the fact that the wires are twisted and therefore the length of the pairs that make up the cable are not equal. Signals sent at the same time will therefore arrive at slightly different times. This difference in arrival times is the Delay Skew, and must be within 50 nanoseconds for Cat5, Cat5e and Cat6 cables.
If Delay Skew is excessive then network devices will have trouble communicating, resulting in very slow or totally non-functioning networks.
All cables are not equal. As network speeds increase new factors have to be taken into account. This is especially important with technologies that use all four pairs of wires in the cable.
Choose a high quality cable to suit the kind of network that you intend to install. Always use good quality connectors and crimps, again making sure that they are of the correct type for the network. Pay attention to how and where you install cables to prevent mechanical damage to the cabling and to avoid areas of electrical interference.
Test and measure the installation in accordance with the standards applicable to the network type. Ideally test the cabling with real network data. If you do this you will be sure that the cabling will handle the data rates required by your client.