Fiber

Some interesting twists about Ethernet cabling

Michael Kassner answers the question: why does Ethernet cabling use twisted-pair conductors? Here is the explanation, including some tips learned from cable installers about proper installation.

In a recent post, Jack Wallen explained how to make Ethernet cable connections. In the ensuing comments, I noticed some interest in Ethernet cabling itself. That's understandable, Ethernet cabling does not look or act like normal wiring.

Differential signaling

To understand why Ethernet cable is built the way it is, it requires familiarity with how the signals are transmitted over the wires. The fancy term is differential signaling. During a digital pulse, each wire in the pair carries a signal that is the same voltage, but opposite polarity. The slide below (courtesy of Wikipedia) exemplifies the process:

The greater the difference between the input pulses, the larger the output pulse, making it easier for the receiver logic to differentiate ones and zeros. I might add that this becomes more important as throughput rates increase.

As you can see in the slide, differential signaling has some noise-canceling capability, but not enough when it comes to Ethernet cabling. That's because two types of interference come into play, electromagnetic radiation from sources such as power wires or crosstalk from other pairs in the same cable. Both introduce noise that reduces the differential. If there is enough noise, the receiver can misread what was sent.

It's all about the twists

The Ethernet cabling most of us are familiar with consists of four pairs of wires, twisted together, and enclosed in a single insulating cover jacket. Those twists are a pain when making connections, but are there for a reason. Transmitting Ethernet signals over a cable that small would not be possible without them.

To understand the importance of twisting the wires requires learning about common-mode rejection. It seems that digital electronics connected with balanced lines, such as twisted pair Ethernet cabling are capable of rejecting noise, as long as the spurious emission is common to both leads in the twisted pair.

It might help to use an example. I mentioned crosstalk earlier. If pairs were not twisted in Ethernet cabling, it is entirely possible for two wires in adjacent pairs to be next to each other for the entire length of the cable. That could skew the results by adding cross talk to only one leg of the pairing.

If the wires in the pair were twisted, both wires in the pair would be affected the same. Then the receiver using common-mode rejection would be able to filter out the crosstalk interference.

Geek trivia alert: This phenomenon was discovered by Alexander Graham Bell. In fact, telegraph wires were the first twisted pair. The wires switched position on the pole after a certain distance (courtesy of Wikipedia):

Another twist

Ever notice that some of the twisted pairs are easier to un-ravel? There is a reason for that. If adjacent pairs have an equivalent twist rate or pitch, the same wires of each pair could be next to each other for the entire run, negating differential signaling. In order to prevent that, Ethernet cable manufacturers use different twist rates (courtesy of Wikipedia):

One last note, according to cabling best practices, pair wires should not be untwisted more than 13 mm. This comes into play when using punch-down blocks.

UTP versus STP

There are two types of Ethernet cabling, Unshielded Twisted Pair (UTP) and Shielded Twisted Pair (STP). Everything that I mentioned up until now, applies to both. STP cabling is used if there is an abnormal amount of electromagnetic interference. STP uses metal-foil shielding that directs any external noise to ground. The metal foil can surround each twisted pair (STP), all the twisted pairs S/UTP), or both (S/STP).

Some installation tips

Every time I work with cable installers, I bug them mercilessly, trying to learn what they consider important. Here are some of their pointers:

  • For solid wire UTP, the minimum bend radius is eight times the outside diameter of the cable. Anything less affects the twist rate, reducing noise rejection.
  • When installing long runs be careful not to stretch the cable, doing so could alter the twist rate, again reducing noise rejections.
  • Due to the high frequency of digital transmissions, the phenomenon of skin effect comes into play. So be careful not to nick the copper wire.
  • Make sure to use plenum-rated Ethernet cabling, if the run is located in a space that is used for air circulation.
  • Run Ethernet cabling at least 15 cm from any high-voltage lines, 30 cm is better.
  • If Ethernet cabling must cross a high-voltage line, do so at a 90 degree angle.

If I missed a tip you feel is important, please share it with the rest of us.

Final thoughts

As digital throughput rates keep increasing the margin for error decreases rapidly. Make sure to keep those pairs twisted.

About

Information is my field...Writing is my passion...Coupling the two is my mission.

40 comments
IT
IT

Please tell me the longest length permitted in varies situation.

mmoran
mmoran

About 20 years ago during my wireless communications days, I was called to a local high school where every time the dispatcher keyed up the 2-way radio base station to talk to a bus, the entire school network crashed. Turned out that some new network drops had been laid above the ceiling side-by-side for about 60 feet with the coaxial cable that connected the base station to its antenna outside. Stranded shield coax does NOT prevent ALL transmit signal leakage.

PCH350
PCH350

If I recall my circuit theory correctly, common mode rejection is due to the differential signaling, regardless of whether the pair is twisted or not. Twisted pairs reduce noise by making sure the induced noise is all common mode. See http://en.wikipedia.org/wiki/Twisted_pair.

CoderWPF
CoderWPF

What is 'high-voltage'? 120v, 240v, or higher?

us
us

Just pointing out the obvious! In the normal situation as Michael points out, the two wires are twisted together. A common problem is that sometimes the pairs are "legged" -- by mis-connection one wire of one pair is paired with one wire of ANOTHER pair. This may be only for a small segment of the total length of the cable and sometimes with a compensating error somewhere along the length so that at the ends of the cable the colors are correct. In this situation the cable will test fine with a DC cable tester but be total disaster with a live signal as effectively there is no twist at all for the length of cable affected Another situation which produces a similar result and is equally hard to find using simple test methods is when there is a dead pair connected in parallel across the active pair. The most simple way to prove either of these conditions is to rebuild the entire length of the cable and re testing with a live signal Great article and so useful to understand more about this

V.H. Scarpacci
V.H. Scarpacci

Around 1983,before ethernet and high speed cables, I was work on the prototype of a test bench design specifically for running diagnostics on a single type of unit. There was a cable that that connected the unit to the computer of the bench and as we were assembling the cable it was a rat's nest of 50 different wires. This cable worked great so we sent it off to get a neatened up version made by the cabling group. When we got it back it looked great, nice assembly job sheathed and all, but it didn't work. We check the pin to pin continuity and all was put together correctly. When we put a scope on the data lines we discovered signals from adjacent lines crossing over to each other. We went about tangling up and making another rat's nest with this cable and sure enough it worked. One of the engineers had just been reading about crosstalk and twisted pairs so we implemented his suggestions and got a power drill and all new wires to rebuild the cable. This time it worked and was neat and fit inside the sheathing. That was a fun project to have worked on and figuring out that cabling was a great puzzle.

kevaburg
kevaburg

Although I enjoyed the article I think there are a few points that I believe make make it a little clearer. Firstly, the tighter the twists in the cable (more twists per inch) the less chance there is of a phenomenon known as "detrimental mutual influence". If a loop or a gap appears between the cables of this pair, an induction loop is created that produces "crosstalk". This crosstalk results in interference that distorts signals traversing the pair. Although this is a very bad thing in communications, it is also one of the principals behind electrical generator where a cable is passed through lines of magnetic flux to induce voltage. It is this specific issue that distorts a signal between two points because the voltage being transmitted will vary according to what is being sent. This variation produces a movement of this magnetic flux effectively passing through the other line. To that end, anyone attaching connectors to the end of the cable should try and keep the twists as tight as possible right up to the point of entry into the plug itself. UTP or STP cabling (although STP to a lesser extent because of its shielding) should therefore also not be laid next to power cables (especially 3-phase high-voltage) because the magnetic field surrounding these AC power lines is very strong and is unlikely to allow a successful data or voice transmission along copper cable. That is why, in areas of high electro-magnetic interference (such as machine shops) fibre-optic cabling is the medium of choice. This is a priciple as well behind some types of data theft or wire tapping. Simply attaching a clamp meter around the cable is enough to detect variations in this magnetic flux so that another device can record and interpret the information. Lying a cable next to an active one can also lead to the same effect if done correctly. Just my ten pence worth..... :)

Michael Kassner
Michael Kassner

But, you have to be careful to include patch cables in that measurement. So, installers typically do not exceed 90 meters on any given run. If you are running contiguous lines between repeating devices, then the cable can be 100 meters.

Realvdude
Realvdude

Started my "tech" career in cash registers. We had a donut shop that called for service frequently. We would find the programming corrupt, but everything else within norms. With the help of the customer's diligent observations, we discovered the culprit was a police officer responding to a call on their portable radio in close proximity of the cash register. When we contacted the manufacturer, they faxed a pre-release tech bulletin to correct a design flaw in the switching power supply. The flaw modulated the 5v power, which was interpreted as logical signals. The fix cost a few cents in parts. When we experienced the same symptoms at a gardening nursery, we looked for the radios, even though the register was from a different manufacturer. The seasonal outdoor register had a walkie-talkie sit against the case.

rmerchberger
rmerchberger

Back when I co-owned an ISP (when 28.8Kbit was considered fast... ;-) ) we offered free hardware technical support for our customers -- even onsite -- none of our competitors offered free onsite tech support for connectivity problems. We had one customer who said that if we couldn't solve his internet disconnection problem, he was just going to give up on it entirely... we were the 4th ISP he'd tried if memory serves, and no one could seem to find his connectivity issues. He would usually connect fine, but sometimes for 5 minutes, sometimes an hour or two, but he would *always* get disconnected at some point. When he brought the computer to our store, it would stay connected for hours on end, so it wasn't the modem. We asked him to call when it would disconnect, keep a log of date/time of drops, etc. There seemed to be no rhyme or reason as to his disconnections. A few days later with no resolution, I went to his home, took apart both the outdoor and indoor phone hookup junction boxes & checked connections (also without help) and so I started tracing his line... He didn't have a phone jack where his computer was, so he poked a small hole in the baseboard from the kitchen where there was an unused jack... and had run his phone line as inconspicuously as possible... Right behind his refrigerator. Every time his fridge would kick on, the interference caused by that nice big inductive motor in that nice unshielded untwisted phone cable would kill his connection every time. I rerouted his cable through the basement away from any big motors & lights, and he never had an issue since. He was our customer for over 10 years... ;-) Laterz!

Michael Kassner
Michael Kassner

A difficult find. As a amateur radio op, I know about this problem big time. I had added a coax connector incorrectly once. A few of the ground shield wires were touching the center element. Oops. Thank you for sharing.

Michael Kassner
Michael Kassner

"To understand the importance of twisting the wires requires learning about common-mode rejection. It seems that digital electronics connected with balanced lines, such as twisted pair Ethernet cabling are capable of rejecting noise, as long as the spurious emission is common to both leads in the twisted pair." Twisting the pair is additional insurance to maintain common-mode.

egamblin
egamblin

See http://www.cabling-design.com/resources/documents/emi/569.shtml The current BICSI recommendations are: Provide the following minimum clearances: 1.2 m (4 ft.) from large motors or transformers. 30 cm (1 ft.) from conduit and cables used for electrical power distribution. 12 cm (5 in.) from fluorescent lighting. IIRC, the original recommendations for separation distances from up to 480v power cables weren't based on actual testing. (The reason this table was removed from later versions of the EIA/TIA 569 standard). In the early 2000s, some modular furniture mfrs sponsored testing to find out exactly how little separation from 120v AC cables was needed. (This applied to the metallic raceway built into mod furniture, where there's usually a metal strip between the power and voice/data cable compartments). I don't have a reference for you, but the recommendation was something like 2 inches minimum. Lucent's recommendations for its Systimax cable are: Lucent Technologies published several tables containing power separation information in its "SYSTIMAX SCS Design Guide" depending on power conductor cross-sectional area, maximum current, cable length etc. In many cases, SYSTIMAX installations of cable lengths of up to 90 m require zero separation distance when adjacent power circuits are powering typical office equipment and limited to 250 Volt, 20 Amperes. Other separation distances vary from 0.02 to 0.8 m. SYSTIMAX SCS Power Sum and GigaSPEED cables require no separation when placed above or adjacent to fluorescent light fixtures. Lucent Technologies tested SYSTIMAX SCS susceptibility to GSM cell phone emissions using 100 Mbps Fast Ethernet and 155 Mbps ATM network protocols. All channels remained error-free even when the antenna of the cell phone was placed directly on the cables or cords. Finally, Siemon has a lot of current info on its website: http://www.siemon.com/us/standards/

Michael Kassner
Michael Kassner

That all of the above apply. I do know that installers try hard to avoid 120/240.

egamblin
egamblin

As you mentioned, on horizontal UTP runs or patch cables a simple continuity test isn't enough. The quickest way I've found to detect split pairs (a miswire condition where a line from each of two pairs forms a pair) is with a cable tester like the Fluke 610. This is a relatively inexpensive unit (sub-$400), especially compared with a fullbore unit for performing all the different tests to certify to Cat5/Cat6 specs. Regarding making your own Cat5 patch cables: have you actually tested these with a test set capable of doing the full range of Cat5 tests? (NEXT, FEXT, etc). I ask because I've come across a lot of homemade patch cables that worked fine for 10BaseT, but failed when used for 100Base-Tx.

Realvdude
Realvdude

Thanks to a Fluke meter and some patience, I questioned the abilities of two field techs, and the culprit was found behind a couple of wall plates. The symptoms were sporadic, thanks to a piece of office equipment that only used intermittantly. When in effect, the ping tests on the meter would fail and the noise would go through the roof.

Michael Kassner
Michael Kassner

I sensed that right away. I love hearing from battle-hardened veterans.

Michael Kassner
Michael Kassner

Dare I say, Ethernet cabling and the science behind it why the Internet works. I also would love to hear more about your efforts. It's quiet people like you that made this amazing thing work.

Papa_Bill
Papa_Bill

...the principle behind radio communications. Antennae are basically unshielded and unbalanced lines to maximize radiation.

alan
alan

Before the internet there was the telephone, and before that was Alexander Graham Bell. Alexander Graham Bell invented twisted pair cables with different rates of twist. See http://en.wikipedia.org/wiki/Twisted_pair Principles do not change, only the application and bandwidth alter.

Michael Kassner
Michael Kassner

Forgive me, I struggled with your post. You mentioned several issues, could you explain in detail what you meant for each?

jcitron
jcitron

Very interesting post here. It brought back some memories for me! About 20 years or so ago, I worked for a company that remanufactured video terminals to Tempest standards. One of the tests we did was to check for "leaks" on some cables and equipment that was being installed. We could see the signals along an unshielded 25-pin serial cable, but they disappeared on the properly shielded one. Later on I worked for a company that used thick-net. Yup, that old thick stuff that required a vampire tap to connect to it. When we did an install, we had to ensure that the cable wasn't bent too tightly because the coax shielding would break, causing signal loss. In some respects this issue still happens with standard Cat-5/6 cables today, and as you pointed out running the cables close to a power source can degrade the performance. One of the biggest culprits is those florescent lights found in office buildings. It can become difficult keeping the the cables away from the ballast boxes and transformers used to power them.

Michael Kassner
Michael Kassner

To figure that out. Kudos. What was the frequency being used?

Michael Kassner
Michael Kassner

I totally love hearing about experiences like this. Thank you for sharing.

Papa_Bill
Papa_Bill

It might be handy to realize that 120V lives are usually *unbalanced* since one side is "hot" (carrying power) and the other side is "ground" (or "earth" to our British cousins). The 240V drop coming to a building is actually balanced since it carries two 120V lines that are 180 degrees out of phase with each other, canceling most radiated power.

egamblin
egamblin

The current equivalent of the product I mentioned is their Micromapper. It does a full wiremap test.

kynth
kynth

Most of what was listed actually occurred during development of early printed circuit boards, esp those for military/ spacecraft and vacuum tube tech example early radar systems and computer systems. When designing circuit boards, crosstalk between parallel runs and em radiation was critical. Most EE's had this knowledge from all this earlier work and as freq increased so did the effects esp. as the increased use of low voltage transistor tech. developed during the 50's and 60's. Many aircraft system boxes worked well in test, but when implemented in the actual system, aircraft and shipboard types the cabling induced errant signals and output enough so total system failure could occur. enough on history the artecle was excellent.

Michael Kassner
Michael Kassner

I have all sorts of radiators, being an amateur radio op. In fact, we are primary licensees on a portion of the 2.4 Ghz spectrum. It's used for sat comms. I am careful not to blast the neighbors too hard.

Michael Kassner
Michael Kassner

That's why I wrote about it and included the slide from the Wikipedia web site.

Michael Kassner
Michael Kassner

I should have linked that in article. I used the paper in my research. Could you tell us what the name of the device is? I would be interested in reading about it.

Michael Kassner
Michael Kassner

I so know about that stuff. It was quite amazing at the time. But, dare I say really tough to work with.

Michael Kassner
Michael Kassner

I appreciate you sharing that information. IT appears to be a matter of degree and not all or nothing.

Andrzej_Ladosz
Andrzej_Ladosz

Actually phases in AC (alternating Current) circuits are made of 3 phases spread 120˚ (degrees - electrically). Phase voltage of 220V (or 231V as per newer European standard) is between one (of three) phases and NEUTRAL. That Neutral is common for all 3 phases in STAR configuration or does not physically exist in DELTA configuration. If you have 220V per phase then between phases you have 380V. Equally 231V per phase equals 400V between phases. Similarly 120V translates to 207.8V, 110V to 190.5V etc. The factor equals 1.7320508075689 =√3 (square root of 3). The frequency 60Hz in N.America or 50Hz rest of the word makes no difference. I don't wan to elaborate about full 3-phase load - symmetrical or not. Sufficient to say that SYMMETRICAL LOAD (theoretically) creates *no Electromagnetic* field around supply wires. Clamp Ammeter is not able to measure anything when put around all (3 or 4) supply wires. But the "ideal" situation does not exist and the best solution is correct separation i.e. don't run these installations (cables) too close. Avoid proximity of big motors (AC or DC) and any rectifiers, VSD (variable speed drives, filters, Power Factor correction capacitors and any other equipment creating harmonics.

Michael Kassner
Michael Kassner

I did not think of that aspect. Thanks for pointing it out.

Michael Kassner
Michael Kassner

Fluke does make nice equipment. I wish I could afford it. I use a simple cable tester that checks the pairing and not much else. I do some network installation for General Mills and they require every Ethernet run and patch cable to be certified.

thegreenwizard1
thegreenwizard1

Yes the devices were from Fluke. I don't remember the model since as an installer we where renting them from the cable provider.