11

There are two types of pinouts for terminating twisted pair cable with a 8P8C (RJ45) connector:

enter image description here

But why must one not simply align the wires serially, e.g:

Green pair - Orange pair - Blue pair - Brown pair

Why do the current standards offer a nontrivial solution?

  • It's not clear which of two different questions you are asking. Are you asking: 1) Why wasn't Ethernet designed to allow aligning the wires serially? or 2) Given the way Ethernet is designed, can you align the wires serially? (And 2 has an obvious answer -- no, you would lose noise immunity for GigE.) – David Schwartz Jan 12 '17 at 18:13
12

UTP cables used in the 586 standard inherited their color code from the 25 pair color code developed by AT&T for cabling used in telecommunication purposes.

The following table shows the combination of colors to be used for each pair

enter image description here

This gives us this set of combinations:

enter image description here

UTP cabling simply used the first four combinations.

The order used in 586a and 586b is compatible with 1-pair and 2-pair Universal Service Order Codes (USOC) pinouts that are a requirement in federal contracts by the U.S. government.

Pair 1 connects to the center pins (4 and 5) of the connector. It gives compatibility with the first line of RJ11, RJ14, RJ25, and RJ61 connectors that all have the first pair in the center pins of these connectors.

The position of the other wires is because signal shielding would be optimized by alternating the "live" and "earthy" pins of each pair.

However it isn't possible because the outermost pair would be too far to meet the electrical echo requirements of high-speed LAN protocols.

That's why only one pair is "untwisted" and used as a shielding for the central pair.

  • 4
    I have my doubts with the shielding argument. When that pinout was set, speeds were so low, you could run them over a wet string, and the few mm of untwisting don't do any much shielding. If you follow the useful alternation pattern and already have 4&5 set, then there is not much other reason necessary, since the split up pair is the only two pins left for the fourth pair. – PlasmaHH Jan 12 '17 at 15:34
  • I didn't follow the untwisting is actually done and why and also outer pair echo impairment. Twisted wires have lower impedance and higher distributed capacitance and lower propagation speeds only if mean dielectric constant is higher. I wonder what explains why Full Duplex never works on my Ethernet at any speed with any cable including CAT6 maybe due to poor PHY Balun CM rejection of Tx or poor CM crosstalk rejection of adjacent pair for Rx or untwisting of wires !! Mine only works on simplex or Auto. at upto 1Ghz. But I certainly understand alternate Tip/Ring on pairs. – Sunnyskyguy EE75 Jan 12 '17 at 16:50
  • 1
    A signal carrying pair cannot shield another signal carrying pair. That's not how EM works. Especially since they are a TX and RX pair, which are by definition uncorrelated signals. The shielding argument is an old misunderstanding, as in my answer, the signals are planned pos-neg-pos-neg for preventing positive signal coupling, but this can be achieved with and without wrapping. – Asmyldof Jan 12 '17 at 17:14
  • There will always be a mutual CM&DM inductance of adjacent pairs with inbalance errors coupling partially from CM to DM in UTP. Finally got my FDX 1GHz working. Since Rx,Tx are uncorrelated it matters not if it was +-+-+- or +--++- as long as the pairs are adjacent. From my tests on PCB, I measure crosstalk in dB/rt( Hz) SNR – Sunnyskyguy EE75 Jan 12 '17 at 20:35
5

It is actually a very trivial and straighforward reason.

The standard starts with a single pair, two pins in the middle.

This expands to a 4 pin plug for two pairs.
What happens if you plan the pairs "straightforward" as you suggest? You break backwards compatibility with systems using only a centre pair.

So you need to wrap the second pair around the first pair, even though this is not good for your signal.

Then, the standard also allows for 4 pairs, but not 3 pairs. So you jump from the 2 pair to the 4 pair, so you obviously choose the set-up where you can wire in the pairs as fully twisted as you can.

The wrapping has, in fact, nothing to do with shielding, as the set-up does not help with any of that. At all.

The only thing to do with signal integrity and signal balance in some of the systems that use RJ-based plugs is the demand for colour-white-colour-white-colour-white...etc.

This actually does shield pairs from coupling into each other positively on the board. But this would have been achieved just as easily in your set-up.

The only reason it not being like that is legacy. That's all.

  • I don't understand - if I'm making my own patch-cord for, say, Fast Ethernet, where does the legacy come from? – Alexander Gonchiy Jan 13 '17 at 8:03
  • @AlexanderGonchiy That everything using RJ connectors is made to be backwards compatible. If you make a new cable now "with your own pairing", the devices will not use it, because they expect the pairing to be as the standard, which has pairing according to the legacy build-up. – Asmyldof Jan 14 '17 at 9:05
2

The current Ethernet 8P8C connector that carries 4 data pairs traces it's heritage to phone jacks carrying a varying number of phone lines. The first line starts in the middle, which allows using smaller plug in a larger socket. You think it shouldn't work? In modern offices, wired for Ethernet it's commonplace to re-purpose an 8P8C line for a phone or a fax and plug RJ11 directly into it. Whenever the RJ11 is wired for 1 (center) or 2 (center and one layer next) it works pretty well.

So the actual question is : why all 4 pairs are not wired center-out as a 4 phone lines would make most sense? If you look at RJ25, it carries 3 lines expectably, in a concentric fashion. Why Rj45 is strange?

The answer is that the original RJ45 was not in fact designed to carry 4 lines as the pincount suggests. It was designed to carry only 1 line for a modem and a little trick: a programming resistor. Value of this resistor would tell the connected modem roughly how long the line is so the modem would set the transmit power accordingly. But now, if they wired this resistor in place where a pair would be normally wired, then something bad could happen if a regular, phone multi-line plug got accidentally plugged there. So they picked pins 7&8, which do not constitute a pair under regular phone-wiring rules. So now even if you would plug a 6-pin RJ25, it would connect the central pair to the only pair, and one pin of the outmost pair to one pin of the resistor. Phew, at least no wrong circuit is completed this way.

And now Ethernet comes to the stage where most of the seats are already taken. So what do Ethernet designers do? They take the only pairs that remain guaranteed unused. The original, 10MB, Ethernet needs only 2 pairs. So it uses the second center line, the one not used for modem, and the other outer line, pins 1&2, not taken by programming resistor.

Long, long time later, the RJ45 phone wiring can be pronounced dead without a doubt. So the remaining "phone" pairs are finally free for the Ethernet to take, as Gigabit Ethernet dully takes advantage of.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.