3

Why does this standard exist? Why can't I just make the same wires go to the same pins on the RJ45 on each end? Why does the order matter?

Inquiring minds want to know! Thanks

  • 4
    Are you aware that the eight wires in the cable are paired (into four twisted-wire pairs)? IOW the cable is not eight individual wires. – sawdust Oct 18 at 23:39
  • I once made a network cable and just made sure both ends were the same. The cable was supposed to transfer 1Gbps, but I did not understand why the connection kept dropping from 1Gbps to 100Mbps until my colleague finally found my error. After redoing the plugs, the 1Gbps was constant. Yes, it was the good old interference problem as suggested by the many answers. – LPChip Oct 19 at 11:57
  • So you're not asking why there is an A and a B standard that only vary by colour used for each position? You're asking why the standard exists at all? – Criggie Oct 20 at 3:06
11

If you just make the same wires go to the same pins on each end, interference will be significant and the connection will not work reliably at high speeds. The reason we have twisted pairs is so that interference will cancel out.

We use differential signal pairs specifically to minimize interference. When we raise the voltage on one side of the signal pair, we lower the voltage on the other side. The receiver measures the difference between the two voltages. This minimizes interference because interference will tend to raise or lower both sides equally and thus will be cancelled out by the receiver.

But for this to work, both sides must receive the interference equally. If you put one half of the signal pair on a different physical twisted pair from the other, then they will receive the interference differently and so the interference will not cancel out.

It's even worse than that. If you don't match pins to pairs, your signals will interfere with each other. Each twisted pair has a different twist rate (from as little as 50 turns per meter to as much as 80) to prevent signals from interfering with each other, but if you map them wrong, you can actually maximize interference.

For example, suppose the positive side of one signal is on the same pair as the positive side of another signal and the same is true of the negative sides of those two signals. Now, neither side can cancel out interference and they're tightly coupled to each other. Ouch.

  • but why couldn't they always use crossover cables like most other cabling standards? – phuclv Oct 19 at 0:34
  • 1
    @phuclv It greatly reduces the chance of technician error if they always terminate every end the same way. This is especially true in cases like in-building wiring where you can't even look at the other end of the cable you're terminating, because it's in a locked closet almost 100m away, and someone else terminated it. – Spiff Oct 19 at 1:52
1

Why can't I just make the same wires go to the same pins on the RJ45 on each end?

Why does the order matter?

Bear in mind that the pairs used by 10Base-T (10 Mbps) and 100Base-TX (100 Mbps) Ethernet are:

  • 1 + 2
  • 3 + 6 (Not 3 + 4)

1000Base-T (1 Gbps) requires all four pairs:

  • 1 + 2
  • 3 + 6
  • 4 + 5
  • 7 + 8

Hypothetical scenario where this would matter:

A damaged 1 Gbps Cat 6 drop stops functioning and needs to be temporarily rewired to support 100 Mbps while a new drop is pulled. Your continuity tester shows that some wires have been severed but at least 4 wires still have good continuity.

If you wanted to temporarily rewire the drop to restore connectivity, albeit at a slower 100 Mbps instead of 1 Gbps rate while a new drop was pulled, then crimping the working wires in the damaged cable into pins 1, 2, 3 and 4 would still result in a completely non-functional cable.

However, if you took the four remaining working wires and crimped them into positions 1, 2, 3 and 6, your damaged drop would then become a valid 100Base-TX cable capable of supporting 100 Mbps and could provide a slower but functional drop in the interim.

The cable shown below supports 100 Mbps data transfer rates, even though only pins 1, 2, 3 and 6 are connected; however, if pin 3 instead terminated to pin 4 at both ends, the cable would no longer work and would be reported as "disconnected" by the NICs at either end.

You can see in the picture that the orange (1 + 2) and green (3 + 6) pairs are wired the same as they would be for a regular T-568B termination, so no crosstalk issues are introduced from split pairs:

enter image description here

This is a potential scenario where being aware of the pins used in T-568A/B could make the difference between a 100 Mbps connection and a completely non-functional cable.

Interference:

The other aspect to this question is the more obvious issue of interference. The pairs are twisted in order to cancel out interference.

Although it is true that ignoring T-568A or T-568B wiring standards and simply wiring from pins 1 -> 8 on both ends would exacerbate crosstalk problems, the impact may not be as pronounced as you might expect.

If you simply wire the same pins from 1 -> 8 on both ends then, electrically speaking, pins 1 + 2 and 7 + 8 would still be in the correct positions.

enter image description here

As shown with the red highlighted section above, a cable that did not confirm to proper Ethernet wiring standards in this way would only see new crosstalk problems on its middle four pins: 3, 4, 5 and 6. Half the cable, pins 1 + 2 and 7 + 8, would still be electrically identical.

I'm not suggesting for a moment that it's okay to ignore wiring standards when crimping cables but it's interesting to note that the issue of crosstalk will only be half as bad as you might expect in this situation.

Half the pins and pairs would still be electrically identical if you terminated both ends to 12345678 instead of 12364578.

  • Only half of the pairs would be in wrong position, but that is still enough to render the network inoperable / very flaky. After all there is no logic to fall over to the working pairs. It's "works or not", not "only half as bad". – jpa Oct 19 at 8:14
  • "If you simply wire the same pins from 1 -> 8 on both ends" -- you're assuming some sort of an order, right? That they would crimp both wires of a pair adjacent to each other? Because if you mix and match freely, or e.g. put all the colored wires first to 1-4, and the colored/whites to 5-8 in the same order, then you've pretty much split all the pairs, which doesn't sound very good. Same if you take just some four working wires and crimp them to 1,2,3, and 6 without regard for the pairing. – ilkkachu Oct 19 at 8:50
  • 1
    @jpa "... but that is still enough to render the network inoperable / very flaky." This is true with the 12345678 termination example used for the purpose of this question. However in realty, you can take a damaged drop which still has 2 working twisted pairs and use them on pins 1+2 and 3+6. As with the earlier example, this would give 100 Mbps, except it would also not have crosstalk issues as the appropriate pairs would still be used, regardless of the fact the colours didn't match T-568A/B and the other pairs were not used. – Mr Ethernet Oct 19 at 8:50
  • It's not that I've actually tested cables with split pairs, though, but I don't expect it to be such a good idea. Might be an interesting exercise. (or not, if it didn't work at all). – ilkkachu Oct 19 at 8:52
  • @MrEthernet, "However in realty, you can take a damaged drop which still has 2 working twisted pairs and use them on pins 1+2 and 3+6. ", yep, that's what I meant with the above. Your answer just talks about "4 wires". – ilkkachu Oct 19 at 8:53
1

There is a functional difference between the two cables. One is "straight" while the other is "crossover".

If you have two computers and no hubs, switches or other gear then a straight cable would mean that transmit pins on a sending device would connect to transmit pins on the receiver and likewise for the receive pins would be wired to receive pins.

The result is that no data can flow because neither transmitter is wired to a receiver.

In this case you need a crossover cable to achieve the necessary swapping:

enter image description here

With a crossover cable it wires transmit to receive and receive to transmit properly.

Hubs and switches on the other hand are dedicated receiving devices. Internally they are wired to swap transmit and receive and therefore need straight cables between the computer and the hub.

Why can't we always just use crossover cables? Well, here's the rub. You cannot daisy chain together crossover cables in even multiples, otherwise the two crossovers means you have a straight cable, you have to have odd multiples of cable (1, 3, 5, etc) to retain the "crossover".

This is also the reason why hubs and switches have their receive and transmit swapped; it's to simplify building wiring. By having one fixed crossover position in the hub then all the connecting cables can be straight and it doesn't matter how many joints or patch panels it goes through, it will always be straight.

But you still occasionally need crossover to connect two non-crossover devices like a computer.

This has largely gone away with computers, hubs and various devices supporting Auto MDI-X. Basically new devices detect what kind of cable and device they are connecting to and internally swap the transmit and receive wires over as necessary. This has largely removed the need for two types of cable, and generally straight cable would be preferred as a result, but there is still old hardware out there and you might need crossover occasionally.

  • "...but there is still old hardware out there and you might need crossover occasionally." True but only one NIC out of the two needs to support it in order for Auto MDI-X to work. All Gigabit NICs have Auto MDI-X support, so you could take the oldest computer you could find (i.e. no Auto MDI-X) and connect it to any computer with a Gigabit NIC and it still won't need a crossover cable. Crossover is only essential when you connect two ancient NICs directly together. – Mr Ethernet Oct 19 at 9:18
  • @MrEthernet it might be true now, but back when the cables were designed it absolutely was necessary to have two types, hence why we have them. The question asked why they exist, not why they are necessary in the modern world which, as my final paragraph implies, they are not. – Mokubai Oct 19 at 9:22
  • 1
    For sure. Just mentioned it because people often talk about Auto MDI-X when discussing crossover cables but the fact that Gigabit NICs always have it and only one NIC out of the two needs it are interesting related facts I like to throw in alongside the main point. – Mr Ethernet Oct 19 at 9:27
  • @MrEthernet fair point. Any side supporting Auto MDI-X does render the whole crossover vs straight question moot. – Mokubai Oct 19 at 9:38

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.