15

I've got in-wall CAT5 wiring in my house. The wall sockets were done by an electrician and the requirement was straight-through T-568B, but I think he got it wrong. When I use a cheap cable tester, I see that one end obviously sends this sequence:
1-2-3-4-5-6-7-8 but the remote unit incorrectly reports with 3&6 swapped:
1-2-6-4-5-3-7-8.

Given that not all wires are used, does this matter? I think it does; my understanding is that wires 1,2,3,6 are used but I'm no networking expert.

Lucky for me, the in-wall cabling ends in a patch panel so perhaps I should just open that up and swap all the 3&6 wires around -- but I don't want to do this unless I'm reasonably sure that'll help.

Also, would this wire problem cause the LAN to work normally at 100Mbit speed but not at gigabit speed, or is that unrelated? (I've asked about that already but just thought it might matter here.)

5
  • 1
    I'm also not an expert but as 100Mbps uses only 4 wires 1Gbps really needs all 8 (all 4 pairs). 100Mbps does use 1-2-3-6 so i'm puzzled as to why you're even getting 100Mbps. If you're sure they are swapped i would test by swapping just one back for testing. The T568B termination states the wires should be "straight through" ( i.e., pins 1 through 8 on one end are connected to pins 1 through 8 on the other end).
    – Rik
    Nov 24, 2013 at 13:59
  • 2
    As to having wires in a pair swapped, some ethernet transceivers can handle this and others can't. It would be far better if the wires were done correctly, even if it does work -- you can't guarantee that all equipment will handle it. Nov 26, 2013 at 22:55
  • 2
    Some adapter can handle swapping transmit and receive signals. But 1&2 are transmit (positive and negative) and 3&6 are receive (pos&neg) and here the positive and negative are swapped. (don't know if that makes a difference). Just re-punch one of the connectors on the patch-panel unless the wall-mount is easier to change and see if you can get 1Gbps. According to the pictures in your other question the patch-panel is correctly color-coded so you can check the wall-mount if the wires there are switched. (Otherwise note the discrepancy in the patch-panel in the documentation of your install)
    – Rik
    Nov 27, 2013 at 0:05
  • @Rik: You should put that as a real answer; it needs upvoting :-) Nov 27, 2013 at 9:47
  • IIRC, for some color coding schemes the color coding for the 3-6 pair is "backwards" in that the striped wire is the first of the pair vs the second (or vice-versa). So it would be easy to get this screwed up if you worked from memory (and your memory is as imperfect as mine). Nov 27, 2013 at 12:43

6 Answers 6

11

First and foremost it is absolutely critical that both ends match (at least with respect to wires that are actually used). If a given color originates on pin 3 on one end that same color must be tied to pin 3 on the other end. (This is contrary to phone cables where opposite ends are mirror images.) If you use a coupler to join two cables the coupler has a built-in "twist" so that this all works out.

Second (not quite as critical at lower speeds) the pairs must be kept together. That is, there is a "send" pair and a "receive" pair, and the wires of the pairs are twisted together. If you get one wire of the "send" pair twisted with one wire of the "receive" pair then you can get "crosstalk" which will introduce errors and slow things down (if not halt things entirely).

The pairs are not assigned to pins in an entirely logical sequence. The pairs are 1-2, 3-6, 4-5, 7-8. That is, three of pairs are adjacent pins, but one pair straddles the adjacent pins of the center pair.

Otherwise, electrons are color blind, so it doesn't matter which colors you tie to which pins, so long as the pairing is maintained. (Some claim that there is an ordering/positioning of the pairs in the cable and using the wrong pair of pairs can cause problems, but this argument is tenuous at best.)

However, it is a good idea to stick to a standard color assignment scheme, to make it easier to get cables terminated the same way on both ends. Unfortunately, there are 2-3 "standard" schemes.

4
  • Color schemes - yes, that's why I've insisted on T-568B throughout; not because I care about which particular standard to use but because I care about having one consistent installation. (I love your statement "electrons are color blind"!) I'll see if I can open a wall socket box - last time I tried, I couldn't figure out how. Nov 27, 2013 at 9:51
  • You could show us a picture of the wall-mount (together with make and model if known). Maybe somebody here knows how to open it easily.
    – Rik
    Nov 27, 2013 at 10:32
  • I'm accepting this answer because it turns out that the basic problem was the wire sequence (solution here), and that's what you are focused on. I still can't explain why it worked at 100Mbit despite wrong cabling though. Nov 27, 2013 at 21:46
  • 1
    @TorbenGundtofte-Bruun - Yep, some punch-down blocks have the pairs paired, vs being in numeric order. It makes cable termination easier (and microscopically less noisy), but can confuse the heck out of you if you don't notice it. Nov 28, 2013 at 4:04
10

I do tech support and the argument of TIA-568A vs TIA-568B vs 'whatever, but straight through" comes up every now and again. You should use the appropriate standard for your application.

Since you're referring to Ethernet, you should use Cat5e or Cat6 cable and terminate with the TIA-568B pinout.

TIA-568A

This standard is acceptable/meant for voice communications. By voice, I mean telephony. You would probably see this on a CAT3 cable, and telecom engineers and electricians are probably familiar with it. Any analog voice systems would probably be using this standard, even if they were using a higher CAT cable like CAT5.

This pinout should not be used for data networks, though it will function to some degree; you may see errors, reduced data rate, or other less-than-desirable results.

TIA-568B

This standard should be used for data networks, packetized video, or other intensive applications (throughput or bandwidth).

I don't think there is a negative consequence to using this standard for voice, but it goes against convention, so I would not use this standard for voice communications.

Random straight through

Sometimes, someone will tell you that as long as the ends match, you don't even need a standard. That may suffice for electrical conductivity, for example if you are using some CAT cable to control a relay, but for communications (tele- or data-), you should use the appropriate, relevant, standard.

Misc notes

  • The actual wires in CAT cables have different lengths and twists per turn (see Wikipedia), so the choice in pinout does matter and has real world consequences.
  • 10/100 mbit Ethernet uses 2 of the 4 pairs, while 1000 mbit Ethernet (Gigabit) does use all 4 pairs of the CAT cable.
  • Wikipedia may say that -A or -B is fine, but from real world experience, use -B for data networks.
  • Telecom engineers, electricians, or other analog-based technicians will probably give you the most static about how the pinout doesn't matter or that "it's all the same anyways". In such a case, they do not understand the complexities of the data communications world. They may not be wrong for the tech they are used to, but data communications are different than what they are used to working with.
  • TIS-568 Standards
  • CAT5 Cable
4
  • 2
    Thank you for giving extra background on why TIA-568B is the correct standard for data. Luckily, it's what I chose :-) Feb 8, 2016 at 22:23
  • This is the correct answer. "Its all the same anyways" will result in slower data transfer speeds.
    – Yllier123
    May 31, 2021 at 21:23
  • This answer sounds like nonsense. But rather than vote down, I will comment with hopes of a good explanation in a comment or improved answer. I've read electrons moving through copper aren't affected by the color of the insulation jacket being used. So why would T568A and T568B differ, as long as consisent on both sides? Answer says "actual wires in CAT cables have different lengths and twists per turn" but I think that refers to Cat5 has more twists per meter than Cat3, not differences in individual cables (e.g. orange has more twists than green/etc.)
    – TOOGAM
    Jul 17, 2023 at 9:52
  • The twists/meter in a cable's length is going to be unrelated to the RJ45 connector, so I don't see how a T568A wiring in the RJ45 vs. a T568B wiring in the RJ45 connector will be significantly impacted by the number of twists/meter of the whole cable length. I wonder if this is caused by a mix-up regarding mixing up TIA/EIA-568B.2 (a standard affecting the cable length) with T568B (a standard affecting the end connector layout)? (In particular, I wonder this after having read superuser.com/a/730201/401839 )
    – TOOGAM
    Jul 17, 2023 at 9:54
4

This is an old question, but since none of the answers are truly correct I will try to answer more clearly.

TL:DR same color pairs must be kept together. The same color wire must be on pins 1&2, 3&6, 4&5 and 7&8. Which color is on which doesn't matter as long as it is the same on both ends.

Reversing the two conductors in a pair does not matter. Ethernet signaling does not care about polarity. Whether white/[COLOR] is connected to pin 4 or 5 doesn't matter as long as [COLOR]/white is connected to the other.

Which color pair is connected to each pair of connector pins does not matter. This is why TIA-568A vs TIA-568B wired sites will both work with data. (A is designed to more closely match telephone colors where pair 1 is blue, pair 2 is orange.)

(The wires do have different number of twists per inch, but this is to minimize the crosstalk between pairs, it does not greatly effect the performance of each pair.)

10/100 used to have one pair as TX, and the other as RX (the other two are unused). Pure 10/100 devices require that pairs 1,2 and 3,6 match on both ends when connecting a switch to a PC (a crossover cable was required if connecting two same devices together). Later HP released the "Auto-MDIX" standard that allows the devices on each end to negotiate which pair will be used for which TX and RX. This eliminated the need for crossover cables when connecting two switches, or connecting two PCs.

When 1GB was released the Auto-MDIX standard was included, and expanded. Now each pair can be used for TX or RX, and which pair is pair 1, pair 2, etc is negotiated. 1GB should work as long as one pair is connected to 1&2, another pair to 3&6, pair three to 4&5 and the forth pair to 7&8. (Negotiating gets tricky and won't work for 10/100 devices if the 10/100 pairs 1&2 or 3&6 gets crossed with the additional 4&5 or 7&8 pairs. If there is 1GB equipment on both ends of the wire even crossing pairs will usually work. )

1

I'm also not an expert but as 100Mbps uses only 4 wires, 1Gbps really needs all 8 (all 4 pairs). 100Mbps does use 1-2-3-6 (like you said) so i'm puzzled as to why you're even getting 100Mbps (with 3&6 swapped).

The T568B termination states the wires should be "straight through" ( i.e., pins 1 through 8 on one end are connected to pins 1 through 8 on the other end) so at least swapping 3&6 would not be "up to the standard".

I know some adapter can handle swapping transmit and receive signals. But 1&2 are transmit (positive and negative) and 3&6 are receive (positive and negative) and if your cable tester is correct your positive and negative (of the receive signals) are swapped. (I'm not exactly sure if that can cause trouble).

As it is (according to your other question and photos) your patch-panel is correctly color-coded so the preferred method would be to check the wall-mounts to see if 3&6 are swapped (according to color).

If 3&6 are really swapped and the wall-mounts are hard to access or difficult to reconnect, you could re-punch just one connector on the patch-panel (for testing) and see if it makes a difference in getting 1Gbps. If you then get 1Gbps you have two options:

  • You could re-punch all the others to get the 1-8 "straight through". In that case your installation would not be standard to color-coding so you would have to make a note of it in the documentation of your install.

  • If you really want "the standard" (although it's not strictly necessary if correctly documented) you could then take the trouble to disassemble the wall-mounts and swap 3&6 there. (swapping that 1 connector for testing on the patch-panel back of course ;)

If the wires in the wall-mounts are correctly color-coded then there is something wrong with the cable-tester. In that case we need to look for other causes.

6
  • Good point about the cable tester! I just verified its correct operation using a single short LAN cable. Nov 27, 2013 at 10:40
  • Maybe I'm getting a 100Mbps signal at all because the in-wall cables don't connect directly to a dumb DSL modem but to a really good switch that can handle the inversed signal... In any case, I'm going to work on this tonight! Nov 27, 2013 at 14:38
  • Because you have a good switch you should be able to get a 1Gbps connection. I did notice the yellow patch-cable from switch to the router. It doesn't have a metallic jacket so my guess it's a cat5 patch-cable. That shouldn't matter because your A1 is 100Mbps max so a 1Gbps connection to the router would not be needed but if you a cat6 patch-cable extra, i would use it). Further... i take it the switch is managed, so you should also check to see if the ports on it (via the web-interface of the switch) are on auto (and not limited at 100Mbps).
    – Rik
    Nov 27, 2013 at 14:49
  • I have no idea what a "managed" switch is :-) Maybe we can meet in chat some time and you can explain it? Not now, though. Thank you! Nov 27, 2013 at 15:00
  • You can read here about managed and unmanaged switches. In short... an unmanaged switch has no configuration settings and is basically plug and play (less expensive). A managed switch has lots of configuration settings from virtual Lans to bandwidth control. It can also set each port 'hard' to i.e. 100Mbps half or full duplex etc. You could login a managed switch via the webbrowser. (In the manual it should say how). (1/2)
    – Rik
    Nov 27, 2013 at 15:28
0

Sounds like they got the crossover idea wrong. Wikipedia has a good article on what was attempted (I guessing) http://en.wikipedia.org/wiki/Ethernet_crossover_cable#Crossover_cable_pinouts . Wire 3 has a positive charge and 6 is a negative charge. You run the risk of shorting out a device that you hook this into. (either the switch or the network card.)

Even small voltage spikes can tank delicate circuits. Fix it or it will add other expenses to your list.

1
  • I think you're on to something as far as they got something wrong, but I don't see that any crossover was attempted because its' just one pair that is swapped, and not two pairs that switch places. Also, in-wall LAN cabling should simply be straight-through and not crossover. Nov 27, 2013 at 9:50
-4

I would like to offer some information that might help. If you are truly running Cat5 and not Cat5e then that would be one reason you are only getting 100Mbps. Cat 5 cabling was only designed to handle 100 meg throughput. There are a few exceptions to that rule, for example if you have a a gigabit switch and a gigabit NIC on your PC, with a short cable run, say for example 25 feet or less, then you might be able to get gigabit speeds, however that is really pushing the limits of the cable.

With that in mind if you want to have stable network connectivity you need to have your cables runs terminated properly. I have seen all kinds or weird things happen on networks that I was brought in to repair because the network runs were terminated incorrectly I will pace a link to one page that can give you some info of the cable types and speeds Network cable speeds, More Cable information, This chart might help understand the relation of network cable speed to the length

All in all, if you want avoid problems with your network you should make sure your cable runs are mapped correctly and in the end you will be happy you did, when you avoid problems like dropped connection or corrupted data as it was transferred across the network or even a problem like you can't see a specific network device even though it is on the network and should be visible.

Also if you are looking at your computers and they are reporting you are getting 100Mbps, this could be a condition of the network card reporting the speed it believes it is connected to the network based on network cable and the switch it is connected to. But what you don't see by looking at that, is whether you are truly even getting the transfer rate of 100Mbps, it might be less as a result of the bad wiring. If you want gigabit speeds you need to make sure the whole network can handle that, meaning the cable, the network switches, the network cards on the network devices and it even helps to make sure the places the cable termination occurs is also capable of those speeds.

I hope this helps you out Cheers

2
  • 3
    This spreads a pervasive false myth that Cat5 is not sufficient for 1000BASE-T. 1000BASE-T was designed for 100 meters of plain old Cat 5.
    – Spiff
    Nov 27, 2013 at 0:52
  • And because references are better than votes here's a link from cisco where it states "1000BASE-T will run on Category 5 and Category 5e cabling at distances to 100 meters." cisco.com/en/US/tech/tk389/tk214/…
    – ndemou
    Jul 26, 2016 at 15:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .