19

I know this question has already been asked a few times but I feel that my case is a little different since I've tried all the solutions provided in the other threads.

I created a small diagram to avoid any potential confusion.

My attempt to solve the kink in the cable, also a diagram how things are laid out currently

So I have a cat6 cable of 80 meters. This cable has worked in the past. However, 2 years later I want to use that cable again and now it only gives me 95 mbps(tested with iperf3), it also gets recognized as a 100mbps link instead of a 1000mbps link. Image 1 provides a visual representation of how everything is laid out at this point.

Parts:

My troubleshooting steps:

  • Redo the connector, router side (changed nothing)
  • Redo the lsa wallplug, pc side (changed nothing)
  • Test the cable with a cable tester. (all wires are working, but still only 100mbps)
  • Check the entire cable for kinks (Found two kinks and cut that piece out. Connected the now 2 cables by adding an rj45 to the ends and connecting them with an ethernet extender, see the after part in the image) (solved nothing)
  • Test the cable again with cable tester (all wires are working, still 100 mbps)
  • Test the cable with iperf from the router to the extender (950 mbps, recognized as 1000mbps link)
  • Test the cable with iperf from the extender to the PC (95 mbps, recognized as 100mbps link)
  • (The previous 2 test results do not change if I do them with or without the extender included.)
  • Redo the connector for the cable from the extender to the pc (nothing changes)
  • Redo the lsa plug for the cable from the extender to the PC (nothing changes).
  • Test again with the cable tester (all wires are working)
  • Nothing changes no matter what I try.

Things I know for sure:

  • The rj45 connectors support solid core cable.
  • The lsa plug supports cat6.
  • I use the T568B standard everywhere.
  • The type of cable I'm using is S/FTP cat6.
  • The cable does not run anywhere near any other cables(except where it connects to the router).
  • Short distance of about 15 meters works just as it's supposed to.
  • Longer distance of about 65 meters gets recognized as a 100 mbps link
  • The entire cable runs inside a conduit.

What else could it be?

The only thing I can think of right now is that the signal isn't strong enough by the time It reaches the pc. So I am still planning putting a switch in the middle of the 65 meter part of the cable. However I would rather not do this since this is outdoors and I don't have a good way of protecting the switch or the not interrupted cable from the elements.

If this also does not help, would upgrading to a cat6a help? I read somewhere that cat6 Is not that reliable after about 50 meters.

PS: If you find any spelling/grammar mistakes, please let me know. I am still learning English ;)

UPDATE I fixed the issue by adding a gigabit switch in the middle. Now I have a 1000mbps link with a PC to PC transfer speed of +-870 mbps. Update

However, there are still a few very interesting answers which you should definitely read if you stumble upon this post in search of a solution to your ethernet problems. Especially the answers from @Zac67 and @The Unix Janitor

  • 20
    Why are you asking the same question again and deleting the old one? Please don't do that. If you have new information you should edit it into the original question instead of asking a new one. – DavidPostill Jun 28 at 10:51
  • 6
    "Connected the now 2 cables by adding an rj45 to the ends and connecting them with an ethernet extender," That is not allowed for UTP cabling. The standard explicitly forbids splices, taps, couplers, etc.. Cables that get kinked (exceeding the minimum bend radius) or exceeding the maximum pulling tension are permanently damaged and should be discarded. The solid-core cable that can b used at such distances is relatively fragile. Even experienced installer have trouble getting Category-6 cabling to pass the test suite; something as simple as the wrong wire in a pair on top can fail the tests. – Ron Maupin Jun 29 at 1:15
  • 2
    See this answer for more information on the required tests. – Ron Maupin Jun 29 at 1:17
  • 3
    @DevShot, an active device. Another thing I noticed is that you are using shielded cable, and all category cables are UTP, except the new Category-8. In any case, with a shielded cable, the shield must be continuous from end-to-end, with all parts, e.g. connectors and terminations also shielded, and it must be grounded on at least both ends. If it is not then the performance will never meet the "category" specifications, and the shield can cause interference. See this answer. – Ron Maupin Jun 29 at 18:35
  • 2
    Modern network cabling is not for amateurs. Back when we had Category-3 cabling, it was pretty easy to DIY, but that is not the case today. It takes a lot more than simple electrical connectivity for modern network speeds and cabling. – Ron Maupin Jun 29 at 18:37
28

Ethernet auto-negotiates using (low-frequency) link pulses. It does not check or train the cable. Link pulses are exchanged on the two 10/100 pairs and for 1 Gbit/s, all four pairs do an additional link check each.

Some Ethernet devices may fall back to 100 Mbit/s when 1000 Mbit/s has been negotiated but fails to link due to one or two missing pairs - 1 Gbit/s requires all four pairs, 10/100 Mbit/s only use two pairs. This is not by standard though and usually limited to Broadcom chips.

So possibly, one of the pairs in the cable has failed. Another possibility is that you've mispaired some contacts when terminating the cable. Cheap cable testers cannot detected split pairs as they only check for continuity and shorts. Double check the color codes inside the connectors - you can (but shouldn't) swap entire pairs or the wires within a pair but you must not swap wires across pairs.

You should note that while Ethernet over twisted pair is generally limited to 100 m - 90 m of those are required to use solid wire cable and only 10 m are allowed to run over stranded (patch) cable. Anything beyond that may throw (more or less) errors or fail at any time.

Also, DIY cables for Gigabit Ethernet can very easily exceed the specification margins and cause transmission errors. You need to make sure that you obey all twisting and pairing rules and generally do a neat job. Connectors must fit the cable (stranded or solid-core) and crimping tools must match the connectors.

| improve this answer | |
  • 6
    Cheap cable testers cannot detect split pairs as they only check for continuity and shorts This is also probably why my tester couldn't identify any problems. Thanks for that information, I had no clue that there were any big differences in testers for ethernet cables. – DevShot Jun 29 at 17:36
  • 4
    @DevShot Yes, testers range from $5 to $2000+, so there are quite a few differences. Measuring crosstalk (FEXT, NEXT, ...) for Cat 6A is quite challenging. – Zac67 Jun 29 at 19:01
  • this an extremely detailed explanation of these fast link pluses. (FLP) dell.com/content/topics/global.aspx/power/en/ps1q01_hernan?c=us – The Unix Janitor Jun 29 at 20:34
  • @TheUnixJanitor That link doesn't seem to work any more. But you can always check the original spec in IEEE 802.3. – Zac67 Jun 29 at 20:53
  • Depending on your country, data and phone runs between buildings may require the lines to have lightning arrestors fitted to be compliant. Not that anyone really audits this sort of thing though. – toxicantidote Jun 29 at 23:06
14

You have a lot of variables here:

  • To get 1Gbit ethernet you need all 4 pairs of the cable that have no faults and are OFC [Oxygen-Free Copper]; any extenders, joins, or couplers are going degraded the signal (to get 100Mbit, you only need 2 pairs working well).

A better solution may be to use fiber and a fiber media converter at each end, as it negates the below and is future-proofed (i.e. 10Gbit), while also avoiding all the below since it's non-electrical, more resistant to water damage, easier to work with, and won't be affected by direct or nearby power surges, with the armored types being very hardy.


Regarding ethernet:

  • Do you have a cable tester so you can verify all 4 pairs match at each end and are not being crossed or disconnected?
    • The pairs needed for 1Gbit could be slightly damaged, so test the electrical resistance of each pair or run a known-good cable to rule this out.
    • Do the extenders actually extend all 4 pairs (some cheap ones only do 2)?
  • Not all network cards are created equal:
    I've had some network cards/devices be able to go 200m, and upon changing the card out, only 50m; whereas others have better noise filtering and power output, so use a high-quality network card if going for a longer range.
    • Cheap routers can have the worst ethernet ports on them for range and power, as they are designed for the home (20m or less).
    • Try setting each endpoint to 1000Mbit non-auto, as sometimes auto-negotiation to 1000Mbit/s fails due to firmware/software problems, or if there are many errors, it falls back to a slower speed.

UPDATE: (Autonegotiation is a requirement for using 1000BASE-T according to Section 28D.5 Extensions required for Clause40 (1000BASE-T). At least the clock source has to be negotiated, as one endpoint must be master and the other endpoint must be slave." Failed autonegotiation means something is broken )

  • You can never rule out interference, even though it's not near anything, as EMF can come from anywhere in the environment.
    • Is the cabled shielded (it should be at that length) and is the shield grounded?
  • Every time you terminate the cable or use an extender, breaking and rejoining the cooper, you are adding resistance and attenuation, which is something you're trying to avoid


If this cable runs outside, there are many reasons why this isn't recommended:

  • Weather, corrosion, and environmental damage (rodents, wind, electrical storms, etc.)
  • EMF from such things as underground powerlines
  • If both endpoints are on different power supplies and have different ground paths, they have different potentials, which can cause ground loops, creating interference, and this voltage difference can build-up, damage network equipment, and, in the worse cases, even shock you.
| improve this answer | |
  • 1
    OP says in the text that it is a SFTP cable, and the two links are to a shielded plug and a shielded wall-socket/jack. Nowhere does it say the shielding was tested successfully though - of my two cheap testers, one does "pin 9" and the other doesn't. – Criggie Jun 29 at 1:51
  • 4
    Try setting each endpoint to 1000Mbit non-auto Disabling autonegotiation violates the gigabit ethernet standard IEEE 802.3ab: "Autonegotiation is a requirement for using 1000BASE-T according to Section 28D.5 Extensions required for Clause40 (1000BASE-T). At least the clock source has to be negotiated, as one endpoint must be master and the other endpoint must be slave." Failed autonegotiation means something is broken. You don't fix a flat tire on your car by flooring the gas pedal and trying to drive faster. – Andrew Henle Jun 29 at 15:29
  • @The Unix Janitor, Regarding the ground loops, is it correct that only 1 side of the cable should be grounded? I have no way of having the two endpoints on the same power supply. – DevShot Jun 29 at 17:39
  • @Criggie My tester also tests for G (ground) and I can verify that the ground portion of the cable conducts, but currently I have no ground connected because I am worried about connecting it wrong (ground loops etc). – DevShot Jun 29 at 17:42
  • @AndrewHenle thanks for the correction!!! I'll add it to my answer. – The Unix Janitor Jun 30 at 15:58
7

This cable has been outside in the weather. Unless you live in the Atacama Desert, rain water has had a chance to penetrate any slight break in the jacket, such as from a kink, and you could now have water at random locations inside the jacket surrounding the insulated conductors. The water has a higher capacitance than the air that is normally inside the jacket, and this can round-off your square waves and smear your signals enough that the cable won't work at 1000 Mbps.

Do you have access to a vacuum chamber? If you could roll up the cable, stick it in the chamber and maintain a vacuum of about 1 torr or less for a couple of hours, any water inside the jacket should evaporate. If you note a prolonged pause in the pressure drop at 4-5 torr, this is indicative of water leaving the cable, and you should continue pumping down until the pressure remains at or below 1 torr.

If the cable now works at 1000 Mbps for its full length, you should be good until the next time it rains.

| improve this answer | |
  • 2
    "rain water has had a chance to penetrate any slight break in the jacket" Actually, indoor cable does not have a waterproof jacket, despite appearances. Water will migrate into the jacket and change the dielectric properties of the cable. – Ron Maupin Jun 29 at 11:56
  • 1
    If access to the cable is available, the best option would be to roll it up and toss it in the trash. CAT6 wire is not that expensive. – spuck Jun 29 at 17:09
  • 2
    This cable has been outside in the weather. I should have been more clear about this, the cable is inside an electric pipe (plastic pipe, don't know the correct English word for it). This pipe was clean and dry on the inside when I checked it, so I don't think that any water has had the change to come in contact with the cable. edit: I just found out that conduit is the correct term. – DevShot Jun 29 at 17:46
4

Your question says you have an "80 meter" cable, but the evidence you provide suggests you have a 15 meter cable and a 65 meter cable.

If you know the 65 meter cable doesn't work at 1Gbps, adding couplers to the ends of it or attaching the 15 meter cable will not make it better.

As another answer states, if the 65 meter cable has been outside and is not rated for outdoor exposure the jacket may be cracked and been contaminated by moisture.

You should replace the known-broken 65m cable with new wire rated for outdoor use. You may also want to consider replacing this run with optical fiber, especially if the connection is between buildings.

| improve this answer | |
2

Adding a switch in the middle to boost the signal worked for me. However, I am still only getting +-870 mbps when testing with iperf3 between 2 PC's but that's more than good enough for me.

This is the current layout of everything: update

| improve this answer | |
  • Does your switch or router or PC have interface metrics? I bet there are port errors going on somewhere. ifconfig in linux, netstat -s | more in windows command prompt, or your switch might have a web page or CLI showing errors. – Criggie Jun 29 at 21:46
  • @Criggie I just ran ifconfig on the pc side of the diagram. This was my output: imgur.com/a/2BjvkmZ Is this what you meant? The switch does not have a webpage or anything else I think, it's just a 30 euro unmanaged switch ;) – DevShot Jun 29 at 22:49
1

You say the plugs support solid-core cable, which implies you've used solid core cable.

Is there anywhere in the cable path that is unsupported? I've found that any flex at all allows solid core cable to fail internally.

Sometimes people use solid core because its marginally cheaper, and is a lot easier to put into plugs before crimping. This is false economy, and the only time you should use solid core is when the cable is inside a wall/ceiling and will never move.

Solid core wire that is cable-tied to a fence will move as wind pushes the fence about.

Your best option is to replace the one run of solid core cat6 with TWO runs of stranded core cat6. Why two? One for now and one for spare - its only very little more work to run two cables over one cable.

If the run is exposed to weather, you should take the time to run it all inside conduit. It is possible to pull two cat6 down a common garden hose provided you use a draw wire, but three is impossible for more than a few metres. Or you can purchase cat6 externally-rated cable, which has an additional casing and a grease fill. It is horridly messy stuff to terminate though.


Another option is to use fibre, but the prices generally rule this out.

| improve this answer | |
  • At the distances in the question, stranded cable will not work. You cannot get very far with stranded cabling and be able to pass the category test suite. – Ron Maupin Jun 29 at 11:58
  • @RonMaupin possibly. The other option is to protect the solid cable better, with rigid waterproof conduit. Depends if OP ever comes back to fill in the missing information, like why there's no way to replace the whole run of cable. Perhaps OP thinks cable is expensive, they're shopping retail prices or....? – Criggie Jun 29 at 12:09
  • 1
    @RonMaupin The price is exactly what is holding me back, I don't want to replace it if I don't really have to, also I hate waste ;) – DevShot Jun 29 at 17:51
  • 2
    @DevShot, 100 meters of cable should not cost more than your 30 euro switch. I appreciate your desire to not be wasteful, but there is a saying in English: “throwing good money after bad”; there comes a time to not spend any more on something broken, and instead to use that money towards its replacement. – spuck Jun 30 at 4:49
0

RJ45 connector itself can impact the link. Be careful that there is some for single-stranded, multi-stranded copper pairs.

There is rj45 for cat5e cat6 cat7 ....

| improve this answer | |
-2

Don't use crimped ends. Period. Won't certify (yes, there are exceptions to every rule, depending on length, and actual RF/ Radiation interference) CAT6/ CAT6a. Use keystone jack's, and patch cables to your desired/ required bandwidth.

| improve this answer | |
New contributor
D M is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
  • Cat6 levels can be reached and qualified with the appropriate (expensive) scanner. Punch down jacks and crimped plugs can and should be in-spec if done properly. – Criggie Jun 30 at 7:08

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.