Ethernet cables do not work, despite continuity.

[amyk]

It should be noted that I cut the 100ft cable I bought into two 30(ish) foot sections and re-terminated it. This does make me curious to know if the cable worked right out of the box.

Probably yes, and you likely tried to use plugs for solid-core whereas pre-terminated cables are usually stranded (and as you found out, far too thin for regular plugs.)

[tooki]

Crimping your own Ethernet cables is a false economy. If you buy preterminated patch cords from Cable Leader or Bestlink Netware, you won't have problems.

You don’t crimp your own patch cables to save money, you do it to get custom lengths that aren’t available, and for situations where you can fit the cable — but not the connector — through a small hole, requiring you to terminate the cable after installation.

[tooki]

It should be noted that I cut the 100ft cable I bought into two 30(ish) foot sections and re-terminated it. This does make me curious to know if the cable worked right out of the box.

Probably yes, and you likely tried to use plugs for solid-core whereas pre-terminated cables are usually stranded (and as you found out, far too thin for regular plugs.)

Patch cords are indeed normally stranded, and since patch cords are the far more common use for RJ45 plugs, the default type of plug is for stranded, not solid. So I would say the “regular” RJ45 is the kind for stranded, not solid. Looking at digikey’s selection of RJ45 plugs, the overwhelming majority are either for stranded wire only, or for stranded and solid wire. Only a tiny percentage (under 4%) are for solid wire only. So the chances that OP has a solid-only plug is relatively small.

The standards require the wire to be 26–22 AWG, but some connectors nonetheless support 28AWG. What does seem to be slightly rarer is support for 22AWG.

[tooki]

1. No CCA for runs over 10 meters.

Don't go cheap on the cable. Buy a known brand, not cheap chinese crap on Amazon.

As many others have said I think this is where I failed. This is the cable I bought: https://www.amazon.com/dp/B01G0LUS90?ref=ppx_yo2ov_dt_b_fed_asin_title&th=1
Not sure why I chose this one, I think I specifically wanted round white cable, when most of the ones that popped up were flat. Either way it apparently was "cheap Chinese crap on Amazon", despite pretty good amazon reviews. When you scroll all the way down to the description it does actually say that it is copper clad wire but not what gauge it is…

FYI, the Cat 5 and higher standards require the cable to be 100% copper, so any CCA wire is by definition not compliant. And if they didn’t follow the materials requirements of the standard, what other aspects of the standard did they also ignore?

At least that Amazon listing is “honest” insofar as they admit to using CCA, though it’s then a lie to claim to be Cat 6 cable. So I doubt it actually meets Cat 6 performance requirements, and your experience with it being susceptible to interference supports this.


Note also that the super-thin flat Ethernet cables also aren’t compliant. They may claim to be Cat 6 or whatever, but they don’t adhere to the standard. They might be able to perform well enough (possibly even enough to pass tests), but nonetheless they’re not complaint to the standards. For example, this one from a reasonably reputable manufacturer is claimed to be Cat 6a, and being a reputable brand it probably meets the performance targets, but it’s also specified as 32AWG, far thinner than the standard allows.

[soldar]

I have terminated dozens of Ethernet cables. I have the crystal connectors and I have the habit of putting connectors on scraps of cable.

Before crimping I very carefully check the colors are correct. I double check and then I crimp. Then I just test the cable between a computer and a router. Half of the times it doesn't work. Then I check the colors and I detect a mistake which I did not detect before crimping. So I cut off the RJ45, install a new one correctly and now it works. I don't know how this happens to me so often but it does.

I have several crates of cables which I do not really need but I still have this graphic on my computer desktop ready for any new crimping.

[BradC]

Something like this is handy : https://www.noyafa.com/products/nf-308s-wire-fault-locator

It checks wire sequencing and gives you a fairly accurate estimate of the length of each pair. We had a rat attack a cable in the roof. It had a run length of > 40M. By checking pair distances from each end I could estimate the break within about 3 roof tiles. Great for a quick check of patch leads after crimping too.

I bought the standard model about 15 years ago. About 5 years ago I upgraded to the "fancy" one with 8 dongles. Just go around the place and plug them all into the wall, then run the tester down the patch bay to identify the sockets. A few friends have moved into houses with un-labeled cables or had connectivity issues. Something like this makes it easy, and is slightly more useful than the "running led" testers.

The warbler has come in handy too, mainly for tracing sprinkler solenoid cables, but also the odd figure-8/coax/whatever.

[J-R]

I've been crimping since the 80s and realistically you can get plenty far with just the $5 cable testers to tell you if you've goofed or not.  With UTP Ethernet, once you establish link, the quality of the cable can become evident if it's supposed to be gigabit but only establishes 100M.  I've also seen plenty of kinked UTP and cables run over dozens of light ballasts and it frequently surprises with how well it works.

I think a lot of this changed with the introduction of 10GBASE-T because it had steeper requirements.  You could do it with proper quality CAT5e over very short distances, but longer required higher and higher cable standards.  Shielded cable was also popular.  Installers were testing and certifying the cable to make sure it met the requirements for 10GBASE-T.  You had customers sometimes going years before discovering that they couldn't actually run 10GBASE-T.  This is because they may not have had 10GBASE-T to the desktop initially.  So the high-frequency certs were important.  Also, the next fallback rate would have been 1GBASE-T which would be pretty disappointing.

Things have changed again with 2.5/5GBASE-T because it can run over CAT5e and will auto-negotiate the link based on what the cabling/switch/device can actually do.  If you have a very large building with thousands of drops, you're not going to want to pull new cable unless you have to.  So you slap new switches and Ethernet cards in (although many computers in the last 5+ years have 2.5GBASE-T).  Whatever you get for the speed is what you get.  Personally in the building I work in, we've seen great results and so far are getting solid 10GBASE-T over CAT5e.  Users with 2.5GBASE-T NICs are also happy with the speed boost.

[soldar]

Getting one of those fancy testers is overkill if you are just a hobbyist doing just a few cables a year.  I agree that a $5 multimeter and a little time are enough to diagnose the most basic problems.

When I terminate a cable I just connect it to a computer and router and if it works I consider it good. If it doesn't work I double check the wires are connected to the correct pins and this is usually the problem.

Very rarely I find it is the cable itself which is at fault. To check the cable, for ease connecting to the RJ45 connectors, I have a couple female connectors with the wires bare where I can connect alligator clips. This way it is very easy to detect continuity and shorts.

Some months ago I detected a bad cable where one faulty wire had no continuity. I have plenty of cables and I did not need to repair this one but, just for fun, I cut it in half and put an RJ45 on the half that was good. Then I cut the other in half and, again, terminated the good half. I did this until the last bad piece was so short it was not worth continuing.


For more serious or frequent work it would be very easy, and basically free, to construct a board with two female RJ45 connectors and some LEDs confirming continuity and correct connections.


And now anecdote time. About 25 years ago I was in a cheap hotel in China and my Ethernet Internet connection was not working. I talked to the young woman at the front desk and she called some maintenance guy and they both came to my room and the guy plugged in some test device into the wall and there were some blinkenlights showing the cable was connected at the other end. The guy did not speak any English and the young woman was translating as best she could, which was not much. The man, who I think did not know much about anything, said the wires had been eaten by rats in the walls and ceiling and the receptionist offered to change me to another room but I said before we do that let us test the Internet connection in the other room ... and it wasn't working either.  The guy insisted there were many rats eating cables. I think he had a basic Ethernet tester but had no idea how to use it. I suggested to the receptionist they reset the router but the guy was adamant. Rats were everywhere. I think he really had no idea what the router was. I told the girl that it did not make a hotel look good to tell guests the place was full of rats.

After some further talk the guy left and after much more talk I convinced the young girl to find out where the router was located because she had no idea and had to ask around. Finally she accompanied me to a small service room in the attic and there we found a woman eating her noodles and, in a corner, the router. I reset it and magically all the rooms had Internet connection again. The cheap router had a habit of hanging because during my stay I had to repeat the operation several more times. The receptionist was grateful that she now could solve the problem so easily even if it meant going to the top of the building.

[Alex_Baker]

With regard to shorting the pins as a test, my point is that if you shorted pins 1 and 2 at one end, but at the other end they were inadvertently reversed (2 and 1 instead), then you'd have continuity, but the cable wouldn't work because it's wired incorrectly.  So it's not a valid test.  Testing each pin individually will be accurate.

Ahh I see, that makes sense, thanks for the clarification. Like most my cable ends are clear so I can see the wire colors to verify them, but you are correct, my test would not be valid in this case.

Note also that the super-thin flat Ethernet cables also aren’t compliant. They may claim to be Cat 6 or whatever, but they don’t adhere to the standard.

Funnily enough this is exactly what I bought in patch cable form to replace the crap that started this thread. In my case I don't expect issues since the cable is just running to a Raspberry Pi that is controlling a 3D printer. I also did not re-terminate any ends and in this instance the flatness makes it look nicer along baseboards and above door frames.

[tooki]

I have terminated dozens of Ethernet cables. I have the crystal connectors and I have the habit of putting connectors on scraps of cable.

Before crimping I very carefully check the colors are correct. I double check and then I crimp. Then I just test the cable between a computer and a router. Half of the times it doesn't work. Then I check the colors and I detect a mistake which I did not detect before crimping. So I cut off the RJ45, install a new one correctly and now it works. I don't know how this happens to me so often but it does.

I was gonna say… if that really is a frequent occurrence, then something is wrong with how you’re “carefully double check[ing]” the colors before crimping, and/or with how you’re moving the connector into the tool to crimp it. (Are wires pulling out and then being pushed back in, potentially in the wrong order, when you put it into the tool?)

I’ve crimped perhaps 20 cables over the years (and installed a bit more with punchdown jacks) and I’ve never had a failure, so this isn’t a “normal” problem.

[tooki]

I think a lot of this changed with the introduction of 10GBASE-T because it had steeper requirements.  You could do it with proper quality CAT5e over very short distances, but longer required higher and higher cable standards.  Shielded cable was also popular.  Installers were testing and certifying the cable to make sure it met the requirements for 10GBASE-T.  You had customers sometimes going years before discovering that they couldn't actually run 10GBASE-T.  This is because they may not have had 10GBASE-T to the desktop initially.  So the high-frequency certs were important.  Also, the next fallback rate would have been 1GBASE-T which would be pretty disappointing.

I recently upgraded my home internet from 600/60Mbps cable to 10Gbps fiber-optic (with WiFi 6 in the new modem/router/gateway), and this means I will eventually also need to upgrade the second access point on the other end of my apartment. (Several heavy masonry walls means WiFi can barely make it from the living room to bedroom.) I have it connected with an Ethernet “backhaul” to the bedroom, which I ran years ago using Cat 6a (maaaybe cat 7? I don’t remember). It’s worked perfectly for gigabit, but until I actually get a 10Gbps access point and Ethernet switch in the bedroom, I won’t actually be able to know whether it actually supports 10Gbps. It’s not a very long run, so I expect it’ll be fine, but still, it’d be great to know.

P.S. I once knew a network admin who had a pet cat named “five”. 🤣

[tooki]

Something like this is handy : https://www.noyafa.com/products/nf-308s-wire-fault-locator

It checks wire sequencing and gives you a fairly accurate estimate of the length of each pair. We had a rat attack a cable in the roof. It had a run length of > 40M. By checking pair distances from each end I could estimate the break within about 3 roof tiles. Great for a quick check of patch leads after crimping too.

I bought the standard model about 15 years ago. About 5 years ago I upgraded to the "fancy" one with 8 dongles. Just go around the place and plug them all into the wall, then run the tester down the patch bay to identify the sockets. A few friends have moved into houses with un-labeled cables or had connectivity issues. Something like this makes it easy, and is slightly more useful than the "running led" testers.

The warbler has come in handy too, mainly for tracing sprinkler solenoid cables, but also the odd figure-8/coax/whatever.

Years ago, I bought a WiFi router (d-link I think, but it may have been netgear) that actually had a cable break test built right in: right in the web interface, it displayed the distance to the end of a cable, or the break inside one. This means it had some form of rudimentary TDR (time-domain reflectometer) built in, presumably a feature of the Ethernet PHY.  I remember it being accurate to within a foot (~30cm) or so. Unfortunately, that router had other issues, so I returned it and got a Linksys (the venerable WRT54G, which I could put custom firmware onto).

If anyone here knows: Is it common for Ethernet PHY chips to have this TDR function, and most devices just don’t use it? Or was it an oddball?

[Monkeh]

If anyone here knows: Is it common for Ethernet PHY chips to have this TDR function, and most devices just don’t use it? Or was it an oddball?

It's very normal (with varying degrees of quality) in managed switch ICs (serious ones, anyway, not one which offer 16 VLANs and a port enable bit), but to my knowledge unusual in end devices - why duplicate the functionality on the other end of the link?

[2N3055]

ADI has some industrial Ethernet PHY that have cable diagnostics.

[madires]

I think a lot of this changed with the introduction of 10GBASE-T because it had steeper requirements.  You could do it with proper quality CAT5e over very short distances, but longer required higher and higher cable standards.

10GBASE-T via proper CAT 5e can work for a length of around 30 m (if lucky 50 m). Beyond that I'd go for glass fiber (OM4). With CAT 7/8 it's possible to run 25GBASE-T or 40GBASE-T for up to 30 m, but getting the installation of CAT 7/8 right isn't easy.

[Berni]

Yeah for short enough lengths of a few meters you can get Ethernet working even over speaker wire.

But at 10Gbit Ethernet it does get pretty sensitive. With crappy CAT5e you might start getting issues at 10 meters, or heck even with proper CAT6 and using crappy enough RJ45 wall sockets it might already start having issues.

For anything above 2.5Gbit it is a good idea to simply go fiber. It can do even 40Gbit speeds over larger distances than you would ever need. At the same time even at only 10Gbit the power usage of the transceivers starts getting ridiculous and it might take 5W of power per each 10Gbit link just to keep it connected over copper pair. Might not sound much but once you have 10 links going that's 50W of power being constantly burned to heat just to keep the Ethernet links up. Using a SFP+ fiber transceiver to do a 10Gbit fiber link has considerably lower power consumption and will go past 100m with ease (or even kilometers on single mode fiber), heck fiber cables are even cheaper per meter than good quality CAT6 since no expensive copper is needed.

[amyk]

If anyone here knows: Is it common for Ethernet PHY chips to have this TDR function, and most devices just don’t use it? Or was it an oddball?

It's very normal (with varying degrees of quality) in managed switch ICs (serious ones, anyway, not one which offer 16 VLANs and a port enable bit), but to my knowledge unusual in end devices - why duplicate the functionality on the other end of the link?

Realtek integrated LAN on many motherboards has this function, can be accessed from BIOS even before booting an OS.

[J-R]

At work we have multiple 100GbE fiber links and are part of Internet2 (400G is coming).   And 40 and 10 all over as well to get that connectivity spread around.  But being able to get 2.5/5/10 over Cat5e for users has been a great upgrade for minimal effort.

At home I started adding 10GbE well over a decade ago but just using peer to peer links.  Cost, power/heat and noise are typical home lab concerns and I was also able to leverage SMB multichannel with 1G which worked well enough for the rest at the time.
But when Mikrotik released their affordable, passively-cooled CRS30x 10GbE switches in 2019, I jumped fully to 10GbE and I now have six of their switches spread around the house in a hybrid ring configuration.
Absolutely, 2.5/5/10 over copper needs power and generates a lot of heat, so I only use it where necessary and almost everything is fiber or DAC.

[Berni]

Yep i also got my first 10Gbit switch in the form of a 8 port passively cooled Mikrotik (most of my LAN is Mikrotik anyway) to start dipping my toes into speeds past the classical 1Gbit since i transfer a lot of files to my NAS and such.

The switch itself is impressively low power, but the RJ45 10Gbit SFP+ modules are so power hungry that it gets too hot to touch. I am using DAC cables where possible, but when it comes to Ethernet ports on actual devices like PCs, NAS servers, WiFi APs..etc they are always RJ45, so you are forced to use twisted pair to connect to them. They never seam to have SFP ports, so i bought some old PCIe network cards with SFP ports to solve that, tho a lot of these cards also run pretty toasty because they are mostly made for servers/workstations with a lot of airflow. Makes sense to use 10Gbit over twisted pair to reuse existing cable installations trough the building, but for connecting just 1 devices on the same desk it is a dumb solution, just let us use DAC cable.

[forrestc]

If anyone here knows: Is it common for Ethernet PHY chips to have this TDR function, and most devices just don’t use it? Or was it an oddball?

It's very normal (with varying degrees of quality) in managed switch ICs (serious ones, anyway, not one which offer 16 VLANs and a port enable bit), but to my knowledge unusual in end devices - why duplicate the functionality on the other end of the link?

Realtek integrated LAN on many motherboards has this function, can be accessed from BIOS even before booting an OS.

Having looked at many PHY datasheets for my day job,  I'm not sure I've seen a recently designed PHY that doesn't have that feature.  I suspect you may get it for free with the cable analysis you have to do with the faster ethernet speeds to make them work, but I haven't dug deep enough into the spec/negotiation to know for sure. (If you have to do a TDR to map reflections to be able to cancel them, then you get the TDR 'diagnosis' function for free silicon-wise).

Now, whether or not it's available to end users vs the hardware being capable of it, that's probably more hit or miss.

[Berni]

Yeah cable fault testing while usually supported by the PHY is not always exposed as a feature to the user.

Depending on the network card you have the cable test results might be in Device Manager for the card, it might require actually installing some utility program from the manufacturer of the card. Sometimes i also seen the cable test exposed in the BIOS menus.

Similarly switches and routers sometimes provide the cable test results while a lot do not.

You can likely do a actual TDR measurement fairly easily by connecting the twisted pair of interest trough a ethernet transformer as a balun and then send a fast edge square wave into it from a signal generator then use a scope to look at the edge. Where the scope trace is not in the middle of the screen that's where the cable is bad.

[J-R]

The Electronic Goldmine has these for $9.95: https://theelectronicgoldmine.com/products/g20107