We all take Ethernet and its ubiquitous RJ-45 connector for granted these days. But Ethernet didn’t start with twisted pair cable. [Mark] and [Ben] at The Serial Port YouTube channel are taking a deep dive into the twisted history of Ethernet on twisted pair wiring. The earliest forms of Ethernet used RG-8 style coaxial cable. It’s a thick, stiff cable requiring special vampire taps and lots of expensive equipment to operate.
The industry added BNC connectors and RG-58 coax for “cheapernet” or 10Base2. This reduced cost, but still had some issues. Anyone who worked in an office wired with 10Base2 can attest to the network drops whenever a cable was kicked out or a terminator was dropped.
The spark came when [Tim Rock] of AT&T realized that the telephone cables already installed in offices around the world could be used for network traffic. [Tim] and a team of engineers from five different companies pitched their idea to the IEEE 802.3 committee on Feb 14, 1984.
The idea wasn’t popular though — Companies like 3COM, and Digital Equipment Corporation had issues with the network topology and the wiring itself. It took ten years of work and a Herculean effort by IEEE committee chairwoman [Pat Thaler] to create the standard the world eventually came to know as 10Base-T. These days we’re running 10 Gigabit Ethernet over those same connectors.
For those who don’t know, this video is part of a much larger series about Ethernet, covering both history and practical applications. We also covered the 40th anniversary of Ethernet in 2020.
The center pins 3,4,5,6 oddity came from phone lines originally
You know what we could use, instead of cat?
USB-C! 😂
So.. USB-C5e?
Not if you want any distance! For full 10 Gb/lane speed, USB-C 3.1 Gen 2 is limited to cable lengths of ONE meter! If you “only” need USB-C 3.1 Gen 1 (5 Gb) — and if you don’t want it to also carry charging power — then you can add a whole second meter.
Looked it up as I always wondered also.
https://superuser.com/questions/981961/what-is-the-logic-behind-the-pin-diagram-of-ethernet-cables
From article: The original concept was that the centre two pins would be one pair, the next two out the second pair, and so on until the outer pins of an eight-pin connector would be the fourth twisted pair.
…but the outermost pair are then too far apart to meet the electrical requirements of high-speed LAN protocols. … Hence the other two pairs are wired to pins 1 and 2, and 7 and 8. Why we have T568A and T568B.
Not only that, but you can physically plug a RJ11 (2 pair/4 wire) or RJ12 (3 pair / 6 wire) connector into a RJ45 socket. Back when RJ connectors were being used instead of DB connectors for serial ports, some of the pinouts accounted for this and gave you ground, TX, and RX on the center pins so you could use the cheaper RJ11 cables if you didn’t need the handshaking lines.
We have RJ-45 phone shockets at office, plugging normal RJ11 into them. They are just CROSed into a different device
I have always wondered why the pairs in the 8 pin modular jack are connected as 1-2 and 3-6. Why skip pins 4 and 5? Now I know.
I really wish either RJ-45 connectors weren’t chosen or at least they would come up with a better locking mechanism. The plastic locking tab is the real weak spot, so easy to break.
True. If you are moving them all the time. But easy enough to clip the connector off, then attach a new connector if becomes a real problem.
yeah. It would have been better if the locking mechanism would have changed sides, locking tab on fixed receiving side and just a ridge on the connector. but alas, that happened in a parallel world somewhere.
its just like fighting cable spaghetti with db25. the locking screws always hook onto other cables… but in case of the rj45, the tabs break off.
As Yaroslav says, there have been a LOT of innovations in RJ-45 over the decades. The most wacky, perhaps (but also the one that best fits your description) is called etherCON and is used by the performance industry (musical instruments, DJ equipment, event video, etc.).
It turns either end into what looks like an XLR cable (big metal tube about 1″ in diameter), but instead of pins for audio channels it has an RJ45 connector floating in the center. the locking mechanism is the same one XLR uses, and captures the outer tube rather than the RJ45 itself. And of course, because the cable ends are cylinders, they can’t possibly snag anything.
They cost close to $100/meter, but I guess it’s worth it when it’s the difference between a live show going off without a hitch, and one that craps out halfway through because a cable fell out of a jack.
There is a hack for that. Eg https://www.cabletiesplus.com/blog/fixing-an-ethernet-cable-with-zip-ties/
Actually.. they have. Just mist of the time you deal with cheaper knock-off but compatible version of sockets and plugs, no retention, no shielding, no locking, etc, which cost 2cents a-piece
The RJ45 is still way better than the BNC from base-2 or AUI, guess you could try to terminate cat5e/6 onto some Token Ring connectors for more robustness
I built my house in 1993. Wired 2 runs of CAT3 (which was the fashion at the time) from each room to four 66 blocks (WWTFATT) in the basement. One run was designated for phone, the other was for data.
Still using them at 10 (the HP Printer) 100 (mostly everything else) and Gigabit (the wifi access point) speeds with no errors. I added a second 110 patch panel for the CAT5 runs in my basement office during COVID.
Bell Operators Give Better Service!
(Blue, Orange, Green, Brown, [Slate])
When we bought this old house in 1996, I pulled CAT3 everywhere I could think of. It happily runs gigabit. I think it is because the runs are fairly short, under 15 meters.
The important thing was that I had an onion on my belt, which was the style at the time.
I wired up my old house with 10base2 (Thinnet) to run my Novell server. I just ran it to 2 rooms at the time.
Later I decided I needed to rewire the whole house for phone, Ethernet, and video, using CAT5E and RG59.
That was a project. When I planned it out I decided on using a specific size of conduit, based on the measurements of all of the cables I needed to pull. Bought all the conduit and installed it. Unfortunately I didn’t get the cables pulled right away. In the mean time, I decided to add a couple more rooms to the network and I switched from RG59 to RG6. So when I went to pull the cables it was a very tight fit in the conduit. I think I used a ton of cable lube getting them pulled.
The moral here is plan your conduit runs on the size of the cable bundle and then double the size of the conduit!
(and, yes, I have some yellow “classic” Ethernet cable as well) I’m currently using a couple of long lengths as feedlines to my HF antennas. Aside from the problems of shared media, the AMP-made, IEEE standard AUI “slide latch” connector was another reason I was glad to see twisted pair Ethernet.
The less said about twisted pair Token Ring, the better.
AUI had twisted pair, as I found out after wiring up a bunch of cables ignorant of that fact.
(Yes, I went back and fixed them.)
Shielded, large gauge twisted pair. I’ve forgotten the details, but once they made it plenum-safe, it was almost impossible to bend. And WAY to unbendable for those poor 15 pin sub-D connectors. And that poor slide-latch never had a chance…
True that!
“The industry added BNC connectors and RG-58 coax for “cheapernet” or 10Base2. This reduced cost, but still had some issues.
Anyone who worked in an office wired with 10Base2 can attest to the network drops whenever a cable was kicked out or a terminator was dropped.”
Not here in Germany. ;)
We had a custom version based on TAE connectors (our phone line connectors).
https://de.wikipedia.org/wiki/EAD-Kabel
To be fair, though, that technology was used in bigger offices.
Home users had used ordinary coaxial cables w/ BNC.
When I first moved to Germany we had those for our phones. They were so chunky compared to the RJ11 I was used to.
Hi, yes, I think that’s understandable.
Gratefully, TAE was still more compact than the previous monstrosities, though.:
https://de.wikipedia.org/wiki/Anschlussdose_(Telefon)#ADo_4_und_ADo_8
Even though in daily life it was annoying that FAX/modem, answering machine and telephone had slightly different TAE connectors.
(TAE-N vs TAE-F; their “nose” was different; they could be modified with a carpet knife to make them fit in the other socket).
We also did have RJ11 at the time, I think, but not on the wall.
RJ11 was rather used as a device connector on fax machines and modems, mainly.
Some landline phones with replaceable cables had used RJ11 on the underside, too. The other end was TAE.
The cord between the handset and the phone sometimes had used an RJ11 connection, I think.
Some brave souls also ran PhoneNet over TAE, it seems.
https://www.brix.de/computer/localtalk_sx.html
Though that didn’t work with a stock installation, I assume.
Our phone cabling in the walls didn’t feature two pairs by default, as it’s common in the states.
Also, some of our vintage phone wiring is so old that it didn’t use twisted-pair yet but something closer to thin loudspeaker cable (with four wires in parallel, two used).
That’s a lot of “fun” in combination with DSL, I think! :D
My brother was an accountant in a small office during the 1980s.
As the first person there with a computer, he became their “guru” when bosses and secretaries got computers for their work. So, it fell on him to wire up the network when that time came. I was pleasantly surprised when he told me that he successfully crimped and ran the ThinNet, and installed the daughterboards and software, as I didn’t think he had knowledge/ability of crimping the connectors. I was a bit embarrassed that in he knew more about computers than I (the family’s electronics guru) did at the time.
Necessity is the mother of invention.
“Misery makes inventive.” (translated) in my mother language. :)
“Necessity is a mother!”
-Frank Zappa
Another thing to notice about twisted pair Ethernet on RJ connectors: the color codes alternate between solid and striped: W/O, O, W/G, B, W/B, G, W/B, B (for 568B). Helps when you’re checking to make sure none of the wires moved as you inserted them into the connector.
I’ve crimped a LOT of Ethernet connectors. And all of the patch panels I’ve seen are wired to 568B
T568B had slightly better tolerance to cross talk. T568A is mostly in federal government systems as that was the standard internally for many departments in order to maintain compatibility with older base establishments.
I suck at that step of making ethernet. “pass through” connectors have helped me a lot
I remember the RG-58 coax segments for thinnet as a young child at my parent’s workplace in the 90s (I’m getting so old.) I believe the coax is still there to this day (of course it’s long gone unused.)
I got into (being able to afford) networking recently enough that the legacy coax flavors just feel so alien.
The slightly touchy and expensive, but superior, cables have just always been optical(with the very limited exception of short DAC patches; which I think are often twinax but are only really standardized in that the captive SFPs should be happy about the situation and so can be whatever they feel like so long as that works); and twisted pair has always been the low-end-but-you-can-shove-it-under-a-desk-and-kick-it-for-a-decade option.
We still use more RF ethernet than ever; but at least consumer tier wifi seems to go out of its way to mostly hide that and discourage you from trying to cable it.
I wouldn’t call Twisted Pair “low end” but rather the workhorse standard.
Aside from the major cost differences in installation, and the fact the vast majority of networking equipment is RJ45 instead of SFP, they’re also MUCH simpler and cheaper to fix. It’s also less common to need runs over 100m uninterrupted which is where the optical would shine.
When I worked on the USNS Comfort doing a rip and renew of the entire network we ran thousands upon thousands of RJ45 terminated Cat5 cables between the switch stacks and the PCs and printers and all that. Those also ran port aggregated Ethernet up to the main routers which did run on optical. Main optical run on the main deck of the ship ran almost the entire length, about 250m, and was fat, maybe 1/2″. They had to lay it flat and then crews worked to get it hung overhead without exceeding the bend radius. Well, about halfway done and one of the Navy personnel onboard was rolling a cart full of printers out of one section and out into the hall. There was an awful cracking sound as he went over the cable. We all knew how much it cost, and were very aware of it and to stay away while they were working. Ouch.
The fact that you can run Cat6 UTP as far as 100 m without any sort of amplifier or active repeater, and still achieve 10 Gb throughput is nothing short of miraculous.
And it wasn’t even the great innovation of 10bT ethernet. Telcos and even telegraph services already knew about the effects of crosstalk, and how twisted wire-pairs could be used to defeat it… over MUCH longer distances. There were old telegraph lines strung up on poles 100 years ago where you could see the wires lazily looping back and forth over each other across MILES of distance.
But analog is a lot more forgiving than digital, in many ways, and 10 Gb over 100 m is still mind-blowing for electrical signaling over a wire that can be bent, wrapped around poles, fished through walls, and tossed down risers.
I mean, HDMI starts to get nervous when it reaches TEN meters, and the rated max is 15 m. USB-C can only handle 4 meters for passive signaling, and when it’s carrying power or needs the full 10 Gb data rate that drops to 1m. UTP ethernet is a wonder.
We bought cat 5 ,10000 feet at a time. Installing the connectors was tricky and it was s 50/50 shot which end was bad if the cable didn’t test. Good times!
I’ve worked with somethung called ArcNet back in the day. Basically used RJ11 and typical telephone cable.
*8P8C Modular Connector
I still do my own cables. Have the crimp tool and tester, so why not :) . Nice to have ‘just’ the length you need :) . That said, I would not want to make cables as my day job… Nope notta. But one once and a while, is ‘ok’.
Being able to make your own cables also comes in very handy when the locking tab inevitably breaks off. You can just cut it off and crimp a new one on instead of ripping out and replacing the cable.
Re: “ripping out” and replacing — the bestest skill I ever acquired during my time doing physical networking, even more valuable than being able to crimp my own RJ45s, was populating punchdown blocks.
If all of your ethernet runs over distance use unterminated cable that gets punched down to a distribution block on the service end, and to the back of a keystone jack on the client end, you’ll never have to “rip out” an ethernet cable ever again.
(Especially since the OTHER bestest skill I learned was, always run twice as much cable as the current wiring plan calls for. Cable is cheap. A “wasted” length of wire stuffed in your walls today, becomes a spare you can swap to when some line inexplicably fails 10 years from now. Or, a drop you already have ready to terminate, when someone suddenly needs 5 jacks instead of the 3 they originally swore would be more than they could ever use.)
Nerdvana is a world where only the short, last-meter connection between a device and its conveniently-located jack uses RJ45. If one of those patch cables breaks, it’s hardly worth the effort to crimp on a new connector.
Lucky you who came in the late 90s and escaped from most of that crap. Hooking up stuff with what were literal tubes was definitely nooo fun, like those blasted Proteon Pronet token-ring boxes.
YT pushed that video to me so hard I thought it was clickbait.
We dabbled with the BNC connectors, but seemed like overnight we were on to using RJ-45. Note to begin with we used ‘hubs’ which did make it nice for spying on ‘all’ the traffic from you desktop, but of course that caused congestion problems. Finally switches came a long to split up the traffic. I could no longer spy on ‘user x’ traffic unless it was heading my way….
To be fair, we also had used null-modem cables (RS-232 serial) and LapLink cables (mostly parallel)..
Laptop users “loved” them.
Little Big Lan on DOS used to be popular in early 90s in Germany (sold as Kirschbaum Netz).
It also supported ethernet (NE2000 cards for 10Base2, ARCNet?) and mixed networks.
Then there were the Macs with AppleTalk over their LocalTalk cables (in the US, PhoneNet was a more popular alternative).
LocalTalk compatible PC cards (ISA) and DOS/Windows 3 software had existed, as well, so PCs could work in a Mac network.
And as if that wasn’t fun enough already, there had been switches, hubs and bridges to combine all sorts of network topologies.
As a useful extra, they did allow to convert from one physical format into another one. 10Base2 to 10BaseT being one popular possiblity.
Other software packages were Novell NetWare, NetWare Lite/Personal NetWare (NV DOS 7) and Lantastic..
And Windows for Warehouses, of course! :D
So it wasn’t dark ages. Rather contrary, the 50 ohms RG-58U cable, as used by 10Base2, was overly popular among CB radio operators and radio amateurs (now it’s obsolete). Merely RG-213 was more sought after.
They could assist in making network cables just easily, thus.
First LAN I made (for hooking together my and my roommate’s PC’s for playing Quake) I used an existing piece of cable TV coax and F-to-BNC adapters, along with tees and terminators. It worked great. However, in order to later support two friends bringing their PCs over for a LAN party, we moved it all to twisted pair. Fortunately, adapter cards at the time had both connectors. We probably used a hub, as they were cheaper than switches at the time.
Well done! The terminators were probably the reason it worked so well, despite impedance mismatch.
Because the sat cable or TV cable normally is 75 ohm rather than 50 ohm that the network cards and CB/ham radios use (old cable TV cable is 60 ohm), so there is a slight mismatch.
On other hand, the cable shielding probably was good here and there was no interference or reflections, so it worked just fine.
So cable was of good quality, likely, a bit above the norm.
I mean, sat cable and US cable TV cable do look quite similar, I think, since both use F connector. So maybe it was low-loss cable back then?
Because in my country, it is (was) a bit different.
Ordinary TV cable (for terrestrial reception) was using 75 ohm and had those Belling Lee connectors since the 1970s or so.
Our cable TV cable looked similar to this and was “merely” UHF capable (up to ca. 900 MHz) rather than sat capable (up to ca. 2 GHz).
https://en.wikipedia.org/wiki/Belling-Lee_connector
Merely for satellite TV reception we used to use the cables with F connector – just like you do for cable TV.
As for the RJ45 connectors themselves (well, more likely the RJ11 that preceded it), rumor has it that it was chosen from among the other competing options when its proponent took one of the cables, laid it on the floor, jumped up and down on the connector a few times, then picked it up and plugged it into a jack, demonstrating that it was still fully functional.
Which is a cute demonstration, if true, but I kind of wish they’d come up with a few more challenges — like pulling one through a nest of wires without snagging the tab on anything and snapping it off.
“men at work” :D
That takes me back to when we had our first LAN parties at the amateur radio club. We used 10Base-T and called them “quake sessions” because we played Quake and the term lan party hadn’t been coined yet (or at least we hadn’t heard about it).
It used to take half the night to debug the coaxial (terminators!), get network drivers on everyone’s machine and figure out the right IP addresses.
Good memories but the introduction of twisted pair and switches has made networking so much easier.