# Was Ethernet ever bidirectional over a single coaxial cable or single twisted-pair line?

Comments below this answer to the question Why was SpaceWire designed with nine wires? involve the history of Ethernet.

Did the earliest forms of Ethernet allow complete bidirectional communication over a single coaxial cable or a single twisted pair (possibly/probably with a shield)?

I've looked at the Wikipedia article on Ethernet and while it appears very well written and informative, I'm a networking noob and I'm having a hard time being sure I'm understanding anything there clearly. I searched for the terms bidirectional and bi-directional without result. In this particular case, I'm turning to Electronics Stackexchange for a concise yes or no, plus perhaps a bit of history.

• There's a subsection that's actually titled "Shared media". – The Photon May 22 '17 at 3:42
• @ThePhoton not everyone asking questions here is a career electrical engineer, nor would they need to be. "Shared media" sounds to me like a USB flash drive. – uhoh May 22 '17 at 3:46
• I'll retract my close vote. – The Photon May 22 '17 at 4:03

The earliest forms of Ethernet were not only bidirectional, all hosts on a network shared a single coaxial cable. This was a defining characteristic of Ethernet and the motivation for the Carrier Sense Multple Access with Collision Detection (CSMA/CD) feature.

According to Wikipedia,

Original Ethernet's shared coaxial cable (the shared medium) traversed a building or campus to every attached machine. A scheme known as carrier sense multiple access with collision detection (CSMA/CD) governed the way the computers shared the channel.

This scheme was maintained in the 10BASE5 and 10BASE2 physical media implementations, both based on coaxial cable. It was not used for twisted-pair implementations.

• Excellent, exactly what I needed, a concise, definitive answer, plus a nice bit of history. – uhoh May 22 '17 at 3:48
• To be clear, early Ethernet over coax was bidirectional but half duplex. That is, a station could send or receive but not both at the same time. 10-baseT and 100baseT could operate in full duplex by using different pairs for transmit and receive. The situation is different than in 1000-baseT -- which operates at full duplex with a station able to transmit and receive over all 4 pairs at the same time. – Evan May 22 '17 at 3:55
• However the thing that really killed that system wasn't the half duplex nature of it. It was that a single issue anywhere on the wire stopped the system from working. In order to connect a new machine you needed to break the cable and insert a T connector, while your doing that the whole network is down. For small static systems it worked fine but it was too fragile to be suitable for large networks. – Andrew May 22 '17 at 8:03
• To be pedantic, the very earliest Ethernet didn't use wire at all, hence its name. But the first widely deployed incarnation did use a single coax cable, with rather expensive taps into it where you wanted to access the signal. – user_1818839 May 22 '17 at 19:07
• en.wikipedia.org/wiki/ALOHAnet pioneered the CSMA technology in Ethernet (the "C" being "Carrier" inherited from its radio origins) and "ether" i.e. wireless medium in the name reflects that origin. To be even more pedantic, you could say there wasn't a formal standard called "ethernet" until it had been ported to cable. – user_1818839 May 22 '17 at 20:54

My 2nd computer at work had a Full Duplex 93 Ohm coaxial network from which I could boot from a Host in our network or locally from a HDD in the early 80's. The network technology used ARCNET which was cheaper & faster than Ethernet but eventually surpassed by coaxial Ethernet which later migrated to UTP.

It was a B22 at Burroughs and ran CTOS from Convergent Technology and supported over 4 programming languages and a 132 column green screen, which was a big improvement over an 80 column screen.

It had UNIX, COBOL, BASIC, FORTRAN and many applications and a mouse maze game but mainly a really good word-processor called "Write One" with different "daisy-wheel" fonts.

Token Ring was the initial popular daisy chain method with 93 Ohm terminations at each end and had faster thruput than ethernet, ran much further than 50 Ohm but eventually gave way to star topology now used in Ethernet.

You can follow the history by starting with ARCNET.

https://en.wikipedia.org/wiki/ARCNET

• The history in the linked article is an interesting read, thank you! Stories of competing/evolving technology are really timeless. Mostly it's the prefix that changes (kilo, mega, giga, tera...) – uhoh May 22 '17 at 4:11
• I was the 5th employee at a company in 1979 which invented a ISDN broadband WAN with DS1 2 way full-duplex 1.544Mb/s , our own graphics weather service, Pay TV with converters outside between poles to serve a dozen homes each, with fire alarm, burger alarm, teleshopping, opinion polling on TV, and digital telephony of course. It was called Project IDA and too far ahead of its time. The company was called Interdiscom Systems Limited and although we proposed Fibre, MTS wanted it on a single coax ,'cause they owned all the twisted pairs AND 200MHz solid coax distribution for TV in the province, – Tony Stewart EE75 May 22 '17 at 4:17
• I designed many of the custom network testers including Bit Error Rate testers (BER) , Subscriber SLIC testers, and modem 2 way time synchronous time slot testers, in addition to riding up in a bucket truck to adjust the modems in Winterpeg Winter and managing the entire network. – Tony Stewart EE75 May 22 '17 at 4:21
• hmm that was Burglar Alarm not Burger ;) Some of my midnite colleagues said I should have been working at NASA. Sleep was highly over-rated when we were having so much fun. – Tony Stewart EE75 May 22 '17 at 4:26
• 6.14 b/s/Hz over a trans-pacific distance is pretty good – Tony Stewart EE75 May 22 '17 at 5:07