Live data from Hacker News

The history of twisted-pair Ethernet [video]

youtube.com

21–27 of 27 posts

Re: The history of twisted-pair Ethernet [video]

#21

Dammit, the video doesn’t explain why we have 4 twisted pairs in our cables instead of just 2! One for Rx, one for Tx, but then why did they keep the other two? Surely they didn’t have 4 dead wires just because of legacy? So why the additional pairs, what were they used for?

Signal integrity basically. Using more pairs means reducing the bandwidth needed on each pair, thus the cost of the cable.

It's also a tradeoff with the cost of the NIC hardware --- look at how much bandwidth xDSL technologies have been able to squeeze into a single traditional twisted-pair phone line, but they use far more complex signal processing techniques than Ethernet.

Re: The history of twisted-pair Ethernet [video]

#22
post #16

My history with networking: - late 70s, rs232 terminals. well, it kind of worked. i spent a lot of time under the desk - early 80s, frozen hose ethernet - i remember not being able to get a sun workstation to sit on the desk because the cable lifted it off too much - late 80s coax ethernet all sorts of termination and other problems - 90s twisted pair and hubs - things started working as they should. - a bit later ch…

Never heard of frozen hose, funny.

I briefly used "thin" coax before the transition to IP cameras, it was a basic NTSC signal into a PCI capture card. 12/24V was pumped over a secondary cable attached to the main coax (siamese)

We were digital plumbers!

Re: The history of twisted-pair Ethernet [video]

#23
post #19
post #5

I found the reddit thread referenced in the video and left comments - On the original ethernet being a mistake - "I have another perspective on this - the bus concept at its roots dates to 1974/75 - a machine that could function as a hub in 1975 or even in 1979 would have potentially been as large and costly as the mini-computers Ethernet was intended to interconnect. A switch at that time would have been extraordina…

Also (as you know), twisted-pair Ethernet with a hub (let alone that STARNET daisy-chain thingy!) just isn't that different from coax Ethernet. Electrically it's largely shared-medium, half-duplex bus Ethernet wearing a mechanical hub-and-spoke disguise, and the hub is by definition just a multiport repeater. It's only when the hub is replaced by a multiport switch that shared-medium Ethernet is decisively left behin…

I very much agree - it has the advantage of not being an electrically shared medium however, which is one of the things that made it superior to 10Base2 or 10Base5

Here is some interesting documentation from the original StarLAN Hub -

https://vtda.org/docs/computing/AT&T/AT&T_Starlan_10_Hub.pdf

There is also some neat and ancient documentation still lovingly preserved on the NCR website -

https://onlinehelp.ncr.com/Retail/Workstations/Wiring/Ethern...

Re: The history of twisted-pair Ethernet [video]

#24

Dammit, the video doesn’t explain why we have 4 twisted pairs in our cables instead of just 2! One for Rx, one for Tx, but then why did they keep the other two? Surely they didn’t have 4 dead wires just because of legacy? So why the additional pairs, what were they used for?

Originally it was more or less one for POTS, one Rx, one Tx and one spare (used for power in ISDN S/T interface, which otherwise shares the same pinout).

Then came 1000-base-T, where the “base” is kind of misnomer and it is actually related to SHDSL. Each pair is essentially a separate full-duplex 250Mbps link with active echo cancellation and four of them is combined to create one 1Gbps link.

Re: The history of twisted-pair Ethernet [video]

#25
post #23
post #19

Earlier quoted context omitted.

Also (as you know), twisted-pair Ethernet with a hub (let alone that STARNET daisy-chain thingy!) just isn't that different from coax Ethernet. Electrically it's largely shared-medium, half-duplex bus Ethernet wearing a mechanical hub-and-spoke disguise, and the hub is by definition just a multiport repeater. It's only when the hub is replaced by a multiport switch that shared-medium Ethernet is decisively left behin…

I very much agree - it has the advantage of not being an electrically shared medium however, which is one of the things that made it superior to 10Base2 or 10Base5 Here is some interesting documentation from the original StarLAN Hub - https://vtda.org/docs/computing/AT&T/AT&T_Starlan_10_Hub.pdf There is also some neat and ancient documentation still lovingly preserved on the NCR website - https://onlinehelp.ncr.com/R…

-> https://onlinehelp.ncr.com/Retail/Workstations/Wiring/Ethern... (NCR Ethernet Communications Wiring Guide pdf 1998)

Re: The history of twisted-pair Ethernet [video]

#27
post #5

I found the reddit thread referenced in the video and left comments - On the original ethernet being a mistake - "I have another perspective on this - the bus concept at its roots dates to 1974/75 - a machine that could function as a hub in 1975 or even in 1979 would have potentially been as large and costly as the mini-computers Ethernet was intended to interconnect. A switch at that time would have been extraordina…

> push-thru modular connectors Yep. Very fond of the "EZ RJ 45" line in particular. The strain relief that's built into them does not interfere with loading the cable and actually gets a good bite into the jacket once crimped. The other advantage I've found is that it's harder to accidentally make defective cable ends. Trying to trim and face yourself you run the risk of one of the lines being too short or pulled bac…

Trim and face is the part thats really hard to learn, and do consistently well - also if you do trim and face well, you're likely untwisting too much wire to get it aligned correctly so it doesn't roll over in the connector while putting it on.

EZ RJ 45 does solve all of these issue, and makes for a reliable cable 100% of the time.

Post reply on HN