9
\$\begingroup\$

1000BASE-T uses all 8 wires in a twisted pair to both RX and TX at the same time. It does it using (what seems to me) a complicated process known as echo cancellation. Details here.

1000BASE-TX dedicates two pairs (four wires) in a Twisted Pair to TX only, and the other two pairs to RX only. Conceptually, this seems far simpler to implement and work with than the alternative.

My question is, why did 1000BASE-T become the de facto standard for Gigabit Ethernet? Why not implement the (apparently) simpler standard of 1000BASE-TX? Why didn't 1000BASE-TX gain more popularity than 1000BASE-T?

Wikipedia had this to say about it:

However, [1000BASE-TX] has been a commercial failure, likely due to the required Category 6 cabling and the rapidly falling cost of 1000BASE-T products.

But I was hoping for a more detailed explanation, or maybe alternative reasons entirely.

\$\endgroup\$
  • \$\begingroup\$ This question would have been more suitable on networkengineering.stackexchange.com \$\endgroup\$ – Fizz Dec 1 '15 at 21:44
  • \$\begingroup\$ @RespawnedFluff I considered that, but I'm not sure NESE deals with the underlying details of twisted pair and their standards. I've had more luck on this stack exchange with my other very physical medium questions. \$\endgroup\$ – Eddie Dec 1 '15 at 21:51
  • \$\begingroup\$ Yeah, but this question is basically "why is cat 6 so expensive". \$\endgroup\$ – Fizz Dec 1 '15 at 21:54
  • 1
    \$\begingroup\$ @Passerby I edited the question to fix the mistake. \$\endgroup\$ – Adam Haun Dec 1 '15 at 21:59
  • 2
    \$\begingroup\$ For amusement purposes: 2002 article predicting the massive success of 1000BASE-TX: cablinginstall.com/articles/print/volume-10/issue-8/contents/… \$\endgroup\$ – Fizz Dec 1 '15 at 22:39
6
\$\begingroup\$

I'm not entirely certain of this, but given who advanced this 1000BASE-TX standard, namely TIA (aka the cable people, the standard being TIA-854) and not IEEE, I have the strong impression that it was largely a promotional exercise for Cat 6... a standard which is also defined by TIA. This impression was reinforced by reading how this standard was described as one of the first Cat 6 applications. I wasn't able to locate any vendor who made a PHY product for this 1000BASE-TX... although I've read in vague terms that something might have existed.

In other words, it seems stuff went like:

  • TIA: hey PHY makers, we've made this standard for a faster cable and here's a PHY spec for making your Ethernet life easier using it: cuts the number of your transmitters and receivers in half (but each needs to be twice as fast).

  • PHY vendors: crunch some numbers on potential sales of PHY for cables probably twice as expensive [back then] but giving the customer the same link speed and shrug at the potential sales. Don't invest in any R&D for this "golden opportunity".

In order for a PHY [standard] to have success, it needs commercial backing... from a PHY maker (not cable guys) In this case it seems none was really there.

It's worth noting that 100BASE-T4 vs 100BASE-TX wasn't entirely the same affair. At least there the latter had a bonus in terms of full-duplex operation. Also 100BASE-TX was actually part of IEEE802.3.

Even today (ok Dec 2014):

For budgeting purposes, and for the sake of this discussion, plan on Cat 6 costing roughly 30% more then cat 5e, and Cat 6A 30% more than Cat 6.

But at least nowadays the latter lets you run 10GBASE-T... up to 37m-55m on Cat 6 (exact max depends on nearby cables) and the usual 100m on Cat 6A. 10GBASE-T also uses all four pairs. There's not even a whiff of a "10GBASE-TX", by the way. Apparently those who design these PHY things concluded that more parallelism is preferable to faster clock... after a certain threshold. This is perhaps not unlike what happened with CPUs.

Also, this 2006 Broadcom presentation [on 10GBASE development] doesn't even mention the [2002 or so] 1000BASE-TX so the latter was probably a complete flop from the start.

enter image description here

Also from later slides of the same presentation, work on 10GBASE-T started in late 2002... so I think the IEEE efforts/focus went into that instead (rather than more gigE copper standards that apparently were not exciting for chip makers).

Oh yeah, the final nail[s] in the coffin are:

  • incompatible with 1000BASE-T; 1000BASE-TX proposed to use TBI (8b/10b) encoding (same as GigE over fiber), whereas 1000BASE-T uses 4D-PAM5. So basically not inter-operable in any way at switch level without a dual PHY. I guess nobody wanted to make that.
  • Probably incompatible with 100BASE-TX as well. Or at least there's no obvious provision for any sort of compatibility. 4D-PAM5 basically has a mode that uses only 3 voltage levels when it's not transmitting data (instead of 5 when it is), so that it can be compatible with MLT-3 used by 100BASE-TX. I don't see any effort in that direction in the 1000BASE-TX spec.

These are just my [semi-]educated guess[es].

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ What is more surprising to me is that 100BASE-TX actually succeeded in displacing 100BASE-T4. From this 1996 Adaptec document "A quick survey of cable costs shows four-pair CAT 3 plenum cable priced at $.09- .25/ft., and four-pair CAT 5 plenum cable at $.23-.37/ft". Also "100BASE-T4 Adapters are about 10% less expensive". The only real selling point there for 100BASE-TX was " Supports Full Duplex mode for up to 200 Mbps in servers". \$\endgroup\$ – Fizz Dec 2 '15 at 16:01
  • \$\begingroup\$ By the year 2000 though Cat 5 was already "the most widely installed LAN cabling infrastructure". \$\endgroup\$ – Fizz Dec 2 '15 at 16:31
  • \$\begingroup\$ Also, Cat5 and 100BASE-TX came during the dotcom bubble (and the liberal spending associated with that) of the latter half of the 1990s. Cat6 and 1000BASE-TX... came right after the dotcom bust in late 2001. So I think people were far more inclined then to opt for a reuse of their [fairly recent] cable upgrades than to redo it all over again for gigabit. In practice Cat 5 could often run at 1000BASE-T for shorter lengths, even if it wasn't over cable that had been tested/sold at Cat 5e standards. \$\endgroup\$ – Fizz Dec 2 '15 at 18:02
3
\$\begingroup\$

I would say that it is a tradeoff between the price of interfaces and the price of using more expensive cables (and having to change the cables in buildings).

  • 100BASE-TX needs Cat5 cable and 62.5MHz bandwidth.
  • 1000BASE-T expects Cat5e cable and 125MHz bandwidth.
  • 1000BASE-TX needs Cat6 cable and 250MHz bandwidth.

Cat5e is certainly much cheaper than Cat6 and was installed for 100Mbps with the prospect of upgradability to 1Gpbs.

The inverse occurred between the Cat3 100BASE-T versus Cat5 100BASE-TX. 100BASE-TX won but the advanced encodings (which allows to reduce cable bandwidth) from 100BASE-T were eventually used in 1000BASE-T.

| improve this answer | |
\$\endgroup\$
-1
\$\begingroup\$

Application point of view, 1000BASE-TX usually a point to point link, packets need not be repeated, therefore no need for collision detection within the domain, just two pair wire, just CSMA part of media access. Used in Avionics.

| improve this answer | |
\$\endgroup\$

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.