8

In 1000Base-T Ethernet connection, why is there a 100 m link length limit?

CSMA/CD links normally are limited to a few hundred meters because of the limitations that CSMA/CD implies (collision domain size). But with full duplex point-to-point links, no limitation other than damping is implied.

So, where does the 100 m limitation come from? When using 1000Base-T in full duplex mode, can this be ignored?

Thanks!

Ron Maupin
  • 98,218
  • 26
  • 115
  • 191
David
  • 213
  • 2
  • 6

3 Answers3

7

The 100 meter limitation comes from physics and the installed base of cabling. One hundred meters is a convenient length due to legacy specifications (10Base-T and 100Base-TX) since the installed cable base was already limited to 100 meters (CSMA/CD, as you pointed out, limited 10Base-T to 100 meters). You may have noticed that every twisted-pair standard strives to meet the 100 meter length for this reason.

The medium, twisted pairs of copper wire in this case, has certain properties that have inherent limitations due to electromagnetism. This is a large subject, but a simple search will yield a lot of results. Basically, you have external and internal noise that will degrade the signal. The longer the run, the worse it is, and this problem gets worse at higher bandwidths.

It becomes a tradeoff: it is better to specify a fixed length and increase the speed rather than keep the speed the same and increase the length.

Ron Maupin
  • 98,218
  • 26
  • 115
  • 191
  • I think it is actually 90 meters? The remaining ten meters is to account in the 5m patch leads at both ends. – user4565 May 17 '15 at 20:13
  • @user4565, that is slightly different. I could have a single cable that connects a PC to a switch, and that could be 100 meters (ignoring the differences between solid-core and stranded cable). In a typical cable installation, you have the <= 90 meter horizontal cable (solid-core), and the two <= 5 meter patch cables (stranded). The solid-core cable has better performance, but is fragile and can't be moved a lot. The stranded cable is less fragile but has poorer performance. The total length is limited to 100 meters. – Ron Maupin May 17 '15 at 20:23
  • I would say"fragile" is a great exaggeration. – Peter Green Dec 02 '15 at 08:50
  • @PeterGreen, the fagility of solid core copper is that if it is bent too aggressively, that will potentially cause severe attenuation at the bend and/or break the copper entirely. Stranded copper is much more resilient to bends and movement (even if individual strands break the "weave" of the strands maintains continuity). – YLearn Mar 02 '18 at 23:16
-1

it is nothing to do with signal degradation is to do with physical propertiy of a signal across the wire. The length is exactly 100 metres as (doing the maths) anything over 100m will be long enough for an ethernet device to transmit a frame and not hear the collission detect back-ff from the far end. so anything over 100m will be subject to late-collisions.

Kevin
  • 1
-3

To use 1000Base-T modules at 1Gb/s you also need Cat 5e or Cat 6 ethernet cable else there is no guarantee that this will work at 1Gb/s.

Also a 1Gb/s link uses all 8 wires in the cable where as a 100Mb/s link only uses 4, in case you have some wires damaged inside the cable it might work well at 100 Mb/s but not at 1Gb/s.

Szombatfalvi Hunor Gellèrt
  • 46
  • 1
  • 3
    You are not answering the question... – Gerben May 18 '15 at 09:52
  • Also wrong, 1Gb/s still only uses 4 wires / 2 pairs. All 8 wires are normally only used for PoE. – Jon Rhoades May 30 '15 at 10:24
  • 1
    @JonRhoades, you are in correct. 1000BASE-T requires four pairs (eight wires). If you try to connect at 1 Gbps with less that four pairs, you will only connect at 100 Mbps. – Ron Maupin Aug 06 '17 at 02:41
  • 1
    @Ron "you will only connect at 100 Mbps" - if at all; fallback is not by standard and only done by some vendors. – Zac67 Aug 06 '17 at 10:48
  • @Zac67, auto-negotiation is a _requirement_ for 1000BASE-T in the IEEE standard. If the two correct pairs are connected, it will negotiate 100BASE-TX. – Ron Maupin Aug 06 '17 at 16:47
  • @Ron The fallback to 100BASE-TX isn't defined by autonegotiation. When link pulses indicate mutual 1000BASE-T, AN is done and the PCS tries to start up the lanes. Failing that, the outcome is not defined. Some ports just try again and again - try older Intel or Realtek hardware. – Zac67 Aug 06 '17 at 16:57
  • 1
    @Zac67, as Ricky Beam pointed out a while back from testing, some of the early 1000BASE-T chipsets were capable of this auto-negotiation, but there was a bug that prevented it from working. The modern chipsets all will fall back to 100BASE-TX. – Ron Maupin Aug 06 '17 at 20:01
  • Thanks, good to know. We've still got several pieces of hardware that don't fall back and our users keep killing the cables. – Zac67 Aug 06 '17 at 20:15