28

We run a Netgear GS724Tv3 switch and part of the configuration per port is an option for "Lower Power Mode". The documentation for the switch lists this feature as a "Green Ethernet Configuration" feature:

A cable test is performed when this is enabled and the port link is up at 1 Gbps: if the cable is less than 10 m, PHYs are placed in low power mode (nominal power)

We have purchased a set of short patch cables of lengths less than 1 m, and we might possibly be having issues with a few of them (testing is still under way). We have this feature disabled at the moment.

Should this feature be enabled if the cables between a NIC / SWITCH are less than 10 m where power consumption is not a concern? As a follow-up question, is there a minimum cable length requirement for Cat 6 gigabit?

Peter Mortensen
  • 2,319
  • 5
  • 23
  • 24
Geoff Dalgas
  • 2,416
  • 5
  • 31
  • 32
  • 10
    I'll be interested to hear the answer to this one. I got into an argument with a former IT vendor for one of my Customers who *swore* that there was a 1M minimum length limitation for gigabit Ethernet but couldn't provide any documentation therewith. FWIW, this "lower power mode" sounds dodgy and I'd avoid it. What are the possible cost savings from an Ethernet switch, per year? Is it worth the risk? (I can't say I'd be using a "Netgear" switch in any kind of corporate capacity either, but...) – Evan Anderson Feb 01 '10 at 20:52
  • 1
    Minimum cable length sounds vaguely familiar to me too. However, I was thinking that it was for fiber optic. Hmmm. Trying to think of a part of the spec where a minimum might be required. – Brian Knoblauch Feb 01 '10 at 21:03
  • 1
    If I had my network gear handy I'd just make a patch lead as short as my crimpers (and fingers) would allow and see what's what. Although I guess if it's *too* short you'll loose the T part of UTP. – Mark Henderson Feb 01 '10 at 21:37
  • 2
    The minimum cable lengths are only relevant for half duplex mode where CSMA/CD comes into play, so they are relevant for 10Mbps and 100Meg half duplex. If you are running your GigE network Full Duplex they shouldn't be relevant. I'm looking for a definitive reference for this but haven't found one yet, when I do I'll answer. – Helvick Feb 01 '10 at 21:43
  • @Helvick. I think you are right, the wikipedia article has a mention about this, but it doesn't say what the minimum length is. It only says that it inserts blank frames in half-duplex mode (GigE only) – Scott Lundberg Feb 01 '10 at 22:48
  • @helvick but do the minimums apply to UTP? I thought that was only for coaxial – Jim B Feb 02 '10 at 20:27
  • @Jim - I didn't think coax was usable for even 100Mbit let alone GigE but that's just speculation on my part, I haven't had to work with coax in over 10 years. Still your point rings some bell somewhere in my head. – Helvick Feb 02 '10 at 21:08
  • Tangential - there are definitely minimum lengths for fibre/fiber cable, mostly to do with the signal level being too high. – Criggie May 04 '21 at 04:11

8 Answers8

23

This question is still missing an authoritative reference, so here it goes:

There is no minimum length specification to a 1000BASE-T segment.

Charles E. Spurgeon - Ethernet, the Definitive Guide, p. 163.

the-wabbit
  • 40,319
  • 13
  • 105
  • 169
9

According to http://www.ctrlink.com/2006_07_01_archive.html there is no minimum length when using a star topology (one node connected to each port of a switch). In the older times, when you had an Ethernet ring, there had to be minimum distances between devices to prevent impedance problems on the ring.

Concerning the low power mode, are you sure that isn't related to POE? I don't see how you can lower the power and still meet the guideline for voltage. Based on the impedance of the cable, the power consumed will be directly related to the voltage put on the cable by the NIC... See Ohms law, which isn't perfect when you are talking high-frequency waveforms, but the same principles still apply.

Peter Mortensen
  • 2,319
  • 5
  • 23
  • 24
Scott Lundberg
  • 2,364
  • 2
  • 14
  • 22
  • 1
    Thanks for the info - POE doesn't exist on this switch. This is an energy savings feature according to the manual. ftp://downloads.netgear.com/files/GS716Tv2_GS724Tv3_usermanual.pdf – Geoff Dalgas Feb 02 '10 at 00:37
  • 1
    Per this link, it's only powering down when the port is idle. Makes sense... http://www.lightwaveonline.com/equipment-design/transport/news/infineon-phy-enables-small-footprint-low-power-for-gigabit-applications-54895962.html – Scott Lundberg Feb 02 '10 at 02:19
  • 2
    Apparently this is part of an upcoming spec for Ethernet. 802.3az spec. http://ieee802.org/3/az/public/index.html – Scott Lundberg Feb 02 '10 at 02:21
5

I've never seen a minimum cable length recommendation. This feature seems like a fairly straightforward power-saving feature. Shorter cable runs incur lower power loss.

According to the NetGear documentation, there are actually two separate power modes, both of which can be enabled. One of the modes is for cable less than 10 meters.

http://kb.netgear.com/app/answers/detail/a_id/11701

"What are the Green Ethernet Features?

"GS716Tv2 and GS724Tv3 support two power saving modes: Auto Power Down Mode and Short Cable Mode. Both features can be enabled at the same time. When Auto Power Down mode is enabled, power will saved for ports that have no link. When Short Cable Mode is enabled, the device will check the cable length and reduce the output power if the cable length is less than 10 meter."

Here is a fairly good overview of the TIA-568-B recommendations and practices:

Greg Askew
  • 34,339
  • 3
  • 52
  • 81
4

I can't find a link to the spec (which isn't suprising as TIA sells it) however seimon has a pretty good selection of summaries. There is no mention of a minumum length but there is one since the minimum bend radius dictates it for the cable (4 times the cable diameter for horizontal UTP cables under no load conditions- I'll let some math wizard figure that arc length out). It's certainly less than 1M. Here is the summary of cable practices here: TWISTED-PAIR CABLING INSTALLATION PRACTICES

There is a guideline in ANSI/TIA/EIA-568-B.1 that says the consolidation point should be located at least 15 meters away from the telecommunications room to reduce the effect of connectors in close proximity.

Also note that (particularly older cat6 cable) some cat6 cable has splines that will prevent the cable from exceeding the minimum bend radius and will most certainly not be useful shorter than about 2 ft. I haven't seen much of that out there but it's out there.

Jim B
  • 23,938
  • 4
  • 35
  • 58
3

To the best of my knowledge there is no minimum cable length beyond the practical (how closely can you get the connectors to each other and successfully terminated).

I have personally used Cat5e cables in the 4 and 6 inch range at Gig speeds without any issues. (This was with Cisco 3500/3550/2960 switches it should be noted).

This is pure speculation but, the "Lower Power Mode" sounds like the switch will energize the conductors/pins with less electric power at shorter distances. I don't know how much power savings this could possibly be -- perhaps their literature has projections?

Peter
  • 5,403
  • 1
  • 25
  • 32
3

I've often wondered about this issue myself. Everything I read online is speculation pointing either to no minimum or 1 m.

I once had an electrical engineer tell me that from an electrical standpoint you would want some distance (he said more than 3 ft (0.9 m)) to allow for the twisted pairs to operate efficiently. I am by no means an electrical engineer though so I can't explain it beyond what he told me.

Peter Mortensen
  • 2,319
  • 5
  • 23
  • 24
einstiien
  • 2,538
  • 18
  • 18
2

Although there is no minimum cable length specification for 1000baseT, any cable still has to meet various requirements like Return Loss, Near End Crosstalk, and the other specifications set forth in the IEEE 802.3ab 1000BASE-T standard, and if the cable is too short there is a possibility that it might not meet some of these requirements.

So essentially, it's not so much the cable length that matters, as it is making sure that the patch cable is certified to meet the standard.

(I'm assuming that this should only be an issue when making your own custom short cables).

(Source: HP ProCurve doc id emr_na-c01139355)

Ben791
  • 21
  • 3
0

International standards - ISO 11801 and EN 50173 / 50174 are your friend. Typically 1m is the minimum for a direct link, but if you’re connecting to patch panels (i.e. the end to end cabling has 3 or more connections) the minimum patch cable length is 2 metres, with a minimum length of 15m for structured cabling, and an extra 5 metres for any structured cable run from an intermediate point, such as area distribution.

This avoids the interference / reflections / cross talk caused by having multiple connectors close together in series.

This is reflected into the warranty/installation requirements for most patching and cabling systems, and can be referenced by local building codes, codes of practice, and legislation, for example BS7671 and BS6701 in the UK.

Example structured cabling guides (assuming you don’t design networks for a living, and thus need to buy the standards) can be found here: http://www.lavancom.com/portal/download/pdf/catalog/Installation%20and%20Testing%20Guideline.pdf

https://www.nexans.co.uk/eservice/UK-en_GB/fileLibrary/Download_540209803/UK/files/General+Installation+guide+2014_V2_1.0+DWI_LR.pdf

Lauren
  • 1