Maximum *Practical* Distance of CAT5e?

The answer is in the first line of my post:

Google Groups didn't list an obvious better place to post it in the alt.internet hierarchy, and as I said, friendly people have helped me here before. I had hoped that someone would politely answer my slightly off-topic question, rather than derail the thread into something about Usenet etiquette.

My apologies.

The correct newsgroup is comp.dcom.cabling. I posted articles there on the topic before.

You're on the right track. One of my fun demonstrations is to put connectors on both ends of a 1000ft roll of CAT5, plug it into my office LAN switch and my laptop and surf merrily. No error shown on the SNMP statistics in the switch. I also have several installations using CAT5 at about 800ft which work just fine. I'm not sure exactly what is the maximum, but my guess is about 1200ft.

As usual, there's a catch. You should use a switch or dedicated NIC port on both ends. 10baseT-HDX (half-duplex) only. Don't bother trying it with 100baseTX. It won't work. 10baseT-FDX (full duplex) is problematic because of NEXT (near end crosstalk).

The orthodoxy in comp.dcom.cabling will claim that the rules are the rules and that 100 meters was selected to insure that it will always work under all circumstances. They are correct. I can create situations where 10baseT-HDX will NOT work at much over 100 meters. Examples on request. However, if you're careful, you shouldn't run into any of these.

Your dumb and unspecified 10/100 NWAY switch will NOT guarantee

10baseT-HDX. It will start at the highest speed and stay there. To force it down to 10baseTX, the easiest way is to insert a cheapo 10baseT hub (not switch) in the line. That will force both ends to 10baseT-HDX. (Hubs cannot do full duplex).

You might also consider using coax cable. I have one system with about 1200 ft of RG-6/u coax, using 10base2 to 10baseT media converters at each end. Works just fine. The 75 ohm to 50 ohm mismatch is barely noticeable as the high cable loss eliminates any reflections from being seen at the opposite end. Radio Shock sells a Type F to BNC adapter which is all you need to make it work with common CATV coax. Do NOT use RG-59/u. It's usually garbage or worse.

The following thread might be worth reading:

Incidentally, if you want to go miles over one twisted pair, it can be done with SDSL. Some pairs of SDSL modems can talk to each other. Flowpoint, Lucent DSL-HST, Elastic, PairGain, Xpeed, etc can work without a DSLAM. 3com and Coppercom, cannot. 768-1500Kbits/sec symmetrical maximum with the older equipment.

Obstacles - wide and tall steel & concrete industrial buildings in between the two areas in question. The more I look into this, the more it appears I should just do it right and run fiber over a ~300 meter more protected route rather than the hazard-filled 150 m direct route.

I appreciate your responses. Thanks.

Why post this to alt.internet.wireless???

snipped-for-privacy@gmail.com wrote in news: snipped-for-privacy@z14g2000cwz.googlegroups.com:

Yes, I'm sure you can exceed the spec by a smidgeon or so, but you are pushing the range by 75%.

Don't think so. The range limitation is related to signal delay times and collision sensing, not the data rate.

That's true today, but Broadband speeds are going up, soon ... ...allegedly :-)

No, see earlier answer. Anyway, autosensing is about backward compatibility, not range optimisation as in wireless LANs.

Don't think it will work, see earlier.

I faced this problem with a customer who'd already laid the cable through underground trunking, and then called me in when nothing worked properly. The cable length was about 200m. Pings worked but with a high drop-out rate, anything else was unusable because connections kept dropping out.

Fortunately the cable was accessible in an intermediate building about half-way. I cut the cable there, reterminated and fed each half into ports of a 100Mbps switch, which cured the problem.

In your case, I suggest you plan to route the cable via a waterproof enclosure about half-way along, into which you can install a 100Mbps switch. You can power the switch from either end using a compatible power-over-Ethernet (PoE) kit, e.g. D-Link DWL- P100 or DWL-P200.

"Pierre" wrote in news:dkttl3$gcv$ snipped-for-privacy@lust.ihug.co.nz:

Whilst technically this would work (at 10Mbps), the downside is that it is based on technology that is already obsolete and, in my experience, very flaky.

snipped-for-privacy@gmail.com wrote in news:1131571941.600961.160360 @z14g2000cwz.googlegroups.com:

Why is that?

[POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]

Not a good answer.

You must not have tried very hard.

First link is in newsgroup comp.dcom.cabling, and several links of possible interest.

Not a good reason.

More than slightly off-topic.

For a link to fall within the definition of Cat5e that is indeed correct. And is also the distance limit in the Ethernet spec for 100Mbs and 1Gb Ethernet.

There is however no specific length limit for 10Mbs UTP Ethernet and in a point to point link round trip delay times wont be an issue.

There is an attenuation and crosstalk test specified for 10Mbs Ethernet and if your using Cat5 cable that equates to somewhere near 180M to 190M of cable.

However as others have suggested you must force either end of the link to

10Mbs, the auto negotiation process is not a qualative check on the speed capability of the cable.

Stuart.

I have run 100 meg at 170m, was a tad intermittent. 150 was fine.

not true - the limit was 100m - but it says Cat3 or better.

in practice 10M will go a long way on good cable with good conditions - see the separate post by Jeff Liebermann

Agreed - but even that is conservative - i.e. it assumes cable that just meets the Cat5 spec, and worst case combination of cabling practice, and reciever sensitivity etc. The various versions of what is claimed to be "Cat

6" and "Cat 7" would give you more margin to play with.

there was a fair amount of slack built in to make everything reliable in the real world - just think on how many patch leads get trodden on, how often someone uses a different type of cable in a patch lead etc.

Agreed. A relaible 2nd hand hub somewhere in the link may be an easy fix.

The old 3Com / Synoptics stuff seems to last forever....

if you must run a long way, then there are "long reach ethernet" products designed to run up to 15 Mbps over several 100 m of poor quality cable (phone wiring) - the cisco flavour is based on DSL type protocols. search for LRE on their site if you want to go this way.

Tis true, there was no specific length limit imposed in the standards for

10baseT, dont believe what you read in books written by so called experts.

The definitive reference document on Ethernet, the standard produced by the IEEE says;

IEEE Std 802.3, 2000 Edition.

14.1.1.3 Twisted-pair media The medium for 10BASE-T is twisted-pair wire. The performance specifications of the simplex link segment are contained in 14.4. This wiring normally consists of 0.4 mm to 0.6 mm diameter [26 AWG to 22 AWG] unshielded wire in a multipair cable. The performance specifications are generally met by 100 m of 0.5 mm telephone twisted pair. Longer lengths are permitted providing the simplex link segment meets the requirements of 14.4. A length of 100 m, the design objective, will be used when referring to the length of a twistedpair link segment.

Note the "Longer lengths are permitted" bit.

Stuart.

Another way to extend the range of ethernet is with LRE's. My employer is using them with 4000 foot cat 3/5 ethernet runs.