How short can a cat5 cable be?
- Thread starter reicherb
- Start date
it's certainly common for cat5 cables to be extremely short (less than a foot) Take for example the case where you have multiple switches stacked vertically in a rack...the only real physical limit is how short can it be before you can no longer fit connectors on it.
I was under the impression that the cable had to be a minimum length to prevent cross talk and other interference. I though that the signal needed to go through a certain number of twists.
is this true?
is this true?
Minimum length for twisted pair cabling (cat 3 or 5) between any two active devices (hub, switch, NIC, router, etc) is 1 meter.
BlueWeasel
Lifer
- Jun 2, 2000
- 15,944
- 475
- 126
Spidey07:
I would assume you mean the 1-meter minimum length for cable is the ideal length. Because I have had cable used just to connect 2 stacked hubs that is 1 to 1.5 feet in length. Communications between the hubs works great.
Is it best to have the cable at least 1-meter long?
I would assume you mean the 1-meter minimum length for cable is the ideal length. Because I have had cable used just to connect 2 stacked hubs that is 1 to 1.5 feet in length. Communications between the hubs works great.
Is it best to have the cable at least 1-meter long?
yeah a meter sounds right, there is a min length.
i forget what it is, but it sure as heck is above a foot.
i forget what it is, but it sure as heck is above a foot.
A meter might be listed in some spec somewhere, but in real life shorter works just fine. I use quite a few one footers between switches and they perform perfectly.
Russ, NCNE
Russ, NCNE
I'm pretty sure the min for Ethernet is 1 meter.
Will shorter ones work? Heck yeah, just like a 130 meter cable will work as well. It is just out of spec.
Has something to do with reflection from the other end. I'll try to find a link.
Will shorter ones work? Heck yeah, just like a 130 meter cable will work as well. It is just out of spec.
Has something to do with reflection from the other end. I'll try to find a link.
There was a minimum length between devices on a network and 24 inches sounds right, but I'm pretty sure that was for thinnet coaxial. I'll check my NetEss books.
The shortest cable I've every been able to get working was 1ft. Shorter than that, and my nice Fluke cable tester showed me why it wasn't working -- crosstalk from heck! But this experiment was only performed to settle a bet which I lost, because I thought 1ft min. distance was arbitrary and ridiculous. Now that I'm older, I've learned that those two qualities describe most of the devices in the network.
I've cut a 15cm cat5 cable to use it between my router and switch without a problem ....
Just make sure the twist is still there to prevent crosstalk..
eRr
Just make sure the twist is still there to prevent crosstalk..
eRr
ERR:
You do have problems, you just can't see them.
Yes you can use an ethernet cable over 100M. You will get link and move frames.
Yes you can use an ethernet cable under 1M. You will get link and move frames.
Heck, you can use coax on a thinnet ethernet without earth grouding one end.
Man, you can even use a cabling plant that is not properly gounded.
BUT, you will have errored frames from crosstalk and reflections. You can check this by looking at the statistics on the switch, hub, whatever active device it may be. Using cables outside of spec will not show up as an obvious problem, but you will suffer performance consequences. Put a sniffer on the line and you will see frames with failed CRC, alignment errors, or even worse late collisions. I've seen it with happen with way too many networks (can you believe it! I actually get paid for telling people "dude, cabling is screwed and too long or your fiber is too weak, call me when you've fixed that"
This is simply the way ethernet works. It is the same with all cabling/network standards (SONET, 10Base-FL, FDDI, token ring, 1000Base-X, etc.) It will work, just not as well as it should.
Hope this clears things up, I'm just a stickler for details.
cheers
You do have problems, you just can't see them.
Yes you can use an ethernet cable over 100M. You will get link and move frames.
Yes you can use an ethernet cable under 1M. You will get link and move frames.
Heck, you can use coax on a thinnet ethernet without earth grouding one end.
Man, you can even use a cabling plant that is not properly gounded.
BUT, you will have errored frames from crosstalk and reflections. You can check this by looking at the statistics on the switch, hub, whatever active device it may be. Using cables outside of spec will not show up as an obvious problem, but you will suffer performance consequences. Put a sniffer on the line and you will see frames with failed CRC, alignment errors, or even worse late collisions. I've seen it with happen with way too many networks (can you believe it! I actually get paid for telling people "dude, cabling is screwed and too long or your fiber is too weak, call me when you've fixed that"
This is simply the way ethernet works. It is the same with all cabling/network standards (SONET, 10Base-FL, FDDI, token ring, 1000Base-X, etc.) It will work, just not as well as it should.
Hope this clears things up, I'm just a stickler for details.
cheers
I don't see how you could possibly get late collisions from having too short of a cable. Late collisions happen when it takes longer to send the signal from one end of a segment to another than to put the entire packet on the network (this most often occurs in a situation where the cable run is too LONG).
Late collisions would happen on a short segment as a result of faulty connectors, damanged cables, bad NICs, etc . . . NOT because the cable was too short.
Also, reflection really has nothing to do with cable length. Reflection is typically caused by improperly terminated cable. The next most common cause is cable bent at too tight of a radius. For completeness, mixing cables with different impedance values can also cause reflections (this doesn't happen too often).
Late collisions would happen on a short segment as a result of faulty connectors, damanged cables, bad NICs, etc . . . NOT because the cable was too short.
Also, reflection really has nothing to do with cable length. Reflection is typically caused by improperly terminated cable. The next most common cause is cable bent at too tight of a radius. For completeness, mixing cables with different impedance values can also cause reflections (this doesn't happen too often).
college guy:
yes, i most certainly agree with you. Late collisions are normally a sign of cable too long. I was pointing out some of the errors that commonly accompany cables or a network "out of spec" short or long.
good point about the impendence mismatches. our entire campus (2000+ nodes) runs on IBM Type1 cable but yet work great. During our testing we painfuly discovered all about impendance, diffence between 100 and 150 ohms makes a HUGE difference. I've actually seen every frame flat out reflected back to the trasmiting port if this isnt taken into consideration and properly balanced. (ex, every bit you transmit, you receive...nasty. plays havoc with spanningtree...you'll actually get a link because the port is receiving its own link pulse even though there is no station attached)
Oh well, I'm still searching IEEE for the spec (not too much time though, this is getting ridiculous)
cheers!
edit - whoops forgot to mention that this is a switched 10/100/1000 ethernet network running on Type1 Cable, gigE is fiber. Squeeze every bit out of that cable investment man!
yes, i most certainly agree with you. Late collisions are normally a sign of cable too long. I was pointing out some of the errors that commonly accompany cables or a network "out of spec" short or long.
good point about the impendence mismatches. our entire campus (2000+ nodes) runs on IBM Type1 cable but yet work great. During our testing we painfuly discovered all about impendance, diffence between 100 and 150 ohms makes a HUGE difference. I've actually seen every frame flat out reflected back to the trasmiting port if this isnt taken into consideration and properly balanced. (ex, every bit you transmit, you receive...nasty. plays havoc with spanningtree...you'll actually get a link because the port is receiving its own link pulse even though there is no station attached)
Oh well, I'm still searching IEEE for the spec (not too much time though, this is getting ridiculous)
cheers!
edit - whoops forgot to mention that this is a switched 10/100/1000 ethernet network running on Type1 Cable, gigE is fiber. Squeeze every bit out of that cable investment man!
Ah, I see
Having cables with different impedances can be a frustrating problem to troubleshoot! That's usually the last thing you thinks about.
The IEEE spec is available on their web site, but you have to be a member to view it. I'm trying to find somewhere that has it posted for free to verify my claim
I assume you are using 1A cable though, correct? 1 is only rated to 16mbps if I remember correctly. It must have been a real b*tch to pull!
Having cables with different impedances can be a frustrating problem to troubleshoot! That's usually the last thing you thinks about.
The IEEE spec is available on their web site, but you have to be a member to view it. I'm trying to find somewhere that has it posted for free to verify my claim
I assume you are using 1A cable though, correct? 1 is only rated to 16mbps if I remember correctly. It must have been a real b*tch to pull!
college:
yeah, 1A (good to 350 Mhz i believe). Use AMP impedance adapters on either end of the type1 cable then patch both ends from there.
Yes, a 3 million dollar network infrastructure is running on type1 cable. Like I said, works great so far.
yeah, 1A (good to 350 Mhz i believe). Use AMP impedance adapters on either end of the type1 cable then patch both ends from there.
Yes, a 3 million dollar network infrastructure is running on type1 cable. Like I said, works great so far.
You should be commended for not pulling all of your hair out! Like you said though, it makes sense to squeeze everything you can out of your existing cable plant.
Somebody with some time and a good cable meter should do what I did a few years ago and find the shortest link with which they can can pass cat5 spec for crosstalk. I used a fluke dsp 2000 or something like that.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter