Network wiring problem - weird one!

[teddysnips]

Awesome. Just what I wanted to know!

Thanks

Edward

[Ron Lowe]

Um, what's a network switch?

Edward

------------

A switch is what's already inside your router. A device to connect 4 network devices together ( 5 if you include the router itself. )

There's nothing wrong with your network layout: it's ideal. You just need to sort out the botched wiring.

[Tim S]

[e-mail] coughed up some electrons that declared:

You'd think so and you'd be right if it were DC signals involved.

However, you're shoving signals with componets in the 100's MHz or GHz range down the wires. Each circuit path must be down a twisted pair for balance and signal immunity, or you'll get complete rubbish out the end, to the point where the equipment won't even give you a link-active light.

Which sounds like the problem you're describing.

You mate with the cheap more-or-less DC buzz-through tool (which is what your description sounds like) proved the pin assignments are correct. Those tools do not prove the cable is wired properly. Proper test kit costs

1000's and does time domain reflectrometry and/or other impressive sounding tests to prove it's right.

As others have said, best get in and check the wire colours. Use a good light - the brown and orange, and the blue and greens are had to tell apart in poor light.

Cheers

Tim

[teddysnips]

Yo! Pear cider! My man Kevin Minchew (well, he's his own man, actually) brews totally awesome pear cider. Single variety, total headf*uck.

Anyway, thanks for the learned exegesis about switches, routers and hubs. I take the point that I've probably over-engineered the damn thing, but I've bought the router now so I'm stuck with it. But it looks very much as if other equally clever bods up-thread have identified the problem (incorrect wiring) so I'll let y'all know how it goes.

Ain't Usenet wonderful!

Edward

[The Natural Philosopher]

Completely WRONG

see here for a very good chart.

& 2 are a pair, 3 and 6 are a pair, 4 and 5 are a pair, and 7 ad 8 are a pair.

Generally it doesn't matter if the pairs are wire arsy versy..as long as the pairs are gong to the correct pins.

The most common faults I have found are badly terminated sockets..the IDC is a bugger to get right without a proper tool, and the next problem that has happened has been wrong grade of socket/plug. That simply doesn't make proper contact.

Ethernet normally uses 3 and 6, and 4 and 5. We used to use the other two pairs for telephony

Thats true. Whats inside may not actually be 12345678 in order.

NO, use the ones in the wiki article: they are correct. Either scheme works, as long as its the same both ends.

No, what the standrd is. You are talking bollox ;-)

Strange that you have managed to get it wrong every time, then.

>

[John Rumm]

No, what you have done it the right way to do it. Had you have taken the wires that connect to the individual sockets straight to the router then you would have needed to terminate each of them in RJ45 plugs. This is not only slow and tedious (much simpler and quicker buying ready made patch leads), but also requires that you take solid core cat5 into the plugs - which while it can be done is not ultra reliable and technically speaking requires a different type of plug.

For bigger posher installs, then you might use a patch panel in place of the 4 way, and a multiport switch in place of the router. That all works nicely when its in a comms rack along with the PBX - then you can patch voice and data to where it is needed, and change it later if required just by re patching.

Yup, when working with signals that are sent as a differential pair, it is vital to make sure they are carried by a twisted pair. Just having the right pins joined together is not enough at these frequencies.

tis indeed! (mostly)

[Tim S]

[e-mail] coughed up some electrons that declared:

It was some foreign stuff (Swedish I think), but at least the brain damage is temporary.

Not quite - you've not wasted your money. You need the router (usually part of an ADSL modem) to get anywhere.

The switch is optional, but does mean that you can get away with running 1 cable half way round your house instead of several, if your layout demands it.

Cheers

Tim

[Ron Lowe]

Yes, Yes.

Obvious Typo. Se my previous posts.

1+2, 3+6. Well spotted.

But my main point was that the COLOUR of the pair doesn't actually matter. What matters is that the pairs are maintained.

1+2, 3+6.

[Ron Lowe]

Demonstrably facually incorrect. Not 'every time', just on that post, as I accept.

My original posts were correct. I'll re-post it just for the avoidance of doubt.

I'd expect it to be:

1 Orange / White 2 Solid Orange 3 Green / White 4 Solid Blue 5 Blue / White 6 Solid Green 7 Brown / White 8 Solid Brown

( The Orange twisted pair may be swapped withthe Green twisted pair on some installations, that's OK )

What's important it the twisted pairs. One pair ( usually orange ) is used for pins 1+2. Another pair ( usually green ) is used for pins 3+6. This is critical. The blue pair and brown pair are only used on gigabit and / or PoE systems, but these are becoming commonplace, and should be wired up correctly too.

[Mike Clarke]

The DG834 has "Auto Uplink" ports which means that it eliminates the need to worry about crossover cables. It will accommodate either type of cable and make the appropriate internal adjustments automatically to make the right connection.

As others have said, it looks like you just need to get the fixed wiring sorted to ensure the wires are paired off in the right order.

[Dave Liquorice]

You better recount your pin numbers, I'm intriged by the common use of pin

3 between two pairs. So on balance I think it's probably best if the OP does not "do what we say". B-)

Looking at the picture of the socket back it does appear that the grn/wht and org/wht wires are transposed. Look at the colour coding blobs half hidden by the wires. It's normal to have adjacent terminals to be a pair not have them split apart. Having them apart plays havoc with the impedance an possibly the balance as well.

[Dave Plowman (News)]

Perhaps it came from the same place as the 'crimp tool' he used?

[Tim S]

Dave Plowman (News) coughed up some electrons that declared:

Silly thing is, TDR type kit would be total overkill for a few wires in a small office or residence. It's very useful when you've laid in 400 cables and you want to prove they're all good for gig (or even 10gig) so you don;t spend the next n-years tracking down obscure computer faults that relate to dodgey cabling.

For simple installations, paying attention to the wire colours and workmanship on the punchdowns is more than adequate to get 1000baseT reliably without much sweat. For final testing, I patch a laptop through to a switch where both can do gig then run a bandwidth test tool (ie flood the link for a few minutes) then look for any errors at the interface level - poor mans test, but at least it's realistic.

Perhaps over-reliance on a test tool in the belief it was testing everything?

Cheers

Tim

[<me9]

It shouldn't matter with reasonably up to date (less than 5yo) hardware, it auto senses.

[<me9]

As long as both ends of the /fixed/ wiring use tthe same standards it shouldn't matter.

[Dave Liquorice]

Probably or lack of understanding of the test tools limitations. A simple DC loop tester will check for correct "pairs" and a single or multiple

*different* wiring errors would be detected. However it won't pickup the *same* wiring error at each end of the cable, that does require far more sophisticated kit to detect.

[Andy Wade]

It's also important to maintain the twisting of each pair right up as close as possible to the IDC terminals of the socket. The 'as installed' picture provided didn't show particularly good practice in this respect.

Now can I ask a supplementary Ethernet question about switches while all the networking experts are around? Can you cascade the cheap unmanaged switches /ad-infinitum/ and maintain communication between all nodes? For example say two ports from a typical 4-port ADSL router, R, each feed remote 8-port switches, A & B. Each switch provides 7 ports used to connect local network devices. In this network will a device on switch A be able to communicate with a device on switch B, via the switch in R?

[John Rumm]

Indeed. You will usually get away with it at 10Mb, and often at 100Mbit, but don't expect gigabit to like it!

With switches - yes, there are no longer any practical limits on chaining segments. With older style hubs there was a limit as they acted as simple repeaters, and kept the whole network as a single collision detection zone. Hence it also had to all run at the same speed. Switches are smarter. While initially they will behave a little like hubs (in the limited sense that they will forward all packets to all ports), they will learn which MAC addresses are on which ports and stop forwarding traffic to segments unnecessarily.

Yup, that would be fine.

[Bob Mannix]

One obvious answer is that you have patch leads with cross-overs in them. Assuming you haven't checked this, get two leads that work connecting the laptop into the router and use one of these as a patch lead. If you have already tried this or it still fails, maybe there's a crossover in the installed cables - try a patch lead with a crossover! Otherwise, in spite of his test, the cabling isn't correct.

[dennis]

Within certain limits yes. Each switch will add a small delay. If the total delay exceeds the protocol timeout you are using it will fail even though the packets are being delivered. In practice it is a lot of switches. I dimension the UK for switched local access once and had it put to the board at BT. It was to give 100M access to every home (and 1G if paid for) but they didn't like it as it wasn't fibre and "the press would crucify us". I won't tell you the estimated cost but it wasn't as expensive as the £28 billion talked about in the press by a long way.

They did steal some of my ideas though..

like putting a wireless receiver as the termination node of the wires and then running a public wireless network from it. I wouldn't be surprised if they make small water proof network switches and drop them down the access holes either as that was what was intended.. string together as many as needed to get to the end point with 100M TP teed off as you went.

The best one was how to power it.. obvious really just get the power from the customers node, its copper so getting a few watts was easy.