Ethernet rewiring/testing

Sometimes, bad cables or connections can partly work, generating loads of retransmits in the protocol stack and low performance. Usually, it's so bad you quickly notice.

Can be useful if something isn't working, or when you first made up the cables. Probably of little value if everything seems to be working.

The socket IDC connectors will have been designed for solid core. If it's working, all well and good. Probably more likely to stop working at some point than solid core would have been.

Probably not so usefule now, but I made up a cheap and cheerful continuity tester (which said nothing about the quality of a connection but told me it was OK).

I used four bidirectional LEDs (OK, eight normal LEDs and four diodes) connected to a patch lead at one end. And battery and four switches at the other. This told me of open circuit connections as well as pair reversals/miswirings.

As I said, cheap and cheerful. But it did catch what problems there were.

Lobster wibbled on Sunday 08 August 2010 15:31

Do you have 2 gig capable devices? If so, test each cable with those. If you've got the 1st 2 pairs right but mess up on the others, it will work but it will fall back to 100 meg.

If you've got them all "right" as in the end points connected *mostly* correctly (there are still ways to do it wrong like splitting a "pair" across 2 logical circuits of the 4) then the devices should negotiate gig. Gig needs all 4 pairs, 10/100 uses 2.

Or you could buy a cheap end-end tester - this just does a continuity buzz through to show basic connectivity but cannot tell you if you split the pairs between circuits or have a bad connection.

If you were doing this commercially, you would now pop out your TDR (time domain reflectometer) tester and certify it. However, 2 things for diy:

If you punched stranded wire (the patch cable generally will be) into sockets designed for solid wire (ie installation cable) you may get problems developing down the line even if it works now. However, it could work forever...

HTH

Tim

In general a continuity tester that plugs into each socket and shorts the various pairs and a LED plus battery thing that plugs in the other end to see that continuity exists, is enough for 99% of all basic faults.

There is one common wiring error commonly found on DIY installs which a continuity tester will not pick up... Split pairs. The twisted pairs are critical to the correct operation.

A common error is to wire it like this:

pins 1+2 - pair1 pins 3+4 - pair2 pins 5+6 - pair3 pins 7+8 - pair4

This may work OK at 10Mbit, and perhaps at 100 over short runs, but will fail miserably over any length.

The correct wiring is:

pins 1+2 - pair 1 ( usually white/orange + orange ) pins 3+6 - pair 2 ( usually white/green + green ) pins 4+5 - pair 3 ( usually blue + white/blue ) pins 7+8 - pair 4 ( usually white/brown + brown )

The actual colours of the pairs is unimportant ( electrons are colour-blind ) but it is critical that the pairing is observed. The wiring I have described is the most commonly used T568B.

Yes...the continuity tester I made detected split pairs, and also crossed over pairs....and I found at least one of each!

In message , Lobster writes

I bought one of these some time ago (when the postage was cheaper!) and it has been marvellous for finding damaged cables in a friend's office. It is cheaply made but it does the job. For us it made it quick and simple to identify which old unlabelled cable went where and then which had been broken by the decorators.

Your tester would have identified crossed pairs with ease, but it must have been more sophisticated than a simple dc continuity tester to pick up split pairs. How did it work?

Not very sophisticated really. At one end, a patch cable into a box. Each pair connected to two LEDs (red and green) fixed opposite ways round.

At the other end, a battery, one side connected to one side of each of the four pairs. Other side connected to other side of corresponding pair via a switch (four switches).

Connect the two boxes, have a helper at the other end in earshot (or on a mobile, for an isolated part of the house). Flip each switch in turn, getting one, two, three then all four on. Then off the same way. Green LEDs should light in turn. Crossed pairs fail to light LED, reversed pairs light the red LED.

At least, that's as I remember it! Haven't opened the boxes recently...

Toolstation manage to knock a quid off Maplins price for walk in convenience.

Cheers Adam

The device in the ebay ad sequences through the wires, so it doesn't need the second person at the far end. I think I paid a total of either £2.98 or £3.98 by choosing the right Chinaman.

I made this 15 years ago...it was the cheapest solution then.

"Lobster" wrote

Not sure how reliable it is, but if you have 2 1Gb enabled laptops, I would expect the network connection to report its speed in the task bar. Usually get something like "Local Area Network Connected at 1Gb".

Phil

How? I don't see how a simple continuity tester (which is all it is) can show up split pairs. A split pair will still be connected at each end but by the wrong bit of copper.

My interpretation of a split pair is the situation where the pins at each end are paired off correctly but not connected by wires twisted in the same pair, e.g. pins 3+6 joined by green and blue wires instead of white/green and green.

None of these cheap testers will detect this type of fault for which you need expensive test kit costing a thousand pounds or more. This is only worthwhile for anyone doing large scale testing. The best way to check for this type of fault on a domestic setup is to rely on thorough visual examination of the terminations.

Opinion endorsed.

Actually, you get degradation at even 10Mpbs.

Correct.

No simple DC test will pick this up. Only a fancy tester or carefull visual inspection.

Or, of course, carefull competent installation in the first place!

That's just a simple continuity test, surely?

So long as:

That's not good enough.

Correct.