Mystery cabling problem

[furball]

Ok, here's a good one for the teccy types.

Yesterday I had a call-out to a shop where one of the tills wasn't connecting to the network (XP PC based system). I tested all the cables and found no faults, I was getting a good link to the switch at the till end and yet the till wouldn't connect.

I figured it must be the till, but my tester could see and talk to it's NIC so that looked OK too.

Next I started experimenting with the till next to it, using a 20m patch lead to see what happened when I plugged till 1 into till 2's data point etc. I eventually concluded that it must be the patch lead that was causing the problem, even though it tested OK, so to prove it I connected the problem till to it's own data point with the 20m patch lead and it worked fine.

Concluding that I had resolved the problem I went and got a brand new patch lead out of the van and.... nothing, no connection. I thought maybe I had a dodgy lead so I took the one off the neighbouring till (a 'known good working' lead) and still nothing. So I plugged the 20m lead back in and it connected immediately.

So the question is this, why would a 20m lead work when a 2m one won't? The only potential clue I have is that some of the wiring for the security tagging system runs up the same pole into the ceiling as the cable for the data point for this till, and every time I plugged/unplugged a patch lead it set the tagging alarm off (which is odd, I assume there is also a problem with the tagging system, but I don't know anything about it so I didn't start messing with it).

[Alchamist]

Stupid question, but were you connecting the patch lead to the same data point (assuming a wall patch point), or directly to the hub?

Also, are there mains supplies going up the same pole? If so then it might be picking up some interference (assuming you are using cat5e patch leads, these can be susceptible to interference from mains, especially on poorly shielded stuff).

And finally ... as has happened before ... on the 20m lead, did it auto-negotiate the speed down to 10Mb/s rather than 100Mb/s?

[james'll fix it]

tagging cable should be ok, if the controller for de-tagger is remote then only data runs down the cable. the de-tagger on the till will emit interference, down aswell as up so make sure nothing too close.
what alarm went off?
there's no connection from the detectors on the door to the till.

[furball]

Stupid question, but were you connecting the patch lead to the same data point (assuming a wall patch point), or directly to the hub?

Also, are there mains supplies going up the same pole? If so then it might be picking up some interference (assuming you are using cat5e patch leads, these can be susceptible to interference from mains, especially on poorly shielded stuff).

And finally ... as has happened before ... on the 20m lead, did it auto-negotiate the speed down to 10Mb/s rather than 100Mb/s?

It was a data point screwed to the woodwork for the till point. The power feeds all go up a separate pole into the ceiling (this particular customer may be a cheapskate but we still have our standards). The two patch leads were both standard excel cat 5e, and my Fluke said there was a full duplex 100 meg connection. That's one of the things that confused me - the Fluke said the connection was fine but the till didn't want to know (cheap NIC).

tagging cable should be ok, if the controller for de-tagger is remote then only data runs down the cable. the de-tagger on the till will emit interference, down aswell as up so make sure nothing too close.
what alarm went off?
there's no connection from the detectors on the door to the till.

I'm thinking the tagging alarm may be a red herring - a problem with the tagging system rather than the network. It uses old fashioned BNC so probably leaks interferance all over the place. It was the main box on the wall rather than the little one mounted by the till that was going off.

I'm wondering if the reason the longer cable worked is if it had some sort of aerial like effect. A case of 'my interferance is bigger than your interferance'

[Alchamist]

It's quite possible ... and the fluke might show 100mb, but when into the hub it might drop it down due to the extra distance. I have had similar before where it would drop to 10mb and work fine, but when running at 100mb there was data, but it was that corrupted that it wouldn't work

[Sir-Doris]

Its not a crossover cable is it? Not all network adapters auto detect crossover cables.

[furball]

It's quite possible ... and the fluke might show 100mb, but when into the hub it might drop it down due to the extra distance. I have had similar before where it would drop to 10mb and work fine, but when running at 100mb there was data, but it was that corrupted that it wouldn't work

Sorry, shoulda said, I had the fluke plugged in at the till end as well as the switch. It's quite a small store so I would think even with the 20m patch lead it's probably no more than a 70m run, so well within tolerances.

It's also a fault that has only recently happened and the store has been trading OK for a good while so I'm guessing they've had some work done on something else that has cabling in the ceiling and something has been damaged or moved. Might even be something as simple as a trapped cable (cat5 being a bit temperamental when it comes to that).