Proper way to install shielded ethernet

(big snip)

I first learned about the problems of grounding looking at a big rack full of electronic equipment, all with the ground connection on the power cords cut off...

Are you asking about EMI from ethernet to other equipment, or EMI from outside into ethernet?

Shielding and grounding is a complicated subject once you have more than one piece of equipment that is line powered. Electrically, the best way is to have only one point connected to earth ground, with everything else grounded through the shield. This often doesn't agree with electrical codes requiring equipment to be grounded through the power line.

Ethernet is transformer coupled, and is pretty immune to outside EMI. The possible exceptions are factories full of arc welders, or inside a building with a high powered radio transmitter. There are possible resonant situations where the shield length might be the appropriate multiple of a half wavelength for a nearby EMI source, or for the ethernet signal itself. In some cases a resistor between the shield and the rest of the system might be needed.

Without knowing the possible EMI sources, it is hard to say. Sound grounding theory says ground the cable to earth at only one point, everything else grounds through the cable shield. There can only be one grounding path between any two pieces of equipment, and that path is through the shield. RJ45 connectors are not very good at making ground contact, so you should connect from the shield to a screw connection on each end of the cable.

-- glen

it would be interesting to learn a bit about your environment that prompted the STP.... EMI from radio, radar, microwave, etc ?

I can't recall exactly all the wiring logistics when we installed the IBM Token Ring (STP) back in the 1980's and the grounding issues, etc...

All I know from general electronics is that floating grounds and ground loops are really bad...

future.)

locations, and

shielded

A few nuggets of information. Just nuggets, not hearty, satisfying pieces, since it's late and I was gonna sleep before I logged on... :-)

What propted the STP? Some speculation about the future, the present, and the past.

THE FUTURE I consulted with several cabling guys whose opinion I value quite a bit, which led in turn to my own online research. I found just enough information about 10 gig to become convinced that it makes sense to take 10 gig specs into consideration when evaluating the scope and breadth of a recabling job. The speculation comes into play because the standard isn't actually ratified yet, so it'll be jiggling for a while yet before it firms up. BUT, some cabling companies already sell cable that meets some version of the draft standard, and it's nearly always STP. Ampnetconnect has a number of interesting white papers on the subject, but they agree that STP clearly outperforms UTP in 10 gig testing. They may come up with good ways to deal with the ANEXT in the electronics, but for now, it's looking like STP has the advantage. If we can spend an extra 30-40% on the cabling job now, and that gives us another whole generation of network speeds in 10 years (or whenever it comes up), then we'll be very happy we spend the extra amount upfront.

THE PRESENT We've had some very odd application errors that have defied normal troubleshooting techniques. It's not the OS on the workstations, it's not the software environment on the workstations, it's not a gazillion other things that we've ruled out. The network is the one thing we haven't been able to provide with a conclusive alibi. Over the past 11 years, any number of people have poked around in our cieling for any number of reasons, and who knows if they all observed structured cabling standards in their poking? So my current cabling infrastructure is a great big question mark.

THE PAST Work that was done as recently as 2.5 years ago (just before I got there) is suspect. Jacks were terminated with only 2 pair, for example. I don't know anything about the work that was done in prior years, or, again, if structured cabling standards were followed (and 2.5 years ago, they certainly should have been punching down all 4 pair, imho). So again, my current cabling is a big question mark. Since I don't know if there ARE some freakish sources of EMI that are inducing network errors (chances are, not...but when all other possibilities have been ruled out, whatever remains, however unlikely, must be considered), and since STP seems a reasonable prediction for 10 gig applications, it seemed best to make no assumptions and play it extremely safe. You have no idea how irksome these bizzare application behaviours have been.

Also, anecdotally, I spoke to another local admin who recently rebuilt his network from the ground up with new cabling and switches, in preparation for a new VoIP phone system. Wouldn't you know it, the wierd behaviour some of his apps had been exhibiting, and which also defied resolution, suddenly disappeared. His switches had not been reporting errors, yet once everything was swapped out, his apps were suddenly rock-solid. Since his switches hadn't cried foul, he's assumed his network was clean and didn't give it another thought, so he was very suprised (and pleased) to find the issues resolved.

All this speculation led to a fuzzy-logic consensus that going STP on the rewire was worth a shot in the short run, and could likely pay off in the long run.

BJ

(oops...it's past my bedtime.)

preparation for

suddenly

this has been a great thread, as I've not seen STP since the old Token Ring days.

With the concept of EMI and various types of "noise" on the line, I was wondering is there any kind of "instrument" that could show exactly what the line was encountering over a period of time ? Especially if the switch is not showing ANY type of errors like CRC, etc. ie - what could the line momentarily experience that would not show up as an error ? hmmm - the circuit overloaded so it looks like a continuous collision ?