Path: utzoo!attcan!uunet!ginosko!uakari.primate.wisc.edu!ames!amdcad!rpw3 From: rpw3@amdcad.AMD.COM (Rob Warnock) Newsgroups: comp.dcom.lans Subject: Re: Grounding Ethernet coax lines. Message-ID: <27682@amdcad.AMD.COM> Date: 7 Oct 89 23:48:04 GMT References: <3824.25292578@zeus.unl.edu> <2128@avsd.UUCP> Reply-To: rpw3@amdcad.UUCP (Rob Warnock) Organization: [Consultant] San Mateo, CA Lines: 123 In article <2128@avsd.UUCP> childers@avsd.UUCP (Richard Childers) writes: +--------------- | network@zeus.unl.edu writes: | >I have heard that it is advisable to ground these [ethernet] lines. | ...[a lot of stuff]... | Now, no doubt I'll get snowed by zillions of corrections by EEs, but that's | about how I understand it and it seems to work. +--------------- Well, I'm sorry, you've just given not only wrong but dangerous advice. +--------------- | >1) Should a line be grounded only on one end, or on both ends? | Both ends. +--------------- Wrong. A given Ethernet (thick or thin) *must* be grounded *EXACTLY ONCE*, no more, no less. To do other wise can *increase* your noise problems or even -- are you ready for this? -- melt your cable! Remember, the original question was talking about between buildings. I have seen situations where there are MANY [is >10 many enough?] volts difference between two adjacent buildings, and if you ground an Ethernet at both ends you have just attempted to short out that potential difference. Now since such differences generally come from different neutral-line currents in *heavy* equipments [elevators, big air conditioners, etc.], that's a very low impedance voltage you're trying to short out, and you can't. What you will do instead is simply create a current in the shield of your Ethernet of magnitude I = E/R, where R is the resistance of the shield. Oh, the shield is a very low resistance? Well then I guess you're gonna get a *very* high current in your shield! Try 10's of amperes (or more!)... In any case, not very good for your Ethernet. Even if it doesn't melt it, it will induce large voltage differences between the center conductor and the shield [remember that the senter conductor is "grounded", as it were, only by the termination resistors at the ends], and this can make noise margins worse, or even make reception impossible, or even destroy transceivers. +--------------- | > Should it be grounded at all? | Yes. To the machines using the cable, only. +--------------- *NO!* *NO!* *NEVER* to the machines using the cable! Most of the cost of an Ethernet transceiver goes to keeping a very high isolation between the controller side and the cable side. Even with thin Ethernet, notice that little ring of ceramic surrounding the BNC jack? That's an INSULATOR! And it's there for a reason. Ethernet cables should be grounded once, preferably *VERY* close to a good earth ground suitable for lightening protection. That is, low resistance, low inductance, and as short as possible. While that will not protect against a *direct* lightening strike [almost nothing will], the most common lightening damage comes from *nearby* strikes which induce large voltages in ungrounded metal objects. Oh, and that's another reason not to ground a cable more than once. Nearby lightening also makes various parts of the ground be something other than "ground" for a while, and various "grounds" temporarily become very different voltage references from each other. If your cable is grounded more than once, it will try to conduct current to "short" the various grounds together. It will lose. [See the above discussion of grounds between buildings.] +--------------- | >2) What about in line taps that go to devices such as a repeater | > or bridge. Meter checks show that this makes a solid ground, | > so does that count as one of the grounds? In these cases do | > we need to add any ground at all? | You shouldn't. The outside of the connector is ground. +--------------- You know, that actually brings up a good question. Repeaters SHOULDN'T ground the cable on either side, but *a* ground for each thin Ethernet must come from *somewhere*, and I doubt that most network installers think to provide one. [It may be the case that these multi-thinwire repeaters ground the "down leg" thinwires to their own chassis ground, which would be the right thing to do in most cases assuming the chassis ground is connected to the third or "green" wire, but they should never ground the "up leg", i.e., the trunk or backbone connection.] +--------------- | >3) BNC Tee connections to a Ethernet card in a PC seem to make a | > very weak ground or no ground at all. Should we ground these | > connections? | They should be adequate, as designed. +--------------- Right. "As designed" there is *NO* ground! Which is correct! +--------------- | >4) Is it advisable to isolate against all unintentional grounds | > as with a connection touching a metal rack by wrapping the | > metal part of the connection? | Preventing accidental grounding can rarely hurt, and almost always helps, | as a preventative measure. +--------------- Always helps. Failure to do so leads to "ground loops", which can lead to the problems enumerated above. +--------------- | >Suggestions and comments are appreciated. | It wouldn't hurt to go find a book on antenna theory and / or ground loops, | if you're interested in getting into ham radio or are really dedicated to | network management, as I have heard ethernet being explained successfully | as an insulated antenna ... +--------------- Sounds like you need such a book, yourself. ;-} Sorry to be so harse, but you gave some really bad advice there... Rob Warnock Systems Architecture Consultant UUCP: {amdcad,sun}!redwood!rpw3 DDD: (415)572-2607 USPS: 627 26th Ave, San Mateo, CA 94403