Consider the following scenario: Meter is on the outside of a house. RMC goes through the wall into an auxiliary gutter. Inside the gutter Polaris connectors are used to split the service three ways. The gutter then nipples into three main-breaker panels.

Where is the correct place to bond the neutral to ground?

All metal parts of the service must be bonded to the neutral, so you'll have to bond the gutter with a jumper to the neutrals. Each metal pipe nipple must be bonded (at one end), either with a grounding locknut that does not encouter any remaing cocentric/eccentric knockouts or with a grounding bushing with the proper sized jumper . Finally, each panel must be bonded, generally using the screw or wire provided by the manufacturer.

In short, everything metal on the line side of the service disconnecting means must be bonded together, to the gounded service conductor, the grounding electrode system, and the feeder equipment grounding conductors.

Keep in mind that 250.6 must be complied with which may require an alteration of the bonding and gronding methods used.

The code is inconsistent with how it handles the flow of electrons on the line side of the service verses the load side of the service. It's on of the many aspects of Article 250 I have never been a big fan of.

In general, the ground / neutral bond is made at the first overcurrent device. So, if the meter has a main breaker (or fuse) as a disconnect, then that's where the two are joined. If there is simply a disconnect - no breaker until you get to the panels ... then the bond is made at those panels.

In general, the ground / neutral bond is made at the first overcurrent device. So, if the meter has a main breaker (or fuse) as a disconnect, then that's where the two are joined. If there is simply a disconnect - no breaker until you get to the panels ... then the bond is made at those panels.

Finally, let's not get confused about that little green screw. All it does is bond the box to the neutral bar. If you have a separate ground buss, you really ought to have a wire connecting the busses together. I believe the 08 actually says this ... but it seems logical; I wouldn't expect that screw's connection to be reliable enough to use it as the only means of carrying the full fault current of a circuit.

So if there are three panels, then you bond the neutral in three places? That would seem to be a violation of 250.6.

From a strictly engineering perspective, what would make the most sense from a "flow of electrons" point of view is to have ground/neutral splice point or bus in the gutter to which you connect the incoming neutral, the neutrals to all the panels, the grounding electrode conductor, bonding wires that go the the ground buses in the panels, and any miscellaneous bonding wires for the gutter & c. But I'm not sure that would be allowed in an auxiliary gutter.

It may seem contrary to the usual practice ... but absent a main breaker for all three panels, you have three 'main' panels, as each is the 'first point' of connection to the utility service.

Now, running the GEC to the ground electrode is another matter. I can see the argument that you need to run three GEC's - one from each panel - to the electrode ..... as well as the counter that one building only needs one GEC.

This is the picture from the 2002 handbook

http://esteroriverheights.com/electrical/gec_tap.jpg

or this

http://esteroriverheights.com/electrical/gec_tap2.jpg

Greg, thanks for the pics ... seems I've been 'close, but not perfect' in my practices.

In that example you will have substantial current flowing in nipples and the metal of the trough. That doesn't seem like it should be correct.

The more I look at that picture (gec_tap.jpg), the more I'm scratching my head. It appears that 250.24(B) in fact requires a ground-neutral bond in each service disconnect enclosure. However, the installation illustrated gec_tap.jpg will have as much as 50% or more of the neutral current flowing in the steel of the nipples and the cabinets. In fact, if there is poor contact at, say, the connection of a nipple to the gutter, it is possible that that point will get dangerously hot.

In any case, it is clearly wrong to have any of the service current, let alone a substantial part of it, flowing through the metal of the raceway. Yet that appears to be what the Code requires.

These are from NFPA so that is what they are thinking. I agree the whole cludge around the meter to service disconnect(s) is one big conductor but they stopped worrying about that sometime around 1999 or 2002. They are specific that you can't reground the neutral on the load side of the service disconnect but they are silent on the line side. The utility will ground it in the meter base if nothing else.
For that matter, you also carry current on the ground electrode system too. We just don't like to think about it.

I guess that's the only advantage of the European TT system - no ground loops. The PoCo grounds the neutral at the substation and that's it, local grounds are only connected to a ground rod. Of course the higher impedance of such a system requires additional ground fault protection, for example 100mA RCDs.

Wait, WHAT? The neutral bar in the gutter makes sense, but I really disagree with separately bonding the neuetral bars in the panels to the enclosure. Neutral current flowing through the enclosures and conduit is completely unsat.

250.24(C) is pretty clear that this is the way NFPA wants to see it, though. The only exception is if all the service disconnects are in the same enclosure.

FYI, NEC 2008 handbook retains gec_tap2.jpg, but does not include gex_tap.jpg.

Thinking more about this, I think I see why this works OK. (And I'm intentionally saying "OK", rather than "well." It's only just, "OK.")

The key is that the white neutral wires going from the neutral bar in the gutter to the neutral bars in the panels are very low impedance paths. That establishes an "upper bound" on the voltage difference between the three neutral bars show, and that upper bound is well under one volt.

All the additional bonding serves to reduce the impedance, making that voltage differential even smaller.

So even though this whole thing is a mess, the maximum voltage differential that results from the mess is substantially less than one volt.

And "substantially less than one volt" is small enough that it doesn't result in any serious problems in practice.

Still, I would much rather see separate neutral and ground bars in the panels, fed by separate neutral and ground wires going to neutral bar in the gutter.