I saw this on a lighting circuit. A 2 pole breaker, in a 3 phase panel, feeding 2 120v circuits sharing one neutral. Knew the neutral wouldn't be 0 and it wasn't, it was less than either hot though. Anything wrong with this practice?

Nothing wrong with that at all. The remaining voltage is the direct result of the two legs not being 180 degrees out of time with each other.

In this case the neutral sees the geometric sum of the current in both phases. If only one phase of a 3 phase system were loaded the neutral would see the full phase current.

A simple rule anytime you run 1Ø off from 3Ø WYE system is that you have to have netural conductor the same size as phase conductor.

I don't excatally recall how I did that forumla it was a while seice I done that but corret me if I am wrong.,

( L1-L2 ) / N * 30°

I am not sure if that work right due the phase shift affect on it.

A true single phase MWBC will cancel netural current due 180° shift as long L1 - L2 load is the same the netrual current will read zero { not excactally but it will show near 0 on it }

Merci,Marc

Think of it this way: assume you have a balanced 3 phase wye load and the currents are 10 amps at 0 degrees, 120 degrees, and 240 degrees. Add the 10 amps at 0 and 120 degrees and you get 10 amps at 60 degrees, which is 180 degrees from the 240 degree load and that adds to 0 amps in the neutral. So the highest load on the neutral can not be higher than any of the phases. Nothing wrong with a shared neutral circuit.

Ichabod

A neutral like this will have less current than if it was a single phase off the 3-phase panel, but typically more than if it was in a residential 120/240. It's normal and OK, and will never exceed 100% ampacity. Well, not in normal circuimstances, at least...

One abnormal circumstance is low pf, leading on one phase and lagging on the other.

I must say, that's a new one! I was more referring to 3-phase 6-step power supplies, but I can your PF problem being especially bad on a circuit with two 208V legs. There's an easy solution, though: swap the legs if that occurs, and the currents cancel, problem solved!

But if the pf is very low, you are only reducing the problem, not eliminating it. And this one can be bad in those extreme cases with 180 degree legs at 120/240V. Fortunately, such cases are unheard of in residential situations.

In Europe larger companies are having all sorts of trouble with large numbers of switching mode power supplies on 3 phase supplies which can raise the neutral current above that of the largest phase current.

So maybe they need to distribute 127/220V or 133/230V or 139/240V around the data center, if they are unwilling to get the low harmonic PSUs, or are unwilling to upsize the neutral, or are unwilling to string up separate neutrals in each phase.

Or there needs to be PSUs designed for 400VAC (and 480VAC might be nice, too).

Of course we in North America can just connect our trashy PSUs line to line at 208V and push the neutral harmonics to the next transformer upstream after the first delta-wye (if the circulating currents in the delta primary isn't a problem). Hopefully by then it will be large enough that all those triplens will not be a significant total. REALLY large data centers won't have that option.

Newer PSUs are supposed to be much lower in harmonic current content. If manufacturer specs are to be believed, the good ones are lower. The cheap $3 ones from the middle kingdom could be a problem.

The main problem: reduced size neutrals in main feeders. They stopped doing that in data centers a few years ago and that pretty much solved the problem.

Same in the US. Neutrals are often sized to 200% in datacenters to deal with the harmonic current.
The problem goes beyond harmonic filtering; the real issue is that switched rectifying power supplies only draw power during the peak of each phase cycle. The current on the neutral is close to a 180Hz stepped wave, which means actual current levels exceeds traditional sinusoidal RMS by a considerable amount. The only surefire way to eliminate this is to use delta power, which eliminates the triplen harmonics, trading them for 5th, 7th & 13th, which are easier to filter out.

Way back in the olden days we avoided the whole mess in computer rooms by never even bringing a neutral into the panel. They used that bus for the IG and all loads were L/L