Problems with dual Athlons (power; popping breakers)

Robert G. Brown rgb at
Thu Aug 1 06:42:40 PDT 2002

On Wed, 31 Jul 2002, Michael Stein wrote:

> > One thing that they did that had me a bit concerned from the beginning
> > was wire the power poles by pulling all three phases to a power pole and
> > coming back with a single common ground.
> Is that 'ground' (grounding conductor?) or 'neutral' (grounded conductor?)?

Sorry, single common neutral.

> > Of course if you have three equal, purely resistive loads on all three
> > hot wires you could in principle grab the common cold wire in your hand
> > as it would carry no current (Kids! Do Not try this at home!)
> Power factor correction (PFC) power supplies should be very close to
> looking like a resistive loads and the currents in the neutral should
> cancel out (assuming balanced loads on the three phases).
> > but in practice it is not at all clear that the load currents of a
> > cluster will all be in phase with the current (or even shifted by a
> > common amount).
> Typical (?) PC power supplies are not PFC and draw a big pulse of current
> on each rising peak (120 times a second for 60 Hz power).  If you have
> a bunch (cluster) of PCs spread over three phases, the pulse from each
> phase doesn't occur at the same time since the phases are 120 degrees
> apart so these pulses can't cancel out in the neutral.

Define "big pulse".  If a supply draws order of 1 A rms average, is the
peak draw 2 A?  4 A?  Depending on the response time of the circuit
breakers, this could explain why they pop well below the rms power
capacity presumed from their current rating.

> > In practice, our racks were popping breakers before they reached 75%
> > of nominal load, making me wonder if they actually had three DIFFERENT
> > phases sharing a ground...
> Breakers should only be in the "hot" lines not in the neutral.  So this
> overload of the neutral shouldn't pop the breaker.  Rather the neutral
> should melt and start a fire...
> I'd worry why you are popping breakers and keep looking for some other
> problem...  And worry about the neutral too.

I'm doing exactly those two things.  They have four circuits per power
pole, for example, not three, but only three phases.  I'm very curious
as to how they are sharing the neutral, as just a silly little mistake
would have two circuits with the same phase sharing the neutral.  Code
is conservative enough, and the run short enough, and the room cold
enough, that this might not start a fire, but we'll all feel better when
they rewire it with a neutral per circuit.

> > Are there published standards for wiring clusters (or similar
> > environments)?  We will soon be having them redo the power pole wiring
> > so each line has its own ground.  Is there anything (relatively
> > inexpensive) we can add or should insist on at that point to condition
> > the power and isolate the harmonics so that they don't bleed through
> > from system to system?  I think dual isolation transformers are beyond
> > our means, but could a high-quality surge strip (perhaps one with
> > limited UPS capacity) serve the same purpose?
> One possiblity I've thought of but haven't tried would be to switch the
> PC power supplies to the 240 setting and run them off of 208 (between
> two phases of the three phases).
> You'll need some special power strips (probably a powerstrip with IEC
> 320 C-13 outlets) and some C-13 to C-14 power cords for the machines.
> The pulses will still exist but won't be going into the neutral (since
> it's not connected).
> I haven't tried this, but I think that some large rack mount systems
> do this...

This is a bit invasive for our setting and we need to keep everything
compliant with code for all possible reasons -- I do some of my own
wiring at home, but Duke requires licensed electricians.  We could have
electricians rewire the poles for 209V and do as you suggest (expensive)
but first I'm just going to give in and a) put an oscilloscope on the
lines; b) put a poor man's GFCI (one I wire out of parts downstairs:-)
inline with a strip and plug in a segment of the cluster to test for
current imbalance.  If the supplies are all drawing a really big pulse
of current in phase and dumping it on neutral in phase, it could be (as
Don suggested offline) deforming the waveforms and dropping the de facto
line voltages at peak to where the push the supplies out of spec.

If the current draw is large enough to drop line voltages on the hot
line (which it does by drawing enough current so that IR on the supply
line is no longer negligible and because of inductive lag in
everything), it is also large enough to create a pulse of line voltage
on the neutral (the same IR in the other direction), which in turn
creates interesting possibilities of a transient ground loop -- too
short to show up on a cheap multimeter, but more than enough to create
all sorts of problems.

Very interesting.  Time to go in and scrounge a scope from my
experimentalist friends...


Robert G. Brown	             
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at

More information about the Beowulf mailing list