


> I understand the common switching power supply for computers switches
> on the current when the voltage is high enough to replenish the tank it
> keeps. So it would see the smaller the computer load relative to the
> power supply capacity, the narrower the current pulses would be. Once
> they are smaller than 50%, you could be going past the 200% mark on the
> neutral, if all your load is these power supplies. For a few computers
> in an office, maybe that would never happen. But for building a large
> data center, I suppose it could get serious.

 So you maintain that, as the load drawn from the computer power supply
 DECREASES, the neutral load INCREASES.
No.
As the computer power supply CAPACITY increases over its ACTUAL LOAD, the
period of time needed to refill capacitors back to full charge is a SMALLER
portion of each AC cycle.
 Lets go with that. I'm open minded.

 What is magic about the 50% current pulses? Why would a <50% pulse result in
 200% neutral current?

 Not arguing your point, just looking for an understanding of the theory
 behind it.
It is easier to illustrate with 33.33% current pulses. Given 3 computers,
each on a different 120 volt leg (phase) of a 208Y/120 volt three phase
service, And assuming identical characterists of these 3 computers (for
simplicity of illustration), the current waveforms might look like:
hot leg A:
__ __ __
/ \________ ________/ \________ ________/ \________ ________
\__/ \__/ \__/
hot leg B:
__ __ __
________/ \________ ________/ \________ ________/ \________
\__/ \__/ \__/
hot leg C:
__ __ __
____ ________/ \________ ________/ \________ ________/ \____
\__/ \__/ \__/
neutral leg (opposite polarity of hot legs):
__ __ __ __ __ __ __ __ __
/ \ / \
/ \ / \
/ \ / \
/ \ / \
/ \
\__/ \__/ \__/ \__/ \__/ \__/ \__/ \__/ \__/
If there are sufficient number of computers on each phase to draw an
average of 30 amps on each phase, then at 33.33% duty cycle, the pulses
would peak at 90 amps. But the average is still 30 amps, so the load
is effectively 30 amps in terms of energy usage, heating, etc. However,
the neutral wire has all 3 sets of 90 amp pulses overlaid on it, so its
overall load is that 90 amps. And it is at 180 Hz (150 Hz in Europe).
Were these loads linear and had sine wave currents, then where the meet
at a common neutral, there would at any instantaneous time be a balance
of all 3 currents, adding positive and negative values, to equal zero.
The worse case in linear loads is when only one hot leg has a load and
so the neutral would have the same load.
But with the nonlinear load, the timings don't allow summing to zero at
instantaneous times in the worst case of 33.33% pulse times or less.
I did say earlier that 33.33% pulse times would be easier to illustrate.
If the times were instead 50%, there would be SOME overlap, and thus
SOME amount of current summation to zero (or current between various taps
on the neutral bus).
The severity depends on just how much of the load is this nonlinear load
and just how nonlinear it is.
A computer power supply that is lightly loaded (for example a 400 watt P/S
unit serving a board that uses 50 watts) will have very little draw from
it's reserve. Thus it needs less time during the cycle peaks to replenish
that reserve.
The concern isn't that the load is low, but rather, that the P/S capacity
being higher exacerbates the nonlinearity at that (constant) load level.
The comparison here would be between a computer board using 50 watts from
a 400 watt capacity power supply, vs, the same computer board using the
same 50 watts, but from a 100 watt capacity power supply.
When enough of those 50 watt loads are piled up on a wye configured circuit
that approachs the configured limit, say for example an average of 100 amps
(300 amp peaks for 33.33% of the time), the neutral has the other polarity
for all 3 circuits for a total of 300 amps (300 amp peaks, but filled in at
99.99% of the time). If the neutral is rated for 100 amps, it's now in a
300% overload condition.
This is the worst case I'm referring to.
Real life cases perhaps don't typically involve such radical departure of
power supply capacity from computer actual usage. Real life cases don't
involve all the load being computers (unless it is a data center).
But lots of these power supplies readily operate up to 240 volts and I am
wondering if that extra voltage capacity means even shorter current times,
making the problem worse for those with such standard voltage.
Note, if running on single phase, 2 computers at opposite poles will have
their current pulses in timed opposition (to the extent that they are
identical) and their summation will be zero on single phase. It is three
phase power systems that have this problem.
200% rated neutrals probably handles most every situation adequately.
But there could be situations approaching 300% in rare cases.


 Phil Howard KA9WGN  http://linuxhomepage.com/ http://ham.org/ 
Click to see the full signature.