 I'm finding this fascinating as I have just enough background in this
 stuff to be dangerous without being helpful...

 so please forgive what may be a dumb question to you both:

 In the case of the 300 watt supply provviding 100 watts
 of power, the diagram, as I read it, demonstrated some
 hefty power peaks. BUT they only seemed to be for
 one third the time, with the other 2/3rds showing nothing.

 Did I interpret it correctly? If so, shouldn't the wiring
 capacity be rated for the averaged out load rather than
 the short term (fractions of seconds) peaks?

 Or am I completely off here... (wouldn't be the first time)
This is not a dumb question. It's a perfectly valid question.
Yes, the pulses are on only 1/3 of the time. Switching power
supplies can do that depending on load.
If you have a current of 60 amps for 1/3 of the time, then in
normal cases you would rate the wiring for only 20 amps average.
The wire under 60 amps peaks would heat up as much as 20 amps
smoothed out.
This applies to both DC and AC, but with AC under a normal linear
load, the peaks on a 20 amp average load goes up to 28.28 amps.
What one needs to look at is the average. If the average is the
normal 20 amps RMS equivalent (though not shaped that way), then
we have 20 amps of heating.
In a normal 2 pole split single phase circuit, with half the load
on each side, the current flow would all be in the same direction
for both sides at the same time (assuming similar power supplies
and similar load ratios). The peak currents on each side would
share the neutral in a way that current would flow between the
two wires joining up at the neutral, but not through the neutral
back to the main service.
In 3 phase power, the problem changes.
With sine wave currents in 3 phase, there is a more complex sharing
of currents on 3 phases joining at the neutral. But they do total
up to zero when all 3 phases are in balance with linear (where the
current has the same sine wave alignment as the voltage) loads.
With nonlinear currents, this changes. What I was showing with
the pulses at 1/3 of a cycle time is the worst case. The pulses
on each phase never occur at the same time as a pulse on either
of the other phases. So during a pulse, the one hot leg plus the
neutral carry the current. The next pulse is on a different hot
leg, but it's the same neutral carrying the other half of the
current circuit.
So the averaging, where you can use 20 amp wiring for 60 amp peaks,
only applies to the 3 hot legs. The neutral is going to get 60
amps of average current load.
Imagine you have wiring rated for 20 amps, but you put 60 amps of
lights on it. You add a motorized switch to turn the lights on and
off so the lights are on only 1/3 of the time. Assuming the rate
is fast enough, the wiring will only heat up based on the 20 amps
of average. You could get away with 20 amps of wiring. Now imagine
you have 3 of those circuits and synchronize the motors so that only
one set of lights is on at the same time. Now you feed all three
sets of lights with individual wires for the motor controlled "hot"
side, but you bring the return current "neutral" back on a shared wire.
That shared neutral wire will need to be rated for 60 amps of current
even though the three hot wires only need rating for 20 amps.
That's a rough analogy to what 3 phase power is doing WHEN the currents
are pulse type currents that don't coincide in time. The neutral has
to serve all 3 phases without any benefit of sharing currents out at
the load. So the neutral has all 3 loads in full.
Mathematically, any harmonic frequencies which are odd multiples of
3 times the fundamental frequency can introduce this problem. Square
waves contain these frequencies, and if narrow enough, pulses can be
composed entirely of these frequencies. In this worst case scenario,
the neutral will be carrying 180 Hz in US/Canada and other places
that use 60 Hz, and carrying 150 Hz in Europe and other places that
use 50 Hz.
The industry refers to these odd multiples of 3 harmonics as "triplens".
This applies to the more common "wye" configuration of 3 phase. The
issues exist with "delta" configuration, too, but not as dramatic
because even sine wave currents on delta configurations require a
1.732 times rating which is already figured in, while the worst case
nonlinear load would push that to 2.0 times rating, which is only
a 15.5% increase over the linear case. So in the worst case with a
delta configuration, you only need to overrate each conductor by
just 15.5%.
And worst case situations are not actually common. In real life the
nonlinear loads are only a portion of the whole load, and the pulses
don't usually get so narrow as to be 1/3 of the cycle.
This document provides a more detailed examination:
http://www.squared.com/us/products/transfor.nsf/07a0210021262d45862564b5006e4f84/38abeddb7b9afe79852565b5004c8ed0 /$FILE/0104ED9501R896.pdf


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