Gene said:
My take on that issue is this:
Everything in the computer is running from the computer's AC to DC
supply, and every one of the DC busses in that supply has filter
capacitors. These will provide current to their loads for enough
milliseconds that the short interruption of the UPS's output will not
matter.
Consider that in the old days before switching supplies, the filters
were reducing ripple due to the 120 Hz output of full wave bridge
rectifiers (it was a bunch of positive half cycles), hence they had to
maintain their output voltage and current for ~8 msec.
I have not addressed the possibility that real voltage spikes can get
through, because I'm not qualified to talk about that.
Switching supplies, like the ATX power supply, are resilient to voltage
value. The incoming AC is converted to high voltage DC (by rectification).
And a switching converter can work over a wide range of DC values. The
only limits on wide range input, are the resulting levels of current
draw. (If the supply draws 300VDC at 1 amp, if the high voltage DC
drops to 100V, then the current draw would be 3 amps to support the
output power. And at some point, the switching transistors will be
exceeding the current they can safely handle. That's why they have
to stop a switching converter, from trying to run when the input
voltage is too low. It would do it, if you let it.)
The ATX supply has a minimum of around 16 milliseconds of holdup time
at full load. This is guaranteed by the input capacitor (the one on the
high voltage DC side). If the ATX supply is not providing full load output
at the time of the AC failure, the power can last a lot longer. On my
current supply, if the computer is sleeping, the +5VSB can "stay up" for
30 seconds, using nothing but the input capacitor as an energy source.
That's without a UPS to help it. But at full load (which my computer
never draws), the supply can only maintain the output for 16 milliseconds.
(Check the spec sheet, as the value can be slightly longer than that.)
The 16 milliseconds of holdup, is enough holdup to support the usage
of a cheap UPS. The UPS transfer time, will be a number less than that.
The original ATX supplies were also waveform tolerant. That's because the
input stage (where cord enters the box) doesn't use a transformer, so
there are no harmonics to worry about. So if the waveform wasn't a
perfect sine wave, the old supplies didn't mind.
That changed, with the advent of ATX supplies featuring either passive
or active PFC (power factor correction). An Active PFC has an expectation
about the wave shape. It tries to make a sinusoidal input current, based
on monitoring the input AC voltage and assuming the voltage waveform
is also a sinusoid. (In other words, it uses the AC voltage as a
template for its behavior.) If a UPS with square wave output is
connected to an Active PFC ATX supply, then a less than ideal behavior
results. The extent to which this is an issue, will depend on
whether the idiot designing those chips, considered cheap UPS
boxes as a potential power source :-(
So in fact, a "true sine wave" UPS (the more expensive kind), is
ready for any kind of computer ATX power supply you throw at it.
Whereas a square wave UPS (or modified sine UPS), those would
work best with a non-PFC ATX supply. This is one reason, the last
two power supplies I bought, I was careful to make sure they had
no power factor correction, as then, they would not be affected
by my crappy UPS (which does what is needed, just fine).
The transfer time of a UPS, can range from zero to one would presume,
close to 16 milliseconds. A UPS would be pretty useless, if every
ATX supply connected to it, "tipped over", so the switching
interval must be shorter than that. A UPS with a zero transfer time,
costs more than one with finite transfer time. My UPS is not zero
transfer time, and the output is probably modified sine wave, rather
than being so horrible as to be pure square.
A double conversion UPS, has a zero transfer time, because it
never transfers anything
You can tell you've got a double
conversion one, because of the loud fan that runs all the time,
the humming and the heat. (We had one at work.) The unit is
basically doing its own power conversion at all times. My UPS,
by comparison, does jack squat 99.999% of the time (bypass mode),
and makes an ugly noise when the AC goes off. And to some extent,
that contributes to its long life (runs ice cold, and will most
likely die of battery failure).
[ Everything in life has tradeoffs ]
Paul
