KR said:
There would be relatively a large "surge" current at start up, when the capacitors
charge up, but doubt this is much of a problem in typical PC and consumer supplies.
This is true. And some supplies, the specs actually state what the surge current will
be. It can range from 40 to 80 amps for a short period of time, like a single cycle.
My ATX supply here, causes the lights to flicker when first switched on at the back,
and that's the surge. The transient is too short, for my UPS to declare the event
to be an "overload". So the duration is short.
That is the purpose of NTCR1 on the upper left of this schematic. It is the
inrush limiter, and has a negative temperature coefficient. To work properly,
when you switch off the supply, you should wait 30 seconds for NTCR1 to cool
off. Then, the next time the back switch is energized, the inrush (surge) will
be limited to the stated value in the spec for the supply. Rapidly toggling of
the rear switch, defeats the protective action of NTCR1, and could lead
to a primary side failure. It needs time to cool off.
http://www.pavouk.org/hw/en_atxps.html
The reason caps fail in switchmode supplies (excluding the mains filter caps) is the high frequency appearing across them. There are large ripple currents involved, over very short periods, and unless the cap is designed for this (Low ESR) it will overheat, bulge and fail. Same with motherboards. Even caps supposedly designed for this can fail over time.
There are two reasons for caps to fail.
The first, is engineering. You have to take the ripple current rating of the
application, the expected temperatures into account, then select the right
capacitor or number of capacitors in parallel for the job. The Arrhenius
equation, with curve fitted exponent, helps predict how long the capacitors
will last, as part of the engineering exercise. One capacitor company claims
you can get up to 15 years from an electrolytic capacitor, before the rubber
bung on the bottom of the cap dries out, and with it, the electrolyte.
So the manufacturer thinks they last for 15 years. Less, if they're constantly
being overheated. Life is very short, if they run at 105C all the time.
The second form of capacitor failure, is purely chemical. Billions of bad
capacitors were made, with an electrolyte formula lacking a stabilizer.
Such capacitors will fail after two years, even if the power supply is
sitting on a shelf, cold. The metal corrodes, and juice leaks out the top.
The pH of the electrolyte is wrong. I had an Antec fail that way, in storage,
and there were four caps leaking inside. The capacitors do not need to be
under bias, or in stress, for a "pure chemistry failure". No engineering
equation would predict it. Because it wasn't intended to work that way.
If the chemistry is right, the caps can last a long time. My 440BX motherboard
still works for example, and must be close to 12 years old.
If you abuse a capacitor in an application (remove an OSCON and replace with
a regular electrolytic capacitor), then the capacitor would be out of its
league, and shouldn't last very long. It wouldn't be rated for the ripple.
Paul
