Lots of fans. Keep in mind that this desktop ATX case was put together in
2001. So however many fans would have been used then.
I've been looking at power supplies at NewEgg -- they range from $11 to $22.
Some have two fans, some one. Some fans are 80 mm and some are 120. Some
advertise ball bearings, others say "low noise, low ripple, super quiet"
etc. Most says 20+4 pin, and a couple say 24 pin. A few are "top mount."
This is an ordinary desktop--no gaming or video. Does it matter much what
kind of a power supply I buy?
<*(((><
The side of the power supply will have a label on it. You want
a supply that matches or exceeds the characteristics. Speaking
in very general terms (without going into what each number means,
and they *are* important), you'd want a 350W or greater supply,
to replace the broken 350W supply.
Buying an $11 to $22 supply is a mistake. Those are "bargoon"
supplies. Read the reviewer comments on Newegg, to see how
many were dead on arrival, how many blew when used, and how long
they last. I remember one bright guy, who bought four of the
same cheap supply sequentially (first one blows, buys a second,
that blows, buys a third...) and never clued into the fact that
he ended up paying a fortune for his supply.
My last replacement supply cost me about $60, runs cool and
behaves well. It is made by Enermax. Even they aren't perfect,
but chances are the supply I got, will last longer than the
$11 one.
To test a supply, you can try this.
1) Disconnect the supply from the computer. We don't know
what defect it has. You didn't say what brand of computer it
is, and you could have an Emachine with a 250W Bestec supply in it.
Those tend to blow, and take out the motherboard when they go.
Many other supplies have defensive circuitry of one sort or
another, reducing to some extent, the odds of that happening.
It is still possible, for the world's best designed supply,
to be damaged by a lightning strike, so there are limits to
what they can defend against.
2) Ideally, you want to place some load on the supply. Some older
supplies, might need a small load to help them maintain regulation.
The ATX supply is not a push pull design - it likes to source
current, but cannot sink current if the voltage rises too high.
I personally haven't had any trouble here with that, and the
supplies I've tried generally seemed to work OK without a load
on them. (Some supplies cheat, and have a dynamic load circuit in
them, that switches on when no external load is present.) I also
have a load box, which draws 100W total or so, distributed over
all the output rails (each rail gets a load resistor, and the
-12V rail has a 50 ohm load and so on). When I buy a new supply,
I leave it simmering with the load box on it for two hours, just
to make sure it isn't a dud. I check the voltages delivered
with a multimeter.
3) So ignoring (2) in the interest of time, you connect PS_ON# to
COM. PS_ON# is usually a green wire, while COM, also known as
ground, is a black wire. When PS_ON# is connected to COM, that
causes the power supply to soft start. The cooling fan should
then start to spin on the supply. See the wiring table with colors
and pin numbers, in these docs. First doc is for 20 pin supplies,
second is for 24 pin supplies.
http://www.formfactors.org/developer/specs/atx/ATX12V_1_3dg.pdf (page 30)
http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf (page 37)
4) If you had a multimeter, you could measure the voltages on
the main power harness, while the PSU is running with the PS_ON#
pin connected to one of the adjacent COM pins.
A similar technique can be used, to turn on a computer which fails
to assert PS_ON#. The motherboard normally grounds PS_ON# when it
wants the PSU to run. You can safely connect PS_ON# to COM,
with the main connector plugged into the motherboard (shorting the
motherboard driver won't hurt it, because it is supposed to be
open collector). But what that also does, is override motherboard
safety features. For example, if the motherboard detects a problem
(like CPU overheat), it needs to be able to turn off the PS_ON# signal.
If you still have your shorting wire present, the power stays on and
the CPU eventually gets cooked. So while the "paper clip trick" can be
used to solve a situation where the PS_ON# driver on the motherboard
is broken, there can be resulting collateral damage if something
else goes wrong.
In terms of 20 pin versus 20+4 pin supplies, the 20+4 pin main
connector has a hinged section which can be removed. For example,
oh my new Enermax, I have the +4 pin section disconnected from
the main connector, since the motherboard is a 20 pinner. There
are a very few 24 pin power supplies, where they neglected to
hinge the four pins on one end, and for those, you can still
plug the connector in, as long as no tall components are in the
way. Pin 1 of the 24 pin, goes to Pin 1 of the 20 pin motherboard
connector. The four "left over pins" would hang over one
end. If they're the hinged type, they can be removed.
The extra four pins don't introduce any new voltages, and are
only present to carry extra current if needed. The most typical
situation where the extra pins are needed, is on computers which
have two PCI Express video cards installed. For most other situations,
the extra pins are not needed.
You can see a picture of the "paper clip trick" here. PSU plugged
into the wall, and switched on at the back, then the paper clip
stays in position for as long as you want it to stay on. It looks
like they connected pin 14 (PS_ON#) to pin 16 (COM). There are four
black COM wires in the vicinity. The green wire is on the side
with the plastic latch.
http://www.chirio.com/IMAGES/atx_012_ps_on.jpg
HTH,
Paul
Very impressive.