gnesterenko said:
Howdy everyone.
Came across this post via a google search as I'm having a similar
problem. Power supply fails to boot PC every so often - though its been
better since I took it apart and cleaned out all the dust that had
collected in it over the last 4 years - in fact it has not failed to
boot once since that cleaning - so I'm hoping that was the issue (it was
REALLY dusty in there). That aside though, I purchased a brand new shiny
Corsair HX750 (the modular model) to replace the aging 420W enermax. It
has never booted my computer properly. All I get is the blinking green
light and nothing else. I figured - faulty PSU - though I haven't done
the plug in a single fan and jump the power connection test yet, gonna
do that sometime over next few days. Here is the thing though - you
mentioned that the blinking led is tied to the 5V rail. That got me
thinking.
My motherboard that I am experiencing these issues on is an Asus
A8N-SLI Premium - which uses the ATX12V 2.0 specification. Now checking
out Wikipedia on the ATX12V power specs, the 2.0 version moved the 5V
circuit to the 24 pin connector and significantly reduced the power
flowing through this 5V circuit. All well and good.
Then I took a look at ATX12V 2.01 specification. Quote "This is a minor
revision from June 2004. The −5 V rail was completely removed from
the specification."
Then I took a look at my new Corsair HX750 specs and saw that it
purported to support ATX12V specs 2.3, and backwards compatible only to
2.01 - not 2.0. Its strange however, as the specs on the PSU DO show a
+5VSB rail as being present.
Could the fact that the 5V rail was removed in 2.01 be the source of my
troubles? Going to test the power supply by jumping the pins and/or with
my dads computer which uses a newer mobo and a modern ATX12V spec. But
any other suggestions are welcome. Was hoping this motherboard would
last until 890 chipset from AMD, don't wanna replace sooner if possible.
On the oldest 20 pin power supply spec - all pins and voltages are present.
Next spec is still 20 pins, but removes the -5V power output. That is minus five volts.
Motherboards no longer rely on -5V being present. You should see a blank
space on the connector, where the removed -5V wire and pin used to be located.
Next spec introduces 24 pins. The four extra wires and pins are there, to carry
more current. No new voltages were introduced on the 24 pin connector. This makes
almost no difference to the motherboard. (The exception is, it helps an SLI
motherboard with two video cards, as some of those combinations draw
slightly more than 8 amps. The rail that needs the current, is 12V.
So the fact there is a 3.3V pin added, and a 5.0V pin added by the
new standard, is practically pointless.)
The +5VSB rail is a standby power source. On your Asus motherboard, it is
monitored by a green LED near the bottom edge of the motherboard ("SB_PWR").
+5VSB is present on all the above mentioned power supply specifications.
+5VSB must work, before the motherboard can turn on the rest of the output
rails via the PS_ON# signal.
When the computer is sleeping, the +5VSB continues to run, and it powers the
RAM sticks. An onboard regulator on the motherboard, drops that voltage to
a level where it can be used by the RAM. Chips such as the LAN chip, are
powered as well, in order that "Wake On LAN" can work.
When all the +5VSB loads are taken into account, perhaps 1 amp is drawn.
The motherboard manual may suggest typical values for that loading (that
is where I got the 1 amp value from).
If you had external USB devices, they might draw power from that rail as well.
A typical supply can provide +5VSB at 2 amps. That is sufficient for a
sleeping computer, plus a couple heavy USB devices. For example, if you
were charging a music player from the USB port, that might represent a
heavy load.
If you overload the +5VSB rail, it is protected, and the power supply
responds by dropping the output voltage to zero. Depending on the supply
design, you may see this happen over and over again, giving a slow blink
on the green LED. The frequency of blink may be irregular.
*******
In terms of verifying the performance of the power supply, you'd be
interested in +5VSB and not the +12V rail (since we can see the green
LED blinking, and have to get the +5VSB working properly, before anything
else will work right). A fan can be used to do a quick test of +12V.
Testing +5VSB is going to be more difficult and complicated to do.
There are no nice loads I can think of, to connect to +5VSB, to test it.
You can buy resistors at Radio Shack. For example, say I were to buy
several packages of 50 ohm 1/2 watt carbon composition resistors, with
wire legs on the ends. The relationship between voltage and power is
P = (V*V)/R 0.5 watt = (5V * 5V)/ R, solving gives R = 50 ohms
So a 50 ohm resistor, is the lowest ohm value 1/2 watt resistor I could
load the +5VSB power supply with. When the 50 ohm resistor is connected
to +5VSB, it dissipates 0.5 watts of heat, and gets real hot. If I
bought a 68 ohm resistor instead, it draws less power (less current too).
So a 68 ohm resistor wouldn't get quite as hot. But I'd need more resistors
to do my tests with.
P = ( 5 * 5 ) / 68 = 0.368 watts
To do a load test, you add resistors one at a time, to the thing to test.
You add the load resistors in parallel, to increment the load. In the
first diagram, one resistor draws 0.1 amp (i.e. 5V / 50 ohms = 0.1 amp).
In the second diagram, two resistors draw 0.2 amp total. The resistor
gets blistering hot. You continue adding resistors, until you hit 2 amps.
That takes 20 resistors at perhaps $0.20 a piece, so our load resistors
cost $4.00 total.
+5VSB ------+ +5VSB ------+-----------+
| | |
50 ohm 50 ohm 50 ohm
1/2W 1/2W 1/2W
resistor resistor resistor
| | |
COM --------+ COM --------+-----------+
In each case, you verify the power supply still is putting out +5VSB
at the full five volts. If the output dropped to zero volts, after
adding 17 resistors, you'd know the real output capacity of the
supply was a little less than 1.7 amps.
So that experiment cost us $4.00 for a handful of resistors, plus at
least $20 for a multimeter to verify +5VSB still has full output voltage
during each test case.
At the power supply factory, that test is automated. Using a Chroma
tester, they program the tester to ramp the load and verify the output
remains within spec. They can do any arbitrary combination of loads
on the rails, so they can verify the "cross-regulation" spec for
the supply. Doing it manually with resistors is cheaper, but a lot
slower to do.
So, say the test results went well above 20 resistors, meaning the
supply was delivering its full 2 amps output. Then, you'd have to
conclude, the electrical load in the computer, was too great for
the supply to handle. You'd get out your clamp-on DC ammeter, and
attempt to measure the current. Since the time the supply remains on
could be fairly short, there is no guarantee the meter will measure
the peak value properly. It all depends on how long it takes the
supply to shut off the +5VSB output.
As you can see, debugging this is a messy proposition, and probably
more than you're interested in doing. And the thing is, many computer
shops may not be set up to do that kind of testing either, preferring
to "swap power supply" and "swap motherboard", until the problem
goes away.
I picked a linear sequence for my load example above, for the purposes
of illustration. It may make it slightly easier to buy resistors,
to do it that way (buy a set of resistors which are all the same
value). You could also buy a set of resistors, such as one each of
50 ohms, 25 ohms, 12.5 ohms, 6.25 ohms, 3.125 ohms, and by using
combinations of them, arrive at a linear load sequence as well. But
odd-ball resistor values (each with a different power rating), would be
much harder to find at the store.
You'll notice, I didn't suggest loading the power supply with light
bulbs. Incandescent bulbs have a non-linear filament characteristic,
and draw more current when they're cold. That can cause the overcurrent
on the power supply to trip out, before the full rated output is achieved.
I picked resistors for my test load above, as they have a more
constant resistance value than a light bulb would have. Even the resistor
has a temperature coefficient, but it is not nearly as bad as a light bulb
would be. (The most constant resistors are made from material such
as manganin alloy wire, but that isn't something you'll find at Radio Shack.)
http://en.wikipedia.org/wiki/Manganin
Paul
