3.3V on ATX power supply doesn't work

[David Peterson]

I am building a pc and it doesn't work, so I got out my voltmeter and
shorted the 14-15 pins on the power supply to turn it on, and tested
all the pins. None of the 3.3V's work (there are 3 of them), they are
all 0 V. I just wanted to check, well this is stupid to ask even, but
the power supply is bad, right? The 3.3V's need to be working, right?
And they should be showing up when I short pins 14/15? Thanks.

 

[w_tom]

If any of power supply voltages don't stabilize, then all voltages
must shut off. If your power supply does not do that, well, how much
did you pay for it and how many other essential functions are missing?

Meanwhile, best is to test with any power supply wires connected.
Test is far more informative if everything is left connected. A
procedure for everything to measure; to get numbers that are also
posted here for additional information is in "When your computer dies
without warning....." starting 6 Feb 2007 in the newsgroup
alt.windows-xp at:
http://tinyurl.com/yvf9vh

For example, what did voltage on the gray wire do? Currently your
post makes no sense. 3.3 volts missing while other voltages exist?
That must not happen. Either your measurements are in error, an
unacceptable problem was created because the power supply was not
connected to motherboard, or ... well too many other possibilities
without all those other numbers.

 

[David Peterson]

OK here's my results, referencing that post:

First measurement is the purple wire from power supply to
motherboard. Black probe attached to chassis and red probe contacts
inside nylon connector body. Measure DC volts. That number must
exceed 4.87.
5.09

Next measure green wire. That voltage must be more than 2 volts
before switch is pressed and drop to well below 0.8 volts when switch
is pressed.

4.8 before, drops to 0

Next is gray wire that must rise well above 2.4 volts within but
seconds of pressing power switch.
5.15

And finally, what happens on one of orange, red, and yellow wires
when and after switch is pressed. Each must rise within a second and
then remain above 3.23, 4.87, and 11.7 volts. Any one that does not
move - that too is a valuable fact.

At this point the power supply would no longer turn on. I hit the
switch and the fan turns a little bit but it doesn't turn on. I turned
it off and on many times and unplugged it from the motherboard and
replugged it and still I can't get it to turn on. I shorted 14/15 and
it does turn on in that case (i.e. not plugged in to the motherboard).
When attached to the motherboard the purple wire is still 5.09.

I guess it should be obvious that this power supply is broken. Is that
the only conclusion?

What happened before, when I first powered up the system, the cpu fan
wouldn't come on. Is my cpu definitely toast?

 

[meerkat]

BLACK Probe into nylon body on a black wire !!

OK here's my results, referencing that post:


4.8 before, drops to 0


At this point the power supply would no longer turn on. I hit the
switch and the fan turns a little bit but it doesn't turn on. I turned
it off and on many times and unplugged it from the motherboard and
replugged it and still I can't get it to turn on. I shorted 14/15 and
it does turn on in that case (i.e. not plugged in to the motherboard).
When attached to the motherboard the purple wire is still 5.09.

I guess it should be obvious that this power supply is broken. Is that
the only conclusion?

What happened before, when I first powered up the system, the cpu fan
wouldn't come on. Is my cpu definitely toast?

 

[David Peterson]

The black probe on system chasis is fine. What's wrong with that? And
I just stuck the red probe into the back of the plug.

 

[w_tom]

When power supply stops turning on, what happens on purple, green,
and gray wires. Voltages for all wires must be taken when power
supply does turn on. Then another voltage list is taken when power
stops turning on. Notice for example, when power does not turn on,
still, some voltages exists. Another fact that is necessary before
accusing a suspect component.

Red, orange, and yellow wire voltages must have existed at during
first measurements because gray wire 5.15 volts implies that some
voltages existed - good or unacceptable. Make measurements for both
conditions. A table of voltages must be listed for all wires when
power supply does turn on and when power supply does not turn on.

Minor point: one number is troubling. Purple wire voltage is 5.09.
Therefore gray wire voltage should rise to less than 5.09. Where did
5.15 volts come from? That may be a problem elsewhere - not in power
supply. A problem we ignore until later during analysis. This
paragraph only demonstrates why those numbers provide others with even
more useful facts.

When power supply stopped turning on, then a whole new table of
voltages from purple, green, gray, etc wires must be taken. Currently
everything in a power supply 'system' is suspect because we don't know
what the purple, green, and grey wires did when power supply did not
power.

Other peripheral issues: Black wire to chassis is a perfectly good
connection.

Is CPU toast? Most likely not. Without details such as CPU
manufacturer, then no one could answer that question anyway. However
limit your attention entirely and only to those power supply
measurements. One step at a time. Other issues such as CPU fan come
later; would only confuse things now.

Don't connect and disconnect anything. When power supply did not
power, then do not disconnect power supply from motherboard. Any
disconnecting only complicates the problem. Don't try to fix
anything. That comes later. Collect facts. Don't change anything
until we have facts. If something changes during the measurements -
ie power supply no longer powers - then start taking all measurements
again during this new situation. With numbers - before and after
power supply stopped working - then we would have a treasure trove of
information.

Shorting 14/15 causes it to power? Well that implies a power supply
'system' problem that is not the power supply. But again, only
implies; we still cannot say for sure ... until we have all those
numbers for both 'does power' and 'does not power' conditions. That
14/15 short tells us something - but still not enough to know anything
useful. Now if you short 14/15 and then collect all those numbers
again, well, that might tell us something useful. But not as useful
as taking all those measurements both when power starts and when power
does not start - and zero system changes.

Ok. Let's assume the power supply stopped powering on. After
collecting all numbers, only then do we change one thing. Disconnect
and restore AC power plug from AC mains. Does that change anything?
Does power supply start or at least try longer to start? If so, then
take meter numbers again.

My suspicion? Not a power supply problem. But that is only a horse
race bet - not yet sufficient numbers to act upon.

 

[larry moe 'n curly]

I am building a pc and it doesn't work, so I got out my voltmeter and
shorted the 14-15 pins on the power supply to turn it on, and tested
all the pins. None of the 3.3V's work (there are 3 of them), they are
all 0 V. I just wanted to check, well this is stupid to ask even, but
the power supply is bad, right? The 3.3V's need to be working, right?
And they should be showing up when I short pins 14/15?

3.3V is used for a lot of the chips on the mobo and plug-in cars. If
this is the only voltage not working, then its diode pack could be bad
(one of the large 3-pin packages bolted to a big heatsink; some PSUs
use a pair of them for some voltages, and usually both in a pair need
to be replaced if one goes bad). But this shouldn't prevent the PSU
from being turned on by shorting the green and black wires together.
If the PSU's fan won't even spin, then the problem is more than a bad
3.3V diode pack.

I have a 350W Ultra PSU made by Wintech, and the first two samples
were defective. The first one's fan sometimes wouldn't start, and the
second one, from a different production run, had bad soldering (blob
of solder on one of the 120VAC wires, a ground wire could be pulled
from the circuit). Maybe the second problem is what's affecting your
Wintech.

 

[David Peterson]

I can't get the PS to turn on any more when it is plugged into the
mobo. If I short 14/15 it does turn on.

Here are the voltage readings I got:

First measurement is the purple wire from power supply to
motherboard. Black probe attached to chassis and red probe contacts
inside nylon connector body. Measure DC volts. That number must
exceed 4.87.

5.09 as before

Next measure green wire. That voltage must be more than 2 volts
before switch is pressed and drop to well below 0.8 volts when switch
is pressed.

4.8 before pressing, dropped to 1.16 momentarily and fan started to
spin. But the fan immediately stopped, and the PS didn't turn on.
Thereafter it is 4.81 before pressing button and 4.81 after pressing
button and the PS doesn't turn on and the fan doesn't move at all.

Next is gray wire that must rise well above 2.4 volts within but
seconds of pressing power switch.

0 before, 0 after.

And finally, what happens on one of orange, red, and yellow wires
when and after switch is pressed. Each must rise within a second and
then remain above 3.23, 4.87, and 11.7 volts. Any one that does not
move - that too is a valuable fact.

None of these change when I press the button. ORA = 0, RED = 0.39, YEL
= 0.36.

I checked ohms to ground, black is 3.31 ohms, orange is also 3.31
ohms!

So, given all these measurements, what can you conclude. (thanks for
the detailed reply w_tom)

 

[w_tom]

4.8 before pressing, dropped to 1.16 momentarily and fan started to
spin. But the fan immediately stopped, and the PS didn't turn on.
Thereafter it is 4.81 before pressing button and 4.81 after pressing
button and the PS doesn't turn on and the fan doesn't move at all.

Power supply controller is not ordering power supply to turn on. It
drops only to 1.16 - not sufficient. First, is there anything that
might cause power supply controller to stop the power supply? For
example, is a CPU fan monitor disconnected or enable in BIOS when no
such monitor connection exists. Is their anything on motherboard that
would tell the power supply controller to turn off power supply? If
not, continue.

We know where the problem is - a connection between power supply
controller and power supply. Two hardware reasons can create this
problem. First is driver inside controller is defective - or does not
tell power supply to stay on. Other is that something has shorted to
the green wire inside power supply - too much current for the
controller's driver.

OK. We are not ready to separate the two suspects; determine with
is 'evil'. Disconnect power cord and then disconnect power supply
cable from motherboard. Then restore power cable.

Set meter to DC amperes - a maximum range. Connect meter black
probe to any black wire into the power supply. Then touch the red
probe to the green wire. This should cause power supply to at least
try to power on.

Now lower the DC amp range until a current is read on multimeter
display. That current should be well below 20 milliamps (0.02
amps). If too high, then something inside power supply has shorted
too much current to green wire. If below 20 milliamps, then driver on
power supply controller was not telling power supply to power on.

Other suspects. Well the power switch can be eliminated. When
pressed, the power supply controller did reply. Therefore if problem
is on motherboard side of green wire (on power suppy controller), then
motherboard must be replaced.

You might kludge a solution (assuming problem is motherboard
controller). You could wire a switch between green wire and black
wire.

Meanwhile, what caused this problem? This is most typically due to
manufacturing defect. Either the defect is in power supply or on
power supply controller. This failure would not be created by surges,
heat, or other external events.

Of course, report back what those currents were. Those numbers and
power supply response as current is taken might provide other useful
information.

 

[David Peterson]

Well, I blew it, literally. I went ahead and got a replacement for the
power supply since I knew it wasn't supplying 3.3V. I got the new one
today and shorted 14/15 and all the voltages were good. Then I plugged
it into the mobo and right when I turned it on I heard a crackle/pop
in the power supply and smelled that fried electronic smell. The CPU
fan did not turn on. I turned it off, shorted 14/15, and, just like
the last ps, the 3.3V was now ground (I get a 0.0V reading where I
should get 3.3V) and the other voltages still read fine. So the
motherboard shorted out this ps and the last one too. Yikes. Seems I
need a new motherboard. Is it possible that my cpu or memory are now
ruined (or, the cause of the short)? I can get those amp readings
tomorrow if that will help.

 

[w_tom]

If motherboard damaged a new power supply, well, did that supply
come with a long list of numeric specifications? If not, then look at
yourself for a source of failure. All power supplies can have all
outputs shorted together and not be damaged. Just another essential
function standard in power supplies 30 years ago. No load can damage
a power supply. No motherboard can damage any acceptable supply.

Since so many computer assembler - since so many certified computer
technicians - do not even know how electricity works, then the market
is full of power supplies missing essential functions. How do you
know it is not acceptable? No long list of specs is necessary to
'forget' essential functions inside a supply - to sell only on dollars
and watts.

That power supply is damaged only if power supply was defective when
designed- a more than 30 year requirement and demanded by Intels own
power supply specifications. This standard is routinely violated
where power supplies are sold only on dollars and watts.

Testing a power supply when not connected must never damage a
supply. However what a supply does without load is a function unique
to that particular design. Some supplies will power; others will not.
More responsible manufacturer may provide that fact in numerical
specs. Least responsible manufacturer will not provide those specs.

Since supply is missing essential functions, then what did it do to
a motherboard? Problem is now exponentially more complex. Each new
variable tends to exponentially complicate the problem. Just another
in so many reasons why shotgunning is bad - although it has caused you
to learn - a good thing.

Hopefully a power supply was only defective when manufactured - not
defective by design. Hopefully the defective power supply did not
damage motherboard. Meanwhile post numbers (as described for old
power supply) to say motherboard was good or bad. We still need
numbers from that 26 Mar post to know what did exist. Those numbers
on 26 Mar would have definitely answered questions. Now numbers only
might answer those same questions. Never do anything until clear
reason exists to do it - which is why 'follow the evidence' is so much
faster and reliable. Get those ampere numbers from original supply's
green wire as described on 26 Mar.

Meanwhile, minimal electrical knowledge demonstrates why the load (a
motherboard) must not damage a power supply - and why.