Really stumped: PS inoperative
Haines Brown
advice regarding the proper procedure for testing out a power
supply. But my problem is more basic than that.
I had an old Enermax power supply in a new box that seemed to work to
the extent I could boot to the MB spash screen, but then it stopped
working altogether (no fan spin, etc.) If I played with the computer
front panel on/off switch (depressing a few seconds, releasing a few
seconds and repeating the cycle), I could get the CPU can to make a few
turns, but that's all. I checked the PS fuse, and it's OK. No visible
signs of trouble inside.
I decided to purchase a new power supply, but when I installed it, same
problem. When I pushed the on switch the first time, the CPU fan rotated
a couple of times, and then nothing. MB LED on, but no fans or anything
else.
I took the old PS by itself, not even installed in a machine, and
connect one of its molex connectors to a fan. I connect power, turn the
supply on, and the fan does not spin. The 115/220 volt switch is in the
right position (115 visible); the power cord does deliver the
voltage. The fan is new and I assume works.
I might be tempted to circuit trace the power supply if I didn't have
the same problem as the new one. Obviously I'm doing something that is
very elementary wrong.
If a power supply is getting juice, the switch is on, and the fuse not
blown, shouldn't the periphery power connectors cause a fan to turn?
Does not at least that part of the power circuit become active as soon
as the supply is turned on?
--
Haines Brown, KB1GRM
Bob Eager
<bro...@teufel.hartford-hwp.com> wrote:
> I raised this issue before in this group and received some very helpful
> advice regarding the proper procedure for testing out a power
> supply. But my problem is more basic than that.
>
> I had an old Enermax power supply in a new box that seemed to work to
> the extent I could boot to the MB spash screen, but then it stopped
> working altogether (no fan spin, etc.) If I played with the computer
> front panel on/off switch (depressing a few seconds, releasing a few
> seconds and repeating the cycle), I could get the CPU can to make a few
> turns, but that's all. I checked the PS fuse, and it's OK. No visible
> signs of trouble inside.
(assuming this is an ATX PSU with the 20 pin connector)
> I decided to purchase a new power supply, but when I installed it, same
> problem. When I pushed the on switch the first time, the CPU fan rotated
> a couple of times, and then nothing. MB LED on, but no fans or anything
> else.
The PSU needs to have a particular pin (PS_ON) grounded to run properly
(and run the fan). It generates a standby voltage on a different pin,
which is enough to run the power control part of the MB, which then
grounds PS_ON when the MB considers itself to be 'on'. The MB may not be
holding that pin at the correct level.
> I took the old PS by itself, not even installed in a machine, and
> connect one of its molex connectors to a fan. I connect power, turn the
> supply on, and the fan does not spin. The 115/220 volt switch is in the
> right position (115 visible); the power cord does deliver the
> voltage. The fan is new and I assume works.
You need to strap the appropriate ground pin to the PS_ON signal.
> I might be tempted to circuit trace the power supply if I didn't have
> the same problem as the new one. Obviously I'm doing something that is
> very elementary wrong.
>
> If a power supply is getting juice, the switch is on, and the fuse not
> blown, shouldn't the periphery power connectors cause a fan to turn?
> Does not at least that part of the power circuit become active as soon
> as the supply is turned on?
A small part does, but it doesn't run the fan.
See here:
http://www.formfactors.org/developer%5Cspecs%5Catx2_2.pdf
and look at chapter 4.
--
Bob Eager
begin 123 a new life...take up Extreme Ironing!
Haines Brown
> The PSU needs to have a particular pin (PS_ON) grounded to run
> properly (and run the fan).
Just what I needed to know to start tracking down the problem. When I
sort pin 20 on the main power connector to the black ground next to it,
the fan runs. So it's beginning to look like I have a problem with the
power-on circuit outside the power supply.
If you are willing, you might help me with a further question. In the
power supply there's apparently a resister between the standby 5+ volts
and ground that holds the Power-On to some level, which does not allow
the power supply to power its rails. When I grounded pin 20 on the main
power connection, that dragged the voltage down to 0, which enabled the
power supply to deliver power to the rails.
What I don't know is how the power on switch on the front of the
computer chassis ties in to this. It is a push-button which that closes
a circuit when depressed and is normally open circuit. Apparently this
temporary closure causes the Power-On connection to the power supply to
toggle its state between being grounded and being open. How is that
done?
I've assumed this control circuit that includes the front panel power
switch is not polarized.
> See here:
>
> http://www.formfactors.org/developer%5Cspecs%5Catx2_2.pdf
>
> and look at chapter 4.
Yes, thanks. Useful.
--
Haines Brown, KB1GRM
Arno Wagner
> I raised this issue before in this group and received some very helpful
> advice regarding the proper procedure for testing out a power
> supply. But my problem is more basic than that.
> I had an old Enermax power supply in a new box that seemed to work to
> the extent I could boot to the MB spash screen, but then it stopped
> working altogether (no fan spin, etc.) If I played with the computer
> front panel on/off switch (depressing a few seconds, releasing a few
> seconds and repeating the cycle), I could get the CPU can to make a few
> turns, but that's all. I checked the PS fuse, and it's OK. No visible
> signs of trouble inside.
> I decided to purchase a new power supply, but when I installed it, same
> problem. When I pushed the on switch the first time, the CPU fan rotated
> a couple of times, and then nothing. MB LED on, but no fans or anything
> else.
> I took the old PS by itself, not even installed in a machine, and
> connect one of its molex connectors to a fan. I connect power, turn the
> supply on, and the fan does not spin. The 115/220 volt switch is in the
> right position (115 visible); the power cord does deliver the
> voltage. The fan is new and I assume works.
That is as expected. An ATX psu gets its poer_on signal from
the mainboard and the power-switch on the front is in turn
connected to the mainboard. In order to turn a PSU on in
"stand alone" mode, you have to short the /power_on
pin to ground. Pinouts for 20pin and 24pin ATX can be
found here:
http://pinouts.ru/Power/atxpower_pinout.shtml
http://pinouts.ru/Power/atx_v2_pinout.shtml
/power_on is supposed to be a green wire, ground is black.
> I might be tempted to circuit trace the power supply if I didn't have
> the same problem as the new one. Obviously I'm doing something that is
> very elementary wrong.
You do.
> If a power supply is getting juice, the switch is on, and the fuse not
> blown, shouldn't the periphery power connectors cause a fan to turn?
No.
> Does not at least that part of the power circuit become active as soon
> as the supply is turned on?
The standby 5V must become active. It is supposed to be
on a purple wire, see also the "5VSB" pin in the diagrams
referenced above.
Arno
Arno Wagner
>> The PSU needs to have a particular pin (PS_ON) grounded to run
>> properly (and run the fan).
> Just what I needed to know to start tracking down the problem. When I
> sort pin 20 on the main power connector to the black ground next to it,
> the fan runs. So it's beginning to look like I have a problem with the
> power-on circuit outside the power supply.
> If you are willing, you might help me with a further question. In the
> power supply there's apparently a resister between the standby 5+ volts
> and ground that holds the Power-On to some level, which does not allow
> the power supply to power its rails. When I grounded pin 20 on the main
> power connection, that dragged the voltage down to 0, which enabled the
> power supply to deliver power to the rails.
> What I don't know is how the power on switch on the front of the
> computer chassis ties in to this. It is a push-button which that closes
> a circuit when depressed and is normally open circuit. Apparently this
> temporary closure causes the Power-On connection to the power supply to
> toggle its state between being grounded and being open. How is that
> done?
Some complex ciecuitry on the maonboard. Complex, because most
mainboards also support power on by other events, e.g. network
packets, modem ring, keyboard, timer, etc..
> I've assumed this control circuit that includes the front panel power
> switch is not polarized.
The front switch is not polarized in any way.
What you can do is depress the front switch, and observe
what happens to the voltage across it. You can also monitor
the PSUs power_on pin while pressing the front switch.
At the moment it looks like your mainboard is shot, but keep
looking a little longer.
Arno
Bob Eager
<bro...@teufel.hartford-hwp.com> wrote:
> "Bob Eager" <rd...@spamcop.net> writes:
>
> > The PSU needs to have a particular pin (PS_ON) grounded to run
> > properly (and run the fan).
>
> Just what I needed to know to start tracking down the problem. When I
> sort pin 20 on the main power connector to the black ground next to it,
> the fan runs. So it's beginning to look like I have a problem with the
> power-on circuit outside the power supply.
>
> If you are willing, you might help me with a further question. In the
> power supply there's apparently a resister between the standby 5+ volts
> and ground that holds the Power-On to some level, which does not allow
> the power supply to power its rails. When I grounded pin 20 on the main
> power connection, that dragged the voltage down to 0, which enabled the
> power supply to deliver power to the rails.
>
> What I don't know is how the power on switch on the front of the
> computer chassis ties in to this. It is a push-button which that closes
> a circuit when depressed and is normally open circuit. Apparently this
> temporary closure causes the Power-On connection to the power supply to
> toggle its state between being grounded and being open. How is that
> done?
My understanding is that circuitry on the motherboard latches the
grounded state. This can be unlatched by the motherboard itself, which
is how a power down and switch-off by software is done. There is a
standby (always-on) power line in that 20 pin plug, and that powers
enough circuitry to work the latch. Failure to latch indicates some kind
of motherboard problem.
> I've assumed this control circuit that includes the front panel power
> switch is not polarized.
No, it's not.
BTW, I bought a useful gadget for testing PSUs. Plug the PSU into it,
and it grounds PWR_ON as displays the state of all of the power lines.
It also provides test points for more accurate measurements. But then I
do quite a lot of system building/repairing. It cost me (in the UK)
about $20 (US).
Hans-Peter Diettrich
> I decided to purchase a new power supply, but when I installed it, same
> problem. When I pushed the on switch the first time, the CPU fan rotated
> a couple of times, and then nothing. MB LED on, but no fans or anything
> else.
I had an similar failure on my machine. In the second repair attempt it
turned out that the CPU socket was the reason, why the cooling
construction had no *permanent* contact with the CPU. That caused the
CPU to turn itself off, about 2 seconds after power on. Perhaps you have
a similar temparature/cooling problem?
DoDi
Haines Brown
Zero voltage across the front power switch when not depressed, and zero
voltage at the Power OK pin on the main power connector, whether or not
the front switch is depressed.
> At the moment it looks like your mainboard is shot, but keep
> looking a little longer.
I hope not. I just bought the board to replace another one. At least if
it's defective, it's under warranty.
--
Haines Brown, KB1GRM
Haines Brown
> On Sun, 13 Apr 2008 19:28:43 UTC, Haines Brown
> <bro...@teufel.hartford-hwp.com> wrote:
>
>> "Bob Eager" <rd...@spamcop.net> writes:
> My understanding is that circuitry on the motherboard latches the
> grounded state. This can be unlatched by the motherboard itself, which
> is how a power down and switch-off by software is done. There is a
> standby (always-on) power line in that 20 pin plug, and that powers
> enough circuitry to work the latch. Failure to latch indicates some
> kind of motherboard problem.
Yes, I suspected as much. Despite what I might have just reported in my
other message, with the PS power switch the green PS_ON wire reads
3.7V. When the front panel switch is depressed, the voltage drops to
zero. If I push the front button again, but hold it for a few seconds,
the voltage climbs back up to 2.5V. So it seems the latching mechanism
on the MB is not entirely dead.
The grey PWR_OK wire on the main power connector reads zero after the PS
switch is off for a while and when it is switched back on. If I now
press the front power button, it causes a little activity at this point,
but it quickly settles back to the 0.004V. This seems wrong. I gather
that the PWR_OK connection should be 5 V initially, and when the power
on button is depressed, drop down to 3.3V. I'm not getting anything like
that.
If my impression of the proper voltages here are correct, then the
question is, what determines them? Is the initial voltage supposed to be
5V, and is that drawn from standby power in the power supply? That is,
is the initial voltage independent of the motherboard?
--
Haines Brown, KB1GRM
Haines Brown
Sorry, but a slight correction to my prior message. The voltage across
the front panel button when the power supply is turned on is 0.06V. When
the switch is pushed, this voltage slowly drops to zero. When the front
switch is released, the voltage goes back up to 0.06V.
Sorry for voltages provided in my first message, for I also happen to
have a faulty front panel switch, and I have to simulate the conditions
for the measurements.
See my note to Bob that suggests maybe the latching mechanism is
working, but that the PWR_ON voltage does not have a sufficiently high
voltage and does not respond to MB power latching.
--
Haines Brown, KB1GRM
Arno Wagner
If your measurements are correct, then it is the board. I would
call your mainboard's power-on circuit stone dead.
Arno
Arno Wagner
> Sorry, but a slight correction to my prior message. The voltage across
> the front panel button when the power supply is turned on is 0.06V. When
> the switch is pushed, this voltage slowly drops to zero. When the front
> switch is released, the voltage goes back up to 0.06V.
It should be around 3-5V. 0.06V is in range for a measurement error
or a short circuit on the board.
> Sorry for voltages provided in my first message, for I also happen to
> have a faulty front panel switch, and I have to simulate the conditions
> for the measurements.
> See my note to Bob that suggests maybe the latching mechanism is
> working, but that the PWR_ON voltage does not have a sufficiently high
> voltage and does not respond to MB power latching.
The power-on voltage is supplied by the PSU itself. The board just
shorts it out to ground to turn the power on. It cannot have
an insufficient level, unless the PSU is broken. You can test
that by turniong on the PSU without the board being connected to it
and then measuring the volgate on /power_on.
I still think the board is dead.
Arno
Arno Wagner
> "Bob Eager" <rd...@spamcop.net> writes:
>> On Sun, 13 Apr 2008 19:28:43 UTC, Haines Brown
>> <bro...@teufel.hartford-hwp.com> wrote:
>>
>>> "Bob Eager" <rd...@spamcop.net> writes:
>> My understanding is that circuitry on the motherboard latches the
>> grounded state. This can be unlatched by the motherboard itself, which
>> is how a power down and switch-off by software is done. There is a
>> standby (always-on) power line in that 20 pin plug, and that powers
>> enough circuitry to work the latch. Failure to latch indicates some
>> kind of motherboard problem.
> Yes, I suspected as much. Despite what I might have just reported in my
> other message, with the PS power switch the green PS_ON wire reads
> 3.7V. When the front panel switch is depressed, the voltage drops to
> zero.
Then it woirks ok.
> If I push the front button again, but hold it for a few seconds,
> the voltage climbs back up to 2.5V. So it seems the latching mechanism
> on the MB is not entirely dead.
Again, entriely ok. Note, however, thet this is /ps_on, and not
ps_on, i.e. it is active low. That means a low voltage tells the
PSU to switch on.
> The grey PWR_OK wire on the main power connector reads zero after the PS
> switch is off for a while and when it is switched back on. If I now
> press the front power button, it causes a little activity at this point,
> but it quickly settles back to the 0.004V. This seems wrong.
It is. It means the PSU could not get the voltages up.
> I gather
> that the PWR_OK connection should be 5 V initially, and when the power
> on button is depressed, drop down to 3.3V. I'm not getting anything like
> that.
No. It shoul be false with the computer off and true with it being on.
> If my impression of the proper voltages here are correct, then the
> question is, what determines them? Is the initial voltage supposed to be
> 5V, and is that drawn from standby power in the power supply? That is,
> is the initial voltage independent of the motherboard?
Ok, first thiese are logic-level signals. <0.8V is false, >2.4V is true.
Power_OK needs to be false if the PSU is turned off and needs to go
to true after switch on (via front-button) within a very short time.
I see an additional possibility here: One or more of the power
lines might be shortened out. This would not even need to be
a problem with the mainboard. Quich check: Disconnect _everything_
except the mainboard and see whether you can get it to start then.
Also remove all add-on cards, except the video card.
A popular way to short out some PSU power lines is also to
mount the mainboard wrongly. To check that, remove the
Mainboard from the case and test in on a pice of cardboard or
other insulating material.
Arno
w_tom
wrote:
> Yes, I suspected as much. Despite what I might have just reported in my
> other message, with the PS power switch the green PS_ON wire reads
> 3.7V. When the front panel switch is depressed, the voltage drops to
> zero. If I push the front button again, but hold it for a few seconds,
> the voltage climbs back up to 2.5V. So it seems the latching mechanism
> on the MB is not entirely dead.
>
> The grey PWR_OK wire on the main power connector reads zero after the PS
> switch is off for a while and when it is switched back on. If I now
> press the front power button, it causes a little activity at this point,
> but it quickly settles back to the 0.004V. This seems wrong. I gather
> that the PWR_OK connection should be 5 V initially, and when the power
> on button is depressed, drop down to 3.3V. I'm not getting anything like
> that.
Important numbers are each wire voltage before and when the power
switch is pressed - as summarized in a previous post. For other's
benefit, what is happening here: isolated parts of that 'less than two
minute procedure' are eventually being performed.
What is called a 'latching circuit' is a power supply controller.
Power switch is only one controller input. Measuring that switch was
unnecessary if numbers from that two minute procedure were provided.
Again, numbers both just before and when power switch is pressed.
Irrelevant are voltages on green and gray wires without first
providing purple wire voltage (both before and when switch is
pressed). Purple wire provides power to the power supply controller
and can make that circuit work intermittently. Purple wire voltage
must measure above 4.87 VDC - and should be posted here because those
numbers may provide other information.
Before switch is pressed, green wire voltage must be well above 2.0
volts. When power switch is pressed, then green wire voltage drops to
below 0.8 volts. Last posts suggest that one part of the power
supply controller is working AND that the power switch is completely
OK. Two minute procedure made power switch testing unnecessary.
Moving on. Not provided were voltages on red, orange, and yellow
wires - each measured as the switch is pressed. Without what happens
on each wire and resulting numbers, useful information has been
withheld making the next paragraph an incomplete reply.
Gray wire must be something near zero before switch is pressed
(again, what was the actual number). When switch was pressed, what
happened to that gray wire voltage in the next 5 seconds? As I read
it, that voltage either starts to rise or never increases. Voltage
must be well above 2.4 volts within 2 minutes. That explains why
replacing a power supply would accomplish nothing. But without
specifically stating what gray wire voltage did and without voltages,
then too many possibilities still exist. No useful reply can be
posted without all those voltages in the two minute procedure.
I might bet one of the orange, red, or yellow wires will indicate
where to look next - what to suspect. But at least five other reasons
still exist which is why nothing more can be recommended until those
numbers are provided.
This paragraph is for others to learn from this discussion. A new
power supply solved nothing because the problem was not identified
first. A two minute procedure with a 3.5 digit multimeter that can
eliminate lost time and labor is "When your computer dies without
warning....." starting 6 Feb 2007 in the newsgroup alt.windows-xp
at:
http://tinyurl.com/yvf9vh
Among many things answered immediately in less than two minutes: is
the power switch good? Is the power supply defective? Are wires
properly connected? Nothing faster could have answered so much so
quickly.
rebel
the mobo are correct? Does it agree with the mobo manual?
Haines Brown
> Have you actually triple-checked that the connections from the front panel to
> the mobo are correct? Does it agree with the mobo manual?
You know, that worried me, too.
1. HDD LED. Here I'm not certain. The manual shows one pin as a square
and the other as a circle, and I was concerned that this implied
polarity, although it shouldn't. As of now I've an orange wire on the
square and a white wire on the circle. Could I be wrong?
2. Reset. OK
3. Power LED. Since the front panel connector accomodates a three pin
socket, and its wires are for pins 1 and 3, it could not be inserted the
"power LED" section of the front panel header, which presumes a two-pin
connector. So instead I put the Power LED connector to another header
that the manual calls "Alternative power/sleep LED", which calls for a
three pin connector. Was I wrong to do this?
4. Front panel audio connector. The wires are only long enough to reach
the side of the board nearest the front panel, while the mb header for
the "front panel audio connector" is on the far side of the MB. Since it
couldn't reach, I just left it off. Some day I may make or find an
extension so that it will reach.
--
Haines Brown, KB1GRM
rebel
Without the board, manual and pooter in front of me I can't really comment on
what you did. But my concern is that the GO button on the pooter may not be
connected to the right mobo header, which could well explain the issues you are
seeing.
Haines Brown
By bad. I simply skipped reporting on that. The front panel power button
goes to a "Power SW" connector, which I have connected to what the
manual calls the "On" pins of the header for front panel connections.
--
Haines Brown, KB1GRM
Haines Brown
voltages before, but I've jury-ricked the equivalent of a front power
switch and here are the results.
1. Purple standy-voltage. When PS is toggled on, this rises to 5.20 V.
2. Green PS_ON. Starts at 3.7V. When front panel power switch is
pressed, drops to 0.001 V.
3. Grey PWR_OK. Starts 0.004 V, when front panel power switch is pressed,
there's some flickering (no voltage to read), but quickly settles back
to 0.004V. This should have remained above 2.4V.
4. Although given the above, there may no point in these, I add them for
completeness.
Orange: 0.00V jumps to 0.80V, but drops right back to 0.00V
Red: Starts 0.00, momentarily jumps to 0.36 but drops back to 0.00V.
Yellow: Starts 0.00, flickers a bit, but remains at 0.00V.
--
Haines Brown, KB1GRM
Arno Wagner
Ok, the mainboard sends the right signals. Looks more and more
like a short somewhere. The orange/red/yellow measurements
migth indicate that +12V (yellow) is shortened out. Incidentially
the CPU voltage is generated from it and a MOS FET (as used in the
switching regulator) can short out a PSU without problem.
Arno
w_tom
wrote:
> Orange: 0.00V jumps to 0.80V, but drops right back to 0.00V
> Red: Starts 0.00, momentarily jumps to 0.36 but drops back to 0.00V.
> Yellow: Starts 0.00, flickers a bit, but remains at 0.00V.
Demonstrated is a point made in most every post. What was
speculated as "no point but added for completeness" was the only
useful information.
. Defect on 12 volts either in the load or in the power supply.
Finally we have taken the first useful step from the original post;
the original problem. Those with best knowledge, for the first time,
are able to post something useful because numbers from the two minute
procedure have been provided. Again, the point for all others who
would learn from this discussion. What was thought not important was
THE most important fact posted. Never filter information. Never
short information from those with technical knowledge. Those
'irrelevant facts' are the only infomation that made a useful reply
possible.
The problem is now isolated to a short list of components. What
uses the 12 volts? Again, keep following the evidence. Problem
exists on something that draws 12 volts.
Clearly not relevant now and not relevant earlier are: 1. HDD LED,
2. Reset, 3. Power LED, 4. Front panel audio connector, or other
missing connections. Those other suggestions demonstrated classic
wild speculation; not how one 'follows the evidence'. Evidence made
woefully obvious by following the evidence: the yellow wire voltage
is shorted out.
w_tom
> Without the board, manual and pooter in front of me I can't really comment on
> what you did. But my concern is that the GO button on the pooter may not be
> connected to the right mobo header, which could well explain the issues you are
> seeing.
What is suggested clearly is not correct - completely ignores the
useful facts provided by the OP - numbers. Voltage on the green wire
both before and when power switch is pressed means that wiring was
always perfectly fine. Also obvious from numbers (once they were
finally provided) was that a missing connection obviously is not the
problem - which is why solutions come from 'following the evidence'.
One reason for so many posts, so much speculation, and no useful
replies were no numbers. All this could have been answered in the
very next post had numbers from that two minute procedure been
provided as requested. Once the OP provided those numbers, quickly
the problem was isolated down to a few possible defects. The OP's
next step is to determine which of few items actually creates the
defect - by again 'following the evidence' and by not speculating that
"it could be this or could be that". Less than two minutes and the
resulting numbers would have made how many posts here completely
unnecessary? Follow the evidence is a concept that many have
difficulty grasping. And 'follow the evidence' is the only reason why
the OP finally has his first useful replies.
Arno Wagner
>> On Tue, 15 Apr 2008 10:25:08 -0400, Haines Brown
>> <bro...@teufel.hartford-hwp.com> wrote:
>>
>>>rebel <m...@privacy.net> writes:
>>>
>> Without the board, manual and pooter in front of me I can't really
>> comment on what you did. But my concern is that the GO button on the
>> pooter may not be connected to the right mobo header, which could well
>> explain the issues you are seeing.
> By bad. I simply skipped reporting on that. The front panel power button
> goes to a "Power SW" connector, which I have connected to what the
> manual calls the "On" pins of the header for front panel connections.
That is the correct way. Also your measurements on the power_on
line show that the power circuit works and is correctly connected.
Arno
Haines Brown
> On Apr 16, 4:20 pm, Haines Brown <bro...@teufel.hartford-hwp.com>
> wrote:
>> Orange: 0.00V jumps to 0.80V, but drops right back to 0.00V Red:
>> Starts 0.00, momentarily jumps to 0.36 but drops back to 0.00V.
>> Yellow: Starts 0.00, flickers a bit, but remains at 0.00V.
> . Defect on 12 volts either in the load or in the power supply.
> The problem is now isolated to a short list of components. What
> uses the 12 volts?
Some obvious questions arise from this.
It sounds like the yellow wire (+12V) supplies the motherboard with 12V
from the PS rail, but no one has said so explicitly yet. And, by analogy,
does the orange wire supply +3.3V and the red wire +5V to the MB from
the PS rails? So it seems, but I want to make sure.
Second question. None of the voltages on any of these wires was correct,
but neither was the reading on the grey wire (PWR_OK). What exactly is
the function of PWR_OK? Is it a signal to the PS to supply voltage to the
rails? How can one infer that this anomaly in the voltage on the PWR_OK
implies a possible short in one of the power circuits? I suspect that
the PWR_OK indicates whether the load on the three rail voltage circuits
is OK, and so its voltage is not a cause but an effect.
Third, since all three (presumed) rail voltages failed to go up
adequately, why single out the orange (+12V) circuit rather than one of
the other two?
Finally, assuming there's a short in the PS or the MB for the +12V
circuit, how to isolate it? Would you kindly critique this scenario?
a) I suppose one should first remove devices from the 12V peripheral
power to see what happens to the voltage on the grey wire.
b) If that shows no improvement, I suppose isolating the front panel
connections might be next. Obviously, for the purposes of test, I would
need to momentarily short out the pins on the front panel header that go
to the front panel ON switch. But I assume that the two LED circuits and
the Reset circuit can be left open.
c) Next, can I remove the 12V CPU power connector? Will doing so
interfere with the power readings I would expect from measurements at
the grey wire of the main power connector?
d) Finally, I suppose I should remove the MB from the case lest there be
some kind of mechanical short between motherboard and case. Are the
black ground wires in the main power connector sufficient to ensure a
common ground between MB and PS?
e) If there's still a problem I still don't know if it is in the PS or
the MB. However, I've another PS that seems to work, and so here I
suppose a swap experiment should be tried - an approach which up to now
you have justly criticized.
--
Haines Brown, KB1GRM
Arno Wagner
>> On Apr 16, 4:20 pm, Haines Brown <bro...@teufel.hartford-hwp.com>
>> wrote:
>>> Orange: 0.00V jumps to 0.80V, but drops right back to 0.00V Red:
>>> Starts 0.00, momentarily jumps to 0.36 but drops back to 0.00V.
>>> Yellow: Starts 0.00, flickers a bit, but remains at 0.00V.
>> . Defect on 12 volts either in the load or in the power supply.
>> The problem is now isolated to a short list of components. What
>> uses the 12 volts?
> Some obvious questions arise from this.
> It sounds like the yellow wire (+12V) supplies the motherboard with 12V
> from the PS rail, but no one has said so explicitly yet. And, by analogy,
> does the orange wire supply +3.3V and the red wire +5V to the MB from
> the PS rails? So it seems, but I want to make sure.
All correct. If you want to know more about the precise workings,
I acan also recommend reading the ATX standard, which is available
on the web.
> Second question. None of the voltages on any of these wires was correct,
> but neither was the reading on the grey wire (PWR_OK). What exactly is
> the function of PWR_OK? Is it a signal to the PS to supply voltage to the
> rails? How can one infer that this anomaly in the voltage on the PWR_OK
> implies a possible short in one of the power circuits? I suspect that
> the PWR_OK indicates whether the load on the three rail voltage circuits
> is OK, and so its voltage is not a cause but an effect.
Bu asstertint pwr_ok, the PSU tells the mainboard that the
voltages have all reached nominal levels an that it is
ok now to start the CPU.
> Third, since all three (presumed) rail voltages failed to go up
> adequately, why single out the orange (+12V) circuit rather than one of
> the other two?
Yellow. Becasue it showed the least signal. The others could also
be shortened.
> Finally, assuming there's a short in the PS or the MB for the +12V
> circuit, how to isolate it? Would you kindly critique this scenario?
> a) I suppose one should first remove devices from the 12V peripheral
> power to see what happens to the voltage on the grey wire.
> b) If that shows no improvement, I suppose isolating the front panel
> connections might be next. Obviously, for the purposes of test, I would
> need to momentarily short out the pins on the front panel header that go
> to the front panel ON switch. But I assume that the two LED circuits and
> the Reset circuit can be left open.
> c) Next, can I remove the 12V CPU power connector? Will doing so
> interfere with the power readings I would expect from measurements at
> the grey wire of the main power connector?
> d) Finally, I suppose I should remove the MB from the case lest there be
> some kind of mechanical short between motherboard and case. Are the
> black ground wires in the main power connector sufficient to ensure a
> common ground between MB and PS?
> e) If there's still a problem I still don't know if it is in the PS or
> the MB. However, I've another PS that seems to work, and so here I
> suppose a swap experiment should be tried - an approach which up to now
> you have justly criticized.
That is not the best approach. If there is a short, you
risk killing some part whenever you pwer up.
Best approach:
1) Remove everything from the PSU and short pwoer_on to grpund.
Measure the voltages on yellow, red and orange. If ok,
procceed.
2) Connect mainboard with only video card in it. Try starting
with push-button. Measure voltages.
3) Add add-in cards and repeat.
4) Connect pheripherials one by one and repeat for each.
Stop when you have found the point where it does not work
anymore. The last component connected is likely the problem.
Arno
w_tom
provided by Arno Wagner. Demonstrated is another reason to perform
that two minute procedure immediately - to ask or learn from questions
asked by Haines Brown.
Green wire orders power supply on. If voltages don't obtain correct
values in less than two seconds, then gray wire says power is bad.
Power supply controller turns off green wire; a closed loop system.
Fundamentals of this design are found in the very first IBM PC which
begs another question: why do A+ Certified Techs not know this?
However some supplies violate this color spec. Therefore, also
provided was:
http://www.hardwarebook.net/connector/power/atxpower.html
Also listed are minimum voltages: 3.23, 4.87, and 11.7. These
numbers are in agreement with ATX specs even though numbers are not
same. Above numbers take into account other factors such as how
multimeters typically measure voltage.
Singled out was the yellow (+12 V) because you said that voltage did
not attempt to rise. How to find the short? Another simple way is
to disconnect devices that consume +12V until +12 volts does rise OR
until only one component remains - motherboard. For example, a short
could be inside a power supply adjacent to the CPU. With a meter, 12
volts may rise slightly where the connector attaches to motherboard,
but not at all at the CPU power supply chips. So with CPU connector
disconnected, then a short in the CPU power supply area would mean a
slight voltage rise where the 20 pin connector attaches and a less or
zero voltage rise as 12 volts gets closer to the CPU.
That previous paragraph demonstrates one way to 'follow the evidence'
- trace the problem to its source without speculating.
A most likely reason for CPU power supply failure is adjacent
electrolytic capacitors. Inspection of anything that connects to 12
volts can be informative. Again, first collect facts before changing
anything. Electrolytics can be replaced for dollars - an example of
how to save money and how to learn by using a meter to ''follow the
evidence'. These are suggestions - examples of how to trace that
failure to its source. How you do it means first collect facts, make
a decision, change a simplest thing, then measure again.
Front panel connections are not relevant to your problem - don't
consume +12 volts. Notice the many peripherals that connect to yellow
wires. These peripherals may use 12 volts.
Also not mentioned before but also relevant: purple wire rises to
5.20 volts. Although not excessive, it is rather too close to maximum
- implying +12 volts may be shorted to the (purple wire) +5VSB
traces. Another fact available only because numbers were provided.
For others - another example of why answers must include numbers to
empower replies from knowledgeable posters.. No numbers means replies
only from those who would only speculate.
Removing a motherboard is another possible test. But don't change
too much at once. Best is to eliminate one possibility, then test.
Change minimum with each test. Therefore the one change that does
eliminate that 'short' only identifies the most minimal number of
suspects.
Motherboard is only removed after 12 volts still does not appear
with only motherboard (nothing else) connected. Yes, the short could
even be a stray metal fragment. That fragment would never be located
if you change too much too fast. Change the power supply and it might
temporarily dislodge the fragment - which shorts again later. First
you want to find the problem. Fixing a problem is the second step.
Same principle demonstrates why that two minute procedure is so
powerful; so informative. Measurements are taken without moving,
disconnecting, or swapping anything. Best way to find a problem
starts by collecting facts while changing nothing. Then change only a
small thing; and measure again.
On Apr 17, 10:30 am, Haines Brown <bro...@teufel.hartford-hwp.com>
wrote:
Haines Brown
progress. The bottom line is that the 12V bus reads correctly if the MB
is removed from the case; not if it is inserted. I probably have a short
somewhere, perhaps something between the MB and its back plate. The next
step will be to remove it from that back place, but that brings up a
question: Is the ground provided by the back main power connector
sufficient to ensure system ground for testing purposes?
In going over the steps to isolate the cause of an undervoltate in one
of the three power busses. I've had to expand or edit your helpful
advice a bit, and perhaps easiest would be for me to provide the result
and have you spot where I go wrong:
1. To isolate the problem to the MB+PSU, remove all cards except the
video card from the MB [why not the video card?], CPU power connector,
all front panel connectors and optionally the power switch, and all PSU
power going to any peripheral drives. Leave connected only the MB main
power connector. Briefly short the PWR_On (green, pin 14) to ground, or
the pins on the front panel header that would connect to the power
switch, or push the power switch if it is connected [strange, this test
condition seems to have made pressing the power switch unnecessary, for
the 12V bus went right up an acceptable value as soon as I turn the PSU
on. Given that I have a jury-rigged power switch, I'll have to make sure
it is not accidentally grounded]. Measure the voltage on the three rail
wires: orange, red and yellow, which should rise to at least +3.23V,
+4.87V, and +11.7V respecively within a second.
2. If these voltages fail to rise properly, remove the MB and PSU to see
if powering up the MB external to the case corrects the problem. [I find
here that since this removal did correct the problem, to determine that
the MB was the culprit, I had to resort to a PSU swap. Is there a better
way?]
3. Or if these voltages do rise properly, then we are in a position to
isolate the problem. Restore the CPU power connection to see if it rises
to about 12V after the power switch is pressed. If there's a CPU short,
there should be only a very slight rise in voltage at the CPU power
connector when the power switch is pushed, but no rise close to the
short itself [I'm not sure quite how to implement this as far as
identifying and accessing test points. My MB, although new, was built
when the electrolytic problem first came came up, and so of course,
inspecting them was an obvious thing to do. I'm not confident that part
replacement is all so easy on a multi-level printed board.]
4) Restore the cards back to the MB and repeat the voltage checks
above. If the voltages are OK, presumably the problem is not that of any
particular card.
5) Reconnect the peripherals one by one, checking voltages each time.
--
Haines Brown, KB1GRM
w_tom
wrote:
> I've benefited from your and W_tom's advice and have made some
> progress. The bottom line is that the 12V bus reads correctly if the MB
> is removed from the case; not if it is inserted. I probably have a short
> somewhere, perhaps something between the MB and its back plate. The next
> step will be to remove it from that back place, but that brings up a
> question: Is the ground provided by the back main power connector
> sufficient to ensure system ground for testing purposes?
Best connection between a motherboard and chassis is only at one
point. Then static electric shocks (et al) do not conduct on paths
through the motherboard - no computer crash. Single point grounding
is one technique used to make electronics so robust. However, many
manufacturers don't bother.
Best place to locate the single point ground near power supply
connector and peripheral cards. One way to see this advantage is with
computer is on a non-conductive surface such as a glass table. Wood
or other table tops are too electrically conductive. Create (painful)
static electric discharges to various corners of a computer's
chassis. Computer that crashes can be fixed by grounding the
motherboard to chassis only at a single point. Then static electric
currents do not pass through motherboard ground plane. Then computer
will be more stable in normal operation.
Each standoff hole should have a copper ring around the hole so that
conductive standoffs do not contact the board on a PC trace and so
that fiberglass board is not damaged by a tightened standoff. If the
standoff contacts a PC trace, eventually that standoff cuts through
the green solder mask to short to that trace.
Haines Brown
plugging cards back in, and then the system dies. I remove the last
item, and that does not correct anything, and I end by completely
deconstructing everything, and then it might work (proper voltages on
the three power rails) or it might not work; somewhat randomly.
At this point I'm back to measuring the standby voltage (purple wire on
the main power connector). For a while it was consistently reading
5.17V, but now reads 4.25V a few times, and without doing anything but
turning the PSU back on, and finally 0.23V and 0.01V. Seems like a
progressive short.
I removed the motherboard to check for a possible short between the
board and aluminum back plane, but nothing even close.
I get the feeling there's an intermediate short somewhere on the
motherboard, and I was being confused because sometimes the standby
voltage was OK and sometimes marginal,and sometimes way off. Moving and
poking components or areas on the MB does not have any effect on the
standby voltage readings.
My inclination would next be to pull the main power connector and
measure the standby voltage arriving at the connector from the PSU. Am I
correct to assume that without any connections, the standby voltage
should be over 4.87V as soon as the PSU switch is turned on, and at
least for a brief measurement, I need not worry about the absence of any
load on the PSU?
I'm also curious as to the normal resistance to ground of the standby
voltage pin on the motherboard header, but I suspect I'll not be able to
find that out. The board is under warranty and it would be tempting to
just return it, but I expect you will tell me that's just a "cop out" ;-)
--
Haines Brown, KB1GRM
Arno Wagner
> plugging cards back in, and then the system dies. I remove the last
> item, and that does not correct anything, and I end by completely
> deconstructing everything, and then it might work (proper voltages on
> the three power rails) or it might not work; somewhat randomly.
> At this point I'm back to measuring the standby voltage (purple wire on
> the main power connector). For a while it was consistently reading
> 5.17V, but now reads 4.25V a few times, and without doing anything but
> turning the PSU back on, and finally 0.23V and 0.01V. Seems like a
> progressive short.
> I removed the motherboard to check for a possible short between the
> board and aluminum back plane, but nothing even close.
> I get the feeling there's an intermediate short somewhere on the
> motherboard, and I was being confused because sometimes the standby
> voltage was OK and sometimes marginal,and sometimes way off. Moving and
> poking components or areas on the MB does not have any effect on the
> standby voltage readings.
> My inclination would next be to pull the main power connector and
> measure the standby voltage arriving at the connector from the PSU. Am I
> correct to assume that without any connections, the standby voltage
> should be over 4.87V as soon as the PSU switch is turned on, and at
> least for a brief measurement, I need not worry about the absence of any
> load on the PSU?
Yes. In fact most modern PSUs run ok without load, the voltages may
just be a bit out of spec.
> I'm also curious as to the normal resistance to ground of the standby
> voltage pin on the motherboard header, but I suspect I'll not be able to
> find that out. The board is under warranty and it would be tempting to
> just return it, but I expect you will tell me that's just a "cop out" ;-)
I would advise you to return it. A non-persistent short can be
_very_ hard to find.
Arno
w_tom
wrote:
> I'm really getting nowhere. When voltages seem within range, I start
> plugging cards back in, and then the system dies. I remove the last
> item, and that does not correct anything, and I end by completely
> deconstructing everything, and then it might work (proper voltages on
> the three power rails) or it might not work; somewhat randomly.
>
> At this point I'm back to measuring the standby voltage (purple wire on
> the main power connector). For a while it was consistently reading
> 5.17V, but now reads 4.25V a few times, and without doing anything but
> turning the PSU back on, and finally 0.23V and 0.01V. Seems like a
An obvious question possibly overlooked. Are you removing AC power
cord every time before disconnecting anything? Purple wire voltage
makes it obvious why power cord must always be disconnected before
making changes. If not, semiconductors may start conducting current
through PN junctions that should never conduct current.
Haines Brown
the voltage that was flakey. So I replaced the PSU and the voltage is
now +5.20V with the main power connector attached to the MB.
So I'm going to reassemble step by step with my fingers crossed.
--
Haines Brown, KB1GRM
Haines Brown
> An obvious question possibly overlooked. Are you removing AC power
> cord every time before disconnecting anything? Purple wire voltage
> makes it obvious why power cord must always be disconnected before
> making changes. If not, semiconductors may start conducting current
> through PN junctions that should never conduct current.
Yes, however your point raises a question. I get my standby voltage only
when the PSU switch is toggled on, not when the power connector is
first attached. The reason I am careful to pull the line voltage is that
the capacitors in the PSU deliver voltage for about ten seconds after I
switch the PSU off.
--
Haines Brown, KB1GRM
Arno Wagner
Arno Wagner
>> An obvious question possibly overlooked. Are you removing AC power
>> cord every time before disconnecting anything? Purple wire voltage
>> makes it obvious why power cord must always be disconnected before
>> making changes. If not, semiconductors may start conducting current
>> through PN junctions that should never conduct current.
The PN-statement is utter nonsens. The thing is that you can get
reverse current only when you have two different active voltages,
which you do not have here. Also "current through PN junctions" is
not a problem in itself, regardless of direction.
> Yes, however your point raises a question. I get my standby voltage only
> when the PSU switch is toggled on, not when the power connector is
> first attached. The reason I am careful to pull the line voltage is that
> the capacitors in the PSU deliver voltage for about ten seconds after I
> switch the PSU off.
Not needed. Switching off on the back is quite enough. The reason
why the PSU should be switched off by switch or pulling the cord
is that a) you may accidentially short something and b) you may
accidentially switch it on otherwise.
Arno
w_tom
> The PN-statement is utter nonsens. The thing is that you can get
> reverse current only when you have two different active voltages,
> which you do not have here. Also "current through PN junctions" is
> not a problem in itself, regardless of direction.
Arno - review an 8080 microprocessor. It, like most digital ICs,
needed -5 volts (a second voltage) so that PN junctions do not
conduct. Starting with the 8085 and Z-80, an onboard DC to DC
converter has been standard - that second voltage created inside the
IC. What happens when ground disconnects first? Two voltages exist
that can cause non-conductive PN junctions to conduct current.
Another symptom created by the same problem was called an SCR effect.
Hot popping (disconnecting something when voltage exists) is
problematic if the ground disconnects before any voltage, for reasons
above and for other unlisted reasons.
You said, "you can get reverse current only when you have two
different active voltages". That onboard voltage converter creates a
second active voltage. Just one of many reasons why 'hot popping' may
create failures. Since earliest days of CMOS (1960s), semiconductors
have been designed more resistant to this effect. Problems no longer
occur anywhere near as frequently. However the symptoms posted by
Haines can be explained by 'hot popping' (and by other reasons).
Haines' post says he disconnects power cord before removing or
installing anything. That is good. Nothing is as reliable as
disconnecting that power cord.
Haines' post also says something unacceptable. "I get my [+5VSB -
purple wire] voltage only when PSU switch is toggled on; not when the
power connector is first attached." Toggled on how? Toggled on when
power switch is pressed? If using a power switch (located on an ATX
power supply), that switch is not for power on. That ATX supply
switch should be left on for all tests. The only 'power on' switch is
a push button switch on cmputer's front panel.
Power supply controller can see that front panel switch only when
+5VSB exists. Power supply controller without +5VSB cannot respond to
the only switch that creates power on.
+5VSB (purple wire) voltage should always appear and always be in
spec when power cord is connected. It should only disappear when
power cord is removed. Computer should not power on when power cord
is connected. Computer should only power on when front panel push
button is pressed after power cord is connected. That green wire
voltage should respond (drop to near zero) only when front panel
switch is pressed. Green wire voltage (during power off) must exist
only when purple wire (+5VSB) voltage exists.
Again, those purple wire voltage changes imply the +5VSB may be
shorted to another voltage - not acceptable. That +5VSB voltage
should never change when green wire (Power On#) voltage changes. That
purple wire voltage should remain same (within less than 0.1 volts) no
matter how many or how few peripherals are connected to motherboard.
Bob Eager
> The PN-statement is utter nonsens. The thing is that you can get
> reverse current only when you have two different active voltages,
> which you do not have here. Also "current through PN junctions" is
> not a problem in itself, regardless of direction.
Indeed.Doesn't he go on and on? Don't get w_tom started on lighting
strikes...
--
Bob Eager
begin 123 a new life...take up Extreme Ironing!