odd shut-down issue


N

Nil

My PC has started yesterday doing something weird: when I shut down
Windows XP, it seems to shut down normally... except where the computer
itself used to shut completely off, now the fans continue to run and
the pilot light on the front continues to glow. The reset button
doesn't respond, which would be the expected behavior if it were
completely shut off. Pressing the power button starts bootup as normal,
and Windows starts normally, no complaints about being shutdown unclean
ly. I have to turn the power supply power switch off to stop the fan,
and when I turn that switch back on, the fans start right up, even
before I press the front power button.

I haven't touched the hardware in weeks in any way that I think would
be related to this symptom. Can anyone explain what's going on and/or
suggest a remedy?


Windows XP Pro SP2
Intel Core 2 Duo E6600 Conroe 2.4GHz LGA 775 Processor
Mitsumi 1.44MB 3.5" Internal USB 2.0 digital card reader
with Floppy Drive
LG GH22NS70 SATA DVD burner
ASUS DRW-2014L1t LS SATA DVD burner
Western Digital Caviar SE16 WD2500KS 250GB 7200 RPM SATA hard disk
ASUS P5L-VM 1394 Socket T (LGA 775) Intel 945G Micro ATX Intel
Motherboard
OCZ Platinum Revision 2 2GB 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400)
Dual Channel Kit Desktop Memory Model
Thermaltake W0101RU ATX 12V 2.0 Version 550W Power Supply
Antec P180 case
 
Ad

Advertisements

P

Paul

Nil said:
My PC has started yesterday doing something weird: when I shut down
Windows XP, it seems to shut down normally... except where the computer
itself used to shut completely off, now the fans continue to run and
the pilot light on the front continues to glow. The reset button
doesn't respond, which would be the expected behavior if it were
completely shut off. Pressing the power button starts bootup as normal,
and Windows starts normally, no complaints about being shutdown unclean
ly. I have to turn the power supply power switch off to stop the fan,
and when I turn that switch back on, the fans start right up, even
before I press the front power button.

I haven't touched the hardware in weeks in any way that I think would
be related to this symptom. Can anyone explain what's going on and/or
suggest a remedy?


Windows XP Pro SP2
Intel Core 2 Duo E6600 Conroe 2.4GHz LGA 775 Processor
Mitsumi 1.44MB 3.5" Internal USB 2.0 digital card reader
with Floppy Drive
LG GH22NS70 SATA DVD burner
ASUS DRW-2014L1t LS SATA DVD burner
Western Digital Caviar SE16 WD2500KS 250GB 7200 RPM SATA hard disk
ASUS P5L-VM 1394 Socket T (LGA 775) Intel 945G Micro ATX Intel
Motherboard
OCZ Platinum Revision 2 2GB 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400)
Dual Channel Kit Desktop Memory Model
Thermaltake W0101RU ATX 12V 2.0 Version 550W Power Supply
Antec P180 case

I checked Newegg reviews, for your power supply, and there are
no reports in a quick check (didn't read all reviews) for PS_ON#
failing to work and turn off power. There are reports of the power
supply outright failing, but that's different.

This is the first report of your symptoms I could find on
vip.asus.com . This person was using Ubuntu, and fans continued
to run after the PC was shut down. The OP here reports a flashing
drive LED as well. The flashing to me, suggest there is more than
a "coma" situation - something could be "looping", but exactly
what isn't clear.

http://vip.asus.com/forum/view.aspx...id=1&model=P5L-VM+1394&page=1&SLanguage=en-us

The OP in that thread, didn't manage to cure the power down symptoms
using BIOS settings. It's probably something related to PS_ON#,
and coming from the motherboard.

You could try:

1) Ubuntu LiveCD. Or better still, a USB stick, as that would
allow disconnecting the CDROM as well.
2) Disconnect hard drive both power and data lines.
To eliminate any (loading) effect.
3) See if Ubuntu listens to request to shut down.

The Asus warranty on your motherboard, will be related to the
manufacture date. There should be a sticker on the motherboard
box, for example

Serial No: 85MOAIxxxxxx

The first digit is the year, 2008, the second digit is month of May.
(Possible month values are 0123456789AB.) My warranty expired on
2011 May (3 years). I didn't own the motherboard for 3 years
however - half the warranty was shot, while the motherboard
stayed in some warehouse. So in terms of calendar time, I had
warranty protection for about a year and a half or so. Managing
warranty with the serial number, reduces fights over using
purchase receipt and so on - it also helps bound the time the
manufacturer must keep stocks of replacement motherboards
(not that they give out new motherboards very often).

If this happened to my motherboard, and the warranty was expired, I'd
probably start with some tests of PS_ON# and see if levels and drive
were OK. A failure to turn off, means the open collector pullup resistor
(to +5VSB), can't be pulling the signal sufficiently above 2.0 to 2.4V
or so. So something is still pulling down on the signal. You can take a
multimeter, and plug the red voltage lead, into the nylon shell of the
main connector, and monitor PS_ON#. The pinout for PS_ON# is
available in one of the formfactors ATX standards, like this one.

http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf (page 37)

You'd be interested in the levels on the green wire. When the supply
is off, the open collector driver should let go, and the voltage
should swing up to 5.0V (or whatever voltage level is available
via +5VSB, which supplies that voltage). When the power supply is
turned on, the signal should swing down to 0.4 to 0.8V or so. That's
the "active" state of the signal. If you attempt to shut down, and
the voltage is around 1.5 to 2V or so, that is "no man's land" and
isn't a valid logic level. Some supplies, when faced with such a level,
will be "half turned on" and power supply outputs will be weak, and
the supply can then slowly oscillate if it wants.

Paul
 
N

Nil

Sounds like a power management setting.

I don't think that's it. If it were, the fans wouldn't spin up after
the power supply was shut off. Also, after I boot up with a bootable CD
(both Knoppix and BartPE) the computer will still only shut down to the
condition I describe. I feel confident that it's not a Windows thing.
 
N

Nil

You could try:

1) Ubuntu LiveCD. Or better still, a USB stick, as that would
allow disconnecting the CDROM as well.
2) Disconnect hard drive both power and data lines.
To eliminate any (loading) effect.
3) See if Ubuntu listens to request to shut down.
The first digit is the year, 2008, the second digit is month of
May. (Possible month values are 0123456789AB.) My warranty expired
on 2011 May (3 years).

My motherboard is almost 6 years old. I assume the warranty expired
long ago.
If this happened to my motherboard, and the warranty was expired,
I'd probably start with some tests of PS_ON# and see if levels and
drive were OK. A failure to turn off, means the open collector
pullup resistor (to +5VSB), can't be pulling the signal
sufficiently above 2.0 to 2.4V or so. So something is still
pulling down on the signal. You can take a multimeter, and plug
the red voltage lead, into the nylon shell of the main connector,
and monitor PS_ON#. The pinout for PS_ON# is available in one of
the formfactors ATX standards, like this one.

http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_b
r2.pdf (page 37)

So, it sounds like you suspect either a power supply or motherboard
(partial) failure, right? I guess I need to take some quality time to
systematically disconnect things and boot up in various combinations
until I narrow it down more.

I did discover after my first post that the CD drive is still alive
after the machine is (incompletely) powered down. I assume now that all
the peripherals are, even though the OS has halted.
 
P

Paul

Nil said:
alt.comp.hardware.pc-homebuilt:
So, it sounds like you suspect either a power supply or motherboard
(partial) failure, right? I guess I need to take some quality time to
systematically disconnect things and boot up in various combinations
until I narrow it down more.

I did discover after my first post that the CD drive is still alive
after the machine is (incompletely) powered down. I assume now that all
the peripherals are, even though the OS has halted.

I would expect if +5VSB was low (it should be 5V +/- 5%), there
would be other symptoms. Based on what I see so far, the problem
is PS_ON#, which is driver by a motherboard chip. The signal does
not come from the front switch directly. There is silicon in the
path, to condition and buffer the signal.

On an open collector interface, there are two ingredients. A
transistor on the driving end, able to pull down to logic zero.
And a power source (+5VSB) and a pullup resistor, pulling the
signal towards 5V (a logic one).

If the transistor on the motherboard broke, you wouldn't be
able to turn the power supply on. If the pullup resistor broke,
you wouldn't be able to turn the supply off. You can also
have conditions which are "half and half", where the transistor
is sorta half on, and the pullup resistor can't really do anything
to assist.

The advantage of open collector circuits, is they're "short
circuit safe". For people who can no longer get their PC to
turn on, you can place a shorting plug between ground (black)
and PS_ON# (green). That would cause the power supply to come
on, in cases where the transistor on the motherboard is broken.
But forcing it, also prevents THERMTRIP from turning off the
system, if the processor is overheating. Such a "hack" is OK
if you know the consequences, and don't really care about
the computer any more.

On the converse side, open collector is not "5V safe". You
can't short PS_ON# to +5VSB, to "stop" the computer. If the
motherboard transistor turns on, it will then be attempting
to "sink" the entire supply, and smoke comes out. The transistor
probably isn't rated for a collector current of 2 amps.

So you're right to surmise the problem is between motherboard
and power supply, but simple minded testing doesn't allow
fault isolation to the nearest component. You have to take
things apart, and test and characterize each end, to gain
more info.

For example, with a little thought, you could probably
come up with some test cases for PS_ON#, while the
power supply is sitting alone on the bench. If you look in
that ATX power supply spec, they have a terse description of
the input characteristic of PS_ON#. Based on that, a
person can arrange a few test cases to check whether the
power supply threshold is where it should be, and whether
the pullup resistor is present or not.

To test the motherboard end, how I'd do that, is buy a
20 pin to 20 pin ATX extender cable, then cut the green
wire on the extender cable. (Using an extender cable and
hacking it, prevents damage to the DUT.) With the wire cut,
that prevents PS_ON# from connecting to the supply. You can
then make measurements of voltages and current flow, to
get a better idea as to whether the motherboard is "driving"
or not, when attempting to make a logic zero. For example,
with the power supply end of the green wire dangling, you
can put the multimeter on "milliamps", move the red wire
to the current measurement hole on the meter, and measure the
amount of current "sinking" to ground. Based on that, you can
tell whether the pullup resistor is intact or not.

The easiest test case, is swapping power supplies, then discovering
the problem remains. If so, it's more likely to be the motherboard.
But if you have the power supply extender cable, you can
cook up other tests, by having some options for cutting things.

Refs:

http://en.wikipedia.org/wiki/Open_collector

*******

And yes, if the fans are still running, the CDROM would be
completely powered (getting +5V and +12V) and the eject
button on the front should work.

Paul
 
Ad

Advertisements

N

Nobody > (Revisited)

Typically, there are power management settings in the BIOS too.

And they can be a bitch.
Horror stories over the S1-S3-Auto settings abound.

This same problem can happen when the +5Vstdby rail in the power supply
gets flakey on shutdown.


--
"Shit this is it, all the pieces do fit.
We're like that crazy old man jumping
out of the alleyway with a baseball bat,
saying, "Remember me motherfucker?"
Jim “Dandy” Mangrum
 
J

John Doe

Nobody > (Revisited) said:
John Doe wrote:

And they can be a bitch.
Horror stories over the S1-S3-Auto settings abound.

I posted one of those horror stories here.
This same problem can happen when the +5Vstdby rail in the power
supply gets flakey on shutdown.

Yup, otherwise, she is probably looking at a hardware problem.

So start swapping hardware.
--
 
J

Jure Sah

Hi,

Problem sounds vaguely reminiscent of the BIOS incorrectly interpreting
the poweroff signal. In the early days before ATX support was
widespread, some computers had their suspend / poweroff signals
backwards and caused the problem you described.

However as the problem started now and worked before, this would seem to
be unlikely. Assuming it's not a PSU problem (easily diagnosed: Swap out
the PSU), it is still possible the BIOS is interpreting the signals
wrong (after all, ACPI firmware implementation is a nightmare: "ACPI is
a complete design disaster in every way. But we're kind of stuck with
it. If any Intel people are listening to this and you had anything to do
with ACPI, shoot yourself now, before you reproduce." -- Linus
Torvalds). Try disconnecting the power completely for a few minutes
(press the power button while the power is unplugged to shorten the wait
period), plug it back in and see if that helped.

Try clearing the BIOS settings.

Try upgrading the BIOS.

LP,
Jure
 
N

Nil

You could try:

1) Ubuntu LiveCD. Or better still, a USB stick, as that would
allow disconnecting the CDROM as well.
2) Disconnect hard drive both power and data lines.
To eliminate any (loading) effect.
3) See if Ubuntu listens to request to shut down.

I tried this, disconnecting all peripherals except the CD drive and booted
up with a Kubuntu live/install CD. Shutdown still behaves the same as it
does with Windows.
If this happened to my motherboard, and the warranty was expired,
I'd probably start with some tests of PS_ON# and see if levels and
drive were OK. A failure to turn off, means the open collector
pullup resistor (to +5VSB), can't be pulling the signal
sufficiently above 2.0 to 2.4V or so. So something is still
pulling down on the signal. You can take a multimeter, and plug
the red voltage lead, into the nylon shell of the main connector,
and monitor PS_ON#. The pinout for PS_ON# is available in one of
the formfactors ATX standards, like this one.

http://www.formfactors.org/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf
(page 37)

You'd be interested in the levels on the green wire. When the
supply is off, the open collector driver should let go, and the
voltage should swing up to 5.0V (or whatever voltage level is
available via +5VSB, which supplies that voltage). When the power
supply is turned on, the signal should swing down to 0.4 to 0.8V
or so. That's the "active" state of the signal. If you attempt to
shut down, and the voltage is around 1.5 to 2V or so, that is "no
man's land" and isn't a valid logic level. Some supplies, when
faced with such a level, will be "half turned on" and power supply
outputs will be weak, and the supply can then slowly oscillate if
it wants.

OK, I finally found time for this test. If I understood you correctly,
with the power supply on and disconnected from the motherboard,
I can measure the voltage between the green wire 16 and black common
(I used pin #1, but can I assume I could have used any of of the 8?)
The voltage at this point measures about 2V, which you describe as
"no man's land" / not "a valid logic level".

So, is that enough evidence to conclude that the power supply is bad?
I would like to swap in a known-good one, but the only one I have available
is in use in a live computer, so I have to wait for a time when I can
temporarily cannibalize it without upsetting someone.
 
N

Nil

However as the problem started now and worked before, this would
seem to be unlikely.

I tend to doubt that its a BIOS thing, as I haven't touched the setting
in quite some time, and it's been working OK for years. But something
happened suddenly a few days ago without any action on my part.
Assuming it's not a PSU problem (easily diagnosed: Swap out the PSU)

I have a feeling this is the source of the problem. I'll swap in
another power supply when I can get my hands on it!
Try clearing the BIOS settings.

Try upgrading the BIOS.

Maybe I'll try that tonight.

Thank you for your thoughts on this. I do hope it's not the
motherboard. Replacing that would be a hassle. Replacing the PS would
be a lot simpler.
 
Ad

Advertisements

J

Jure Sah

....
I doubt it's the BIOS, since it's worked for years. I'd vote for
hardware. I'm not sure about the PS but if anything is mis-interpreting
the shut down, I'd vote for MoBo. I hope it's the PS.

This problem is mostly in laptops where there is extra power management
sophistication in the BIOS (power state control of other devices via
SMBus) -- I can only assume it is possible that PCs suffer from the same
problem:

BIOS holds some state information regarding power management across
shutdowns (S0), which may be screwed up and causing a problem,
especially if there is an actual BIOS bug in the equation. That ACPI
nightmare Linus spoke of referred to a special protocol between the
operating system and the BIOS for changing power states (standby in
particular), where the Operating System interprets pseudo-code provided
by the BIOS and uses it to control the power state of various devices --
meaning it is possible for the operating system to do something wrong
resulting in some very low-level irregularities that survive reboots.

The example on laptops I had were USB ports that refused to power up
until the BIOS state information was blanked by removing the laptop
battery for a little while. BIOS upgrade fixed it.

LP,
Jure
 
Ad

Advertisements

P

Paul

Nil said:
I tried this, disconnecting all peripherals except the CD drive and booted
up with a Kubuntu live/install CD. Shutdown still behaves the same as it
does with Windows.


OK, I finally found time for this test. If I understood you correctly,
with the power supply on and disconnected from the motherboard,
I can measure the voltage between the green wire 16 and black common
(I used pin #1, but can I assume I could have used any of of the 8?)
The voltage at this point measures about 2V, which you describe as
"no man's land" / not "a valid logic level".

So, is that enough evidence to conclude that the power supply is bad?
I would like to swap in a known-good one, but the only one I have available
is in use in a live computer, so I have to wait for a time when I can
temporarily cannibalize it without upsetting someone.

If the voltage on PS_ON# green wire is 2V, that really isn't high enough.
With nothing connected to the main ATX power cable, the power supply
is in control of that voltage value.

Based on finding that value, I'd try another supply.

The pullup resistor inside the power supply, is connected to +5VSB.
If all was "normal", an open main cable sitting on the desk, would have
the +5VSB pin sitting at 5.0V (+/-5%) and the PS_ON# wire should be
the same value or very close to it.

+5VSB
|
Pullup
Resistor
|
---> PS_ON# >------+------------- Control portion of
/ ATX supply
Mobo ___|/
Cct |\
\
GND

The motherboard pulls the signal down, with it's open collector output
transistor. When the transistor is turned off, the pullup resistor is
supposed to pull PS_ON# back up to 5.0 volts. If you find 2.0 volts
or perhaps 1.5V etc., this is a non-valid voltage value. The logic
voltages are either very high (>2.4V) or very low (<0.8V). The separation
between the two voltage values, provides some noise immunity. The
accepted input and output values, are listed in the ATX spec (with
separate specs for input threshold, and output driven values).

Try another supply. You may get lucky.

I like to keep a spare supply on hand, but with supplies dying on me,
and "temporary builds", I never seem to have one when I need it :)
I'd have to pull the supply out of my most pathetic computer and use
that :)

Paul
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top