PC (ATX power supply) won't start from soft-power switch

if shorting the green wire to any black wire on the ATX24 connector starts the PSU, I'd say that points to a motherboard issue.

n.b. the green wire is PS_ON which is pin 16 not 14.

My mistake. I counted the pins wrongly (*). Yes, the green wire.

Am I right that the motherboard needs standby power to sense the power switch being closed at the header, in order that it can drive PS_ON (pin

I'm now wondering whether the PSU is sometimes not producing +5V SB (pin

In the meantime I'll fit a replacement spare PSU and try that after leaving it with no mains.

Intermittent problems are always a bugger to known whether you've fixed them...

(*) I've worked out my mistake: the pins are numbered differently depending on whether the connector has the extra 2x2 pins (+12, +3.3,

I rather thought that the power button from case to the motherboard would just be connected straight to PSU pin16 to short it to ground, rather than going through any active circuitry?

I thought ATX20 connectors were mostly extinct ...

You can never know. Perfect trouble free operation is simply an intermittent fault that hasn't shown up yet...

The motherboard power button is 'soft' - it sends an interrupt to the embedded controller on the motherboard. That handles the differences between long presses and short presses, etc. It remembers the state so that the button doesn't need to latch.

The EC then tells the PSU what to do via a 'hard' PS_ON: if it's connected to ground the PSU powers up, if it's not grounded the PSU turns off.

If the EC is not correctly powered it won't be able to power up the board. So lack of 5V_standby is a problem.

(if you try to use the 'soft' button in a 'hard' way, eg by connecting a rocker switch to it, you may get a boot loop. Board powers up, then sees the 'power button being held down for 30 seconds' and turns off. Then it sees the power button is pressed and powers up, then it powers down after 30 seconds... repeating until the rocker is flipped off. Then it stays either on or off depending on what part of the cycle it happened to be in. The same happens if the plastic of your momentary button gets wedged in the pressed position)

A lot of PSU cabling is still 20+4, although you may have to break off the

Theo

Does the EC keep the PS_ON low *all the time* that the PSU needs to be turned on, and let it be pulled up high when the PSU should turn off? Or is PS_ON also "momentary" - assuming a long "moment"?

I ask because when I shorted PS_ON to ground and the PSU started up, I accidentally broke the connection within a couple of seconds and the PSU turned itself off. But if I kept it low for about 30 seconds, I was

Is that normal behaviour or does it indicate a PSU fault?

The reason I'm asking is that if it is the EC that has gone faulty, and the PSU is behaving itself as it should, one "solution" may be to install a hard (but not latching) switch to drive PS_ON low, and instruct that the button is kept in for an appropriate period of time - as an alternative to getting a completely new motherboard or declaring the PC "unrepairable".

It should be stateless: grounded = on, not grounded = off.

That's surprising. I've not heard of that happening before.

The way to turn the PSU off is to open-circuit PS_ON. If that doesn't work then it seems like the PSU is jammed permanently on, and there's nothing the motherboard can do to turn itself off. That doesn't sound like intended behaviour.

(there is a typically a 30 second period of holding the 'soft' power button to make the EC force a power off. But this sounds different)

When you said 'stayed on', do you mean the PC stayed running? Or the PSU in isolation (without the motherboard connected) stayed on? It can take some time for voltage rails to decay away under no-load conditions - there are hefty capacitors on them. Also I wouldn't be surprised if some PSUs continue to run the fan to cool things down for a while after PS_ON is removed.

Given you say that 5V_SB is sometimes not there, it does rather imply a PSU fault.

I would swap out the PSU in the first instance before worrying about the motherboard.

Theo

The PSU connected to the motherboard continued to run the CPU fan and the HDD. I was being lazy and running the PC without a keyboard, mouse and monitor connected, so I can't be certain what Windows was doing; I know from when I had a monitor connected that there is a delay of about

I'm wondering whether my "5V_SB not there" error was a false alarm - maybe a poor connection with my voltmeter which is measuring 5V_SB and ground.

Every time I've measured it since then, it's been present as long as the PSU has mains.

I'm giving the old PSU one last chance. Having started the PC by shorting PS_ON to ground, then then long-pressing the soft power button until the PC stops, I'm going to leave it overnight *with mains connected to the PSU* and see if soft power will start it in the morning. Yesterday's test was to start it and stop it as above, and then leave it overnight with mains disconnected.

It's looking as if the PC's soft power circuitry may be dodgy, but only if the PC is left turned off (ie not booted) for a long time.

Having "jump-started" the PSU by manually shorting PS_ON for a while, the PC then responds normally to long and short presses on the soft-power button. This is the case if it is left unbooted for several hours, but not if it is left overnight with mains off and then mains on in the morning (which supplies 5V_SB again). The final test is to leave the PSU with mains (and hence 5V_SB to the motherboard) and see if the motherboard with no interruption to 5V_SB can respond to the soft power button.

Given that 5V_SB *is* being reliably produced whenever the PSU had mains (ie that I was wrong earlier about it being intermittent), I *think* I'm right in saying that the PSU is OK. I may try tomorrow with a different PSU to see if it behaves any differently.

I would also change the CMOS coin cell. Weird booting/power issues can sometimes be because the CMOS battery is on the way out. It's the only wear part on the motherboard so it's worth doing as a matter of course.

Theo

Thanks, everyone, for a really useful thread. When I eventually unbury from behind a heap of stuff, the PC which controls my darkroom things, I'm pretty certain I'm going to find that's why it stopped working and this thread will be a useful starting point.

Cheers,

Nick

Oh, I like that. May I steal it to use as a .sig in another place?

Nick

With old CMOS battery:

- if PC is left overnight without mains power (ie no 5V_SB) it fails to respond to soft power switch (even though 5V_SB is present as soon as I turn on mains, before trying soft power)

- if PC is left with mains power (but PSU not running) it responds to soft power

- mains can be removed for a few hours (but not 12 hours overnight) and it responds to soft power

I've changed the CMOS battery just in case, though old one was reading 3.1 V. Maybe that will make a difference, maybe it won't. I'll try it again tomorrow. Having to wait overnight is a bummer. I don't want to try it too soon because as soon as PSU turns on (either by soft switch or by taking PS_ON low) it resets the "error timer" and I'd have to wait longer to test it!

Just to give you an update, after a bit of wrong information (from me) earlier in the thread...

With old CMOS battery:

- if PC is left overnight without mains power (ie no 5V_SB) it fails to respond to soft power switch (even though 5V_SB is present as soon as I turn on mains, before trying soft power)

- if PC is left with mains power (but PSU not running) it responds to soft power

- mains can be removed for a few hours (but not 12 hours overnight) and it responds to soft power

I've changed the CMOS battery just in case, though old one was reading 3.1 V. Maybe that will make a difference, maybe it won't. I'll try it again tomorrow. Having to wait overnight is a bummer. I don't want to try it too soon because as soon as PSU turns on (either by soft switch or by taking PS_ON low) it resets the "error timer" and I'd have to wait longer to test it!

I'll report whether changing the CMOS battery made any difference by reproducing the no-mains-overnight situation tonight. It's a bummer having to wait so long to see if it affects the problem...

Of course.

Thanks for that. Instead of letting these usenet messages expire as usual I'm locking this thread so any further updates here from anybody will be saved locally.

Nick

Bad news. After leaving the PC without mains power for 11 hours (9 AM to

It looks very much as if there is a capacitor somewhere which is storing enough charge for the soft-power switch to work (providing the PSU has mains) for a few hours after the PSU was last fully on, but the voltage gradually decays beyond a critical point.

I might fit a second momentary-action switch wired between PS_ON and ground, with instructions that if the soft power switch doesn't work, press that for a second or so (to allow the power logic to behave properly) and then press the soft-power switch.

It might be worth putting an oscilloscope on the 5VSB supply and on the Vss & Vdd pins of the PSU control IC's

You might find it is rippling. I simple muiltimeter will not show dc ripple unless its fairly low Hz.

On many PSU's they have a PSU control chip which is fed with 5 V and this will have a capacitor across its Vss and Vdd pins.

Its common to see 10 uF or so electrolytics adjacent to such ICs across their Vss & Vdd rails and its purpose is to eliminate the ripple so the IC functions as expected.

The capacitor has probably dried out increasing its ESR (Effective Series Resistance)

I usually use Panasonic FM or FG series of capacitors as they are low ESR and are rated to 105 °C operation.