ATX PSU PS_ON "leaking" issue

[spilihps]

Hello,

I have a ATX PSU that has a weird issue with how it handles the PS_ON signal.

I know these are cheap and can easily be replaced, but I tend to try and repair them, as there is always something to be learned while doing it.

The symptoms:

PSU, provides VSB fine without issue, it also turns on normally when PS_ON pin is grounded.
However, when it sits turned off, and plugged in to a motherboard. It starts to "leak" voltage on the power rails. Like 7V on the 12 and 3.5 on the 5v.

It is easy to blame the motherboard here, but checking the PS_ON voltage with it unplugged compared to it plugged in I get a voltage drop from 4v down to 3.5v.
This is far from from TTL 0v that is in the spec for PS_ON signal, so the PSU should absolutely not start to present this voltage during these conditions.

I did experiment a bit on the PSU and when loading it with different resistors I can make this happen without a motherboard connected. And it is analog in a sense. Different resistance on the PSON signal, will result in different voltage on the rails.
However, this should function as a switch as far as I know.

I did some digging around in the PSON circuit on the PSU,

Schematics so far:

PSON pin straight to the base of a C945 transistor.
Base connected in parallel with GND through a 150kOhm resistor and also a 33kOhm resistor to pin 13,14,15 of the PWM controller DBL494.
It probably goes further than those pins, this was just as far as I have come on my last session.

Was hoping that the C945 was the culprit, so i pulled it, but it checked out on the standard diode multimeter test.

Voltage on the transistor when sitting idle in standby:

Emitter 3.3388
Collector 5.028 (This comes from VSB)
Base 3.994

The reason for me to suspect either the transistor or circuit around it is that I can make this problem occur by only messing with the PSON pin, and this guy is the only one connected directly through it.
Might be completely incorrect ofcourse.

At first i was working with the theory that the internal pullup for PSON in the psu was incorrect. But thinking about it, we do have clearly voltage above TTL 0 when it starts to turn on.

Before continuing on my hunt, I wanted to check if I'm on the right track or if another theory might be more probable?

The PSU is one of these standard ATX PSUs from late 90s early 2000s.
Brand: ATNG (Might be ALTNG, the logo is not really clear)
Model: AT-230S

[coromonadalix]

is it the same subject from this thread ?   https://www.eevblog.com/forum/repair/atx-psu-ps_on-issue/msg5426477/#msg5426477

if it is, this is not a behaviour i appreciate, you should have stayed in the related thread

now it become a tendency to replicate / force a thread  to get some answers

[spilihps]

is it the same subject from this thread ?   https://www.eevblog.com/forum/repair/atx-psu-ps_on-issue/msg5426477/#msg5426477

if it is, this is not a behaviour i appreciate, you should have stayed in the related thread

now it become a tendency to replicate / force a thread  to get some answers

Hm first of all, the link you provided is to this thread?

If you are referring to the other thread started relating to troubleshooting issue with the motherboard, it is not the same.
That thread was started because I initially and incorrectly suspected the motherboard to be the culprit.
However, since it was not in fact the motherboard but issue with PSU instead I decided to open a new thread.

This is for several reasons, such as it being much much more simple for anyone that might have the same type of PSU issue down the road to find the correct thread.
It will also be much more simple for anyone with expertise on SMPS to find the thread, a subject of motherboard troubleshooting is something completely different.


As I cannot delete the topic, I did change it to Solved, as it was in fact not the motherboard.

[asis]

Hi,

Here's a similar diagram.
In any case, replace 4.7 mF x 50V (C10).

[spilihps]

Hi,

Here's a similar diagram.
In any case, replace 4.7 mF x 50V (C10).

Hey! thank you for the diagram, super helpful.

I see, do you have the reasoning behind the replacement? Is it a changed value or do you suspect it is faulty all togheter?

Found C10 on the diagram, now I just need to locate it on my board 

[Medved]

When looking at the schematic, according to me the real culprit is the (unpublishable word) design how the PowerOn signal is processed to form the desired functionality.
So not any faulty component, but just the way how it was designed. So to really fixing it you would need to really patch the circuitry (add a 2-NPN and a few resistors schmitt trigger circuit,...)...

It may work when generated by a 5VSB supplied CMOS circuitry (i.e. High=5V or very near, Low=0V or very near), but otherwise the response is very questionable. Mainly when some boards generate it from 3.3V CMOS logic levels (many MBs have onboard 5VSB -> 3.3VSB regulator on the MB itself, as the 3.3V higher integration logic is way cheaper than a 5V one since the late 90's; these chips have 5V tolerant IOs so may accept 5V from other circuitry, but are not able to generate 5V levels themself), assuming all cirvuitry has true TTL / 3.3VCMOS (both are the same 0.8..2V threshold) compatible logic inputs with a sharp response.

But in any case, this "leaky zombie mode" is extremely dangerous as a fire hazard: The "PON" voltage may be high enough to disable the overcurrent protection  (keep the Q11/Q15 latch permanently resetted via the Q12 and Q13 follower), but not enough to pull the duty cycle limit input (pin 4 of IC1, via D24 and the input Q13 follower) high enough to really shut the converter down. So in case of some output short circuit (a fault on the motherboard,...), there would not be anything to limit the output fault current (neither in value, nor time). And as the thing is designed to normally deliver about 20A, it won't have that big issue to keep feeding 100A for pretty long time, enough to really set something on fire.

[asis]

@spilihps I would ask for the IC5 type.

Well, evaluate the condition of the “2525” 3V batteries that power the CMOS PCB.

[HwAoRrDk]

It may work when generated by a 5VSB supplied CMOS circuitry (i.e. High=5V or very near, Low=0V or very near), but otherwise the response is very questionable. Mainly when some boards generate it from 3.3V CMOS logic levels (many MBs have onboard 5VSB -> 3.3VSB regulator on the MB itself, as the 3.3V higher integration logic is way cheaper than a 5V one since the late 90's; these chips have 5V tolerant IOs so may accept 5V from other circuitry, but are not able to generate 5V levels themself), assuming all cirvuitry has true TTL / 3.3VCMOS (both are the same 0.8..2V threshold) compatible logic inputs with a sharp response.

What are you talking about? The ATX power supply PS_ON signal is active-low, so the MB will have an open-drain/open-collector output for which the logic level is irrelevant, because it is just switching PS_ON to ground.

[Medved]

Yes, it is Active low and it is expected to be driven by an open drain (where, indeed,  the supply of the logic does not matter once it is 5V tolerant).
But there is nowhere written it can not be driven by a push-pull logic output holding the output at 3.5V or so when outputting "high". Or some may have used just a plain 3.3V open drain, but with an ESD diode towards its 3.3V supply, so clamping the voltage to 3.7V or so. All perfectly compatible with the TTL specification (High is supposed to be generated as 2.4V and above, which it is).

The point is even when the voltage is held at 2V and above (the TTL high level minimum input level), the power supply is supposed to respond to it as a proper "PowerOff" command.

And this PSU design clearly does not do that, because the PSON input is designed the way it is there.

And regardless what input levels the PSU is designed for, wrong by itself is the fact that just a wrong voltage level at the control signal may cause situation when the PSU generates high current/power output without any overload/short circuit protection.

[m k]

So first IC1 pin 12 must have a voltage, then if pins 8 and 11 are pulsing the chip is on.
But if pin 12 is pulsing there is a fault and the chip is possibly not on.

When IC1 pins 8 and 11 are pulsing T2 is driving T1.
But if 12V feedback is not present through D10 the drive is weaker.

Back in the day I learned that anything but obvious is too time consuming.
Schematics situation wasn't good either.