Hostile is right. Go to EEVBLOG and tell them you have an old, rusty and pigeon poop covered Tek scope found in barn and you want to gut it to make a YT prop (or worse yet, a cat bed) out of it. Watch the fireworks, with tar and feathers and the neck rope. They do spend a lot of money on junk (a LOT judging by photos of some collectors there), they just don't consider early 90's PC machines to be old or valuable. To each their own I guess but they can be pleny toxic too.

Deunan wrote on 2024-04-02, 23:09:

cross wrote on 2024-04-02, 19:53:

Not entirely sure why a PSON circuit like this that should and could only be either off or on, would be setup with a sort of amplifier circuit? (if im getting the setup correctly)

Emitter-follower configuration, also known as common collector, has voltage amplification of 1. In fact lower than that because there is a drop on the base-emitter junction equal to a typical silicon diode - about 0.6V. What is amplified is the base current, therefore the source of the control voltage is not being loaded much. And that's why the transistor is there at all, to make sure that even some sort of high impendance (weak in other words) output that drives the PS ON signal will be able to properly drive whatever circuitry there is in the PSU.

So the emitter voltage will be always (well, as long as you don't exceed the 5V supply to the collector) Vbase - 0.6V. If your mobo floats that input at about 4V then you'll see 3.4V at the emitter and that is what will drive the PSU internally. Assuming the trick used here was to connect the emitter to the dead time control pin (#4), the datasheet says the zero duty cycle is at around 3.3V at that input. So you are probably just starting to turn the PWM controller on (this voltage will also depend on temperature, chip aging, manufacturing tolerances, etc). So if it is like this then I was right and the circuit is simply not designed up to spec but rather simplified and hoping the mobo will keep the PS ON close enough to 5V for this not to be an issue.

Again, thank you so much for the write-up. Its slowly becoming clearer 😁 I now also understand a bit further on how the PWM IC is utilized here. Reading the datasheet it was not clear to me how the pson signal is handled. Now I have a bit better understanding.

So I did write up some further circuit:

Emitter from previously mentioned transistor is connected via a zener to pin4 on the IC (as you suspected? 😁 ) Parallel from this it is connected via a 4,7kOhm resistor to the base of another C945 which in turn is emmiter GND collector -> zener and also pin 4.

Maybe this I can make it clearer like this:

C945 (number 1)
Emitter -> Zener -> Pin 4 on IC
-> 4,7Kohm - Base C945 (number2)

C945 (number 2)
Emitter -> GND
Collector -> Zener -> Pin4

The second transistor is some sort of overvoltage protection for the PWM controller - in the event that PS ON wire actually gets external power fed to it somehow. I don't quite get the zener diode on the emitter of the first transistor, makes no sense to me.
Try to change that pull-up and perhaps that will work, otherwise this bit would have to be redesigned - preferably by replacing the transistor with proper logic gate(s) like HCT14 for example.

cross wrote on 2024-04-02, 15:06:

Hello momaka, […]
Show full quote

Hello momaka,

Thank you for your reply.

I did actually open a post on eevblog forum,specifically for the PSU: https://www.eevblog.com/forum/repair/atx-psu-ps_on-issue/

Feel free to check it out!

Cool, just did that. Glad to see some pictures.
Caps look visually fine indeed, but I noticed the smaller ones are "CS-logo" branded, which are notorious for going bad.
There's a 47 uF 50V small cap right by the secondary heatsink visible in the top-center of the picture. I believe that one should correspond to C25 on the schematic diagram that was posted on EEVLOG by user asis. If so, then that's the filter for the secondary side aux. rail that supplies power to the DBL494 IC. If this cap is bad, the high ripple could cause all sorts of "strange" issues with the IC.

I notice the other caps are the same brand too, so they are really all worth a check if you have an ESR meter. In particular, if the corresponding cap to C31 and C38 on the diagram have failed with high internal leakage current, those can also mess with the logic level perceived as "On" by the IC. So all of the small electros should really be replaced or checked.

Now, you mention you could re-create the problem with playing with resistor combinations on the PS_ON pin. If so, that's indeed a very poor logic level design. Given how copied this design is, I'm curious how many old PSUs with only a 494 / 7500 controller have this issue. Speaking of which, DBL494/TL494 and KA7500 are 100% pin-compatible and interchangeable ICs, so if the DBL494 datasheet is not clear (at least the one I have on my PC seems like an ancient scan from eons ago), you can also reference the KA7500 / KA7500C datasheet.

cross wrote on 2024-04-03, 13:35:

Maybe this I can make it clearer like this: […]
Show full quote

Maybe this I can make it clearer like this:

C945 (number 1)
Emitter -> Zener -> Pin 4 on IC
-> 4,7Kohm - Base C945 (number2)

C945 (number 2)
Emitter -> GND
Collector -> Zener -> Pin4

Hmmmm.. are yo sure those Zener diodes are actually such?
Going by the description of your connections above, it is essentially the same thing as the schematic diagram posted by user asis on EEVLOG, except on that schematic, the "Zener" diodes you describe are just standard regular 1N4148 diodes (D24 and D25 to be precise). 1N4148 are usually small, red, glass devices in appearance... which is what perhaps made you confuse them with Zener diodes??

Either way, I'm just trying to confirm here how close your PSU is to the schematic diagram posted.

I also see a 7805 regulator on a heatsink on your pictures, which suggests the 5VSB uses an ancient self-oscillating design with no voltage feedback (open-loop). Ironically, this design is safer compared to the slightly "newer" 2-transistor self-oscillating designs with feedback (presence of an optocoupler next to the 5VSB transformer), since the 7805 regulator has quite the input voltage range before things go haywire. In any case, the presence of the 7805 reg makes the posted schematic diagram one step closer to what your PSU is.

The one main difference I see between the schematic and your PSU is that yours probably uses a linear-regulated 3.3V rail (derived from the 5V rail via MOSFET on the secondary heatsink) rather than a mag-amp regulated 3.3V rail (presence of a 2nd toroid on the output)... unless I missed seeing that 2nd toroid for the 3.3V rail (or perhaps it's a "PI" coil, like on some cheaper older PSUs.) This all relates to the main PSU rails (5V, 12V, 3.3V) and has nothing to do with the PS_ON signal and the PSU half turning On... but again, just mentioning it to have as a comparison to the posted schematic.

...
Anyways, going back to the PS_ON signal pin of your PSU...
If you have a few common spare junk parts around, I ****think**** the easiest circuit fix would be to use a common PNP transistor like 2N3905.
To do so, disconnect PS_ON pin from PSU and connect Emitter of PNP transistor to where the PS_ON pin was. Then connect the Collector of the PNP transistor to ground. Finally, connect a 10k resistor in series with a white or blue LED (green LED may be OK too or might need an extra diode in series) and have this duo connected on one side to PSU's 5VSB and the other to the Base of the PNP transistor. The LED should be forward-biased (Cathode pin) towards the Base of the PNP transistor. Since white/blue (and pink!) LED's have an approximate voltage drop of about 2.7-2.8v minimum (and about 3.3-3.5V nominal), the PNP transistor shouldn't turn On until the voltage at it's base drops to 5V minus the LED voltage drop (i.e. ~3V), so this should bring the PS_ON threshold voltage around 2-2.5V now. You can further lower this by adding in another (regular) diode (so another ~0.6-0.7V drop) to the LED-10k-resistor series connection. Furthermore, the setup with the PNP transistor mentioned above is a bit more of a "all or nothing" when it comes to logic levels. Thus, the invalid "in-between On/Off" state of the PSU should no longer be possible no matter how you mess with the PS_ON signals.

**** I have not built this circuit for this particular use before, so can't 100% confirm it should work as theorized above. But if wired properly, there won't be any danger to the motherboard or the PSU for trying it.

rasz_pl wrote on 2024-04-03, 05:19:

Badcaps is more tolerant of wasting money on junk 😀 They have whole section for bad supplies https://www.badcaps.net/forum/troubleshooting … r-supply-design

Well, my views might be biased 😁 , since that used to be my main place to hang around for many years (and still is to a point.)
Not all PSUs are a waste of money/time to attempt to fix. There are sometimes very proprietary stuff that's hard or impossible to find, so the only way to go is to fix the original.

Of course, if you dig a little further back through that section on badcaps, you will find quite a few topics of people fixing & upgrading utter junk ATX PSUs, all just for fun. Actually, there's a good few of those started by yours truly here. 🤣 But like anything else, it can be a fun hobby if you have the "knack" for it.

So with that said, I stand by this:

Deunan wrote on 2024-04-03, 08:44:

To each their own I guess

Deunan & Momaka,

Thank you both for your help and expertise with this.

You are most likely correct with my being incorrect about the zener diode. When checking it again I most likely got it incorrect. I will dig out the microscope to confirm exactly what it is.
Fully agree that it seems strange for it to be a Zener 😁 Sorry for this.

First of all, I found some time to work on the supply this weekend. Got fed up with it for a while 😜
I did replace the 33k pullup with a 20k. I did notice that it was more stable when "floating" than before. Like when testing my resistor on PSON pin. It was for sure less sensitive.
But, while testing it on the motherboard, the issue was still present unfortunately.

Cool, just did that. Glad to see some pictures. Caps look visually fine indeed, but I noticed the smaller ones are "CS-logo" bra […]
Show full quote

Cool, just did that. Glad to see some pictures.
Caps look visually fine indeed, but I noticed the smaller ones are "CS-logo" branded, which are notorious for going bad.
There's a 47 uF 50V small cap right by the secondary heatsink visible in the top-center of the picture. I believe that one should correspond to C25 on the schematic diagram that was posted on EEVLOG by user asis. If so, then that's the filter for the secondary side aux. rail that supplies power to the DBL494 IC. If this cap is bad, the high ripple could cause all sorts of "strange" issues with the IC.

I notice the other caps are the same brand too, so they are really all worth a check if you have an ESR meter. In particular, if the corresponding cap to C31 and C38 on the diagram have failed with high internal leakage current, those can also mess with the logic level perceived as "On" by the IC. So all of the small electros should really be replaced or checked.

Now, you mention you could re-create the problem with playing with resistor combinations on the PS_ON pin. If so, that's indeed a very poor logic level design. Given how copied this design is, I'm curious how many old PSUs with only a 494 / 7500 controller have this issue. Speaking of which, DBL494/TL494 and KA7500 are 100% pin-compatible and interchangeable ICs, so if the DBL494 datasheet is not clear (at least the one I have on my PC seems like an ancient scan from eons ago), you can also reference the KA7500 / KA7500C datasheet.

Thank you for the tip on the specific caps to check. I will see if I can hunt them down. I do own a ESR meter. This is usually my go-to to check SMPS like this.
However, this kind of failure is new to me, and since i wanted to check how the PSON was handled in a circuit like this, I started a bit backwards 😜
Please let me know if any further pictures might be of interest for this. It was kind of difficult to get good top pictures, since the heat sinks are pretty large.

Anyways, going back to the PS_ON signal pin of your PSU... If you have a few common spare junk parts around, I ****think**** the […]
Show full quote

Anyways, going back to the PS_ON signal pin of your PSU...
If you have a few common spare junk parts around, I ****think**** the easiest circuit fix would be to use a common PNP transistor like 2N3905.
To do so, disconnect PS_ON pin from PSU and connect Emitter of PNP transistor to where the PS_ON pin was. Then connect the Collector of the PNP transistor to ground. Finally, connect a 10k resistor in series with a white or blue LED (green LED may be OK too or might need an extra diode in series) and have this duo connected on one side to PSU's 5VSB and the other to the Base of the PNP transistor. The LED should be forward-biased (Cathode pin) towards the Base of the PNP transistor. Since white/blue (and pink!) LED's have an approximate voltage drop of about 2.7-2.8v minimum (and about 3.3-3.5V nominal), the PNP transistor shouldn't turn On until the voltage at it's base drops to 5V minus the LED voltage drop (i.e. ~3V), so this should bring the PS_ON threshold voltage around 2-2.5V now. You can further lower this by adding in another (regular) diode (so another ~0.6-0.7V drop) to the LED-10k-resistor series connection. Furthermore, the setup with the PNP transistor mentioned above is a bit more of a "all or nothing" when it comes to logic levels. Thus, the invalid "in-between On/Off" state of the PSU should no longer be possible no matter how you mess with the PS_ON signals.

**** I have not built this circuit for this particular use before, so can't 100% confirm it should work as theorized above. But if wired properly, there won't be any danger to the motherboard or the PSU for trying it.

Amazing write up! If all else fails I might give this a shot.

To learn as much as possible I would like to empty all possibilities of broken components before giving up and accepting its a bad design. Can commercial products relly be that bad? *wink wink*

I think next steps will be to check caps mentioned to see if we can be so lucky to solve it by this.

If the PSU is nicely built (though I somehow doubt that given the design issues) then I'd try to mod the circuit with 74HCT14. Or LS14 because CMOS is more susceptible to ESD - but that's not a big issue here, long wires and big capacitor on input would prevent most nasty shocks from damaging the chip. And LS would most likely require a pull-up on output as well so all in all HCT is a better pick.

Looking at the schematic from EEVBLOG (which might not be of this exact model mind you) I'd remove the C945 on input, and it's base pull-up, use two gates (since '14 is a negation) in series to replace it. What was base is going to be the input of the gates, output on emitter, use collector as +5V power for the chip and just find a nearby GND to complete the circuit. Preferably connect 5V or GND to the inputs of unused gates, and if it works the chip can be potted or otherwise insulated and just stuffed on short wires on the low volate side of the PSU somewhere.

cross wrote on 2024-04-07, 20:44:

First of all, I found some time to work on the supply this weekend. Got fed up with it for a while 😜

No worries at all.
That's really the best way to work on stuff - take a break from them when you get stuck. Often you'll find your brain to still be doing calculations "at the back"... and sometimes that's how you get new ideas to try.

cross wrote on 2024-04-07, 20:44:

I did replace the 33k pullup with a 20k. I did notice that it was more stable when "floating" than before. Like when testing my resistor on PSON pin. It was for sure less sensitive.
But, while testing it on the motherboard, the issue was still present unfortunately.

Yea, I think the way that PS_ON circuit is build (1st C945 transistor wired as a voltage follower) may indeed be sub-optimal.

cross wrote on 2024-04-07, 20:44:

Thank you for the tip on the specific caps to check. I will see if I can hunt them down. I do own a ESR meter. This is usually my go-to to check SMPS like this. ...
If all else fails I might give this a shot.

To learn as much as possible I would like to empty all possibilities of broken components before giving up and accepting its a bad design.

Sounds good!
And while you check the caps and other components, I'll try to see how my idea (and possibly the whole PS_ON circuit) behaves in a Multisim simulation. I think it may not actually work that well on a 2nd though. So hold on before building that circuit. I had to pull out my PC with multisim, since I have this Mirage Nano S8 subwoofer that needs fixing. It actually works OK, but it has no speaker protection relay or output mute delay, so it makes a very loud POP at turn-On. Thus, I'm currently designing a circuit to take care of that. So while at it, I'll give this PSU's PS_ON circuit a sim try too.

cross wrote on 2024-04-07, 20:44:

Can commercial products relly be that bad? *wink wink*

Yes, not too uncommon.
A lot of times, the (cheaper) manufacturers just aim for "good enough" / working at lowest possible cost. So if a circuit appears to work well enough, you can end up with these problematic circuits easily that start misbehaving when their components age a little.

That being said, the PSU you have doesn't appear that badly built. Sure it uses an old (and probably not very efficient) half-bridge design, but the heatsinks appear overall large and there are no missing caps on the output, so it's not a low-end junk unit. Probably adequate for 200-250W, which is usually more than enough for most retro PCs.

Hello again!

So I put some work in today, removing and checking the mentioned caps.
First of all, the one that is 47uF 50v and close to the heat sink that was mentioned before.
I removed that one and capacitance measured spot on, however, the ESR was 2.1ohm. Which I think might be out of spec. Difficult to know exactly besides these "generic" ESR-tables.

C38 in the schematics (Which co-incidentally? was also C38 on the board) measured ESR 4.3ohm.

Currently pulling and checking all electrolytes surrounding the controller IC. Idea here was to either identify a culprit or replace all of them. If error goes away, then replace all electrolytic caps on the board.

What is slightly more concerning is that I cannot locate the mentioned C31, i cannot find any electrolytic cap conneted directly to PSON line. Maybe this design is missing it all togheter?

So far this is the ESR measurements of the pulled caps (all have spot-on capacitance):
(Board markings)
C30 47uF 50v 2.1 ohm
C31 0.68uF 50v 21 ohm
C33 4.7uF 50v 16 ohm
C34 1uF 50v 6.6 ohm
C38 2.2uf 50v 4.3 ohm (pin 4 on the IC)

Any thought about the readings here?

cross wrote on 2024-04-13, 16:02:

So I put some work in today, removing and checking the mentioned caps.
First of all, the one that is 47uF 50v and close to the heat sink that was mentioned before.
I removed that one and capacitance measured spot on, however, the ESR was 2.1ohm. Which I think might be out of spec. Difficult to know exactly besides these "generic" ESR-tables.

Well, problem is that a lot of unknown / garage cap brands are not easy to find a datasheet for... and often in the cases that one does, it's found that the caps are likely not low ESR type. I suspect this is exactly the case with these GL caps - they are 105C -rated, but they are not low ESR - not even entry-level low ESR. So if that's the case (and I think it probably it is), then 2.1 Ohms is not too high of a reading for a non-low ESR cap... which is a bit of a cheapskate move on the PSU manufacturer's side, since all caps filtering a switched output should be at least entry-level low ESR.

cross wrote on 2024-04-13, 16:02:

So far this is the ESR measurements of the pulled caps (all have spot-on capacitance): (Board markings) C30 47uF 50v 2.1 ohm C31 […]
Show full quote

So far this is the ESR measurements of the pulled caps (all have spot-on capacitance):
(Board markings)
C30 47uF 50v 2.1 ohm
C31 0.68uF 50v 21 ohm
C33 4.7uF 50v 16 ohm
C34 1uF 50v 6.6 ohm
C38 2.2uf 50v 4.3 ohm (pin 4 on the IC)

Any thought about the readings here?

For small general-purpose e-caps, which these 5x11 mm cans are, up to 2-3 Ohms ESR I'd consider passable with 4-5 Ohms ESR being about the borderline. Anything above that, I'd say the caps are starting to go high ESR. Now, if the caps were low-ESR type, I'd expect to see lower numbers for the ESR - typically 1 Ohm or less.

Given the above numbers, it indeed very much looks like these GL caps are just plain old general purpose 105C caps, hence their >1 Ohm ESR. That said, I think C31 and C33 are definitely bad. And even C34 I'd say is not healthy.

Since the equivalent of C25 in the schematic (the 47 uF 50V cap) appears to be relatively OK, I'm starting to think the whole "PSU half-On" issue is not due to bad caps (though replacing those caps would still be advisable.)

As to the missing 2.2 uF 50V cap from your PSU (C31 in the schematic): it actually doesn't make that much of a difference if that cap is there or not, at least when the driving source is a PC motherboard. I think it was included in the design of the schematic to suppress switch bounce, if say, a mechanical switch was used to ground the PS_ON wire. But PC motherboards usually switch PS_ON either via open Collector output logic or CMOS logic... so switch bouncing shouldn't be an issue.

I also did the simulation in Multisim. In the first iteration, I forgot to add C31. But it didn't make any difference. I used a slow *ramp* function for the simulation (similar to what you did with the resistors) to see how the PS_ON signal of the PSU behaves. What's interesting is that the simulation gave out almost exactly the same "open-circuit" PS_ON voltage level that you measured without anything connected to the motherboard. And furthermore, the signal going to pin 4 of the TL494 IC had the same exact slow linear ramp of the PS_ON signal, which I think is what allows the PSU to get in these "in-between" On/Off states.

So in short, it really looks like this PSU's design might just be a bit on the sub-optimal side. I'll try to post the results of both circuits either tomorrow or Monday, as it might take me a few minutes to type up the details of that post (and also a diagram of the modified circuit I came up with.) The modified circuit I cam up with is not that much more complicated than the series resistor-LED circuit I suggested above. And at least from my simulation, it should be rock-steady.

I suppose just change those three caps with the highest ESR and give the PSU a test (with your resistor experiment) just to see if that changed anything. I highly doubt that it would. But if would be good to confirm before we go modifying the PSU.

Alright!

So, I actually had new caps on hand, which was a first, after years of ordering too many parts 😁

But yes, nothing did change, it did behave exactly the same.

I have decided that this PSU, as it is otherwise in good condition will continue its life in a older machine that has a hardwired PSON-switch.
The adapter cable is on its way, and most likely it will work OK, as the switch will open and close the PSON pin.

Thank you to everyone for helping out and special thank you to momaka for your extremely good explanations and information on every aspects of this and attempts to make a correction circuit for it.
The main goal was achieved and that was to learn what the issue was. Now one more PSU is in the books and iam one fault wiser 😁

Now I need to move on to my next projects that are piling up. Just found out today that two motherboards from the core 2 duo period died in storage. Worked perfectly before being put away, but this is for another thread! 😁

cross wrote on 2024-04-17, 20:57:

I have decided that this PSU, as it is otherwise in good condition will continue its life in a older machine that has a hardwired PSON-switch.

Sounds good to me. 😀

cross wrote on 2024-04-17, 20:57:

Thank you to everyone for helping out and special thank you to momaka for your extremely good explanations and information on every aspects of this and attempts to make a correction circuit for it.
The main goal was achieved and that was to learn what the issue was. Now one more PSU is in the books and iam one fault wiser 😁

You're welcome!

Speaking of the issue with the original PS_ON circuit, here are all of the details below.
... but first, a schematic cut of the Octek X25d AP-3-1 250 Watt ATX PSU that was posted on EEVlog, so we can see the PS_ON circuit more clearly:
download/file.php?mode=view&id=191164

And here is a somewhat simplified diagram of the original circuit I made in Multisim:

About the above schematic:
Q1 corresponds to Q13, Q2 to Q12, and Q3 to Q11. Q15 was omitted, since it's only used for the short-circuit and OVP protections. The rest should be relatively trivial to figure out what's where with all of the resistor values.
To show what happens with the simulation, I set up a TS (transient step) function with voltage source V1. It starts at 5.5V and stays at that level for 2 seconds, then linearly drops to 0V in another 2 seconds , then stays at 0V afterwards. It basically replicates what would happen if a potentiometer was used on the PS_ON signal. Diode D3_TS is not part of the PSU. I added it only to isolate the PS_ON signal from the TS function at "idle" state. This will hopefully become a little more clear in the graph below.

So how does the original PS_ON circuit of the PSU react to a "non-ideal" PS_ON signal? Short answer: NOT GOOD. See below graph.

What do we see here?
- The purple line shows the TS function.
- The red line shows the PS_ON signal at the PSU's PS_ON pin. Note that with nothing connected (i.e. when the TS function is at 5.5V), the "open-circuit" PS_ON voltage level of the PSU is at approximately 4V, which is pretty much exactly what @cross measured in post #3 of this thread on his PSU. Once the voltage from the TS function drops about 0.5V below 4V (due to D3_TS diode), note that the PS_ON signal begins to follow the shape of the TS function and also drops linearly.
- The blue line is the one of most interest here, as this is the signal going to pin #4 of the TL494 / DBL494 IC. Note that this signal also follows the shape of the TS function and PS_ON signal... which is NO GOOD, since pin 4 inside the TL/DBL494 IC is the input to the DTC comparator. Going by the TL494 datasheet, we can see on page 6 that zero duty cycle (i.e. PSU main rails in OFF state) is typically at 3V... but could be up to 3.3V due to temperature, chip variations, and etc. Meanwhile, pin #4 @ 0V = 100% duty cycle (i.e. PSU fully On.) To AVOID a situation where the PS_ON signal is controlling the duty cycle of the PSU and making it "halfway turned On", the signal at pin #4 of the TL494 IC should NOT follow the shape of the PS_ON signal. Rather, it should be a defined high or low signal. Since that's not the case here and we can clearly see how TL494 pin #4 follows the PS_ON signal, it's absolutely no surprise the PSU acts the way it does.

So it indeed seems to all all boil down to a poor design of the PS_ON circuit.