Viper22A 5VSB circuit

[прямо]

Ok, that was weird. For a couple of tries, 3.3V rail was 1.68V, TL431 cathode was 1.25V and reference was 1.68V

But now 3.3V outputs 2.25V, the same as TL431 cathode, reference remains at 1.68V

Interesting point, without the TL431, I get the same voltages.

What next? Change the final opto coupler that is connected to this 3.3V FB circuit or change the TL431 reference resistor?

[прямо]

I replaced the last opto coupler, but before I could even start the PSU, just as AC mains came on board, I heard a tiny spark. Don't know what caused it and I wasn't able to see which component it came from.

PSU started. Rails dropped by 1/3. All TL431 feedback circuits had like 11.5V on the reference and cathode, then I turned on and off the PSU a few times, same thing.

Now the rails except 5VSB no longer output any voltages, and my bulb is blinking bright.

I don't see any blown or burnt component. Will have to test them again one by one. FUN!!!!

[прямо]

So the reason why the outputs are very low is because UC3845B VREF pin is 1.2V instead of 5V.
It's connected to the opto coupler collector pin, and despite I have replaced it (again!), nothing changes. I'm not surprised being that the feedback is pretty much also screwed up (TL431s cathode and reference reads a few hundreds mV instead of the minimum 2.5V)

I lifted one leg of every passive components that are on the main PWM and both feedback circuit to measure them individually and they're ok.

Not sure what else to do.

[temyth]

I'm using UC3845B not UC3844.

I replaced the opto coupler, and..... et voila! 12V is now 12.5V and 5V is 5V!!! but now 3.3V has dropped to 1.68V

my fault, i misreaded it. well, thats good. did you tested the voltage comparator of 3.3v? if i'm well remembered there are 3 optocouplers: 1 for stdby, 1 for 12v and another for 3.3v. try checking the last one and follow the trails

[прямо]

12V TL431 cathode was shorted to anode.
I also put in the wrong TL431 with 1.24V REF.

Fixed that and 12V, 5V output rails are back to normal (yay!)

3.3V is still very low, because its REF is at 1.47V (1+5K1/15K) x 1.47V = 1.968V
if REF is at 2.5V, the 3.3V rail would be normal.

Not sure what caused it to drop. Could be a bad TL431.

Unfortunately I don't have a spare TL431 with 2.5V REF anymore. I still have some 1.24V REF ones though, so I guess I could I use it and change the resistors divider network instead, something like this: (1+10K/5K6) x 1.24V = 3.3V

[temyth]

12V TL431 cathode was shorted to anode.
I also put in the wrong TL431 with 1.24V REF.

Fixed that and 12V, 5V output rails are back to normal (yay!)

3.3V is still very low, because its REF is at 1.47V (1+5K1/15K) x 1.47V = 1.968V
if REF is at 2.5V, the 3.3V rail would be normal.

Not sure what caused it to drop. Could be a bad TL431.

Unfortunately I don't have a spare TL431 with 2.5V REF anymore. I still have some 1.24V REF ones though, so I guess I could I use it and change the resistors divider network instead, something like this: (1+10K/5K6) x 1.24V = 3.3V

yeah, you can just change the resistors. i would place a ponteciometer instead of a resistor, so you could regulate it easier. by the way, how do you know the voltage reference of each TL431? i didn't knew about this spec of TL431, just thought that every TL431 was identical (beside of after letters)

[прямо]

They are not identicals.
AZ431A has 0.4% voltage tolerance.
AZ431B has 0.8% voltage tolerance.
AZ431L has a lower minimum VREF @ 1.24V/1.25V

[прямо]

Well that didn't work at all. Quickest way to test the theory, I swapped the resistors divider around, thus:

VOUT = (1+15K/5K1) x 1.24V

which should be 4.8V, but it doesn't.

No matter what, 3.3V rail will not go above 1.99V (2V)

Interesting point, I do see 3.3V going to the opto coupler pin 2. I have no idea why would there be a 3.3V

It's only a single trace connected to the supervisor IC (which is dead when 12V aka VCC was 33V) pin 3 (FAULT PROTECTION) via a 470 ohm resistor and this pin has to be shorted to ground to start the PSU without the IC.

[прямо]

So I took a TL431 from a working 32V 4A power supply board and I put that in.

I also put back the resistors divider. 5K1 divides by 15K.

While at it, I took the time to test the main transformer RC snubber and Schottky diode. All checked out OK.

I injected 3.5V from my bench supply at the 3.3V rail and I could see the TL431 working as it should. 3.1V is also present at the supervisor IC VS33 monitoring pin. Which means there are no faulty components that would bring 3.3V down to 2V.

Perhaps the reason why it can't go higher than 2V is because I'm still using UC3845B instead of TL3845P. When the part gets here, I will confirm if that's indeed the culprit.

Addendum:
Reconnected the lifted leg of a 180 ohm 1W resistor that gotten hot before and brought back the -12V rail. Now 3.3V rail has increased to 2.6V and that resistor still gets very hot to a point I can smell the heat. I don't like it.

I also don't understand how 3.7V (if without supervisor IC) can appear at the opto-coupler pin #2 even though the other end of a 470R resistor in series with it is 0V. Can someone explain this to me??? EDIT: Nevermind, I figured it out. Pin #2 is cathode of the opto coupler LED, so 3.7V appears because of 5VSB to anode, and the LED has 1.3V voltage drop.

[stj]

does the 3v3 diode get hot with a load?
diodes can look good on a meter and screw up with higher voltage or current on them.

[прямо]

stj

No it doesn't get hot. There's already a 27 ohm 1W resistor connected to the 3.3V rail on board, acting as a basic load.

Previously with 2V output, it didn't conduct because it's below the min 2.5V REF, now it does since output has increased.

[harp]

No matter what, 3.3V rail will not go above 1.99V (2V)

How much current can you get from rail, how much amps is declared to 3.3v?
Can you observe switching signal acording to load?

[прямо]

harp,

30A for 3.3V (dual rectifier is MOSPEC 40A 45V)

PSU is currently not loaded at all, aside from the basic 1W resistor the factory put in.

When it previously had a bad opto coupler, 12V and 5V were over volting up to 3 times if I didn't use my bulb current limiter, however 3.3V was only went as high as 4.2V

12V and 5V are back to normal. 3.3V is drifting between 2.5V to 2.6V because the TL431 is stabilizing it, but changing the voltage divider resistors to set a higher VOUT doesn't do anything. I could only make it go lower than 2.6V, and that was with a 1.24V REF TL431L.

Don't have a scope to monitor the switching signal.

Obviously the PSU isn't worth fixing, and I'm just doing it to satisfy my ego that's saying "I CAN FIX THIS!" 🤣

[прямо]

Bloody hell! I FIXED IT GUYS!

I was watching a YT video about UC3845 and out of curiosity I decided to measure its duty cycle. Was surprised to see it's only at 4%, so that got me thinking, is this PSU experiencing what Richard (dicky96) had in one of his ATX PSU repair videos. Turns out it is! All I have to do is just add a load to the 12V rail to push the PWM chip working a bit harder, so I attached the fan, then I couldn't believe what I saw.

3.3V rail is now 3.3V
Thank you all! especially harp

[temyth]

Bloody hell! I FIXED IT GUYS!

I was watching a YT video about UC3845 and out of curiosity I decided to measure its duty cycle. Was surprised to see it's only at 4%, so that got me thinking, is this PSU experiencing what Richard (dicky96) had in one of his ATX PSU repair videos. Turns out it is! All I have to do is just add a load to the 12V rail to push the PWM chip working a bit harder, so I attached the fan, then I couldn't believe what I saw.

3.3V rail is now 3.3V
Thank you all! especially harp

congrats champs!!! happy you fixed it, I'm sorry i coudn't help more than i did

[прямо]

congrats champs!!! happy you fixed it, I'm sorry i coudn't help more than i did

You did helped me, otherwise I wouldn't be able to fix this because of the missing zener diode at the primary circuit. Without it the VCC supply for Viper22A and UC3845 will be unstable and when it goes higher than max voltage allowed, they will get blown up again.

So thank you for that

[прямо]

BTW, what can I use to replace a 180 ohm resistor that connects the 12V rail to the -12V rail?
I tried upping it from 1W to a 2W size and it's generating about the same heat. Can't put my finger on it for more than a few seconds.
I don't like it because it's next to the output caps.

EDIT: OOPS, my bad. The resistor is a dummy load for the 12V rail. Now it makes perfect sense why it gets quite hot.

12.3V / 180 = 0.07A
0.07 x 12.3V = 0.8W

cutting it close, I'm surprised it's not dead.

[temyth]

BTW, what can I use to replace a 180 ohm resistor that connects the 12V rail to the -12V rail?
I tried upping it from 1W to a 2W size and it's generating about the same heat. Can't put my finger on it for more than a few seconds.
I don't like it because it's next to the output caps.

thank you very much for your consideration, it was a pleasure helping you, I learned a lot with it! maybe you can use a current limiter diode

[прямо]

I had the pleasure of testing the PSU with a 200W TDP graphics card. When the card is working at 100% load, consuming power about 88W, the 12V rail drops to 10.77V (slightly worse than a 10% tolerance)
Not sure what to look for to remedy that.
Main mosfet not good enough? it is after all, a part that I salvaged from junks.
Or the output dual rectifiers are getting old and tired?

[прямо]

Apparently I forgot to replace the bridge rectifier.
I was still using what I had at the time in my stash, a couple of 4A diodes.

Changed them to a 10A bridge recifier, and the 12V rail now stays at 11.9V when the graphics card is under full load pulling approx. 202W