Is the input fuse blown (the round thing that says T2A 250V)? If so, you also have a short somewhere on the primary, making the whole situation even more problematic.
Personally, I would try to fix it out of curiosity and to test my skills, but it certainly isn't worth it for £30+. The caps should not be that expensive...

I would start by diagnosing the short on +5VSB. There is no point in replacing anything before that is fixed.
After that, I would replace all the electrolytic caps with good brands (except maybe the big one the primary) and hope for the best.
Since this PSU obviously has no PFC, it is relatively safe to work on it with the classic light bulb in series trick.

watson wrote on 2024-05-18, 16:26:

Is the input fuse blown (the round thing that says T2A 250V)? If so, you also have a short somewhere on the primary, making the […]
Show full quote

Is the input fuse blown (the round thing that says T2A 250V)? If so, you also have a short somewhere on the primary, making the whole situation even more problematic.
Personally, I would try to fix it out of curiosity and to test my skills, but it certainly isn't worth it for £30+. The caps should not be that expensive...

I would start by diagnosing the short on +5VSB. There is no point in replacing anything before that is fixed.
After that, I would replace all the electrolytic caps with good brands (except maybe the big one the primary) and hope for the best.
Since this PSU obviously has no PFC, it is relatively safe to work on it with the classic light bulb in series trick.

Thanks for this - the fuse didn't look damaged so I hadn't actually tested it - there is no connection across it - so it must be blown!

Any tips for tracing the fault on primary? The cracked thermistor next to the fuse is reading about 9R so that seems OK at least? The bridge rectifier does test as short between two of the pins (I understand each combination of pins should either read closed or diode drop of ~0.5 ) - could that be the problem?

The cart cost was creeping up toward £30 because I'd included a new large capacitor (120uf 400v, £5.20 each) and other components come in minimum 5 or 10, or I've added at least one spare of each etc. I'd kind of rather bulk out the order to get to the minimum of £40, than pay a £9.99 handling charge on an order less than that! (might chuck in another SSD or something ...)

justin1985 wrote on 2024-05-18, 23:13:

The bridge rectifier does test as short between two of the pins (I understand each combination of pins should either read closed or diode drop of ~0.5 ) - could that be the problem?

Shorted rectifier is usually effect and not a cause, that is to say you should also check (desolder if necessary) the switching transistor(s). When the secondary caps vented it might have caused a spike that upset (or even killed) the PWM IC, which in turn overloaded the transistor. But it could be just the bridge, I've seen those go bad but that's rare.

The stuff vented from electrolytic caps is both corrosive and conductive (even the residue) so wash the PCB well. And dry, obviously, take your time with it. And be careful if you haven't repaired a few of these before, it can be dangerous due to line connection and charged primary cap. I suggest you use 100W bulb as input current limiter, this should be enough to let the PSU turn on without (or minimal) load but will protect any possible replacement bridge and transistor from failing again if you missed something. Also is safer for you (except touching the primary cap).

Deunan wrote on 2024-05-18, 23:54:

Shorted rectifier is usually effect and not a cause, that is to say you should also check (desolder if necessary) the switching transistor(s). When the secondary caps vented it might have caused a spike that upset (or even killed) the PWM IC, which in turn overloaded the transistor. But it could be just the bridge, I've seen those go bad but that's rare.

The stuff vented from electrolytic caps is both corrosive and conductive (even the residue) so wash the PCB well. And dry, obviously, take your time with it. And be careful if you haven't repaired a few of these before, it can be dangerous due to line connection and charged primary cap. I suggest you use 100W bulb as input current limiter, this should be enough to let the PSU turn on without (or minimal) load but will protect any possible replacement bridge and transistor from failing again if you missed something. Also is safer for you (except touching the primary cap).

Really useful, thank you! I'll try desoldering and testing those today, before I commit to ordering parts. (I did, of course, carefully check there was no charge using a meter before touching any parts directly!)

Thinking about alternatives, I was looking at PicoPSUs and the handy looking 'vintage computer' breakout boards designed for them (I like the fuses!), but when the board only takes 12v and standby 5v, that seems like unnecessary complexity ...

USB-C PD with a trigger board set to 12v, plus a buck converter for the 5VSB seems like a possibility? (I noticed the board wouldn't boot unless 5VSB was connected when I was lashing it up to an ATX supply)

The old PSU is only rated for 100w total anyway, so comfortably within the rating of USB-C PD.

Or perhaps a 12v+5v open frame Meanwell type power supply board?

Either way, is there any simple way to replicate the ATX soft power on?

Well I figured the easiest and cheapest way forward is a generic PicoPSU and hacking up a 24pin ATX extension cable. Figure I can cut off the other end of the cable and solder on only the connectors I need, and probably still be able to hide the whole thing within the original PSU case.

It feels like powering the drives direct from the PicoPSU, rather than via the step down circuitry on the motherboard might help be a bit kinder to the old board too?

Pick whatever suits you best. I would try to fix the original PSU - but I do have some (if limited) experience fixing these. Pretty much everything can be replaced except the transformers, and sometimes you can try to swap those as well but fo custom designs like this one it probably wouldn't work.

Deunan wrote on 2024-05-19, 23:35:

Pick whatever suits you best. I would try to fix the original PSU - but I do have some (if limited) experience fixing these. Pretty much everything can be replaced except the transformers, and sometimes you can try to swap those as well but fo custom designs like this one it probably wouldn't work.

Thanks for the encouragement! I will keep it, and keep replacement caps etc in the Farnell basket for when I next have other things to order. But I am put off by the prospect of ordering one set of parts only to find there is a problem with something else as well.

It's annoying there doesn't seem to be a UK electronics component supplier that's both comprehensive and doesn't penalise small orders ...

I guess that least it is most likely the 5vSB short is caused by one or more of the failed capacitors? But that wouldn't necessarily explain the fuse being blown?

justin1985 wrote on 2024-05-20, 07:09:

I guess that least it is most likely the 5vSB short is caused by one or more of the failed capacitors? But that wouldn't necessarily explain the fuse being blown?

IMO any dead short on the output rails should be detected as overcurrent by the PWM IC, and should trip the PSU into shutdown. Now the standby rail is bit different but either it's limited pulse width anyway or a separate free-running circuit. Neither should have the ability to blow the fuse.

My guess is the caps on the secondary main line (5V usually) went bad and this caused very unstable operation, which might just upset the IC enough to kill the switching transistor. These usually go short, blow the fuse and possibly short the rectifier as well. Then they go open by exploding. If there is no PCB damage I'd replace the fuse, rectifier, check and possibly replace the transistor(s), and any caps in the base drive circuits - as well as the obviously dead ones and anything else tired or marginal I can find. Primary side caps are usually OK and the ESR isn't really critical here. Bring the PSU up with isolated variac with 100W lightbulb in series. If it fails look at it again (replace the IC if cheap and available), if it works remove the 100W bulb and try some 12V car lightbulbs as load. Verify voltages when loaded.

Deunan wrote on 2024-05-20, 10:57:

My guess is the caps on the secondary main line (5V usually) went bad and this caused very unstable operation, which might just upset the IC enough to kill the switching transistor. These usually go short, blow the fuse and possibly short the rectifier as well. Then they go open by exploding. If there is no PCB damage I'd replace the fuse, rectifier, check and possibly replace the transistor(s), and any caps in the base drive circuits - as well as the obviously dead ones and anything else tired or marginal I can find. Primary side caps are usually OK and the ESR isn't really critical here. Bring the PSU up with isolated variac with 100W lightbulb in series. If it fails look at it again (replace the IC if cheap and available), if it works remove the 100W bulb and try some 12V car lightbulbs as load. Verify voltages when loaded.

I'd put this away (and not visited Vogons for ages) and hadn't seen this reply until after I'd already bought and fitted replacements for all the capacitors from Farnell.

With all fresh electrolytic (and disc) capacitors, and the other components that looked physically damaged replaced (thermistor, bridge rectifier) - it still has a dead short to ground on the 5VSB rail. 12v rail has no short, so I guess that narrows down the fault, at least?

There wasn't any damage to the PCB, other than slight marks around the caps that had burst (which I've cleaned up).

So it definitely sounds like the transistor(s) - just as you said!

Would the main suspect be the secondary side main transistor with the heatsink? Or the small ones near the lower 5VSB connection? Would they all need desoldering to test?

(Thinking about it seems weird that 5VSB is connected to both edges of the PCB (in this photo) - it's made trying to follow the short with a meter really complicated!)

I've invested enough time (and components) in this now that I'm really determined to fix it! 😀

justin1985 wrote on 2024-07-23, 08:57:

it still has a dead short to ground on the 5VSB rail

There is only one transformer on this mobo, no extra inductors, the 5VSB must be generated by the main circuit then.

The Q1 is the main switching transistor but it's on the primary side, and while it can also be damaged you first need to look for the obvious source of the short on the secondary side. Since this PSU only makes 12V and 5VSB the main diode D11 on the heatsink must be for 12V. That leaves that weird D10 diode combo I guess? Not sure if that's a factory job or an attempt at previous repair, I haven't seen these PSUs in person before. Check those. Could be there are two to make the forward drop a bit lower and split the heat dissipation between them to improve reliability.

Note, depending on the configuration these two diodes might test as near-short to GND. But not zero. Try measuring that in diode drop mode if your meter has such a function, and also reverse the test leads. If it tests as close to zero but say something like 100mV drop or so, and much higher resistance with the leads reversed, then good chances are the diodes are OK. In that case try powering the PSU (with no load) through a lightbulb and see what it does.

Deunan wrote on 2024-07-23, 10:32:

The Q1 is the main switching transistor but it's on the primary side, and while it can also be damaged you first need to look for the obvious source of the short on the secondary side. Since this PSU only makes 12V and 5VSB the main diode D11 on the heatsink must be for 12V. That leaves that weird D10 diode combo I guess? Not sure if that's a factory job or an attempt at previous repair, I haven't seen these PSUs in person before. Check those. Could be there are two to make the forward drop a bit lower and split the heat dissipation between them to improve reliability.

Note, depending on the configuration these two diodes might test as near-short to GND. But not zero. Try measuring that in diode drop mode if your meter has such a function, and also reverse the test leads. If it tests as close to zero but say something like 100mV drop or so, and much higher resistance with the leads reversed, then good chances are the diodes are OK. In that case try powering the PSU (with no load) through a lightbulb and see what it does.

Thanks so much for this advice! I'm confident with soldering and generally OK with simpler electronics from my other hobby (model railways) and messing around with Arduinos etc., but I'm certainly out of my comfort zone with this circuit!

I had been suspicious of those 'stacked' diodes before. *From memory*, I recall testing them in circuit and in continuity-test mode on the meter they were beeping as shorts in both directions - I desoldered them and tested again, and (if I remember correctly) they seemed to work as I'd expect a diode to work (i.e. continuity beep in one direction, high resistance in the other). So I put them back - at the time I assumed that meant there was another source of a short in the circuit with them - but you think this test result could be a feature of the overall circuit design? (implication seems to be that the short could just be a spurious 'feature' of the design when no power?) Is it worth desoldering the stacked dioddes again to test in diode drop mode?

Would it also be worth desoldering and testing Q2 (right hand edge) and Q3 (lower right)? Also not sure what U2 is? They all look like what I remember from school GCSE level electronics as 'transistor', but reall difficult to read any of the text on them.

Is the light bulb trick is simply to add a traditional incandescent bulb in series between the mains and the PSU? I guess I could just splice a spare ceiling 'rose' light fitting into a kettle lead, basically? But probably the only incandescent bulbs I have in the house are fancy 'squirrel cage' type ones probably around 40w. In fact I think higher wattage traditional bulbs are now banned from sale in the UK! I guess an LED or CFL bulb wouldn't really do the trick ... The idea is just to limit the current and so cut the chances of it all blowing again if there is still a short?

justin1985 wrote on 2024-07-23, 12:29:

they seemed to work as I'd expect a diode to work (i.e. continuity beep in one direction, high resistance in the other)

Well in that case leave them be. These are high current diodes, they take a lot of abuse to break. So if the out of PCB test says they're good there is little point in suspecting them now. Note the 5VSB goes to some other parts of the circuit on the PSU as well, possibly it's also the power source for the PWM chip. In that case there might be some load on that line that some meters will show as short. The beep usually kicks in at 100 ohms or less, and that's not yet a proper short.

I'll see if I can draw some of the schematic based on the copper traces. U2 is probably some voltage reference, or perhaps temperature sensor. I do wonder what that power part is that has no heatsink on it, right next to the yellow/oragne wires.

EDIT 2: Duh, that must be the MOSFET that turns on/off the 12V line, since the rail will always produce something if the 5VSB is to be maintained.

justin1985 wrote on 2024-07-23, 12:29:

Would it also be worth desoldering and testing Q2 (right hand edge) and Q3 (lower right)? Also not sure what U2 is? They all look like what I remember from school GCSE level electronics as 'transistor', but reall difficult to read any of the text on them.

Eventually, when you can't figure out what is shorted, yes, but don't remove parts just to test them. It's work that doesn't need to be done and there is always some risk of part or PCB damage that you want to avoid.

justin1985 wrote on 2024-07-23, 12:29:

Is the light bulb trick is simply to add a traditional incandescent bulb in series between the mains and the PSU?

Yes, and it needs to be ye olde tungsten wire bulb. Basically it will act as current limiter, greatly lowering the chances that you'll blow something in the PSU again if there is another problem with it. And it also nicely shows the health of the PSU while doing so. You'd expect the bulb to just flash once when power is applied and the primary capacitors charge up. Then it should go dark. It might glow faintly with no load on PSU if it's a lower wattage bulb, something like 50 or 25 Watts. This is because the PSU will pull some current to maintain the 5VSB and it's own power supply.

If the bulb is bright for more than 2s or keeps flashing, turn the thing off - there is a problem and it needs to be fixed. BTW do not leave this PSU under mains power without a fan for more than a minute or so. It might need some airflow even witn no load. And don't touch anything (esp. the primary side!) while it's powered, obviously.

EDIT: You know what, those two big, pinkish power resistors are probably a dummy load to keep the PSU stable with no or light load. These might easily be 100 ohms or less (especially for the 5 volts). So it doesn't have to be an actual short.

Deunan wrote on 2024-07-23, 13:52:

Note the 5VSB goes to some other parts of the circuit on the PSU as well, possibly it's also the power source for the PWM chip. In that case there might be some load on that line that some meters will show as short. The beep usually kicks in at 100 ohms or less, and that's not yet a proper short.

Testing resistance between ground wire and 5VSB gives a value that fluctuates around 30 ohms.

Deunan wrote on 2024-07-23, 13:52:

EDIT: You know what, those two big, pinkish power resistors are probably a dummy load to keep the PSU stable with no or light load. These might easily be 100 ohms or less (especially for the 5 volts). So it doesn't have to be an actual short.

Do you mean the ones tucked in around the purple 5VSB wire top right? There are 3 close to each other, all slightly different values, all under 150 ohms. The one closest to the 5VSB wire is ... 30 ohms! Could this be the apparent short on 5VSB?

(The two big pink blocks by the input inductors are 0.22uf 275v capacitors, which I replaced, probably unnecessarily, because well they're capacitors. When I tested capacitance after removing the old ones, one had a much lower value).

Deunan wrote on 2024-07-23, 13:52:

Yes, and it needs to be ye olde tungsten wire bulb. Basically it will act as current limiter, greatly lowering the chances that you'll blow something in the PSU again if there is another problem with it. And it also nicely shows the health of the PSU while doing so. You'd expect the bulb to just flash once when power is applied and the primary capacitors charge up. Then it should go dark. It might glow faintly with no load on PSU if it's a lower wattage bulb, something like 50 or 25 Watts. This is because the PSU will pull some current to maintain the 5VSB and it's own power supply.

If the bulb is bright for more than 2s or keeps flashing, turn the thing off - there is a problem and it needs to be fixed. BTW do not leave this PSU under mains power without a fan for more than a minute or so. It might need some airflow even witn no load. And don't touch anything (esp. the primary side!) while it's powered, obviously.

Many thanks for this! I was just about to try when I realised that the only tungsten bulbs I have are Edison screw types, and of course the only spare bulb fixtures I have are Bayonet type! Might take a while to track down either a bayonet tungsten bulb, or an E27 bare light fixture...

justin1985 wrote on 2024-07-23, 21:06:

There are 3 close to each other, all slightly different values, all under 150 ohms. The one closest to the 5VSB wire is ... 30 ohms! Could this be the apparent short on 5VSB?

Yes. This is a simple solution to keep the rails loaded and somewhat balanced (esp. when the PSU is in standby mode). Not great for efficiency though. 30 ohms is low (I'd expect something like 51 or 47) but acceptable if that's what it takes to make sure the voltage regulation circuit will work properly.

I found an incandescent bayonet bulb in an abandoned desk lamp at the back of the stationery cupboard at work today, so could test the PSU with this in series.

Deunan wrote on 2024-07-23, 13:52:

Yes, and it needs to be ye olde tungsten wire bulb. Basically it will act as current limiter, greatly lowering the chances that you'll blow something in the PSU again if there is another problem with it. And it also nicely shows the health of the PSU while doing so. You'd expect the bulb to just flash once when power is applied and the primary capacitors charge up. Then it should go dark. It might glow faintly with no load on PSU if it's a lower wattage bulb, something like 50 or 25 Watts. This is because the PSU will pull some current to maintain the 5VSB and its own power supply.

If the bulb is bright for more than 2s or keeps flashing, turn the thing off - there is a problem and it needs to be fixed. BTW do not leave this PSU under mains power without a fan for more than a minute or so. It might need some airflow even witn no load. And don't touch anything (esp. the primary side!) while it's powered, obviously.

It worked exactly as suggested here! The bulb quickly got progressively brighter and then quickly faded back out within a second or two. I went ahead and tested the purple 5VSB wire with a meter, and got a nice stable 5V. Turned it off and then tried again with the PSON green wire shorted to ground with a paperclip, and the fan span up nicely, and the 12V line tested also a nice stable accurate 12V. Success!

The entire time the mains cable is connected there is a quite loud high pitched whine though - I guess from the coils? Is there anything that can be done about that? It reminded me that this kind of whine was much more common from PSUs in the 90s than it is today!

Is there any other testing I should do before connecting to the PC?

Coil whine in standby might get better when the mobo is connected and it adds more load, but probably not much. A closed case will also dampen it a bit. It can also change (for better or worse) with time, as the caps fully form and age a bit.
There are some ways to try and get rid of the whine, not always fully though, but to do it properly you'd need to desolder the transfomer (it's probably the main source of noise) and impregnate it again. That requires special varnish, preferably a vacuum jar, correct drying/baking procedure... a lot of work and materials needed. And if the new layer is too thick it will make the transfomer run hotter, which is also not great.

If the voltages test OK with no load I'd say go ahead and connect it. Do not try to limit the current with a bulb when the mobo is connected, you'll only make it harder for the PSU to start and run. If it didn't blow up yet it shouldn't with the load either. You might want to do a final pass with a brush to clean any possible metallic debris from soldering / cutting leads, and dust (I see some fuzzy stuff on Q1) before you put it back in a case.

Deunan wrote on 2024-07-24, 22:22:

Coil whine in standby might get better when the mobo is connected and it adds more load, but probably not much. A closed case will also dampen it a bit. It can also change (for better or worse) with time, as the caps fully form and age a bit.
There are some ways to try and get rid of the whine, not always fully though, but to do it properly you'd need to desolder the transfomer (it's probably the main source of noise) and impregnate it again. That requires special varnish, preferably a vacuum jar, correct drying/baking procedure... a lot of work and materials needed. And if the new layer is too thick it will make the transfomer run hotter, which is also not great.

Thanks so much for all of your help and advice - REALLY appreciated!

Servicing the main transformer like that sounds like a right nightmare, so I won't be tackling that.

One thing I did notice though is that the small inductor coil next to the output wires was originally thoroughly coated in goop. But that goop had dried right up and all flaked off. Covering this with hot glue might be worth a shot?

justin1985, kudos for tackling this project. What a strange little board!

justin1985 wrote on 2024-07-25, 13:03:

Covering this with hot glue might be worth a shot?

Yes, but if you don't hear much of a difference remove the glue. The problem is again adding any extra layers will interfere with heat dissipation, and while this coil shouldn't get hot (might be warm) the caps nearby will want every bit of airflow they can get to last longer. And the wires are already in the way. So you don't want to create any extra airflow restrictions there if that's avoidable.