Check the diodes on the secondary side, both 12V and 5V. Open diode on the 12V rail would cause pretty high drop and a lot of ripple. Check the soldering too, could be cracked solder somewhere.
This is unrelated but I don't like the scratches near the switching transistors - you might want to put some insulating coating on that to prevent any possible arcing when dust builds up. And speaking of dust, you can wash that PCB with running water plus soap and some soft brush - just give it a day or two to dry out.

In my experience the primary side caps are rarely bad but the secondary side ones are often toast so replacing them was a good idea even if they looked OKish.

Deunan wrote on 2023-09-30, 16:50:

Open diode on the 12V rail would cause pretty high drop and a lot of ripple. Check the soldering too, could be cracked solder somewhere.

Good advice. The +12V rail in this supply is rectified by the heatsinked double-diode D22 (behind R40), labelled "C2588 / 02C". There are two 12V AC output pins on the transformer, which have opposed polarity: When one of these pins is around +12V, the other one is around -12V. The supply relies on the fact that one of the two diodes in the double diode is conducting nearly all the time, providing a steady source of current for the +12V rail. If one of the two paths from the transformer to through that diode does not work anymore, low +12V is the expected consequence.

But there might be another issue: The voltages in the supply are coupled quite closely, and the same regulation applies to all of them. Usually, the +5V output is monitored, and all output voltages follow the +5V output. If the fans (use +12V only) and the hard drives (modern ones?) are the only load, there is way too little load on +5V. This means +5V would be getting too high, and regulation kicks in and turns the whole supply down to not exceed the +5V target. Check the label. I would not be surprised if the label requests a minimum load of 1.5A or 2A on +5V (any 286 motherboard would provide enough load), and your current configuration just draws something like 0.2 to 0.3A on +5V.

Thanks for the suggestions. I have had this supply connected to a 486 mainboard for load and the 5,12 voltages were the same, however the motherboard would not post with this supply and the floppy controller chip on the mainboard got super hot. However with a more modern AT supply I had that same 486 board did post. I do know that C2588/02C has the same 10.3 volts on the center pin. I will check the outside legs for AC and see what I find.

Deunan,

I will get those scratches taken care of with some UV solder mask. Thanks for pointing it out. There was a huge dirt dobber nest right there and I didn't know at the time, but my copper brush was maybe not the best thing to use to get it off there. This thing was a barn find and in pretty bad rusty shape on some of the parts.

Well went from bad to worse. I checked the two ends of C2588/02C and had 32 volts on one leg and 58 volts on the other, however shortly after I finished the reading a loud firecracker pop happened and now the power supply won't turn on at all. Fuse did not blow, nothing looks burned or cracked, so not sure what exactly popped.

lowlytech wrote on 2023-09-30, 18:31:

Well went from bad to worse. I checked the two ends of C2588/02C and had 32 volts on one leg and 58 volts on the other,

Those values sound completely off. One would expect a reading of zero volts in DC mode, or maybe a slightly negative voltage (like -0.8V) there, because it is AC. If you measure in AC mode, and the meter does sensible RMS conversion, you would expect 11V in your case and 12.5V in a working suppy. If the meter does a "peak value / sqrt(2)" conversion to generate an AC reading (that's how the 10$ meter class works), you would expect a false reading of around 8V. Most importantly, you would expect both outer pins of that double diode should behave the same.

If there is no really strange stuff going on when you measured the voltages, it seems like that double diode is broken beyond repair. Possibly this is all what was wrong before...

lowlytech wrote on 2023-09-30, 18:31:

however shortly after I finished the reading a loud firecracker pop happened and now the power supply won't turn on at all. Fuse did not blow, nothing looks burned or cracked, so not sure what exactly popped.

... this happend. Let's assume that the AC readings are high because there are spikes from the switch-mode operation and the meter does in fact do a peak value measurement, and the side that displayed 32 volts is "behaving correctly", there is a severe overvoltage on the side that measures 58 volts. This is likely outside the design specification of that double diode, which might have shorted out, which is an even stronger argument for desoldering D22 and testing it out-of-circuit in diode mode. If the +12V rectifier is shorted, the short-circuit protection feature might kick in and prevent startup of the supply.

Well around the sections of the C25 (88) 02C everything is shorted. Tried to find what was shorting out the area and it seems like it is the transformer. Unfortunately there is no marking or values on this transformer. Hope this isn't a dead end on what to use here as a replacement. The D83-004 and C258802C seem to test fine in diode mode. I don't know how to test a transformer but when it was removed the shorts stopped. Here are my findings on the transformer in Ohm mode

pin 1 to..
pin2= 0.10 ohm
pin3=0.10 ohm
pin4=OL
pin5=0.09 ohm
pin6=OL
pin7=0.11 ohm
pin8=OL
pin9=OL

pin2 to
pin1=0.11ohm
pin3=0.10ohm
pin4=OL
pin5=0.11ohm
pin6=OL
pin7=0.10 ohm
pin8=OL
pin9=OL

pin3 to
pin1=0.11 ohm
pin2=0.10 ohm
pin4=OL
pin5=0.11ohm
pin6=OL
pin7=0.09 ohm
pin8=OL
pin9=OL

pin4 to
pin1=OL
pin2=OL
pin3=OL
pin5=OL
pin6=0.10 ohm
pin7=OL
pin8=OL
pin9=OL

pin5 to
pin1=0.10 ohm
pin2=0.11 ohm
pin3=0.12 ohm
pin4=OL
pin6=OL
pin7=0.12 ohm
pin8=OL
pin9=OL

pin6 to
pin1=OL
pin2=OL
pin3=OL
pin4=0.11 ohm
pin5=OL
pin7=OL
pin8=OL
pin9=OL

pin 7 to
pin1=0.10 ohm
pin2=0.09
pin3=0.09
pin4=OL
pin5=0.11 ohm
pin6=OL
pin8=OL
pin9=OL

pin8 to
pin1=OL
pin2=OL
pin3=OL
pin4=OL
pin5=OL
pin6=OL
pin7=OL
pin9= 0.30 ohm

pin9 to
pin1=OL
pin2=OL
pin3=OL
pin4=OL
pin5=OL
pin6=OL
pin7=OL
pin8= 0.30 ohm

That really looks like the guts from a Packard bell PSU I had, what is the branding?

As for a bad transformer that is usually the one part you don’t want to break.

Long ago I knew a real repair guy that mostly worked on DEC and minis that had “stuff” to help with winding transformers but good lord that could suck, miles of human hair thickness wire.

Usually glued together making manual counting, disassembly and rewinding nearly impossible

No branding that I can find anywhere. There isn't even a FCC id on the case. It basically looks like a 5170 AT clone case basically. The power supply has SPS-200 otherwise everything else is totally generic. Date codes look like 88-89 era.

Board's filthy .... also...

Pulled R4 and it is out of spec to a degree. Looks like it is 2.2ohm and it measured 2.8ohm.

That is 25% out of spec ~ it's done, replace it. Looking at things I believe that's base drive for the chopper transistor beside it, which may not have been turning fully on. This is probably not what went pop (although the transistor should be checked)....output transformer looks OK

So I found a shorted power transistor, it was the 2sc2625 @ Q1. All three pins are shorted together. Pulling this out got one of the two smaller transformers to not be completely shorted on all pins. Do you think the transformers are good? Is there a way to test them if there is no model numbers on any of the transformers. Maybe I need to make a new thread since the fault is no longer low 12 volt supply?

giantclam wrote on 2023-10-01, 00:57:

That is 25% out of spec ~ it's done, replace it. Looking at things I believe that's base drive for the chopper transistor beside it, which may not have been turning fully on. This is probably not what went pop (although the transistor should be checked)....output transformer looks OK

Thanks I didn't see this before my last post. So the big transformer looks ok with those particular pins showing continuity? So currently we have on our list to change 1 resistor (2.2) and a 2sc2625?

Still nervous about something not being good. That pop was pretty loud, I can't believe I can't see some physical signs of damage anywhere.

Understand I treat this like a job for repair that turned up on my workbench, and I've done what visual inspection I can from the images provided ... and from here, in AU, I could spot a cracked resistor ...why only me?

1. the board is filthy ~ this is me saying that's the first job here ; clean it.... all that dust and lint laying about is just asking for leakage/shorts on the HV side
2. the resistor is way out of spec ....I said 'check' it's associated transistor, not replace it outright just because that resistor is bad
3. I've got no visuals of the board after the 'pop' ... I'm blind so to speak ....but I've had PSUs go 'pop' just because step 1 above wasn't done first =)

Thanks giantclam for the help. Here are some current pictures. The board is cleaned and I have some parts still removed.

Parts I hope are OK...

Parts that tested bad/out of spec..

Much better ....

If Q1 went short, it will likely have popped the fuse ~ no other signs of explosion (I can't see solderside) except black soot in glass fuse ; if it's blown, that's likely what you heard...

....if Q1 suddenly failed, and given the age of the unit ... both bulk caps should be replaced (you can remove & measure them if you like, but after ~30years I just replace them) ~ if they pass on excessive ripple to Q1/Q2 it's typically a bad happening...

...replace BOTH chopper transistors, not just the measured bad one...

....check diodes/resistors and replace those caps if not done already (inside blue circle)...

....replace C11, C12, C13 if not done already (yellow circle)...

...the IR3M02 IC is equivalent to a TL494 PWM controller it seems...but more expensive (if it was just a TL494 I'd replace it on sight 'coz they're only a few dollars and it's quicker than testing the driver outputs =) ... but even if the IR3M02 is pin for pin compatible with TL494 doesn't mean it's electrically the same (one would need consult datasheets) ... I would stick a 'scope on it's outputs and check it's switching phases correctly. (could be just me being paranoid, and the outputs should be protected by the driver transformer...but, you know....Q1 is dead, and if that IC had something to do with it you'll blow the new part at switch-on) ...you do this testing without choppers installed of course..

I had recapped the entire board (except the two big filter caps) before the pop incident. If you think they were degraded from the surge I can swap them out again in those areas. The 2.2Ohm resistor on the other side of Q1 was also showing 2.9Ohm, so replaced that with a new 2.2 resistor as well.

I will check all the parts in the areas you circled.

I put the parts back in that I thought tested okay and what is bothering me is the short in all secondary pins on the main transformer 1-7 came back when I reinstalled the transformer. They were not shorted until then and I installed the transformer last. On the attached picture pins 1-2 are on one trace, 3-4 are on another trace, then 5 ,6 are separate, then pin 7 goes to the ground wire.

Edit.. Also the fuse didn't blow, it still shows continuity amazingly.

To properly measure SMPS output transformers, you have to be able to read resistance down to milliOhms ... ie; you measured 0.10R ...100 mOhms ....you need that sort of resolution (plus inductance should be measured at operational frequency)....blablabla, short answer is they look like shorts if you're not aware of the low resistance of these windings. Likewise the fast rectifier diodes usually measure very low on diode-check (0.1 -> 0.22), and that can throw you off if you're unaware.

The pop noise could've come from underneath (or from inside) one of the big bulk caps ... you can't trust old caps (not bulged by dried out)

Replacing all the parts just because "old" is not very economical, even if perhaps the safest thing to do. That being said time is also money so for anyone unwilling to invest all that effort into diagnosis and repair it might be better to just replace parts. Then again the problem with this approach is you learn nothing and if the actual problem has not been fixed you will blow up another set of parts, eventually, trying to get the PSU to run.

These are still pretty simple SMPS - there two sides, primary and secondary. Usually primary side should not die because of a short or overload on secondary side, and even some parts failure would not outright kill the PSU (for example in most cases you can't drive DC to the swiching transistors so it's not possible to turn them on permanently). But every time primary side dies it's a very good idea to check secondary as well.

Switching transistors can go open or short. Open is rare and more often then not is a result of a short that has vaporized parts of the smiconductor structure. So you need to look for what went open if the fuse didn't - the possible parts are the rectifier bridge (always check), the surge protection (some sort of NTC varistor usually) and possibly the coils on the noise filter. Sometimes the copper on the PCB can get vaporized but that is obvious to find and in general quite spectacular when it happens.
If there's 2 transistors usually both go, you can try and leave the other one if it tests good but frankly it might be a good idea to replace it, I would not trust it. With both transistors removed the entire input section, along with caps, can be easily tested - you just connect it to the mains via a 40-60W light bulb. The bulb should light up briefly as the caps charge and then go off. If not you have a short (could be the caps or the rectifier, or parts in the input filter). It's safe and perhaps even recommended to leave the caps reforming with the lightbulb for 30min, nothing should short or smoke, the circuit must be perfectly stable with no load. If the bulb is not on you can add miliamp meter in circuit, modern caps should have next too no leakage after reforming so the current should be below 1mA, if it's higer then either the filter or the electrolytics should probably be replaced because it'll only get worse.

NOTE: There are (or should be) resistors to discharge the primary side caps when there is no load, but that will take time. After you disconnect the PSU from mains do not touch anything on the PCB until you are sure the residual charge is low enough - test with a volt meter to be sure, the resistors could be open for example. Or missing because cheap PSU.

Once you are happy with the primary side, except the transistors, check the secondary side. Make sure all the replaced caps are soldered in correctly polarization wise. If in doubt locate the secondary GND (black wires) and go back from there. Check each rail for a short (note, some PSU can have below 100 ohm load on 5V or 12V lines, that is OK, a short is when you see below 10 ohms on a rail with nothing connected).

If everything checks out and you replace the switching transistors and the PSU dies again, then you can suspect the transformer or the PWM chip. And at this point you probably want another PSU since replacing the parts again is just a waste of money. Unless the PSU is somehow special (different voltages, or form factor) and you badly want it working again. BTW first PSU tests after the repair should also be done with the lightbulb (though 40W might be too small for an old PSU with poor efficiency) - it should prevent permanent damage. You try the PSU with no load and the bulb will be on but dim (or at least not full brightness). You can try up to 100W lightbulb and if that is not enough to get the PSU going by itself, you have more repairs to do.