Pulled the main filter caps off, they seem okay, both tested almost identically, but if I have some 470 /200v caps in the drawer I will go ahead and swap them. The bridge rectifier looks good as well with a basic diode check.

My only meters are a fluke 45 and a fluke 289. The 289 has a Lo Ohm setting, but ends up still showing 0.08~0.09 ohm between all the transformer pins on the secondary side.

Unfortunately the power supply/s non standard form factor compared to an IBM power supply might require repair to be able to use the case. This supply is quite a bit higher than the IBM and the fan holes don't line up. Maybe worst case can retrofit some kind of meanwell supply in here if it gets to that point, but negative/positive and power good lines have me dreading that idea.

Here in a few hours I should be able to start pulling off components in the circled areas and checking them out for any strange issues.

That is actually AT power supply form factor. Standard stuff back in the day.

Bottom PSU is a PC and XT form factor again standard stuff back in the day, also.

Cheers.

Ah, thanks for the info. I was focusing on the power side switch instead of the more common 4 spade connectors to a push button switch in the front like i was use to back in the day when messing with AT's.

A quick ebay search does indeed pull up a power supply that looks exactly like this form factor. They just want 4 times the price for it than i paid for this whole case.

Hi lowlytech,

That transformer is OK, these readings are normal, they are of the order of miliohm.

Check transistors, diodes and electrolytic capacitors on the primary side (the little ones: C7, C...).

while you are working on it, check some of the solder, two places look like they could short. There could be others....

Thanks Horun, I will get that cleaned up. Also thinking I am gonna replace that extra ground wire. I can see where it is going over components and worried it too may be making a short.

Back to the transformer which I can't seem to keep from bringing up, 🤣 ... Now I have noticed the square metal bracket that goes around the coils has broken loose. It isn't like coming off, but it wobbles around now. Should I just use wood glue, super glue, or epoxy, or something else to secure it back?

AT/ATX half bridge topology has gone through a very long history, becoming incredibly economically efficient.
Its voltage stabilization (power redistribution) is carried out by a common mode choke (circled in yellow), and the power required for regulation is determined by that voltage divider of 3 resistors (red on the left), which are in contact with pin 1 of the 494 PWM controller.
Check that there:
- are resistors with +5V and +12V on 1 pin of the 494 controller,
- so that it itself produces +5V reference on pin 14.
- and it powering 20-30V on pin 12 (the circuit is circled in red in the center. The capacitor is replace immediately and check the integrity of the 2-10 Ohm resistor).

Cool, thanks Shevalier for the schematic and the focused places to look on the pcb.

Went through all the diodes, caps, and resistors and didn't find anything shorted. So far Q1 is a dead short on all 3 pins, but haven't found anything else but a few out of spec resistors. Got the 2SC2625 transistors ordered, but may take 10-15 days to get them. Still need to check the remaining TO-92 Transistors, but in the meantime I will build up a lightbulb tester and wait til I can replace both Q1 and Q2 and report back.

lowlytech wrote on 2023-10-02, 17:10:

Went through all the diodes, caps, and resistors and didn't find anything shorted. So far Q1 is a dead short on all 3 pins, but haven't found anything else but a few out of spec resistors.

If nothing in the input is shorted then the reason Q1 died would be either overheating (which should not happen with no load, even when the fan is not running) or the main transformer saturating (which would imply open diodes on the main rails +5V and/or +12V). Were you testing the PSU with no load? If the feedback for regulation is not great in that particular unit/design then it could also happen because of that. But one would think self-destructing SMPS were a thing of the past in late '80s alerady.

The lighbulb on input should protect your PSU from blowing parts again but if you see it misbehaving then try adding some light load to the PSU first. A simple solution is 5W 12V car lightbulbs on both main output rails. That should be enough to prevent any accidental transformer saturation in case the built-in resistors are open or too high value. In fact you can switch these bulbs for 21W ones (or even 55W) to test the PSU on the bench, without risking any PC parts. Do keep in mind this will also require the input protection lighbulb to be higher wattage or it will introduce too much of voltage drop for PSU to run properly. This will also help you with the ouput voltage testing, the 12V rail might simply need a decent load (at least 10% of the PSU rating) to regulate properly - it's often the case with cheap power supplies. In other words, if the voltage on 12V rail gets in spec with 21W or 55W load then the PSU is good even if the voltage is otherwise a bit too low with little or no load. It would only mean you need a lower rated PSU for your setup, and/or this particular one is just very poor design in general.

Deunan wrote on 2023-10-03, 11:11:

lowlytech wrote on 2023-10-02, 17:10:

Went through all the diodes, caps, and resistors and didn't find anything shorted. So far Q1 is a dead short on all 3 pins, but haven't found anything else but a few out of spec resistors.

If nothing in the input is shorted then the reason Q1 died would be either overheating (which should not happen with no load, even when the fan is not running) or the main transformer saturating (which would imply open diodes on the main rails +5V and/or +12V).

Saturation of the magnetic circuit is impossible due to the presence of a 2.2 μF * 400 V capacitor in series with the primary winding of the transformer. This is its main function.
Also, given that the AT power supply starts in auto-generation mode, if the power transistors are faulty, it will not start at all.

Let's say evasively, the Q1 transistor behaved badly during repairs and was punished.

shevalier wrote on 2023-10-03, 12:54:

Saturation of the magnetic circuit is impossible due to the presence of a 2.2 μF * 400 V capacitor in series with the primary winding of the transformer.

Maybe it went short. After Q1 CE junction became permanent conductor something had to absorb all that mains power and the cap internal short vaporized and "unshorted" itself. Maybe even has some capacity left afterwards.
Q1 should not randomly fail, usually parts fail right away or work for a long time. Sure, it was old so anything can happen but in general I try to look for possible causes of the failure.

Deunan wrote on 2023-10-04, 09:43:

Q1 should not randomly fail

The power supply turned on, but did not stabilize the voltage.
With a damaged power transistor, it wouldn't start at all.

lowlytech wrote on 2023-09-30, 16:14:

5 volts seems fine at 4.93 when loaded down with 3 hard drives and 3 cooling fans, but the 12 volt rail comes in at 10.3

Something just went wrong during the repair.
Perhaps the main capacitors were not fully discharged during manipulation.
Died and died, it's just a transistor.

Well got the power transistors in and installed, plugged into dim bulb tester and the power supply came to life. 12 v rail is at 10.8v, 5 rail is at 4.9v. Haven't tried the PS without connecting it thru the bulb tester, but it will not turn on with 2 hard drives connected. The bulb pulses a few times then it shutdown ( only with 2 hdds connected). Now that were back up gonna go check voltages around the system and study those schematics.

Checked the power supply while plugged into a motherboard with 1 hard drive connected and off the dim bulb tester. Have 11.6V and 4.9V. Think this looks better. Didn't change any other components except the 2.2Ohm resistors on Q1 and Q2 and Q1 and Q2 2625's themselves.

After messing a bit more around with this power supply I think I know now what killed Q1 in the beginning. I didn't realize but the 2625's are not insulated from the heat sinks. Found this out by brushing my hand up against one of the heat sinks when it was on. Gave me a good little buzz. Testing on AC I have about 70-74 volts AC when probing the heat sinks.

When I originally was testing values while this supply was on I guarantee that the heat sink came in contact with the power supply case since I had the board flipped around upside down. I was just concerned about components touching the case and totally didn't think about heat sinks. That must be what popped Q1 in the beginning.

Thanks again for all the help you all have been walking me though ideas and troubleshooting steps. I would say it has been a good learning experience and a success.

lowlytech wrote on 2023-10-09, 16:13:

After messing a bit more around with this power supply I think I know now what killed Q1 in the beginning. I didn't realize but the 2625's are not insulated from the heat sinks. Found this out by brushing my hand up against one of the heat sinks when it was on. Gave me a good little buzz. Testing on AC I have about 70-74 volts AC when probing the heat sinks.

If there is no insulating silicon pad and an insulating plastic shim at the screw, the heatsink is "live". Some old IBM supplies even had big warning labels inside calling this out: "ATTENTION: HEAT SINK IS LIVE!". Most AT supplies I have seen use a common heat sink for Q1 and Q2, though, so they have to insulate the heat sink from the transistors. The metal tab of the transistor is at the collector, and the collector is not shared between Q1 and Q2.

This also explains why you couldn't find any popped component after the "pop" sound: The pop sound was from an arc between the heat sink and the supply case. You might find a spot in the case, though, which looks a bit like a spot welding mark.

lowlytech wrote on 2023-10-09, 16:13:

Checked the power supply while plugged into a motherboard with 1 hard drive connected and off the dim bulb tester. Have 11.6V and 4.9V. Think this looks better.

12V is in spec now, so this is definitely an improvement. 11.6 is still "a bit low". If I am correct with my hypothesis that this supply requires some load on the +5V line, adding extra cards to the ISA bus, like a VGA card, should raise the +12V line even closer to the desired voltage.

As both the 5V and the 12V line are low at the moment, you might also get an improvement by adjusting the output voltage trimmer, if this supply has one. This will raise/lower all voltages of the supply. Before you re-trim the supply, make sure your meter is accurate!. For example, measure the +5V line on a modern ATX supply currently connected to an operating computer (easiest way: plug the probe leads into a 5.25" drive molex power connector), and verify that you get a value quite close to 5V on that 5V line. As 11.6 on the +12V line less than 4% low, I have no better idea for a sufficiently accurate common reference voltage you definitely have at home you can use to check you meter's calibration.

To raise 12V, you can replace the double diode with a Schottky diode 100V 40A in TO-247 like VS-40CPQ100-N3
Or a pair of them connected in parallel, if the printed circuit board allows.