If you can (find them), do use at least entry-level low-ESR capacitors - something like Rubycon YXJ/YXF, Nichicon PS/PM/PW, and United Chemicon LXV/LXY/LXZ. Anything a notch better should be OK too, e.g.: Rubycon YXG, Nichicon HE, and UCC KY/KYB. Going above that should be considered experimental, at least when it comes to recapping old half-bridge platforms like this Logysis. On that note, you could also use 105C-rated general purpose caps in a pinch too. It's generally not recommended as the GP caps may not last as long (especially if they are not good quality brands), but they will still work.

That aside, I'm more worried about the white smoke you saw. Can we see more pictures of the rest of the PSU? Reason I ask is because some Logysis PSUs are not really worth recapping/repairing.

I presume this is for a 5V-heavy PC? If not, then you could easily use a more modern PSU.
And if it is for a 5V-heavy PSU, check the size of the 5V rail rectifiers, as that will really dictate how much current (max, theoretical) you have available on the 5V rail.

Hi Momaka! 😀

I'd follow Momaka's advice, this gentleman knows his PSUs.

No idea where the white smoke came from, but I inspected the board and saw no damage aside from two severely bloated Canicon and AsiaX caps. One of the big 680uf 200v no brand (likely Fuhjyyu) smoothing caps is slightly bloated on top as well. I turned off the power supply immediately when it began smoking to prevent any further damage, heard no pops at all. This is actually a very rare power supply nowadays so I would really like to get it working for a socket 462 system I plan on building soon.

Another angle:

kupa97 wrote on 2026-07-12, 15:16:

... This is actually a very rare power supply ...

It is indeed.

When I first got the power supply, it would power on for about 2-3 seconds then turn itself off. It finally stayed on after a few power cycles, but started producing smoke.

I took the heatsinks off and checked the values of the diodes and transistors using my multimeter. Readings all appear within spec, no physical damage. I did notice what appears to be burn marks near the big transformer, it could also be old burnt flux from manufacturing.

kupa97 wrote on 2026-07-13, 03:28:

I took the heatsinks off and checked the values of the diodes and transistors using my multimeter. Readings all appear within spec, no physical damage. I did notice what appears to be burn marks near the big transformer, it could also be old burnt flux from manufacturing.

The attachment 20260712_214249.jpg is no longer available

Pretty sure it's flux residue, burn marks would appear different.

kupa97 wrote on 2026-07-12, 15:16:

No idea where the white smoke came from, but I inspected the board and saw no damage aside from two severely bloated Canicon and AsiaX caps. One of the big 680uf 200v no brand (likely Fuhjyyu) smoothing caps is slightly bloated on top as well. I turned off the power supply immediately when it began smoking to prevent any further damage, heard no pops at all.

Yeah, it was probably the Canicon caps venting. They are absolute craps. Asia'X too.
I don't see any burn marks around the 5VSB area or the PWM chip... and looking at the PSU's 5VSB, it doesn't appear to use a critical cap for the 5VSB circuit. So more than likely, you are right and nothing has gone bad aside from the capacitors.

In regards to the big 200V Fuhjyyu caps on the primary side and particularly the one that has a slight "bulge", press on the plastic cap/sleeve on top of the cap and see if it gives/goes down or if it really does feel hard, indicating a real bulge underneath. If the plastic cap goes down, then the cap is probably OK. But if it can't, you can cut out and remove the plastic cap sleeve on top to verify. If it's bulged, then you'd probably have to replace it. More often than not, it's just the the plastic that bulges. While Fuhjyyu are also pretty bad in terms of quality, I haven't had any issues yet with their primary caps in any non-APFC PSU and still continue to run them in several PSUs.

Looking at the rest of the pictures, I do see quite a few of the very small caps are Asia'X... and seeing that the PSU does indeed look worthwhile to recap, I'd suggest to change those small caps too, or as many of them as possible.

The 5VSB output filter cap also looks a bit small - probably a 10V 1000 uF or 16V 470 uF or similar judging by the size. I would up that to 1500 uF as it looks the 2nd filtering cap for the 5VSB is probably one of the 6.3x11 mm cans in that jungle of small caps

kupa97 wrote on 2026-07-12, 15:16:

This is actually a very rare power supply nowadays so I would really like to get it working for a socket 462 system I plan on building soon.

TELVM wrote on 2026-07-12, 17:33:

It is indeed. […]
Show full quote

kupa97 wrote on 2026-07-12, 15:16:

... This is actually a very rare power supply ...

It is indeed.

Oh wow, a fully transparent acrylic case?! That is indeed pretty rare. Would go phenomenally in one of those all-acrilyc see-through PC cases if you can even find them anymore. And late socket A is perfectly fitting for the era, as that's when all of this movement started gaining traction.

Yeah, the PSU is definitely worthwhile fixing, and not just for the looks. I see it has overall pretty good parts inside. The 5V rail output rectifier is in a TO-247 case, so it's probably rated for either 30 Amps or 40 Amps (I'm guessing the former, though.) Since this is a half-bridge platform, that means you can pull currents up to what the rectifiers are rated for... and in this case, 30 Amps for the 5V rail is quite nice. The main transformer is proper ERL35 -type (35x41 mm core) and the heatsinks are very generous. The 680 uF caps should be good enough for about 350 Watts total... which is where I expect the rest of the components to top out at too. I couldn't see what the primary-side BJTs are (the transistors attached on the primary heatsink right behind the big 200V caps), but if the two for the main power supply are in TO-220 case, then probably the limit is actually closer to 300 Watts. If they are in TO-3P or TO-247 case, then the PSU might be capable of more. The output toroids - both the 3.3V rail and the 5V/12V rail common toroid look pretty well sized too and with lots of copper on them, which is always good to see. The 3.3V rail one actually uses the more expensive Micrometals -52 core, so that's nice... and means you can probably use even lower ESR caps for the 3.3V rail, like possibly the equivalent of Rubycon ZLQ or Panasonic FM/FR/FS or Nichicon HV or UCC KZE. I also like how they put the thermistor for the fan controller right inside the toroid, as with a 5V heavy load from a powerful socket A system, you can expect that toroid to heat up quite a bit when the 5V currents go pretty high. So the engineers/techies at the factory did the right thigh there. The secondary heatsink is usually the other hottest part... but no one should care that much about it, as rectifiers actually tend to become more efficient (lower voltage drop) at higher temperatures.
And lastly, any PSU manufacturer that bothers to put a fuse in a fuse holder and a connector for the AC input to make the PSU easy to take apart makes the PSU always worth fixing in my book. 😀

tehsiggi wrote on 2026-07-13, 05:18:

Pretty sure it's flux residue, burn marks would appear different.

I agree. Looks 100% like flux to me as well.

TELVM wrote on 2026-07-12, 13:08:

Hi Momaka! 😀

Hey TELVM!
Long time no see. Nice to see familiar names online. 😀

kupa97 wrote on 2026-07-13, 03:28:

I took the heatsinks off and checked the values of the diodes and transistors using my multimeter. Readings all appear within spec, no physical damage ...

^ Cool, kudos to you! 👍 If switching transistors and rectifier diodes passed the test, then I'd try a recap.

With a little bit of luck the smoke was just some caps venting out.

Yep, that 2nd +5VSB outpur filter cap is rachitic:

Me, I'd replace both with 10mm 10V 2200uF to bombproof +5VSB.

This Logysis was indeed "a piece of art" 😀 :

momaka wrote on 2026-07-13, 10:57:

Hey TELVM!
Long time no see. Nice to see familiar names online. 😀

Greetings man! Nice to see you too! 👍

This is all very helpful information, thanks. On the tiny black and white diodes on the secondary side of the board, I'm getting the expected half volt on the forward bias, but I'm also getting a reading on the reverse bias in some cases over 1 volt. Is this because they're still in circuit and the interference is giving me a misleading value?

I would first replace all the larger caps and only then if the PSU isn't working properly replace the smaller caps. Generally the smaller caps aren't stressed as much as the larger ones directly on the power rails.

I would also see if you can de-solder the two large mains capacitors, even if they look okay. They can leak out the bottom, but when they are glued down around the base with goop like in your case, any leakage can be contained and not visible. If you see no signs of wetness underneath them after removal, they're probably okay and you can reinstall them. The glue around the base has probably gone all hard and crusty, so you might have a job chipping it off, so be careful! 😀

kupa97 wrote on 2026-07-13, 18:02:

... I'm also getting a reading on the reverse bias in some cases over 1 volt. Is this because they're still in circuit and the interference is giving me a misleading value?

Probably some resistors / coils in the vicinity are altering the reading. Testing in diode mode, in circuit, if the diode was open you'd get OL both ways. If it was shorted you'd get a close to zero voltage reading both ways.

If you happen to have an old style analog multimeter (needle display), set it at low resistance (~10 Ohm or so) and test the in circuit diode both ways . One way it should display some low value resistance, the other way max resistance (needle all the way to the left). [This test doesn't work with digital multimeters, only analog ones].

To be 100% sure, unsolder one leg of the diode to test it off circuit.

If when off circuit and testing in diode mode you get anything other than OL in reverse bias, diode is open or leaky and should be replaced.

All the voltages look good, even tested the big filter caps and they checked out. Ordered some really nice Rubicons, gonna look real nice when it's all back together soon and hopefully it won't explode. 🙏

Keep in mind that testing voltages doesn't confirm the capacitors are good, your multimeter is just testing the average voltage, it cannot display rapid fluctuation (ripple voltages). If you want to test them in circuit, a better way is to use an oscilloscope.

The reverse voltage across the diode, especially if it's climbing fast and then stabilizing, is likely leakage current from a nearby capacitor. You can sometimes reveal this by waiting until the measured diode voltage stabilizes, then removing the test leads and quickly connecting them again. If the voltage goes up from a higher starting point than before, it's because the capacitors are still charged from your previous attempt.

Great news! The power supply works absolutely perfectly now. Getting the correct voltages on all the rails and it hasn't blown up. This thing looks unbelievable in the dark, the pictures really don't do it justice.

I miss the era of transparent and UV - it has so much charme.

You can see from the outside when it has filled up with dust and you have to clean it!

majestyk wrote on 2026-07-19, 05:53:

You can see from the outside when it has filled up with dust and you have to clean it!

I recall dust filters being a thing, even back in the day.
Cheap variant was pantyhose fixated over the intake fans. Wild days..