Kasreyn wrote on 2026-11-02, 20:27:

Time for some additions to the previous photos of Zasilacz Akyga Basic 700W (AK-B1-700).. […]
Show full quote

Time for some additions to the previous photos of Zasilacz Akyga Basic 700W (AK-B1-700)..

Three out of six #1 #2 #4 I could not get a photo of before because they sit in very inaccessible areas behind other bulky components. #2 was in view but the text wasn't legible.
#1 #2 #4 have now been successfully photographed which you can see in the photos at the end.
In the first overview picture counting from left to right, bottom to top, they are:

#1 CTS P515S4 S30C45F - Located to the left on the left heat sink (bottom)
#2 HS 5SL010 MBR30 100CT - Located to the left on the left heat sink (middle)
#3 CTS P515S4 S30C45F - Located to the left on the left heat sink (top) (seen clearly in previous post)

#4 HS 5SL010 MBR30 100CT - Located to the right on the left heat sink (middle) (this sits on the opposite side in line with #2)

#5 HS 2521 DS09L - Located on the right heat sink (top) (seen clearly in previous post)
#6 HS 2521 DS09L - Located on the right heat sink (bottom) (seen clearly in previous post)

So there you have it. A bit more than previous photographs could reveal. To be honest I should have taken more time on the photos but I didn't want to take the PSU apart (and limited to 5 uploads). Unscrewing the board and lifting it out from the case is sort of required.

Kasreyn I'm still unable to detect any PI coils on the secondary filtering +12V / +5V / +3.3V outputs. Looks like there aren't any present (bad news for ripple suppression 👎 ), with the exception of this one on the +5VSB rail:

Kasreyn wrote on 2026-10-28, 17:08:

I opened up a Zasilacz Akyga Basic 700W (AK-B1-700) and took some pictures.

Better let people judge these components for themselves. But please comment with opinions.
Whether or not they want this to be the keeper of their retro treasures 😁

Alright, since I haven't posted in this thread for a while, let me take a stab at this. 😉
The label promises 700 Watts of power, with the 3.3V rail capable of 35 Amps, 5V rail for 38 Amps, and 12V rail 32 Amps according to the website's PSU label. Let's keep all of these in mind. 😀

We start with these two pictures:
download/file.php?id=251679&mode=view
download/file.php?id=251677&mode=view

The main thing to observe here is the two NPN BJTs (D209L), the "ERL-35" main transformer, and the two small transformers next to it. From this alone, I can confirm with 99.999% certainty that this PSU is based on a half-bridge PWM topology - it's old (mid-late 70's design, IIRC) , it's not very efficient (almost guaranteed to never go above 80%... so 80_Minus (tm)?? 🤣 ), but it's also well-understood and simple to repair & maintain... a bit like an old carburated car, really. Nothing wrong with these designs when done/constructed with good parts, but they are just outdated.

What to know about this design:
Given the two BJTs, this PSU will have a relatively low switching frequency (mid 25 t0 35 KHz, typically). Low switching frequency means the output power will be lower for a given main transformer size compared to PSUs that use higher switching frequencies and/or more modern design. And in this case, the "ERL-35" main transformer is probably going to limit the total output power of the 3.3V+5V+12V+(-)12V rails to 300-350 Watts.
BUT!
All of the above is assuming all of the parts used in this PSU are of good, honest quality. Can we give this assumption a pass here?
- I think not. In fact, I'm almost certain of it, and here's why:
For starters, that "ERL-35" transformer might appear like any other, but what's often the case with these no-name manufacturers, especially the more modern ones, is they will try to cut every corner possible to save on costs. While the core size might adhere to the standard "35" size (35 x 42 x 12 mm EE or EI type), you can bet the copper windings inside are absolute bare minimum (Copper is a relatively pricey metal!) and/or not wound optimally to try and squeeze more turns in there for higher performance. As such, I wouldn't be surprised if the transformer core saturates below 300 Watts... never mind 350 Watts, which is what a proper ERL-35 core should withstand. More likely, I'd rate this PSU at 250 Watts maximum continuous. Anything more should be considered a gamble.
The D209L BJTs are also likely to be counterfeit and not from a genuine semiconductor manufacturer. In fact, this whole PSU reminds me of the KDMPower MIPC MI-X8775CD "775 Watt" PSU I showed some years back on BCN forums (link here... note: it's a very long read).
If anyone read through that thread, you should see the similarities between that PSU and this one pretty easily. In the case of the KDMPower MI-x8775cd, the two BJTs were a pair of E13009 's... which is more or less on par with the D209L 's... yet one blew open before the 3.25 Amp main fuse did. Does that tell you anything? For those who aren't into PSUs and power electronics, this highly suggests that the E13009 in the KDMPower PSU were bad counterfeits. A proper E13009 or D209L in such large packages (TO-3P and/or TO-247) should easily pass more current and blow such small fuse to dust if it ever shorted. Very likely these counterfeit parts have much smaller dies inside and are nowhere near capable of the specs listed on their labels. Remember this, because this is going to be a recurring theme in such cheap PSUs.

Next, we analyze the secondary (output) side with these three pictures:
download/file.php?id=252122&mode=view
download/file.php?id=251679&mode=view
download/file.php?id=252125&mode=view
And given this info:

Kasreyn wrote on 2026-11-02, 20:27:

#1 CTS P515S4 S30C45F - Located to the left on the left heat sink (bottom) #2 HS 5SL010 MBR30 100CT - Located to the left on the […]
Show full quote

#1 CTS P515S4 S30C45F - Located to the left on the left heat sink (bottom)
#2 HS 5SL010 MBR30 100CT - Located to the left on the left heat sink (middle)
#3 CTS P515S4 S30C45F - Located to the left on the left heat sink (top) (seen clearly in previous post)
#4 HS 5SL010 MBR30 100CT - Located to the right on the left heat sink (middle) (this sits on the opposite side in line with #2)

#1 S30C45F is for the 3.3V rail. So total "raw" secondary-side limit for the 3.3V rail is 30 Amps
#2 and #4 MBR30 100CT 's are in parallel and for the 12V rail. So total "raw" secondary-side limit for the 12V rail is going to be roughly 40-45 Amps Amps
#3 S30C45F is for the 5V rail. So total "raw" secondary-side limit for the 5V rail is 30 Amps

And right away, I can tell you something is fishy up there, which tehsiggi spotted correctly and noted with the following comment:

tehsiggi wrote on 2026-10-29, 09:24:

And this is from the MOSPEC datasheet.. that up there is not a genuine mospec part, so I suspect it'll not be better.

Not only are none of the above non-genuine MOSPEC (or other known semi manufacturer) parts, but look at those S30C45F rectifiers again - they are in a TO-220FP (isolated) cases! 30 Amp rating? From a cheap no-name semiconductor manufacturer?? Not a snowflake's chance in hell!
These are more likely to be 20 Amp rectifiers with sanded off part numbers and new labels put on. (Yes, you'd be surprised to what lengths some counterfeit manufacturers will go to.) Or at best, they might be (don't bet on it) 30 Amp parts, but factory rejects... so no telling what they are capable of really.

But let's leave the above aside for a moment.

The reason I mentioned the words "raw power" in my previous text above is because this is the maximum limit of the rectifiers. Of course, one has to remember that if the PSU output filter (output caps and inductors) are not large enough / not up to the task, then the ripple and noise on the output could (and I bet it would!) go out of spec way before those "raw" limits are reached... meaning the PC / attached hardware will not be able to pull such large loads.

And that's the part we go into next - the output filtering. Starting with this quote:

TELVM wrote on 2026-11-04, 10:59:

Kasreyn I'm still unable to detect any PI coils on the secondary filtering +12V / +5V / +3.3V outputs. Looks like there aren't any present (bad news for ripple suppression 👎 )

There aren't any indeed (5VSB excepted).
And from what I can see, the 3.3V rail and 5V rail are both filtered by only a single electrolytic capacitor... presumably 2200 uF(?)
If that's the case, don't expect to pull more than 10 Amps from these before the ripple and noise goes out of spec.
I would personally say even this is too optimistic. I had a cheapo L&C PSU before - actually, very similar to the one that TELVM posted above, except mine was an even more "bare bones" version, with only one 2200 uF cap per rail. When I used that PSU 20 years ago to try and power up a Pentium II 400 MHz PC with nothing more than a single HDD and a very low power S3 Savage 4 (later changed to a Radeon 7000) GPU, I'd get frequent system lockups. Changing the PSU immediately solved those issues. The system hardly used more than 40 Watts from the 5V rail... but that was already enough to cause the noise and ripple to go out of spec.

So if we have to be realistic in regards to the 3.3V and 5V rail outputs of the Akyga AK-B1-700 shown here, I'd rate both of them at 10 Amps max each... and again, even that is probably a bit too optimistic *** (see below)

Furthermore, we can see from the construction of the output side (and also from the Youtube video of the 500W Akyga earlier) that despite the ancient half-bridge design, this PSU was geared/tweaked towards supplying most of the output power from the 12V rail. In that regard, we see the 12V rail has 3 electrolytic caps filtering it. I can't see what their values are... but even if they are only 1000 uF each, that 12V rail should still be capable of at least around 20-ish Amps of clean power (taking into account that there are no PI coils here.) With bigger caps, more could be pulled... but only up to a point.

This is where we move to analyzing the output toroid - that round thingy with all of the wires wound onto it. I can't quite gauge what size it is from the pictures, but I doubt it will be a T130-size core (i.e. 1.3 inches output diameter... or 33 mm for those of us (rest of the world??) who prefer metric 😉 ). More likely, it's T103-size (1.03 inches / 26 mm). Being also of the more lossy -26 type core (yellow color with one side painted white) and not the better and more expensive -52 core (light green with one side painted blue), that also kind of puts a combined limit on the 5V and 12V rails to about 200 Watts continuous and 250W peak.

So conclusions:
- this is a 250 Watt PSU * on a good day *. I'd be worried to try and push it to or past 300 Watts... which of course won't be possible, because the ripple and noise on the output will probably go out of spec before then... except *maybe* the 12V rail. And with those "Hangkon" output caps, don't expect them to last more than 5 years reliably. Well, maybe they will or maybe they won't - that's the problem with cheapo electrolytic caps (you just never know.)

*** In regards to the "too optimistic" comment above...
From my experiments with the KDMPower MI-x8775CD PSU and also from the Youtube video of the 500W Akyga, we can see the cross-regulation of these cheap no-name PSUs is absolutely dismal. Load the 12V rail a bit more or don't load the 5V rail enough and the 5V rail goes too high / out of spec. Conversely, load the 5V rail even a smidgen without much load on the 12V rail, and the 12V rail goes too high on voltage. This is NOT GOOD!

So in regards to using it for a retro PC - that would be a hard NO in my book.

Well, that's all from me for today - hope you enjoyed the wall-of-text a little. 😀

CharlieFoxtrot wrote on 2026-10-28, 10:26:

This is a new one for me, a scrap metal PFC!

Heh, that's a relatively old trick in these cheapo PSUs. Started around the mid-late 2000's, I think.

CharlieFoxtrot wrote on 2026-10-28, 14:26:

Faking passive PFC choke is something else. It's not like it is an expensive component and even manufacturing that fake one isn't free. Everything on that PSU is probably on a level that it barely functions.

Nothing to argue against that last sentence. Seeing what I did in my KDMPower PSU, the design is badly copied and there are quite a few glaring mistakes on the PCB, starting from the AC input. Mix in questionable / possibly counterfeit components, and it's no wonder why so many of these blow up "randomly".

As to the remark regarding PPFCs - yes, the PPFC inductor IS relatively expensive in regards to the other components in the PSU because it contains many winding of copper. Copper is not cheap. Around $14 per kilogram, currently. Compare that to steel, which is... eh, depending if you get any kind of plating or not on it.. about $1.00 to $1.40 per kg max. If one PPFC nets ~100 grams of copper inside, then 10 of them are about $14 worth of copper. When you include the labor to make them, that's raises the price even further. And FWIW, one of these cheap PSUs costs no more than $2-3 to make in China, but can easily sell for $7-10... often even $15-20 in richer countries. So you can see the numbers start to add up quickly and why corners are cut everywhere.

As for why fake the PPFC coil and not completely go without it - that's because it might be a requirement in some markets, like Europe. Of course, unless someone with the right knowledge and tools bothers to check, most of these cheapo PSUs pass a visual inspection to the authorities upon importation. But if/when they get caught: change the company name, change the label - hey now it's a totally different product! (until someone notices again).

CharlieFoxtrot wrote on 2026-10-28, 14:26:

Efficiency of that unit is also atrociously bad, perhaps around 50% or so, even less.

You'd be surprised. These PSUs will be more like in the low to mid 70's at more sane loads and sometimes peak or touch the 80% mark.
Granted this is far from great, I admit too.

But for powering an average Pentium II or 3 PC like they did back in the day, it's not a big deal. Those PII/P3/AMD K6 systems hardly used more than 60-70 Watts under load (the P3's more like 50 Watts with a mid-range DX7 GPU, like a GeForce 4 MX 440, and a single HDD.) At these low loads, modern (and much more powerful) PSUs also tend to have poor efficiency in the high 70's or low 80's. Dropping to low-mid 70's is not that big of a deal at 70W system load. A 70% efficient PSU will draw 100 Watts for that load. An 80%-efficient PSU will reduce that to about 88 Watts. So we are talking about a difference of 12 Watts. That's two average LED light bulbs worth of power consumption. I'm not saying it should be disregarded, but it's also not worth making a big fuss about either, IMO... unless you run that PC 24/7 and you live in a hot climate and/or somewhere with really expensive electricity.

Mike_ wrote on 2026-10-02, 21:18:

I'm using a 10Ω 25W resistor for 12V line and 3Ω 50W resistor for 5V line. The latter is perhaps somewhat light load, so I also […]
Show full quote

I'm using a 10Ω 25W resistor for 12V line and 3Ω 50W resistor for 5V line. The latter is perhaps somewhat light load, so I also tried a 1Ω resistor for 5V line. The whine is still there, but at slightly different pitch. Also, this time heatsink for the BJTs got quite warm really fast (I touched it only after removing the wall plug!). I also tried removing Cdt, but that had no effect. 🙁

Now that I think about it, I actually measured this PSU last winter, and it was fine then. I even used same load for testing. So something has broke down since then... But I think I'm fed up with playing with this thing for now, I'll have another look at it if I actually ever need it.

However, using 1Ω load for 5V line made the other PSU behave. There's no longer 1kHz sine wave in 12V line, and ripple for 5V line is also reduced. So at least I got two of the three recapped PSUs to work properly.

Anyways, thanks for help with debugging these! This got a lot more complicated than I expected a recap job to be. 😀

I returned a bit to this oddly behaving PSU. I probed both outputs 1 and 2 (or C1 and C2), and it looks like output 2 is not behaving quite as it should... it's as if C2 sometimes goes halfway down. Maybe the controller or some other components had got damaged somehow?

Also, it looks like C1 and C2 are both on for part of the cycle.

I set a bit different offsets for channels 1 and 2 so that it's easier to see which is which.

momaka wrote on 2026-11-05, 14:14:

#1 S30C45F is for the 3.3V rail. So total "raw" secondary-side limit for the 3.3V rail is 30 Amps
#2 and #4 MBR30 100CT 's are in parallel and for the 12V rail. So total "raw" secondary-side limit for the 12V rail is going to be roughly 40-45 Amps Amps
#3 S30C45F is for the 5V rail. So total "raw" secondary-side limit for the 5V rail is 30 Amps

Small correction: The two parallel MBR30100CT are on the 5V rail and the S30C45F are one for the 3.3V rail and one for the 12V rail.
So the 5V rail is closer to it's rating with the components, though pretty is different. 3.3V and 12V are still under spec.

It's not a good PSU. It is a very bad PSU for the claimed 700W. If they would just slap a 300W or 250W sticker on it, I might not even be too much against it. But the 700W are just a blatant lie.

momaka wrote on 2026-11-05, 14:14:

Alright, since I haven't posted in this thread for a while, let me take a stab at this. ;) The label promises 700 Watts of power […]
Show full quote

Kasreyn wrote on 2026-10-28, 17:08:

I opened up a Zasilacz Akyga Basic 700W (AK-B1-700) and took some pictures.

Better let people judge these components for themselves. But please comment with opinions.
Whether or not they want this to be the keeper of their retro treasures 😁

Alright, since I haven't posted in this thread for a while, let me take a stab at this. 😉
The label promises 700 Watts of power, with the 3.3V rail capable of 35 Amps, 5V rail for 38 Amps, and 12V rail 32 Amps according to the website's PSU label. Let's keep all of these in mind. 😀

We start with these two pictures:
download/file.php?id=251679&mode=view
download/file.php?id=251677&mode=view

The main thing to observe here is the two NPN BJTs (D209L), the "ERL-35" main transformer, and the two small transformers next to it. From this alone, I can confirm with 99.999% certainty that this PSU is based on a half-bridge PWM topology - it's old (mid-late 70's design, IIRC) , it's not very efficient (almost guaranteed to never go above 80%... so 80_Minus (tm)?? 🤣 ), but it's also well-understood and simple to repair & maintain... a bit like an old carburated car, really. Nothing wrong with these designs when done/constructed with good parts, but they are just outdated.

What to know about this design:
Given the two BJTs, this PSU will have a relatively low switching frequency (mid 25 t0 35 KHz, typically). Low switching frequency means the output power will be lower for a given main transformer size compared to PSUs that use higher switching frequencies and/or more modern design. And in this case, the "ERL-35" main transformer is probably going to limit the total output power of the 3.3V+5V+12V+(-)12V rails to 300-350 Watts.
BUT!
All of the above is assuming all of the parts used in this PSU are of good, honest quality. Can we give this assumption a pass here?
- I think not. In fact, I'm almost certain of it, and here's why:
For starters, that "ERL-35" transformer might appear like any other, but what's often the case with these no-name manufacturers, especially the more modern ones, is they will try to cut every corner possible to save on costs. While the core size might adhere to the standard "35" size (35 x 42 x 12 mm EE or EI type), you can bet the copper windings inside are absolute bare minimum (Copper is a relatively pricey metal!) and/or not wound optimally to try and squeeze more turns in there for higher performance. As such, I wouldn't be surprised if the transformer core saturates below 300 Watts... never mind 350 Watts, which is what a proper ERL-35 core should withstand. More likely, I'd rate this PSU at 250 Watts maximum continuous. Anything more should be considered a gamble.
The D209L BJTs are also likely to be counterfeit and not from a genuine semiconductor manufacturer. In fact, this whole PSU reminds me of the KDMPower MIPC MI-X8775CD "775 Watt" PSU I showed some years back on BCN forums (link here... note: it's a very long read).
If anyone read through that thread, you should see the similarities between that PSU and this one pretty easily. In the case of the KDMPower MI-x8775cd, the two BJTs were a pair of E13009 's... which is more or less on par with the D209L 's... yet one blew open before the 3.25 Amp main fuse did. Does that tell you anything? For those who aren't into PSUs and power electronics, this highly suggests that the E13009 in the KDMPower PSU were bad counterfeits. A proper E13009 or D209L in such large packages (TO-3P and/or TO-247) should easily pass more current and blow such small fuse to dust if it ever shorted. Very likely these counterfeit parts have much smaller dies inside and are nowhere near capable of the specs listed on their labels. Remember this, because this is going to be a recurring theme in such cheap PSUs.

Next, we analyze the secondary (output) side with these three pictures:
download/file.php?id=252122&mode=view
download/file.php?id=251679&mode=view
download/file.php?id=252125&mode=view
And given this info:

Kasreyn wrote on 2026-11-02, 20:27:

#1 CTS P515S4 S30C45F - Located to the left on the left heat sink (bottom) #2 HS 5SL010 MBR30 100CT - Located to the left on the […]
Show full quote

#1 CTS P515S4 S30C45F - Located to the left on the left heat sink (bottom)
#2 HS 5SL010 MBR30 100CT - Located to the left on the left heat sink (middle)
#3 CTS P515S4 S30C45F - Located to the left on the left heat sink (top) (seen clearly in previous post)
#4 HS 5SL010 MBR30 100CT - Located to the right on the left heat sink (middle) (this sits on the opposite side in line with #2)

#1 S30C45F is for the 3.3V rail. So total "raw" secondary-side limit for the 3.3V rail is 30 Amps
#2 and #4 MBR30 100CT 's are in parallel and for the 12V rail. So total "raw" secondary-side limit for the 12V rail is going to be roughly 40-45 Amps Amps
#3 S30C45F is for the 5V rail. So total "raw" secondary-side limit for the 5V rail is 30 Amps

And right away, I can tell you something is fishy up there, which tehsiggi spotted correctly and noted with the following comment:

tehsiggi wrote on 2026-10-29, 09:24:

And this is from the MOSPEC datasheet.. that up there is not a genuine mospec part, so I suspect it'll not be better.

Not only are none of the above non-genuine MOSPEC (or other known semi manufacturer) parts, but look at those S30C45F rectifiers again - they are in a TO-220FP (isolated) cases! 30 Amp rating? From a cheap no-name semiconductor manufacturer?? Not a snowflake's chance in hell!
These are more likely to be 20 Amp rectifiers with sanded off part numbers and new labels put on. (Yes, you'd be surprised to what lengths some counterfeit manufacturers will go to.) Or at best, they might be (don't bet on it) 30 Amp parts, but factory rejects... so no telling what they are capable of really.

But let's leave the above aside for a moment.

The reason I mentioned the words "raw power" in my previous text above is because this is the maximum limit of the rectifiers. Of course, one has to remember that if the PSU output filter (output caps and inductors) are not large enough / not up to the task, then the ripple and noise on the output could (and I bet it would!) go out of spec way before those "raw" limits are reached... meaning the PC / attached hardware will not be able to pull such large loads.

And that's the part we go into next - the output filtering. Starting with this quote:

TELVM wrote on 2026-11-04, 10:59:

Kasreyn I'm still unable to detect any PI coils on the secondary filtering +12V / +5V / +3.3V outputs. Looks like there aren't any present (bad news for ripple suppression 👎 )

There aren't any indeed (5VSB excepted).
And from what I can see, the 3.3V rail and 5V rail are both filtered by only a single electrolytic capacitor... presumably 2200 uF(?)
If that's the case, don't expect to pull more than 10 Amps from these before the ripple and noise goes out of spec.
I would personally say even this is too optimistic. I had a cheapo L&C PSU before - actually, very similar to the one that TELVM posted above, except mine was an even more "bare bones" version, with only one 2200 uF cap per rail. When I used that PSU 20 years ago to try and power up a Pentium II 400 MHz PC with nothing more than a single HDD and a very low power S3 Savage 4 (later changed to a Radeon 7000) GPU, I'd get frequent system lockups. Changing the PSU immediately solved those issues. The system hardly used more than 40 Watts from the 5V rail... but that was already enough to cause the noise and ripple to go out of spec.

So if we have to be realistic in regards to the 3.3V and 5V rail outputs of the Akyga AK-B1-700 shown here, I'd rate both of them at 10 Amps max each... and again, even that is probably a bit too optimistic *** (see below)

Furthermore, we can see from the construction of the output side (and also from the Youtube video of the 500W Akyga earlier) that despite the ancient half-bridge design, this PSU was geared/tweaked towards supplying most of the output power from the 12V rail. In that regard, we see the 12V rail has 3 electrolytic caps filtering it. I can't see what their values are... but even if they are only 1000 uF each, that 12V rail should still be capable of at least around 20-ish Amps of clean power (taking into account that there are no PI coils here.) With bigger caps, more could be pulled... but only up to a point.

This is where we move to analyzing the output toroid - that round thingy with all of the wires wound onto it. I can't quite gauge what size it is from the pictures, but I doubt it will be a T130-size core (i.e. 1.3 inches output diameter... or 33 mm for those of us (rest of the world??) who prefer metric 😉 ). More likely, it's T103-size (1.03 inches / 26 mm). Being also of the more lossy -26 type core (yellow color with one side painted white) and not the better and more expensive -52 core (light green with one side painted blue), that also kind of puts a combined limit on the 5V and 12V rails to about 200 Watts continuous and 250W peak.

So conclusions:
- this is a 250 Watt PSU * on a good day *. I'd be worried to try and push it to or past 300 Watts... which of course won't be possible, because the ripple and noise on the output will probably go out of spec before then... except *maybe* the 12V rail. And with those "Hangkon" output caps, don't expect them to last more than 5 years reliably. Well, maybe they will or maybe they won't - that's the problem with cheapo electrolytic caps (you just never know.)

*** In regards to the "too optimistic" comment above...
From my experiments with the KDMPower MI-x8775CD PSU and also from the Youtube video of the 500W Akyga, we can see the cross-regulation of these cheap no-name PSUs is absolutely dismal. Load the 12V rail a bit more or don't load the 5V rail enough and the 5V rail goes too high / out of spec. Conversely, load the 5V rail even a smidgen without much load on the 12V rail, and the 12V rail goes too high on voltage. This is NOT GOOD!

So in regards to using it for a retro PC - that would be a hard NO in my book.

Well, that's all from me for today - hope you enjoyed the wall-of-text a little. 😀

CharlieFoxtrot wrote on 2026-10-28, 10:26:

This is a new one for me, a scrap metal PFC!

Heh, that's a relatively old trick in these cheapo PSUs. Started around the mid-late 2000's, I think.

CharlieFoxtrot wrote on 2026-10-28, 14:26:

Faking passive PFC choke is something else. It's not like it is an expensive component and even manufacturing that fake one isn't free. Everything on that PSU is probably on a level that it barely functions.

Nothing to argue against that last sentence. Seeing what I did in my KDMPower PSU, the design is badly copied and there are quite a few glaring mistakes on the PCB, starting from the AC input. Mix in questionable / possibly counterfeit components, and it's no wonder why so many of these blow up "randomly".

As to the remark regarding PPFCs - yes, the PPFC inductor IS relatively expensive in regards to the other components in the PSU because it contains many winding of copper. Copper is not cheap. Around $14 per kilogram, currently. Compare that to steel, which is... eh, depending if you get any kind of plating or not on it.. about $1.00 to $1.40 per kg max. If one PPFC nets ~100 grams of copper inside, then 10 of them are about $14 worth of copper. When you include the labor to make them, that's raises the price even further. And FWIW, one of these cheap PSUs costs no more than $2-3 to make in China, but can easily sell for $7-10... often even $15-20 in richer countries. So you can see the numbers start to add up quickly and why corners are cut everywhere.

As for why fake the PPFC coil and not completely go without it - that's because it might be a requirement in some markets, like Europe. Of course, unless someone with the right knowledge and tools bothers to check, most of these cheapo PSUs pass a visual inspection to the authorities upon importation. But if/when they get caught: change the company name, change the label - hey now it's a totally different product! (until someone notices again).

CharlieFoxtrot wrote on 2026-10-28, 14:26:

Efficiency of that unit is also atrociously bad, perhaps around 50% or so, even less.

You'd be surprised. These PSUs will be more like in the low to mid 70's at more sane loads and sometimes peak or touch the 80% mark.
Granted this is far from great, I admit too.

But for powering an average Pentium II or 3 PC like they did back in the day, it's not a big deal. Those PII/P3/AMD K6 systems hardly used more than 60-70 Watts under load (the P3's more like 50 Watts with a mid-range DX7 GPU, like a GeForce 4 MX 440, and a single HDD.) At these low loads, modern (and much more powerful) PSUs also tend to have poor efficiency in the high 70's or low 80's. Dropping to low-mid 70's is not that big of a deal at 70W system load. A 70% efficient PSU will draw 100 Watts for that load. An 80%-efficient PSU will reduce that to about 88 Watts. So we are talking about a difference of 12 Watts. That's two average LED light bulbs worth of power consumption. I'm not saying it should be disregarded, but it's also not worth making a big fuss about either, IMO... unless you run that PC 24/7 and you live in a hot climate and/or somewhere with really expensive electricity.

^ I was hoping for Momaka scrutinizing that gutless wonder's innards with his characteristic surgical precision. 😀 👍

tehsiggi wrote on 2026-11-06, 05:34:

Small correction: The two parallel MBR30100CT are on the 5V rail and the S30C45F are one for the 3.3V rail and one for the 12V rail.
So the 5V rail is closer to it's rating with the components, though pretty is different. 3.3V and 12V are still under spec.

No, I'm pretty sure I am correct on those rectifiers.

45V rectifier will simply not work on the 12v rail of most half-bridge PSUs - it will short-circuit almost instantly due to the very high reverse voltages. The AC PWM pulses at the transformer output of the 12V rail can easily exceed 25V... and since these half-bridge PSUs are center-tapped, you can essentially have 50+ Volts tap-to-tap. Depending on the manufacturer's specific design, some can even exceed 60V tap-to-tap. This is why even 60V rectifiers may get damaged on half-bridge PSUs.
So there's a reason you'll almost never see a rectifiers with anything under 100V reverse voltage rating on the 12V rail.

But if you don't believe me, follow the trances on the picture that shows the solder side.
There you can see that the two MBR30100's center (Cathode) pin gos into two solder points for the output toroid. And if you look at the output toroid, you can see it has the 12V rail winding consisting of two parallel wires (the orange ones). These two orange wires then exit to the 12V rail output side where the filtering capacitors are (again, can be seen on the solder side picture). The thicker red wire is for the 5V rail, and it has fewer turns, as it should.
Back in the days of old half-bridge PSUs, this used to be the opposite: the 5V rail was rated higher than the 12V rail and thus had two thicker parallel wires on the output toroid, whereas the 12V rail only got a single wire. But in the case of these more modern cheapo PSUs, the manufacturer(s) have "tweaked" this to gear the PSU towards a 12V-heavy load. Unfortunately, they also tweaked a few other things, which is why these "modern"cheapo PSUs do so poorly with load regulation at any load.

TELVM wrote on 2026-11-06, 12:12:

^ I was hoping for Momaka scrutinizing that gutless wonder's innards with his characteristic surgical precision. 😀 👍

Thanks, glad you liked it. 😀

As an additional note to the above PSU, I forgo to mention that the only good thing about these modern cheapo PSUs is that they finally started using proper PWM / PWM-FET offline ICs for the 5VSB generation, which not only makes the 5VSB more efficient, but also much less likely to damage attached PC hardware should something go wrong with the 5VSB caps.
But so much for that when the rest of the PSU is build like crap.

Mike_ wrote on 2026-10-02, 21:18:

Also, this time heatsink for the BJTs got quite warm really fast (I touched it only after removing the wall plug!).

That's not good (well, obviously 😁 ) - a properly working half-bridge PSU can usually power 10-40 Watts of load without the primary (BJT) heatsink getting too hot, even without a fan. Usually, only the output / secondary-side heatsink gets warm/hot due to losses from the output rectifiers.

The whining/whistling you hear along with the odd noise on the oscilloscope tells me the BJTs on the primary side are not driven properly.

Before anything else, first check the primary input caps. If one has gone open or very low on capacitance, that can cause all sorts of issues. Generally, the primary caps very very rarely fail (almost never, really... but *not* never 😀 ).

Also, what capacitors (brand and series) did you use for recapping the PSU? I mention this because I've had issues with old CWT PSUs not working well with very low ESR electrolytic caps.

And while on the topic of caps, I can assume you also changed the two small electrolytic caps close to the BJTs?
If the PSU worked fine before, but the new problems appeared after the recap, it's more likely the problem was cause by or during the recap process (e.g. cracked solder joint somewhere.)
Since the MPP cap on the primary side got damaged (most likely by the improperly-driven BJTs), you might also want to check the ceramic snubber cap + resistor on the primary. It's usually something like a 2 nF + 22 to 100 Ohm resistor in series. Check both and maybe even just plain change the ceramic snubber cap "blindly". Sometimes, testing such high voltage parts on a meter is not adequate, because they can be breaking down under load / at the operating voltages.
On that note, same goes for the (usually) ceramic snubber caps (with series resistors) on the output of the PSU - I've seen these measure fine but breakdown when the PSU is actually operating. Just replace and see if that helps.

If that still doesn't do it, replace the entire BJT driver section - both on the primary and on the secondary side. The primary usually consists of (-per- BJT) one or two diodes (not counting the protection diode, though), one electrolytic cap (1 to 10 uF typically... which I mentioned above), and two resistors (one usually 1 to 2.2 Ohms, and the other around 2.2 to 4.7 KOhms.) As for the drive circuit on the secondary side, that's usually made up of two c945 TO-92 BJTs, some resistors, some diodes, and typically an electrolytic cap or two. Make sure the resistor supplying the center-tap of the driver transformer is not reading high resistance. Also check the protection diodes across each c945. Better yet, just change the protection diodes and the c945 driver BJTs altogether. Sometimes, they might appear "fine" in circuit and even so outside. But a leaky protection diode there can cause all sorts off issues for the drive signal.

If none of that helps anything, change the PWM IC as a last resort. Usually these don't fail unless their Vcc goes way too high (can happen on ATX PSUs with 2-transistor 5VSB circuits that like to go over-voltage when their critical cap fails.)

I don't really like the "shotgun" approach to fixing PSUs by changing multiple parts... but sometimes, that could be the more effective method to find what's causing the issue. Ideally, if changing any of the parts above fixes the issue, do the switch again and put the old part(s) back in place one by one to see which one was the cause - that is, if you care to know which part caused the issue. I always do, because it gives me insights and/or knowledge about how various parts can fail.

momaka wrote on Today, 08:29:

45V rectifier will simply not work on the 12v rail of most half-bridge PSUs - it will short-circuit almost instantly due to the very high reverse voltages. The AC PWM pulses at the transformer output of the 12V rail can easily exceed 25V... and since these half-bridge PSUs are center-tapped, you can essentially have 50+ Volts tap-to-tap. Depending on the manufacturer's specific design, some can even exceed 60V tap-to-tap. This is why even 60V rectifiers may get damaged on half-bridge PSUs.
So there's a reason you'll almost never see a rectifiers with anything under 100V reverse voltage rating on the 12V rail.

Didn't see the second pictures where we can actually see the colors of the output coil going to the PCB, my bad. That makes this whole thing even more weird.. going for the parallel diodes on the output for the rail that least needs it for it's rating. Lovely.

One more reason to believe they just rolled the dice on the sticker. 😁 Basically they took a low tier "12V centric" PSU and slapped a "I can do everything" sticker on it. If you'd consider it normally being rated for, let's say, 20A max per 3.3V/5V rail with a combined max of 150W or something, it wouldn't look too bad at least from the rectifier point of view.. leaving the last 150W of this 300W PSU in disguise for the 12V.

The longer I look at it, the less impressed I become..

momaka wrote on Today, 08:29:

That's not good (well, obviously 😁 ) - a properly working half-bridge PSU can usually power 10-40 Watts of load without the primary (BJT) heatsink getting too hot, even without a fan. Usually, only the output / secondary-side heatsink gets warm/hot due to losses from the output rectifiers.

The whining/whistling you hear along with the odd noise on the oscilloscope tells me the BJTs on the primary side are not driven properly.

Yes, I also suspected that they are not driven properly as output C2 on KA7500B doesn't go down properly. Instead it goes just halfway down on some cycles. There's an oscilloscope picture of both C1 and C2 outputs (E1 and E2 are grounded in the PSU) in this post.

Re: Any sense buying modern PSU for old hardware?

momaka wrote on Today, 08:29:

Before anything else, first check the primary input caps. If one has gone open or very low on capacitance, that can cause all sorts of issues. Generally, the primary caps very very rarely fail (almost never, really... but *not* never 😀 ).

It looks like the primary caps are Panasonic/Matsushita based on the M logo and "Japan". So I left them in place, here's a picture of them I found from my phone.

On a side note, one of the primary caps on another AT PSU I recapped had started to bulge. But that one still worked before and after the recap. 😀

momaka wrote on Today, 08:29:

Also, what capacitors (brand and series) did you use for recapping the PSU? I mention this because I've had issues with old CWT PSUs not working well with very low ESR electrolytic caps.

And while on the topic of caps, I can assume you also changed the two small electrolytic caps close to the BJTs?

Yes, I replaced all electrolytic caps except the primary caps. I used Rubycon ZLH series for output caps, and YXF series for most of the small caps apart from the 4.7µF ones, which are some Vishay's general purpose series. ZLH series is low ESR, but not ultra-low ESR so I assumed they would be fine on a PSU.

momaka wrote on Today, 08:29:

If the PSU worked fine before, but the new problems appeared after the recap, it's more likely the problem was cause by or durin […]
Show full quote

If the PSU worked fine before, but the new problems appeared after the recap, it's more likely the problem was cause by or during the recap process (e.g. cracked solder joint somewhere.)
Since the MPP cap on the primary side got damaged (most likely by the improperly-driven BJTs), you might also want to check the ceramic snubber cap + resistor on the primary. It's usually something like a 2 nF + 22 to 100 Ohm resistor in series. Check both and maybe even just plain change the ceramic snubber cap "blindly". Sometimes, testing such high voltage parts on a meter is not adequate, because they can be breaking down under load / at the operating voltages.
On that note, same goes for the (usually) ceramic snubber caps (with series resistors) on the output of the PSU - I've seen these measure fine but breakdown when the PSU is actually operating. Just replace and see if that helps.

If that still doesn't do it, replace the entire BJT driver section - both on the primary and on the secondary side. The primary usually consists of (-per- BJT) one or two diodes (not counting the protection diode, though), one electrolytic cap (1 to 10 uF typically... which I mentioned above), and two resistors (one usually 1 to 2.2 Ohms, and the other around 2.2 to 4.7 KOhms.) As for the drive circuit on the secondary side, that's usually made up of two c945 TO-92 BJTs, some resistors, some diodes, and typically an electrolytic cap or two. Make sure the resistor supplying the center-tap of the driver transformer is not reading high resistance. Also check the protection diodes across each c945. Better yet, just change the protection diodes and the c945 driver BJTs altogether. Sometimes, they might appear "fine" in circuit and even so outside. But a leaky protection diode there can cause all sorts off issues for the drive signal.

Yes, it must be something that happened during recap...

I actually soldered out secondary side driver BJTs (the two C945s to right from the chip) and measured their threshold voltages with a multimeter, and I also checked the diodes around them without soldering them out. They seemed to be fine, but it could be that the problem doesn't appear until they are at operating voltages.

I assume driver transformer is the small transformer on this picture? Note that this picture was taken before I had changed the 0.68µF cap.

momaka wrote on Today, 08:29:

If none of that helps anything, change the PWM IC as a last resort. Usually these don't fail unless their Vcc goes way too high (can happen on ATX PSUs with 2-transistor 5VSB circuits that like to go over-voltage when their critical cap fails.)

I actually tried this, as C2 output looked a bit fishy and I happened to have a TL494 on hand. So I soldered the old chip out, and put a socket there instead. Unfortunately changing the chip had no effect, so I put the old one back. 🙁

momaka wrote on Today, 08:29:

I don't really like the "shotgun" approach to fixing PSUs by changing multiple parts... but sometimes, that could be the more effective method to find what's causing the issue. Ideally, if changing any of the parts above fixes the issue, do the switch again and put the old part(s) back in place one by one to see which one was the cause - that is, if you care to know which part caused the issue. I always do, because it gives me insights and/or knowledge about how various parts can fail.

I have stuffed the PSU back into storage for now, but I'll probably dig it up sometimes later and try other things you suggested. I'll let you know if I get it to work, and thanks for the informative post! 😀