You can somewhat calc the watts by using the Vf and amps. Divide that by how many in parallel for each. At 10A be about 4.5w each for 2 or 3w each for 3.
I do not think you will need heatsinks but they will get very warm..comparing to a AMD 286 cpu at about 3w but has larger surface area....

Looking through some double diodes datasheets I found this one for SF160xG.

I also found SF1608G diodes on chinese fleabay for dirt cheap, so I'll just change the two SF1603G (Vf ~ 1.0 V at 8.0 A) double diodes with SF1608G (Vf = 1.7 V at 8.0 A).

analog_programmer wrote on 2025-01-05, 16:05:

And now the question:
I haven't touched the VR2 trimmer, which is on the main PCB, because I have no idea how it will affect the +12 V, possibly the +3,3 V (this voltage at least is stable - about 3,35 V and not affected by VR1 on the small PWM PCB) and +5 V voltages. From the PSU schematic it looks to me, that VR2 on the main PCB is also relevant to regulating the output voltages. If I'm sure that VR2 only affects the +12 V line, then that would be the easiest solution to the output voltages regulation problem.

So, there's no actual way to only adjust the 12V rail and leave the 5V rail alone and vice versa. It's a group-regulated design, so these two go up and down together only... but seeing how far you've looked into the matter, I think you already know this now. 😉
As for VR2, its location on the diagram suggests that it's probably for adjusting the over-power protection trip point. However, that schematic diagram of the FSP ATX-350PNR provided above doesn't seem to make a lot of sense the way it's drawn. Unfortunately, I can't find a datasheet for the "3528" PWM controller, so I can't verify if my suspicion is correct. In any case, the driver / "middle" transformer in these old half-bridge PSUs has a dual function (actually, tripple!):
- 1) the PWM controller drives the primary-side BJTs via two small transistors on the secondary side, but...
- 2) there's also a winding on this driver transformer that is in series with the main transformer, so when one of the BJTs starts to turn on, it amplifies it's own base signal drive and turns on even harder (so essentially, it's a positive-feedback circuit), allowing the BJT to saturate and turn on fully.
- 3) the pulse created from the transformer winding mentioned in 2) above gets rectified and sent back to the PWM controller as a rudimentary way for it to "see" what the primary side is "doing" - a crude over-power protection (OPP), of sorts. You can see that with diode D10, and resistors R15, R17, R18, and VR2. VR2 essentially sets the "trip point" of the over-power protection for the primary side. As to what doesn't make sense in the schematic above, it's where point "K4" goes to - pin 16, 12V rail sense input. But without a datasheet for the 3528, I can't know for sure.

In short, I wouldn't recommend messing with VR2... or if you do, at least make sure to mark its original position before doing anything. Better yet, pull it out and measure the resistance between its variable resistance output leg and the other two, just so you also have an idea of what resistance it should be in that spot on the circuit as "factory". But again, I suspect it's for primary-side OPP / main transformer saturation protection, so probably best to leave it be for now.

analog_programmer wrote on 2025-01-05, 19:40:

Actually, in my first comment I saved some additional explanations on the root of the problem and unsuccessful repair attempt be […]
Show full quote

Actually, in my first comment I saved some additional explanations on the root of the problem and unsuccessful repair attempt because of the "wall of text" effect. I found an YT-video* from beyond "the (new) iron curtain" in which there's explanation for the very same problem, but the given solution didn't work for me.

In these series of Fortron/FSP power supplies of the time period in question, the core (ferrite) of the output choke (inductor) is made of not very good quality material and changes its ferromagnetic parameters over time, resulting in the observed effects with a change in output voltages due to altered impedance and inductance of the choke. Also, the windings of the choke in question are not factory wound across the entire ferrite ring and the advice from the youtuber was to rewind them with an even distribution across the entire core of the choke, and also removing 1-2 windings from the +12 V winding with the increased output voltage. I did all of this, but for my PSU it did not give any noticeable result with improvement in the output voltage values in the problematic +5 and +12 V lines.

I'm assuming that the output choke of my PSU with its almost 20 years of aging and with plenty of heating and cooling cycles has already reached some new stable ferromagnetic properties of its core and the whole circuit just needs new values for some of the components according to the settled new inductance of the choke.

* - The YT-video is in russian language, so I don't feel the need to post the link here (if anyone understands the language and would be interested in seeing it and hearing the explanations, I can post the link).

analog_programmer wrote on 2025-01-06, 19:08:

Here it is the video with the solution, that did not help in my case 🙁

P.S. And this is the first video with the explanation of the off- +5 and +12 voltages problem for the same PSU.

Nice video finds on the subject regarding this PSU. My Russian is... let's call it "super-basic" so I don't embarrass myself too much 😁 ... but I can see (and understand a tiny bit) what the guy did there.

Indeed, it looks like the output toroid may be going marginal in your PSU like the guy shows in his video. As mkarcher noted, the output toroid has both the 12V and 5V rail windings together to reduce cross-load regulation between these two rails. Actually, it's more than just that - the output toroid is literally the output converter of a PWM-driven SMPS: it's an energy-storage component that converts the square-wave PWM voltage pulses from the main transformer into triangular-wave pulses of current. In mathematical terms, it integrates. The reason both the 12V and 5V rail have windings on it is so that both of them can contribute (equally) to the magnetic flux of the toroid's core. If one rail contributes less than the other, the toroid core can transfer some energy to the less-contributing winding, essentially also helping to balance out the voltages on the output.

That being said, I don't see how rewinding the core with the turns distributed evenly around (as suggested in the video) would help better balance the voltages, because the core's magnetic flux should be equal everywhere... and your experiments also confirm this. In the rare case that it does, then there's a good chance the core is starting to fail in a bad way.

On the other hand, if the core of the toroid is starting to degrade, then I re-affirm the idea of trying to "play" with the number of turns for the 12V and 5V rails in hopes of a solution there. My suggestion here, however, would be to add extra turns to either (or both) rails rather than removing turns. Removing turns reduces the magnetic flux density in the core, so each rail will have less effect on the other. It also reduces the inductance that each winding presents to the output of the PSU, which isn't desired here. If there is one thing that may happen to these inductor cores as they age, their inductance factor (Al) may go down. So to (attempt to) make up for that, you need to add extra turns. As to how many, that will depend on how many turns there are for each rail originally. Looking at the notes for a T130-26 inductor I rewound for a Macron Power MPT-3512 ATX PSU, the 5V rail had 5(.5) turns, the 12V rail had 13(.5) turns, and the -12V rail had 15(.5) turns (the 0.5 after each number is due to the fact that each winding ends on the other side of the toroid of where it started, so this adds an extra half-turn.) This made the ratio of these windings equal to 13.5 / 5.5 = 2.454545454[45 in period]. Now, if you take 12 and divide it by 5, you get 2.4. So you can see how that's reasonably close. In case of the toroid I rewound, I could also do 6(.5) turns on the 5V rail, 15(.5) turns on the 12V rail, and 17-18[.5] turns on the -12V rail (it's a loosely-regulated rail, so it shouldn't matter too much here). In fact, I think that's exactly what I did based on the pictures I've taken from the repair. I didn't do the change to fix any cross-regulation issues, though, but more as an experiment. The PSU's old toroid was burned out to a crisp (including the wires) due to poor airflow, so I bought and rewound it a new toroid.

Now, if you have another spare/scrap ATX PSU on hand, you may be able to swap the toroids between the two to see if that fixes your PSU's issues. The toroid of the donor PSU doesn't have to match the size of the one in yours. Only the PSU needs to be of the same design/topology - i.e. half-bridge type (3 transformers in the middle of the PSU, typically.) And you just need to match the pinout of the windings (i.e. 5V, 12V, and -12V rail "in" and "out" pins.)

From what I can see from the Russian videos above, and from pictures of my own old half-bridge based-FSP PSUs, the toroids in these are usually T130-26... that is, around 33 mm in diameter (1.30 inches, hence the 130 in the model number), and the core is Micrometals iron powder mix type "26". These and T106-26 are probably the most widely used toroid cores in group-regulated ATX PSUs, especially back in the day. The other popular one is T130-52 and T106-52 - essentially the same size cores, but from core mix type 52. The two are directly compatible, but 52 allows for higher frequency use and has lower core losses (less heat generated inside the inductor with AC currents.) Type 26 is yellow in color with one side painted white, while type 52 is light green with one side painted in blue. So again, if you have any other scrap PSUs, now you'll know which options can work. 😉

Once you swap the inductor in your PSU... or add turns to the original... don't set VR1 based on the n0-load voltages. Instead, set it with some constant load - preferably around 2 Amps on the 5V rail and 1 Amp on the 12V rail (the 3.3V, -12V, and 5VSB don't need a load.) One easy way to do this is with 12V car/auto light bulbs. Incandescent types rated 20-50 Watts should do the trick. Halogen type - maybe (some PSU's short-circuit protection circuits may trip from the very low cold resistance of halogen bulbs.) I myself use 12V MR16 halogen bulbs rated for 20 Watts (approximately 1 Amp load on 5V and 1.67 Amps on the 12V rail): two on the 5V rail and one on the 12V rail. Once you do this, then set VR1 so that the 5V rail is as close to 5V as possible, or slightly over. Then check that the 12V rail is well within range or re-adjust if not. In general, the 5V rail has more "weight" on the PWM controller's error amp than the 12V rail or 3.3V rail (and you can see this by the resistances of resistors R14, R15, and R23 in the FSP ATX-350PNR schematic going to pin 4 of the PWM controller, with R14 for the 5V rail having a lower resistance than the others.) So setting the 5V rail closer to target 5V is more important than setting the 12V rail close to 12V.

analog_programmer wrote on 2025-01-08, 11:30:

Ok, let's try the electrician's way :) […]
Show full quote

Ok, let's try the electrician's way 😀

Maybe not the smartest idea, but what if I add some 35+ A power diode on the +12 V line after the scre*ed-up output inductor? Its V-drop will reduce the output voltage by 0.6-0.7 V, which will be enough for my case:
...
The problem here is, that I can't find any not so bulky power diode for more than 35 A current. The label on the PSU "says": +12V1 = 18.0 A; +12V2 = 16.0 A. Any suggestions for a suitable less bulky diode?

Well, you don't really need a diode capable of 18+16 (34) Amps of current. If you look at the label again, it should say somewhere on it what the combined power rating of the 12V rail is. I have an FSP AX400-PN PSU rated for 400 Watts, and the power for 12Va + 12Vb rails is given as 300 Watts max... i.e. 25 Amps max. I imagine your 350 Watt PSU will probably have a slightly lower rating than that.
In any case, probably a standard 20 or 30 Amp Schottky rectifier (i.e. SBR2040CT, MBR2060CT, SBL3045PT, and etc.) out of another scrap PSU may do the job just fine. Just wire the rectified output from the inductor to the two anodes of the (dual) diode, and the cathode to the output. If you do use one of these, though, you will need to couple it to the heatsink. Otherwise it will run too hot and die. On the other hand the FX2000A looks like it can handle the load without an extra heatsink, so that is a possible option too.

analog_programmer wrote on 2025-01-08, 19:35:

Actually I was also thinking of two or three smaller diodes in parallel, but maybe I should pick them from a pile of diodes, so that their parameters are nearly equal - I don't want to get an overload by current on one of them.

If the diodes are not well thermally-coupled together, one will always run higher and probably "run away". In the case of the FX2000A diodes, if you want to parallel two of them, then twist their leads together as far back as possible (but not too tight to break them out of the diode body) and solder together - both on the anode and the cathode side. Their leads will transfer heat between each other and better keep the junctions closer in temperature. Speaking of which...

analog_programmer wrote on 2025-01-08, 21:22:

Great! My last concern is their operating temperature and cooling, since I can't attach any heatsinks to them and they will only be cooled by the airflow from the PSU's fan. On the other hand these diodes will not actually work in rectifying mode. I just don't have an idea what amount of heat they will produce in that case.

In the case of the FX2000A diodes, you can solder a strip of copper (or steel-anodized aluminum strip from a scrap laptop heatsink) to the anode and cathode leads - the larger, the better. A steel strip could work too, though will have much worse heat transfer.
In terms of heat output, the datasheet seems to indicate about 0.75V drop for 1-2 Amps of current, and about 0.85V drop for ~10 Amps of current at 25C ambient room temperature. The latter load will generate about 8.5 Watts of heat... which is quite a big deal. Think how hot a 10 Watt hot glue will run (approximately 150-160C). So here, I think the diode(s) will easily reach over 100C without airflow and heatsinking. With airflow and heatsinking, you'll probably be lucky to keep them under 90C. With that said, you'll definitely want these located far away from other parts and closer to the PSU exhaust for extra airflow.

Good info on the inductor ! Some Good ideas too but since most Schotky have a very low Vf of about 0.4v @10A (and much less at 5A) they most likely will not drop the 12v far enough imho as there is a 0.9v differential

4.75 V no load, and the +12 V line about 12.65 V

. OP already mentioned if he gets the +12v much higher than that 12.65 then it trips the OVP , not sure if a 0.2-0.4v drop would be enough...
Agree in theory adding a few winding on the 5V windings of torroid would help it absorb more mag flux from the higher 12v peaks and should boost it slightly, not lessen it (thinking about it as a transformer/load balancer).

Hi together,

I've read this thread so far and would also like to understand the root cause of the matter.
I also have one of these FSP ATX-350PNF PSUs showing the same degradation.
According to my notes from some time ago the voltage values are:

rail: no load / load (optical drive)
--------------------------
+3.3: 3.40 / 3.40
+5V: 4.80 / 4.76
+12V: 12.93 / 13.08
-12V: -12.33 / - 12.50

To summarize the facts written here so far for me personally, it seems the ferrite core of the output filter (L1.x in the schematic posted in the beginning) is the main suspect. Please correct me someone if I'm wrong.

Regarding this I have some questions:

1) Wouldn't a reduction in inductance of the bespoken ferrite core influence all rails the same way (e.g. all going up or down, respectively) or, would it rather lead to a shift in "weight" of the individual rails on the regulation which would then lead to the effects seen (i.e. the rails drifting away in opposite direction)?

2) If it really is the core itself, would it be possible to remove the hole part from the PCB, remove the windings from the core, replace the core with a suitable new one and rewind it using the old copper wires? Or would this likely damage the copper wire's insulation? That way you could restore the original parameters without messing up too much by using improper replacements. New ferrite cores are relatively easy to obtain, for the wires one would need the exact diameters, lenghts, turns, etc. So I could imagine replacing the core as the easier task if no complete replacement is available (which may also have degraded already if it's not a new one).

3) If this inductor's ferrite core has degraded over time then what about this other one marked in red (may be L5/6/7 in the schematic)? Could this also have degraded and have an impact on the output rails?

Seeing only one wire on the core and taking the thickness into account I could think of it beeing L5 in the +12V rail. But haven't had a closer look to the PCB for any markings.

I have also found a note regarding thermal aging of such iron powder cores in an old micrometals catalog stating a permanent decrease in inductance caused by high temperatures above 75°C.
https://www.micrometals.com/design-and-applic … ons/literature/
-> "Micrometals Iron Powder Cores for Power Conversion Catalog - Issue L 2007", page 5
This would support this thesis, wouldn't it?

Regards.

momaka, thank you very much for your detailed answer and all the explanations and practical tips!

I can't find any datasheet for FSP3528 PWM controller and protection chip, but I found this web page (sorry, it's russian and you'll have to use some online translator).

The other resources about FSP3528, that I keep on my HDD for years, are these (I don't even remember from what russian sites I got them many years ago):

I managed to understand that +3.3, +5 and +12 V lines are adjusted by VR1 all together, but the VR1's impact on the +3.3 V line voltage is much smaller than on the +5 and +12 V lines (probably because of +5 and +12 V electromagnetic "connection" through the problematic output inductor). I also managed to understand not to mess with VR2 settings - thanks for the conformation on this!

The guy from the YT-videos explains, that ferrite material of this particular FSP output inductor is not good.

I think I can't add windings on +5 V coil of the inductor, nor remove more windings from its +12 V coil, 'cause it will probably break apart (at least its plastic base, I already broke one leg when desoldered one of the +12 V winding ends) after one more repair attempt.

Unfortunately I only have one similar spare inductor from junky Codegen PSU which has thinner wires for windings and is smaller in size compared to original FSP inductor. It's not necessary to mention that Codegen's PSU power ratings are even faker than the JNC's (another well known sh*tty chinese brand) 😀

As for my not so clever idea to increase V-drop on +12 V line by adding some diodes... it's really not so clever as there's not a suitable space inside the PSU case where to put those very hot diodes without causing some thermal damage to nearby components.

I finally came to this (slightly more clever) idea:

analog_programmer wrote on 2025-01-12, 19:40:

Looking through some double diodes datasheets I found this one for SF160xG.

I also found SF1608G diodes on chinese fleabay for dirt cheap, so I'll just change the two SF1603G (Vf ~ 1.0 V at 8.0 A) double diodes with SF1608G (Vf = 1.7 V at 8.0 A).

Yeah, I'll live with some more power decrease (increased power losses due to additional V-dorp on +12 V line), but I'll be happy as the PSU still gives me more than 300 W in total.

What do you think about it?

P.S. momaka, I don't want to bother you with explanatory private messages, but I generally understand and write best in your native language - на български ще се разберем най-добре 😉

ChrisK, in short - the root of the voltage troubles with these particular FSP PNF power supply models (also applies to PNR models) is the bad material used for ferrite core of the output inductor, as seen (heard) from the explanations in the YT-videos, which I found some time ago and posted in one of my previous comments.

The answer to your second question is also in the YT-videos. You can rewind the output inductor (I partially did this, but without some sensible effect), but you'll have to try to do it better than me. It seems that I've not overlapped enough windings of the +5 V and +12 V lines, so the electromagnetic induction between them to increase.

The easiest solution for the problem is also shown in the videos. Replace the bad FSP output inductor with a suitable one from another (non-FSP) PSU model.

As for your third question - probably ferrite cores used in smaller chokes L5, L6, etc. are made of "Ok quality" material or even more likely their inductances don't have much impact on the output voltages as they're more of a filtering parts. See the momaka's explanations on "the big bad output inductor" - it's more than just a filter.

OK, got it.
So first task is getting a new ferrite core as I don't have any parts like that in my parts bins.
I'll try to have look into that asap and report back.

ChrisK wrote on 2025-01-13, 16:01:

OK, got it.
So first task is getting a new ferrite core as I don't have any parts like that in my parts bins.
I'll try to have look into that asap and report back.

Yeah, you can find some suitable inductor replacement from another PSU, this is the easiest solution. Or if you can't find any spare suitable inductor (like me) - you may try my latest idea to change +12 V line double diodes with ones, which have higher Vf value and then use VR1 to slightly rise all the voltages to acceptable values. Now I'm just waiting for approval from momaka and Horun (or someone else who is more into electronics), as they're not against my previous idea for adding some more V-drop on +12 V line by using additional diodes 😀

P.S. A few days ago I was about to give up on discussing the weird problem with this PSU model at all and send the damned thing as spare parts donor in the closet. But eventually I still shared my stupid idea about adding an extra diodes on +12 V line with no hope to get any opinions on it... and Horun responded with comments, and yesterday momaka also wrote his very detailed post... And finally now you came out with confirmation, that this model PSU really does have the voltage problem set out of the factory... Should you have remained "silent" for so long, we could have discussed and resolved the problem much earlier? 😀 I wish you luck with your repair 😉

ChrisK wrote on 2025-01-13, 10:41:

Regarding this I have some questions: […]
Show full quote

Regarding this I have some questions:

1) Wouldn't a reduction in inductance of the bespoken ferrite core influence all rails the same way (e.g. all going up or down, respectively) or, would it rather lead to a shift in "weight" of the individual rails on the regulation which would then lead to the effects seen (i.e. the rails drifting away in opposite direction)?

2) If it really is the core itself, would it be possible to remove the hole part from the PCB, remove the windings from the core, replace the core with a suitable new one and rewind it using the old copper wires? Or would this likely damage the copper wire's insulation? That way you could restore the original parameters without messing up too much by using improper replacements. New ferrite cores are relatively easy to obtain, for the wires one would need the exact diameters, lenghts, turns, etc. So I could imagine replacing the core as the easier task if no complete replacement is available (which may also have degraded already if it's not a new one).

3) If this inductor's ferrite core has degraded over time then what about this other one marked in red (may be L5/6/7 in the schematic)? Could this also have degraded and have an impact on the output rails?

Seeing only one wire on the core and taking the thickness into account I could think of it beeing L5 in the +12V rail. But haven't had a closer look to the PCB for any markings.

1) A reduction of the inductance for each winding on the toroid means each rail will be contributing less to the magnetic flux in the toroid's core, and so each rail will have a lower effect on the other in terms of balancing - i.e. if the 12V rail goes up, the 5V rail will not go up if it is not well-coupled (via inductance) to the toroid.
Alternatively said, imagine if you have two coils of wire, each forming an air-core inductor. If you bring the two close together, a current in one will induce a current in the other. The further you pull the coils apart from each other, the less they will induce currents on one another. Now, these powder iron core toroids are essentially very much like an air-core inductor, but with a [much] higher permeability core (it is iron powder and not air after all.) However, if the iron powder core starts to loose its properties, the effect of the core is the same as the example with the two air-core inductors getting pulled apart further and further from each other. So that's how a degraded core can negatively impact the 5V & 12V windings.

2) Yes, it should be possible to reuse the wire from the old toroid and wind it onto a new one. The tricky part here is that the new toroid inductor needs to have the same core type and size. If you use a different core type and size, you may need fewer or more turns to achieve the same inductance... and there's actually even more than just the inductance. Different core materials have different initial permeability and Al values. This can affect other parameters of how the toroid behaves at various output loads... and certain loads (too high or too low) may break the normal operation of the entire circuit. So the core type needs to be chosen very carefully. As noted in my previous post, generally you can go back and forth between Micrometals mix types -26 and -52 without issues.

3) The one you circled in red is the 3.3V rail filter toroid and corresponds to L3 in the schematic provided on the previous page. It has the same function as the toroid for the 5V and 12V rails. But the reason the 3.3V rail isn't wound on the same toroid for the 5V/12V rails in this case, is because the 3.3V rail on this particular PSU uses what's called a mag-amp (short for magnetic amplifier) circuit. It's essentially a voltage conversion circuit, but it's not exactly DC-DC - it's more AC-AC type. If you look at the picture you provided of your PSU, you can see there is an even smaller toroid on the secondary side near the small middle transformer. This is the mag-amp's AC-AC conversion coil (corresponds to L2 on the schematic.) It's connected to one of the main transformer 5V rail taps, and with the help of a small PNP BJT transistor (usually), it transforms the 5V rail's PWM AC pulses into lower amplitude ones. These lower AC amplitude pulses are then rectified (by rectifier BD2 on the schematic) and sent to the the 3.3V rail filter toroid, L3.

In short, the 3.3V rail has it's own regulation circuit. So that's why it wasn't affected from changes made to VR1 resistance.
Yes, the 3.3V rail toroid can degrade too, but it's extremely unlikely that it will, generally due to much lower loads on the 3.3V rail compared to the 5V and 12V rails. Actually, the only time I have seen a failed 3.3V rail toroid was in a Bestec PSU with failed 3.3V rail output caps. Essentially, the failed output caps caused much higher AC currents to go through the toroid, which caused it to burn out.

ChrisK wrote on 2025-01-13, 10:41:

I have also found a note regarding thermal aging of such iron powder cores in an old micrometals catalog stating a permanent dec […]
Show full quote

I have also found a note regarding thermal aging of such iron powder cores in an old micrometals catalog stating a permanent decrease in inductance caused by high temperatures above 75°C.
https://www.micrometals.com/design-and-applic … ons/literature/
-> "Micrometals Iron Powder Cores for Power Conversion Catalog - Issue L 2007", page 5
This would support this thesis, wouldn't it?

Regards.

Correct!
Powder cores can indeed start to loose their properties around 80°C, and going further up, totally get destroyed.
In the case of the Macron Power MPT-3512 I fixed, I believe the output toroid failed (overheat and fried all of the wires on it) due to insufficient airflow from the overhead 120 mm fan getting blocked by silly fin design of the secondary heatsink.
But that's not the only reason the main output toroid can fail. Having the wrong turns ratio (or one not as close as possible as) the turns ratio of the 5V and 12V rails of the main transformer can lead to the output toroid trying to "make up" too much for the differences in the voltages from the main transformer... and this can cause the core to overheat and fail too. Or, it could be a combination of this and insufficient airflow, causing the toroid to always run hot (but not extremely hot), making it loose its parameters as it ages.

analog_programmer wrote on 2025-01-13, 10:41:

I can't find any datasheet for FSP3528 PWM controller and protection chip, but I found this web page (sorry, it's russian and you'll have to use some online translator).

The other resources about FSP3528, that I keep on my HDD for years, are these (I don't even remember from what russian sites I got them many years ago):

Awesome, thank you very much for these!
Looking at the first one and the last one of these, I see they show pin 16 ("+12V" / "OVP+12V") is indeed connected to the 12V rail through R19. So there's a small error on the schematic of the ATX-350PNR shown on the first page - R19 should have its left side connected to pin 16 and not pin 15 on the FSP3528 IC.

analog_programmer wrote on 2025-01-13, 10:41:

I managed to understand that +3.3, +5 and +12 V lines are adjusted by VR1 all together

Only +5V and +12V are adjusted by VR1.
3.3V can be adjusted by changing VR1a and VR1b: increase value of VR1a and 3.3V rail voltage will go down, or decrease VR1a and 3.3V rail voltage will go up.

analog_programmer wrote on 2025-01-13, 10:41:

The guy from the YT-videos explains, that ferrite material of this particular FSP output inductor is not good.

Not good, as in cheap?
-26 mix type is indeed the cheapest toroid core type for ATX PSUs... but I haven't seen other PSUs have so much problems with it so consistently like apparently FSP seems to be here. So I don't imagine that's the reason.
Perhaps it's not a genuine Micrometals product and some cheaper production from China?? But then, that wouldn't make much sense either, since FSP is a large PSU manufacturer and is more likely to get higher quality parts, especially compared to the really cheap and nasty "gutless wonder" ATX PSUs.

So I do have to wonder if it really is the inductor that's not good here or if FSP goofed up somewhere on the design with this PSU, and perhaps that's why the inductor is failing.
The only other known consistent output inductor failure I know of is with the Corsair CX750M, model 75-002019.

analog_programmer wrote on 2025-01-13, 10:41:

I think I can't add windings on +5 V coil of the inductor, nor remove more windings from its +12 V coil, 'cause it will probably break apart (at least its plastic base, I already broke one leg when desoldered one of the +12 V winding ends) after one more repair attempt.

No biggie if the plastic base breaks apart. 😉 It's only there to hold the wires so that the inductor is easier to insert in place during manufacturing. So if it breaks off, you can just solder wire extensions to the inductor and then solder those in.

analog_programmer wrote on 2025-01-13, 10:41:

Unfortunately I only have one similar spare inductor from junky Codegen PSU which has thinner wires for windings and is smaller in size compared to original FSP inductor. It's not necessary to mention that Codegen's PSU power ratings are even faker than the JNC's (another well known sh*tty chinese brand) 😀

Well, if it's not too much of an inconvenience, you can still try out the spare inductor from the Codegen PSU. The thinner windings and smaller size means it will only be limited to lower currents on each rail (most cheapo garbage PSUs can still usually manage about 200 Watts... so that's still plenty.) In terms of inductance parameters, it should be fine to just drop in, as these half-bridge PSUs don't vary much in terms of the main power design.

I actually did this exact thing with a "550 Watt" BFG PSU (more like a 400 Watt PSU really). It had a burned output toroid inductor due to badly failed output caps. I didn't have one at the time (it was a few years before fixing the Macron Power PSU mentioned above), so I put one -ironically- from a junky JNC PSU that I scrapped over 2 decades ago as my first desoldering learning experience. 😁 The BFG PSU worked fine with it... though I do plan to eventually fix it properly again with new and rightly-sized toroid. (It's just another of the many projects on my To-Do list.)

analog_programmer wrote on 2025-01-13, 10:41:

As for my not so clever idea to increase V-drop on +12 V line by adding some diodes... it's really not so clever as there's not […]
Show full quote

As for my not so clever idea to increase V-drop on +12 V line by adding some diodes... it's really not so clever as there's not a suitable space inside the PSU case where to put those very hot diodes without causing some thermal damage to nearby components.

I finally came to this (slightly more clever) idea:

analog_programmer wrote on 2025-01-12, 19:40:

Looking through some double diodes datasheets I found this one for SF160xG.

I also found SF1608G diodes on chinese fleabay for dirt cheap, so I'll just change the two SF1603G (Vf ~ 1.0 V at 8.0 A) double diodes with SF1608G (Vf = 1.7 V at 8.0 A).

Yeah, I'll live with some more power decrease (increased power losses due to additional V-dorp on +12 V line), but I'll be happy as the PSU still gives me more than 300 W in total.

What do you think about it?

Sounds good and worth trying. And you won't have to do any crazy wiring to add an extra diode in there.
I've actually done the reverse before on an old PSU - upgraded the 12V rail to a low Vf schottky from a high[er] Vf ultrafast, because the 12V was always a bit too low. After the change, it was much closer to spec.
So I think your idea should work fine.
Only thing is, the secondary heatsink might become even hotter than before - but that's only if you use the PSU with a PC that pulls heavily from the 12V rail. If not, then it should all be the same, more or less.
Speaking of which, have you looked where the fan thermistor is placed in your PSU? If it's coupled to the secondary heatsink, the fan will (or should) run louder too... which may be a good thing for the output toroid. 😉

Also, I just hope the cheap Chinese SF1608G diodes aren't some re-marked garbage parts with much lower ratings. But that's always a possibility with fleabay and ali(baba)... and the 7 theifs. 🤣

analog_programmer wrote on 2025-01-13, 10:41:

P.S. momaka, I don't want to bother you with explanatory private messages, but I generally understand and write best in your native language - на български ще се разберем най-добре 😉

Hey, no worries, we can do that too if you like. 😉
Actually, I just see now, I have two PMs from you... but from last year??
I vaguely remember chatting with someone here, but not sure what happened afterwards as I didn't get any further message notifications.

analog_programmer wrote on 2025-01-12, 19:40:

Looking through some double diodes datasheets I found this one for SF160xG.
I also found SF1608G diodes on chinese fleabay for dirt cheap, so I'll just change the two SF1603G (Vf ~ 1.0 V at 8.0 A) double diodes with SF1608G (Vf = 1.7 V at 8.0 A).

I too think is a wise choice and should work better/easier than series diodes.

momaka, kudos for all the info on the inductor ferrite core materials! It's always a pleasure to learn something new from someone with knowledge and practical experience.

momaka wrote on 2025-01-13, 23:26:

Looking at the first one and the last one of these, I see they show pin 16 ("+12V" / "OVP+12V") is indeed connected to the 12V rail through R19. So there's a small error on the schematic of the ATX-350PNR shown on the first page - R19 should have its left side connected to pin 16 and not pin 15 on the FSP3528 IC.

There's no official manufacturer's documentation available, so all these schematics and diagrams are based on someone's "reverse engineering" and mistakes are possible. At least on the big PSU circuit diagram (I don't like the way it's drawn either) the power lines and the components related to them look Ok. As for control circuits and components - I'm not into this and I can't check every connection between all those SMD components on the small PWM controller PCB, but you catch the error at first look.

momaka wrote on 2025-01-13, 23:26:

Only +5V and +12V are adjusted by VR1.
3.3V can be adjusted by changing VR1a and VR1b: increase value of VR1a and 3.3V rail voltage will go down, or decrease VR1a and 3.3V rail voltage will go up.

I agree, my tests on the PSU also confirmed this. As I mentioned, I'm not much into electronic circuits design, so I thought if power lines control circuits are connected at some point (as in some of the schematics for 3528 PWM chip), VR1 adjusts all the three positive voltage outputs. From your detailed explanations now I've managed to understand why in reality this is not valid for the +3.3 V line.

momaka wrote on 2025-01-13, 23:26:

Not good, as in cheap? -26 mix type is indeed the cheapest toroid core type for ATX PSUs... but I haven't seen other PSUs have s […]
Show full quote

Not good, as in cheap?
-26 mix type is indeed the cheapest toroid core type for ATX PSUs... but I haven't seen other PSUs have so much problems with it so consistently like apparently FSP seems to be here. So I don't imagine that's the reason.
Perhaps it's not a genuine Micrometals product and some cheaper production from China?? But then, that wouldn't make much sense either, since FSP is a large PSU manufacturer and is more likely to get higher quality parts, especially compared to the really cheap and nasty "gutless wonder" ATX PSUs.

So I do have to wonder if it really is the inductor that's not good here or if FSP goofed up somewhere on the design with this PSU, and perhaps that's why the inductor is failing.
The only other known consistent output inductor failure I know of is with the Corsair CX750M, model 75-002019.

The russian guy from the YT-videos just mentions cheap chinese ferrite materials and faster ferrite core parameters degradation for cores made of non-quality materials as a possible explanation on the inductor caused voltage problems.

As for the quality of Fortron/FSP products, I don't have any good impressions from their old models PSUs. This one is my second and last ever used PSU from FSP, and it's... well, it's Ok quality overall, but practically it turned out to be a sh*t, because of it's problematic output inductor. The first one Fortron branded PSU, that I bought (before this one) suddenly died into flames and smoke with no obvious reason 😁 It killed a motherboard and a HDD after its two years warranty period had expired. So, to me, in therms of their overall reliability, these old Fortron/FSP PSUs are slightly above bad chinese brands PSUs and below mid-range quality PSUs. It wouldn't surprise me at all, if FSP once used cheap output inductors made of non-quality materials for certain models or batches of models PSUs (like PNF and PNR models).

momaka wrote on 2025-01-13, 23:26:

No biggie if the plastic base breaks apart. 😉 It's only there to hold the wires so that the inductor is easier to insert in place during manufacturing. So if it breaks off, you can just solder wire extensions to the inductor and then solder those in.

I know I can disassemble the inductor once more, if I'm extra careful, but I prefer not to bother with it after my first (unsuccessful) repair attempt 😀 Those +12 V copper (or copper alloy??) windings wires are really hard ("hard" as "made from a firm material") and it's really hard to bend and rewind them - I don't think they're made of typical wire copper material.

momaka wrote on 2025-01-13, 23:26:

Well, if it's not too much of an inconvenience, you can still try out the spare inductor from the Codegen PSU. The thinner windings and smaller size means it will only be limited to lower currents on each rail (most cheapo garbage PSUs can still usually manage about 200 Watts... so that's still plenty.) In terms of inductance parameters, it should be fine to just drop in, as these half-bridge PSUs don't vary much in terms of the main power design.

I actually did this exact thing with a "550 Watt" BFG PSU (more like a 400 Watt PSU really). It had a burned output toroid inductor due to badly failed output caps. I didn't have one at the time (it was a few years before fixing the Macron Power PSU mentioned above), so I put one -ironically- from a junky JNC PSU that I scrapped over 2 decades ago as my first desoldering learning experience. 😁 The BFG PSU worked fine with it... though I do plan to eventually fix it properly again with new and rightly-sized toroid. (It's just another of the many projects on my To-Do list.)

I know I can do this just to confirm once more, that the problem comes from the big output inductor. I have an old healthy JNC ATX PSU labeled for 300 W (in reality maybe about 200 W), so I don't need a second 200 W ATX sh*t 😉

momaka wrote on 2025-01-13, 23:26:

Sounds good and worth trying. And you won't have to do any crazy wiring to add an extra diode in there. I've actually done the r […]
Show full quote

Sounds good and worth trying. And you won't have to do any crazy wiring to add an extra diode in there.
I've actually done the reverse before on an old PSU - upgraded the 12V rail to a low Vf schottky from a high[er] Vf ultrafast, because the 12V was always a bit too low. After the change, it was much closer to spec.
So I think your idea should work fine.
Only thing is, the secondary heatsink might become even hotter than before - but that's only if you use the PSU with a PC that pulls heavily from the 12V rail. If not, then it should all be the same, more or less.
Speaking of which, have you looked where the fan thermistor is placed in your PSU? If it's coupled to the secondary heatsink, the fan will (or should) run louder too... which may be a good thing for the output toroid. 😉

Also, I just hope the cheap Chinese SF1608G diodes aren't some re-marked garbage parts with much lower ratings. But that's always a possibility with fleabay and ali(baba)... and the 7 theifs. 🤣

Thanks for your approval of my revisited not-so-clever idea, I'll go this way with the repair 😀 The thermistor for the fan's RPM regulator is actually soldered on main PCB close to the second bigger heatsink and it touches the lower part of the heatsink. Those SF160xG double diodes are obsolete components. On ali (baba) and his 40 thieves sometimes pop-up genuine old parts for cheap price. I'll try my luck 😀

momaka wrote on 2025-01-13, 23:26:

Hey, no worries, we can do that too if you like. 😉
Actually, I just see now, I have two PMs from you... but from last year??
I vaguely remember chatting with someone here, but not sure what happened afterwards as I didn't get any further message notifications.

Last year I've asked you for advice about electrolytic capacitor full replacement on a failed Chieftec PSU. It turned out that it also lost its "power good" signal (I'm not sure if this was result of my re-cap interventions), so it's still in my unfinished repairs list 😁 I have one (mostly non-workling) laptop PSU "brick" with strange behavior, so maybe I'll contact you later by PM for advice on how to find its defective component(s), after I take some pictures of it in disassembled state as I can't find any circuit diagrams or schematics for it.

Thank you once again for all the help and information!

Horun wrote on 2025-01-14, 03:04:

I too think is a wise choice and should work better/easier than series diodes.

Thanks for your approval on my not-so-wise second idea 😉 I'll do this as the easiest option in my case, as I don't have any suitable spare output inductor replacement.

momaka wrote on 2025-01-13, 23:26:

1) A reduction of the inductance for each winding on the toroid means each rail will be contributing less to the magnetic flux i […]
Show full quote

ChrisK wrote on 2025-01-13, 10:41:

Regarding this I have some questions: […]
Show full quote

Regarding this I have some questions:

1) Wouldn't a reduction in inductance of the bespoken ferrite core influence all rails the same way (e.g. all going up or down, respectively) or, would it rather lead to a shift in "weight" of the individual rails on the regulation which would then lead to the effects seen (i.e. the rails drifting away in opposite direction)?

2) If it really is the core itself, would it be possible to remove the hole part from the PCB, remove the windings from the core, replace the core with a suitable new one and rewind it using the old copper wires? Or would this likely damage the copper wire's insulation? That way you could restore the original parameters without messing up too much by using improper replacements. New ferrite cores are relatively easy to obtain, for the wires one would need the exact diameters, lenghts, turns, etc. So I could imagine replacing the core as the easier task if no complete replacement is available (which may also have degraded already if it's not a new one).

3) If this inductor's ferrite core has degraded over time then what about this other one marked in red (may be L5/6/7 in the schematic)? Could this also have degraded and have an impact on the output rails?

Seeing only one wire on the core and taking the thickness into account I could think of it beeing L5 in the +12V rail. But haven't had a closer look to the PCB for any markings.

1) A reduction of the inductance for each winding on the toroid means each rail will be contributing less to the magnetic flux in the toroid's core, and so each rail will have a lower effect on the other in terms of balancing - i.e. if the 12V rail goes up, the 5V rail will not go up if it is not well-coupled (via inductance) to the toroid.
Alternatively said, imagine if you have two coils of wire, each forming an air-core inductor. If you bring the two close together, a current in one will induce a current in the other. The further you pull the coils apart from each other, the less they will induce currents on one another. Now, these powder iron core toroids are essentially very much like an air-core inductor, but with a [much] higher permeability core (it is iron powder and not air after all.) However, if the iron powder core starts to loose its properties, the effect of the core is the same as the example with the two air-core inductors getting pulled apart further and further from each other. So that's how a degraded core can negatively impact the 5V & 12V windings.

2) Yes, it should be possible to reuse the wire from the old toroid and wind it onto a new one. The tricky part here is that the new toroid inductor needs to have the same core type and size. If you use a different core type and size, you may need fewer or more turns to achieve the same inductance... and there's actually even more than just the inductance. Different core materials have different initial permeability and Al values. This can affect other parameters of how the toroid behaves at various output loads... and certain loads (too high or too low) may break the normal operation of the entire circuit. So the core type needs to be chosen very carefully. As noted in my previous post, generally you can go back and forth between Micrometals mix types -26 and -52 without issues.

3) The one you circled in red is the 3.3V rail filter toroid and corresponds to L3 in the schematic provided on the previous page. It has the same function as the toroid for the 5V and 12V rails. But the reason the 3.3V rail isn't wound on the same toroid for the 5V/12V rails in this case, is because the 3.3V rail on this particular PSU uses what's called a mag-amp (short for magnetic amplifier) circuit. It's essentially a voltage conversion circuit, but it's not exactly DC-DC - it's more AC-AC type. If you look at the picture you provided of your PSU, you can see there is an even smaller toroid on the secondary side near the small middle transformer. This is the mag-amp's AC-AC conversion coil (corresponds to L2 on the schematic.) It's connected to one of the main transformer 5V rail taps, and with the help of a small PNP BJT transistor (usually), it transforms the 5V rail's PWM AC pulses into lower amplitude ones. These lower AC amplitude pulses are then rectified (by rectifier BD2 on the schematic) and sent to the the 3.3V rail filter toroid, L3.

In short, the 3.3V rail has it's own regulation circuit. So that's why it wasn't affected from changes made to VR1 resistance.
Yes, the 3.3V rail toroid can degrade too, but it's extremely unlikely that it will, generally due to much lower loads on the 3.3V rail compared to the 5V and 12V rails. Actually, the only time I have seen a failed 3.3V rail toroid was in a Bestec PSU with failed 3.3V rail output caps. Essentially, the failed output caps caused much higher AC currents to go through the toroid, which caused it to burn out.

@momaka:
This I call a reply! Wow!
Thank you for your overwhelming explanations!

Before going on there's one little note I'd like to throw in between:
This topic is mostly a learning session for me rather than a desperate rescue operation. I don't need that specific PSU for anything important and I also don't consider it high end in any terms. That's why there's no real need on my side for repair and thing's will go on reallllly slow on here. Especially ordering repair parts does usually take quite a long time for me because I always collect stuff first for several projects before acutally ordering. So please have that in mind, for the most time beeing ahead this will be a purely theoretical topic for me.

Anyway, to the questions:
1) This should be clear now.

2) I think we agree that it only makes sense to get the exact same core type. IF I ever dare to try that kind of operation I'd only do so if I could source a 1:1 match.
Do you know if the color codes of the toroids follow some kind of standard? So if there's a yellow one with white marking, is this always material -26, independent of the manufacturer?
As far as I have seen -52 would be preferable to -26 according the specs regarding thermal performace, losses etc., right?

3) I just guessed about the 12V because of the huge wire size used and the 25A rating of that rail. But 3.3V is 22A-rated. So not much less...
OK, then again a good ol' mag-amp here. I have dealt with that concept before in another PSU, so this is not a totally blank slate for me, although I have to admit it's working concept wasn't that easy to grab for me back then. Also, it's circuit is not that easy to see in the way that schematic here is drawn if you are not to overfamiliar with PSU circuit design.
But again, you answered the core point of my question already more than enough.

One other thing that came to my mind regarding the diodes (without having really thought about it yet):
Could it also be a possible solution instead of increasing the drop over the 12V diodes to decrease the drop over the 5V diodes? That way the 5V should come out higher while the 12V could be lowered via VR1.
I wouldn't feel really comfortable with increasing the overall thermal losses by using diodes with higher Vdrop for noise and long term reliability concerns. Also I think this would only be a temporary fix working for now but what if the inductance of the toroid core decreases further? Will that still be enough?
It may be possible to use some low dropout schottky diodes in the 5V rail (I recall you mentioned schottky diodes somewhere here already). They don't need to have that high blocking voltage and the higher reverse current of schottkys shouldn't be that critical, too. What do you think?

analog_programmer wrote on 2025-01-13, 16:19:

P.S. A few days ago I was about to give up on discussing the weird problem with this PSU model at all and send the damned thing as spare parts donor in the closet. But eventually I still shared my stupid idea about adding an extra diodes on +12 V line with no hope to get any opinions on it... and Horun responded with comments, and yesterday momaka also wrote his very detailed post... And finally now you came out with confirmation, that this model PSU really does have the voltage problem set out of the factory... Should you have remained "silent" for so long, we could have discussed and resolved the problem much earlier? 😀 I wish you luck with your repair 😉

😀 Thanks!

ChrisK wrote on 2025-01-14, 15:52:

One other thing that came to my mind regarding the diodes (without having really thought about it yet):
Could it also be a possible solution instead of increasing the drop over the 12V diodes to decrease the drop over the 5V diodes? That way the 5V should come out higher while the 12V could be lowered via VR1.

Yes, actually this is the ideal most simple solution, except... You know, there's always an "except" 😀 I can't find any double schottky diode with lower Vf than the original SBL2045 (Vf = 0.45 - 0.5 V at 10 A at Tj = 125 degree C) on the +5 V rail.

I'm pretty sure that there are no diodes with Vf as low as 0.2 V, but if someone can help with any suggestion, I'll replace SBL2045 double diode instead.

P.S. I know this is not a high-end PSU, it's not even a mid-range quality PSU, but it's an old ATX standard PSU, which (if working properly) is Ok for testing of Athlon XP/P4 or older motherboards. And I don't want to go into the topic "what modern PSU I can use with my ancient PC configurations?". And also I don't like to throw away usable things, if they still can be repaired 😉

Ah ok, my fault. I wasn't aware that there's already a schottky and not a plain rectifier / non-schottky diode in there. Should have had a closer look before. Then this "idea" probably leads nowhere.

Not easily throwing things away if they are still working / repairable is my also my attitude 😀

The plot thickens, as they say.

I got an FSP ATX-300PNF (or was it a PNR, I forgot) for $0.50 from the local flea market last weekend. Haven't had time to "play" with it yet, or even open it for cleaning. But maybe this or next week I'll get to it. Then will test it with my resistive load bank to see how it does.

On a stray note, I really ought to purchase a thermal camera adapter (either for my phone or as a stand-alone camera.) Might come in handy in cases like these, so we can see what exactly is going on with the output toroid in terms of thermals.

ChrisK wrote on 2025-01-14, 15:52:

Before going on there's one little note I'd like to throw in between:
This topic is mostly a learning session for me rather than a desperate rescue operation. I don't need that specific PSU for anything important and I also don't consider it high end in any terms. That's why there's no real need on my side for repair and thing's will go on reallllly slow on here. Especially ordering repair parts does usually take quite a long time for me because I always collect stuff first for several projects before acutally ordering. So please have that in mind, for the most time beeing ahead this will be a purely theoretical topic for me.

Hey, no worries. Most of my projects are exactly like that too. I have some stuff that have been open for close to 10 years now - embarrassing, really! But they do get done.... eventually.

ChrisK wrote on 2025-01-14, 15:52:

Do you know if the color codes of the toroids follow some kind of standard? So if there's a yellow one with white marking, is this always material -26, independent of the manufacturer?

Looks like it, though I don't know if these color codes are necessarily official. More likely, Micrometals company developed the various core types and used the colors to designate them... then the cheap Chinese companies simply copied the design and also stuck to those same color codes for simplicity and less confusion.

ChrisK wrote on 2025-01-14, 15:52:

As far as I have seen -52 would be preferable to -26 according the specs regarding thermal performace, losses etc., right?

Correct.
-52 material has better high-frequency spec and lower losses. But for half-bridge PSUs, which this FSP is, the switching frequency is relatively low (under 30 KHz, usually), so using -26 material core probably isn't that much worse than -52.

ChrisK wrote on 2025-01-14, 15:52:

I wouldn't feel really comfortable with increasing the overall thermal losses by using diodes with higher Vdrop for noise and long term reliability concerns. Also I think this would only be a temporary fix working for now but what if the inductance of the toroid core decreases further? Will that still be enough?

Good point.
I suppose we will have to see what happens. Hopefully we'll all come back here and post our findings if one of us tries this solution.

analog_programmer wrote on 2025-01-14, 11:31:

momaka, kudos for all the info on the inductor ferrite core materials!

My pleasure. 👍
SMPSes, while mundane and boring to most people, I have always found them interesting and exciting. So fixing/repairing them is really another hobby within my retro PC hobby. 😁

analog_programmer wrote on 2025-01-14, 11:31:

The russian guy from the YT-videos just mentions cheap chinese ferrite materials and faster ferrite core parameters degradation for cores made of non-quality materials as a possible explanation on the inductor caused voltage problems.

Yeah, that's quite possible.
What I find strange, though, is that some of the really cheap and nasty PSUs don't seems to suffer from this problem, despite using ferrite materials from who knows what manufacturer in China. So it could also be a design error, where FSP perhaps didn't calculate the number of turns on the output toroid correctly, leading to the core running hotter and getting damaged over time.
Sadly, these are questions that probably go too deep and might not be possible to answer without an x-ray microscope or some other expensive equipment to analyze the material properties in detail.

analog_programmer wrote on 2025-01-14, 11:31:

As for the quality of Fortron/FSP products, I don't have any good impressions from their old models PSUs.

Same here.
And it looks like they pushed that old half-bridge design for way too long. Now, an InWin/Powerman PSU I have (which, IIRC, is made by FSP) seems to be built much better, as in uses much newer double-forward converter design and better-acting protections.

analog_programmer wrote on 2025-01-14, 11:31:

Those +12 V copper (or copper alloy??) windings wires are really hard ("hard" as "made from a firm material") and it's really hard to bend and rewind them - I don't think they're made of typical wire copper material.

Should be solid copper. They are just really thick relative to the core's thickness, making them hard to bend.

analog_programmer wrote on 2025-01-14, 11:31:

Last year I've asked you for advice about electrolytic capacitor full replacement on a failed Chieftec PSU. It turned out that it also lost its "power good" signal (I'm not sure if this was result of my re-cap interventions), so it's still in my unfinished repairs list 😁

Oh I remember now. And saw the old PMs. Sorry about the disappearance on my end. Last year was a bit hectic with a few things I was trying to do and so my time on online forums was sporadic at best.

ChrisK wrote on 2025-01-15, 10:22:

Not easily throwing things away if they are still working / repairable is my also my attitude 😀

Add me in X3 here. 😀
Heck, I probably go too far with way too many things. The other day, for example, I mended a pair of underwear / boxers that had a loose waistband. 🤣 🤣

momaka wrote on 2025-01-29, 00:08:

The plot thickens, as they say.

I got an FSP ATX-300PNF (or was it a PNR, I forgot) for $0.50 from the local flea market last weekend. Haven't had time to "play" with it yet, or even open it for cleaning. But maybe this or next week I'll get to it. Then will test it with my resistive load bank to see how it does.

Well, now you have the opportunity to personally see the problem of these PSU models (PNR = PNF minus PFC reactor) and investigate it 😀 I'm in wait for the delivery of three pieces SF1608G diodes. I've ordered them on chinese fleabay twice as expensive as the price of your "new" FSP PSU 😁

momaka wrote on 2025-01-29, 00:08:

On a stray note, I really ought to purchase a thermal camera adapter (either for my phone or as a stand-alone camera.) Might come in handy in cases like these, so we can see what exactly is going on with the output toroid in terms of thermals.

IR thermal camera is a costly purchase, but such a tool will be very handy in electronics repairs workshop.

momaka wrote on 2025-01-29, 00:08:

ChrisK wrote on 2025-01-14, 15:52:

I wouldn't feel really comfortable with increasing the overall thermal losses by using diodes with higher Vdrop for noise and long term reliability concerns. Also I think this would only be a temporary fix working for now but what if the inductance of the toroid core decreases further? Will that still be enough?

Good point.
I suppose we will have to see what happens. Hopefully we'll all come back here and post our findings if one of us tries this solution.

I'm aware of this and I know that my diode change solution is horrible in terms of wasted power and may turn out to be a temporary solution, but I'll give it a try as it's the easiest variant for me.

momaka wrote on 2025-01-29, 00:08:

SMPSes, while mundane and boring to most people, I have always found them interesting and exciting. So fixing/repairing them is really another hobby within my retro PC hobby. 😁

The vast majority of users neglect and underestimate the PSU as a fundamental part which provides stable operation of the entire computer system. That's why even here we sometimes encounter discussions about instability in systems with slight CPU overclock.

momaka wrote on 2025-01-29, 00:08:

Oh I remember now. And saw the old PMs. Sorry about the disappearance on my end. Last year was a bit hectic with a few things I was trying to do and so my time on online forums was sporadic at best.

I don't know if it would be appropriate to ask you for diagrams/schematics of two mostly disassembled old cr*ppy PSUs, for which I can't find anything matching their actual PCBs and components (fortunately I have kept their main parts like transformers, inductors, MOSFETs, power diodes, heatsinks etc.). I also have one problematic laptop power supply to repair, but I can't find any circuit diagram for it and the root of its problem.

momaka wrote on 2025-01-29, 00:08:

Heck, I probably go too far with way too many things. The other day, for example, I mended a pair of underwear / boxers that had a loose waistband. 🤣 🤣

Nothing wrong here - if something can be fixed and used, you fix it and keep using it (also applies to underwear 😁) 😉

Finally I got those chinese TS (Taiwan Semiconductor??) SF1608G double diodes (I ordered 3 pcs just in case... and of course the chinese ali-baba-gipsy-seller sent only 2 pcs instead of 3). At least they look brand new. I measured their Vf with my trusty multimeter in diode mode and it is only 0.435 V for each diode in both pieces. I don't know if this Vf will rise to about 1.7 V under load at high temperatures. This weekend I'll desolder the original SF1603G diodes from the PSU's PCB, so I'll be able to compare their Vf values and change those with these SF1608G.