Sorry, I can't edit my published posts. I meant those SF1608GS diodes which I bought and intended to use as a replacements for the original SF1603GS on the +12 V line, not on the +5 V line.

Why did I come to this thread, I wanted to ask.

Some PNFs with PASSIVE PFC have a weird circuit with diodes and electrolytic capacitors. To demagnetize the choke or something.
Newer ones just have a film capacitor in parallel with the PFC choke.
If anyone has one, what is its capacitance and voltage rating?

I'm still thinking for adding of some dummy load on +12 V line when PSU is used with older systems - from 5 or 10 W ceramic resistors to small 12V auto bulbs, but all these solutions will be outside the PSU case even if I use 4-pin +12 V CPU connector.

shevalier wrote on 2025-06-02, 13:30:

Without load (short green&black wire), the output voltages will be determined by the ratio of load resistors R42/R41, which you […]
Show full quote

Without load (short green&black wire), the output voltages will be determined by the ratio of load resistors R42/R41, which you have circled absolutely correctly.
They are located next to the AC power input (and OCP board if it`s present) and have a darkened pcb underneath. 🙁
(Bad advice)
Put a 10W ceramic resistor in position R41 at 51 Ohm, and the voltages will be fine without load.

This idea seems fine to me as the PSU's fan is big enough and has temperature regulator, but how did yo came with the value of 51 Ohms for R41 resistor? Could you explain the relation between R41 on +12 V line and R42 on +5 V line and how to calculate their values to adjust the load on these power lines?

shevalier wrote on 2025-06-02, 13:30:

This idea seems fine to me as the PSU's fan is big enough and has temperature regulator, but how did yo came with the value of 51 Ohms for R41 resistor?

This is a stupid idea.
It just shows how voltages depend without a load.
These resistors are needed only to ensure that the voltage without load does not trigger the overvoltage protection. And the power supply has reached at least some stable operating mode.

Could you explain the relation between R41 on +12 V line and R42 on +5 V line and how to calculate their values to adjust the load on these power lines?

No way.
The power distribution along the power lines is included in the calculation of the power supply.
Accordingly, the required additional load for a given power supply is not a calculated value, but a practical one.

Enlight hpc-250 - example of a power supply for working with ATX 1.0
External signs - no 4 pins 12 for the CPU, 20 pins and not 20+4 for the motherboard
Inside just 1 capacitor 2200uF for +12V
Unpaired diodes for +12V
The group choke is wound with a wire so that there are two of them for +5, and one for 12V.
Therefore, the energy in it is redistributed so that more power goes to +5V.
P.S. And an electrolytic capacitor 1uF in the control of the standby voltage +5V, which after drying leads to complete burnout of the circuit, including the transformer. 🙁
What is needed for systems based on Pentium 3

shevalier wrote on 2025-06-03, 05:33:

This is a stupid idea.
It just shows how voltages depend without a load.
These resistors are needed only to ensure that the voltage without load does not trigger the overvoltage protection. And the power supply has reached at least some stable operating mode.

It's not as stupid as my idea to change the schottky diodes on +12 V line with the ones with more V-drop 😉 Simple doesn't mean stupid and I like to keep things as simple as possible. Both of these ideas will work at the expense of some more released heat inside the case, but I'm sure the PSU's fan will cope with it.

shevalier wrote on 2025-06-03, 05:33:

The power distribution along the power lines is included in the calculation of the power supply.
Accordingly, the required additional load for a given power supply is not a calculated value, but a practical one.

Well, then I can trust your practical experience as a starting point.

I can't find a single 51 Ohm 10 W resistor, but I can get one 47 Ohm 10 W, one 56 Ohm 10 W or two 27 Ohm 5 W (two in series for 54 Ohm 10 W) resistors from my nearest electronic components shop. Which one will be the most suitable as value?

I think that two 5 W 27 Ohm resistors in series will be better cooled than single 10 W 56 or 47 Ohm resistor:

shevalier wrote on 2025-06-03, 05:33:

Enlight hpc-250 - example of a power supply for working with ATX 1.0 External signs - no 4 pins 12 for the CPU, 20 pins and not […]
Show full quote

Enlight hpc-250 - example of a power supply for working with ATX 1.0
External signs - no 4 pins 12 for the CPU, 20 pins and not 20+4 for the motherboard
Inside just 1 capacitor 2200uF for +12V
Unpaired diodes for +12V
The group choke is wound with a wire so that there are two of them for +5, and one for 12V.
Therefore, the energy in it is redistributed so that more power goes to +5V.
P.S. And an electrolytic capacitor 1uF in the control of the standby voltage +5V, which after drying leads to complete burnout of the circuit, including the transformer. 🙁
What is needed for systems based on Pentium 3

I don't know if you're giving this is as a suggestion or as an example, but I don't want to buy an old ATX 1.0 PSU's in unknown condition and to deal with more potential problems. I'll try to modify mine (it's freshly recapped) with additional PCB on its back side of the PSU case (the grilled one) and option to switch between the original R41 and the new ones (about 51 Ohm - as suggested).

analog_programmer wrote on 2025-06-03, 06:43:

It's not as stupid as my idea to change the schottky diodes on +12 V line with the ones with more V-drop 😉

Its not Schottky, its ultrafast diod.
Although the snubber on the PFN is set up well and there are no voltage surges on the diodes.
100 V Schottky diodes work reliably at +12V rail.
But 2 assemblies of ultrafast diodes in parallel give a greater effect than one Schottky diode.
This way you not only change the efficiency, but also redistribute the energy.
If you remove the second ultra-fast diode on the +12V rail and replace on the +5V Schottky rail with a modern parallel one, you will get a redistribution of about 1-3%.
Somewhere from 12.8 to 12.5V.
Fine-tuning the existing design.

I can't find a single 51 Ohm 10 W resistor, but I can get one 47 Ohm 10 W,

Resistance is 51 Ohm - so the readings will be good without any load.
For normal PC operation, the load on +12V should be 1.5~2 or more amperes.
And this is 6 Ohm 25 Watts.
https://imrad.com.ua/userdata/modules/product … HS-10-05.45.pdf
On a radiator with a fan.

Or an assembly of 5-6 pieces of 10 watts each.
https://imrad.com.ua/userdata/modules/product … ssives-CR-W.pdf
Wire wound resistors are a must.
If you use a 10W resistor rating at 10W power consumption, it will be about 100 degrees Celsius (boiling water).
Thus, you represent the heat death of the universe, senselessly converting electricity into heat. 😀

Honestly once it reached the point of implementing workarounds for primarily 12v group regulated power supply on primarily 5v system my question would be - is it really a good idea?

Abnormally high and low voltages like this are not a good thing and trying to work that around by manually adding/removing load is not really safe. This old hardware can be relatively valuable at this point - blowing it up would not be nice and just one incident will more than cover the cost of proper power supply...

So why not get a proper PSU? Either a good vintage one intended for 5v systems, or modern DC-DC one in which every voltage is properly regulated and which should be able to safely handle most systems, apart from may be the most power hungry socketA ones...

Archer57 wrote on 2025-06-03, 08:58:

...
So why not get a proper PSU? Either a good vintage one intended for 5v systems, or modern DC-DC one in which every voltage is properly regulated and which should be able to safely handle most systems, apart from may be the most power hungry socketA ones...

Because:

analog_programmer wrote on 2025-06-03, 06:43:

I don't want to buy an old ATX 1.0 PSU's in unknown condition and to deal with more potential problems. I'll try to modify mine (it's freshly recapped)...

Archer57 wrote on 2025-06-03, 08:58:

that around by manually adding/removing load is not really safe. T

From a technical point of view, the method is absolutely acceptable.
But not excellent.
Although if these are 3 LED strips of 12V and 3-4 fans of 120mm, then there will be the same additional load.
Even almost useful.

shevalier wrote on 2025-06-03, 08:05:

Its not Schottky, its ultrafast diod. Although the snubber on the PFN is set up well and there are no voltage surges on the diod […]
Show full quote

Its not Schottky, its ultrafast diod.
Although the snubber on the PFN is set up well and there are no voltage surges on the diodes.
100 V Schottky diodes work reliably at +12V rail.
But 2 assemblies of ultrafast diodes in parallel give a greater effect than one Schottky diode.
This way you not only change the efficiency, but also redistribute the energy.
If you remove the second ultra-fast diode on the +12V rail and replace on the +5V Schottky rail with a modern parallel one, you will get a redistribution of about 1-3%.
Somewhere from 12.8 to 12.5V.
Fine-tuning the existing design.

Yeah, my mistake and that's what often happens when I'm doing a couple of similar things simultaneously. Those SF1603G diodes on the +12 V line (BD4 and BD5) are not schottky and now I have two SF1608G from my previous idea to change them with suitable diodes with higher V-drop (I got SF1608G).

Off: And now is a perfect time to express my thanks to the moderating script, that took away my "edit button" for "good" 😁

shevalier wrote on 2025-06-03, 08:05:

Resistance is 51 Ohm - so the readings will be good without any load.

Hmmm. But for the "good readings" I have to put back the load into the shape from the diagram you've linked - you know what I mean. I think, someone else also mentioned before, that R41 and R42 are some kind of "dummy load" resistors when PSU is starting. And they are exactly this - a "dummy load" resistors as one end of each of these two resistors is connected to ground, the other - to corresponding output voltage line. So if I could force the PSU to produce "good voltages" wit just this "small internal dummy load", it also has to produce "good voltages" with more external load on +12 V line (HDD, CD-ROM, FDD - there are no other serious consumers on +12 V line for the old systems, right?).

shevalier wrote on 2025-06-03, 08:05:

For normal PC operation, the load on +12V should be 1.5~2 or more amperes. And this is 6 Ohm 25 Watts. https://imrad.com.ua/user […]
Show full quote

For normal PC operation, the load on +12V should be 1.5~2 or more amperes.
And this is 6 Ohm 25 Watts.
https://imrad.com.ua/userdata/modules/product … HS-10-05.45.pdf
On a radiator with a fan.

Or an assembly of 5-6 pieces of 10 watts each.
https://imrad.com.ua/userdata/modules/product … ssives-CR-W.pdf
Wire wound resistors are a must.
If you use a 10W resistor rating at 10W power consumption, it will be about 100 degrees Celsius (boiling water).
Thus, you represent the heat death of the universe, senselessly converting electricity into heat. 😀

Six pieces of 1 Ohm 5 W resistors (the non-ceramic ones are small enough) in series with no radiators on +12 V line sounds good for the revisited additional internal PCB idea. 30 W seems to much for dummy load, probably 20 W will do the job when PSU is used with old ATX/AT systems. I can try a combination of two 1 Ohm 5 W + two 2 Ohm 5 W resistors in series or... I'd better off to try the R41 mod with the additioanal resistors on a PCB.

And of course I can always go back to my not so clever idea with changing of BD4 and BD5 (SF1603G) diodes to SF1608G diodes, but this is not a flexible solution.

Pictures are good 😀

Let's assume that the green "filled" area covers up to pentium-class systems consumption. And I have at least two rock-stable ATX PSUs for this, so it's not of big interest to me to get the PSU just there. But then the newer systems without any +12 V load (like P2/P3, early Athlons and similar in which I'm interested to use this PSU) are into the green "striked" area and there is no way to get there with just adding more resistive load to +12 V line producing more and more excessive heat.

Now in theory my only option is to try the BD4 and BD5 mod and how it will change the PSU output voltages under load with P2 or P3 system.

shevalier wrote on 2025-06-03, 10:27:

From a technical point of view, the method is absolutely acceptable. But not excellent. Although if these are 3 LED strips of 12 […]
Show full quote

From a technical point of view, the method is absolutely acceptable.
But not excellent.
Although if these are 3 LED strips of 12V and 3-4 fans of 120mm, then there will be the same additional load.
Even almost useful.

The issue is - it only works if you specifically know the load you will be using and apply appropriate extra load as needed. Otherwise even with that extra load going outside of normal operation range is still possible. This is a huge hassle given the use case (testing different systems) with large probability of user error and then the only thing that can prevent frying stuff is hope that those "vendor specific behavior" includes some over/under voltage protection.

To me doing something like this would make sense only if there was absolutely no other choice in some critical situation. Otherwise it is a nasty hack which will cause something to blow up eventually.

That's just my opinion though, the only reason i've posted here is that whole last page of this discussion feels like someone asking about doing something which is bound to end badly and people joking about solutions. So i thought may be it is a good idea for OP to stop and think about what are the benefits and the risks/downsides of doing it, but may be i am completely wrong - feel free to ignore me...

In my previous post I mean AT-PSUs for up to pentium-class systems, not ATX-PSUs.

Once again: Many thanks to the moderating script, that took away my "edit button" for "good"! 😁

Archer57 wrote on 2025-06-03, 11:07:

The issue is - it only works if you specifically know the load you will be using and apply appropriate extra load as needed. Otherwise even with that extra load going outside of normal operation range is still possible. This is a huge hassle given the use case (testing different systems) with large probability of user error and then the only thing that can prevent frying stuff is hope that those "vendor specific behavior" includes some over/under voltage protection.

To me doing something like this would make sense only if there was absolutely no other choice in some critical situation. Otherwise it is a nasty hack which will cause something to blow up eventually.

And that's why I don't like this PSU and I rate it as Fortron/FSP branded junk, even if it's not the absolute bottom junk like all the Codegen or similar cr*ps.

Archer57 wrote on 2025-06-03, 11:07:

That's just my opinion though, the only reason i've posted here is that whole last page of this discussion feels like someone asking about doing something which is bound to end badly and people joking about solutions. So i thought may be it is a good idea for OP to stop and think about what are the benefits and the risks/downsides of doing it, but may be i am completely wrong - feel free to ignore me...

The benefits are some PSU modifications practice and eventually the damn thing to become usable in P3 and older systems.

Risks/downsides...? I'll test the freaking PSU if it works fine with newer Athlon 64 x2 system and if it's still usable there, it will be just the PSU limited cr*p tor systems of this type.

Thanks to the shevalier's comments now I'm convinced that this Fortron/FSP ATX-350PNF PSU just can not be used with an ATX-to-AT power adapter even with some kind of modifications.

It turns out that in general it's not a good idea to use any kind of ATX ver. 2.xx or newer PSU as a power source for older AT or ATX ver. 1.xx systems. A bunch of "which modern PSU model I can use for my retro PC..." questions now can go to trash bin for good.

I'm gonna try to revive a partially scraped cr*ppy ATX ver. 1.xx PSU (it has -5 V line, "300 W" on label) for P2/P3/Arhlon(XP) motherboards/systems tests, but not in this thread.

I did have luck with one ATX v2.03 using an AT adapter but that might just have been the board/psu combo at the time

Horun wrote on 2025-06-04, 23:36:

I did have luck with one ATX v2.03 using an AT adapter but that might just have been the board/psu combo at the time

According to this diagram the leftmost parts of the curves with combined loads that guarantee normal operation of ATX ver. 2.03 and ver. 2.2 power supplies are very similar, so your case may indeed have been in the "manufacturer specific behavior area" and a matter of luck. Not all manufacturers strictly follow the ATX versions standards/specs, so that is a factor in the "lottery" as well.

After some more testing it indeed looks like the 5V and 12V rails go more into spec values when there's "enough" load on the 12V rail...
On my device it changes from 4.8V/12.8V (sometimes worse like 4.65V/13V) to something in the region 5.05V/12.2V, which is "ok".
I think I had it in one or two systems before and the voltages were always a bit on the edge. Maybe these were still too underpowered to get it fully into regulation.

But even with these normal values it still is one of the worst PSUs I have. Every other one shows perfectly regulated voltages even at complete no external load conditions, is more silent and/or runs with better efficiency. Now that I know that this is like that not by a fault but per design the case is closed for me.
I know where / how to use it and got a fresh thinking into PSU design again.
At least I'm quite confident that it *could* deliver the 350W it has written on the case looking at the PCBs backside and seeing all that thick traces full of additional solder and bridge wires.
Compared to other examples (looking at you, 400W Excellent Power) this leaves more confidence.

shevalier wrote on 2025-06-03, 04:44:

Why did I come to this thread, I wanted to ask. […]
Show full quote

Why did I come to this thread, I wanted to ask.

Some PNFs with PASSIVE PFC have a weird circuit with diodes and electrolytic capacitors. To demagnetize the choke or something.
Newer ones just have a film capacitor in parallel with the PFC choke.
If anyone has one, what is its capacitance and voltage rating?

Only electrolytic on my side.

Edit: Forgot to add picture...

ChrisK wrote on 2025-06-05, 05:38:

After some more testing it indeed looks like the 5V and 12V rails go more into spec values when there's "enough" load on the 12V […]
Show full quote

After some more testing it indeed looks like the 5V and 12V rails go more into spec values when there's "enough" load on the 12V rail...
On my device it changes from 4.8V/12.8V (sometimes worse like 4.65V/13V) to something in the region 5.05V/12.2V, which is "ok".
I think I had it in one or two systems before and the voltages were always a bit on the edge. Maybe these were still too underpowered to get it fully into regulation.

But even with these normal values it still is one of the worst PSUs I have. Every other one shows perfectly regulated voltages even at complete no external load conditions, is more silent and/or runs with better efficiency. Now that I know that this is like that not by a fault but per design the case is closed for me.
I know where / how to use it and got a fresh thinking into PSU design again.
At least I'm quite confident that it *could* deliver the 350W it has written on the case looking at the PCBs backside and seeing all that thick traces full of additional solder and bridge wires.
Compared to other examples (looking at you, 400W Excellent Power) this leaves more confidence.

So, now we can close the topic about this model improvements for good. Everything's fine, this Fortron/FSP ATX-350PNF PSU is just very limited in its use case.

shevalier wrote on 2025-06-03, 04:44:

Why did I come to this thread, I wanted to ask. […]
Show full quote

Why did I come to this thread, I wanted to ask.

Some PNFs with PASSIVE PFC have a weird circuit with diodes and electrolytic capacitors. To demagnetize the choke or something.
Newer ones just have a film capacitor in parallel with the PFC choke.
If anyone has one, what is its capacitance and voltage rating?

Only electrolytic on my side.
[/quote]

I confirm that there are no film capacitors near the PFC choke connections on the PCB in my ATX-350PNF- only a couple of small electrolytic capacitors.

I bought a Sparkle Power ATX-250GT yesterday.

Replaced 3 leaky caps (2x 1000uf 10V which I replaced with 1500uf because I didn't have 1000uf available and 1x 3300uf 10V).

I've tested with no load and I get:

-12V: 11.5
+12V: 11.5
+5V: 5.3
+3.3V: 3.4
5V SB: 5.2

With a Hard disk attached, I get:

-12V: 11.8
+12V: 11.8
+5V: 5.1
+3.3V: 3.4
5V SB: 5.2

When I add a second hard disk, the -12 and +12 rise to 11.9V.

I haven't tested with a motherboard yet but I presume if I actually use it for a PC, the values will align more closely with the expected values.