Converting an ATX PSU into an adjustable split supply

[Th3_uN1Qu3]

I've long needed a proper adjustable supply but was always too lazy to build a custom job from the ground up. And it eventually occured to me that one of the many ATX PSUs lying around would be a perfect candidate for this.

I grabbed an "EC" 200 watt supply. No one needs a 200W supply no more, moreover, its secondary caps were all good, no troubleshooting required. Besides the non-existing input filtering, it actually looks decent. It matches this schematic perfectly with the only mention that the switching transistors are MJE13007s.

For now i put in two MBR20100 diode packs. But they will be changed as i will need a negative voltage. It is planned to be adjustable to +/-30v and current limited as well.

Current progress: an adjustable supply in its crudest form. I turn a trimpot and the voltage changes. I still need to tackle error amp compensation (the transformer hisses at certain voltage values, and load regulation leaves to be desired), i need to rewind the transformer and inductor, i need 35 volt caps...

The existing protection circuit has been bypassed entirely and i will build a new one from the ground up. Since i only have two rails to monitor for current limiting, it's likely that i will be using the same LM393 comparator that is already in there.

[Th3_uN1Qu3]

Re: Converting an ATX PSU into an adjustable split supply

Finished rewinding the transformer... honestly i doubted it'll fit, but all went fine eventually. The existing primary was wound with 21AWG, a total of 40 turns. It was split with the secondaries wound inside it, but not equally - the outside layer was 12 turns, the inside 28. This splitting of the primary is done so that radiated interference is minimized. Clock frequency of the TL494 is 78kHz, but since this is a halfbridge supply the actual switching frequency of the transistors is half that. 34kHz also fits the bill for maximum operating frequency of 21AWG wire before skin effect kicks in.

With 7.75A/mm^2 this solved to 3.18A or a maximum power of 280W at the minimum input voltage, so i kept it the way it was. The new secondaries are 22 turns each, center tapped (that's how it's required for half bridge). Rated output is to be 4.5A on each winding at full voltage (60*4.5=270W) but i wanted a little bit more output at lower voltages so the secondary windings were calculated for 5A DC, which is 3.45A RMS in the transformer. I used 6 strands of 0.3mm wire for the secondary.

In the meantime i also beefed up the input caps and added all missing filtering components. I plugged it in and... it didn't blow. So far both voltages are still positive, i'll deal with the negative rail later. Voltage regulation totally, absolutely, SUCKS, the control loop oscillates like crazy, but hey, i have a working power supply.

As i advanced the pot, various things that weren't suited for that voltage started smoking and catching fire. Yeah, i forgot to disconnect the -12v and -5v rails... Ah, the glorious smell of burning electronics. They eventually settled, and they actually made my life easier - if it's gone black, i know i have to remove it. Now since i have the transformer in its final shape, i can go on and calculate the compensation components for the error amplifier.

If anyone is interested in the math i'll post it.

[momaka]

Re: Converting an ATX PSU into an adjustable split supply

Nice project again.
I always wanted to do something similar, but I don't think I have enough knowledge yet (and right now I'm trying to clean up my "fix" pile first ).

I was thinking, though - instead of adjusting all of the outputs, why not leave them alone and add another output winding that does, say +/-30v? Then have separate rectifiers and MOSFETs to regulate to whatever voltage you need. Or is that not a possible option?
Otherwise, this would likely give you cleaner voltages and better regulation.

If anyone is interested in the math i'll post it

If you want to. I don't think I'll be looking it over anytime soon, though. Just had some exams today and I'm not sure if more math would do me any good .

[Th3_uN1Qu3]

Re: Converting an ATX PSU into an adjustable split supply

I can't leave the original windings alone simply because there is no room on the transformer. Anyway... I quickly learned that 2x 30v + crossload + group regulation = BOOM. Since i'll be using only the positive output most frequently, looks like simply modding a PC power supply is not an option.

Looks like i'll be back to my original plan - building two separate 2-transistor forward converters (i don't like half bridge anyway), one for the positive rail, one for the negative rail. Which means two transformers, four MOSFETs, two controllers. The only thing that i will keep (more correctly, steal) from the power supply will be the little flyback converter that does the standby supply, which FYI also feeds the controller with 20-some volts.