Things that can go wrong when modding an ATX PSU
This isn't "The ATX PSU Mod Guide", that is due soon when i'll finish my own build. It's just 'round the corner though, the hard part is done, i just have some little quirks to iron out.
First of all, grab this schematic. Better yet, print it. You're going to need it. I have a couple more i'll post later.
All Deer/L&C/Allied/whatever PSUs with TL494 and LM339 or LM393 use the same schematic mostly, but depending on what the engineer was smoking on at that time, there are some bewildering quirks to each model. From ultra crappy layout to absolutely pointless schemes. Alas, the one i picked for modification seems to have both.
First of all, always use a lightbulb in series with the PSU when testing. This will prevent most spectacular failures. Preferably make a proper limiting device: mains plug on one end, double-pole switch, lightbulb in series after the switch, mains outlet on the other side. You can use a single pole switch if you make double sure you're switching the LIVE wire, but better safe than sorry.
Sure, an ATX PSU can be modded into something like a big 13.8v source without major circuit modifications. All you have to do is rewind the transformer and coil, and you're done. But when you're trying to do +/-60v like i do, you're going to run into some interesting circuit behaviors. I'll note some of them.
For some reason, this particular power supply of mine doesn't have the 494 feed off a higher voltage winding of the 5vsb supply. 5vsb does kick it off, but then it derives its own supply from the base drive transformer. Short version: Whatever you're going to have on the highest voltage output of your power supply, you're gonna have on the 494. I discovered this only when the (open-loop) output went to 70 volts and the 494 failed.
After replacement, i also closed the feedback loop. At 60 volts the 494 seems to be holding up - it has a 40v absolute maximum rating, but it appears to work fine at 60. Since this is the maximum voltage i intend my supply to produce, i won't have to worry, right? Wrong... With the lightbulb in series, even a small load was only supported by the power supply for a few seconds, then the transformer would start making a weird noise, and if i left the load connected for longer, the bulb would light up and all voltages would drop.
I assumed this was due to the bulb, so i wired up the supply straight to the mains. It held the small load fine, so i upped the load. Next thing i know - same noise, and before i could disconnect the load, the output went to zero volts. First thought the transformer saturated and blew the power switches - but, it has the same number of primary turns as the original one. I couldn't have been farther from the truth. The power switches were okay.
The fault turned to be the C945 driver transistors. "Collector-Emitter breakdown voltage: 50v". Aha. It now becomes apparent why the high voltage driver supply - the main power transistors require several amps of base drive current, while the C945s are really, really, really lame. 150mA is all they will do. So, to make it work, they use a base drive transformer with high turns ratio, and a high driving voltage to obtain the required current output.
I'm going to fix this lazy man's way and put a dropper resistor and 33 volt zener to feed the 494 and driver transistors, but i'm still going to be changing out the drivers for 2N2222s. The lame driving scheme is also one of the reasons for the low efficiency of those junkers. And it still makes no sense why they didn't make a nice and stable supply from the 5vsb transformer to feed them.
Anyway, hope this helps those who are on the same quest as mine.
First of all, grab this schematic. Better yet, print it. You're going to need it. I have a couple more i'll post later.
All Deer/L&C/Allied/whatever PSUs with TL494 and LM339 or LM393 use the same schematic mostly, but depending on what the engineer was smoking on at that time, there are some bewildering quirks to each model. From ultra crappy layout to absolutely pointless schemes. Alas, the one i picked for modification seems to have both.
First of all, always use a lightbulb in series with the PSU when testing. This will prevent most spectacular failures. Preferably make a proper limiting device: mains plug on one end, double-pole switch, lightbulb in series after the switch, mains outlet on the other side. You can use a single pole switch if you make double sure you're switching the LIVE wire, but better safe than sorry.
Sure, an ATX PSU can be modded into something like a big 13.8v source without major circuit modifications. All you have to do is rewind the transformer and coil, and you're done. But when you're trying to do +/-60v like i do, you're going to run into some interesting circuit behaviors. I'll note some of them.
- Various resistors burning - mostly non-critical, you can either replace them with higher values, or let them burn till they fail open.
- Small capacitors blowing up - you changed the output caps, all others should be running off the standby supply, no? Wrong.
- TL494 blowing up - you assumed the TL494 was fed from the standby supply. Wrong again.
- And my latest - driver transistors blowing up (the two C945 next to transformer T2 in the above schematic). Those aren't fed from the standby supply either. And some manufacturers like to put them in the most obscure places.
For some reason, this particular power supply of mine doesn't have the 494 feed off a higher voltage winding of the 5vsb supply. 5vsb does kick it off, but then it derives its own supply from the base drive transformer. Short version: Whatever you're going to have on the highest voltage output of your power supply, you're gonna have on the 494. I discovered this only when the (open-loop) output went to 70 volts and the 494 failed.
After replacement, i also closed the feedback loop. At 60 volts the 494 seems to be holding up - it has a 40v absolute maximum rating, but it appears to work fine at 60. Since this is the maximum voltage i intend my supply to produce, i won't have to worry, right? Wrong... With the lightbulb in series, even a small load was only supported by the power supply for a few seconds, then the transformer would start making a weird noise, and if i left the load connected for longer, the bulb would light up and all voltages would drop.
I assumed this was due to the bulb, so i wired up the supply straight to the mains. It held the small load fine, so i upped the load. Next thing i know - same noise, and before i could disconnect the load, the output went to zero volts. First thought the transformer saturated and blew the power switches - but, it has the same number of primary turns as the original one. I couldn't have been farther from the truth. The power switches were okay.
The fault turned to be the C945 driver transistors. "Collector-Emitter breakdown voltage: 50v". Aha. It now becomes apparent why the high voltage driver supply - the main power transistors require several amps of base drive current, while the C945s are really, really, really lame. 150mA is all they will do. So, to make it work, they use a base drive transformer with high turns ratio, and a high driving voltage to obtain the required current output.
I'm going to fix this lazy man's way and put a dropper resistor and 33 volt zener to feed the 494 and driver transistors, but i'm still going to be changing out the drivers for 2N2222s. The lame driving scheme is also one of the reasons for the low efficiency of those junkers. And it still makes no sense why they didn't make a nice and stable supply from the 5vsb transformer to feed them.
Anyway, hope this helps those who are on the same quest as mine.
That's quite a feat considering this platform usually blows up at ~250W, and i don't even have the proper output caps installed yet.
The driver transformer will have to be rewound as well, but at least you know what you'll be putting in there. You will also need to hack the standby supply to give at least 15v, but that's as easy as changing a resistor. And the driver transistors will DEFINITELY need to be 2N2222s or even better if you can (higher current).
The funny thing is, i don't have any other 494-based supply to hack, the Hipros are way too crowded to bother... I shoulda just gone with cooking my own from the get go. But don't worry, this ain't over yet. 
Overkill, but I love it when a plan comes together.
Comment