The "proper" AT-to-split-supply mod!

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

@ kaboom: Thanks a lot for the explanation.
@ momaka: Yeah, shipping was $8. Just the bare boards, but i had enough housings anyway. Some fans would've been nice tho.

I'm leaving this away for a while because i'm experiencing some issues that point to transformer saturation, even though this shouldn't be the case, and another transformer from a half bridge supply (not the original one as unfortunately i broke that when i tried to take it apart), behaves the same. There's something funky going on but i don't exactly know what.

I've decided to first make a PCB for the supply i am designing and building from scratch, and then come back to this one.

[momaka]

Re: The "proper" AT-to-split-supply mod!

Any PSU using tl494 is a good candidate for mods. The gutless ones are better coz there's less stuff to remove.

In that case, I'm good then . Got quite a few old Deers.

12 ATX PSUs for $15? Not bad. I'm assuming most of that $15 is for the shipping .

[kaboom]

Re: The "proper" AT-to-split-supply mod!

The '494 gets its VCC after the 12V rectifier but before the output inductor. Hence no pulse length integration-to-variable-voltage. That's why you had 28V on the '494 no matter what the 'output' was.

When power is first applied, the primary switches are slightly biased into conduction. Almost immediately, one saturates. Current flows through the power transformer, as well as a one turn winding in the driver tx. The driver tx couples this winding to two windings to the bases of the switchers, one base winding 180 degrees out from the other. Shortly after, the driver tx nears saturation and this is coupled to the other transistor. The one that was saturated drops out and the other picks up. After a few cycles of self-oscillation (by way of driver tx), the cap(s) feeding the '494 charge up. Since that point is before the output inductor, this happens even before the 'big' output caps are fully charged. Now the '494 has power and drives the bases of the 2SC945s feeding into the driver tx. At that point, the supply has bootstrapped and the bases of the switches are fed from their own output via the driver tx. Add a 4148, some small caps and resistors, and the driver tx can give a turns ratio derived reflected voltage that represents the primary current. After all, the primary current flows in the one turn 'kicker' winding in the driver tx.

So that 'driver' tx has the possibility of doing triple duty- nice for parts saving and leads to 'interesting' circuits.

Have fun.
-Paul

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Update. The driver transformer only starts the thing... the TL494 power is actually derived from the main transformer winding, and i learned it the hard way. I'll spare you the long story but the TL494 wouldn't start and its power supply were at 1.38v or so. I thought there was something wrong with that, but nope... i had a shorted 494. Now having a good idea as to why it died, i replaced it and added a 30v zener (actually two 15v in series) between TL494 Vcc and ground. Success!

The power supply now starts up just fine. I have to go to the post office to get a package (12 ATX PSU boards with everything on them except the wires, $15 for the whole lot ), when i'll be back i'm wrapping the whole thing up and taking pics.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Once i overcome my laziness i will... But i already said that the Vcc pin of the 494 holds steady at 28v while i vary the 12v rail from 7v to 28v. It only goes down (to ~5v) when the protection trips hence my suspicion it's somehow pulled from the driver transformer. If there is a buck-boost converter, i don't see it. And i highly doubt there is.

Edit: Confirmed. It has a little self-oscillating affair feeding the 494, which like i thought is coming from the driver transformer.

[Pyr0Beast]

Re: The "proper" AT-to-split-supply mod!

12V rail probably. Follow the supply trace.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

I know that. But i don't know where exactly the 494 is powered from.

[Pyr0Beast]

Re: The "proper" AT-to-split-supply mod!

Self-oscillation after which TL494 takes control.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

If I knew this PSU was this capable (in terms of mods), I would have definitely fixed it.

Awesome work, keep it up.

Any PSU using tl494 is a good candidate for mods. The gutless ones are better coz there's less stuff to remove.

What isn't clear to me yet is how the 494 is powered. Vcc is a steady 28 volts regardless of the output voltage of the former 12v rail. I didn't find any extra windings on the main transformer so i suspect it's powered from the base drive trafo somehow.

We'll see what happens when i install the new main transformer. If the supply doesn't start at all it means i was wrong.

[momaka]

Re: The "proper" AT-to-split-supply mod!

This is the ATX version (also search for Meico):
https://www.badcaps.net/forum/showth...ighlight=m-tec

I have the ATX version as well. JNC. White label with green letters. It's in pieces unfortunately - I took out every single component off of the PCB as a learning experience in desoldering because it was broken (so it looks more gutless than yours ). The 5vsb circuit destroyed itself. It was one of those self-oscillating wonders with no feedback. You can see where IC701 is not populated on yours - that's where a 7805 regulator would go for 5vsb. And Q501 is for the 3.3v MOSFET.
If I knew this PSU was this capable (in terms of mods), I would have definitely fixed it.

Awesome work, keep it up.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Believe it or not it ran fine for years. A Pentium MMX wasn't that demanding...

• Current progress: Installed positive/negative rectifiers.
• I'm going to make the 5v rail into the negative rail.
• Found a bigger heatsink for the diodes but i'll have to drill a hole in it, as the only heatsink i had that had the mounting pins in the right place, doesn't line up in terms of holes.
• Rewound transformer, used an ER-35W core this time. Should be good for 500W if the rest of the supply can take it.

Now, time to patch it all together and fire it up. Stay tuned.

[TheLaw]

Re: The "proper" AT-to-split-supply mod!

Gutless AT Picture

HOLY SHIT! That's got to be one of the worse ever.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Update. I have replaced the primary caps (2x 330uF CapXon) with 2x 470uF CS. Not the best but they'll do for this one. I have also replaced the lame PR1007 primary diodes (1A!!!) with 5A ultrafast ones while i was at it, and also installed a 4 amp bridge rectifier instead of the 2A diodes. These parts were quite a tight fit, this board really wasn't intended for quality. I've also done away with the big bulky external power switch and installed a regular single pole one. For now, it can only dream of an EMI filtering stage. I know i had one of those filtering boards that attach to the AC receptacle, but i can't find it.

I have also successfully tested the positive/negative rectifiers. But i realized one thing while trying to make the output filter - iron cores start losing ground fast at high output voltages.

So what's the next logical step? Of course, ferrite. Since i have tons of EI33 cores anyway (and a dozen more PSU boards coming in the mail), i took one of the untampered bobbins and hooked the primary (40 turns) then a half primary (20 turns) for an inductor. Results? Excellent. With 40 turns the core worked even with no air gap, and got just moderately warm after a 10-minute run at 100W. The core was out of the reach of the fan btw. And i'm sure that is 100% copper loss since the primary wire is 22AWG or maybe even thinner. Keep in mind the inductance is a few millihenry, more than 10 times higher than can be economically obtained from an iron powder core! But indeed this is too much inductance for my application and will cause overshoot.

At 20 turns the core saturated without a gap and made this horrible screeching noise, then the power supply shut down. A gap made from a sheet of copy paper bent over 4 times corrected the issue, and the inductor runs even cooler since there is now just half the copper in the way. Problem solved. Now all i need is a new transformer and job done.

I'll make some measurements of the actual inductance tomorrow.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

The negative rail diodes are in plain view here. Btw, i just lifted the 5v rectifier and (apart for the fact that i now have to cobble up something for the fan), the power supply still works.

Here you have a pic. And before you ask, yes the 5v rectifier is still in place, i was too lazy to remove the transformer or heatsink to take it out. I just unsoldered it. And the -5v diodes too. Believe it or not, this PSU didn't have a lot of parts even before i started on it.

[Pyr0Beast]

Re: The "proper" AT-to-split-supply mod!

Correct.

Only problem is that sometimes those diodes hide under the main coil, but you need to rewind that one as well, so no prob.

-5V and -12V usually aren't regulated, tho they sometimes are monitored for under/overvoltage.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Well, removing the negative rails proved too much of a hassle (at least for -12v), but while tinkering with them i figured something out. I don't think they're even in the feedback loop but i think they power some of the driver circuitry.

"Wait... aren't those derived FROM THE SAME TRANSFORMER WINDING???" Why yes they are, it's just that the diodes are the other way round. So that means... yup. Just beef up the negative rectifiers, rewind transformer and coil, and tada, i have a split supply with perfect regulation.

[Th3_uN1Qu3]

Re: The "proper" AT-to-split-supply mod!

Did the 'JNC' label have red stylized letters on a white background?

Nope, it's blue stylized letters on white background.

The fact that you got such high voltages out of it by manually poking the feedback is interesting. This would indicate that unusually low duty cycles were used to 'regulate it down' to the 'puter voltages. I'll guarantee that thing was discontinuous, even till a few tens of watts out! Makes for nice high peak ripple current through the coupled inductor and caps. High output ripple voltage, too- what with no PI filters.

With a ~100W load i can still achieve 22.5v maximum. That's quite a bit, and seems to reinforce your theory. Strange thing however is that the highest measured efficiency is 60%, and with 100W dissipation the thing should have been burning in minutes. That doesn't happen. I think that due to the small capacitors i currently have, the ripple is high enough to fool one of the two cheap meters on the output (as i only have one TrueRMS meter and it's obviously used for monitoring current drawn from the mains).

Btw, over power protection only kicks in at 2.2A input. That's almost 500W! This is a 200W unit, right? Either the lousy efficiency numbers are real (but i can't figure where all the heat is going in that case), or it was simply designed to work only as short circuit protection. I don't know. The positive side of things is that i won't have to mod it.

They left the cap over the RC combo out 'cause they didn't want to AC comp it. Perhaps because of the stupidly low duty cycle?? The output inductor and capacitance of the rectifiers ring every time the 'spikes' from discontinuous operation excite it. The AC comp cap would feed these parasites into the error amp and upset it, so they left it out. They're counting on the extra phase margin...

That's interesting. I put in a 680pF cap with no issues. I also changed the original RC pair from 10nF/24k to 4.7nF/24k. Anyway, the thing starts oscillating when it's turned lower than 10 volts, because the duty cycle required is so low that the 494 starts skipping cycles!

I traced the -12v rail and it seems to be summed with the -5v rail before going somewhere. I assume it'll be going to the feedback node. I haven't traced the rest of the circuit as it's messy, but since i already left the 5v rail out of the regulation loop, it looks like it'll be safe to remove -12v and -5v too. Before, if the -12v and -5v caps were removed the supply wouldn't start. We'll see what it does now.

[kaboom]

Re: The "proper" AT-to-split-supply mod!

A CFL example...

The electronic ballast, closer to a constant current supply, will attempt to deliver the same current to the lamp. Even if it means suicide.

I'll use a lamp current of 200mA (thin tubing) and a lamp voltage of 65V. For the purpose of this example, I'll ignore 'discharge lamp characteristics.'

A new tube has low voltage drop across the cathodes since the coating hasn't sputtered all away.

Most of the 13W goes into the discharge, maybe a watt or two lost in each cathode. A few years later, the fun starts. Regardless of what the discharge voltage is now, the cathode fall is much higher. Our emissive coating has sputtered away. Instead of maybe 5V across each cathode (5*.2=1W), it could be 20-25V now (25*.2=5W). The ballast is completely happy delivering the constant current, but there's now 5W at each cathode. The plastic base is now discoloring, possibly melting, where the tube goes inside.

Of course, if the lamp (cathode+arc) voltage goes higher than what the ballast can sustain, the lamp goes out. Then the ballast can burn up (if it didn't already) from trying to start the lamp in series resonance too long.

http://www.pavouk.org/hw/lamp/en_index.html

-Paul

[Pyr0Beast]

Re: The "proper" AT-to-split-supply mod!

Lamps draw more power when they age. Some would draw so much that the supply would collapse with output shottky diode cooking and so with the caps. Needs about double the heatsinking than original and it is moderately ok.

[kaboom]

Re: The "proper" AT-to-split-supply mod!

Also, to an extent, the current. I've seen a monitor that called for a 12V/3.5A PS, but it actually pulled 2.4A at 12V. This was with the backlight all the way up.

Sounds right, 1A for each inverter at full brightness and a little bit for the signal/logic/panel.