Hacking an AT/ATX power supply.... / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Hacking an AT/ATX power supply....

previous 1 2 3 next

Move Thread

LAN_403

m4ge123

Sun Jan 22 2012, 04:49AM

Registered Member #4118 Joined: Mon Oct 03 2011, 04:50PM
Location: MD
Posts: 140

You can only draw half the rated current because originally it's split between both halves so each half is pulling 15A half the time, and if you pulled 15A through the whole thing that would be 15A all the time.

Neet Studio

Sun Jan 22 2012, 05:33AM

Registered Member #4037 Joined: Fri Jul 29 2011, 03:13PM
Location:
Posts: 86

Here is a circuit diagram for a 200W ATX supply:

Hopefully it would give you some idea of roughly what's on the board. I haven't seen more modern schematics. The primary driver stages might be a bit different, but the secondary side would probably be very similar. Things to watch out for is the over voltage protection circuit might kicks in if you forget to take care of that.

I have done modification to multiple power bricks before by reconnecting the parallel winding to serial, trimming down output voltages etc. These days, I would consider just making a buck-boost converter like a SEPIC connected to the 12V output and not even bother to mod the PC power supply. Safety/insurance are my reasons for not doing stuff like that. (Signs for getting old and worrying about liability etc.)

The transient performance of microcontroller based PWM outside of specialized digital supply with DSP isn't all that great compared to an analog solution as you have very low bandwidth in your feedback path. I would go with injecting a DC bias via a DAC to the feedback path with traditional analog PWM controller.

Steve Conner

Sun Jan 22 2012, 07:30AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I think in mage123's explanation, one-half should be 0.707, since it's RMS current that heats wires.

stepher

Sun Jan 22 2012, 09:56PM

Registered Member #4357 Joined: Thu Jan 19 2012, 05:32PM
Location:
Posts: 8

Wow!! Guess I came to the right place!! Thanks for the help!

I'm much more of a digital guy and, while I understand the basics of power supplies, magnetics are not my forte' and there must be something here I'm missing. So, bare with me while I do some thinking out loud and, hopefully, your feedback will help me get it.

So, I pull a typically designed ATX supply out of the box (no mods, yet) with the following differences from a "normal" ATX supply: the primary has been over-designed and can handle pretty much whatever power the secondary needs and it has only 4 outputs: +/-12VDC and +/-5VDC with the following specs:

+5VDC @ 20A, +12VDC @ 10A, -5VDC @ 3A and -12VDC @ 2A

Total power to secondary: 260W, give or take. Let's say the xformer can handle up to 300W drawn by the secondary. Let's assume here that I do not plan to use the +5VDC (except for maybe a 5W load to keep the supply active)

Now I want to modify it. I'd like -12VDC to supply as much current as the +12VDC supplies, 10A. Based on comments made here, the lower windings (those "below" the C.T. providing the source for the -5V and -12V lines) essentially are the same thickness as those of the upper windings (the ones "above" the C.T. used to provide +5V and +12V). If this is valid, then my sense is that I could change the diode(s) and caps in the -12V line and create a mirror image of the +12V @ 10A. So, now I have +12V @ 10A and -12V @ 10A. No apparent violation of the power limits of the xformer (240W vs 300W limit). Should work. OK to this point?

So, now, let's assume I want 24VDC from this thing. Can I use the -12VDC as "ground" and the +12VDC as "hot" and get 24V @ 10A? (Assume I isolate the ATX supply's C.T. ground from the load). If not, why not?

So, onto what may be the key question: How would that differ from using the xformer end points (-12V line and +12V line) to create a 24V @ 10A supply? I sense that using only a half-wave diode scheme may impact current (out of the box, the +12V circuit uses half-wave rectification to get 10A). If I used full wave, my guess is I should be able to get 12V @ 20A, but since the +/-12V lines are designed for 10A, that would be my limit (tho, the filtering might be less demanding? With a full wave bridge, would that get me to +24V @ 10A? What about just a bridge rectifier (no C.T.)?

Thanks for your patience with me. I really appreciate the help on this.

Cheers....Steph

Steve Conner

Sun Jan 22 2012, 10:50PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

stepher wrote ...

So, now, let's assume I want 24VDC from this thing. Can I use the -12VDC as "ground" and the +12VDC as "hot" and get 24V @ 10A? (Assume I isolate the ATX supply's C.T. ground from the load). If not, why not?

Yes.

wrote ...

So, onto what may be the key question: How would that differ from using the xformer end points (-12V line and +12V line) to create a 24V @ 10A supply?

I don't see any difference other than the choice of ground reference. By uprating the diodes in the -12V rail you created a bridge rectifier, and as I said in an earlier post, that would be the way to make a 24V supply.

There are some issues with the output filter inductor that I skipped over. The -12V rail also has an inductor and that would need uprated to 10A. Or you'd need to rearrange things so that the full 24V output from the bridge rectifier went through one inductor. (Floating the centre tap of the transformer and grounding what used to be the -12V rail would take care of that.) The inductor would then need more turns to handle the extra voltage, and the voltage regulation setpoint would need doubled too, or it would just keep on putting out 12V.

Finally, it's not like one half of the winding generates +12V and the other half generates -12V. They change places every half-cycle.

Neet Studio

Mon Jan 23 2012, 12:32AM

Registered Member #4037 Joined: Fri Jul 29 2011, 03:13PM
Location:
Posts: 86

stepher wrote ...

Based on comments made here, the lower windings (those "below" the C.T. providing the source for the -5V and -12V lines) essentially are the same thickness as those of the upper windings (the ones "above" the C.T. used to provide +5V and +12V).

That's a really really bad assumption. The cross section area of a transformer is always a premium. No sane commercial design would waste extra wire thickness for a winding if it does not require to handle the current. So your scheme would not work as is.

What is true is that the high current windings are done by having multiple thinner wires in parallel. If a winding requires less current, then fewer wires are used in parallel for that winding. So for example, someone would use a certain wire thickness for 2A wire. So the -12V @2A winding might have 1 of those wires where as the +5V @20A one would use 10 wires in parallel. (Sometimes they might use a thicker wire too.)

If you don't believe me, please take apart a transformer in question. I have done so in my past.

As for full wave vs half wave. You do realize that there are 2 diode drops vs 1 for a full wave. Would you hire a designer that use extra diode(s) and have a lower efficiency for a product that has low profit margins?

Steve Conner

Mon Jan 23 2012, 07:12AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Well, that's somewhat of an oversimplification. A full-wave rectifier only has one diode drop compared to a bridge's two diode drops. But:

The diodes in the full-wave circuit need twice the voltage rating for a given output voltage. Higher voltage diodes cost more and have a higher voltage drop.

The transformer winding needs 40% more copper because only half of it is in use at any time.

So it comes to a tradeoff between diodes and copper. Nevertheless, in whatever switchers I've taken apart, the full-wave circuit is used for voltages below roughly 100V, and the bridge is used above that.

It's interesting to see that in line frequency power supplies, the corresponding voltage is 10V or less, because the 50/60Hz transformer contains so much more copper than a SMPS one, and the slow diodes are cheaper.

I don't think any designer of low-cost SMPS was bothered about efficiency, since the end user pays for the losses. Well, until the Energy Star thing, I think it mandated a minimum efficiency.

Neet Studio

Mon Jan 23 2012, 11:33AM

Registered Member #4037 Joined: Fri Jul 29 2011, 03:13PM
Location:
Posts: 86

Oops. Sorry. Ignore what I said, I was thinking of bridge for some reasons. For some reasons I thought of PC supplies only have 1 winding, not using a CT for some reasons. Must have been thinking of bricks/or ahead of myself.

If there is a CT, then it is possible to use both the 2 windings to get twice the voltage. Going single wave is one way, but not likely to get you full power. That's probably where I thought of bridge here.

The diodes in a PC power supply tends to be using 40V schottky OR'ing diodes, so going bridge would need 3 packages.

stepher

Mon Jan 23 2012, 07:29PM

Registered Member #4357 Joined: Thu Jan 19 2012, 05:32PM
Location:
Posts: 8

Once again, thanks to all for the insight. I discovered I'm really not as confused as I thought I was. I realized because of the way I asked my original question, I got responses that I wasn't quite expecting. My posting was really intended to focus much more on the xformer/magnetics and much less on the rectifier/filter part of the circuit. My apologies for misspeaking :(

If you'll bear with me just for a bit more, I want to truly clarify a couple of things.

OK, back to the "typical" ATX supply.This one has a CT secondary to ground but is designed to provide only +/- 12VDC, i.e. 1/2 the windings designed to deliver +12 @ 10A and the other 1/2 of the windings for -12V @ 10A (voltages as read after diode and filter). If I lift the CT from ground (and ignore it), remove the -12 rectifier/filter from the circuit and disconnect the xformer lead from this circuit (-12V) and connect it to ground, I would expect to read (at the orig +12VDC output) +24VDC and be able to draw 10A from the circuit...essentially, series windings (Steve, I believe you've already confirmed part of this).

Now, if I could put the windings in parallel (ignoring phasing for the moment), I would expect my orig +12VDC output to read +12VDC, but I would expect to be able to draw up to 20A from the circuit. Not that I plan to do this, but my sense is to physically connect the windings in a parallel arrangement on an ATX-style supply, I would have to break the CT apart to be able to get the right connections to the right places to get them in phase (otherwise I would expect a nasty AC short to appear). For anyone in the know, am I right about this?

OK, if I'm on a roll here, my last question: I understand x-sectional area of the windings determines the current capabilities. And since these supplies typically use Lizt-style wire, then the higher current outputs would use (many?) more 18 gauge strands in a winding "wire" (increasing equivalent x-sectional area) than would be used in lower current outputs. Simple economics. So, again, for those in the know, given that the minus voltages (-12VDC, -5VDC) have so much less current demand placed on them, can I imply that the number of strands used in these windings is much fewer (thus limiting max current) than used in their plus counterparts (+5VDC, +12VDC). If this assumption is correct then, if I lift the CT in an ATX supply (with all the compliment of voltage outputs), while I might be able to get +24VDC out of it (using the "end" points), I would be limited in the amount of current I could draw due to the windings for the minus voltages. So, instead of being able to get the 10-15 amps (+12V windings being the weakest link on the plus side), I would only be able to get 2-3 (maybe 4 in a stretch) amps because that is the limit for either the -12 or -5 windings.

Or, as he hopefully asked ;) do the -5 windings use the same number of strands as the +5 windings (and same question for -12/+12)...which means I could realistically get the 10-15 amps out of the supply?

Again, I really appreciate all the help (and your patience with me) as I sort this out.

Cheers....Steph

m4ge123

Mon Jan 23 2012, 09:31PM

Registered Member #4118 Joined: Mon Oct 03 2011, 04:50PM
Location: MD
Posts: 140

As Steve Conner said, "It's not like one half of the winding generates +12V and the other half generates -12V. They change places every half-cycle." The reason the -12V rail is rated so low is because of the diodes. What you are calling the negative winding carries just as much current as the "positive" one. Each conducts current for half a cycle.

previous 1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob