Some USB fast chargers are pretty impressive, if not quite a single chip. Honestly, 100W from a tiny plug in brick?

I wonder if the biggest problem for switching supplies is getting the output inductor size down? I know many years ago I had to design a buck dc-dc supply to fit on a daughter board, so it had very tight height constraints (think it was 8mm). Biggest problem was finding one with high enough inductance and current capability, and could only find one, which nearly scuppered it since production people hate (for good reasons) single source parts. But switching speeds have gotten a lot faster since then, so if I remember rightly, that means outputs need less smoothing, so can have smaller value components. Quick look suggests that the same supply now would only need an inductor about 4mm x 4mm x 2mm. Which is pretty awesome. Don't know how high the switching speeds would have to be to get the required inductor value down to something that could be packaged within a chip though. Looks like those high power USB chargers are using gallium, so they're probably pushing switching speed as fast as they can.

SiC semiconductors is starting to show up. Less on resistance.

Cheers,

Isn't Intel trying to make power supplies just a single voltage and then the motherboard cuts it up further? Once you have only 1 specific voltage you can make much simpler power supplies.

Unknown_K wrote on 2021-12-03, 00:44:

Isn't Intel trying to make power supplies just a single voltage and then the motherboard cuts it up further? Once you have only 1 specific voltage you can make much simpler power supplies.

It's already the case on some low-profile motherboards. A single DC 12V input powers everything. Depending on the board the power may be connected either via internal 4-pin (ATX 12V) or a barrel connector to an external 12V power adapter.

snufkin wrote on 2021-12-02, 22:39:

Some USB fast chargers are pretty impressive, if not quite a single chip. Honestly, 100W from a tiny plug in brick?

I wonder if the biggest problem for switching supplies is getting the output inductor size down? I know many years ago I had to design a buck dc-dc supply to fit on a daughter board, so it had very tight height constraints (think it was 8mm). Biggest problem was finding one with high enough inductance and current capability, and could only find one, which nearly scuppered it since production people hate (for good reasons) single source parts. But switching speeds have gotten a lot faster since then, so if I remember rightly, that means outputs need less smoothing, so can have smaller value components. Quick look suggests that the same supply now would only need an inductor about 4mm x 4mm x 2mm. Which is pretty awesome. Don't know how high the switching speeds would have to be to get the required inductor value down to something that could be packaged within a chip though. Looks like those high power USB chargers are using gallium, so they're probably pushing switching speed as fast as they can.

Chargers using GaN are gaining popularity as it allows smaller form factors (for the same power and port count). It seems USB Type-C connectors supporting Power Delivery additionally allows carrying other voltages (including 12V).

This configuration of using 12V to power most of things is to cut down on number of wires. This is already done on OEM computers in low end business computers for at least 10 years.

Only exception is 5V for powering 2.5" drives and optical drives if optioned.

Cheers,

Yes one of Intel newest ATX specs ATX12VO (2019) supply just +12v. The other new ATX12 spec: ATX12V v2.53 (June 2020) does not remove +3.3 and +5v but requires Bronze effiency minimum iirc.
Yes micro boards have had just a single supply for quite a few years. The ATX12V spec will not disappear for a while, you cannot put that much conversion on board and keep it small and reliable if you want to run an Intel or AMD with 8 to 16 with cores.
just MHO....ATX12v will be around for at least a few or dozen more years 😀 BTW Intel quit making motherboards in 2014 so they do not really care if it puts more work on the OEM board makers (again just my opinion).

Ha, thinking of on board power generation, I've mentioned before that the Abit KA7 generates its own 3.3V supply from the ATX 12V, and doesn't use the PSU 3.3V at all (as far as I can tell). And that was over 20 years ago.

I used to think the size limit for switching supplies would be the output inductor, which can be smaller if the supply switches faster, but that means you need lower gate capacitance (otherwise switching losses become big). But knowing I'm out of touch, I just went and looked, and DC-DC is more awesome that I thought: https://www.farnell.com/datasheets/3208420.pdf
40Amps@3.3V, @>95% efficiency (so supplying around 132W and burning 7W), with no external inductor (it's built in), in a 15mm x 30mm x 5.2mm package. Even higher current at lower voltages, but the efficiency drops off.

Maybe that's the sort of thing that rmay was talking about for what's currently possible, and the question is more about doing AC-DC in a small package. Maybe the problem is that FETs able to maintain isolation of 340V (peak of a 240Vac supply) are too big to switch fast enough?

I quite like the idea of house level 24V DC distribution. One good high efficiency AC-DC adapter by the fuse box, then run DC everywhere, plus AC to anything that really needs the power (kettle, oven, kettle, washing machine, iron. Kettle.). New devices then just need to do low voltage DC-DC. Would need some DC-AC adapters to connect some old transformer based equipment.

Yes that is the sort of thing I was talking about but if you go into the military environment there are some that also output a couple voltages along with others up to 200 amps

Externally Look like a giant hunk of silicon on a heat spreader (but internally might be a bunch in parallel)

I would imagine something a little less beefy could output common currents for our favorite motherboards

Or perhaps even today using different models of the device you linked to could be combined into a home made industrial at/ATX PSU but we would still need to rectify the input voltage to something suitable

My main thought is quality/ lifespan over cost to make something that is truly solid state with few discrete analog parts like caps to fail
but likely doesn’t make sense (yet)

rmay635703 wrote on 2021-12-03, 14:29:

Or perhaps even today using different models of the device you linked to could be combined into a home made industrial at/ATX PSU but we would still need to rectify the input voltage to something suitable

My main thought is quality/ lifespan over cost to make something that is truly solid state with few discrete analog parts like caps to fail
but likely doesn’t make sense (yet)

Isn't that effectively how modern ATX PSUs work, convert mains to 12VDC, then have multiple DC-DC converters to generate everything else from the 12V? I've only had a quick look around but it looks like the tricky part is getting the AC-DC bit small. And safe. Mustn't forget about safe. Ish.

Hmm, this is slightly thinking out loud, but given the problem I've read of using modern PSUs with limited 5V current with old motherboards, maybe there's a use in having a small active adapter that plugs in to the ATX (or AT) connector on the old motherboard, that on the other side accepts both an ATX connector from a modern PSU, but also the 8 pin CPU 12V connector, and it uses that to generate the 5V. 8 pin CPU 12V is (I think) rated for 24 amps, so that'd be around 50 amps maximum on 5V (allowing for efficiency losses), which is more than the 5 5V ATX power pins can handle safely. Maybe crossloading would be a problem as it wouldn't then be drawing anything from the PSU 5V. But most modern supplies are (as far as I know) fine with crossloading since all the supplies are independent. And there'd still be floppy/optical/hard drives using PSU 5V, unless there's a signaling noise problem of having each end using a slightly different 5V supply.

That way someone could get a modern 400W PSU with a maximum of 70W on 5V (for example), and be able to use >200W at 5V. That also frees up the combined 5V+3.3V current limit most PSUs have, so it can be used just for 3.3V.

Effectively balance the load of an old system to look a bit more like a new system.

It is, sort of, but there is more complexity than you would want on the input side (AC) to save cost as dumb as that sounds and no not every rail has to be driven off a single DC supply, PSUs as built are becoming more standardized as time marches but in the desktop world are still a bit of a snowflake, some are designed with perplexing component combinations that don’t make sense as we race to the bottom with random component combinations saving a few Pennie’s on materials. In my case my power supplies seem to fail in one of two ways with it usually being in the AC / DC power section.

Some DC-DC devices can handle 24 or even 48 volts and limit right on down to 3.3v or 5v (but usually with a much lower amperage)
These higher voltage DCDC could theoretically run off a reliable laptop supply but unless you use multiple laptop supplies are capped at about 300watts. Interesting sidebar to run mains supply off dc to avoid those “issues “

My hope was at some point we could avoid the AC deficiencies and simplify the whole thing to become more robust making it easier to rebuild proprietary supplies with new and less problematic “guts”. Some PSUs are just poorly designed and not worth repair.

Still a fantasy

I do agree however that we could make a robust single voltage supply to help with running vintage gear on modern supplies.
I would argue for it being completely separately supplied off the mains so it doesn’t over stress a questionable rated 12v or 5v line

rmay635703 wrote on 2021-12-03, 17:59:

I do agree however that we could make a robust single voltage supply to help with running vintage gear on modern supplies.
I would argue for it being completely separately supplied off the mains so it doesn’t over stress a questionable rated 12v or 5v line

I have a deal with mains electricity. I don't touch it, it doesn't cause my arm to slam backward and cut myself.

So I figure let someone else deal with getting down to sensible voltages (<24DC) and take it from there. I'd probably go for using a decent quality 80+Gold ATX PSU (Seasonic or the like; SilverStone use them for some models I think?). Good quality and thanks to mass (relatively) manufacturing much cheaper than industrial 24V supplies. Even the 400W models provide way more 12V than would be needed, so why not convert some of it to 5V instead. Another possible benefit of local +12 to +5 is that it'd run the PSU closer to its peak efficiency, compared to getting a massively overpowered one running at 20% capacity just to get a few extra amps on the 5V.

I've no use for such a thing right now, but I'm glad I now know there are some really good high current DC-DC converters, so design should be simple.

I guess it is only a matter of time until we get complete hybrid switching regulators in the TO-220 package that can replace linear regs and electrolytics.

OT:

snufkin wrote on 2021-12-03, 11:40:

I quite like the idea of house level 24V DC distribution.

Not so good if there is humidity somewhere.
There are still most tramways using DC, and DC leak causes the operators a *lot* of pain.

snufkin wrote on 2021-12-03, 18:54:

I have a deal with mains electricity. I don't touch it, it doesn't cause my arm to slam backward and cut myself.

There are good reasons why for example clinics use IT (Insulated Terra) networks.

Trying to reduce a PSU to a single chip would hit the limit on diminishing returns on how much you can shrink components and hit a hard limit on how much heat it could dissipate with such a small surface area. You would then require a lot of active cooling for said chip to keep it within an efficiency envelope and have therefore defeated the purpose of shrinking the PSU to begin with.

Far cheaper to just make the PSU deliver 12v to the motherboard and have the motherboard step the voltage down as required via small but highly efficient IC's. (Funny I think a huge semi conductor corporation is trying to do exactly this hmmm I wonder who it could be)