Why is 12V used for power distribution in electronic devices?

[NiHaoMike]

Exactly where did the norm of using 12V for power distribution inside electronic devices come from? And why was it chosen so that the tolerance sits right inside the working voltage range of a 12V lead acid battery, complicating the design of a DC/DC converter to get a regulated 12V rail from a 12V lead acid battery?

[T3sl4co1l]

You rather answered your own question...

Back in the day, the most demanding application, with respect to the regulation of 12V (or lack thereof), was probably the vacuum tubes made to deal with it.  Automotive regulators were poor (mechanical), so automotive radios used tubes with heaters specified for 12.6V nominal operation which would have adequate emission (and freedom from cathode poisoning) down to, whatever, 11V or so, and adequate lifetime (and freedom from diffusion and whatever) up at 14 or 15V.  There were also tubes meant for direct operation at 12V (plate supply), awful specs but, for what they were used in, better than a stupid vibrator supply!

And the history goes back a bit further than that, still, with 6.3V being more common in autos (sometimes positive ground, at that), and 2V and 6.3V being common A battery ratings (again, yep, tied to lead acid), going back to the beginnings of radio itself.

And what do you really need regulated 12V for, anyway?  A lot of regulated supplies are really just arbitrary, and the circuit either can do just fine on variable power, or can be made to behave (e.g., an amplifier using op-amps, instead of discrete transistor circuits with shit PSRR).

Anyway, if you must have a DC-DC converter, there's SEPIC, Zeta or buck-boost ("flying inductor") for that.  You'll probably want something like that anyway for a much wider input range, accounting for dips (down to 6V during cranking, say) and swells (15V max charging, but also >30V in the rare load dump condition).  Nothing at all wrong with doing that, pretty standard.  If you require efficiency, then a DC-DC at all is required; and if you can't choose a lower supply (like 5 to 10V) to use a buck, then just change topology, probably choose a different controller, and get on with it.

Mind, also, 12V is completely arbitrary, outside of the nominal lead acid 6S discharge voltage.  14V might be a better tradition nowadays, with lithium batteries being dominant.  Or you could use 11V, or 10V, or anything else just as well.  MOSFETs are typically spec'd for Vgs(on) = 10V, though 12V is commonly used to supply drivers.  Anything from Vgs(on), up to Vgs(max), could just as well be used.  Why is 12V used, indeed?

Tim

[NiHaoMike]

I was thinking of PC hardware that requires a fairly tightly regulated 12V, in many cases using 16V capacitors which leave too little margin for the 14.4V or so lead acid charging voltage. Google has used some motherboards tweaked to accept a wider range in order to simplify the addition of battery backup in the servers.

[Siwastaja]

Well, in this case, you answered your question quite well already...

To save cost, tighter regulation is needed. When the bus voltage is specified tightly, they can use the lowest rated MOSFETs, capacitors, etc. everywhere. In masses, this does matter. 16V capacitors, 20V MOSFETs, 16V absolute maximum ratings in ICs... This goes down to silicon fabs optimized to a certain voltage.

Now, with cheap PC hardware, the supply is almost always a switch mode converter from mains. Because these converters are regulated anyway, they can easily produce tight enough regulation, to allow using tightly specified components downstream. For the accuracy needed, they can use the cheapest possible regulation feedback, and it's still good enough.

So, the answer is cost. Reduce the margins to minimum.

Why exactly 12V? Arbitrary choice and tradition. If it was anything else, you'd still ask the same question. It has to be something.

[GeoffreyF]

What others said but it also comes to battery chemistry.  We end up with something that is some integral number of cells and then with a voltage high enough so that as the battery runs down, there will still be enough regulated voltage to do the job.

Different types of semi-conductor technologies impose voltage requirements on how they function.  I think this is tied to the physics of different types of transistor or FET but I can't explain that authoritatively. 

[Mechatrommer]

Exactly where did the norm of using 12V for power distribution inside electronic devices come from?

where is it required? or as stringent prerequisite specification? i know one, PC PSU, but the power source comes from mains. and most digital/logic circuit comes from 5V supply or further stepped down to 3.3V. afaik 12V only for motor or power drive which still works plus minus 2V or more?

complicating the design of a DC/DC converter to get a regulated 12V rail from a 12V lead acid battery?

why is it required to regulate a 12V battery to 12V? or is it hard to boost and then buck if even you have to do it? given your long membership and experience and as associate to Tiffany Yep, i dont think this type of question should come out of you

[David Hess]

A better question might be why PC power supplies started off using 12 volts (and -12 volts) instead of 15 volts (and -15 volts).  Actually, that is not such a mystery and it has nothing to do with 12 volt lead-acid batteries. 

Early MOS memories and processors commonly required a +12 volt (and -5 volt) supply in addition to the +5 volt logic supply.  The last examples of this include the Intel 8080 microprocessor, Intel 16kbit DRAMs, and Intel 8kbit EPROMs.  Parts in the generation after these only required +5 volts which was a major improvement but the +12 volt auxiliary power supply stayed for things like mechanical mass storage and RS-232.

These days PC designers would prefer that the +12 supply be a little higher to ease requirements on the low voltage point-of-load regulators which produce very low voltages at high currents but not too much higher.  +48 volts (1) would be the obvious choice and some non-PC systems use this but the buck switching regulators would need to operate with a duty cycle of about 2% which is difficult.  More commonly the +48 volts is stepped down to +12 volts and then to the low voltage.

(1) The nominal battery voltage of the POTS (plain old telephone service) standard is 52.1 volts based on a 24-cell lead-acid battery.

[NiHaoMike]

why is it required to regulate a 12V battery to 12V? or is it hard to boost and then buck if even you have to do it? given your long membership and experience and as associate to Tiffany Yep, i dont think this type of question should come out of you

A buck boost converter is more complex than just a buck or a boost converter and there are considerably fewer controller chips and ready made modules for that. And many buck boost converters were basically nonisolated flyback converters connecting the inductor to the input then the output, until LT came out with some chips to implement converters that do it right - dynamically transitioning between buck and boost modes.

I also don't see how Tiffany Yep is relevant when she mostly views power electronics as black boxes that make (electrical) noise as a side effect of their operation - only tolerated in the system because the processor cannot run her code without a power supply.

[T3sl4co1l]

I wonder how much in a PC is actually dependent on 12V anyway.  Are hard drive spindle motors usually powered directly from 12V?  Most loads are themselves switching regulators, which would be perfectly happy down to, eh, probably 9V, assuming UVLO isn't set higher.  Upper limit probably below 20V, because 20V transistors or 18V drivers are fairly common and that's the "eee popped it" limit.

Would be an interesting exercise to see what typical hardware can tolerate what voltages.

The safe option, of course, would be to regulate!

Tim

[mariush]

They added 20v with usb power delivery and some motherboards are already designed to work directly from an external dc brick with 18..18.5v dc output , which should be accepted by devices powered in the future from 20v usb ports.

I personally wish the computer psu manufacturers would get together and make up a new standard since intel refuses to make changes to existing one. 

For example make separate connector (to reduce risk of people accidentally damaging parts inserting wrong connector) for 20v (like pci-e 6pin or 8pin) but make them 4 pin with option to join together connectors for 4+4+4...  remove 12v from "main" connector and just reuse a pci-e 6pin if needed (make 12v optional, allow motherboards to reduce wires by creating 12v if needed from 20v, for itx boards or boards with no pci-e connectors)

I'd drop 3.3v , -12v and replace power good and power ok wires with usb data signals and have the psu report "health" statistics and current monitoring and all that and also receive through same wires commands like change colors of rgb fans, have fan spin at higher than default speed or disable turn off fan if psu is below some temperature etc etc
Instead of a big fat 24pin connector we could have a 6-8 pin connector with 5v standby and 5v and  the communication channel (keep it 5v to make it easy to create "adapters" for backwards compatibility with atx... make an adapter with a micro or something that creates the power good and power ok signals and a dc-dc converter to make 3.3v from 12v or 20v)
 
@T3sl4co1l  : fans, rgb strips now use 12v ... some mechanical hdds use 12v motors but quite a few use 5v. But i suppose it would be easy to manufacture fans working at other voltages like 20v, or just use 24v fans and change the stickers on them to say 20v or 18v or whatever.

[David Hess]

I wonder how much in a PC is actually dependent on 12V anyway.  Are hard drive spindle motors usually powered directly from 12V?  Most loads are themselves switching regulators, which would be perfectly happy down to, eh, probably 9V, assuming UVLO isn't set higher.  Upper limit probably below 20V, because 20V transistors or 18V drivers are fairly common and that's the "eee popped it" limit.

12 volts has become the de facto standard for distribution to all of the low voltage high current point-of-load regulators and they are not that picky about input voltage.  3.5" hard drives still use it for the spindle motor *and* the voice coil but 2.5" hard drives get away with just the +5 volt supply.  Most cooling fans use 12 volts.

I personally wish the computer psu manufacturers would get together and make up a new standard since intel refuses to make changes to existing one.

There is not much reason to change it.  The only real improvement would be to have a higher voltage high current supply for the point-of-load regulators to more easily support high power CPUs and especially GPUs but 12 volts is not terrible in this regard.  Bigger systems do use a higher voltage for this but they do not use commodity PC power supplies or motherboards.

I'd drop 3.3v , -12v and replace power good and power ok wires with usb data signals and have the psu report "health" statistics and current monitoring and all that and also receive through same wires commands like change colors of rgb fans, have fan spin at higher than default speed or disable turn off fan if psu is below some temperature etc etc

The power good signal is required to control -reset on the motherboard to prevent unintended operation.  Relying on USB for this would be a really bad idea.

[mariush]

I personally wish the computer psu manufacturers would get together and make up a new standard since intel refuses to make changes to existing one.

There is not much reason to change it.  The only real improvement would be to have a higher voltage high current supply for the point-of-load regulators to more easily support high power CPUs and especially GPUs but 12 volts is not terrible in this regard.  Bigger systems do use a higher voltage for this but they do not use commodity PC power supplies or motherboards.

I'd drop 3.3v , -12v and replace power good and power ok wires with usb data signals and have the psu report "health" statistics and current monitoring and all that and also receive through same wires commands like change colors of rgb fans, have fan spin at higher than default speed or disable turn off fan if psu is below some temperature etc etc

The power good signal is required to control -reset on the motherboard to prevent unintended operation.  Relying on USB for this would be a really bad idea.

I do. There's lots of motherboards that don't really need the 24pin connector not to mention a separate 4 pin or 8 pin cpu connector , here's one example of such board: https://www.newegg.com/Product/Product.aspx?Item=N82E16813157728&ignorebbr=1

These are boards that probably don't go over 30-40w and most of that is probably pulled from the 12v wires in the connector, is there really a need to have 4 pairs of wires carrying 3.3v and 5 pairs of wires carrying 5v PLUS the 5v stand-by ?
On such boards, they could have only the 6-8 pin connector and no optional connector for 12v and 20v and have a tiny dc-dc converter creating 12v for the pci-e x1 slot (~10-20w) and fans. Then maybe we would have modular STX style power supplies in the 150-200w range and you'd have only one cable between psu and board.

Remove the connector and use the space for more functionality like in the case of the board above maybe a second memory slot or a m.2 connector


As for power good signal ... I assume on most motherboards the chipset is powered from 5v stand-by coming from the psu which is isolated circuit from the rest of the psu. It shouldn't be too hard to isolate the data wires for some simplified usb or some spi or i2c or canbus (whatever bi-directional standard is agreed upon) and then have a set of commands created (like FTP or the SATA commands or whatever) ... i just said usb because i figured since chipsets already have usb functionality built in, it wouldn't be hard to change chipsets to add a usb 1.1 or usb 2 or some simplified usb controller in the chipset that would only accept devices called "power supply" and then learn about their capabilities (wattage, number of connectors, can they provide 20v or not, do they have 12v or not and if not and the motheboard can't work without 12v then refuse to start)

If the chipset "asks" the psu if it's ready and the psu dosn't reply or says no or says wait... that's your power good... same,  psu could send a data packet saying psu must shut down... if the chipset has a polling rate of 1000hz or something like that, it can receive status from psu within 1 ms and react.

oh... and psu could also send messages like "fan failed" and "psu overheating, derating output to xxx watts, bios should throttle cpu down, notify user that psu needs to be serviced or replaced" ... just some thoughts.

And another thought... if they add 20v, then a single 4 pin cpu connector could power a whole system with a office cpu, like the ryzen 1200 or athlon 200ge which average around 50w of power consumption... even a single awg18 pair can safely carry let's say 8A ... so 2 pairs of wires could carry 300w+ into a motherboard that'll only use ~ 100w

[David Hess]

As for power good signal ... I assume on most motherboards the chipset is powered from 5v stand-by coming from the psu which is isolated circuit from the rest of the psu. It shouldn't be too hard to isolate the data wires for some simplified usb or some spi or i2c or canbus (whatever bi-directional standard is agreed upon) and then have a set of commands created (like FTP or the SATA commands or whatever) ... i just said usb because i figured since chipsets already have usb functionality built in, it wouldn't be hard to change chipsets to add a usb 1.1 or usb 2 or some simplified usb controller in the chipset that would only accept devices called "power supply" and then learn about their capabilities (wattage, number of connectors, can they provide 20v or not, do they have 12v or not and if not and the motheboard can't work without 12v then refuse to start)

If the chipset "asks" the psu if it's ready and the psu dosn't reply or says no or says wait... that's your power good... same,  psu could send a data packet saying psu must shut down... if the chipset has a polling rate of 1000hz or something like that, it can receive status from psu within 1 ms and react.

I will repeat it again; removing the power good signal or replacing it with something other than a single digital bit is an incredibly bad idea.

Another function of the power good signal is to pull the entire motherboard and everything on it into reset if the power on the DC side glitches or is about to glitch.  Between the time of the glitch and being able to interpret some data stream, things can go very wrong.

oh... and psu could also send messages like "fan failed" and "psu overheating, derating output to xxx watts, bios should throttle cpu down, notify user that psu needs to be serviced or replaced" ... just some thoughts.

Some power supplies support this through USB or some other link.

And another thought... if they add 20v, then a single 4 pin cpu connector could power a whole system with a office cpu, like the ryzen 1200 or athlon 200ge which average around 50w of power consumption... even a single awg18 pair can safely carry let's say 8A ... so 2 pairs of wires could carry 300w+ into a motherboard that'll only use ~ 100w

This is not a bad idea and some higher power custom systems work this way although 20 volts is probably a little low.  The disadvantage is that now a whole mess of cables have to go to the motherboard which now must supply low voltage power for various mass storage devices which are still going to require 3.3, 5, and 12 volts.

[T3sl4co1l]

Speaking of high power in PCs, I'm a bit surprised blade connectors never caught on.  They're always the old, old (and I mean 1960s old) Molex circular or rectangular pin and socket connectors, or something similar thereto.

I know blade connectors tend to be expensive in general; wouldn't be surprised if production did look into them, and just never could get a good enough price on them.

...With an obvious catch: battery contacts tend to be blade type, which is nice, good for them.  Probably one of the few such types with enough commodity production to be worth it?

Tim

[David Hess]

I have seen a few Molex style motherboard power connectors catch on fire but reliability wise they seem pretty good.  I have also seen blade connectors get loose and catch on fire so I doubt they would be any better.

Were you thinking of a blade connector inside of a connector housing?  I am not sure they would be any better.