.

[fullmoon6661]

.

[mariush]

Read this  :  https://www.ti.com/lit/ml/slup129/slup129.pdf

TI's Magnetic Amplifier Control for Simple Low-Cost, Secondary Regulation by Bob Mammano

[taste_tester]

The above link is good if you want to learn the theory behind group regulation (actually I'm quite thankful for it), but this discussion maybe provides a better answer in context of why it matters for your PC:

https://superuser.com/questions/1051024/under-what-conditions-will-a-group-regulated-psu-damage-a-modern-computer

To  drastically oversimplify it, modern lower power PC's don't really draw anywhere as much +12V as older PC's did, and as time goes on +12v is even less needed however +3.3v much more. Lots of components that historically use 12v + 5v  together now only use 3.3v, and in the last decade largely only 5v and no 12v.  A circuit that needs a certain level +12V load to be stable is not a good idea if you barely sip +12V and need really accurate and large amount of 3.3V and 5V.

[magic]

This article doesn't exactly describe common ATX power supplies and doesn't answer OP's question about shared inductor.

One common type of old group-regulated PSU is found on page 6 here:
https://www.onsemi.com/pub/Collateral/TND350-D.PDF

This is combined with a forward topology primary. The secondary windings of the transformer output two square waves in approximately 12:5 proportion but higher than 12V/5V, the positive halves are applied through diodes to two inductors in 12:5 proportions wound on a common core which implements a sort of dual buck converter. Output voltage is regulated by varying duty cycle like in any buck, but the two voltages are locked together in approximately 12:5 ratio.

Another option found in low power or really cheap PSUs is flyback topology. Then, pulses from the secondary directly charge output capacitors through diodes and the dual buck coil is not present.

The 3.3V rail is generally derived from 5V, either by means of a mag amp buck converter (described in both PDFs linked here) which requires two coils (the mag amp coil and an ordinary buck coil) or by a switcher (the mag amp is replaced with a transistor) or by linear regulation (no mag amp, no buck coil, only one transistor and terrible efficiency). This choice is independent of how the 12V/5V rails are generated, although there tends to be a correlation: the cheapest PSUs use flyback and linear 3.3V rail, the better use forward and buck/mag amp.

So the number of coils can be anywhere from zero to three and these are all group regulated designs. Sometimes much smaller inductors are used for ripple filtering, placed inbetween two capacitors.

To  drastically oversimplify it, modern lower power PC's don't really draw anywhere as much +12V as older PC's did, and as time goes on +12v is even less needed however +3.3v much more. Lots of components that historically use 12v + 5v  together now only use 3.3v, and in the last decade largely only 5v and no 12v.  A circuit that needs a certain level +12V load to be stable is not a good idea if you barely sip +12V and need really accurate and large amount of 3.3V and 5V.

It's the exact opposite. Computer chips used to run on 3.3V and 5V straight from the PSU, but now they use much lower voltages which are generated on the motherboard by small bucks drawing power from 12V. Some chips may still use 3.3V for I/O interfaces specified at this level, and the largest remaining consumer of 5V is USB and disks.

[HwAoRrDk]

To  drastically oversimplify it, modern lower power PC's don't really draw anywhere as much +12V as older PC's did, and as time goes on +12v is even less needed however +3.3v much more. Lots of components that historically use 12v + 5v  together now only use 3.3v, and in the last decade largely only 5v and no 12v.

Yeah, this is nonsense.

The majority power draw on a modern ATX desktop PC is from 12V. As evidenced by the fact that for at least a decade now we have had secondary 12V-only connections to motherboards (EPS/12V2 connectors) used pretty much exclusively to power the CPU (through further local VRM conversion), and PCIe power connectors for GPUs, etc. (again, with local down-conversion).

In fact, Intel have developed and been pushing to PC OEMs their new "12VO" ATX standard, where the PSU will only supply 12V.

[mariush]

They still use 5v for the RAM vrm/dc-dc converter (and I'm not sure but I think they may switch between 5v and 5vSB to keep powering ram while system is in sleep/stand-by)

I think some motherboards also use a 5v-3.3v dc-dc converter to power the M.2 connectors, so that the voltage drop on the 3.3v through the motherboard won't cause issues.

Yeah, the 12v0 standard drops 3.3v and 5v, and they did it mostly for stand-by and idle efficiency, and possibly to make it easier to design power supplies that can turn on and off hundreds of times a day like it can happen with laptops.
Don't like it through, mainly due to them adding new connectors (different connector for motherboard-sata and motherboard-molex cables to power drives) and for making it optional to add a pci-e 6pin like extension to the 10-12 pin connector, basically losing all the benefits of not having a bit 24pin power connector.   

They added 3.3v in the M.2 connectors, there's 10w worth of 3.3v in pci-e slots, but i feel like it's easier to drop 3.3v from power supply compared to dropping 5v, as 5v is still used in usb connectors and sata and mechanical hard drives.

I would have loved to see a modern power supply standard that adds 20v so that motherboards could then have good usb power deliver ex 5v/12v/20v and power a monitor or external video card through usb 20v

[MrAl]

Hi,

I havent read any of the links but i would suppose that group regulation simply refers to the technique of regulating several voltages using only one feedback.  That means all the windings are on one transformer core and the bunch of the outputs (after rectifying) are all regulated by sensing only one of the outputs or some simply addition of the outputs.  When doing it that way the output voltage of say the 5v line may drop or increase when the 12v line is loaded, so there will be some interaction between voltages as they are loaded.  It shouldnt be too much though.

The +12v line is used for a LOT of things these days in the PC.  That's especially true of high performance graphics cards and CPUs.  The CPU cores usually have their own individual buck to regulate the current to each core.  They would get the power from one or more of the +12v lines.  When it is buck'd down, they get more current at the lower voltage to run the CPU.

So i dont ever see the +12v lines going away.  It is true that the SSD's use MUCH MUCH less power than the HDD's but other things still draw current.  Also, with the advent of USB charging at higher powers like 100 watts we will see buck circuits converting 12v down to 5v to power charge phones and laptops or whatever.

[taste_tester]

To  drastically oversimplify it, modern lower power PC's don't really draw anywhere as much +12V as older PC's did, and as time goes on +12v is even less needed however +3.3v much more. Lots of components that historically use 12v + 5v  together now only use 3.3v, and in the last decade largely only 5v and no 12v.

Yeah, this is nonsense.

The majority power draw on a modern ATX desktop PC is from 12V. As evidenced by the fact that for at least a decade now we have had secondary 12V-only connections to motherboards (EPS/12V2 connectors) used pretty much exclusively to power the CPU (through further local VRM conversion), and PCIe power connectors for GPUs, etc. (again, with local down-conversion).

In fact, Intel have developed and been pushing to PC OEMs their new "12VO" ATX standard, where the PSU will only supply 12V.

I see how I was incorrect, I was thinking in terms of stuff like M2 expansion cards which run off of 3.3v but was not considering that said voltage levels are not provided directly by the ATX PSU.