Can you run a computer on a linear power supply?

[JZatopa]

One of the larger complaints of computer audio is the noise generated by the SMPS of a computer. There are many good quality PS where this has become a non-issue but many old timers who are accustomed with old computer tech refuse to believe that a SMPS could have negligible noise. Because of this I started thinking about what could be possible solutions. One idea I have would be to build a linear PS with a toroid that is up to "audiophile" standards. Has this been done before, I have yet to find anyone who has tried this from my search. What are the problems with this, I am sure there must be some because I cannot find any info about someone doing this before. I understand that size and heat are a factor but is there something else I am missing?

 

[theAnonymous1]

[snip] is there something else I am missing?

The noise generated by the motherboard itself?

 

[KSTR]

Yes, this is possible and has been done in comercial products.

Look eg this review of the Audiovolver II, which is based on regular PC hardware with a soundcard but sans frontends, running under Linux.
http://www.audiodata-hifi.de/pdf/AuVo-SD1108.pdf
(german, but look at the pic on pg.3)

- Klaus

 

[DigitalJunkie]

I once ran an older AT motherboard from a linear supply I rigged up with some 78xx/79xx regulators,using 2N3055's as pass elements.I ran the whole works from a lead-acid battery.It worked well,and ran almost all day from a 17AH batt. At one point I even bypassed the 12V regulator,and ran the +12V rail directly from the battery.
I derived the negative rails from a small +5V input switcher,and a 7905/7912.

The main issue was heat.The +5V regulator ran pretty warm,despite a big heatsink. Assuming an ATX board,you'll need to add a beefy +3.3V regulator also,and it will probably run even warmer.
The +5Vstandby (constant power) could be handled by another 7805 and a separate small 6.3V transformer or something.

Edit: Here's the crude old drawing of the supply I built:
http://www.users.qwest.net/~ptaylor/Electronics/DC-PCsupply/PC-DC-AT-supply.jpg

 

[abraxalito]

What are the problems with this, I am sure there must be some because I cannot find any info about someone doing this before. I understand that size and heat are a factor but is there something else I am missing?

A mid-range CPU nowadays (65nm process) might have a core voltage around 1.3V. It'll also have a TDP (total design power) from 35 - 95W. Good luck with a linear power supply to deliver up to 70 amps to the CPU which meets the very stringent load regulation specs. Your task will be easier if you go for one of the new Atom CPUs as these require less than 3W I believe for the single core version.

 

[jackinnj]

I run a Dell laptop with a linear power supply -- just an LM317 with a pass transistor and fold-back current limiting -- the switch mode supply died and a new one was more than the computer was worth.

 

[DigitalJunkie]

Re: Re: Can you run a computer on a linear power supply?

A mid-range CPU nowadays (65nm process) might have a core voltage around 1.3V. It'll also have a TDP (total design power) from 35 - 95W. Good luck with a linear power supply to deliver up to 70 amps to the CPU which meets the very stringent load regulation specs. Your task will be easier if you go for one of the new Atom CPUs as these require less than 3W I believe for the single core version.

It seems that 'Vcore' is usually derived from a buck converter running from one of the higher voltage rails(3.3/5/12V),so you'll need to make sure they are 'stiff' enough.

Edit: I just finished up a linear supply for my laptop.A friend had loaned me a SMPS that worked,but was a bit low on current (only 2.5A,versus 4.5A) Now my lappy charges faster.

 

[SunRa]

Well, I guess that even laptops are having switching regulators on the motherboard so you still have HF noise injected. In fact, even the CPU will probably inject HF noise...

I guess the smart way to do things is to find a way to decouple the audio source from the PC. Be it a PCI or a USB/firewire/ethernet soundcard...

 

[HaLo6]

So the Audiovolver is just using a Pico PSU connected to PCB mounted toroids instead of using a wall wart... easy enough... I have been using a Pico PSU on a VIA Epia PX10000G and it works great! USB out into a TPA Buffalo32 ...

I actually plan on bumping up to a ITX mobo using a 775 processor and this is exactly my plan to make it look like! awesome pic now I have a good reference of a nice layout...

 

[analog_sa]

a linear PS with a toroid that is up to "audiophile" standards. Has this been done before,

Have a look at the computer audio asylum. Not only glowing opinions of the sonic abilities of linear supplies but also of after-market PS cords. And rather forget the "audiophile" toroid and use a proper EI tranny.

 

[dangus]

I ran my first computer, an Apple ][+ clone off a linear supply that I built in my bedroom. I can't say I noticed any particular audio benefit from doing this, but then that quasi-fruit only had 1-bit audio out. I later bought a "black beauty" switching supply for it because my linear one didn't have enough 12V capacity to run a floppy drive. (5V regulator was a 5 amp 78H05)

It might be worth providing clean linear-regulated power to the analog stuff on an internal sound card, but it would be far more sensible to just use S/PDIF and a good external DAC (or ADC).

 

[Fenris]

All current generation motherboards rely primarily on the +12v line. It's just easier to run 20 amps at 12v than 80 amps at 3.3v - cable thicknesses get to be extreme at those currents. There are several multi-phase buck converters to supply the various lower voltages for the CPU, IO, PCI bus, etc. The drawback with a linear supply is the loss of effiency, so while it is possible, it may not be beneficial. Go with as efficient a system as possible (lower speed Core 2 duo, LE series AMD, underclocked processor, or Atom) that meets your requirments. A SSD/flash drive or laptop hard drive to minimize power draw and noise would help. But you're still left with the problem that the internal switching noise of all the transistors and voltage converters can and will be radiated to some extent. Right now, I can audibly hear my computer (crappy Dell Optiplex) emit a whine at a few hundred Hz. It's not even audio related - no speakers attached, and it's not fan related. It varies with hard drive and mouse activity, evidence that there are some sympathetic vibrations in physical materials caused by electrical signals. I can only imagine how bad the RF/EMI and noise is on the inside of the case...