Building a silent audio PC. First attempt.

After long procrastination i finally built a dedicated audio PC following in general the Zuma recipe Computer Audiophile - Computer Audiophile Pocket Server C.A.P.S. v3 Zuma

Same case, MB and disk and a Nano 150W PS from Streacom. As the processor is a low consumption i3 3220T even the heatsinking arrangement of the Streacom case turned out to be an overkill.

The new PC is headless and replaces a nice Dell notebook feeding a USB dac with optical isolation and fifo reclocking.

So, one may wonder, how does the sound compare to the notebook running the same OS with JRiver 18 using ASIO drivers?

In short it sucks. Not a huge difference, but obviously inferior to what i remembered from the Dell. Latency turned out to be about half compared to the Dell, likely because of the lack of antivurus and other background processes - the Dell is a general purpose PC and no attempt has been made to optimise it for audio.

So, what then?

Plugged the Dell in, listened. Slightly better but definitely not what i remember. Maybe something else in the system has taken a dive. Or it's just one of those days...

The nano PS comes with a gigantic mains brick providing 12v @12.5A. It made me a bit suspicious so i switched it off. Now the Dell is sounding finally normal. It appears the PS is polluting either the room, the mains, or both.

This is what my scope shows with the probe shorted to ground and resting on top of the PS. No electrical connection and purely common mode radiated noise.


  • nano psu.jpg
    nano psu.jpg
    979.1 KB · Views: 355
Just finished listening after replacing the little horror with a bench PS @12.5v. Average current consumption seems to hover around 1.5A.

And what a difference! There is ambience, dynamics, hf details galore. Tomorrow will be building a linear supply. So, even with a nasty little switcher supplying the ATX voltages out of the incoming 12v, the raw dc still makes a huge difference, at least in my system and to my ears.

This exercise seems to prove (to me if to no one else :)) the importance of clean power compared to the absence of fans and hard disk drives and the reduced latency.
Last edited:
No, i am not :) Just sharing some personal observations. All this has been done before. Red Wine Audio per example sell a lithium battery system for powering PCs. Not cheap.

Of course, even though supplying cleaner power to a switcher seems to make a pronounced difference in my case, it is much better to get rid of the "nano", "pico", or whatever other switchers generate the ATX supplies out of the incoming 12v and instead source the required voltages from linear supplies. For a low power system, such as the one i use it is not hard. The problem is that most motherboards are fussy about the sequence of applying the various voltages, so some intelligence may be required for the PS.

Detractors are always fast to point out that in any case some switchers will still be left on the mobo itself. Probably too hard to bypass those. It seems though that motherboards are designed to keep switcher interference well controlled, while external switchers are much more damaging.
This is getting a bit frustrating. To recap what is being compared:

1. Dell XPS i5 Win 7 64 laptop

2. i3 3220t in a fanless case with SSD Win 7 64, power for ATX12v is supplied from a little switcher board called "nano" from incoming 12v.

Replacing the original 12v switching supply with a bench supply brought considerable improvement to the silent PC. Replacing that with a linear supply using a 1084 regulator and plenty of good capacitance brought a lateral change, rather than improvement.

Dell laptop still sounds better overall, which tbh is rather annoying. I see three options right now

- nano is a useless POS with excessive noise level. It would be a good idea to replace it with a low noise standard ATX supply for a test only atm i don't have a free one. Ideally the whole lot should be powered from linear supplies but that is definitely not a weekend project

- despite prevailing theories a low power processor is just not sufficient. There is a low consumption i7 which seems like a good candidate.

- motherboard is cursed by the audio gods and will never sound good. Time for attempt N2
Last edited:
Detractors are always fast to point out that in any case some switchers will still be left on the mobo itself. Probably too hard to bypass those. It seems though that motherboards are designed to keep switcher interference well controlled, while external switchers are much more damaging.
Interesting point, yes local on board SMPS's should be laid out properly and to the device data sheets, plus they are much smaller than the main computer SMPS's. Also a lot of the main supplies can leave a lot to be desired in both quality and EMC compliance, we once shipped a container load over and ended up skipping them because of the emissions. A good quality SMPS on the front end will be no different than a linear supply for resultant noise though, as a lot of PC electrical noise is created by the PC its self and the logic switching.
Again on motherboards and local supplies a good design will only have 2 SMPS's one after the other, then for the lower current local supplies use LDO's.
DELL generaly seem to be good quality, we use them in the CAD department for reliability.
Replacing the 12v dc with batteries is something i can relatively easily try out. I can see that compared to a linear supply it will help by limiting the RF hash of the nano going back into the mains and providing connecting wires are screened, also limiting radiated HF. But it will not in any way improve what the nano does to the pc supplies.

It may, of course turn out that the real problem is not how much the pc rails get polluted, but rather RF contamination of external audio components.
I have an isolator on the i2s side followed by a fifo and a reclocker (Ian's fifo) . Theoretically, there is no possible way the transport side can affect the sound. From what i know, there is only one option available as USB isolator Adnaco-S3B – USB 3.0 Over Fiber Optic Expansion System. I am not at all sure it will work with the EXA usb board.

What test results? I guess i could measure PS rail spectra with different supplies. How meaningful is that?