Building a linear ATX PSU for an audio PC. Need help & advice

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Although I can't speak for it as I haven't tried myself, but many who have tried a USB optical isolator (like the Adnaco S3B) have said it's better than a regular motherboard USB connection, but not better than some specialty, linear powered USB cards; most of which cost less than the Adnaco. They claimed a lack of timbre, realism, or nuance to the music, in comparison.

You compare apples with pears. Both are good but the taste is different.
In fact you compare diferent (IC) DAC's with diferent output stages.
For an exact comparison you must use "exactly" the same DAC, the only difference should be how you isolate the connection (ground loop) or the power supply type, but not both of them simultaneously changed.
 
So just to be clear, you are worried about using a switching power supply to feed something that only consists of switching circuits?


Absolutely! RF radiation off a competently designed MB is not a serious issue, where the pico/nano appear to radiate a lot worse than even a standard ATX supply.

I don't pretend to understand what exactly is going on with noise -whether is the radiated noise, the ps noise, or the noise appearing on the usb data lines, which makes the most difference soundwise.
 
- In fact is worse because you lose the BIOS control and complicate the things.
- The only thing that can affect the audio are the power circuit of the CPU, chokes oscillate at 300-400Khz, but your MB is a really good one

In my opinion don't change anything related with the MB.
Thanks for telling me that, I never considered that if I did power my RAM with a linear supply I'd lose BIOS control. You're absolutely right. Hmm, I'll have to think about it now. Thanks :)

In fact you compare different (IC) DAC's with different output stages.
For an exact comparison you must use "exactly" the same DAC, the only difference should be how you isolate the connection (ground loop) or the power supply type, but not both of them simultaneously changed.
I agree, it only makes sense to change one at a time, or else you won't be able to accurately tell what made the difference.

And actually, what I said was that I hadn't tried it before, but some other people had. When they were talking about it, I assumed, they kept their system the same and only changed the USB->DAC connection.

In that case they first listened to
1. standard motherboard USB
and then
2. Adnaco SB3,

Then they tried
1. standard motherboard USB
and then
2. Specialty Audio USB card

And finally
1. Specialty Audio USB card
and then
2. Adnaco SB3.

Their results were:
Best: Specialty Audio USB card
Better: Adnaco SB3
Normal: standard motherboard USB

So, I haven't tried this myself, but I do agree than logically the optical isolation of USB should be a good thing. From what I've read it is a good thing, but not the best thing. I have since bought a Specialty Audio USB card, so that's the path I'm going down now. No optical for me, unless someone I know buys a unit and lets me test it out in my system :)

Absolutely! RF radiation off a competently designed MB is not a serious issue, where the pico/nano appear to radiate a lot worse than even a standard ATX supply.
Yeah, people who've really given the PICO a going over haven't touted its sonic attributes. It is damn convenient, though, isn't it?! I REALLY wish I could use it, it would make life A LOT easier, but it's just not in the cards.

I don't pretend to understand what exactly is going on with noise -whether is the radiated noise, the ps noise, or the noise appearing on the usb data lines, which makes the most difference soundwise.
Me too, I've got not idea, and I haven't had much opportunity until now to start trying it out myself. I can only go off what others have said here and there and then hope connect a dot or two. Let's hope this road I'm going down now teaches me a bit about the topic.


How about those transformers, though? Have any of you guys done anything with the AMB - Sigma11 or Sigma22 boards?
 
So, nobody has any experience with making or buying some kind of voltage timing board? One that could supply certain voltages in a certain sequence?

And no one out there has any experience with the AMB σ11 or σ22 boards? Or picking what kind of toroidal to go with for a certain voltage?

Or how about recommended toroidal transformer manufacturers?
 
Moderator
Joined 2002
Paid Member
FWIW I did this on a much smaller scale, using a linear supply for a bedroom audio PC based on a Cedar Trail processor. I think they're called Cedar Trail because they're made of wood chips, they're that slow!

Anywhoo, it made a mild difference, but maybe that's because the Coolermaster 'universal' laptop supply I bought for it was really junk. This is one of those 19V motherboards referenced to earlier.
 
BonkM, I have done this and wouldn't look back. To the cynics here I will say that I was with you until I tried it and now I'm no longer a cynic. Make sure you have low wattage CPUs and it's a lot easier if you have a motherboard that is not to fussy about sticking to the ATX spec power supply. I use an Intel DH61DL and this doesn't need any of the ATX control voltages. Literally supply +12, +5 and +3.3 to the appropriate ATX MB connections and ground the power button header connector to turn it on. I've got a wapping great toroid that supplies the voltages and I used a PSU circuit that I found on audio asylum to build the different voltage rails.

Once you have that working have a play with underclocking the ram and CPU. Some people have reported that you will find a sweet spot (ie sounds best) when playing with reducing the voltages - go to far and it detracts from the sound...before not booting at all ;-). Once you're happy with it all then check what voltage the RAM is using in the BIOS and then look at the MB to find out where the RAM supply comes from. Then power this separately. I haven't bothered to do this with my current set up - it's icing on the cake stuff - but worth it if you've done everything else.

Instead of the RAM you can try other things:

1) Get rid of all fans, use passive heatsink
2) Get rid of hard disk and replace with ssd, CF or apparently OS on a USB stick is good but I haven't tried it
3) Power SSD and the USB that you use for 'audio out' separately. Who knows why but the quality of power to the OS SSD changes the sound a lot...yes, I know how this is starting to sound(!)

I have gone from linear supplies to now feeding the OS SSD and the USB card using lifepo4 power. Best I've had yet.

Good luck!
 
I tried the Anker battery packs with a USB socket. Very easy to get 5v out of them by chopping up a USB lead. This was an improvement over a decent linear supply and very strangely using this power on the OS SSD had the most benefit. I then used various bunk-boost circuits (IIRC) and a few regulators to see if I could get 5v out of what are widely recognised as the best for audio lifepo4's:

Lithium Ion Cells | Cylindrical Cells | 26650 Lithium Cells

The best sounding - if probably not the best for the cells - was to just wire a decent 5v linear supply in parallel with 2 x cells in series, and the load. The cells need to be discharged to around 2.6v each. So that you end up with 5v from the linear supply and about 5v from the cells. Seems to work fine for the moment and none of the hassle charging up either. I'm still testing all this though!
 
10A for the 12V rail? That's a lot. I have an ASRock Z87E-ITX with Haswell 4570T processor. I've measured the current on the 12V linear supply I have feeding the PicoPSU. Playing music including multichannel SACD it uses 1.1A. Playing video with JRiver set to Red October HQ it consumes 2.2A. Obviously if you want to add a stonking great video card to the setup it will consume more power but I haven't found a need to yet.

I was asking a similar question re quality toroids the other day. Somewhere on this site is a list of recommended manufacturers (Google DIYaudio recommended toroid transformers). Toroidy in Poland is recommended. Getting the required VAC is straightforward - just leave yourself plenty of headroom.

Have you decided what rectifier diodes you are going to use? Snubber on the secondaries and how to optimise? Which smoothing caps? Regulator?
 
It is not a good sounding solution at all. Something similar to the Pico (Nano) prompted me to build a linear supply. Admittedly, for a very low power i3, but it made a very substantial difference soundwise. There was a thread about it somewhere...

Didn't you find that the greatest gain you experienced was from supplying the PicoPSu with clean 12V and that the next step into a full linear PSU was a modest further gain?
 
The for toroidal transformers for the power, I'm still trying to figure it out, but have so far found something like this:

Transformer #1.
30VCT (15v-0-15v) @ 10A (from Primary)
10V in parallel @ 7A or 5VDC (from Secondary)

Transformer #2.
9V in parallel @ 5A or 3.3VDC (from Primary)
8V in parallel @ 3A or 1.65VDC (from Secondary)

I'm still trying to understand the actual Primary and Secondary output part, but I've contacted a custom toroidal manufacturer and explained what I'm looking for, so hopefully he can correct any misconceptions I have. Feel free to do so here, though, in case you see something out of place.

This scares the hell out of me! If you don't understand the difference between the primary side of the transformer and the secondary I would not be messing about until you do.

A useful resource can be found here:

DIY Audio Articles
 
-12V .. for the devices what need simetrical power supply , like onboard audio ... very low current no more than 1A
5vsb - that is the standby rail ... usually power up the USB and some other devices like ethernet ... arround 3.5A

You need to bypass two signal to get working the computer with your DIY PSU:

>"power good signal" grey cable that signal is generated by the monitor IC of the PSU, is a 5V signal with is measure by the motherboard ... if the signal lasts less than 500ms means that the PSU is OK ( Temperature, voltajes, current ect ...)
faq04.gif


>PS_ON the green cable .. 5v is a signal that the motherboard connected to the GND when you push the POWER UP BUTTON and turns on the PSU

But you can remove a IC monitor chip from a cheapo PSU and use it or buy one

Example: http://www.fairchildsemi.com/ds/FA/FAN7680.pdf.

Lexer98, given the FAN7680 is discontinued, presumably this could be used in its place?

http://uk.mouser.com/ProductDetail/Texas-Instruments/TPS3510PE4/?qs=sGAEpiMZZMunEhqKs81nFOCeV9XTAN0H74BSfCHXZRA%3d

I presume one can ignore -12V for an all-digital device?
 
Last edited:
SGK,
According to the transformer section of that very handy guide,

"When an electric current is passed through a coil of wire, a magnetic field is created - this works with AC or DC, but with DC, the magnetic field is obviously static. For this reason, transformers cannot be used directly with DC, for although a magnetic field exists, it must be changing to induce a voltage into another coil."

So then how is it possible for a battery to power a transformer, if the battery is DC? I was sorta thinking of this:
Full Spectrum P1 Lithium Iron Phosphate (LiFePO4) Battery | SEPBatteries.com

It's a nanophosphate LiFePO4 12V battery. :D
 
A battery cannot be used directly with a transformer...It is impossible in fact as a battery is DC.....
You would have to convert the dc to ac to convert it back to dc.
As for powering ram with a linear supply, I would not recommend it with todays DDR memory, no benefits plenty of problems.....
 
Last edited:
A battery cannot be used directly with a transformer...It is impossible in fact as a battery is DC.....

Darn, you spoiled it - I would have loved to see the face of the guy at the electronic store when somebody asks for a DC transformer... :)

As for powering ram with a linear supply, I would not recommend it with todays DDR memory, no benefits plenty of problems.....

There is only so many times we can keep repeating that message. Facts are useless against the "linear good, switching bad" audiophile superstition...
 
There is only so many times we can keep repeating that message. Facts are useless against the "linear good, switching bad" audiophile superstition...

You've made your opinion abundantly clear on this subject across a range of fronts. Others, including some posting here who have implemented a linear PSU for this application, are very much of a different opinion.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.