Audiophile PC with true linear power supply

[bobopich]

Hi,

 

I want to introduce to you my project of audiophile PC powered by true linear psu. The PSU is the most important part of the project and was designed and build in Bulgaria it is the prototype on test board implemented into my PC but it will soon be offered as a product for sale. The result is astonishing and the sound and picture quality is by far superior compared to the same PC powered by switching mode PSU, take a look at the pics and later i will post more details and info.

 

http://bobopich.snimka.bg/radio-amateur/htpc.671317.26962420.big

[ItemAudio]

Hi,

 

I want to introduce to you my project of audiophile PC powered by true linear psu. The PSU is the most important part of the project and was designed and build in Bulgaria it is the prototype on test board implemented into my PC but it will soon be offered as a product for sale. The result is astonishing and the sound and picture quality is by far superior compared to the same PC powered by switching mode PSU, take a look at the pics and later i will post more details and info.

 

http://bobopich.snimka.bg/radio-amateur/htpc.671317.26962420.big

 

Great work! Every time I read a post like this - which, you should be aware, will be censored shortly because of it being advance promotion for a commercial product - I metaphorically punch the air. The PSU is the single biggest influence on the performance of a computer transport. It's all well and good to obsess over software and USB devices and 'isolation', and digital cabling . . . but getting past those low rent switching components is the biggest hurdle.

 

I should warn you, though, that three-way linear PSUs are a bit 2010 . . . you need four to cut the mustard in 2012!

 

Good luck.

[bobopich]

Sorry, I didn't mean to make advertise if my post is really considered like one. The PSU will most probably be sold only in Bulgaria and it is not actually developed by me but by a friend of mine. I was the lucky guy to have the test board. Unfortunatelly this is why i don't have the schematics. In general the PSU has two rails 12V and 5+3V powered by torodial transformer 230V/ 11 + 19V AC, two rectifiers and filter caps. It can supply up to 5-6A per rail but mine is with a cut off protection at 3A as the PC consumes very low power. The measurements of the rail consuptions are :

 

 

Voltage Music (A) Movie(A) Max (A) Stand by (A)

+12VDC 0,460 1,150 1,420

+5VDC 1,050 1,580 2,120

+3.3VDC 0,450 0,440 0,420

+5VDC SB 0,450 0,380 0,500 0,175

-12V - not used

 

 

The motherboard is ASUS E45M1-I Delux with 2x2GB Kingston DDR3 and a HDD. I am using a USB DAC for the audio and also connected a receiver with HDMI for the video.

 

The aim of my post was to exchange experience with someone that is also runing a linear powered PC. Why do you think the PSU should be four way ? I think it has enough power for a mini ITX based PC.

[blueixus]

I agree with you, this is so important for a computer dedicated to Audio. I am intrigued why you chose to put this into the PC box rather than house totally externally which would make it more flexible and easier to house those heat sinks!

 

Good luck!

[Julf]

I want to introduce to you my project of audiophile PC powered by true linear psu.

 

I find that interesting, as I have found a well-designed SMPS has better regulation and is much better at rejecting power supply noise than a linear one. A linear power supply *generates* less noise than a badly-designed SMPS, but is not very good at dealing with the power supply noise caused by the digital circuits in a computer.

[sandyk]

Julf

The problem with most ATX PSUs these days as I see it, are that they have a single +12V high current rail.

If you Google ATX PSU schematics for example, you are likely to find a 300W supply that has a single +12V 16A rail.

The +12V output is via a choke, and filtered by a single 1,500uF Low ESR capacitor.

The ripple is likely to be relatively high, but the problem here is that typical onboard voltage regulators

have very little rejection at the ripple frequency.

A typical voltage regulator may have only 20dB rejection at 100kHz for example,compared with 70dB or more at 100Hz.

I would be interested to see if anybody has experience of replacing a single +12V ATX supply with a dual +12V ATX supply.

Would one +12V supply be reserved for the motherboard, and the other +12V supply for ancillaries ?

Alex

[Julf]

Alex,

 

The +12V output is via a choke, and filtered by a single 1,500uF Low ESR capacitor.

 

True - that is a fairly common configuration.

 

The ripple is likely to be relatively high, but the problem here is that typical onboard voltage regulators have very little rejection at the ripple frequency. A typical voltage regulator may have only 20dB rejection at 100kHz for example,compared with 70dB or more at 100Hz.

 

I agree - but that is also why complementing those onboard linear regulators with a linear power supply won't actually improve the situation.

 

I would be interested to see if anybody has experience of replacing a single +12V ATX supply with a dual +12V ATX supply.

Would one +12V supply be reserved for the motherboard, and the other +12V supply for ancillaries ?

 

Another approach would be to simply have a separate choke/cap pair for ancillaries and motherboard. But even more interesting would be to find out what parts of the circuitry (if any) are actually sensitive (in terms of sound quality) to power supply noise.

[sandyk]

"Another approach would be to simply have a separate choke/cap pair for ancillaries and motherboard. But even more interesting would be to find out what parts of the circuitry (if any) are actually sensitive (in terms of sound quality) to power supply noise."

Julf

Agreed. However that's not entirely practical.

I have found in my own PC with a 550W ATX PSU, that even fitting a DIY C-L-C filter, or a JLH PSU Add-on in line with an internal LG BR writer, although improving rips, (let's not go there!) often leads to small glitches on playback from memory, when activities like scheduled email checks occur.

Alex

[Julf]

I have found in my own PC with a 550W ATX PSU, that even fitting a DIY C-L-C filter, or a JLH PSU Add-on in line with an internal LG BR writer, although improving rips, (let's not go there!) often leads to small glitches on playback from memory, when activities like scheduled email checks occur.

 

That sounds more like a software/interrupt priority issue rather than a power supply issue.

[sandyk]

Julf

I agree. But the glitches were not noticable before either of these were fitted. A member of another forum also mentioned increased sensitivity to "glitches" while doing other things when listening to music after fitting a JLH PSU Add-on tin line with the additional molex connector for his Asus soundcard, although he was quite happy with the improvement to Analogue Out.

Alex

[Julf]

But the glitches were not noticable before either of these were fitted. A member of another forum also mentioned increased sensitivity to "glitches" while doing other things when listening to music after fitting a JLH PSU Add-on tin line with the additional molex connector for his Asus soundcard, although he was quite happy with the improvement to Analogue Out.

 

Could it be that the PSU add-on restricts peak current so that the voltage drops too much when the CPU (or some other part) uses peak power?

[sandyk]

Julf

The JLH normally only drags around 16mA or so after initial start, which is used for modulating purposes in opposition to ripple etc. It does have a simulated capacitance of around 1 Farad though.

However, a similar problem was noticed with the C-L-C filter in line with the internal LG BR writer.

(100uF + 100nF at input, series 100uH 5A 50 milliohms, 4,700uF + 2,200uF at output in both +5V and +12V feeds.)

Regards

Alex

[manisandher]

I would be interested to see if anybody has experience of replacing a single +12V ATX supply with a dual +12V ATX supply. Would one +12V supply be reserved for the motherboard, and the other +12V supply for ancillaries ?

 

Hi Alex, I have two PSs (both switch-mode) installed in my CaseLabs W8 case, with the intention of doing just this. Just having gotten around to trying it yet, but will report back as soon as I do.

 

Nowadays, I'm quite happy to use SMPSs (as opposed to linear PSs), but feed them from a 1KW isolation transformer. My AC noise detectors detect no noise being injected back into the AC mains, which has always been my main consideration with SMPSs. My Linux machine is still powered by a top-end Paul Hynes PS, but only because I had the PS lying around anyway.

 

Mani.

[sandyk]

Hi Mani

I will be interested to see what you find. I really would love to be able to have a separate +12V supply for the M/B and fit a couple of JLH PCBs to further filter the +12V areas like both Optical Writers, and aux. Soundcard supply.

I have been told that improved aux . PSU to Asus Essence STX not only improves Analogue Out, but coax SPDIF Out as well.

Improved PSU isolation to HDD and SSD may also make both types sound the same too ?

Regards

Alex

[Wavelength]

Hi,

 

I want to introduce to you my project of audiophile PC powered by true linear psu. The PSU is the most important part of the project and was designed and build in Bulgaria it is the prototype on test board implemented into my PC but it will soon be offered as a product for sale. The result is astonishing and the sound and picture quality is by far superior compared to the same PC powered by switching mode PSU, take a look at the pics and later i will post more details and info.

 

http://bobopich.snimka.bg/radio-amateur/htpc.671317.26962420.big

 

Bob,

 

Just putting a linear supply onto a PC mother board is hardly enough. There are at least 20 DCDC converters on the motherboard it self.

 

You really need to look into having some one design this part as it is the crucial part to real success.

 

Thanks

Gordon

[Edgaronline]

Then I see a big artic 8L fan... That would surely take some of the musical joy away...

 

Anyway, in my project, I powered the motherboard with the switching PSU it was supplied with, used a separate lineair PSU for the DAC (basically the MF VPSU).

 

 

The next step I am thinking is to supply the USB/SPDIF converter with power which is not coming from the motherboard, so, a separate 5V DC supply which I want to take from the lineair 12V supply I already have...

 

Nothing for sale here.. don't want to upset Chris his sponsors...

 

projects...projects....

 

[bobopich]

I agree with you, this is so important for a computer dedicated to Audio. I am intrigued why you chose to put this into the PC box rather than house totally externally which would make it more flexible and easier to house those heat sinks!

 

Good luck!

 

Actually it is just a custom box it was not designed especially for PC and i wanted to put all toghether in order to save space. As the PSU generates a lot of heat it is not a bad idea to be put in separate box.

[bobopich]

Then I see a big artic 8L fan... That would surely take some of the musical joy away...

 

Anyway, in my project, I powered the motherboard with the switching PSU it was supplied with, used a separate lineair PSU for the DAC (basically the MF VPSU).

 

 

The next step I am thinking is to supply the USB/SPDIF converter with power which is not coming from the motherboard, so, a separate 5V DC supply which I want to take from the lineair 12V supply I already have...

 

 

 

I've put the fan just in case. These were older pictures without the FAN control i have designed. This is the new picture of the PC with the fan control : http://bobopich.snimka.bg/radio-amateur/htpc.671317.27072879.big

 

The temperature of the CPU is around 45C when in idle or listening to music and the fan is not working at that point. It switches on only when the temp gets to 60C at 5V with it's own regulators and at 70C at 12V and goes down with hysteresis . I can even adjust the exact temperatures with the trimmer. So basically i have two speeds for the fan and when listening to music it is not working at all. Sometimes it switches on when watching a movie or doing some intensive CPU job.

 

I am also thinking of supplying the USB/SPDIF converter with separate 5V from the linear rail to check if there will be any audiable difference but accoring to my measurments the MB is supplyiing the same 5V to the USB port that comes from the linear PSU so they are basicly clean. But i will make the test anyway.

[bobopich]

I find that interesting, as I have found a well-designed SMPS has better regulation and is much better at rejecting power supply noise than a linear one. A linear power supply *generates* less noise than a badly-designed SMPS, but is not very good at dealing with the power supply noise caused by the digital circuits in a computer.

 

I partly agree with you but what about a well-designed linear regulated PSU will a well-designed SMPS still be that good. The PSU in my PC has stable voltages that has 0.0001% shift no matter of the load up to the allowed current cut off.

[bobopich]

The PSU has direct voltage tracking of the rails which controlls the linear stabilizers themself. No digital or any other kind of "noise" could affect the clear voltage of the PSU.

[sandyk]

"The PSU in my PC has stable voltages that has 0.0001% shift no matter of the load up to the allowed current cut off. "

 

While I definitely support the use of Linear PSUs for at least certain areas of a computer, I am having a great deal of difficulty taking the above claims at face value.With a regulated +12V supply, a shift of only 1.2mV from no load to maximum load, over the wide range of temperatures involved, and ambient temperatures as well, would be accepted by most as an extremely good result.

However, unless my quick stab at the maths is very wrong, then the claim here is for a change of only .0012 Millivolts

change under these extreme conditions. I doubt that there would be too many laboratory standard PSUs that could claim to meet those specifications.

Alex

[bobopich]

"The PSU in my PC has stable voltages that has 0.0001% shift no matter of the load up to the allowed current cut off. "

 

While I definitely support the use of Linear PSUs for at least certain areas of a computer, I am having a great deal of difficulty taking the above claims at face value.With a regulated +12V supply, a shift of only 1.2mV from no load to maximum load, over the wide range of temperatures involved, and ambient temperatures as well, would be accepted by most as an extremely good result.

However, unless my quick stab at the maths is very wrong, then the claim here is for a change of only .0012 Millivolts

change under these extreme conditions. I doubt that there would be too many laboratory standard PSUs that could claim to meet those specifications.

Alex

 

 

Hi Alex,

 

Yes it is my mistake i should have written accuracy 0.0001V not % , i.e. 0.1mV but it is still within extremely good specs. I can request for a block diagram of the PSU if you're interested but i cannot take the schematics due to the reason i explained above.

[phofman]

The PSU has direct voltage tracking of the rails which controlls the linear stabilizers themself. No digital or any other kind of "noise" could affect the clear voltage of the PSU.

 

Well, every control feedback has a delay

[1audio]

Hi Alex,

 

Yes it is my mistake i should have written accuracy 0.0001V not % , i.e. 0.1mV but it is still within extremely good specs. I can request for a block diagram of the PSU if you're interested but i cannot take the schematics due to the reason i explained above.

 

You can't get .1 mV without remote sensing. Even with remote sensing 1 inch from the connection you will get more voltage variation. In the grand old days we used linear supplies for computers. The were heavy and had noisy fans. We had remote sensing or they would not work. Today's computers have much more stringent requirements for power and linear supplies won't work ultimately. The motherboard has on board "point of load" switching supplies that regulate down to the actual voltages used by the chips. A modern processor or memory module will run at 1.1V (or at least the i7 I'm using) and draw up to 100 W. That's 100 A. 10 AWG wire is 1 milliOhm per foot. At 100A that drop would be 100 mV for 1 foot of wire. 1000 times the spec above. From a purpose built supply that would be probably 200W dissipation in the supply to provide the 100W. Most commercial power amps would barely be able to handle this much heat, and they have the heat sources less concentrated.

 

I think Mani is right for the most part. Use a very good isolation scheme between the ac source and the computer and supply sensitive internal circuits from isolated low noise supplies. And per Gordon's remark, not easily done by the DIY'er since it takes a lot of research to figure out what to isolate and how to isolate it. Without a systems approach the supply isolation may be obviated by the radiation of noise from the power wires between the supply and the PC for example.

[One and a half]

Well, there's not we can really do as far as audio enthusiasts to fiddle with the interiors of a PC. The amount of noise generated varies by motherboard designs and the case/power supply. To cater for all combinations will be astronomically difficult.

 

What we can do is 'isolate' the PC from all sources of supply and the signal, the PC becomes an island electrically so to speak, so the network can't see the PC as far as noise is concerned. With the right filtering and experiments that can be achieved through either DIY (keeping in mind live mains voltages which can kill you) or designs available commercially.

 

The signal end for the USB bus, Audiophilleo have a great battery isolation, Red Wine Audio works off battery as well, there are ways and means of getting the PC to be an island. If someone can come up with a DSD/PCM768 pass-through jitter immune optical system, then that's a walk in the park.

 

BTW, PC= Windows, Linux & OSX