High quality power supplies for single board computers

[nbpf]

With the introduction of transports like the Allo DigiOne and USBridge, single board computers requiring power supplies able to deliver about 3A at 5VDC or less are becoming increasingly popular as stand-alone music server or as S/PDIF and USB network renderers. Some contributors to these forums have argued that improvements in sound quality that can be achieved by powering single board computers with high quality power supplies. What are suitable high quality power supply for single board computers? Has anybody compared UpTone Audio's JS-2, Custom HiFi Cables' DC3, Paul Hynes power supplies, etc. to less expensive wall wart power supplies like the ifi iPower and the Allo 5V/3A? If so, please, report your findings!

[tapatrick]

I recommend Paul Hynes SR4, an excellent PSU for the price which uses much of the technology of his SR7 and I didn't have to wait long for delivery. I have 2 which I use for SBCs.

[nbpf]

53 minutes ago, tapatrick said:

I recommend Paul Hynes SR4, an excellent PSU for the price which uses much of the technology of his SR7 and I didn't have to wait long for delivery. I have 2 which I use for SBCs.

Thanks tapatrick! Have you noticed any improvements against other PSUs when powering your SBCs with the SR4s?  

[tapatrick]

7 hours ago, nbpf said:

Thanks tapatrick! Have you noticed any improvements against other PSUs when powering your SBCs with the SR4s?  

I haven't specifically compared different PSUs running SBCs but I had run dual in series LT3045 voltage regulators after a modified Teradak DC30 and also a LPS1which improved sound significantly. When I got the SR4 I found I did not need the LT3045s afterwards as they did not make any further improvement which convinced me of its quality.

[nbpf]

5 hours ago, tapatrick said:

I haven't specifically compared different PSUs running SBCs but I had run dual in series LT3045 voltage regulators after a modified Teradak DC30 and also a LPS1which improved sound significantly. When I got the SR4 I found I did not need the LT3045s afterwards as they did not make any further improvement which convinced me of its quality.

Very nice, thanks for your assessment!

[barrows]

I can make power supplies optimized specifically for these types of loads, with standard leading low noise and output impedance.

But, I would caution folks to understand that one is never going to get the best performance for audio by using the commercial computer gears suggested here.  Even with an ultra low noise pre-regulated supply, these boards are going to have noisy DC/DC switching regulators onboard, so they are going to make a lot of noise on the board.

Higher performance solutions are available, built for audio purposes, which do not use noisy DC/DC converters.

[nbpf]

39 minutes ago, barrows said:

I can make power supplies optimized specifically for these types of loads, with standard leading low noise and output impedance....

Thanks, but the focus here is to compare existing high quality power supplies (JS-2, DC3, etc.) to less expensive wall wart solutions for SBCs.

[nbpf]

42 minutes ago, barrows said:

...

But, I would caution folks to understand that one is never going to get the best performance for audio by using the commercial computer gears suggested here.  Even with an ultra low noise pre-regulated supply, these boards are going to have noisy DC/DC switching regulators onboard, so they are going to make a lot of noise on the board.

Higher performance solutions are available, built for audio purposes, which do not use noisy DC/DC converters.

Good point but I am not aware of currently available S/PDIF server or network renderer solutions that are designed to match or exceed the performance of the DigiOne. If I am overseeing some products, please let me know! Of course, we do have excellent network renderers and, lately, servers with USB outputs. But these devices need to be complemented with USB to S/PDIF interfaces in order to deliver S/PDIF streams. The quality of the results depends then crucially on the quality of the interface and of its power supply! Other major limitations of currently available devices built for audio purposes are their very limited customizability and the lack of certifications for publicly available operating systems: many users would not mind investing in high performance servers built for audio purposes if they knew that, beside running the manufacturer's proprietary OS, these devices can also boot a standard Linux distribution. For users that need S/PDIF outputs and/or some control and customization options, commercial SBCs and dedicated interfaces like the ones mentioned in my original post are, if not optimal, at least viable solutions.

[rikhav]

10 hours ago, barrows said:

I can make power supplies optimized specifically for these types of loads, with standard leading low noise and output impedance.

But, I would caution folks to understand that one is never going to get the best performance for audio by using the commercial computer gears suggested here.  Even with an ultra low noise pre-regulated supply, these boards are going to have noisy DC/DC switching regulators onboard, so they are going to make a lot of noise on the board.

Higher performance solutions are available, built for audio purposes, which do not use noisy DC/DC converters.

 

Do laptops have same noisy dc/dc converters ?

If yes then what would kind of computer would be best for audio for someone who can't spend big bucks on music servers or likes of it 

[barrows]

1 hour ago, rikhav said:

 

Do laptops have same noisy dc/dc converters ?

If yes then what would kind of computer would be best for audio for someone who can't spend big bucks on music servers or likes of it 

Yes.  All commercial computers are built to compete in very competitive markets and to the lowest possible price point.  And all computers need various voltages at the main board level.  To get these different voltages they use switching DC/DC converters (regulators) at the board level.  this is especially the case wit laptops, as switching converters are more efficient than linear ones, and battery life is a very big consideration for laptop makers.  The commercial computer companies will do whatever is most affordable and still able to work (and pass EMI emissions standards), they have no reason to use more expensive, ultra low noise linear regulators as they are not building computer products for the bets possible audio quality.

 

This is why we have high end music servers and Ethernet Renderers, to overcome the limitations of standard commercial computers.  (But even most "high end" servers still use an off the shelf motherboard with noisy DC/DC converters on board as it is too expensive for most high companies to invest enough R and D into developing their own MoBo).

 

The most affordable solution to get the best possible sound at the lowest price would be to use an affordable Ethernet Renderer (like the Sonore microRendu) combined with ordinary commercial computer gear you already own.  the microRendu then provides a good deal of noise isolation from the computer and feeds the DAC a very low noise, very clean USB stream.  Be sure to place the noisy commercial computer gear well away (in another room) from the audio gear.

 

Putting an expensive power supply on a commercial computer product can help, but it is a kind of backwards approach, like putting a band aid on a fractured leg.

[barrows]

11 hours ago, nbpf said:

Good point but I am not aware of currently available S/PDIF server or network renderer solutions that are designed to match or exceed the performance of the DigiOne. If I am overseeing some products, please let me know! Of course, we do have excellent network renderers and, lately, servers with USB outputs. But these devices need to be complemented with USB to S/PDIF interfaces in order to deliver S/PDIF streams. The quality of the results depends then crucially on the quality of the interface and of its power supply! Other major limitations of currently available devices built for audio purposes are their very limited customizability and the lack of certifications for publicly available operating systems: many users would not mind investing in high performance servers built for audio purposes if they knew that, beside running the manufacturer's proprietary OS, these devices can also boot a standard Linux distribution. For users that need S/PDIF outputs and/or some control and customization options, commercial SBCs and dedicated interfaces like the ones mentioned in my original post are, if not optimal, at least viable solutions.

There are High End Ethernet Renderers which include SPIDF output: Like Bricasti and dCS, and Sonore used to make the Sonore Signature Rendu which is a superb SPDIF source if you can find one on the used market.  Both of these will easily outperform ordinary commercial computer solutions like Allo, etc.

 

Additionally, SPDIF basically sucks, but that is another topic...  It is too bad that some DACs do not put a well engineered USB input in their DAC... especially now as the way to do it is fairly well understood.

[nbpf]

22 minutes ago, barrows said:

... Additionally, SPDIF basically sucks, but that is another topic...  It is too bad that some DACs do not put a well engineered USB input in their DAC... especially now as the way to do it is fairly well understood.

Agree but ... what can you do? There are some good old DACs that have S/PDIF inputs ...

[barrows]

7 minutes ago, nbpf said:

Agree but ... what can you do? There are some good old DACs that have S/PDIF inputs ...

Agreed, if you must use SPDIF then use one of the Renderers which has a SPDIF output, or use a good USB-SPDIF converter.  The converter, if well designed, offers the additional advantage of another stage of noise isolation (for example the Sonore ultraDigital is isolated between USB receiver section and SPIDF output), and the converter also performs a s a re-clocker.

[chauphuong]

Is it true that Sotm SMS 200 still uses switching dc-dc converter?

[barrows]

2 minutes ago, chauphuong said:

Is it true that Sotm SMS 200 still uses switching dc-dc converter?

You would have to ask SOtM about that...

[nbpf]

25 minutes ago, barrows said:

There are High End Ethernet Renderers which include SPIDF output: Like Bricasti and dCS, and Sonore used to make the Sonore Signature Rendu which is a superb SPDIF source if you can find one on the used market.  Both of these will easily outperform ordinary commercial computer solutions like Allo, etc.

Perhaps yes but I am more interested in solutions that can act as integrated server and renderer. Maybe the new Antipodes CX or the announced Innuos Signature will be interesting devices. For the time being, I have to rely on off-the-shelf headless computers with dedicated Linux distributions and USB to S/PDIF or, better I2S to S/PDIF interfaces. By the way, the DigiOne is actually not bad, as far as I can say. And over at the Naim forums, a few users have been testing DigiOnes to stream Qobuz and Tidal to their NDS streamers and reported good results. It is not an NDS, of course. But it's quite good!     

[nbpf]

13 minutes ago, barrows said:

Agreed, if you must use SPDIF then use one of the Renderers which has a SPDIF output, or use a good USB-SPDIF converter.  The converter, if well designed, offers the additional advantage of another stage of noise isolation (for example the Sonore ultraDigital is isolated between USB receiver section and SPIDF output), and the converter also performs a s a re-clocker.

Yes, I have tried many USB-SPDIF converters, including the Mutec MC-3+ USB and the Schiit Eitr. To tell you the truth, I wa not impressed by the Mutec. I still have an Eitr connected to the RPi that hosts the DigiOne. Thus, I can easily compare the two by just switching inputs on the Naim DAC. The DigiOne also provides galvanic isolation and reclocking. I am not sure that the latter is very important in my case: the old Naim DAC apparently overrides the clocking of the SPDIF stream with its internal clocks before this enters the DSP and DAC stages. Alas, just put a memory slot and a wireless board on the top of the ultraRendu and we do not have to play around with off-the-shelf computers any longer!

[barrows]

58 minutes ago, nbpf said:

DAC apparently overrides the clocking of the SPDIF stream with its internal clocks before this enters the DSP and DAC stages

I doubt that is the case with your DAC.  Even if it internally resamples the data to the DAC clock, it still has to reconcile the DAC clock with the SPDIF clock.  Typically this is done with a PLL, and these circuits effectively couple the incoming SPDIF jitter to the DAC clock.  This is different form async USB, where the DAC clock can be the true master and does not have to sync with any incoming clock (because async USB is two way, and the rate the data is being sent can be adjusted by the DAC).

 

There are some DACs with fully async SPDIF receivers, but they are very rare.

[Ralf11]

curious, so I started a new topic:

 

 

[nbpf]

59 minutes ago, barrows said:

I doubt that is the case with your DAC.  Even if it internally resamples the data to the DAC clock, it still has to reconcile the DAC clock with the SPDIF clock.  Typically this is done with a PLL, and these circuits effectively couple the incoming SPDIF jitter to the DAC clock.  This is different form async USB, where the DAC clock can be the true master and does not have to sync with any incoming clock (because async USB is two way, and the rate the data is being sent can be adjusted by the DAC).

 

There are some DACs with fully async SPDIF receivers, but they are very rare.

From https://www.naimaudio.com/sites/default/files/products/downloads/files/naim_dac_august_2009.pdf I understand that PLL is only used to select one of the internal clocks. This is then used to clock the buffered SPDIF data for the DSP and DAC sections. I might be missing something, of course.

 

I also have an old M2Tech hiFace Evo that, if I am not mistaken, uses asynchronous USB. The M2Tech is connected to an headless CompuLab fitPC3 running a minimal Debian system. Both the M2Tech and the fitPC3 are powered by Teddy Pardo LPSUs. This has been my default system for about five years. Now I have the impression that the DigiOne sounds a little bit better than the M2Tech. 

[Qhwoeprktiyns]

7 hours ago, barrows said:

Tmost affordable solution to get the best possible sound at the lowest price would be to use an affordable Ethernet Renderer (like the Sonore microRendu) combined with ordinary commercial computer gear you already own.  the microRendu then provides a good deal of noise isolation from the computer and feeds the DAC a very low noise, very clean USB stream.  Be sure to place the noisy commercial computer gear well away (in another room) from the audio gear.

 

If the DAC is not using the 5v does that change anything? 

[barrows]

7 minutes ago, hopkins said:

 

If the DAC is not using the 5v does that change anything? 

No.  Having a really clean, low noise, USB data stream is still just as important.

[Qhwoeprktiyns]

3 minutes ago, barrows said:

No.  Having a really clean, low noise, USB data stream is still just as important.

 

Thanks for your reply. I would assume having a self-powered DAC would probably still help (assuming the power supply within the DAC is good). 

 

One more question, aside for the voltage regulators within the PC doesn't the processors themselves generate noise? In that sense, why would a Microrendu, or any SBC, behave any differently than a regular Mac/PC? 

[barrows]

3 minutes ago, hopkins said:

 

Thanks for your reply. I would assume having a self-powered DAC would probably still help (assuming the power supply within the DAC is good). 

 

One more question, aside for the voltage regulators within the PC doesn't the processors themselves generate noise? In that sense, why would a Microrendu, or any SBC, behave any differently than a regular Mac/PC? 

Yes, absolutely, the processors generate a lot of high frequency noise.  Also, the more powerful the processor, the more noise they make.  In the case of the Sonore Rendu products, the processor is chosen such that it has the lowest possible noise footprint (by using a specially selected processor and by using one with just enough power to do the job).  A general purpose computer will use a much more powerful processor, as it has to be able to do much more than what is required just to serve music files from Ethernet.  So in a purpose built design you can tailor the processor to be just sufficient for the task at hand.  Additionally, by taking care of all the power supply details (using ultra low noise linear regulators and multiple power supply rails) the processor noise is kept contained (for the most part) and not allowed to propagate to the rest of the circuitry (example: each section of the circuit has its own dedicated ultra low noise linear regulator, providing circuit isolation from the different sections).  As commercial computer builders are not building in order to produce high end sound, they only do the minimum necessary to make a computer work and pass EMI standards.

[Qhwoeprktiyns]

Thanks again. I assume these improvements should translate into better SQ. There is a lot of processing going on within the DAC as well, but that is another subject. 

 

One side of things that I find frustrating in computer audio is the reliance on Alsa, for Linux systems at least. It is a black box, and I suspect there would be some benefit in designing better software for USB audio without relying on alsa. Perhaps an operating system designed from the ground up for audio purposes would be something to consider..