Amir's Music Server Build Thread

[Sal1950]

Very nice box Amir! I've been looking at building a new totally silent box myself and this has given me some pointers though I'd like to spend a little less $ this time around. Checking into some less expensive options for a passive cooled CPU.
il saluto

 

[March Audio]

Nice PC!. I used to build a lot of PCs myself, however for some years now I find laptops do everything I need. As it is switched on most of the time, I do find the lower power consumption of laptops beneficial.

Low end Acer, about $350 AUD IIRC manages Roon core and Jriver for video serving just fine.

 

[Fitzcaraldo215]

I agree, nice build, Amir.

Much of the specialization of computer audio starts with the fan noise problem in the room. From there, many have tried to insist that these silent, audiophile PC's also sound way better than ordinary PC's, heaping much folklore on top of that. I do not think that is true.

My own solution to the fan noise problem has just been to build a powerful workhorse I7 tower, an ordinary one with fans, and put it an adjoining room. Fortunately, that is a small utility room. Then, I run 5 meter USB and HDMI cables to my DAC and TV monitor in the main room through a small hole in the back of a closet. Cost: around $1,000.

It works great, looks great and sounds great. To load optical disks, which I do only occasionally, I just have a short walk to the next room. No big deal.

 

[Mivera]

Amir uses a competent USB to AES/EBU bridge so the noise from the SMPS won't be as much of an issue for him.

 

[Fitzcaraldo215]

Yes, LPS is the "party line", reinforced by gazillions of computer audiophile anecdotal testimonials, all in totally uncontrolled subjective listening. Most of them read something like: "I replaced my SMPS with an LPS, and it blew my socks off". We are unclear on how much time elapsed between hearing each one while the new power supply was being installed and anything else about the comparison. But, any truly objective proof, measurably or otherwise and bias controlled, that LPS inherently sounds better than SMPS, particularly in audio digitally emanating from a computer or even the output from a DAC, seems to be totally lacking. Or, am I missing something?

Note that much Pro audio gear used in making recordings uses SMPS. Also, note that Soulution, the Swiss mega amp makers, recently "upgraded" their top monoblocks to SMPS from LPS. Also, note that Bruno Putseys' highly praised Class D amp designs use SMPS.

Footnote: for those unfamiliar with the acronyms, LPS = Linear Power Supply, essentially analog, SMPS = Switch Mode Power Supply, essentially digital.

 

[amirm]

Even stranger is buying a linear power supply, then feeding it to a module inside the PC that generates all the various voltages required using switchmode power supply! They sell the idea as "creating a moat" around the noisy things in the PC. Yet many folks in audio world worry about what goes back through their linear power supply in their amps and such, by investing in cables, outlets, etc.

 

[Mivera]

Yes, LPS is the "party line", reinforced by gazillions of computer audiophile anecdotal testimonials, all in totally uncontrolled subjective listening. Most of them read something like: "I replaced my SMPS with an LPS, and it blew my socks off". We are unclear on how much time elapsed between hearing each one while the new power supply was being installed and anything else about the comparison. But, any truly objective proof, measurably or otherwise and bias controlled, that LPS inherently sounds better than SMPS, particularly in audio digitally emanating from a computer or even the output from a DAC, seems to be totally lacking. Or, am I missing something?

Note that much Pro audio gear used in making recordings uses SMPS. Also, note that Soulution, the Swiss mega amp makers, recently "upgraded" their top monoblocks to SMPS from LPS. Also, note that Bruno Putseys' highly praised Class D amp designs use SMPS.

Footnote: for those unfamiliar with the acronyms, LPS = Linear Power Supply, essentially analog, SMPS = Switch Mode Power Supply, essentially digital.

Yes because their hardware is specifically designed for very high PSRR with high end audio grade LDO's and gain stages. General purpose computer gear is not designed this way. However a proper digital interface for a DAC does reject the switching noise. Only problem is most aren't designed this way.

 

[Mivera]

Even stranger is buying a linear power supply, then feeding it to a module inside the PC that generates all the various voltages required using switchmode power supply! They sell the idea as "creating a moat" around the noisy things in the PC. Yet many folks in audio world worry about what goes back through their linear power supply in their amps and such, by investing in cables, outlets, etc.

Compare the USB to AES bridge I sent you with your Berkeley. If computer noise is a non issue, they should both sound and measure identical. With a large footprint noisy motherboard with high powered CPU like you used, it probably makes less of a difference using an LPS than it does with a small low noise footprint board like the NUC board I sent you.

 

[Fitzcaraldo215]

Yes because their hardware is specifically designed for very high PSRR with high end audio grade LDO's and gain stages. General purpose computer gear is not designed this way. However a proper digital interface for a DAC does reject the switching noise. Only problem is most aren't designed this way.

Evidence of audibility?

 

[Mivera]

Evidence of audibility?

Amir has the gear and tools to obtain it.

 

[amirm]

Compare the USB to AES bridge I sent you with your Berkeley. If computer noise is a non issue, they should both sound and measure identical.

Why? The accuracy of the internal clock oscillator still matters. As does the implementation of galvanic isolation.

 

[Mivera]

Why? The accuracy of the internal clock oscillator still matters. As does the implementation of galvanic isolation.

Yeah but those should measure similar with any computer connected. Try the NUC powered by clean LPS on both bridges vs your server connected to both bridges. Connect the AP to the AES/EBU outs and run the full gamut of tests

The galvanic isolation is one of the measures only some USB interfaces have. It's an extra measure to help battle server demons. The bridge I sent doesn't have it.

 

[amirm]

Amir has the gear and tools to obtain it.

I can measure but don't have evidence of audibility. Here are two async USB interface measurements. The DAC is Peachtree and I either use its internal USB implementation or Berkeley USB to S/SPDIF:

The Berkeley in blue has less correlated noise than USB input of the Peachtree. But both are at very low levels and below audibility. Here is the zoomed version:

We are talking about peaks around -120 dBFS. While the Berkeley is better engineered, I can't with a straight face say there is an audible difference.

 

[amirm]

Here is another article on my comparison of two async USB interfaces ,the Berkeley and Audiophilleo: http://www.audiosciencereview.com/f...nce-pc-server-interfaces-async-usb.8/#post-10

The less performant Audiophilleo in red still has distortions below -110 dBFS.

 

[Mivera]

I can measure but don't have evidence of audibility. Here are two async USB interface measurements. The DAC is Peachtree and I either use its internal USB implementation or Berkeley USB to S/SPDIF:

View attachment 1653
The Berkeley in blue has less correlated noise than USB input of the Peachtree. But both are at very low levels and below audibility. Here is the zoomed version:

View attachment 1654

We are talking about peaks around -120 dB SPL. While the Berkeley is better engineered, I can't with a straight face say there is an audible difference.

If you do the test I suggested, it should be clear if your server introduces more noise into the system vs the NUC powered by clean LPS. It will also prove that the noise isolation measures in the Berkeley are good. And if the noise from your server combined with the cheap bridge is below the threshold of audibility, it will prove that there's no need to spend that kind of money on a bridge if a $200 bridge can do just as good of a job.

 

[amirm]

OK I see that. Thanks.

 

[Mivera]

Here is another article on my comparison of two async USB interfaces ,the Berkeley and Audiophilleo: http://www.audiosciencereview.com/f...nce-pc-server-interfaces-async-usb.8/#post-10

The less performant Audiophilleo in red still has distortions below -110 dBFS.

Looks to me that test proves server noise is a real issue.

 

[Mivera]

OK I see that. Thanks.

Since the Roon core can run perfectly fine on a $145 fanless NUC, perhaps if you have a DAC with a USB interface worth its salt, maybe that's all that's required to get the highest performance out of Roon. Especially if the small footprint Linux version is used that dedicates the OS specifically to Roon and runs no GUI.

 

[amirm]

Looks to me that test proves server noise is a real issue.

That's not the case because with a different DAC the correlated noise/distortion is different.

Here is an AVR comparing its HDMI input to Berkeley over S/PDIF both driven from my laptop:

Two very different responses.

I actually tried to induce noise from the laptop by creating graphics activity but I could not.

 

[Mivera]

That's not the case because with a different DAC the correlated noise/distortion is different.

Here is an AVR comparing its HDMI input to Berkeley over S/PDIF both driven from my laptop:
View attachment 1655

Two very different responses.

I actually tried to induce noise from the laptop by creating graphics activity but I could not.

It looks to me like the Berkeley is doing a better job.