A novel way to massively improve the SQ of computer audio streaming

[bobfa]

1 hour ago, austinpop said:

 

Good questions. Let's start with the defaults. I did some poking around. Going by this file:  https://github.com/ralph-irving/squeezelite/blob/master/squeezelite.h

• #define STREAMBUF_SIZE (2 * 1024 * 1024)
• #define OUTPUTBUF_SIZE (44100 * 8 * 10)
• #define ALSA_BUFFER_TIME  40
• #define ALSA_PERIOD_COUNT 4

This suggests the defaults are:

• IB (inputstream buffer) is 2MB
• OB (output buffer) is ~3MB
• -a 40:4::

Note that the first value in the -a option has a change in units at 500. For 0-499 it has units of ms, and 500 or greater, it's in bytes.

 

So back to your question, Bob. We know that the high settings sound great, but are glitchy:

• input buffer IB = 2GB
• output buffer OB = 2GB
• ALSA buffer AB = 50MB

We know the defaults have lower SQ but no glitches:

• input buffer IB = 2MB
• output buffer OB = ~3MB
• ALSA buffer AB = 40ms

So the search space for best SQ while avoiding glitches lies between these values.

First report.  I set to the Large Buffer settings that you had suggested.  I find that the music just stops playing.  Switching between albums seems to aggravate this.  I am not ready to talk about sound quality in detail.  I am now switching back to the defaults and will see how that plays and if it is stable.  I am going to read the GitHub page and see what I can glean from getting more informed

 

[austinpop]

25 minutes ago, bobfa said:

First report.  I set to the Large Buffer settings that you had suggested.  I find that the music just stops playing.  Switching between albums seems to aggravate this.  I am not ready to talk about sound quality in detail.  I am now switching back to the defaults and will see how that plays and if it is stable.  I am going to read the GitHub page and see what I can glean from getting more informed

 

 

Thats the glitchiness at work. Sometimes when you hit play nothing happens. Stop, wait for the shiny dot next to the track name to disappear, and retry.

[bobfa]

4 hours ago, austinpop said:

Thats the glitchiness at work. Sometimes when you hit play nothing happens. Stop, wait for the shiny dot next to the track name to disappear, and retry.

We have been running for about 4 hours with SL set at defaults with no issues.  The SQ is better, my son says the bass response in his set of drum and bass tracks is better refined than before.  For me there is less high end glare and in some of the tracks I have talked about before there is again more information shown to my ear.  I will study the software some more and decide what I want to try.  I was thinking about a binary tree.  Split the difference between default and the big buffers that were causing some issues.  

 

We will see, have to clean house tomorrow first!

[ferenc]

Can be interesting to know about these:

 

https://www.networkworld.com/article/3386142/intel-unveils-an-epic-response-to-amds-server-push.html

 

"In addition to its hot rod Xeons, Intel also rolled out the Xeon D-1600 series processors, a low power variant based on a completely different architecture. Xeon D-1600 series processors are designed for space and/or power constrained environments, such as edge network devices and base stations."

[seeteeyou]

Xeon D-1500 / D-1600 / D-2100 could only be soldered-down processors since we're talking about FCBGA here, therefore we'll only get a few choices from ASRock / Gigabyte / Supermicro etc.

 

IMHO only the most expensive ones with 8 cores are competitive with the upcoming i9-9900T with 35W TDP, though a motherboard should easily cost around $1,000 since both Xeon D 1649N and 1653N are bloody expensive to begin with

 

https://ark.intel.com/content/www/us/en/ark/products/codename/190862/hewitt-lake.html

https://www.anandtech.com/show/14147/intel-launches-the-xeon-d-1600-family

 

If I were spending $705 for something with 8C16T and ECC RAM support myself, I could save a buck and get an entire motherboard instead

 

https://www.gamepc.com/shop/products?sku=M11SDV-8C-LN4F

https://www.servethehome.com/supermicro-m11sdv-8c-ln4f-review-amd-epyc-3251-mitx-platform/

 

Even Xeon D-1500 Series offered options with 12 cores and even 16 cores, it's hard to tell why they're going backwards while on average we're still 1.5MB L3 cache / core

 

https://ark.intel.com/content/www/us/en/ark/products/series/87041/intel-xeon-d-processor.html

 

Since we couldn't purchase Xeon D-1600 Series processors and motherboards separately while they're costing so much to begin with, just keep an eye on that motherboard from SOtM and see how it goes.

 

However, we still dunno if running Xeon or Core i3 with ECC RAM were actually any good for an audiophile yet. 

[bobfa]

On 4/6/2019 at 4:28 PM, austinpop said:

 

Good questions. Let's start with the defaults. I did some poking around. Going by this file:  https://github.com/ralph-irving/squeezelite/blob/master/squeezelite.h

•  

i am just starting to look at it.  And this week is not going to be kind to my time.  

 

Here is the source to the man page also.  I think I will load this up on my Mac and look at the code more closely.

 

https://github.com/ralph-irving/squeezelite/blob/master/doc/squeezelite.1

 

 

I am thinking I will just cut your buffers in half and see if it is stable.  Then do some listening again.  I have not gone back to read the section on the thread where you were testing before.  i will go do that also.

 

 

bob

 

 

[austinpop]

4 minutes ago, bobfa said:

i am just starting to look at it.  And this week is not going to be kind to my time.  

 

Here is the source to the man page also.  I think I will load this up on my Mac and look at the code more closely.

 

https://github.com/ralph-irving/squeezelite/blob/master/doc/squeezelite.1

 

 

I am thinking I will just cut your buffers in half and see if it is stable.  Then do some listening again.  I have not gone back to read the section on the thread where you were testing before.  i will go do that also.

 

 

bob

So far, I have tried

1. -b 2097152:2097152 -a 52428800:4::
2. -b 1048576:1048576 -a 26214400:4::
3. -b 524288:524288 -a 13107200:4::
4. defaults (nothing in the Euphony expert field)

 

Only 4. is glitch free. 

 

So the answer lies beyond.

 

 

[bobfa]

11 hours ago, austinpop said:

So far, I have tried

1. -b 2097152:2097152 -a 52428800:4::
2. -b 1048576:1048576 -a 26214400:4::
3. -b 524288:524288 -a 13107200:4::
4. defaults (nothing in the Euphony expert field)

 

Only 4. is glitch free. 

 

So the answer lies beyond.

 

 

We have two or three  interactions here that are possible. Is Roon doing something to cause issues from the server.  I suppose we could find a LMS to try to eliminate that.  Second is the code not written to handle the larger buffers or are we creating a problem with the changes?  A final possibility is the Linux mixer..

 

I have not spent enough time looking at the source code to understand it at all.  

[Dev]

1 hour ago, bobfa said:

We have two or three  interactions here that are possible. Is Roon doing something to cause issues from the server.  I suppose we could find a LMS to try to eliminate that.  Second is the code not written to handle the larger buffers or are we creating a problem with the changes?  A final possibility is the Linux mixer..

 

I have not spent enough time looking at the source code to understand it at all.  

 

Earlier I have extensively tried LMS/Squeezelite combo. Even with very large buffers (the max I can go with 8gb ram and AL/ramroot), there were zero issues with LMS. I even tried upsampling to DSD256 (using sox) to see how it sounds but there were zero glitches. I think the problem is on the Roon side and probably to do with the messaging protocol (someone said earlier it's different than LMS). Also note that how Roon limits sampling rate to 192K and DSD64 only. There is no such limitations with LMS. 

[austinpop]

1 hour ago, Dev said:

 

Earlier I have extensively tried LMS/Squeezelite combo. Even with very large buffers (the max I can go with 8gb ram and AL/ramroot), there were zero issues with LMS. I even tried upsampling to DSD256 (using sox) to see how it sounds but there were zero glitches. I think the problem is on the Roon side and probably to do with the messaging protocol (someone said earlier it's different than LMS). Also note that how Roon limits sampling rate to 192K and DSD64 only. There is no such limitations with LMS. 

 

Exactly.

 

Luckily it is easy enough to switch to LMS on AudioLinux. Euphony does not currently come with LMS. It would be useful if it did.

[Summit]

Have anyone had the chance to listen to the Auralic Aries G2 yet?

 

Maybe even compared it to other network streamer/servers like Pink Faun or Zenith SE?

[bobfa]

8 minutes ago, Summit said:

Have anyone had the chance to listen to the Auralic Aries G2 yet?

 

Maybe even compared it to other network streamer/servers like Pink Faun or Zenith SE?

 

[Summit]

9 minutes ago, bobfa said:

 

 

Thanks I have read it before. I was wondering if some else than CA has listen to the Auralic Aries G2 and compared it other top network streamer or servers.

[Pink Faun]

On ‎4‎/‎9‎/‎2019 at 8:13 PM, Summit said:

Have anyone had the chance to listen to the Auralic Aries G2 yet?

 

Maybe even compared it to other network streamer/servers like Pink Faun or Zenith SE?

Here's a review of the Pink Faun streamer 2.16x and comparison with other music servers:

https://www.hifi-advice.com/blog/review/digital-reviews/music-server-reviews/pink-faun-2-16x-1-3/

 

[Summit]

20 hours ago, Pink Faun said:

Here's a review of the Pink Faun streamer 2.16x and comparison with other music servers:

https://www.hifi-advice.com/blog/review/digital-reviews/music-server-reviews/pink-faun-2-16x-1-3/

 

 

Thanks. Punter is an awesome reviewer. I would love to read a compare between the Auralic Aries G2, the Pink Faun 2.16X and other SOTA network streamer/servers.

[Ossanboy]

Hi Guys,

 

Halo, I just joined this forum recently I stumbled upon it when researching about network switcher for streaming audio.

 

I only found out about JSSG 360   yesterday. I want to try it first with my power amplifier internal wiring. The cable I use is Mogami W3104 11AWG (speaker cable). It sounds amazing but I want to try to JSSG360 every single one of them. Will this be beneficial ?

 

 

Thank you

[sandyk]

4 hours ago, Ossanboy said:

Hi Guys,

 

Halo, I just joined this forum recently I stumbled upon it when researching about network switcher for streaming audio.

 

I only found out about JSSG 360   yesterday. I want to try it first with my power amplifier internal wiring. The cable I use is Mogami W3104 11AWG (speaker cable). It sounds amazing but I want to try to JSSG360 every single one of them. Will this be beneficial ?

 

 

Thank you

 

Personally, I would only bother trying it on the input cables. Even then, because the amplifier is normally inside a metal case it may not achieve too much as radiated RF/EMI from external sources will be pretty well excluded already, and proper internal layout with good quality shielded cables should be adequate to protect against PSU related RF/EMI ingress. (e.g. rectifier diode switching artifacts etc.)

The amplifier output leads are usually pretty short, and may even be direct from onboard PCB sockets.

[Ossanboy]

Thanks for the reply!

 

To make it clear, here are the photo of my power amp.

 

The power is separate, with a huge transformer! 

The RCA input cable already double shielded braid but it is not in JSSG configuration. I will add it. 

 

[rickca]

On 4/3/2019 at 7:37 PM, rickca said:

I'm looking forward to Computex in late May.  AMD is expected to formally announce Ryzen 3000 CPUs and x570 chipset featuring PCIe V4 with availability in July.  AMD stock has tripled in the past 12 months. 

And now it looks like Intel will announce their 9th gen 35W CPUs on May 15.

 

Do we understand yet whether it's better for a server to have more cores with a lower base clock vs fewer cores with a higher base clock?  What's the best trade off?

 

https://www.anandtech.com/show/14214/intel-9th-generation-core-35w-cpus-launch-may-15th-kikatek

[soares]

No words to describe the results from using REW for room correction. I am completely astonished how good the SQ became and this without manual correction of the results. Lot’s of room to improve... The Program is free (but consider donation) and you just need to buy a compatible mic. Give it a try. Cheers. Jorge

[Ricardo007]

@ausstinpop

What does mean bridging a server and how To achieve this?

On 5/9/2017 at 7:23 AM, austinpop said:

 

Yes, it is direct - in the sense that the music server is bridged, and the direct connection from the bridge is going into the Zyxel (modded) switch, along with the sMS-200 "ultra."

 

In my simplistic way of thinking, this "ultra" switch's role is nothing but an Ethernet "regen" with the ultra clock.

 

[mourip]

7 hours ago, Ricardo007 said:

@ausstinpop

What does mean bridging a server and how To achieve this?

 

 

There is a helpful index at the top of every page of this very long thread. One of the links found there is for network bridging.

[seeteeyou]

Fanless by default, and we've got ECC RAM support this time according to their PDF file

 

https://www.asrockind.com/news/index.asp?iD=2

https://www.amd.com/system/files/documents/R1000-product-brief.pdf#page=3

 

ASRock Announces iBox-R1000 AMD-Powered 'NUC'
https://www.tomshardware.com/news/asrock-announces-ibox-r1000-amd-nuc,39113.html

 

ASRock unveils iBOX-R1000 mini PC with AMD Ryzen R1000 processor
https://liliputing.com/2019/04/asrock-unveils-ibox-r1000-mini-pc-with-amd-ryzen-r1000-processor.html

 

AMD announces Ryzen Embedded R1000 series with the R1606G and R1505G processors; 3DMark 11 performance comparison unearthed
https://www.notebookcheck.net/AMD-announces-Ryzen-Embedded-R1000-series-with-the-R1606G-and-R1505G-processors-3DMark-11-performance-comparison-unearthed.417920.0.html

 

AMD Ryzen™ Embedded R1000 Series
https://www.amd.com/en/products/embedded-ryzen-r1000-series

 

Ryzen Embedded R1606G - AMD
https://en.wikichip.org/wiki/amd/ryzen_embedded/r1606g

 

Ryzen Embedded R1505G - AMD
https://en.wikichip.org/wiki/amd/ryzen_embedded/r1505g

[Ossanboy]

16 hours ago, mourip said:

 

There is a helpful index at the top of every page of this very long thread. One of the links found there is for network bridging.

 

Does anyone know if I can bridge NAS synology running roon server to Roon endpoint (Trinnov Altitude) directly?

 

If yes how?

 

Thanks

[seeteeyou]

On 11/12/2018 at 9:16 AM, romaz said:

It turns out the HDMI outputs in these NUCs can output digital audio (you would have to turn on this feature in the BIOS) and using this Kanex HDMI to Toslink/SPDIF device, it works very well (up to 24/192 PCM).  In fact, using an optical HDMI cable and powering this Kanex unit with my LPS-1 (set to 5V), SQ is quite good going out to either SPDIF coax or Toslink.  Having said that, USB still sounds better resolved and more finessed but who knows what might happen if the clocks on this cheap unit were replaced.

 

Thread gone quiet lately? I found this box for less than $450 and it does support HDMI input / coaxial output with DoP

 

https://item.taobao.com/item.htm?id=590187330024

 

Microsoft makes Windows 10 1903 available on MSDN

 

https://www.zdnet.com/article/microsoft-makes-windows-10-1903-available-on-msdn/

 

For audio output via HDMI, here's latest version of Intel Graphics DCH drivers for Windows 10 1903

 

https://downloadcenter.intel.com/download/28731/Intel-Graphics-Windows-10-DCH-Drivers

https://downloadmirror.intel.com/28731/a08/dch_igcc_win64_26.20.100.6709.exe

https://downloadmirror.intel.com/28731/a08/dch_igcc_win64_26.20.100.6709.zip

https://downloadmirror.intel.com/28731/eng/ReadMe.txt

https://downloadmirror.intel.com/28731/eng/ReleaseNotes_DCH_iGCC_26.20.100.6709.pdf

Quote

Intel® Display Audio Driver (10.26.00.06)