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

[austinpop]

1 hour ago, Advieira said:

 Ok, I will wait @austinpop and @romaz confirming its findings.

 

Hi Adveira,

 

I sympathize with the difficulty in keeping up with this thread. It has gotten pretty wild in the last few months! 

 

BTW - I am not picking on you - I quoted your post to represent a sentiment I get a lot via PM which goes along the lines of - "So what are the conclusions?" Again, I completely understand the sentiment, but it's a question we can't answer. 

 

This thread is a journey with no foreseeable "conclusion!" Yes, I know Roy, Larry, and I talk in moments of excitement of how this is endgame stuff, but this thread is, and will always be, an exploration for more SQ, and most of time with mechanisms that don't have clear objective explanations. That's OK.

 

Take my setup, for example. When Roy started this thread on Jan 1, 2017 - yes, almost 2 years ago - I was using an Aries Mini with Auralic LPS as my endpoint, with MinimServer on my NAS. I thought this was about as endgame as it got. Then I tried:

• FMCs
• bridging
• a plethora of PSUs
• sMS-200
• sCLK-EX with the famous trifecta of switch-sMS-tX
• Reference clock - first Cybershaft, then Mutec Ref-10
• then I moved from distributed to standalone (Zenith SE)
• more stuff
• then I tried AL/RAM on my SE
• Now I am back to distributed with NUC/AL as endpoint.

Is this the conclusion? For me and many of us here - you know it isn't! I have zero doubt we will learn more, and we will continue to improve the SQ of our systems. Ultimately this thread stays popular because a critical mass of people try these findings for themselves and report their findings. As Larry pointed out, not everyone hears the same effect, or even a positive change. That's OK too. We all start from a different baseline, and we all hear things differently. 

 

The only "conclusion" is to ask what people are using currently, and I'll post a diagram soon, but my system is till in flux. 

[Advieira]

8 minutes ago, austinpop said:

The only "conclusion" is to ask what people are using currently, and I'll post a diagram soon, but my system is till in flux. 

 

But maybe you find some kind of holy grail of the computer audio, I mean, a device to completely and 100% turn the mediaserver unimportant to SQ. Could a clocked switch, before NAA, be this device?

 

Rob Watts said to me its new M Scaler do it (make server unimportant).

 

I mean, if we can isolate the server, its one thing less to worry about.

[LTG2010]

20 minutes ago, davide256 said:

Given the good result I can get with  LPS 1.2/overclocked RPI3/AL as endpoint I'm not buying a "more power/more RAM" argument for better endpoint.

I'm certainly not against giving the RPI at try but with only 1gb of RAM I assume Ramboot is not possible? and for me the difference between Ramboot and non Ramboot was not subtle therefore I can't agree with your findings.

In fact you will see that some commercial; enterprises are now licensing Audiolinux Ramboot for their streamers, which would not be happening if the difference was small.

[rettib2001]

2 hours ago, LTG2010 said:

Carry on for now, there's so much more to discover especially on the server side, eg:

How powerfull does the server need to be?

Impact on power supply on server, atx or linear ?

Type of server NUC with linear power supply or standard high power setup?

Effect of better clocks on NUC.

All these questions have been racing through my mind today. 

 

Also wondering if going' fanless' on the server side is worth it or not. I guess it is even if its just to reduce temps. 

[austinpop]

9 minutes ago, Advieira said:

 

But maybe you find some kind of holy grail of the computer audio, I mean, a device to completely and 100% turn the mediaserver unimportant to SQ. Could a clocked switch, before NAA, be this device?

 

Rob Watts said to me its new M Scaler do it (make server unimportant).

 

I mean, if we can isolate the server, its one thing less to worry about.

 

Repeat after me: there is no holy grail.   Everything still matters.

 

Somewhere upthread, Roy posted, and I have confirmed, that even with a reclocked, isolated switch (SOtM in his case, TLS in mine), the server still matters. Search for his posts in the last 3 months - there aren't many, but on the flip side, they are incredibly informative.

[LTG2010]

5 minutes ago, rettib2001 said:

Also wondering if going' fanless' on the server side is worth it or not. I guess it is even if its just to reduce temps.

Go fanless in any type of server / endpoint. The fan introduces electrical noise, emi, vibration and it just blows hot air around onto other components. Plus the processor is monitoring temperature and adjusting its speed etc.

[Advieira]

17 minutes ago, austinpop said:

 

Repeat after me: there is no holy grail.   Everything still matters.

 

Somewhere upthread, Roy posted, and I have confirmed, that even with a reclocked, isolated switch (SOtM in his case, TLS in mine), the server still matters. Search for his posts in the last 3 months - there aren't many, but on the flip side, they are incredibly informative.

 I see.

But what has more impact in Sound Quality in these new findings: a NUC+Roon server or a NUC+AL endpoint?

[austinpop]

2 minutes ago, Advieira said:

 I see.

But what has more impact in Sound Quality in these new findings: a NUC+Roon server or a NUC+AL endpoint?

 

Search is your friend. Please search back on this thread, we have posted this multiple times.

 

Yes, the consensus is the NUC (or endpoint) running AL in RAM provides the bulk of the improvement. Whether it is 60% or 80% or some other number is based on subjective opinions.

 

The server still seems to matter. Running AL in RAM on the server matters. System speed seems to matter. But bigger systems with their noise add harshness too. Yes, even through the switches. Where is the sweet spot? What are the best CPU options? Is TDP a guiding metric? Where should music be stored? Pros/cons?

 

These all need research. So get to work!

[Windows X]

From the tests with my friends, I recommend to use low powered design for better sound quality with AudioLinux. Maybe something like Celeron or i3 processor rather than Intel i5-i7 processor. Unless you can resolve strong effect of digital glare in Linux platform, I don't recommend using Ramroot or RAMOS feature as it makes the glare more stronger too.

[davide256]

24 minutes ago, Windows X said:

From the tests with my friends, I recommend to use low powered design for better sound quality with AudioLinux. Maybe something like Celeron or i3 processor rather than Intel i5-i7 processor. Unless you can resolve strong effect of digital glare in Linux platform, I don't recommend using Ramroot or RAMOS feature as it makes the glare more stronger too.

Can you elaborate on this? Glare is what is driving me not to use Pentium NUC as AL endpoint... Magnepans seem to magnify glare 

[Windows X]

12 minutes ago, davide256 said:

Can you elaborate on this? Glare is what is driving me not to use Pentium NUC as AL endpoint... Magnepans seem to magnify glare 

 

More processor power tends to produce high frequency noise in Linux OS. Using low powered device, high frequency noise will be less apparent in Linux OS. Windows address these issues better so high processing power processor won't cause glare effect as much as Linux one.

 

Such issue in Linux can be redeemed through Alsa audio driver configuration and kernel configuration to reduce glare effect but they're mostly hardware dependent like adjusting some parameters to perform better in specific hardware.

[rickca]

24 minutes ago, Windows X said:

Windows address these issues better

How does Windows do this?

[Confused]

3 hours ago, LTG2010 said:

Yes as a previous owner of an Ultra, but why? I'm not sure but there are some differences between the 2.

First,, the NUC in general has more processing power.

Second the NUC processors mentioned have a much larger processor cache.

Third, you can use audiolinux booted into RAM on the NUC, rather than an OS on a micro sd card. I dont think there is enough memory on the rendu/ultra to boot audiolinux in RAM, that would level the playing field for testing.

The sMS-200ultra has 2 GB DDR3 RAM.  I had assumed that loading AudioLinux (or Jussie’s similar HQPlayer image) on a sMS-200ultra would be impossible?

[Windows X]

18 minutes ago, rickca said:

How does Windows do this?

 

Before I answer, do you use high powered processor and have digital glare issue with AudioLinux?

[rickca]

1 minute ago, Windows X said:

 

Before I answer, do you use high powered processor and have digital glare issue with AudioLinux?

No.  I have an i7-6700K system running Windows 10 with a considerable amount of optimization.  I power an MLC SSD with the OS separately with a 5V LPS.  I also own an Innuos Zenith SE which certainly does not have digital glare.

[lmitche]

2 hours ago, Windows X said:

From the tests with my friends, I recommend to use low powered design for better sound quality with AudioLinux. Maybe something like Celeron or i3 processor rather than Intel i5-i7 processor. Unless you can resolve strong effect of digital glare in Linux platform, I don't recommend using Ramroot or RAMOS feature as it makes the glare more stronger too.

How are you avoiding the use of flash memory during playback on your endpoint?

[Dev]

1 hour ago, Windows X said:

 

More processor power tends to produce high frequency noise in Linux OS. Using low powered device, high frequency noise will be less apparent in Linux OS. Windows address these issues better so high processing power processor won't cause glare effect as much as Linux one.

 

Such issue in Linux can be redeemed through Alsa audio driver configuration and kernel configuration to reduce glare effect but they're mostly hardware dependent like adjusting some parameters to perform better in specific hardware. 

 

Can you elaborate technically ? How does Windows handle glare better ? I have a low power Celeron dual core NUC in a fanless case as well as 8th gen i7 NUC in the stock case with noisy fan - running them as endpoint has different results. Before I tell which one sounds better (in my system off-coarse), I am interested to know what Alsa driver parameters and kernel configuration can reduce glare effect ? Maybe we can incorporate them in AL.

[LTG2010]

57 minutes ago, Confused said:

The sMS-200ultra has 2 GB DDR3 RAM.  I had assumed that loading AudioLinux (or Jussie’s similar HQPlayer image) on a sMS-200ultra would be impossible?

For audiolinux you need 4gb RAM min. But with HQP daemon the image is much less so I would imagine it will work and might be a super way to upgrade the unit, burn it to a micro sd and boot.

[mourip]

55 minutes ago, Windows X said:

 

Before I answer, do you use high powered processor and have digital glare issue with AudioLinux?

 

 Are you the developer of Fidelizer, a program sold to optimize Windows for audio?

 

I imagine that we all would be interested in your answer to @rickca's question. Perhaps it will give us ideas for Linux although personally I have found AL in ramroot to not have anything like glare, which I am very sensitive to.

 

Have you looked into optimizing Linux?

[Blake]

3 hours ago, austinpop said:

Is this the conclusion? For me and many of us here - you know it isn't! I have zero doubt we will learn more, and we will continue to improve the SQ of our systems. Ultimately this thread stays popular because a critical mass of people try these findings for themselves and report their findings. As Larry pointed out, not everyone hears the same effect, or even a positive change. That's OK too. We all start from a different baseline, and we all hear things differently. 

 

The only "conclusion" is to ask what people are using currently, and I'll post a diagram soon, but my system is till in flux. 

 

Well stated.  This should be a sticky!  This area is rapidly changing.  At some point, you have to decide whether you want to continue to spectate, or jump on board and buy/try something.  The only thing certain is that a year from now, all will likely have changed to something else.

 

For me personally, I have decided to continue to monitor things until a friendly consumer version is available that will not require frequent visits to the NUC/AL troubleshooting thread.  

 

Oh, and the other thing that jumps out to me- be cautious about spending a lot of money on this part of your system or servers, clocks or DACs for that matter.  Things are changing too quickly and today's treasure is tomorrow's "meh" product and the value of used items plummet.  Very much like computers.

 

However, that is what is great about this NUC/AL solution- this stuff is cheap if you are DIY.

 

 

[mourip]

2 hours ago, austinpop said:

Yes, the consensus is the NUC (or endpoint) running AL in RAM provides the bulk of the improvement. Whether it is 60% or 80% or some other number is based on subjective opinions.

 

The server still seems to matter. Running AL in RAM on the server matters. System speed seems to matter. But bigger systems with their noise add harshness too. Yes, even through the switches. Where is the sweet spot? What are the best CPU options? Is TDP a guiding metric? Where should music be stored? Pros/cons?

 

I run AL in ramroot on my endpoints but Windows 10 Pro on my Roon server for now. I found that when I installed Audiophile Optimizer it made a big difference in overall sound quality which lends support to the general consnsus that the server OS and tweaks still do matter.

[austinpop]

Just now, mourip said:

Are you the developer of Fidelizer, a program sold to optimize Windows for audio?

 

 

@Windows X is this true?

 

First of all, respect. I haven't personally used Fidelizer, but have heard great things about it. I have used Audiophile Optimizer and Process Lasso on Windows. While all of these products made a positive difference, the effects were subtle compared to what I am hearing with Audiolinux running in RAM.

 

I for one do not hear even the hint of glare with AL/RAM! Quite the opposite.

 

Please realize that when you come into a thread like this making general negative statements against a competing product, you will come across as biased.

[Johnseye]

10 hours ago, Em2016 said:

 

Hi John

 

Did you end up comparing the NAA ramroot bootable image @Miska made, against Audiolinux ramroot NAA?

 

Any differences to your ears?

 

No sound quality differences to my ears...

 

Yes, I tried both.  If there is any difference it's so minimal that bias comes into question.  I'd say both are equally as good for an HQPlayer NAA.  If you need a Roon Bridge or other as NAA, obviously Jussi's OS won't work.

 

 

10 hours ago, Em2016 said:

 

But Jussi's image wins on performance for me - when playing new tracks there are pops just at the beginning of tracks with Audiolinux (and nearly every Linux distro I've had running NAA before).

 

With Jussi's bootable image, it's dead silence changing tracks, so Jussi's image easily wins for me. These pops between tracks (between, never during) are way more noticeable on headphones (which I use a lot) than speakers but they are still noticeable on speakers, again between tracks (not during for me).

 

Unexpected pops on headphones are never fun. Jussi's bootable image is flawless in this regard (as one would expect since he has full control of the bootable OS). And of course Jussi's bootable image comes free versus Audiolinux cost (assumption is someone already has a HQP license of course).

 

Interested to hear your impressions about sound quality though.

 

 

This is very interesting because up until last weekend I had not experienced any pops, except when I swap between DSD and PCM.  Now for whatever reason I'm hearing a pop between every track.  Because this wasn't happening before, I question whether it's a software setting or other change I made.  I will put the HQPlayer NAA back on the NUC and see what happens.  I was also using an iso Regen over the weekend.

 

With my Audio Alchemy DAC I used to get these pops all the time, and some very loud.  It actually blew my B&W tweeters.  It was one of the reasons I bought a new DAC.  When I got the Holo Spring they stopped, until now.  They are very small and quiet pops now however.

 

This is what Peter Madnick described as the cause of the pop.  This is different than what we're talking about, but I question why it's happening, how it's impacted by the hardware and if any mute relay is involved.

 

"When you power a product on and off (ours or anyone else’s) the power supplies must charge up, or discharge, as the case may be.  The raw power supply voltages run through regulators to stabilize them at some known number (+ and – 18 Volts, in our case).  Total capacitance and current draw varies amongst the circuits and cannot be guaranteed to stay at the exact same voltage during the charge and discharge cycles.  An imbalance between the + and – can cause noise.  Again, all products are subject to this phenomenon.  The volume control is essentially the “valve” which controls a preamp’s output level.  If it is all the way up then any instantaneous noise which could come out before the MUTE relay closes can be heard as a POP. "

 

[Johnseye]

1 hour ago, Windows X said:

 

Before I answer, do you use high powered processor and have digital glare issue with AudioLinux?

 

1 hour ago, rickca said:

No.  I have an i7-6700K system running Windows 10 with a considerable amount of optimization.  I power an MLC SSD with the OS separately with a 5V LPS.  I also own an Innuos Zenith SE which certainly does not have digital glare.

 

I have a high powered processor and no glare with Audiolinux.

 

I have used Windows with both Fidelizer and Audiophile Optimizer so I have enough experience to compare.

I hear less digital glare, high end, sibilance, etc with the NUC and Audiolinux in ram than with any other configuration.

[Windows X]

19 minutes ago, austinpop said:

 

@Windows X is this true?

 

First of all, respect. I haven't personally used Fidelizer, but have heard great things about it. I have used Audiophile Optimizer and Process Lasso on Windows. While all of these products made a positive difference, the effects were subtle compared to what I am hearing with Audiolinux running in RAM.

 

I for one do not hear even the hint of glare with AL/RAM! Quite the opposite.

 

Please realize that when you come into a thread like this making general negative statements against a competing product, you will come across as biased.

 

I did optimizations on both Windows and Linux platform. Fidelizer for Windows and Purist ROM for Linux on DAP. I considered Purist OS but didn't make it because I prefer Windows myself.

 

If you think I came to this thread to make general negative statements to compete against AudioLinux, I'll quietly leave this thread to preserve the atmosphere of your moderation.

 

I came to share my findings about AudioLinux with my friends and some who has that issue asked me to elaborate it. I didn't meant to disturb the peace and harmoney of AudioLinux platform at all. That's not my intention and have a nice day.