Jump to content
austinpop

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

Rate this topic

Recommended Posts

On 11/2/2018 at 2:51 PM, seeteeyou said:

A little bit of clarification would be great, do you have more than one HP Z820 by any chance? Dual E5-2687W V2 in 2017 and then single E5-2687W V2 in 2018?

 

Yes, my office has a few HP Z820s and even an older Z800.  We really like these machines because they seem to last forever but they are extremely noisy and so I eventually replaced the one I was using at home with the Mac Pro which is essentially just as powerful for my purposes (4k editing in Premiere Pro and After Effects) and also whisper quiet.  The Z820 I previously had in my home was a dual CPU model with a total of 16 cores but the unit I brought home from the office recently for testing was one of our single CPU workstations.  What was not initially intuitive to me is that due to inefficiencies, there is an overhead to incorporating 2 CPUs on a single board to the extent that a 2nd CPU doesn’t make the computer twice as fast but rather only about 20% faster.  A single CPU with 16 cores would be a better way to go with multithreaded apps.  Today, I believe you can get a Xeon with as many as 24 cores although I suspect that for RoonServer, that would probably be way overkill.

Share this post


Link to post
Share on other sites
On 11/3/2018 at 1:00 PM, beautiful music said:

Since there are many folks here are using Roon with NUC running Audiolinux, so what about the jRiver?

 

How did you find jRiver with NUC mentioned here?

 

Also curious to know is there a superior benefit in SQ from Roon over jRiver, because if so I will switch to Roon eventually.

 

 

I own JRiver and the last time I compared it against Roon was back when I owned a CAD CAT which utilized Scot Berry’s optimized version of Windows 8.1 and so this was back in 2016.  During my comparisons, I did find that JRiver sounded better than Roon although I got my very best SQ using JRiver only as a UPNP server while using JPlay Streamer as a UPNP renderer all on the same machine and so it seemed to me that JPlay Streamer was an even lower latency player than JRiver but nevertheless, Roon came in last.  There were other platforms that I felt were also better sounding than Roon at the time including Audrivana+ and Bughead Emperor but I didn’t find the SQ differences to be night and day better and I found that Roon’s vastly superior user experience with regards to its interface, ease of managing a large library as well as the integration of Tidal to be worth the SQ hit and so I paid my lifetime subscription and I haven’t looked back.

 

As time has progressed, Roon has improved dramatically but so has JRiver probably and so who knows if JRiver is still better.  What I can say is that with Roon distributed among 3 machines (as RoonServer, RoonBridge, and Roon Remote) and with a powerful RoonServer (which is by default headless), a low power RoonBridge, and my iPAD Pro as Roon Remote, the SQ improvement I am hearing compared against Roon running on a single machine is greater than the SQ gap I recall hearing when I last compared Roon against JRiver.

 

Having said that, since JRiver can run on AudioLinux, it would be useful to give it a try and to compare it against Roon and report back.  It’s possible it may still be better than Roon although I am quite pleased with what I am getting from Roon these days and with the likely integration of Qobuz, coming soon, it would have to take something extraordinary from JRiver to get me to switch.

Share this post


Link to post
Share on other sites
On 11/4/2018 at 1:07 AM, Superdad said:

 

So very true, Alex, and I believe I acknowledged your pioneering contributions very early on in this thread:

 

 

Larry deserves credit as well:

 

 

But even your original USB Regen was preceded by the Schiit Wyrd and Sonore's microRendu was preceded by SOtM's sMS-100.  As you said, "and still we progress."

Share this post


Link to post
Share on other sites
13 minutes ago, romaz said:

As time has progressed, Roon has improved dramatically but so has JRiver probably and so who knows if JRiver is still better.  What I can say is that with Roon distributed among 3 machines (as RoonServer, RoonBridge, and Roon Remote) and with a powerful RoonServer (which is by default headless), a low power RoonBridge, and my iPAD Pro as Roon Remote, the SQ improvement I am hearing compared against Roon running on a single machine is greater than the SQ gap I recall hearing when I last compared Roon against JRiver.

 

Having said that, since JRiver can run on AudioLinux, it would be useful to give it a try and to compare it against Roon and report back.  It’s possible it may still be better than Roon although I am quite pleased with what I am getting from Roon these days and with the likely integration of Qobuz, coming soon, it would have to take something extraordinary from JRiver to get me to switch.

 

@romaz are you up-sampling by any chance ? I was using Roon end-to-end and just started to experiment with HQP in the mix with Roon just managing the library and using NAA instead on the streamer (NUC). I need to up-sample to DSD256. Though CPU usage is lot more taxing with HQP than Roon, it appears that the SQ might be better. I am still experimenting.

 

Share this post


Link to post
Share on other sites
9 minutes ago, Dev said:

 

@romaz are you up-sampling by any chance ? I was using Roon end-to-end and just started to experiment with HQP in the mix with Roon just managing the library and using NAA instead on the streamer (NUC). I need to up-sample to DSD256. Though CPU usage is lot more taxing with HQP than Roon, it appears that the SQ might be better. I am still experimenting.

 

 

No, I'm not.  I have tried it in the past with HQP combined with Roon.  I know this is a great option for many but I'm using Chord's M-scaler for upsampling instead.  For people with Chord DACs, it's a no fuss way to upsample and the results have been very satisfying.  You would do well to see what @flkin has been doing as he does this with his Pink Faun 2.16.

Share this post


Link to post
Share on other sites
9 minutes ago, romaz said:

 

No, I'm not.  I have tried it in the past with HQP combined with Roon.  I know this is a great option for many but I'm using Chord's M-scaler for upsampling instead.  For people with Chord DACs, it's a no fuss way to upsample and the results have been very satisfying.  You would do well to see what @flkin has been doing as he does this with his Pink Faun 2.16.

 

Ok, got it. @flkin, are you using Pink Faun to up-sample ? if so, with HQP or Roon ?

Share this post


Link to post
Share on other sites
1 hour ago, romaz said:

With regards to buffers, this is baffling for many of us.  There are buffers in our NICs and network switches, our hard drives and SSDs have memory buffers, our CPUs have 3 levels of buffers (L1, L2, and L3), and of course, RAM serves as a buffer.  Buffers are supposed to minimize wait states so that the faster components are not waiting needlessly on the slower components and yet, despite all of these buffers, latency is never zero meaning there will probably always be room for improvement.  

 

With Roon, the latency could further be reduced by caching the music files in the ram to start with. I see that with roon playback, it constantly accesses the usb drive which stores the songs. After a bit of search in Roon forums, it seems like they don't. Dany said they will probably think about it and implement it at some point. In order to find out if its beneficial to SQ or not, we could create a small ramdisk with couple of music files and add it to the storage configuration - it would be easier to then compare the effect with and without. Between, JRiver already does this.

 

Share this post


Link to post
Share on other sites
41 minutes ago, Dev said:

 

With Roon, the latency could further be reduced by caching the music files in the ram to start with. I see that with roon playback, it constantly accesses the usb drive which stores the songs. After a bit of search in Roon forums, it seems like they don't. Dany said they will probably think about it and implement it at some point. In order to find out if its beneficial to SQ or not, we could create a small ramdisk with couple of music files and add it to the storage configuration - it would be easier to then compare the effect with and without. Between, JRiver already does this.

 

 

Very interesting.  I'm glad to know the guys at Roon are focusing more on SQ.  Your ramdisk suggestion for music files should be easy enough to try out.  Deep down, I'm hoping it doesn't result in a significant improvement because I'm loving the simplicity and versatility of what I have now.

Share this post


Link to post
Share on other sites

The DAC has a buffer too. As long as the DAC buffer never empties how in the world could latency anywhere up the chain make a difference? I am a networking guy and networking protocols are specifically designed to handle a wide range of latency. So the latency one might see in the computer can't matter as long as the buffer in the DAC never empties. It never does on a properly running setup with a properly running network. If the DAC buffer never empties, the DAC, nor the data being transferred, has any notion of the latency anywhere upstream. It can't matter.

Share this post


Link to post
Share on other sites

Strange, with my sonicTransporter i5 with HQPlayer Embedded and Roon, I can use these filters Dithers/Noise Shapers for PCM:

 

Filters

none

IIR

FIR

asymFIR

minphaseFIR

FFT

poly-sinc-lp

poly-sinc-mp

poly-sinc-short-lp

poly-sinc-short-mp

poly-sinc-hb

poly-sinc-ext

poly-sinc-ext2

poly-sinc-mqa-lp

poly-sinc-mqa-mp

poly-sinc-xtr-lp

poly-sinc-xtr-mp

ASRC

polynomial-1

polynomial-2

minringFIR-lp

minringFIR-mp

closed-form

closed-form-fast

closed-form-M

sinc-M

 

Dithers/Noise Shapers

none

NS1

NS4

NS5

NS9

RPDF

TPDF

Gauss1

shaped

 

With my DAC, I prefer upsampling everything to 352.8 and I use the poly-sinc-xtr-mp filter with the NS5 noise shaper.

 

Note that using no upsampling, no filter, and no noise shaper with HQPlayer still sounds better than using Roon without HQPlayer to my DAC through an ultraRendu. That tells me that there is something better about using HQPlayer and NAA versus Roon's RAAT.

 

Scott

Share this post


Link to post
Share on other sites
2 hours ago, romaz said:

Straight out of the box, even without the Ref10 and despite the initial harshness, I was simply blown away by this switch.  What is amazing is I thought I knew what to expect based on my experience with the prototype and yet my expectations were still exceeded.  As good as the TLS switch is, this switch is operating several notches higher.  This switch isolates, that's for sure, but more impressively, it acts like a gain stage.  Dynamics are unreal.  Everything is just bigger and badder but this switch also portrays the subtlest of nuances and the most delicate textures.  To my ears, this sNH-10G is their most impactful device that I have heard.  More impactful than the sMS-200ultra Neo, tX-USBultra, or sCLK-OCX10 master clock and so if I were forced to choose a single SOtM product, this would be it.

 

Hi Roy,

 

Many thanks for your recent contributions including the above on the sNH-10G

 

I've been in that rare audiophile state of being (fairly) content with my system and SQ for a while now. But funnily enough I spent a couple hours on Sunday morning rereading both yours and Rajiv's reviews of the sNH-10G prototype. Posted here for convenience of others that are interested:

 

 

 

 

 

I currently have the AQVOX SE in my system and am generally pretty happy with the SQ from Tidal. I also have the Zenith SE after following your findings on that and am generally pretty happy with the SQ of locally stored files (although annoyingly more with the LMS player rather than my preferred Roon).

 

But. And yes, there is ALWAYS a but! The prospect of the sNH-10G further lifting that Tidal playback is intriguing and there are two thoughts bouncing around my mind that have me intrigued (or perhaps that should be bothered!).

 

Path 1: Tidal into AQVOX SE > Zenith SE > TX-USBUltra > DAC

 

I'm intrigued by the prospect of enhancing this to:

 

Path 1: Tidal into sNH-10G > Zenith SE > TX-USBUltra > DAC

 

The sNH-10G comes with the option of me connecting it to my sCLK-OCX10 (which on paper at least) suggests it's likely to outperform the AQVOX.

 

(Also having spent the significant outlay on the reference clock there is an appeal to utilizing it more than I currently do).

 

But then there another path or two which become potentially very interesting. I have a DAC that accepts ethernet in. So by adding a switch that is disciplined by a reference clock two other paths open up that might just surpass my heavily optimised USB path:

 

Path 2: Tidal into sNH-10G > Zenith SE (LAN output) > DAC

 

or

 

Path 3: Tidal into sNH-10G > Zenith SE (LAN output) > sNH-10G > DAC

 

In other words I can pass 'clean/clocked' ethernet into the SE and then directly onto the DAC. Or I could pass 'clean/clocked' ethernet into the SE and then pack to the sNH-10G to be 'cleaned/clocked' again before being passed onto DAC.

 

Roy - my use-case is a lot closer to @austinpop's than yours but your observations of the sNH-10G: 

 

2 hours ago, romaz said:

switch isolates, that's for sure, but more impressively, it acts like a gain stage.  Dynamics are unreal.  Everything is just bigger and badder

 

Well that's put the cat amongst the pigeons for me. Especially in Path 3 where crucially both Tidal and locally stored files pass through the sNH-10G ?

 

Whilst I'm nowhere near as adventurous or experimental as many on this thread I'm tempted to give it a go out of scientific discovery as much as out of the hope of raising the SQ further. But then again it's spending another 2 grand on a system that I currently have no complaints about.

 

The SoTM website says the sNH-10G doesn't start shipping until 26th of November so I guess I have some time to ponder, pontificate and otherwise generally drive myself up the wall with such thoughts! ?

 

Then of course there is the other 'maybe hold off' factor. Which is @Superdad's upcoming audiophile switch. Alex - I don't suppose you and John S are any closer to announcing a much awaited release date for yours?

 

Anyway - pardon my ramblings but since I'm driving myself to distraction I thought I'd share in case any of the gang have any thoughts or insight.

 

Cheers,

Alan 

 

 

Share this post


Link to post
Share on other sites
4 hours ago, MrUnderhill said:

A few notes on where I am up to:...

This really is putting the cat amongst the pigeons!

 

Thanks M, I was wondering how you are getting on :)

 

I'm also still running mine in and experimenting while waiting for 5.5/2.5 plugs so I can power with a PH SR4 at 12v. I will report more later when fully assessed.
 

I have to say that this setup has moved the goal posts for me by quite a margin already in the best possible way.

Share this post


Link to post
Share on other sites
8 hours ago, romaz said:

RAAT is only applicable in a Roon Ready device as Bridge doesn't use RAAT.

Maybe I'm just lucky, but my Roonbridge definitely has a RAATserver process running next to it.

 

Larry

Share this post


Link to post
Share on other sites
5 hours ago, BigAlMc said:

But then there another path or two which become potentially very interesting. I have a DAC that accepts ethernet in. So by adding a switch that is disciplined by a reference clock two other paths open up that might just surpass my heavily optimised USB path:

 

Path 2: Tidal into sNH-10G > Zenith SE (LAN output) > DAC

 

or

 

Path 3: Tidal into sNH-10G > Zenith SE (LAN output) > sNH-10G > DAC

I'm a bit confused about this.  If your DAC is ethernet attached to your SE, then of course the TIDAL files have to pass through the SE to get to the DAC.  I'm not so sure if the streaming music goes through the SE if your DAC is attached to the switch as in Path 3.

 

I am confused because if you were using DLNA/UPnP in Path 3 the music stream would go directly through the switch into your DAC.  What I don't understand is how this works if your using LMS/Squeezelite with a control point like iPeng.  Maybe @left channel can educate me, please?

 

And maybe I just haven't had enough coffee yet today.

Share this post


Link to post
Share on other sites
7 minutes ago, rickca said:

I am confused because if you were using DLNA/UPnP in Path 3 the music stream would go directly through the switch into your DAC.  What I don't understand is how this works if your using LMS/Squeezelite with a control point like iPeng.  Maybe @left channel can educate me, please?

 

Hi Rick,

 

Actually I think you're right.

 

In path 2 the Tidal files go through the SE because the LAN cables are connected: Switch > SE > DAC

 

In path 3 the Tidal files would probably actually go Switch > DAC (and the only thing doing via the SE would be the server commands).

 

So with this in mind I guess the question is not whether the Tidal files get 'cleaned/reclocked' twice by the sNH-10G in Path 3, but whether Path 3 is better than Path 2 because it skips the SE and goes directly to the DAC from the (sCLK-OCX10 disciplined) sNH-10G.

 

Good catch!

 

Thanks,

Alan

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×