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

[austinpop]

9 hours ago, romaz said:

If one wants to hear what true silence sounds like, then play back a file from your internal storage and unplug your device from the network because how can any network configuration sound better than "no network."

This is now my preferred mode of listening. Cue up a local album in HQPlayer, hit Play then physically disconnect the network. For good measure I quit JPLAY and Roon Server. And no second guessing about the effect of the upstream network.

 

Just stunning!

[flkin]

Good to see others are coming around to the idea that the network needs to be physically disconnected to see a baseline of the effect of the network connection has on the sound and this would then be the target to achieve with network optimisations. Anything else is added flavour to the network sound and for some might actually sound better which indicates that their downstream devices needs further work. 

 

I would like to repeat/add that in the current version of Euphony, both Tidal and Quobuz files are totalling downloaded into the streamer's buffer before playing. So disconnecting from the network will still play fine (for up to the end of the next track). However this doesn't seem to apply to the embedded HQPlayer from within Euphony.  The current playing track isn't totally buffered and only plays for a few seconds longer before stopping. Nor is the following track buffered. So in Euphony/HQPlayer's case, disconnecting from the network is not possible.

[R1200CL]

On 7/25/2019 at 5:26 AM, austinpop said:

A Mini Review of the Farad Super3 Power Supply

Post 14586
Did you ever try the Paul Hynes SR4T version ? Or any plan to do ?

The “turbo” versions claims to be very close to SR7.
 

Also the 12V rated 6A SR5T should be a good option. 

[ASRMichael]

28 minutes ago, flkin said:

Good to see others are coming around to the idea that the network needs to be physically disconnected to see a baseline of the effect of the network connection has on the sound and this would then be the target to achieve with network optimisations. Anything else is added flavour to the network sound and for some might actually sound better which indicates that their downstream devices needs further work. 

 

I would like to repeat/add that in the current version of Euphony, both Tidal and Quobuz files are totalling downloaded into the streamer's buffer before playing. So disconnecting from the network will still play fine (for up to the end of the next track). However this doesn't seem to apply to the embedded HQPlayer from within Euphony.  The current playing track isn't totally buffered and only plays for a few seconds longer before stopping. Nor is the following track buffered. So in Euphony/HQPlayer's case, disconnecting from the network is not possible.

Would be good if Euphony could find a way to automatically disconnect/shutdown LAN card/traffic whilst playing songs. Would be nice feature to have. 

[hols]

18 hours ago, romaz said:

 

Thanks for explaining.  If you are going to compare an HQP setup (intended for DSD upsampling) versus a machine like the Extreme that is not set up at all for HQP upsampling, then the results with be expectedly different.  If you really like the sound of DSD 256 ASDM7EC, then I would expect that your server that gives you this is going to sound better than a server that doesn't.  Your post really isn't about the Extreme versus your DIY build but rather your preference for DSD upsampling versus NOS. 

WOW!!  What a surprise. So many comments in such a short while. And its really great to learn so many things from you guys here. 

 

Let me try to explain a bit more on how I did my comparisons. First I was using Denafrips Terminator Plus during the comparison. My extreme(using jplay and connected to network with optical fibre) is sitting on the top platform and my server(using HQplayer ASDM7EC dsd 256 and connected to NAA computer through the audio switches) is sitting on the rack just below. Both are optimized with their best power cords and USB cables playing same piece in their local nvme storage. After playing from my server then we change the USB cable to the extreme to play the same piece. So basically both extreme and my server are playing in their own most optimised state and connected to same DAC and amplifier system. I don't find anything wrong with such an arrangement. We are looking for the same musical qualities coming from the same piece of music. Listening to music is a subjective thing at the last moment. And you are just using a critical listening rather than doing it for pleasure.  If you are saying that one cannot compare a bit perfect player with a DSD upsampling player, then would you be saying that you should not compare a tube amplifier with a solid state amplifier? If you don't compare how can you decide what to buy? Any comparison is solely applicable to the system at that moment. It may or may not be applicable to other systems. 

 

While I don't agree to the idea that it is inappropriate I can understand that some may like to narrow down the comparison to a more like to like state. To this end I can only try to install HQplayer onto extreme and try use PCM upsampling in both servers and have a more apple to apple comparison. The other way round is not possible because my server is solely designed for HQplayer embedded and not suitable for doing other things. This is going to take a few days because my Terminator plus is away in another place and these few days I am using the Chord Dave with and without Blu2 and I need to install driver to extreme which to my understanding is not as easy as installing in my desktop computer and need to try out the different settings in HQplayer4 desktop. I will surely report back when it is done.

 

Just another thing I can relate is about PCM upsampling. Although I don't deny I am more inclined to dsd yet I have also explored PCM upsampling with HQplayer. As some of you may know that Denafrips Terminator or Terminator plus can do PCM1536. And I have tried to upsample to PCM1536 with HQplayer and the sound is surprisingly analogue and warm and is in a different league compared with PCM768. And our impression then was the sound of DSD1024 and PCM 1536 are very similar and is it possible that when the resolution gets higher and higher the difference between DSD and PCM becomes minimal. Will need more PCM guys input on this point. Cheers.

 

 

[austinpop]

46 minutes ago, hols said:

To this end I can only try to install HQplayer onto extreme and try use PCM upsampling in both servers and have a more apple to apple comparison.

 

This would be a very pertinent comparison, well within the capabilities of both machines.

 

46 minutes ago, hols said:

And our impression then was the sound of DSD1024 and PCM 1536 are very similar and is it possible that when the resolution gets higher and higher the difference between DSD and PCM becomes minimal.

 

Very interesting point, and one that I've wondered myself. If so, it also gives a path forward (with PCM1536) that does not require heavy lifting in terms of CPU utilization.

[Nenon]

11 hours ago, austinpop said:

This is now my preferred mode of listening. Cue up a local album in HQPlayer, hit Play then physically disconnect the network. For good measure I quit JPLAY and Roon Server. And no second guessing about the effect of the upstream network.

 

It's amazing, isn't it?

This is why I lost interest in network tweaks recently. But that recently tweaked Buffalo switch brought up some new sparks.

 

Have you tried the BMC module yet? I never got to test whether unplugging the cable was an additional improvement over disabling the NIC in Windows and using the BMC interface for management. 

 

I still believe there are two completely different ways the network harms our server quality:

1. The electrical noise that gets transmitted over the network cable. That can be significantly reduced by using fiber optic isolation, but then the devices that convert fiber back to copper also generate some noise.

2. The effects that processing network traffic has on your server. We get too focussed on #1 and often miss this important point. Every packet that arrives on the network needs to be processed by our server. That process, even if it is relatively small to everything else the server is doing is interrupting the server and using valuable resources.

 

Unplugging the network cable solves both #1 and #2.

Leaving the cable plugged in, disabling the NIC, and using the out of band management access (BMC) solves just #2.

It's an interesting test I forgot to do.

 

While on this topic, I just received an interesting report from an Extreme user that replacing the DC cable on his switch had a quite audible effect on playing local music. I deliberately highlighted the word local. We are not talking about streaming Tidal/Qobuz but local music stored on his Extreme NVME storage. His has a dedicated fiber Internet line just for his Extreme and the laptop that controls it. And the laptop is isolated with sonore optical module powered by battery power supply. I find this quite interesting! It's obvious that network devices modulate the noise. And if a DC power cable for your switch can modulate the noise on the switch, then it all goes back to your system. Fun, fun, fun...

 

Lastly, I appreciate @hols's attempts to cheer up our DIY efforts and give some hopes that our DIY efforts might be as good as the Extreme. Sorry to tell you that, but no, we are not even close. My best efforts are pretty close to an Extreme server running Euphony. But when you add all the Taiko software tweaks, the Extreme is on another level than anything else I have heard. 

The good news is that Emile is considering creating some products for the DIY market instead of developing an Extreme mini. I really hope that happens. It must be a hard decision for him, so I will encourage everyone who likes the idea to hit the LIKE button on his post:

https://whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-235#post-667366

That would help him gauge the interest and make a decision. 

[Exocer]

36 minutes ago, Nenon said:

The good news is that Emile is considering creating some products for the DIY market instead of developing an Extreme mini. I really hope that happens. It must be a hard decision for him, so I will encourage everyone who likes the idea to hit the LIKE button on his post:

https://whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-235#post-667366

That would help him gauge the interest and make a decision. 

 

This is fantastic...The DIY community will benefit greatly from this decision and I do not think it will affect the TAE sales that much. A win win for everyone IMHO.

[matthias]

On ‎8‎/‎17‎/‎2020 at 7:44 PM, Nenon said:

The good news is that Emile is considering creating some products for the DIY market instead of developing an Extreme mini. I really hope that happens.

 

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-238#post-668057

 

Matt

[ASRMichael]

Has anyone done this? Or have their setup this way. 
 

 

[edwardsean]

I was really looking forward to the Extreme Mini. It is disappointing to hear that it might not be released. Still, Emile has to do what is best for his company, of course. 

 

That is an amazing problem to have: that is, it might be too successful and overwhelm their ability to provide quality service to all their customers!

 

It says a lot about what he has accomplished and his dedication to maintain such stellar standards. 

 

 

[MarkusBarkus]

...before ditching the TV, since it looks like he has to move an acoustic panel to watch anyway, I would have tried a light drapery over the TV re: reflections.
 

I have my gear between/slightly behind my speakers. I would prefer not to, as it is clearly a high pressure and vibration zone. Seems like it could impact imaging, no?

 

For me, the trade was *much* longer cabling and very awkward gear placement in the room.

And whatever I can do for vibration control (CF platforms, footers, wooden shelves).

 

My top choice would be to pass a large conduit through the wall: gear on one side, speakers on the other. 
 

I built the room, so that would have been easy to do. I deemed the additional expense of a divorce attorney to be cost prohibitive, so the plan was altered to current config.

[drjimwillie]

Are used to be an architect and over a ten-year period we built, and I designed all of the radio facilities in New York City. Our tagline was humanizing the media workspace. Davy Jones of the Monkees said that my WPLJ studio was the best feeling studio he had ever been in. I was fortunate to work with a really brilliant acoustic engineer. Once when hiding some big box speakers in a soffit he told me to align the face of the speaker with the plane of the soffit. Doing this essentially makes the adjacent soffit part of the cabinets reflective surface. 
in the picture below I have used the face of the fireplace as part of the speaker system. 
until about two weeks ago I had all of my gear tucked behind the left speaker. If you recall recently I showed a picture of the mahogany rack I built. Here is a picture of it in the basement, below the speakers. 
moving my gear to the basement has allowed me to separate the pieces into three racks. One rack for my power supply equipment one rack for my network equipment and the mahogany rack with the azure plinths for my gear. 
because of the nature of this quadruple isolation upgrade I cannot tell you the effect of each modification but cumulatively it is stunning. The sound is much quicker and the music is much more relaxed. 
I hooked up an infrared repeater so I can use my remote with my preamp. 
my friend said now I really do have a man cave for my headphone listening. 

[hols]

As promised earlier I have done a comparison of my server with my extreme on PCM oversampling with the Chord Dave in the past few days. If you like a short answer, yes , the extreme is the winner in this comparison. And if you don't mind spending some more time please read on because many a time it is the details that matters. I narrowed it down to use 705.6 and sinc-M filter in the comparison. I have also taken off my SOtM Tx-USB Ultra and also narrowed down to use only one SOtM switch(taken off the Buffalo switches). This can eliminate some multiple switch bias and it is also easier for the Dave to handshake with the NAA. Sometimes there is difficulty in handshaking with the Dave during change of server too rapidly and may need multiple attempts to turn on and off the NAA.

 

First one to come up is my server in PCM setting as above. The sound is faster than what I am used to have and the attack a little brisk. the soundstage detail is more pinpoint. The image a little smaller than before. The overall musicality not bad except mid range is still a little thin and bright. Next server on is the extreme with HQplayer alone. The sound I would say is not much different from my server. If you want to split the hair then the low frequency is lighter than my server and the high frequency a bit more beautiful. During the attacks it is usually the mid range that comes off more prominently and less well backed up by low frequency. Actually when I installed my HQplayer on to extreme there are warnings that some of the processes are not carried out so I approached Emile and see if is that a problem. And he kindly helped to assure me that it is OK and he further offered to help me install a NAA in the same extreme computer making use of the two CPUs available. And I certainly would like to give it a try.  So next up is the extreme HQ+NAA. And that's a WOW. The performance just leaps to another level. The presence , the focus as well as the soundstage all goes to another level. the string quartet now just sits right in front of you. The improvement in orchestral piece is not as obvious in the chamber music or jazz or unaccompanied voice. But it is the clear winner.  This confirms my experience of using a NAA together with HQplayer is always beneficial and a NAA using optimised  computers like the extreme actually helps tremendously to boost the SQ.  It has all along been my experience that the better computer used as NAA the better is the sound. (I understand that this may not be agreed by all).

 

After testing with Chord Dave I have also changed to DAC to my old Terminator not plus ( Terminator plus is temporarily away in friend's place) with a view to test result with different DAC and also get a glimpse of what PCM 1411.2 can sound. ( I only use the multiple of 44.1 for lighter CPU consumption). The result is basically similar to with Chord Dave but with the more sumptuous bass from Terminator my server seems to take a lead over the extreme HQplayer alone. The attacks are more sumptuous and well supported. The extreme HQ+NAA still leads by a wide margin especially in the flow of music and the cozy feeling. What about PCM1411.2 The sound is in a class of its own. The sound is smooth without edges(which are sometimes evident in poorly executed PCM). The instruments have body and the nuances are clearly revealed. It is not so much as the same as DSD which still gives a more sumptuous body and well harmonized soundstage. But the PCM1411.2 is best described as giving a perfect flow of music.  The details are more and the presentation are a bit different  from 705.6. You see more of individual instrument's body and can also sense the layering more in the orchestra but you may not be able to feel a block of harmonized sound. Or maybe it is another way to say that you may need to take time to get accustomed to listen to so many detail.  

Hope when the COVID-19 becomes more stable we can arrange more people to audition this PCM upsampling sound when my Terminator Plus is back. Will give an update if we can arrange a group audition.

[austinpop]

Very interesting findings @hols. Thank you very much for going back and doing this comparison.

 

I am also enjoying HQP+NAA on the Extreme, and there is indeed an SQ bump from separating the noisier (relatively, since PCM upsampling is very light) HQPlayer process doing upsampling and the NAA doing rendering onto different processors. This is just another example of why having 2 physical NUMA nodes (aka processors) enables a raft of optimizations that were just not possible on my single processor custom server.

 

I'm intrigued by the benefit of PCM1411.2/1536. Did you compare just PCM705.6 vs. PCM1411.2 — holding everything else constant? It sounds like you did. And that suggests that there is a benefit to DACs going to this higher speed in future versions.

[ted_b]

Hols, what is your NAA computer?  And will you report back when you do the NAA daemon on the other Extreme cpu instead?

 

Also, does anyone have HQP V3 on an Extreme and can compare to a reinstall of V4 (being that V3 runs on both cpus and V4 runs, currently, on only one)?  I realize auditory memory can be fleeting, especially after the time it takes to reinstall V4.

[hols]

22 hours ago, austinpop said:

I'm intrigued by the benefit of PCM1411.2/1536. Did you compare just PCM705.6 vs. PCM1411.2 — holding everything else constant? It sounds like you did. And that suggests that there is a benefit to DACs going to this higher speed in future versions.

Yes, all others are constant during comparison of 705.6 and 1411.2  I think there is some very worthwhile benefit in going to 1411.2 but it probably needs some tweaking (maybe different filters) before its full potential can be confirmed.

[hols]

21 hours ago, ted_b said:

Hols, what is your NAA computer?  And will you report back when you do the NAA daemon on the other Extreme cpu instead?

 

Also, does anyone have HQP V3 on an Extreme and can compare to a reinstall of V4 (being that V3 runs on both cpus and V4 runs, currently, on only one)?  I realize auditory memory can be fleeting, especially after the time it takes to reinstall V4.

My comment on the NAA computer also affects the SQ comes from my past experience. I have three NAA computers using similar Asrock motherboards Pro4 but three different CPUs , Intel 8700T, 9700K, and 10700 and I have compared the three  with the same HQplayer computer and the SQ is in the same ascending order from 8700T to 9700K to 10700, clearly audible though not night and day difference. And I am quite convinced that the NAA installed in extreme would result in an excellent SQ.

It is unlikely that I will conduct any more further NAA computer comparison because of my past experience explained above.

[ted_b]

1 hour ago, hols said:

My comment on the NAA computer also affects the SQ comes from my past experience. I have three NAA computers using similar Asrock motherboards Pro4 but three different CPUs , Intel 8700T, 9700K, and 10700 and I have compared the three  with the same HQplayer computer and the SQ is in the same ascending order from 8700T to 9700K to 10700, clearly audible though not night and day difference. And I am quite convinced that the NAA installed in extreme would result in an excellent SQ.

It is unlikely that I will conduct any more further NAA computer comparison because of my past experience explained above.

Sorry I didn’t look up your previous posting experience. I will do a search to find out more about your NAA (OS, etc)

 

edit: your build post doesn’t say whether your NAA is also Linux-based, like your non-Extreme server build

[Gavin1977]

On 8/21/2020 at 4:55 PM, hols said:It is not so much as the same as DSD which still gives a more sumptuous body and well harmonized soundstage.

Great feedback @hols  what do you mean by the above sentence... do you mean DSD still sounds better than PCM upsampled at the same rate?  Is the difference large enough to warrant the huge processing requirements for DSD?

 

I ask because I run T+A DAC 8 DSD, I can’t do NOS PCM, and am presently CPU limited (like most) to DSD256 for demanding filters.

 

Elsewhere a report of someone preferring straight PCM: 

 

[kennyb123]

Ok, I’ll admit it: I have a problem when it comes to fuse upgrades for my Spectral amp.  Moving to the latest color from Synergistic has always been a worthwhile move that easily justified the price.  Over the years the progression has been: stock fuses, HiFi Tuning Supreme, SR Red, SR Black, SR Blue, and now SR Orange.

 

Each of those steps had brought worthwhile gains.  What each of the steps leading up to the Blue had in common was a reduction in harm that the fuse itself was causing by overlaying a signature onto the music.  Each fuse’s signature didn’t become evident until the next fuse showed up.  There was always this surprise that the prior fuse wasn’t as transparent as it has seemed to be when it replaced its predecessor.  I had thought, for example, that the Black had totally vanished as it seemed that way relative to the Red.  But then the Blue showed up and made it evident that the Black had been overlaying some harm onto the music.

 

I wondered if it was finally safe to say that the Blue totally vanished as I couldn’t imagine how SR could advance things even further.  But then the positive reports started rolling in on the Orange fuse, so I figured I had to eventually take the plunge.  I swapped the Orange fuses into my amp a week ago but have only been able to get my arms around their contribution the last couple of days.  The portable AC that I was forced to use in my listening room was a bit too noisy.  At present it’s a lovely 73 degrees with no AC thanks to cooler outdoor temperatures.

 

It turns out that I might have been partly right about the Blue in that it had vanished - at least as far as overlaying harm onto the music.  With the Orange, I hear no evidence that a signature had been stripped away.  What I do hear though is a very nice step forward in openness and naturalness that’s well worth the price I paid (I took advantage of the buy two get one free promotion). “More organic” is how I described the improvement Orange fuses brought to each of my Super3 power supplies.  I hear that with the substitution of Orange in my amp too.

 

I’ll summarize all this by saying that if you are using any fuse less than a Blue, it’s likely doing harm and you should audition either a Blue or Orange.  If your component has a Blue fuse now, it’s in a really good place. But if you really liked what you heard from the Blue, it’s very likely you’ll like an Orange even more.

[sig8]

I am considering T+ and Gaia, but based on @hols reporting above if higher speeds in PCM and perhaps in DSD sound better, then Gaia will limit you to DSD512, and PCM to 768 kHz. Don't know if T+ at higher speed is better than T+ and Gaia combo at lower speed. Thoughts? Experience?

[HeeBroG]

On 8/22/2020 at 12:21 AM, austinpop said:

Very interesting findings @hols. Thank you very much for going back and doing this comparison.

 

I am also enjoying HQP+NAA on the Extreme, and there is indeed an SQ bump from separating the noisier (relatively, since PCM upsampling is very light) HQPlayer process doing upsampling and the NAA doing rendering onto different processors. This is just another example of why having 2 physical NUMA nodes (aka processors) enables a raft of optimizations that were just not possible on my single processor custom server.

@hols and @austinpop

 

Hi guys,

 

Can you please elaborate to a computer nube like me what you mean by an NAA on the Extreme?

Is this another program to install when using HQP4?

 

Cheers,

 

G

[ted_b]

NAA (Network audio adapter) is an HQPlayer audio daemon (application) that resides on a different box than the hard-working HQPlayer.  Its intent is to isolate the dac interface from all that heavy lifting, and thereby reduce noise.  All that runs on the NAA box is the DAC interface (driver).  In the case of the Extreme, and the fact that, currently, HQPlayer V4 only occupies one of the Extreme cpus, the thinking is to install NAA on the OTHER cpu....thereby virtually creating a networked HQPlayer-to-NAA environment.

[hols]

23 hours ago, ted_b said:

Sorry I didn’t look up your previous posting experience. I will do a search to find out more about your NAA (OS, etc)

 

edit: your build post doesn’t say whether your NAA is also Linux-based, like your non-Extreme server build

Sorry I haven't posted about previous experience about NAA. I only mean that I have already had an experience of comparing different NAA computers. My NAA are Linux -based.