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

[Hauser]

Although I haven't had much to contribute to CA, I have nevertheless found the threads very helpful, especially those of Romaz.  He has shown a deep commitment to research and posting his results here.  Reading that he will no longer be posting is a significant loss to the community.  Thankyou for all the effort and time you have contributed Roy.

 

Martin.

[TopQuark]

To those going back to the CAPS type set-up, is it because you prefer Windows over Linux?  Is that the only reason?  For Roon, I thought it is a lot more convenient to run the trifecta with ROCK.  If one wants to bypass the renderer because it is likely a Linux box (e.g. sMS-200ultra), then the Windows server is burdened by doing 2 things at the same time, i.e. Roon + Renderer.  If one switches the DSP, room correction, and upsampling in Roon, then that is even worse.

 

If not running Roon or non interested in other digital enhancements, then I can understand. But I thought Romaz runs his system with Roon(?)

 

[ElviaCaprice]

53 minutes ago, TopQuark said:

 

 

If not running Roon or non interested in other digital enhancements, then I can understand. But I thought Romaz runs his system with Roon(?)

 

Huh?  Why would you want to upsample in Roon if you got the best upsamplers in Chord?  Why use a renderer if you have an sCLK-EX server (CAPS) mod?  You don't use Linux because it has limitations.  You don't bypass anything, it's server direct USB to DAC, doesn't get any simpler, can add a fixer if so inclined.   I think you need to go back and reread the last post of Romaz about his latest server setup.  It's all about the sCLK-EX and where you want to modify/add it in your system.

 

[austinpop]

3 hours ago, TopQuark said:

To those going back to the CAPS type set-up, is it because you prefer Windows over Linux?  Is that the only reason?  For Roon, I thought it is a lot more convenient to run the trifecta with ROCK.  If one wants to bypass the renderer because it is likely a Linux box (e.g. sMS-200ultra), then the Windows server is burdened by doing 2 things at the same time, i.e. Roon + Renderer.  If one switches the DSP, room correction, and upsampling in Roon, then that is even worse.

 

If not running Roon or non interested in other digital enhancements, then I can understand. But I thought Romaz runs his system with Roon(?)

 

 

I think we need to all step back here, and realize that what Roy is describing is his journey, and his choices. I think he would be the first to agree that he's hardly handing down the ten commandments on stone tablets.

 

Just because he eventually chose to abandon the network endpoint approach, it does not negate that approach! Some people have a tendency to think of this as a zero sum game, where if one approach is right, the other must be wrong. It doesn't work that way.

 

In fact, back when Roy first talked about his decision to switch to a direct server approach, he wrote:

 

On 7/2/2017 at 7:52 PM, romaz said:

It is sometimes hard to fully convey in written words the extent of how much I like something and so I'm sure there are times when people think I like something to a greater degree than I intended.  With straight USB vs the sMS-200 stack, it's not good vs horrible, it's more my personal preference as to which compromises I'm willing to live with.   Having lived with your sMS-200 Ultra stack for a while now, you can attest to how wonderful it is.  If I had chosen to remain with that stack, trust me, absolutely no one should feel sorry for me because that is one incredible trifecta, especially for the type of music I listen to, and so I would never discourage anyone from going down that route nor would I disparage anyone who is already there.  For those who feel they want more meat on the bones on their music than what this stack provides, I would also suggest adding the IR somewhere in that chain.  If there is no intention to address the motherboard, it's actually my belief that going down the sMS-200 or mR route is the better way to go.

 

I discussed it further with him at RMAF, in the context of where I go next from my trifecta medley. In no way did he suggest I abandon the network approach and adopt his direct approach.

 

If you are already on a direct USB configuration, then Roy's current approach will be very attractive, either using his build, or @ElviaCaprice's NUC approach, or other variants. If.on the other hand, you prefer to go the network endpoint route, I still maintain the "trifecta" of modded switch, modded sMS-200, and tX-USBultra is a tremendous bang for the buck. Of course you can take this to its extreme like Roy and @Bamber have done, where you have sCLK-EX end to end. The latter is undoubtedly incredible SQ, but a very expensive path, and not in everyone's reach.

 

To quote Larry - there is more than one way to skin this cat.  

[austinpop]

On 10/9/2017 at 5:52 PM, austinpop said:

Darn it, @lmitche and @unbalanced output, now you've got me thinking.

 

The last time I tried HQPlayer was before the clock chains, when I was on the sMS-200. I should try it again, to confirm for myself what I'm hearing now. But before I do, a couple of questions.

 

What percent of your listening is Redbook content? In my case, I may be atypical. While my library is about 70% Redbook, my actual listening tends to be <50% Redbook. I gather the benefits are most pronounced with lower-res tracks?

 

Also, Larry, regarding this:

 

 

Last time I tried it, I remember 4 parameters (none, none, none, none) that you had to manually tweak via pull-downs to vary the settings. Can you tell me how (maybe via PM) one can set up HQP to auto-detect PCM or DSD tracks, and conditionally upsample PCM tracks to 384, and DSD tracks to 128, without intervention?

 

I'm not sure I can download another trial version of HQP, without having to buy it.

 

Here is an update on my revisited evaluation of HQPlayer for my use. My original (pre-trifecta) evaluation was in this post on another thread:

 

This time, I was able to run another evaluation with the latest 3.18 version of HQPlayer. The difference is I am now on a far more resolving digital chain with far better SQ. Would this enable me to hear differences that I didn't previously deem significant? I'd like to thank @lmitche and @auricgoldfinger for their help with optimizing my settings, and general HQPlayer navigation.

 

Since this is such a well-known and well-liked product, I am going to cut to the chase - I am going to buy myself a copy of HQPlayer.

 

Key facts:

1. Recall my DAC does DSD over DoP, so supports only DSD128, while supporting PCM384
2. My music listening is > 50% hi-rez (DSD or 24/44.1-384)

Summary of Findings:

1. DSD64 tracks: no benefit, either upsampled to DSD128 or PCM384
2. 24-bit tracks: the SQ benefit by upsampling to PCM384 seems to be inversely related to sample rate. So 24/44.1 benefits the most, while 24/384 the least. That seems logical. Also, upsampling to DSD128 was not as pleasing. It seemed like all the edges had been rounded off. Polite, but not engaging.
3. Redbook 16/44.1 tracks: universal improvement across the board, but the magnitude of improvement varied by track.

Overall, I'd still say the improvements were modest, not earth-shattering. Still, for the very reasonable cost, it is a no-brainer. Until I switch DACs, I can accommodate the above findings by setting all PCM tracks to upsample to PCM384, while checking DirectSDM to bypass upsampling for DSD.

[Johnseye]

7 minutes ago, austinpop said:

 

Here is an update on my revisited evaluation of HQPlayer for my use. My original (pre-trifecta) evaluation was in this post on another thread:

 

This time, I was able to run another evaluation with the latest 3.18 version of HQPlayer. The difference is I am now on a far more resolving digital chain with far better SQ. Would this enable me to hear differences that I didn't previously deem significant? I'd like to thank @lmitche and @auricgoldfinger for their help with optimizing my settings, and general HQPlayer navigation.

 

Since this is such a well-known and well-liked product, I am going to cut to the chase - I am going to buy myself a copy of HQPlayer.

 

Key facts:

1. Recall my DAC does DSD over DoP, so supports only DSD128, while supporting PCM384
2. My music listening is > 50% hi-rez (DSD or 24/44.1-384)

Summary of Findings:

1. DSD64 tracks: no benefit, either upsampled to DSD128 or PCM384
2. 24-bit tracks: the SQ benefit by upsampling to PCM384 seems to be inversely related to sample rate. So 24/44.1 benefits the most, while 24/384 the least. That seems logical. Also, upsampling to DSD128 was not as pleasing. It seemed like all the edges had been rounded off. Polite, but not engaging.
3. Redbook 16/44.1 tracks: universal improvement across the board, but the magnitude of improvement varied by track.

Overall, I'd still say the improvements were modest, not earth-shattering. Still, for the very reasonable cost, it is a no-brainer. Until I switch DACs, I can accommodate the above findings by setting all PCM tracks to upsample to PCM384, while checking DirectSDM to bypass upsampling for DSD.

 

Are you using the xtr filter?  What are your settings? 

 

I found this interesting from Miska. 

 

"I created it for people who like Chord Dave -style filters. It is about 5 times longer than 
poly-sinc and thus about 5x heavier. Offered in linear- and minimum-phase variants. Stop-band 
attenuation exceeds needs of 40-bit PCM, being over -240 dB. Normal poly-sinc filters are created to reach excess of 32-bit PCM resolution (> -192 dB) stop-band attenuation."

[austinpop]

8 minutes ago, Johnseye said:

 

Are you using the xtr filter?  What are your settings? 

 

I found this interesting from Miska. 

 

"I created it for people who like Chord Dave -style filters. It is about 5 times longer than 
poly-sinc and thus about 5x heavier. Offered in linear- and minimum-phase variants. Stop-band 
attenuation exceeds needs of 40-bit PCM, being over -240 dB. Normal poly-sinc filters are created to reach excess of 32-bit PCM resolution (> -192 dB) stop-band attenuation."

 

Yes, I used xtr. I didn't experiment with filters, but I will eventually. I used the recommendations I got. For this eval, the relevant settings were:

 

 

and 

 

[mozes]

28 minutes ago, austinpop said:

Overall, I'd still say the improvements were modest, not earth-shattering. Still, for the very reasonable cost, it is a no-brainer.

The interface is what keeps me away from HQPlayer.  I am used to the convenience of Jriver, but I always keep my eyes open. Have you tried no upsampling to check if HQplayer is still sonically superior playing files at native sampling rates?

[Johnseye]

8 minutes ago, mozes said:

The interface is what keeps me away from HQPlayer.  I am used to the convenience of Jriver, but I always keep my eyes open. Have you tried no upsampling to check if HQplayer is still sonically superior playing files at native sampling rates?

 

You can use Roon with HQP. No need to use its native interface. 

[ElviaCaprice]

I despise over paying or even paying for software.  The only one I buy for bit perfect audio, great library and video capabilities, JRiver.  And that's only every two or three releases for $16 if you catch the pre-release offer (initial 1st time buy is $50, I think, it was many moons ago). 

Windows 10 - free, you don't need it to be licensed to work properly.   Everything you need to optimize Windows 10 can be done free on your own. 

[mozes]

4 minutes ago, Johnseye said:

 

You can use Roon with HQP. No need to use its native interface. 

Yes that’s a good alternative, but I did A/B testing a while back between Roon and Jriver and Jriver was clearly more transparent sounding to my ears. Of course I didn’t try Roon with HQP.

[mozes]

6 minutes ago, ElviaCaprice said:

I despise over paying or even paying for software.  The only one I buy for bit perfect audio, great library and video capabilities, JRiver.  And that's only every two or three releases for $16 if you catch the pre-release update (initial 1st time buy is $50, I think, it was many moons ago). 

Windows 10 - free, you don't need it to be licensed to work properly.   Everything you need to optimize Windows 10 can be done free on your own. 

Jriver is a great value!

[ElviaCaprice]

Mozes, it doesn't matter, your never going to use HQP having a DAVE.   Nor am I within the Chord family.   If I was going to use HQP, I would think about DAC's that have been shown to work best with it.  Not the other way around.  Let alone the entire system would be redesigned around it.  Mobo, etc.

  It is an option and I still say that the sCLK-EX server would be a huge benefit to a HQP system.  I await for someone to give this a go???

[Johnseye]

I own JRiver as well and have for... Many moons. A lot to like. You can try Roon with HQPlayer for free. At least you'll know. 

[mozes]

1 minute ago, ElviaCaprice said:

Mozes, it doesn't matter, your never going to use HQP having a DAVE.   Nor am I within the Chord family.   If I was going to use HQP, I would think about DAC's that have been shown to work best with it.  Not the other way around.

Even with my Brooklyn, I used to prefer no upsampling. I felt that playing the music natively has more of an effortless flow.

I also like my PC not to be pushed hard, I belong to the low power low noise school  

[tedwoods]

12 hours ago, unbalanced output said:

However, the Lampi in DSD is a completely different that I could describe as taking a big screen off the front of the speakers.

I agree.

What I gathered from listening through a wide variety of systems was that the format/player/DAC  topology has to be viewed as a chain of sorts.

DSD through the chipless Lampi has an uncanny ability to present things in a most natural and intuitive way.

With the SOtM chain this trait has been elevated to almost frightening yet utterly exciting levels of palpability.

[TopQuark]

7 hours ago, ElviaCaprice said:

Huh?  Why would you want to upsample in Roon if you got the best upsamplers in Chord?  Why use a renderer if you have an sCLK-EX server (CAPS) mod?  You don't use Linux because it has limitations.  You don't bypass anything, it's server direct USB to DAC, doesn't get any simpler, can add a fixer if so inclined.   I think you need to go back and reread the last post of Romaz about his latest server setup.  It's all about the sCLK-EX and where you want to modify/add it in your system.

 

Huh? Not all of us are using Chord. The renderer, sMS-200ultra in this case, includes the sCLK-EX.  The sMS-200ultra is the PC running low power ARM in Linux.  The ROCK is just needed to feed the renderer or one can just attach the drives directly to the sMS-200ultra and skip the ROCK or a NAS.  So in essence, it is a direct USB to DAC set-up.

 

What I was asking is a general question.  Barring any sCLK-EX that can be added in any set-up, if using Roon, I don't see any reason why anyone want to go with the CAPS set-up except for the OS.  Even if the upsampler or other exotic digital enhancements are not used in Roon, the library database assembly alone can affect rendering the music if only a single box is used.

 

 

 

[TopQuark]

5 hours ago, austinpop said:

 

I think we need to all step back here, and realize that what Roy is describing is his journey, and his choices. I think he would be the first to agree that he's hardly handing down the ten commandments on stone tablets.

 

Just because he eventually chose to abandon the network endpoint approach, it does not negate that approach! Some people have a tendency to think of this as a zero sum game, where if one approach is right, the other must be wrong. It doesn't work that way.

 

In fact, back when Roy first talked about his decision to switch to a direct server approach, he wrote:

 

 

I discussed it further with him at RMAF, in the context of where I go next from my trifecta medley. In no way did he suggest I abandon the network approach and adopt his direct approach.

 

If you are already on a direct USB configuration, then Roy's current approach will be very attractive, either using his build, or @ElviaCaprice's NUC approach, or other variants. If.on the other hand, you prefer to go the network endpoint route, I still maintain the "trifecta" of modded switch, modded sMS-200, and tX-USBultra is a tremendous bang for the buck. Of course you can take this to its extreme like Roy and @Bamber have done, where you have sCLK-EX end to end. The latter is undoubtedly incredible SQ, but a very expensive path, and not in everyone's reach.

 

To quote Larry - there is more than one way to skin this cat.  

 

I can understand Roy's journey.  What I cannot understand is the use of only 1 box for library assembly and rendering unless he decided to move away from Roon.  As I understand from his previous post, he was using Roon.  So it remains that, if he is still using Roon, that only means library assembly that uses a good amount of work on the CPU is not affecting rendering.  I wonder if this was the conclusion.

 

Edit:  He mentioned about the use of the great tX-USBexp.  I'm not sure if he went to 1 box just for this card but I have my doubts that SOtM designed the $1,200 sMS-200ultra to be inferior to the $350 tX-USBexp.

 

[ElviaCaprice]

53 minutes ago, TopQuark said:

 

Huh? Not all of us are using Chord. The renderer, sMS-200ultra in this case, includes the sCLK-EX.  The sMS-200ultra is the PC running low power ARM in Linux.  The ROCK is just needed to feed the renderer or one can just attach the drives directly to the sMS-200ultra and skip the ROCK or a NAS.  So in essence, it is a direct USB to DAC set-up.

 

What I was asking is a general question.  Barring any sCLK-EX that can be added in any set-up, if using Roon, I don't see any reason why anyone want to go with the CAPS set-up except for the OS.  Even if the upsampler or other exotic digital enhancements are not used in Roon, the library database assembly alone can affect rendering the music if only a single box is used.

 

 

 

Sorry to have been so rude in my first reply.  Most of the answers you seek have already been addressed.   Yes, the renderer is the low power server, almost, that is limited by proprietary software.  Problem is that the data being streamed to the renderer seems to still have a qualitative effect on SQ, whether from another server, router or as you mentioned hard drive.  If you are capable and fine with the sCLK-EX renderer and it's limitations, then by all means stick with it as Austin pop points out above.

 

Roon is not an OS.  So either you go with Linux or Windows.  Roon, which I don't use, to my knowledge works fine on a server only Windows OS.  Thus you would enjoy the benefits of a custom DAC Windows driver which as Roy pointed out sounds better than the generic Linux driver used on the Renderer.  In his case that would make the Server direct to DAC via USB a more favorable system.

 

I've never encountered a SQ effect with a library database on a single server.  But I use JRiver and it is not an issue.  Track selections are loaded into memory for playback.  Remember, just because you can use a dual setup with your renderer, doesn't mean you negate all the effects of the damage of the audio stream from the server.

 

So in theory one still needs to address the server, which is what I have done to an even greater degree than a renderer, by using 3 of the sCLK-EX points on the low power mobo and the 4th point on the attached tXUSBexp PCIE usb card.  The theory of KISS comes into play here.  By dealing with the problems of the original streaming source I am nullifying the need for the fixers down stream and I am not pushing out the problems further by using a NAA or renderer.  If the problems of clocks and power can by solved at the point of origination, then KISS comes into play making the sCLK-EX server via usb to DAC a much more simple system that has equal or better SQ to a more complex system.  Let alone the cost saved in such a simple system, flexability of software usage, and universal format capability.

[octaviars]

1 hour ago, ElviaCaprice said:

So in theory one still needs to address the server, which is what I have done to an even greater degree than a renderer, by using 3 of the sCLK-EX points on the low power mobo and the 4th point on the attached tXUSBexp PCIE usb card.

 

Perhaps you have written this somewere but it is hard to find information in this large thread about specific systems so my question is what mobo do you use?  

[sig8]

Rajiv: You can use JRiver as front end for HQ Player by using Geoffrey's method. Works well, and easy to implement.

 

 

[ted_b]

and therefore JRemote too.  And you can set up HQPlayer/Jriver to change filters/settings based on music too (i.e genre, etc).  The setup looks a little daunting but in fact is quite simple once you get the pieces.

[Johnseye]

12 minutes ago, sig8 said:

Rajiv: You can use JRiver as front end for HQ Player by using Geoffrey's method. Works well, and easy to implement.

 

 

 

Going to have to try this.  I hadn't realized it was even possible, thanks!

[ted_b]

Geoffrey began this little project as an experiment, and has consistently asked if more worthy script writers could take over and add functionality (and maybe an easier install) to this JRiver as-front-end-to-HQP idea.  But the main connector is JRiver's flexible file association setup (in tools->options-> file types). 

[auricgoldfinger]

9 hours ago, austinpop said:

I discussed it further with him at RMAF, in the context of where I go next from my trifecta medley.

 

What were Roy's suggestions for your next step?