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

[Dev]

16 minutes ago, bibo01 said:

This is the setting the developer of C-3PO (marcoc1712) uses for non-DSD:

-n squeezelite-R2-o hw:CARD=J20 -x -m 00:e0:4c:68:ce:78 -r 44100-768000 -c aif,wav,pcm -b 1048576:1048576 -a 499:3::0 -f /var/log/squeezelite-R2/squeezelite-R2.log

 

I do not want to divert traffic on another site, but you guys having problems with LMS/Squeezelite are really better off posting your problems (you can post in English) on Italian NextHardware, in this thread.

You will find many developers and users of Squeezelite, HQPlayer... and myself! 

 

@bibo01, thanks.

 

Quick question - will removing the C-3PO plugin disable upsampling/conversion of the original files ? I have few DSD256 content but my DAC supports only DSD128. How do I down-sample in this case ?

[Dev]

42 minutes ago, Bricki said:

Does anyone know how to resolve an issue with my audiolinux nuc running roon bridge but not appearing in roon when I have my core on audiolinux as well? If I run my core on sonicorbitor then I can see it and it works fine but then I boot my sonicTransporter into audiolinux and I can't find my roon bridge NUC. 

 

I am using different usb sticks to boot from (for server and bridge).

 

Anyone else had this problem? 

 

See if you can ping the NUC from sonicorbitor...it looks to be network issue. Raat protocol is generally very stable - I never had issue with Roon server recognizing a bridge on the network.

[Dev]

NUC8i7BEH update:

 

Earlier in the morning I set up HQP on the NUC to quickly test its performance. With xtr-mp filter and upsampling to dsd128, the cpu started throttling big time but it kept going without drop outs. Can't tell I could discernibly hear any difference when throttled. However, it looks like dsd256 could be a burst on this NUC. I will test this sometime later.

 

Secondly the fan is very irritating. My rack is 15ft away and I could hear it loud and clear.

 

Not sure how NUC8i7HVK would compare but it will be more noisy if you are planing to place it in the audio room. A Xeon E3 with passive cooling seems like a far better option as a server.

 

After google'ing around a bit similar performance was reported here:

https://www.notebookcheck.net/Intel-NUC-Kit-NUC8i7BEH-i7-8559U-Mini-PC-Review.360356.0.html

[Dev]

18 minutes ago, austinpop said:

The Roon Device for the squeezelite instance allows a max sample rate of PCM 192. I don't know if this is a hard limit set by Roon, of if there is any way to bump that up to PCM 384k.

 

Can you check if DSD is also limited to dsd64 if your DAC supported DSD ?

[Dev]

Roon Core/Roon Bridge vs LMS/Squeezelite vs Roon Core/Squeezelite:

 

With the bigger buffer size (I can only set to 1Gb since my NUC is fitted with 4Gb at the moment), the LMS/Squeezelite just sounds so much better than the Roon Core/Bridge combo. Bigger soundstage, more air around vocals, more of everything. This is without up-sampling anything. Just native resolution directly to the DAC. In all the test the server was NUC8i7BEH and NUC7CJYH as the streamer all running AL.

 

Later moving to RoonCore/Squeezelite combo (thanks to @austinpop), it sounds about the same as LMS/Squeezelite combo. With much better user interface of Roon and far better SQ of the squeezelite, you have the best of both the world but keep in mind PCM is limited to 192KHz and DSD to 64. Also, as @austinpop has noted earlier, sometime it skips to the next song. It didn't happen with LMS/Squeezelite combo.

 

I am yet to throw in HQP in the mix. My earlier experiment with HQP vs Roon, Roon again fell short, even without employing up-sampling but as far a audio memory serves, the LMS(or Roon)/Squeezelite is the best combo so far I have heard in my system.

 

[Dev]

11 hours ago, LTG2010 said:

I don't think its suitable as a DSD upsampling machine you'll need some serious power and these CPU's are mobile units., its better as a roon server etc with the smaller unit as the renderer. You could try the HQP settings without upsampling bit perfect ( some users prefer that) and the the smaller NUC as a networkaudio renderer.

 

DSD up-sampling is fine depending the sample rate. For example, dsd128 with xtr-mp work fine though cpu's were throttled as core temp was high. The -2s should be far less taxing. I haven't tried dsd256 but my gut tells it will do fine with -2s. Don't think it can handle dsd512 remotely.

 

11 hours ago, LTG2010 said:

You could also test it as a renderer with your PC as server it may outperform the smaller NUC, that comparison would be very useful. 

 

At some point I plan to do this as well, now that I have more or less settled on the music player s/w.

 

 

[Dev]

6 minutes ago, austinpop said:

 

Could you please clarify the various squeezelite versions? Is R2 a different branch? What are the benefits of using it vs. the Ralph Irving latest 1.9.0-1 version? Especially if you don't need:

• HQPlayer
• transcoding via C-3PO?

 

The v1.8.4 (R2) didn't seem to handle the -b option though the help specifies it. When I used that version with -b option of 1Gb, I didn't see any change in free memory. Roon vs Squeezelite sounded about the same and I thought what the heck is all this fuss about ? until I used the 1.9.0-1 from you.

[Dev]

 

22 minutes ago, beerandmusic said:

Nuc is from intel mfr....I have bought ASUS vm60s i5's for anywhere from $60 to $150 at most and that included mem and hd and windows os.  I stated why i prefer them to nucs already (sd slot, easier accessibility to hd, sdslot, spdif, 6USB slots and price)  Another reason i didn't mention is that since i have acronis images, i can have one image and it work on all of my vm60s in under 10min from scratch.  I have nothing against nuc's i just prefer asus vm60's for several reasons.  My only inquiry about nuc's was because i was curious if there was something unique to them where they may offer better SQ than an ASUS vm60.

 

Not sure if you are aware but Intel has been following this thread for a while before they released the Celeron NUC this year that many of us here like it for audio. Get one while you can ?

[Dev]

Just now, beerandmusic said:

 

Great to hear...maybe they will read my message and come out with a slightly larger footprint with a little more functionality (wink).  It won't be the first time manufacturers take suggestions from me.  I absolutely love the build of the vm60s...hope they pick one up...the removable top is awesome for those that swap ssd''s a lot.

 

Trust me, they are very very picky from whom they take suggestions from....

[Dev]

59 minutes ago, jean-michel6 said:

Thank you , very interesting results which confirm the results of Rajiv but in a dual pc setup. 

Were are located the music file library in your system ?

 

My setup in this regard is a bit unorthodox ? I use a mini-itx with JCAT Femto NET card and a 4tb 2.5" hdd attached, running AL and powered by LPS. This box acts a NAS as well as an audiophile switch. The JCAT ports are bridged and connected to the streamer and server.

 

[Dev]

On 11/23/2018 at 12:50 PM, Dev said:

Roon Core/Roon Bridge vs LMS/Squeezelite vs Roon Core/Squeezelite:

 

With the bigger buffer size (I can only set to 1Gb since my NUC is fitted with 4Gb at the moment), the LMS/Squeezelite just sounds so much better than the Roon Core/Bridge combo. Bigger soundstage, more air around vocals, more of everything. This is without up-sampling anything. Just native resolution directly to the DAC. In all the test the server was NUC8i7BEH and NUC7CJYH as the streamer all running AL.

 

Later moving to RoonCore/Squeezelite combo (thanks to @austinpop), it sounds about the same as LMS/Squeezelite combo. With much better user interface of Roon and far better SQ of the squeezelite, you have the best of both the world but keep in mind PCM is limited to 192KHz and DSD to 64. Also, as @austinpop has noted earlier, sometime it skips to the next song. It didn't happen with LMS/Squeezelite combo.

 

I am yet to throw in HQP in the mix. My earlier experiment with HQP vs Roon, Roon again fell short, even without employing up-sampling but as far a audio memory serves, the LMS(or Roon)/Squeezelite is the best combo so far I have heard in my system.

 

 

Roon/Squeezelite vs LMS/Squeezelite:

 

I decided to do some careful listening this morning between the Roon and LMS, in which Squeezelite was running on the streamer end. After back-n-forth with certain tracks, I preferred LMS. It has bit more airy sound compared to Roon which has slightly thicker color - this was done without any up-sampling and raw bits straight to the PSA DS. Later I decided to try on the Purestream which is optimized for DSD256 playback. Since Roon maxes out at dsd64, I could only compare it with LMS doing dsd256, both using clans-7 sdm filter. This is where LMS made a bigger difference than Roon. I think I will stick with LMS for now but I will dearly miss the usability of Roon - as a long time Roon user, LMS interface sucks big time.

 

All this while my server running AL in standard mode. While I didn't dare to run the server NUC in extreme mode (the stock fan cooling may not be up to it), switching to ramroot on the server made an unexpected incremental bump in SQ - I would say its not subtle and made bad recordings more enjoyable ?

[Dev]

1 minute ago, greenleo said:

I think it's you want to know if loading music to the memory is possible.  Technically, any access of local HDD is loaded into the RAM first. 

 

You want to avoid local HDD access during playback and hence avoid page faults as much as possible. The page faults increases latency. Playback s/w like Roon accesses local storage during playback all the time.

[Dev]

3 hours ago, austinpop said:

Very interesting. And I think I agree with you - even with both using squeezelite as the player, LMS seems to sound just a tad more open and relaxed than Roon.

 

Yes, exactly Rajiv.

 

3 hours ago, austinpop said:

My issue is with Roon/squeezelite is the occurrence of skipping. Sometimes it happens 5 seconds into a song - other times not at all. Also, I notice sometimes the track transitions are not gapless.

 

Same here.

 

3 hours ago, austinpop said:

Piero's suggested priority tweaks for rtirq and rtapp

 

Cool. I haven't given much thought into it yet but looks like there is still scope of improvements.

 

3 hours ago, austinpop said:

playing with the -a parameter. I use "-a 16:4::" I'll see if I can change the behavior by varying these.

• I did try "-a 499:4::" as to mimic the settings @bibo01 posted. It changed the sound. 16 sounded better. 499 seemed to smooth out the transients. I want to see if a small change can help with the skipping, without affecting the SQ.

 

 

Ok, I haven't played with -a option also. I let it run default. However, I am currently running with 4Gb buffer (w/ 8G ram). 

 

3 hours ago, austinpop said:

Beyond this, it may just be that Roon's LMS emulation is not ready for prime time.

 

Yes, I don't think Roon guys are any serious about enhancing their implementation, removing their limitation on 192k and dsd64, neither keen at fix bugs nor improving their own SQ. I think we are at the crossroads now ?

[Dev]

7 minutes ago, Superdad said:

 

Bump @Dev.

Can you compare HQ Player Desktop>AL/NAA versus LMS>AL/Squeezelite?

 

Ops, I didn't realize your post earlier. This thread spins so fast its sometimes difficult to keep track. Yeah, I plan to compare them very soon as its also of interest to me but I will be using the embedded version for the time being ?

[Dev]

10 minutes ago, bibo01 said:

Why don't you do a simple test with enabling LMS and use that in place of Roon, so you can verify if the problem is due to Roon?!

 

From the limited experience I have so far with Roon/Squeezelite integration, the skip issue happens more often with Roon. Occasionally I have also seen the LMS will skip track from the playlist and when there is a sample rate change. 

[Dev]

22 hours ago, ronfint said:

@Dev,

 

Sorry to drop back 6 pages, but did you happen to try the NUC8i7BEH with AudioLinux as a Roon server only (i.e without HQPlayer)? I'm interested in whether it is powerful enough to upsample in roon to DSD512. — Thanks!

 

 

 

DSD256 is fine with Roon. I don't have DSD512 capable DAC, so can't say for sure but unlike HQP, Roon is generally light on compute when up-sampling.

[Dev]

21 minutes ago, afrancois said:

Can somebody who has Audiolinux lxqt check if the HQPlayer Embedded fingerprint of the NUC changes with every reboot?

 

I just paid for an HQPe license but it is now useless because the fingerprint of the NUC changes during reboot.

 

It won't change as long as you are not adding/removing h/w components as below:

CPU, motherboard and network interfaces are part of the fingerprint. USB
interfaces as separate cards, GPUs, HDD/SSD and RAM are not included.

 

Also, with AL lxqt, you should be able to use HQP desktop, unless you are running the lxqt in command line.

[Dev]

1 minute ago, afrancois said:

This doesn't seem to be the case.

As I said it changes at every boot. Of course, I don't change any hardware in the meantime.

 

The fingerprinting is similar to AO on Windows. I am on a trial, so didn't pay much attention to it but I did ask Jussi about this before. 

[Dev]

1 minute ago, afrancois said:

I did also ask how it worked before buying the license. I now see that the fingerprint changes while no hardware changes have been made.

 

If its possible, revert back the AL image to the trial version and see if it changes on every reboot.

[Dev]

@austinpop, by any change did you try with LMS/Squeezelite combo ?

 

Perhaps if Roon implemented caching to ram before playback, these effects could have been minimized.

 

[Dev]

2 minutes ago, austinpop said:

I was chatting via email with an old classmate of mine from engineering college. He shares an interest in audio, but is just getting into computer audio.

 

I told him that with all those programming and systems courses we took, did we even think the day would come when we would ask the question: which of these sound better?

• squeezelite -n zenith-se-player -o hw:CARD=Interface -a 16:4:: -D

or:

• squeezelite -n zenith-se-player -o hw:CARD=Interface -b 1572864:1572864 -a 16:4:: -D

 

 

I graduated to -b 2097152:2097152 ?

 

Between, a CL16 sounds different than CL17, not by a big margin but they do and please don't ask me why ?

[Dev]

5 minutes ago, austinpop said:

Sure - CL17 is one whole cycle slower. By itself, no big deal, but a cycle here, a cycle there, and pretty soon you have a Tour de France!

 

The higher the running frequencies, the more effect it can have on generated noise pattern (or SI stuff). I hardly doubt it will have anything to do with latency but who knows ?

[Dev]

2 minutes ago, rickca said:

I'm looking at the squeezelite command flags, and I really don't get what this one is saying.  Also note that the last optional specification is whether or not to use mmap (memory map) ... I wonder what that might do for sound quality.

 

By default it will try to mmap if the device used in "hw" parameter is memory mapped I/O, as in my case

 

  82200000-8220ffff : 0000:00:15.0
    82200000-8220ffff : xhci-hcd

 

card0    USB Audio card        usb1  --> 2616:0110    PS Audio PS
card0    STATUS  --> access: MMAP_INTERLEAVED format: S32_LE subformat: STD channels: 2 rate: 352800 (352800/1) period_size: 1411 buffer_size: 5644

 

[Dev]

25 minutes ago, jean-michel6 said:

My understanding is that you need to have Daphile lms installed on the server side. 

Then on the lms screen ,once you have selected the songs you want to play there is a small icon on the bottom right side of the screen to load your playlist to ram. 

 

My my friend has 8gb and he can load full albums .

I will try to get screenshot from him. 

 

I had used Daphile many moons ago since it was the only free s/w to integrate DSD-toDSD upsampling using sox (from mansr). But when Roon introduced their upsampling and later with 1.5, I dropped Daphile in favor of Roon. At the time, I knew it used LMS/squeezlite in the backend but I don't remember seeing the setting you are talking about. Its possible that this feature was later introduced. When conversing with the developer, Kimmo, I recollect he said that there are number of changes made to LMS for Daphile. Not sure if those changes made to the mainline code or not. I believe loading playlist to RAM would be an LMS feature and it would be excellent if we can get more some info on this and how can this feature be available into LMS mainline.

 

@hifi25nl can you take a look at the specifics please ?

 

[Dev]

5 hours ago, austinpop said:

 

huh...looks like you have a very busy weekend ahead ?