Jump to content
IGNORED

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


Message added by The Computer Audiophile

Important and useful information about this thread

Posting guidelines

History and index of useful posts

Most important: please realize this thread is about bleeding edge experimentation and discovery. No one has The Answer™. If you are not into tweaking, just know that you can have a musically satisfying system without doing any of the nutty things we do here.

Recommended Posts

 Joining u all here. I am not even sure what topics are supposed to be discussed here. I read from months ago a discussion about sCLK vs OXCO. Does anyone know any difference in SQ or has compared betweem the pinkfaun and SOtM USB bridge? Pinkfaun uses TOCO but then can be upgraded to OXCO 10MHZ, where as SOtm can be upgrad3 with sCLK. SOtM with clock is about $300 cheaper than Pinkfaun with OXCO 

Link to comment
On 11/3/2018 at 2:43 AM, One and a half said:

Isn't the concept of the lower power the better concept already existing with the renderers? I mean, the smaller power footprint equates to less noise. If you need more cores, then a regular mobo with all its noise is back again.

If the object is to have a lowest possible overhead on the OS, wouldn't Fidelizer or Audio Optimizer do that job already? If a modular Intel based single board computer like an Odroid H2 could be interfaced to a Jplay USB or Ethernet Card, there could be some interest.

 

 

Interesting point. I am thinking of using i5 7th gen intel shi h perhaps is overkill but I would be running M-ch and likelu M-ch DSD in future, and perhaps upsampling/down sampling. There is thought that higher CPU makes it more efficient to run and actually sounds better. I will use euphony audio which will be s very quiet OS anyway. 

Link to comment
10 hours ago, flkin said:

The main difference is one uses an external reference clock and the other has the clock on board. It's from the idea of the advantage of having the clock at the needed frequency at the point it's needed rather than a reference signal far away that requires a regeneration to the required frequency from the reference one. On the other hand a reference clock may be able to reclock many points and is thus more flexible in use.

 

I have the tX-USBultra that contains the sCLK-EX card which by itself sounds good but turns great when adding an external 10Mhz reference clock. Reference clocks can be pretty expensive depending on which one you choose and the sound reflects the quality of the clock added. In addition you'll be needing a few power supplies and cables to connect the lot, the cost of which may even exceed the hardware for top grade ones.

 

I also have the Pink Faun USB card with OCXO on board in my Pink Faun 2.16x streamer. I'm not using an external power supply. It sounds best by itself and doesn't benefit further from passing through a reclocked tX-USBultra.

 

As for comparing the two directly it's difficult to do for me but prior to using the PF streamer, the tX-USBultra w/clock had always improved the sound of whatever I attached. But not anymore.

 

I havent tried the TCXO version but I know that Pink Faun will be offerng a higher grade OCXO module for their usb card in a month or so. Probably as an option and not a replacement.

 

 

Thanks for this info. Very helpful. I guess the  OCXO from pink faun acts like reference clock? I think it makes sense to get pinkfaun bridge first and then upgrade with the OcXO clock later. TheSOtM seems a bit less cost effective, though cheaper initially. Pinkfaun also seems to have great service and more upgrades available lae on. I presume their newer OCXO woudl be evem more costly ?

Link to comment
4 hours ago, misterspense said:

What are the experiences with the original Intel NUCs with respect to the noise of the fan? I'm following this thread and contemplating buying a NUC, but am not keen on a fan equiped computer in my livingroom. I know there are fanless cases available aftermarket, like Akasa, or fanless mini pc's from other brands.

Do these Celeron NUCs produce some noise? Or are they virtually inaudible?

There are some fanless NUC and industrial fanless PCs. I am not sure how quiet they are,if good enough for PCs. There is also HdPlex fanless PC which I am planning to use. They are sold as DIY only.  If you plan to do large files like DSD256, DXD, M-ch and/or upsampling/resampling from PCM--> DSD, then you need a more powerful PC, probably like at least 7th gen i5. 

If you play only low res or only up to 24/192hz something like Raspberry/small low power PC probably is adequate & much quieter than regular NUC. 

Link to comment
On 9/18/2018 at 4:21 AM, seeteeyou said:

 

This site is acting strange, I get these quotes stuck on and cannot delete. Above is an error.  Ok, maybe it would be a good idea to have people express their grievances and arguments by PM (via this site) to the other person directly instead ?  

 

2 hours ago, austinpop said:

 

I think @ElviaCaprice is wise to reflect before jumping off the fence! The benefits of running in memory with NUC hardware seems to be most striking with an endpoint, which has no state to preserve or persist. Things do get more complex for a server, though. I see 2 areas that need careful navigation.

  1. Memory footprint: I will guess that to run AudioLinux AND JRiver in memory will need more than 4GB of RAM? I also guess that can be fairly easily remedied with the purchase of more RAM.
  2. Persistence: This one is trickier. Any write I/O's the server does will not be preserved when running in ramroot mode. I wouldn't count on ramsave as a safe persistence mechanism. There may be some workarounds, analogous to what we've done on Roon Server, which involve replacing the location of the database and executables with symbolic links to persistent storage. Has someone found a way to do this with JRiver?

For the memory issue 4 RAM should be adequate for most playback on a system like linux. I have no experience of J River on Linux. J River Mac is fine with 4 RAM but I did not use up/down sampling or conversion from PCM  __> DSD.  Usually more than 4 RAM is needed mainly for upsampling/resamping/conversion from PCM---> DSD, esp if the track is all loaded into RAM.  For low res to 24/96, 4RAM should be fine. DSD may cause issues with Mac, not sure about Windows, but   DSD runs on my linux/euphony OS, even at DSD256 (native, not upsampled), and with 4 RAM only.   

Link to comment
On 9/18/2018 at 4:21 AM, seeteeyou said:

 Ignore the top, for some reason, it keeps pasting this automatically.

 

Has anyone tried HdPlex PC + its LPS? 

Link to comment
1 hour ago, sandyk said:

 A good idea.

 BTW, this happened to a friend of mine.

 However, HDPlex did the right thing by him, also providing a later model  replacement PSU PCB with protection.

ok, I am planning to get the latest Hdplex LPS as sold on web now. Would there be issue with this later model? I am planing to use internal SSD 2.5 Sata as boot drive (euphony OS).

Link to comment
5 minutes ago, sandyk said:

 

 I would expect that current models all have overload protection incorporated.

 Normally, the manufacturer will prevent this happening in the design of the PCB by restricting the amount of adjustment possible, or at least provide warnings in the User Manual.

 

Irrespective, it is never a good idea to adjust the output of a voltage regulator PCB way below the maximum output voltage.

Strange the manufacturer would allow such adjustment if it would actually cause overheat issue. I would check with Hdplex later when I get the LPS.

Link to comment
1 minute ago, sandyk said:

 

  Use the 5V output and you will have no problems !

In many cases the adjustable 19V rail was left unused, so some who needed extra outputs tried to use it adjusted down to +5V,

 which is fine if only using a low current device such as a Regen etc.

Ok, noted, I will keep that in mind. BTW, I don't think the HdPlex 200W LPS can drive 2 devices that need 19V at the same time? 

Link to comment
1 minute ago, octaviars said:

 

I use all four outputs on my 200W HD Plex. 

 

5V ISP fiberconverter. 

19V router (5-19V)

12V switch 

19V NUC with Roon ROCK 

 

But it gets hot and I mean really hot so I have a speedcontroled 120mm fan cooling it (all my equipment is in a cabinet so that will of course affect the heat)

Cool, thanks for the info

Link to comment
1 hour ago, Johnseye said:

Audiolinux NUC Update

 

After a lengthy search I thought I found a NUC7CJYH, placed an order, only to learn weeks later it was on back order for many months.  I suspect it was out of stock, CDW didn't show that, and not likely to be produced by Intel due to next gen releases.  This was probably a good thing because I just ended up buying a BLKNUC7i7DNK1E which contains the NUC7i7DNBE board with 8M i7 processor cache.

 

Background

 

My current system is an i7 mini ITX motherboard with system and ethernet clocks modified by the SoTM sCLK-EX.  This server also has a tX-USBexp with sCLK mod providing USB out to a tX-USBultra which feeds a Holo Spring L3 DAC.  Power is fed by a two rail SR7 with DR, LPS-1 and LPS-1.2.  I've been using this system since about May of this year.  Prior to that, something similar with a Celeron proc that couldn't upsample.  This i7 mini iTX board has been able to upsample DSD512 via Roon and HQPlayer.  I have been very pleased with the sound quality to date and complacent in making any adjustments, thinking it was about the best I could get.

 

Current and Evolving State

 

Based on Larry and Roy's recent discoveries, which are always valuable ideas worth chasing, I implemented Audiolinux with a NUC as endpoint.  Like many here I've owned an mR and I still have an sMS-200.  My single server build sounded better than what I heard with these endpoints.  I did install Audiolinux on my server before getting the NUC and ran it in RAM complete with Roon database.  I upsampled to DSD512 successfully and was pleased with what I was hearing.  It was a minor improvement.  I wasn't blown away, but it was better and absolutely worth the $30.  Since I power my SSD separately I was also able to free up an LPS-1.2.  I've been enjoying Audiolinux in that system for about a month or so.

 

Then I got the NUC.  After a slight road bump where I used the same USB stick to load the NUC as I did the server, Larry straightened me out pointing out that it creates a duplicate host name.  I had installed the headless version on the NUC in the mean time which also resolved the issue, but it was very helpful information to know I need two different USB sticks.  Running Audiolinux as a headless NAA for HQPlayer was simple and running in minutes.

 

I first set things up where I went from the NUC to the tX-USBultra to the DAC.  The Linear Solution switch connects my mini ITX server with NUC and NAS.  It's currently powered by the TLS LPS provided with the switch.  This was a stunning ear opener.  BAM!  It was like turning on a switch.  A big improvement.  The soundstage was not only wider  but it was deeper.  This allows for instruments to have more separation which allows for better distinction.  There is better resolution and vibration of the music.  Then I removed the tX-USBultra.  It's not going back.  I'm still going to have a couple more back and forth sessions with and without it, but based on what I've heard so far it only collapses the soundstage and resolution.

 

Because my SR7 was powering my ITX server I need to figure out how I'm going to power it, or if it even matters.  The SR7 is now powering the NUC.  I'm going to see if the sPS-500 can power the server.  If not I'll probably use the HDPlex supply.  The other SR7 rail is going to start powering the TLS switch and the LPS-1.2 will power the sCLK-EX.  I'm going to pull out the tX-USBexp and shelve it with the tX-USBultra.  The Akasa fanless case is on order.  I'll be pulling the board out of the NUC and putting it in that case.  I expect another improvement then.

 

Folks, this is a big deal.  I'm using the Spring L3 in NOS mode and upsampling with HQPlayer because the ability to define my sound through those filters.  I find this granularity incredibly powerful and like not having the DAC putting its signature filters on, then upsampling with another filter.  Personal preference, we all have them.  So where I would hear subtle but audible differences between filters it's now become significantly audible differences.  I love this.

 

When I removed the tX-USBultra I felt surrounded by the music.  It wasn't just in front of me in different dimensions, it was all around me, blooming.  My skin tingled with goosebumps.  I've got a lot of good music listening to come and will update as I make additional changes.

 

Sorry I can’t quite follow clearly. U mean when u are using SOtm USB-exp + sClk-ex alone, it is better than together with USBultra? Is the sCLK a 10Mhz reference? 

Link to comment
1 hour ago, Johnseye said:

 

I have been using the tX-USBexp out to the tX-USBultra, both clocked by the sCLK-EX.  I am removing the tX-USBexp and tX-USBultra because the NUC alone with Audiolinux loaded into RAM is better sounding.  My sCLK-EX is using the Mutec REF10 as 10MHz master clock.

 

I should also point out that I have 8GB of RAM in both the server and NUC.

Interesting that external ref clock is not always better. It is perhaps introducing additional noise? 

Link to comment
On 11/23/2018 at 12:19 PM, austinpop said:

 

Hey B&M,

 

The point about booting from USB is that yes, while it is slow, you can remove the stick (and associated noise) once booted. While an SSD is faster, you can't (easily) disconnect it after boot. 

 

However, you seem to keep missing the point, no matter how many people try to explain it to you. The combination of NUC endpoint running Audiolinux in RAM was arrived at by experimentation, and to some extent, luck. No one here can explain exactly why it's such a great sounding combo. Read this paragraph again a few times until it sinks in, please.

 

There are plenty of theories, but this is not the thread to debate them. May I suggest this thread? Although I see you seem to have worn out your welcome there too. Or start another thread.

 

Ok, sorry bur I am still very confused about this AL running RAM. U mean the entire OS is somehow loaded into RAM, or the software player is loaded into RAM and then also playing the entire music file loaded in RAM? Or just the music file in RAM? How much RAM would one need for that? Sorry if this has been discussed but it is really hard to trace back with 1000s of responses on multiple topics.

Link to comment
1 hour ago, greenleo said:

If the AL is not using as the server but an NAA, I doubt a music file is totally loaded into the RAM.  On the other hand, if AL is using as a server and local access of the files is utilized, then by making the partition in the USB disk bigger, the RAM in the computer bigger that may hold the complete partition, probably the music files may be loaded into the RAM as well.

 

I've never tried this before.  You may ask Piero directly as he participates in this thread as well.

Oh yah, I meant mainly if audiolinux is running as OS in a harddrive where the music is also stored and so the music files can be loaded into RAM but it likely means the music player also needs to be inside the same linux OS? Not sure, since from the audiolinux web and here it talks mainly about Roon and NAA and NAS which means the music files are being processed off site somewhere else so the music files won’t be loaded into the same RAM where the Audiolinux is also loaded? The website stated HQplayer can be embedded in the AL, I presume meaning it is also in the same linux OS, as opposed to  off site via NAA, am I correct? 

Link to comment
  • 2 weeks later...
1 hour ago, afrancois said:

I've copied some FLAC's to a directory in AL ramroot and couldn't hear a difference between "in memory" play and play from the NAS using RoonServer. I did even shutdown the NAS to be sure the NAS wasn't interfering somehow. Perhaps my system is not resolving enough. I haven't tested with a NUC endpoint yet.

I talked to guy at Exasound (canada) and he thinks the fuss with NAS, NAA may or may not work any better. I try to use a single Music server with all music in the SSD or external drive . If audiolinux claims to be a great OS then it shouldn’t matter much. Perhaps a crappy OS may benefit from separate decoding of music using NAS or NAA. But then I could be wrong, as I have never tried those metthods. Isn’t roon mainly invented for convenience so u can play music anywhere u want, even works well on inferior devices?  

Link to comment
4 hours ago, RickyV said:

 

What do you think the streamer makes on total sq?

 

 

Streamer is as important if not more than DAC, and that includes the software player, and the OS as well as the hardware itself. For eg Mac OS sucks compared with Linux. Audirvana beats J river on Mac OS from my experience 

Link to comment
22 hours ago, rickca said:

Any idea why AMD Ryzen is more suitable for Linux than Intel?

Not sure, AMD Ryzen has more cores like 8 cores in general. Pinkfaun stated the Intel won't work well with I2S PCle cards. It may be more task focused as it has no graphics at all so perfect for headless audio

Link to comment
  • 3 weeks later...
On 1/25/2019 at 2:47 PM, Metnoc said:

Hi my first post on this forum with my own impressions rather than a question.

So obviously I wanted to try Audiolinux as well since its held in such a high regard by many.

I have a Zotac 620 computer running Roon core, it has passive cooling, external power supply and two ethernetports, I use one of them for direct connection to my sms200ultra.
Usually I run roon core on win10 with fidelizer in purist/streaming mode, so I compared this to Audiolinux in ramboot mode.
Well, I must say that I agree with the guy behind Fidelizer with what he as written about the sound difference between the two OS.

https://www.fidelizer-audio.com/extreme-dual-pc-computer-audio-setup-shootout-between-audiolinux-and-windows-server-2019/
Audiolinux might have better transparency and sound a little snappier but it also sounds thin with quite poor harmonics. As a whole, AL doesn't sound very natural at all.
So I prefer win10/fidelizer over AL. 

Has anyone tried running without root or NAS/NAA but just using software for playback in the same computer running AL our windowns+fidelizer. I don't do Roon etc. This I presume would better test the OS to compare there 2. Also with hi-res such as DXD/DSD either native and/or upsampled. I find it is harder to play Hi_Rers in general, and the better OS tends to do much better  

Link to comment
  • 4 weeks later...
2 hours ago, EnjoyTheMusicNow said:

The x10 is my router/internet and is in a closet 100 feet away from my listening room. It is connected to the internet cable coming into the house. Previously I had an RJ45 cable running from the router to the listening room, connecting to the SOtM switch. This was replaced by fiber. I gained galvanic isolation from my sonictransporter (also 100 feet away connected to the X10). I also think I gained a cleaner (less loss) signal but I have no way to confirm that - just the sound. The X10 is just a beast. It is a 1.7GHz Quad Core Processor. My only guess is that similar to the NUC findings, the extra processing power ensures that it can handle network traffic very well - i.e. it is not struggling to perform tasks.

So this optical fiber is a type of ethernet fiber cable  or some other cable  with RJ45 connector?

Link to comment

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



×
×
  • Create New...