HQPlayer's Network Audio Adapter

[Superdad]

Google Translate

 

Win NAA sounds better, perhaps due to better Win ASIO drivers ?

[/url]

 

Hi Edward:

I get a kick from Google translate of that page. To wit: "Well the sauce as NAA does his hair and beard to CUBOX…" Awesome!

 

But seriously, I think we need to be careful about drawing conclusions from that report because he was comparing two major variables at the same time. To compare a CuBox-i as Linux NAA versus a full Intel Windows machine as NAA and to attribute the differences to drivers is not prudent. Better would be to run Linux/NAA on that same Intel machine and compare that versus the Windows NAA. Even then, care should be exercised in the test.

 

Besides, it is no surprise to me that CuBox-i does not make the best NAA. You know I made my own fair hardware comparison between CuBox-i4Pro and i7 Mac mini--both as text-only Linux/NAA using Miska's images, and both run from my good JS-2 LPS. Mac was far better.

 

I know you have already made your own tests, preferring Windows greatly on the MacPro versus OS X--when running HQP direct to DAC.

But to get closer to answering the question of better OS/drivers for NAA, I have been thinking about installing Windows on a drive partition in my Mac mini (mainly because I also want to finally take a listen to XXHighEnd, as well as to HQP desktop). I could then compare the new Windows NAA versus the same machine booted from the same Linux USB/NAA image (as used in prior testing). To me, that would more directly address the Linux versus Windows issue.

But I won't be able to do this any time soon as I am busy and behind on other work (darn addicting CA forum! )

 

--Alex C.

[Superdad]

We compared HQ Player Desktop running in WS2012 + AO in MSM and the same DH61DL i7 3770 PC running just NAA program, with the NAA being fed by HQ Player on a Mac Pro Win 8.1, Mac Pro Yoshemite, Mac Book Air, and Surface Pro 3

 

Hi Edward:

 

Forgive me if you have already reported about it, but do you feel that NAA run under WS2012 + AO in MSM gives better sound than the SAME machine running Linux/NAA? That comparison would be important to me and others in deciding on an OS for NAA. And of course WS2012 is more expensive then a free Linux dist.

 

If you have made the comparison, please report details of OS, set up, and SQ.

 

Many thanks!

--Alex C.

[Superdad]

I think a Fitlet-B will be a good platform to make a decision between Win NAA vs Linux NAA. I will test slimmed Win8.1, WS and Linux.

Furthermore, IMHO it's important to have a LPSU for NAA. I think EuroDriver was not running his i7 based NAA on a LPSU.

Another test that we are running on our Italian forum, which I suggest others to try, is to have a Win8 NAA system slimmed with CAD script running in SAFE mode with no GUI.

 

Those are all GREAT ideas Gianluca. Any idea when Fitlets are actually going to ship? Have bee a Compulab fan for years, and the Fitlets look terrific for general audio purposes.

 

Funny that you mentioned safe mode for Windows and NAA. I was recently asking Miska about looking into the commands and framework requirements to launch the OS X NAA from the Mac's text-only single user mode (cmd-S at startup and you are at at Apple's UNIX command line). I know he is not an OS X fan, so I was not surprised when he demurred.

 

And indeed, to get the most benefit from using an NAA, it, as the device connected to the DAC, needs to be as "quiet" as possible, well isolated, and feed really good power. (BTW, Eurodriver is also running his DAC on the stock 12V/1.6A SMPS brick it came with. Ted_B and others have already reported solid improvements with a linear on exaSound.)

[Superdad]

The Fitlet is only 0.5A @ 12V.

 

Maybe at 10-20% it will be, but I think at 90% usage it will slot in at around 12W, so at 12V that would be 1A. So I would want to have a bit more muscle in reserve and would plan for at least a 3A supply.

[Superdad]

The possible NAA platform Fitlet-B, for either Windows or Linux, is ready for order:

Order fitlet

 

Oooo, soooo tempting. I just can't decide which to get!

[Superdad]

It's mainly about noise and keeping noise sources under control

- PS supplying NAA

- SSD inside NAA

- RF quiet CPU

- Noise comming from LAN connection

- choice of OS / optimizations of OS

- quality of USB driver implementation

- match with DAC's USB receiver set up

 

SOTM SMS-100 ticks a lot of boxes, you can feed it with a LPS, there is no SSD, Linux is a small OS, with a RF foot print to match. I don't know how well your UD501 works with Linux

 

Fitlet - b looks pretty attractive on paper. Will probably be able to run WS2012+AO which despite it's massive size and complexity is one of the best sounding OS out there. No SSD and Compulab has a track record of building low RF PC's

 

Cubox - I looks good on paper but has repeatedly lost out in SQ comparison tests against

- Linux NAA on Mac mini

- Win NAA

- Mac OS NAA on Mac mini

the achille's heal of the Cubox might be its sensitivity to noise coming over the LAN cable. Mac mini's have a filter on their LAN input, and perhaps the Cubox does not have a similar filter. An optical LAN isolator would definitely improve the performance of the Cubox but up to what level ?

 

low powered CAPS / Win PC

one question yet to be answered is how much SQ advantage does WS 2012 + AO have over a slim Win 7 or Win 8 when running NAA. Running HQ Desktop the SQ difference is unbelievably huge, but NAA is quite different in it's resource requirement and calls to the OS for services. I plan to do a SQ comparison between NAA on WS2012+ AO and Win 8 in 2 weeks time. Also what is not known is the SQ difference between SSD with independent power supply compared to system with no SSD.

On the Nexthardware forum, they report consistent SQ improvements with RAM disk with NAA

 

Mac mini

OS is turning out to be a lot more customize able and it's possible to boot off SD card and dismount SSD. Negative is that it's not easy to run Mac mini Fanless. A DC fan with no PWM control still has a RF signature that can be audible

 

Laptops

Stay away, screen and video are sources of noise you can not get rid of easily

 

which ever road you choose, order a pair of fibre media converters, optical patch cord and a LPS for the FMC connected to the NAA ;-)

 

 

PS

I have a Anker 2 and it was very noisy as a power source for my E20 DAC

 

Great condensed summary of your past few months research Edward!

 

Only thing you left out is something that is hard to compare/quantify between different host devices, and that is the quality/signal integrity of the USB audio output from the computer:

Virtually no DAC's USB input PHY and processor is immune to ground-plane and packet-data noise and will be effected by the incoming signal--which will make the PHY's decoding work more of less hard thus causing some unique "fingerprint" of activity inside the DAC which can and will get past even the best digital ("galvanic") isolators after the USB processor--jittering the master clock somewhat.

 

Thus products such as the SOtM or Paul Pang PCI USB boards, or the forthcoming UpTone/Swenson USB REGEN device (single port hub with ultra low-noise regulators and good USB clock--plugs in right at the DAC) will also have an impact on SQ with NAA or standard set ups. I am sorry I fried my CuBox-i4Pro just as I was setting up to test my REGEN on it with it running Miska's Linux/NAA image.

 

Please also remind me if you, Gianluca (bibi01), or anyone over on the NextHardware forum has yet directly compared text-only Linux NAA to the variations of Windows NAA with the exact same hardware chain. I recall you saying that Linux driver for your exaSound DAC is not available. Is that correct or am I misremembering things? (so much to lee track of these days )

 

Thanks,

ALEX

[Superdad]

another interesting way to go is Corning's optical USB cable which converts the USB signal to optical and then converts it back to electrical. The + 5 volt and ground are also in the cable so that could still be a significant conduit for RF noise. However a string ( series ) of appropriately selected ferrite beads might be able to effectively take care of any RF over a chosen range of frequencies.

 

Thanks for your thoughts and additions Edward. Just remember though, the Corning USB cable has a PHY and a clock at BOTH ends, and the power it uses to run them is still the computer's 5VBUS. So more stuff and mostly done to extend USB distance.

[Superdad]

If we were to do a NAA OSX test comparison, we would need your OSX slimming script :-)

 

I'd rather you compare the text-only Linux NAA image for Mac that Miska gave me about a year ago (but does not give out much anymore). Not sure if he ever updated it for the Network Audio Daemon 3.0.0. He refers to mine as "naa-205-lite" (AMD64). Ask Jussi if it would be okay for me to send it to you--or if he has a newer version (which I would love to get). I think that since he released NAA for OS X and Windows he may have stopped trying to support the Intel/AMD versions for Linux. Was too much of a headache. But with the Fitlet coming, there would seem to be a more compelling case.

 

I must correct myself here: I know that Miska still offers Linux NAA for AMD/Intel, etc.--but has people install their own Debian Jessi or Ubuntu Studio, etc. to run it. I was above referring to the heavily stripped-down, text-only disk images that he offered--and still does for CuBox-i--where it just boots and loads the NAA with nothing extraneous and the user does nothing. Such a plug-and-play, optimized appliance that is so easy--when it works. And it sounded much better to me than the Linux/NAA install that I did on the same mini.

 

Ciao,

--Alex C.

[Superdad]

In recent discussions with Jussi (Miska) about setting up a fitlet-b with his NAA 3.0 Ubuntu (Trusty) version, Jussi recommended:

 

"using Ubuntu Server (Trusty) for NAA use precisely because it doesn't default to GUI. The custom kernels I have built work also on the server version. Or if you don't need the custom kernels I've built, you can install "linux-image-lowlatency" and "linux-headers-lowlatency" instead of the server kernel to have a kernel better suited for this purpose.

 

When the server installer asks what additional software to install, select either nothing or at most just SSH server (for headless remote maintenance).

 

This keeps the installation slim."

 

Thanks for that info!

Can you tell me more about what the "custom kernels" Jussi built are for? Under what circumstance might I need or not need them?

 

I am really a novice when it comes to installing Linux components via command line and really just wish that Miska was able to provide clean and simple bootable Linux/NAA images for each of the different platforms. But I understand that there are both Linux distro licensing legal issues as well as somewhat wide variation in both host hardware and connected DAC support requirements.

 

Still, it is pretty clear--after NAA being available for several years--that it is never really going to see wide adoption without installation becoming simpler for novices. And while the recent OS X and Windows NAA releases are an attempt at easing use and broadening the range of compatible platforms, I fear that the sonic advantages of an NAA booted from a very trim Linux are not equaled by NAA under Core Audio (OS X) or under the range of Windows installs.

 

None of the above is meant to take anything away from the terrific research and experimentation that the rest of you are doing with HQP and NAA, nor from the great support that Miska provides for his wonderful performing software. My comments are meant only as encouragement towards greater success with those of us who don't have the time or knowledge to dink about too much with operating systems.

 

--Alex C.

[Superdad]

Shadowlight and Eric (ericuco): Many thanks to you both. I'll stick my toes back in the Ubuntu waters as soon as I have a chance.

[Superdad]

Out, out, ye vile running daemons!

 

Wow, you are on a roll these past couple of days Jud. (All that sparing with Priaptor in that nasty other thread much have sharpened your wit a bit. )

[Superdad]

The simplest way for me to use an NAA would be to run an Ethernet cable from my MacBook Pro to the NAA and connect the NAA to the DAC. Both the BeagleBoard-xM I have now and the ODroid-C1+ I'm thinking of would be running Linux, and I assume the Linux version of networkaudiod. Any problem at all for HQPlayer on the MacBookPro to "see" the daemon running on the NAA in that case?

 

Hey buddy:

It's not the OS difference that will give you any problem, rather it's that direct Ethernet cable from your MBP to the NAA that will give you fits. NAA needs some sort of DNS happening and HQP desktop may not "see" the NAA. When you boot the NAA, watch the screen, and just before it all finishes loading (guess it depends if you are booting a text-only Linux), look t see if the machine got an IP address.

I tried a bunch of stuff (including bootp) to have my desktop Mac act as the DHCP server for a directly connected NAA, but even when I managed to get the NAA device an IP address, HQP desktop still would not see it. It has been a while since I did this, and it was with earlier versions of both HQP and the NAA (2.0.5).

 

In the end I gave up and went back to just remote controlling my music Mac mini running HQP on it. I did get that machine to boot a stripped-down Linux/NAA image that Miska gave me, but the only way I could get it and my desktop machine to hook-up and play was to have the NAA's Ethernet cable go into my DSL modem (it hands out IP addresses for my network, but is slow and the Linux/NAA boot was too impatient to wait for an IP to come through to my Cisco switch). So for me, my direct Ethernet (Blue Jeans Cat6a) cable from desk mini to SD-card-booting tweaked mini sounds as good or better than the NAA through the router.

 

So if you do figure out how to get HQP/NAA working via direct Ethernet cable (with a Mac on one end and Linux NAA on the other) please LET ME KNOW HOW!

 

Thanks,

 

Alex C.

[Superdad]

So a $100 Cubox (never heard of this...it's So cute! lol) acts as a simple 'buffer' b/w my M.Mini and my DAC? And you said it wasn't hard to setup? Seems like an affordable way of keeping the SI as clean as possible.

 

Though something like the forthcoming microRendu from Sonore is going to sound a whole lot better as an NAA than a Cubox.

[Superdad]

I want to try this out on a mac mini, what exact image am I supposed to use? And do I have this right: load to SD card and boot off this?

 

Index of /bins/naa/v3/images/

 

../ 04-Jan-2016 23:10 -

foss-src/ 15-Mar-2016 00:01 -

naa-311-beaglebone.zip 04-Jan-2016 23:46 10964994

naa-311-cubox-i.zip 04-Jan-2016 23:46 13001710

naa-311-raspberrypi.zip 04-Jan-2016 23:46 34120640

naa-311-raspberrypi2.zip 04-Jan-2016 23:46 34510873

naa-321-beaglebone.zip 15-Mar-2016 00:02 22323996

naa-321-cubox-i.zip 15-Mar-2016 00:02 27157870

 

None of those images will work for an Intel Mac mini.

 

And if someone ever figures our how to get a Linux install to boot a Mac mini from an SD card, please PM me instructions. USB sticks, no problem. SD card, no way that I have ever found--and I tried a bunch.

[Superdad]

NAA On OSX – just run the NAA app:

Index of /bins/naa/v3/osx/

 

 

I found that OS X NAA does not sound as good as Linux NAA--probably because CoreAudio is still involved on the NAA.

I was able to make a fair comparison on my 2012 Mac mini because over a year ago Miska gave me an image of a his stripped-down, text-only Linux/NAA that would boot my mini (don't ask him for it as I don't think he will give it out and mine is now very out-of-date as it uses networkaudiod 2.0.5).

[Superdad]

Hi Miska,

 

Any plans to get NAA on MSB's renderer module?

 

Hi Tranz:

Given the chipset used on the DLNA modules that MSB licensed, there is likely no way the NAA code could ever be ported to it. The module doesn't even have an OS per se.

[Superdad]

...or through any DAC by Merging Technologies (Horus, Hapi, NADAC).

 

But that's not using Miska's NAA architecture, whose purpose is to separate the heavy processing computer from the DAC-connected device. And with Merging's Ravenna implementation you have to use their "virtual sound card" software and the compromises of its communication with the OS and the player app are not known.

[Superdad]

I was really hoping to compare a lightweight NAA against a regular computer as I was considering a microU or something equivalent.

 

Owning both (Cubox-i4Pro and Sonore MicroRendu) I can assure you that comparing you Cubox to you computer will tell you NOTHING about how a MicroRendu will sound. Would be about as valid as driving a Yugo up a hill and concluding that all compact cars are poor performers uphill.

[Superdad]

3 hours ago, jabbr said:

Is networkaudiod expected to pick up a new USB device when it’s plugged in ie hotplug? 

 

Seems like it does sometimes but other times times needs to be restarted

 

Linux does not generally support hotplugging of USB hubs, so if you are referring to something like the ISO REGEN you are testing then you should expect to have to restart. (Not the case for Windows or macOS.)

[Superdad]

1 hour ago, Miska said:

Oh yes it does, I can hotplug as much as I want. Sometimes for some hardware, either hardware is buggy or Linux kernel driver support for the hardware is buggy and then something may not work. But on x86 xHCI has been working fine for me for a long time...

 

Okay, that's great for your set ups!  ? But apparently a lot of Linux-based streamers/renderers do not have a properly working Udev and with those people have not been able to hot-plug our ISO REGEN.

Recognition of the hub (the ISO REGEN) only occurs if it is attached and powered on before the Linux device boots up.  Once the hub is seen people can plug/unplug the DAC on the downstream side all they want and the DAC will be seen.

I'm just reporting based on the support we provide to hundreds of users...  And I am the first to admit that I don't understand Linux much at all! ?

 

[Superdad]

On 9/5/2018 at 2:37 PM, jabbr said:

I reconfigured my wireless mesh network (ASUS) and am able to get DSD512 via Raspbery Pi 3 B+ as NAA without dropouts. raspbian 4.14.62. Its on the 5Ghz band.

 

Color me impressed.  

[Superdad]

Yup, DSD256 on my Spring L3 is mighty sweet.

I also greatly enjoy PCM 352.8 with poly-sinc-short an NS5.  Amazingly dynamic and lifelike. The trick is getting DAC-bits setting dialed in just right (hint, it should be much lower than you would expect).

[Superdad]

6 hours ago, Bob Stern said:

 I'd guess 19-bit would be ideal for most state-of-the-art DACs.

 

Well I have not had all that many DACs through there—and certainly no parade of expensive “state of the art” units —but your 19-bit guess is spot on for the Spring and for the custom PCM1704 DAC I own.

 

And I am always curious about what setting others have found best.  Getting DAC-bits setting right in HQP really allows the benefits and differences of the various dither settings to be heard.  I think this is a pretty important part of dialing in HQP for best SQ.

 

[Superdad]

1 hour ago, firedog said:

How do you have an idea what this is? Listening tests?

 

Yep, just listening and bracketing (starting by swinging wide to get a sense of what changes, then back and forth, closer and closer until it becomes apparent).  That's also how I used to tune Advanced iZotope filter settings in A+ before I switched to HQP a few years ago.  That's also how I do speaker positioning down to about 1/8-inch. 

 

P.S. I just looked at my HQP prefs and see that with the Spring I actually have been listening at the same 20-bits Miska recommends, not 19 (that must have been with my other DAC).  Sure sounding great these days, especially with my new 47--position Slageformer for level control (replacing my $12K Hovland preamp); Bass to the basement and nuance galore!

[Superdad]

5 hours ago, Miska said:

Etherregen is the bottleneck there, since it limits the maximum speed to 100 Mbps.

 

The EtherREGEN is fine up to DSD512.

The extreme of DSD1024 is too much for the 100Mbps ‘B’ port of our switch. 
 

I can not speak to the high-rate limitations of the [iMX6-based] Sonore Rendu series.