Jump to content
IGNORED

Audiolinux Server configurations, Software, Hardware, and Listening Impressions


lmitche

Recommended Posts

What a long strange trip this will likely be!

I’m about to join you all down this rabbit hole and have a few questions as I get started.

 

First of all, thanks to all of you who have done the hard work getting this all initiated, evolved and documented.

I’ve been following at a distance until things looked ripe.

 

Second, my existing setup for streaming:

 

HQPlayer-embedded running on a NUC, together with my SageTV system, under Ubuntu 18 with no specific optimizations. This is one connection into my netgear 24 port router, which also connects to my Sonore opticalRendu operating as an NAA for HQPlayer. I use the HQPDcontrol app on an android tablet to select my music. The NUC is powered by the default brick SMPS while the oRendu is powered by a Paul Hynes SR-4. Between the router and the oRendu is a Sonore opticalModule powered by a uptone audio lps-1.2 power supply. 50Ft of single mode fiber connects the oRendu (in my living room) and the oModule (in my basement at the other end of the house). This NUC will remain my SageTV machine and basement computer.

 

Some system observations that may be of interest:

1. the oRendu was a (unexpectedly) small step up from the ultraRendu (by itself, fed by SFP in netgear router).

2. the oRendu + oModule together was a big step up from the ultraRendu, suggesting that cleaning up the ethernet signal with the oM was critical in my system.

3. I use HQP without any signal processing, upsampling, filtering or dither. Simply separating the renderer (at NUC) and NAA (at oRendu) has made a big difference in my system.

4. In my system, the difference in playing back WAV versus FLAC is significant. As an engineer, this makes me want to pull out my hair and I’ve always wondered what was going on--maybe eliminating processing at the server/HQP or interupts or who knows what is helping. Certainly, moving processing off the Rendu (i.e. NAA mode) made a big difference. If WAV versus FLAC can make a difference, I’m ready to try optimizing the server further. Q: Has anyone else compared WAV and FLAC? Does any of the AudioLinux stuff eliminate/reduce this sensitivity?

 

My choices seem to be either a NUC with AudioLinux running HQPlayer-embedded or a NUC running Jussi’s HQPlayer image directly. I’m planning on trying both and comparing them. Q: has anyone compared the AudioLinux/HQP combo versus running the NUC off Jussi’s image?

 

So here I am with a lot more questions, even after reading through 2 complete AudioLinux threads and large parts of the mother-thread, taking cut/paste notes of anything that caught my eye and ending up with a 29 page document to jog my memory re: AudioLinux lessons learned.

 

I plan to obtain a new NUC (which one depends on the answers here) for this experiment and I have a HDPLEX 200 ordered and scheduled to be delivered next week.

Here is my first round of questions.

 

1. Bridging—I have been wondering, even after reading, exactly what is going on with this, at least partly because my brain has trouble picturing the connections made up of dots and dashes. Then I ran across a post (for beginners) that said that if you have a ‘star’ configuration (i.e. the server and NAA are in separate boxes and are each connected to the router, not to each other) there is no advantage to bridging. I had thought that somehow the bridging was keeping the traffic separate and that this was advantageous, but I guess I was wrong and that it is ok if the traffic coalesces in the router. Or maybe this was only an issue when the server and endpoint were co-habited in the same NUC? Am I understanding this correctly? If so, my job is much simpler as I was looking at easy ways to get a second ethernet port on my NUC. So, is anything to be gained from bridging if I have a star configuration and my NUC will only host the server/HQPlayer?

 

2. Which version of AudioLinux? I had assumed that I wanted to run headless but I see people moving back and forth between lqxt and headless. Is this for convenience (I’m not an expert at Linux but am not a newbie either—both M$ and Apple are banned from the house and I run Ubuntu 18 on my desktop NUC and basement NUC/HQP/SageTV server, Ubuntu 19 on my Dell laptop) or did this change as Piero’s menus got more sophistocated?

 

3. H/W configuration, part 1: I’m planning on 16G of RAM for ramroot (hqplayer’s library.xml file is currently only about 12MB for 6TB of music in WAV format). Any sound quality differences noted or recommendations on RAM specs?

 

4. H/W configuration, part 2: Any pros/cons for USB stick versus m.2 optane to use for loading ramroot? From what I read, I concluded that m.2 optane and no internal HDD/SSD is quieter and sufficient to run HQPlayer-embedded.

 

5. For the music storage, I read that 3.5” drives are better than 2.5” drives (quieter) and there was a recommendation to put a 3.5” drive in an external usb3.0 case powered with a good PSU. I have some of those around but currently have my music on an 8TB Seagate external usb3.0 drive. Would a good PSU (e.g. HDPLEX) on the Seagate be as good/quiet as putting a drive into an external case? Any recommendations on one path versus the other? Any comparisons of external usb3.0 drive versus NAS storage of music?

 

6. Are there any router configurations/flow control settings known to be important?

 

7. Given that I’m only running HQPlayer-embedded headless on AudioLinux or only using Jussi’s HQPlayer-embedded image, how much processing power do I need and which NUC should I get?

If I’m running headless, I don’t need powerful graphics processing, so is there a NUC which has been found to be quietest in headless mode? And if I’m not doing any upsampling/filtering/dithering in HQP, my processing needs are likely quite modest, so what NUC should I get?

 

Thanks in advance for the great work so far and your help going forward.

 

 

 

 

Link to comment
  • 1 month later...
  • 2 weeks later...

time for a report after heading down this rabbit hole with you all.

 

old system/baseline:

hpqplayer embedded running on 5i6 NUC under ubuntu, wall wart switching power supply, shared PC/NUC with sageTV server, music on usb HDD, ethernet to netgear router, ethernet to sonore opticalModule (lps-1.2 powered), fiber to sonore opticalRendu (in NAA mode, SR4 powered), usb to totalDAC.

 

new system:

hqplayer embedded running in ramroot under audiolinux on new 7i7 NUC (4G Apacer ram, 8G m.2 optane) powered by HDPlex200 at 19v with external usb3.0 HDD powered by HDPlex200 at 9v, ethernet to netgear router (and all the rest similar)

 

i had a audio buddy over last week and we started by listening to some of my standard evaluation tracks with the old baseline system:

 

Bonnie Raitt-Angel from Montgomery

Weavers-Guantanamera

Paul Simon-Diamonds on the Soles of their shoes

Alison Krauss-Forget about It

Beach Boys-Caroline No

Rickie Lee Jones-Danny's All Star Joint

Joan Baez-It's All Over Now Baby Blue

Muddy Waters-My Home is in the Delta

Duke Ellington-Concerto for Cootie

John Renbourn-At the break of Day

 

Next, I switched over to the new (i.e. optimized over the past few weeks) system and the improvement was staggeringly large.

In my friend's words, the improvement was "revelatory." 

 

Now I had achieved this new optimized state with a lot of individual steps over a period of weeks, after picking the brains of this thread and a few others to identify the factors most likely to improve things here based on my assessment of consensus and applicability of various reports.

 

It is of course impossible to absolutely identify which factors were the most significant contributors since not all combinations could be tried in a sane amount of time or in all combinations, nor did I have the patience.  Here are my performance observations in the order in which they happened.

 

> (was better)

>> (was much better)

 

new 7i7 NUC w/8G RAM and usb3.0 external 8TB HDD, both powered by HDPLEX

AL set to extreme realtime mode, hyperthreading and turbo turned off

 

Jussi's hqplayerd image on usb stick > old baseline system

hqplayer embedded under AL on usb stick > Jussi's hqplayerd image on usb stick

hqplayer embedded under AL in ramroot >> hqplayer embedded under AL on usb stick

hqplayer embedded under AL in ramroot w/ Apacer 4G ram >> hqplayer embedded under AL in ramroot

 

the last 2 really boggle my engineer's brain, but there is no doubting what i hear and that is the bottom line.

i would LOVE to understand from a scientific point of view how/why ramroot and Apacer ram can make such a difference to the sound quality but you guys have nailed it and i'm down the rabbit hole with you.

 

Impact:  better imaging, huge decays into silence between notes, greater detail w/o harshness, more and clearer bass lines

 

i will just mention that a few years ago, i had achieved a really fine level of digital playback (magic even!) using hqplayer and an ultraRendu in NAA mode powered by a SoTM battery power supply............then i changed a few things, lost the magic, sold a few things, swapped out a few more, and never regained the magic.......until now.......is this really all about power?

maybe all the other remaining wallwart/switching power supplies need to go?

 

 

Link to comment
  • 2 weeks later...
On 10/30/2019 at 9:52 PM, cat6man said:

can i assume the size of ramroot will stay the same, with the same features enabled as previously?

 

i ordered the update to AL2.0 but (unusual for me) have been waiting for the dust to settle before jumping in on the (b)leading edge.

 

now that folks have some experience, let me ask the above question again regarding RAMROOT.

 

i am running hqplayer only and have 4G of RAM.

i understand there is a very small increase in size (about 0.1G) but apparently the ram defaults values are now different

and 4G will not work with the new default values.

 

what are the new recommended settings for ram values that will run AL2.0 with hqplayer only with 4G of RAM?

with the number of people ordering/using 4G of RAM, would it be worthwhile to have a menu option to enable this simply?

 

what exactly were the default values for RAM as used in the 1.0 version of AL?

Link to comment
4 hours ago, hifi25nl said:

Version 2.1.0 is out!


1) Full system update
2) New AUR package manager yay
3) Removed all audio applications installed. Now you can install only what you need, reducing the size of the image (2.7 GB after drive installation)
4) Menu updated to version 206 with option to set MPD music directory, "RAMROOT configuration" in System menu, UPDATE menu completely re-designed. Now you can install/reinstall/remove all audio applications.

 

Previous version 2.0.0 can be updated with the last menu 206

 

great work!

thanks for the quick update

Link to comment

i just got my AL2.1 system up and running and am scanning my library right now.

 

however, i was not able to get ramroot working due to my current settings (defaults for ram) and 4G of ram available.

when i boot, it says that it cannot load into ram as is have only 3810M available and need 4061M.

i see there is a configuration menu option for ramroot.

what values would you propose i try?

i am only running hqplayer with no signal processing/upscaling/dither.

 

thanks

 

p.s.  i also ran afoul of the 'forgot which password to use' glitch..........took way too long before i remembered that root password was not the new download password nor the menu password but the old root password 'audiolinux0'...........D'UH

Link to comment
13 hours ago, elan120 said:

Update to menu 207 is the first step.  After that, pending what application(s) you want to install will determine the final memory requirement.  In my case, I am using NAA only, and total memory required was trimmed down from ~4.8GB to ~3.3GB.

 

that may work for the NAA but not for hqplayer-embedded at the other end where i get

only 3810M available and need 4061M.

 

therefore, the only option i see is to adjust the min_ram and min_zram settings.

 

@pierro

 

can you recommend values to use that will get me down 251M = 4061M - 3810M?

 

or is there anything else that can be deleted?

Link to comment

i now have hqplayer running in ramroot with 4G of ram and AL2.1

 

since i was originally short by 251M = 4061M (needed) - 3810M (available), the suggestion of

"try 200 zram 350 RAM " only bought me 200M (default was 250/500) so i further lowered

the values to 150M/280M which got me to 3772M (or 38M to spare).

 

two questions:

1.  would you expect any performance issues with this ram allocation?

2.  if i put some of the 38M available to better use, how should i apportion it between the two values i lowered?

 

thanks

 

@pierro

even though you don't have a 4G ram system, you could always use the memory requirement check on boot to see if a particular system configuration fits in a given amount available ram.  that wouldn't though tell you if there was a performance issue, just that ramroot would load.

 

 

 

 

Link to comment

i'm finding the system pretty slow responding, so i think i pushed it too far........unless there is some other way to reduce the size of the software in ramroot (so i can put the ram minima back up to defaults), i'll probably just upgrade to 8G of ram...........too bad, as 4G of ram was perfectly fine with AL1.0

 

 

Link to comment
On 11/14/2019 at 1:17 PM, cat6man said:

i now have hqplayer running in ramroot with 4G of ram and AL2.1

 

since i was originally short by 251M = 4061M (needed) - 3810M (available), the suggestion of

"try 200 zram 350 RAM " only bought me 200M (default was 250/500) so i further lowered

the values to 150M/280M which got me to 3772M (or 38M to spare).

 

two questions:

1.  would you expect any performance issues with this ram allocation?

2.  if i put some of the 38M available to better use, how should i apportion it between the two values i lowered?

 

thanks

 

@pierro

even though you don't have a 4G ram system, you could always use the memory requirement check on boot to see if a particular system configuration fits in a given amount available ram.  that wouldn't though tell you if there was a performance issue, just that ramroot would load.

 

 

 

 

 

 

great news!
the latest image loads hqplayer server (aka hqplayer embedded) into 4G ramroot without even lowering the default values for ram and zram!

 

super job Pierro!

 

 

Link to comment

one issue found here.

 

with AL1.0, i was very happy to use the HQPD controller app on my android phone and tablet to choose my music.

 

with AL2.1, i can no longer get the HQPD controller app to connect to the NUC running hqplayer embedded.

i know that the devices can connect via a browser.

a direct browser login to HQPlayer-embedded at 192.168.1.xxx:8088 works fine but HQPD controller running on the same devices now fails to connect, so something changed in how hqpd controller app needs to access hqplayer-embedded in 2.1

 

cheers

 

 

Link to comment
18 minutes ago, hifi25nl said:

HQPlayer is playing from HQPDcontrol now as I am writing this post. 

 

Please check Host in settings and refresh database in HQPDcontrol.

It is also better to stop HQPlayer and restart it after a change of configuration.

 

 

host is fine, but i get a 'connecting to 192.168.1.xxx' constantly at the bottom of the page.

normally, that message comes up and disappears after 1 second.

 

i'm re-scanning my whole music library now, and will try to refresh database in hqpdcontrol when it finishes.

Link to comment

re-booted with AL1.0 and the app connects and works fine

re-booted wtih AL2.1 and now app connects and works fine

 

no idea what the problem was..............i had tried rebooting AL2.1 multiple times without success, but after this sequence it worked.

 

seems ok now

 

Link to comment
29 minutes ago, lmitche said:

Over here, I've just started a comparison between AL2 kernel 5 with and without the BFQ scheduler on my NUC endpoint vs. Euphony Stylus/EP. The two AL configs do sound different and curiously without BFQ AL2 sounds very much like Euphony Stylus/EP.

 

I haven't changed my preference for AL2 BFQ but it is early days. Stay tuned.

 

Larry

 

 

BFQ scheduler?

i assume that is default ON in latest AL2?

probably somewhere in the thread, but where in menu is this selectable?

Link to comment
  • 3 weeks later...
  • 1 month later...

i'm running hqplayer inside audio-linux also, and it sounds great and is very stable.

 

the only issue i have, and it is pretty minor, is that after a period of inactivity or the first use of the day, the command from the android app HQPDcontrol has to be sent twice...........the first time i try to play an album, it seems to be waiting to connect, the second time i try to play the album, it loads and plays beautifully.

 

if i am playing music and change to a different album, there is no delay or need to try a 2nd time, only after a period of inactivity.

Link to comment
  • 4 weeks later...

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...