AudioLinux on Pi - a Quick Review

Discussion in 'Computer Audiophile: Software, Configs, Tools' started by Woland, Aug 16, 2021.

  1. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    TLDR Get AudioLinux for an easy interface to technical settings and for personal support.

    AudioLinux is, as the name hints, a Linux distribution tuned for Audio.

    The website is pretty busy, with a huge amount of information that is not easy to absorb. I do not get a good sense for the product from the website.

    I got an understanding from buying it ($69 with a year's support) and using it for a few weeks. I used it on my own Pi 4 with a HiFiBerry Digi 2 HAT, and with the loaner Pi 4 with Pi2AES. The main difference between those two is the connectivity options to DACs.


    How do you install it

    See my earlier post


    Why would you use it?

    One reason is that it has a lot of scripts that allow easy setup and configuration of the audio hardware. I reached for AudioLinux because I was experimenting with HQPlayer Embedded and was frustrated with the official HQPlayerOS. AudioLinux made WiFi setup easy, and makes switching between configurations easy. So I have switched between configurations as an HQPlayer upsampler, Roon Endpoint, Spotify Connect - that's been great. There's also the ability to do CD ripping, use the Pi as a file server and so forth, but I haven't explored those.

    [​IMG]


    A second reason is the support. Piero is almost on-call and you get rapid response and great help with any query. I can see this being the killer feature for a lot of SBAF people.

    A third reason is because you believe sound quality improves with streamer software. AudioLinux has a lot of tools that let you customise the setup to be optimized for audio. It was pretty easy to setup my Pi so that it has core isolation with audio applications having ownership of 3 of the 4 CPU cores, and with the whole OS loaded into RAM for snappier response.


    Audio Impact

    Using a DAC with good SPDIF processing and reclocking, I have dived pretty deep into changing hardware and software configurations in recent months. I haven't noticed any meaningful difference between changes in streamer hardware or software, once they work reliably. The difference is on the magnitude you get by changing interconnect cables, the ambient temperature or the orientation to the earth's magnetic fields and personal chakras.

    The one exception is that you might get the streamer to run much better upsampling than the DAC does internally. AudioLinux runs HQPlayer resampling (PCM only was tested) very well.


    Usability impact

    I have greatly preferred using AudioLinux vs Volumio. I like the high level menu system and scripts for doing the big stuff, and access to a proper Linux distribution for other tweaks and customization. I also like that Piero has been there when I have questions.

    AudioLinux is not quite as polished and reliable as RoPieeeXL, which is probably the best 'set and forget' Pi installation for most users and is also free.


    Conclusion

    AudioLinux gives a neat interface for tweakers, and awesome support.
     
    Last edited: Aug 16, 2021
  2. Thad E Ginathom

    Thad E Ginathom Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    12,050
    Trophy Points:
    113
    Location:
    India
    I have something of an oddity: a firewire audio interface. I was having huge problems with DPC Latency, and it didn't get a lot better when I switched to Linux. But I did make that switch, then and for-ever after, first to Ubuntu, then to Linux Mint.

    Initially, getting firewire audio to work was pure hell! I eventually discovered a distro called KXStudio. It made Firewire and Jack just work. Like magic. Comes with a lot of stuff to optimise for audio, and a heap of audio/visual software, of which one may install some or all.

    Back in the day, it was necessary to get realtime kernels. Then low-latency kernels. Then the generic kernel did the job anyway.

    Don't want to get further off the topic of the thread distro, just establishing a background to the question of why people are being charged money and paying for audio distros now? One is easily spoilt by Linux, and getting almost everything free: I am! Although I used to send beer money to the KXStudio guy until he stopped accepting it.

    Mind you, although I retired from being a techie and forgot lots, I used to earn my living on the Unix command line, and am certainly not afraid of it. On the other hand, one of the nice things about KX is the fairly-polished graphic front-end to JACK.

    @Woland: always interested to read anything that you post which is related to audio and any form of Linux! Hence, hear I am, albeit sans Raspberry Pi! I did used to use the [originally]-Logitech software to stream to a Logitech box. Sad day when the duh-what's-linux management of Logitech killed that line.
     
  3. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    That's fascinating! I wasn't paying much attention to audio over the years when the transition happened, but have encountered a lot of discussions about realtime kernels, especially when playing around with Ubuntu Studio version. But along the lines you suggest, realtime is currently tagged as being problematic and unnecessary for low-latency audio.

    That's kind of you! I'm largely self-educating as I undertake this audio adventure, and finding many more dead-ends than successes. It's surprising how much of the journey ends up being about learning history (like your realtime to mainstream description) or about the quirks of the people involved, rather than the technology per se.

    Maybe you could try running Audiolinux on a pc, and install Logitech Media Server (LMS) with it! It's not free, but it's not a huge indulgence.
     
    Last edited: Aug 17, 2021
  4. Thad E Ginathom

    Thad E Ginathom Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    12,050
    Trophy Points:
    113
    Location:
    India
    My audio is all on the desk these days, and I don't have a "proper" hifi setup. So there isn't anything for me to use LMS to stream to. But I used to. Did you know that there is a whole forum of people dedicated to the LMS stuff and devices? I'm fairly sure it is still active. I'll dig for the link ... I dug... slimdevices.com

    Another useful resource is the LInux Musicians forum. Except for ripping LPs, I didn't ever do much actual recording of music. I'm no musician. But it is a useful site covering the technology.
     
  5. Merrick

    Merrick A lidless ear

    Friend
    Joined:
    Jan 6, 2016
    Likes Received:
    11,077
    Trophy Points:
    113
    Location:
    Portland, OR
    Is this worthwhile as a replacement to Moode even if I’m not going to use HQPlayer? Can I UPNP to it with Sox resampling?
     
  6. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    Hi Merrick

    It looks like MPD (supported) has capability for UPNP and SOX. I haven't tried it though.

    Do you have SOX settings you use now?

    Edit: I have MPD UPNP installed from the menu, and it works well. However, it looks like SOX is not installed for upsampling.
     
    Last edited: Aug 17, 2021
  7. Gazny

    Gazny Friend

    Friend Contributor
    Joined:
    May 11, 2020
    Likes Received:
    1,598
    Trophy Points:
    93
    Location:
    open sky
    I would guess its major improvement is RT linux kernel. 60 usd might be worth your time instead of compiling it on a rpi.
     
  8. Merrick

    Merrick A lidless ear

    Friend
    Joined:
    Jan 6, 2016
    Likes Received:
    11,077
    Trophy Points:
    113
    Location:
    Portland, OR
    Currently I just have it set to in Moode to resample everything to 24/88. I notice there is a resampling feature built in, curious how good that is in comparison.
     
  9. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    I'm also using a Gungnir Multibit. I've found HQPlayer's resampling to make a major improvement to it and a Magnius, but haven't had any positive experiences with SOX. If you're interested, try running the HQPlayer Desktop trial on your computer.. it has a player app (clunky, but it works) so you can get started in minutes. The trial works for 30 mins at a time.
     
    Last edited: Aug 17, 2021
  10. Merrick

    Merrick A lidless ear

    Friend
    Joined:
    Jan 6, 2016
    Likes Received:
    11,077
    Trophy Points:
    113
    Location:
    Portland, OR
    I did try putting it on my SD card with Etcher but never got it to load.
     
  11. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    Try out the sound quality from the desktop version on your computer before doing the fiddlier stuff with a pi's sd-card.
     
  12. Thad E Ginathom

    Thad E Ginathom Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    12,050
    Trophy Points:
    113
    Location:
    India
    I get it that the thread is about the software, not the cost, and I don't want to detract from that. But different kernels are available (eg for Ubuntu-based stuff) without compiling your own.

    Mind you, someone has to make it all work together. I installed low-latency on my Linux Mint recently, and my audio stuff, far from improving, collapsed! But that's what distros are about.
     
  13. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    A new observation..

    I recall getting the most computationally intensive PCM upsampling, Sinc-M, working on HQPlayerOS.

    I expected the system tweaks on audiolinux to boost capability a little ... However, I'm getting audible artifacts if I try to use the sync-M PCM filter on AudioLinux, and even sinc-L may have problems. When I use a system monitor, there doesn't seem to be much stress on the system -- go figure. Maybe this is the same issue @Thad E Ginathom found with using a realtime kernel.

    Edit: This might not be about Audiolinux - it might be because I'm using a different audio HAT, for example.. I'll test this hardware with HQPlayerOS and verify.
     
    Last edited: Aug 18, 2021
  14. Thad E Ginathom

    Thad E Ginathom Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    12,050
    Trophy Points:
    113
    Location:
    India
    No, my problem was that the Cadence/JACK/firewire stuff simply wouldn't run. It was the low-latency kernel I was trying. I wanted to get the interrupt optimisation working. That is what can cause clicks and dropouts. That is fairly basic to audio optimisation, but it seems it wasn't working for my linux. It started to happen again recently, after a complete motherboard/processor/guts upgrade.

    Yes, I've had audible problems with sample-rate changing, both up and down. I try to avoid it. I try to keep all my music 44.1. Even if there is a valid rate for higher sampling rates, it would not apply to the current state of my hearing.
     
  15. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    You can't avoid upsampling.. either it's done digitally by the DAC chip or by an analog filter smoothing after the DAC chip. Somehow the digital dots have to be linked to make analog movements in the air. They're not going to quantum leap to the music!
     
  16. Thad E Ginathom

    Thad E Ginathom Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    12,050
    Trophy Points:
    113
    Location:
    India
    What the DAC does is all part of the magic of the DAC. People have tried to explain stuff to me like xsync, and I *sort-of* see it, but my basic innumeracy severely limits my ability to understand digital sound, despite having an interest.

    It is resampling to a different sample rate in software, before sending to the DAC, that I have noticed problems with.
     
  17. coin777

    coin777 Acquaintance

    Joined:
    Feb 8, 2019
    Likes Received:
    34
    Trophy Points:
    18
    Location:
    poland
    There are NOS DAC that don't upsample audio. There are even NOS DAC that dont have an analog filter like Holo Audio May
     
  18. Woland

    Woland Friend

    Friend
    Joined:
    Jan 13, 2021
    Likes Received:
    1,192
    Trophy Points:
    93
    Location:
    a friendly land
    I'm playing around with Audiolinux and HQPlayer's Sinc filters.

    Typically the upsampling is 4x (44100 to 176400, or 48000 to 196000), all PCM.

    Computational intensity for 44100:

    Sinc-S 16,384 taps
    Sinc-Mx 250,000 taps
    Sinc-L 524,280 taps
    Sinc-M 1,000,000 taps


    sinc-S sinc-filter with adaptive number of taps (4096 x ratio). Very sharp cut-off and high attenuation. Integer apodizing
    sinc-M sinc-filter with one million taps. Very sharp cut-off and high attenuation. Integer apodizing
    sinc-Mx Constant time version of sinc-M filter. Filter length is constant in time, with million taps at 16x PCM output rates. Integer apodizing
    sinc-L sinc-filter with adaptive number of taps (131070 x conversion ratio). Extremely sharp cut-off but average attenuation. Integer Non-apodizing


    Here's what I'm finding:

    Sinc-S will run on even the most conservative AudioLinux settings (Green profile)
    Sinc-Mx requires standard settings (Orange profile)
    Sinc L and M get taxing. Red profile + overclocking to 2GHz seem to be enough. Here's how to do the overclocking.
     
    Last edited: Oct 31, 2021

Share This Page