WTB cheap Roon endpoint

[Grover Gardner]

Jim,

I am having a problem with Bubbleupmpd or upmpdcli at track transitions I need to find out. I get some dropouts when reading a new track. May be related to filling the buffer (or lack of a buffer).

David

I am also having the same issues. It’s the buffering issue. It only happens when I change the track, but no drop outs in a playlist. It happens on Redbook and DSD.

I am using Audirvana —> upmpdcli —> mpd (gentooplayer).

I had this problem with Audirvana for a while. I think I increased the MPD buffer and that solved it at that point, but with the latest build of gentooplayer it's gone away. You might also try updating upmpdcli, or even reverting to a previous version, which gentooplayer allows you to do.

[Grover Gardner]

David,

I was suspecting the buffering between audirvana and upmpdcli as the problem. I am using default config for both audirvana and gentooplayer.

Both Audirvana and GentooPlayer are specialized for audio playback. The end user should not have to tinker with configuration to get "acceptable" quality output from them.

Well... Gentooplayer is still a work in progress and is a little tweakier than Moode or Volumio. Moode never gives any problems, IIRC, but I don't like the sound as much.

[Cogito]

Shashi,

There is NO buffer configuration setting for upmpdcli (I checked).

David

Thanks. I will check gentooplayer this evening.

[David McGown]

Grover, Shashi,

I think with any of these players built on opensource applications (mpd, upmpdcli), is that they may not be changing the underlying application code, but adjusting their parameters either in config files or custom building the applications with changes in default parameters values to tailor to the distribution. If they actually change the code (fork), they have to distribute the source code for those components per open source licensing. Of course, if they can do it if they feel it is warranted. Then you get into how much of the Linux system overhead functionality you want to retain, whether you run RT vs standard kernel, and how much kernel functionality you want to have. Alot of these things can influence performance, and getting a tight system that runs on limited memory and CPU resources can increase the availability of the CPU and memory for audio.

I guess I am not making much of a point, except to say that tweaking mpd and upmpdcli is perfectly legitimate to do, you are not going to break things if you know your way around a Linux system (keep a copy of the original config file). You can SSH into the system, sudo for commands needing root functionality (for changing *.conf files), of course have an editor you can use (vi or nano) at the command line. Reboot or restart your daemons to see if the change helps or hurts. If something breaks, just SSH in and reverse it (or figure out your mistake in the editing - more a problem with vi). Nothing broken that cannot be fixed.

David

[Cogito]

Shashi,

There is NO buffer configuration setting for upmpdcli (I checked).

David

Thanks. I will check gentooplayer this evening.

In addition to skipping, gentooplayer sounded bland compare to Ap-Linux. It lacked the dynamics and energy of AP-Linux. The major cause of this sonic difference between the two systems is how their kernels were configured.

Gentooplayer gives option of 5 kernel profiles. Default is Profile 5 which really does nothing. Profile 1 is the best for audio playback. It dedicates thes CPUs based on type of tasks (see below):

Profile1:
Move all system and kernel processes to CPU0
Isolate CPU1,2 and 3
Move process and USB IRQ to CPU1
Move process and LAN IRQ to CPU2
Move the Player(s) to CPU3
Give priority to processes and players

Once the kernel was reconfigured to Profile 1 specs, the gentooplayer totally changed. Now, there is more life in the music, sounds closer to AP-Linux. And all buffering issues are also gone (I did not change mpd config).

Lesson: AP-Linux comes pre-configured with best kernel settings for audio, where as gentooplayer lets you experiment and discover (kinda rediscovering the wheel).

[Grover Gardner]

Shashi,

There is NO buffer configuration setting for upmpdcli (I checked).

David

Thanks. I will check gentooplayer this evening.

In addition to skipping, gentooplayer sounded bland compare to Ap-Linux. It lacked the dynamics and energy of AP-Linux. The major cause of this sonic difference between the two systems is how their kernels were configured.

Gentooplayer gives option of 5 kernel profiles. Default is Profile 5 which really does nothing. Profile 1 is the best for audio playback. It dedicates thes CPUs based on type of tasks (see below):

Profile1:
Move all system and kernel processes to CPU0
Isolate CPU1,2 and 3
Move process and USB IRQ to CPU1
Move process and LAN IRQ to CPU2
Move the Player(s) to CPU3
Give priority to processes and players

Once the kernel was reconfigured to Profile 1 specs, the gentooplayer totally changed. Now, there is more life in the music, sounds closer to AP-Linux. And all buffering issues are also gone (I did not change mpd config).

Lesson: AP-Linux comes pre-configured with best kernel settings for audio, where as gentooplayer lets you experiment and discover (kinda rediscovering the wheel).

Thank you for that tip, Shashi. I will try that. Using GP with the Allo USBridge is tricky because not all the kernels work well with it. Maybe I should try a NUC.

[Cogito]

In addition to skipping, gentooplayer sounded bland compare to Ap-Linux. It lacked the dynamics and energy of AP-Linux. The major cause of this sonic difference between the two systems is how their kernels were configured.

Gentooplayer gives option of 5 kernel profiles. Default is Profile 5 which really does nothing. Profile 1 is the best for audio playback. It dedicates thes CPUs based on type of tasks (see below):

Profile1:
Move all system and kernel processes to CPU0
Isolate CPU1,2 and 3
Move process and USB IRQ to CPU1
Move process and LAN IRQ to CPU2
Move the Player(s) to CPU3
Give priority to processes and players

Once the kernel was reconfigured to Profile 1 specs, the gentooplayer totally changed. Now, there is more life in the music, sounds closer to AP-Linux. And all buffering issues are also gone (I did not change mpd config).

Lesson: AP-Linux comes pre-configured with best kernel settings for audio, where as gentooplayer lets you experiment and discover (kinda rediscovering the wheel).

Thank you for that tip, Shashi. I will try that. Using GP with the Allo USBridge is tricky because not all the kernels work well with it. Maybe I should try a NUC.

NUC may not work. Low power CPUs are used in NUCs for easier heat deception which means they may not have more than 2 cores. To implement GP Profile 1, you need 4 cores (real cores, not threads) which means you would need desktop version of i5 CPU. And, I would turn off hyper threading, if any, in the BIOS.

[Grover Gardner]

In addition to skipping, gentooplayer sounded bland compare to Ap-Linux. It lacked the dynamics and energy of AP-Linux. The major cause of this sonic difference between the two systems is how their kernels were configured.

Gentooplayer gives option of 5 kernel profiles. Default is Profile 5 which really does nothing. Profile 1 is the best for audio playback. It dedicates thes CPUs based on type of tasks (see below):

Profile1:
Move all system and kernel processes to CPU0
Isolate CPU1,2 and 3
Move process and USB IRQ to CPU1
Move process and LAN IRQ to CPU2
Move the Player(s) to CPU3
Give priority to processes and players

Once the kernel was reconfigured to Profile 1 specs, the gentooplayer totally changed. Now, there is more life in the music, sounds closer to AP-Linux. And all buffering issues are also gone (I did not change mpd config).

Lesson: AP-Linux comes pre-configured with best kernel settings for audio, where as gentooplayer lets you experiment and discover (kinda rediscovering the wheel).

Thank you for that tip, Shashi. I will try that. Using GP with the Allo USBridge is tricky because not all the kernels work well with it. Maybe I should try a NUC.

NUC may not work. Low power CPUs are used in NUCs for easier heat deception which means they may not have more than 2 cores. To implement GP Profile 1, you need 4 cores (real cores, not threads) which means you would need desktop version of i5 CPU. And, I would turn off hyper threading, if any, in the BIOS.

Ah, I see.

[Cogito]

Shashi,

BTW, I was looking at the HDPlex cases, and will likely go that route when I setup my "SuperServer" with a motherboard and PCIe cards.

Yesterday I was at MicroCenter in Fairfax to pickup a SSD. They had a large display of computer chassis and cooling systems geared towards gaming industry. It got my curiosity and I checked out one full size tower chassis. It had a PS and 3 120mm chassis fans (mounted inside) in it, all running. Standing next to it, I could not hear any fan noise. I could see the fans running, so opened the transparent side panel and still could not hear any noise. I had to move my head within aboud 10" before I could hear the noise.

Guess what, they are Magnetic Levitation (not a typo) fans, each costing about $130. They are 24dB silent. I remember seeing CPU fans, Power Supplies also of the same brand.
Technology has improved so much, fan noise is not an issue anymore in audio rooms.

Check them out.

[David McGown]

Shashi,

It is more than the physical noise. Running a fan, particularly with PWM control, introduces electrical noise into the computer and its power supply. Of course there is already noise from the DC to DC converters, but that is unavoidable to address the numerous voltages that on-board component need. A secondary rationale for passive cooling is to, as much as possible, reduce as many of those electrical noise sources as possible. Also using an external LPS for powering the main board and internal components (incl. USB and Ethernet cards) gets the noise from a big fan cooled switching power supply from polluting the computer power bus.

If one can get by with passive cooling, I feel it is better overall. But it will limit how powerful of a CPU you can use due to heat dissipation unless you find an case that can dissipate the heat under load.

David