[LAU] MOTU AVB discussion from LAC

[Fernando Lopez-Lezcano]

On 3/30/19 10:37 AM, Max wrote:
> On 30.03.19 04:52, Paul Davis wrote:
>> On Fri, Mar 29, 2019 at 4:57 PM Max <abonnements at revolwear.com 
...
>> fernando mentioned a post on "linux musicians" which describes each one.
> 
> for reference, I guess the thread is this one:
> https://www.linuxmusicians.com/viewtopic.php?f=6&t=18046

Yes, that is the one. Emailed to me by Anders Hellquist when he read my 
lac paper (thanks!).

> It's a rather long thread with some interesting tidbits here and there. 

It has a LOT of information!

> Someone even wrote a test program, but one has to be signed in to see 
> attachments on this forum.
> 
> Three takeaways from this thread:
> 1. AVB might be supported by the kernel at some point, which would allow 
> to use the AVB interfaces without the USB connection.

That is what I am waiting for... A very long time ago I managed to get 
the OpenAVnu code compiled and got a linux machine to recognize and sync 
with a Motu's clock - never got further (and that was not particularly 
easy).

 From the thread it appears that at least one user has working code for 
streaming one AVB stream (8 channels) back and forth. Sounds fantastic.

There was also a paper in LAC about syncing jack with an AVB clock so 
that is also a step in the right direction.

Still, for systems like the one in our Listening Room and Stage going 
the USB way as an interface for end users seems like the best option. I 
would probably not want to expose the whole AVB network behind the scenes...

> See: https://github.com/L-Acoustics/avdecc for a multiplatform 
> implementation.
...
> The old firmwares are downloadable here:
> http://motu.com/techsupport/technotes/firmwarechangelog/
> 
> 3. There is no way to change the number of channels through ALSA or via 
> the configuration panel of the web-interface. It's static.
> 
>>     Why can't I tell ALSA to use only the first 2, 4 whatever channels of
>>     the device? I can only open the device if all channels are
>>     connected. Is
>>     this always like that or is that a limitation of MOTU's 
>> implementation?
>>
>> ALSA doesn't work that way. You open a device. You tell it how many 
>> channels you want to use.
> 
> That's what I thought, but apparently it only succeeds to open an ALSA 
> device if the number of channels match.

That is true if you use the ALSA "hw" interface, which is what you want 
to do if you use Jack (ie: direct access to the hardware with nothing in 
between). The number of channels you open has to match what the hardware 
can do.

If you use instead the ALSA "plughw" interface you could open it with 
less channels and ALSA would insert stuff in the middle to account for 
the discrepancy but you don't want to go down that rabbit hole with 
Jack. You want direct access to the hardware.

Other stuff I found in the thread (or in comments from Anders):

With a usb kernel audio driver quirk you can access the proprietary usb 
streaming endpoints (not class compliant) of the audio interface, but 
that does not seem to help.

Another one is an incantation for switching the interface between 
"proprietary mode" and "class compliant mode". That might explain why 
sometimes a Linux laptop does not see the interface after a Mac has been 
using it and probably also the other way around. When I hit this I never 
had time to investigate. I would probably need a GUI exposed button to 
fix this...

And probably other juicy stuff, I have to re-read the thread again... I 
was busy at the time with other stuff. I have to sign up to be able to 
see the patches and other stuff.

Anders emailed me a beta version of the firmware for the UltraLite that 
him best performance - I think it did fix problems for someone that was 
at LAC...

-- Fernando