Re: [LAU] Experiences with miniDSP AVB products ?

From: Len Ovens <len@email-addr-hidden>
Date: Mon Oct 01 2018 - 20:17:47 EEST

On Mon, 1 Oct 2018, nik@parkellipsen.de wrote:

> Hmm but you wrote in a previous mail that there might be a chance to make it
> work if sb is willing to put the effort in?
>
> Are there any resources on that ? I only found a thread in the ardour
> forums, but couldn't extract any info about where to start. (tbh i just skimmed
> it quickly).

Both put the effort in and have the hardware to test with.

In my mind there are two paths one could take, jack or alsa.

ALSA: in the case of ALSA one would create a driver that would start up at
system start or manually after start (manually could mean using a GUI).
The device once connected would be static. It would have a fixed number of
audio ports and possibly break if that device went away (same as a USB
device). This would be fine for many uses. It would have to be all
inclusive though.

Jack: This case is or could be much more flexable. In this case I would
use the NIC's clock (assuming an i210 or similar that has such a thing)
and use it to time the dummy backend for jack. There would be some code
involved to either make this clock master or slave. I don't know if this
already happens with the code as it stands. The i210 has some extra pins,
it would be great if one of them could be programed to put out wordclock
as well for syncing a local audio interface to the whole works.

A GUI would list known interfaces that can be connected (those that have
the right SR at least) either as a grid (like Ardour's audio port
connection GUI) or as a list like qjackctl's Connections window. I suspect
as things get more complex the patchage style might be hard to use but
that is another possibility. This GUI would not just make a jack
connection, but would have to first start a jack client that is connected
to the remote device and then connect it. As many AVB devices only seem to
deal with 8 channel streams at a time, connecting one port would mean that
8 ports would show up on the jack graph and the GUI would connect any
other port for that stream without creating another jack client. Devices
could be added or removed at any time and if an AVB client is no longer
needed (has no connections for more than 10 seconds) it could vanish.

I do not know if there is already code that allows a jack client or
connection GUI to advertize stream availablity to the world but, as with
all things AVB, that would be yet another piece of code.

Going this route would also leave things open for AES67 using the same
clock. It would seem possible to connect to both AVB and aes67 devices at
the same time. Assuming a suitable aes67 jack client was available.

I had thought to work on something like that and have the i210 NIC but not
A) another computer to act as an end point or B) (better) an AVB endpoint
to play with. In my case I would prefer inputs.

Another solution, though probably more costly, is to purchase something
that has both USB and AVB and use it to access a remote audio device. The
Motu 8D for example (8 aes3 i/o) is 800CDN or the 8A (8 audio i/o) is
1050CDN. The advantage is that you end up with local audio i/o that is in
sync with your remote. (well sort of, you may find aes3 inputs go through
a SRC stage unless you provide sync to any aes3 input box)

If all you want is a cheap network audio solution... an old computer with
8 outs and runs jack... use zita-njbridge... be happy. Almost all HDA
mother boards have 8 outs for a number of years, but it may be almost as
cheap to use a new atom based MB with 16v PS as the minidsp 8n. (or a
nuc... well maybe not, they have 7.1 through hdmi but only 4 physical on
the $200 model) Pulse also offers networked audio... but I have no
experience with that and pulse is known to be lossy (as in samples just
get lost)

--
Len Ovens
www.ovenwerks.net
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@lists.linuxaudio.org
https://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Mon Oct 1 20:15:02 2018

This archive was generated by hypermail 2.1.8 : Mon Oct 01 2018 - 20:15:02 EEST