[Jack-Devel] AES67 / SMPTE ST 2110-30

[Chris Caudle]

On Mon, October 2, 2017 5:12 am, Philippe Bekaert wrote:
> I’m also in contact with people at
> Merging regarding their VSC driver in progress for Linux.

Did you get a copy of the ALSA driver?  The current git head does not
compile, I waited to see if merging would correct that on their own, but
still not corrected after a month, so I need to ask what is happening
there.  Not encouraging that it is actually getting any attention if the
source can go more than a month in a non-buildable state.

> I'm aiming for a jack client application, and perhaps a jack audio driver
> (like the net drivers). The latter allow to have jack running at AES67
> clock, and being fully transparent (no sample rate conversion). But the
> former, a client application, is probably better if you just want to
> receive and monitor / send AES67 on / from a computer having a sound card
> already.

I recommend a jack audio driver.  If you want to monitor from a local
sound card you can always load zita-j2a and zita-a2j resampling bridges,
that work has been done so no need to implement twice what Fons has
already implemented.  The advantages of a net driver is that you can in
principle collect streams from different devices that are in the same
clock domain and have a large virtual sound card.  For example, in a home
recording environment you could have a Ravenna output device in your room
with a computer, have a Ravenna in/out device in a different room with
someone playing keyboards, and a Ravenna in/out device in the house large
room playing guitar, and all three of those physically separated devices
could be treated as one large audio interface.
I am not sure if jackd will let a driver load and then change the number
of ports at a later time, so how configuration would work is a task that
needs to be investigated early (e.g. would you have to discover and
configure Ravenna streams first, then load the net driver, or could the
net driver be loaded with some arbitrary number of ports declared to
jackd, and various Ravenna streams connected and disconnected from those
declared ports at run time).

I am still struggling to find reasonable cost network connected
interfaces.  The small device from Hasseb I linked last time is reasonable
cost, but very low feature, I would prefer at least a duplex stereo
device.  Focusrite have a new two channel Dante device which could run in
AES67 mode, but it is output only, and I think is around US$400.  The 4
input 2 output Cymatic device is still not available yet, and even though
the 24 channel Cymatic device has reduced price by 50% lately, that is the
base USB version only, so by the time you purchase the Ravenna option card
it is still around US$1000.  Not a bad price for 24 channels, but still
could be considered quite a lot to spend on a hobby investigation project.

If a Windows computer is available, the Lawo R3lay virtual sound card for
Windows is the best option to get started.  I tried it and was able to
send audio between two Windows laptops.  You will need a PTP master on the
network, that VSC will not act as clock master, but I am running a
BeagleBone Black as PTP master (uses chronyd to sync time to ntp, then
provides the ntp time to the network using linuxptp).  You could also use
your existing linux machine to act at PTP master, it would just have more
jitter in the PTP performance if you do not have hardware timestamping
capability (the BeagleBone has hardware timestamp in the MAC layer but not
the phy, not ideal but still fairly high performance).

-- 
Chris Caudle