Thank you for your comments. I will look into the zita-njbridge. I had not heard of it.
My application is SDR (Software Defined Radio). Of the recent commodity products, the
demodulated audio or I/Q are made available to software apps via:
1) Conventional audio streams. For example, I use a LP-PAN 2 panadapter attached to my
transceiver along with a Xonar U7 to get 192 KHz/24-bit stereo (I/Q) data to various apps
running on both Windows and Linux. I guess that most ham radio operators these days have
some digital modes software that they use on a computer using audio I/O, and lots of hams
do use the 192 KHz sampling that a few sound cards offer.
2) Various USB-based protocols for up to 10 to 15 Msamp/sec 16-bit complex data (320 to
480 Mbit/sec).
3) Various IP-based protocols up to gigabit network limits or about 20 Msamp/sec of 16-bit
complex data.
I have been looking for a way to unify the multitude of hardware interfaces to a single,
IP-based protocol. Ethernet seems like the way to go for many reasons.
One of the responders mentioned something about 10gbit Ethernet. Prices have come down
amazingly - you can buy a 10gbit switch now for <$500, so I would not rule that out. My
specific requirements are not that onerous. 1 gibabit Ethernet is probably adequate for
most amateur radio operators.
It does seem to me that thinking of audio as something that a human being can directly
hear is only one perspective. Cheap computers are easily fast enough to handle streaming
multiple MHz of bandwidth over the network, and doing all sorts of
modulation/demodulation.
I think that the professional audio and SDR worlds are really very similar. SDR operators
are not interested in what audiophiles think of as "effects", but nevertheless
we are very interesting in decoding and encoding all kinds of analog and digital
modulations. It seems to me that both communities could use much of the same
infrastructure for sharing data between applications and over networks.
When I started looking into what exists today in the way of standards for communicating
streaming audio over Ethernet, Jack seemed like a very interesting option. I am still
investigating.
James
________________________________________
From: Len Ovens [len(a)ovenwerks.net]
Sent: Sunday, December 18, 2016 10:23 AM
To: Schatzman, James
Cc: linux-audio-user(a)lists.linuxaudio.org
Subject: Re: [LAU] Jack Audio Appliance?
On Fri, 16 Dec 2016, Schatzman, James wrote:
Is there a Jack Audio Appliance commercially
available?
If not, I would like to build one. Here are my requirements
1) Accepts 4 or more input audio streams at Xonar 7 rates (24-bit
channels at 192KHz).
So not audio then. What is the actual frequency you are interested in?
Audio is 20 to 20k, so 48k sample rate. Even ADCs that will output 96k
(let alone 192k) may have the input frequency range tied well below 30k by
the analog circuitry. So an off the shelf solution for what you are doing
is unlikely.
2) Makes the audio streams available to Jack clients
on the network.
May I suggest zita-njbridge for network clients? It seems more stable than
jack-net if SRC is ok.
3) Is compact and reliable.
As you would be building it, size and reliablility are in your hands :)
What would be a reasonable compact platform to host
such a device
(processor speed/cores, memory, etc.)?
One of the multicore Atom chips would be my choice, they deal well with
low latency audio in my experience (sometimes better than i3-7 and much
easier to deal with than arm)
Also, is it feasible to jack up the data rates even
higher - say to
several megasamples per second?
This is the statement that made me feel that it was not audio you are
interested in. It also makes me wonder why the 24 bit spec, I suspect 16
would be fine. high frequency ADCs that I have looked at are mostly 8-12
bit. Fine for making an oscilliscope for example. If you are doing rf
work, I don't know how far down the noise floor needs to be to keep BW
down... I just havn't played with it at all. That would determine the
needed bit depth.
--
Len Ovens
www.ovenwerks.net