[LAD] FOSS Ethernet Soundcard - IETF Ethernet AVB

Folderol folderol at ukfsn.org
Wed Dec 2 23:06:46 UTC 2009


On Wed, 2 Dec 2009 21:49:10 +0100
fons at kokkinizita.net wrote:

> On Wed, Dec 02, 2009 at 07:21:46PM +0000, errordeveloper at gmail.com wrote:
> 
> > their XS1 chips seem to be really great,
> > their are basically every innovative and open-source minded.
> > the official toolchain is LLVM-GCC based.
> > 
> > you can use C, C++ or their own XC.
> > XC is basically C with some stuff omited (like goto and floats)
> > and XMOS IO stuff added, don't just say WTF, look at it first!
> 
> It's not just IO stuff added. Their 'C' has parallelism 
> at the statement level, and this supported directly by
> the hardware - no need for such thing as threads supported
> by an OS.  
> 
> The chips are really a re-invention of the INMOS Transputer
> of 25 years ago, and the special 'C' language can be traced
> back to Occam, the Transputer's native language. 
> 
> It's an immensely powerful combination, and IMHO years ahead
> of anything else on the market - even if the basic ideas go
> back at least 30 years.
> 
> Ciao,

Well, I've just done some (not so) light reading, and this does indeed
seem to tick all the boxes. There seems to be quite widespread support
for AVB in the industry, which is always good, although I can only find
this one vendor producing development kits.

Also, there is one very real problem. The core chip itself is a
massive BGA device. This would mean we would ether have to always use
development boards (expensive and clunky) or would have to find a
contractor that would mount these, possibly on a carrier board/socket of
some sort. This would only be practical for bulk orders.

As a dedicated audio chip it would however, make life a lot easier (not
withstanding the single source availability). If we were to redirect to
this, it would leave a couple of major questions unanswered.

How is the link to Jack implemented in the computer?

How do we make the AD/serial conversion.

What control protocol should we use (OSC)?

Oh dear, that's three :)

It occurs to me to wonder how clear and well defined the AVB protocol
is, and if it is possible/practical to implement it ourselves on a
platform of our choice. This is sort of half-way between wheel
re-inventing, and non-dependence on a single source.

What do others think?

-- 
Will J Godfrey
http://www.musically.me.uk
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.



More information about the Linux-audio-dev mailing list