Re: [Jack-Devel] Writing an IEEE 802.1BA (AVB) compliant network backend for Jack2

On Tue, March 1, Torben wrote:
> i dont really understand this.
> you mean streaming from harddisk to multiple endpoints,
> all playing phase synced ?

Torben,
I couldn't really tell if you understood the scenario proposed.
Consider the situation of a DAC at the mix position playing back, and a
separate DAC farther away, playing the monitor mix for the musician.  Two
separate endpoints that you would want in sync.
The ravenna page even describes audio over IP direct to the monitors, so
you have an Ethernet cable to each speaker.  Definitely want those two
endpoints to play phase synced.

> with soundcards that generate their own clocks, and dont allow
> for rate control ?

No, of course the endpoints have to allow for rate control. That is why
all the talk of IEEE-1588 or 802.1AS.  Network based clock
synchronization. All the endpoints synchronize to the PTP clock, and
generate sample clocks based on that.

> if one allows 3ms of diff, we can do this with netjack

1us or better is the goal.

> with a reasonably good ptp timer in the kernel.
> this is all easy stuff.

Florian's point was that "reasonably good" is context sensitive, and in
the case of Ravenna, the goal is to replace the word clock or AES11 synch
signal with network time clock. AES11 requires that output samples are
within +/- 1us of the nominal frame synch signal at 48ksamp/s, or +/-0.5us
at 96k.  I guess maybe 0,5us if you are in Europe.
I don't fully understand the implications of that spec for packetized data
delivery (as opposed to AES3 or AES10 or AES50 which stream the data
continuously),  which is why I think maybe the software implementation in
the Linux kernel will be OK as long as the average frequency has no
difference, and the maximum error is bounded, but perhaps not quite as
tightly as 1us.

-- 
Chris Caudle

1299040407.27429_0.ltw:2,a <4db49c86164a406e62c8041fcbbf592e.squirrel at email dot powweb dot com>