On 2014-04-18, Sampo Syreeni wrote:
What's not to like? ;)
(Obviously how you'd synch the clocks of millions of converters.
But there we then have the kind of back channel algorithm available over
distributed analog networks which doesn't work too well over the digital
ones. There is a certain analog, stochastic, swarm algorithm, derived
from our understanding of how swarms of certain bioluminescent insects
synchronize their light. It can be guaranteed to either take everybody
into global synch, or exactly into 180 degree phase discrepancy with
their neighbours. So, using some simple logic and control theory, and
compounding the basic protocol, you can not only guarantee full global
synch starting from whichever condition, but also exponential global
convergence to zero mutual phase shift. Subject only to noise, which you
can then too engineer into reasonable and scalable levels, using old
school control theory.
This is why I say you need a backchannel in order to synch the
converters: you could well do upto about 32 or 128 channels simply by
engineering your clock distribution tree *just* right. At about those
numbers, you could also still run a high time constant PLL over
ethernet, and eventually land at sample accurate registration between
multiple, separated converter boards. But beyond that, you just can't do
it easily via any other means than a dedicated, backchannel swarm
algorithm.)
--
Sampo Syreeni, aka decoy - [email protected], http://decoy.iki.fi/front
+358-40-3255353, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2
_______________________________________________
Sursound mailing list
[email protected]
https://mail.music.vt.edu/mailman/listinfo/sursound
