>Well, I am a big fan of feedback loops, and the ability to make them
>with paul's model will allow us to go where fripp and eno could only
>dream of going. My simple mind says that there is no need to break the
>simple, single "process (nframes)" model if every route connected in a
>feedback loop contains a delay line >= n samples. If this is correct
>then the only issue is to address the situation where a route does not
>contain a delay line >= n samples. Since this is a low latency
>environment, n will correspond to something less than 5 msec.
I think you're missing the very obvious situation I was thinking of
Tom.
Imagine a data stream from disk coming into an Ardour mixer strip
(using the visual model here), and then being sent as part of a submix
to a bus. Another mixer strip reads from the bus, does "global"
processing on the submix, and then sends it out to a physical
channel.
At the moment, the data emerging from the channel will be delayed by
one engine cycle. Not good.
>It *IS* the simplest way to think about it.
>It is *THE* simplest way to think about it.
>It is the *SIMPLEST* way to think about it. (can you tell that I am
>enthused?)
>
>This diagram represents the best model that I have ever seen. It is so
>simple that it is also deceptive. Please encourage it.
Your enthusiam has been noted :)
--p
