[linux-audio-dev] Project: modular synth editor

[Dave Robillard]

On Fri, 2004-01-16 at 19:05, Fons Adriaensen wrote:
> On Fri, Jan 16, 2004 at 02:46:25PM -0500, Dave Robillard wrote:
> 
> > I'll just assume you have your reasons.  But just because AMS is going
> > to have metadata doesn't mean some modular synth can't use LADSPA
> > without doing so.
> 
> Nor did I say so. To illustrate this point, let me mention the problem of a
> voice controller (VC) in a polyphonic patch. How is the VC to know when a voice
> has ended and can be re-allocated to a new note ? This requires some feedback
> from (typically) the envelope generators. So either you make this explicit
> (a solution I do not dislike, but definitely non-standard), or you use a hidden
> data path, and in a modular system that means metadata.

Good example.  I'll grind away at it in my head over the next few days
and see what happens. :)  I'm thinking voice allocation could be handled
without really caring what each voice is actually doing at the moment
(it's like a scheduling problem really), but more on that later.

> > > OK, that maps to the requirement that almost everything that has a slider
> > > must also be 'voltage-controlled'.
> > suppose, yes.  Is this a problem?  The added flexibility would be huge.
> 
> Depends on how it's done. If a slider corresponds to a control input, then
> one would expect the slider to follow the control voltage if it is connected.
> That could mean a *lot* of CPU load if you have hundreds of sliders dancing the
> night away. Two solutions: 1. Add the control voltage to the slider value
> instead of replacing it, but not all parameters can be 'added' in a way that
> makes sense, or 2.  eliminate the slider and all other widgets, and turn
> each of them into a separate module that can optionally be created and 
> connected to the control input. That means a completely different GUI concept.

Oh yeah, the sliders still responding to input is definately not what I
was thinking.  It's an 'or' thing.. slider _or_ control port.  SSM does
this kinda thing come to think of it, if you plug into the port (all are
there by default), the control disappears.

> > > Strange. I'm using AMS in poly mode with a real MIDI controller almost
> > > daily. I'm sure J.S. Bach would complain if some of his notes fall out :-).
> > > Anyway there's no such thing as "dead on simultaneous' as MIDI is a serial
> > > protocol. And even if you play a MIDI file, chords will be serialised at
> > > the ALSA port. Waiting for your example.
> > 
> > Pedant. :)  That's odd.. I'll put one together early next week and look
> > into it - no other synth prog I have has the problem, so it's not my
> > system configuration.  I have a feeling this one isn't going to be fun
> > to root out. (I'll take this to the ams list)
> 
> What I wanted to point out here (whithout being pedantic :-) is the following:
> MIDI input is looked at once for each process cycle, so everything that happened
> during the previous cycle is 'dead on simultaneous' as far as AMS is concerned.
> And since AMS surely can hande more than one note-on per cycle, you could be
> right in suggesting that this one will be fun to root out...

Oh yay.  well, I guess no more discussion about this one is possible
without a test/case and/or looking at the code.