Re: [Linuxsampler-devel] Gigasampler and Key Groups

[Christian S. <sch...@so...>]

Es geschah am Sonntag 12 September 2004 00:59 als Sebastien Metrot schrieb:
> What you call keygroups is usualy refered to as exclusive groups. What

Yeah, but AFAIK in the Gigasampler world it's called keygroups. Probably 
because the gig Format only allows a key to have one region, unlike e.g. the 
DLS format which allows several regions for the same key.

> you have to do is to simply kill the currently playing voice of an
> exclusive group when the user starts playing a new voice on the same
> group. (Of course you have to take care of the potential click and fade
> out the rendering). You can picture this as a monophonic sub layer
> inside a layer...

Ok, already guessed that, just wanted to be sure. Thanks!
I commited it to CVS minutes ago.

Es geschah am Sonntag 12 September 2004 04:09 als Vladimir Senkov schrieb:
> I think we have already discussed this topic of "simply killing" the
> voice back when we were talking about voice stealing.
> Is it just a quick release and if so where do we get timing parameters
> for it? Any ideas?

I simply use a very quick exponential fade down (see 
EGADSR::CalculateEndCoeff()) and every state can now immediately transit to 
the 'end_stage' which actually does the quick fade down (btw, we have to kill 
denormals there, I forgot to do that). Here the latest state diagram:

 http://www.linuxsampler.org/doc/engines/gig/EGADSR.pdf

> Can the same mechanism be used for voice stealing?

You can use the new Voice::Kill() method to kill a voice (e.g. for voice 
stealing). It takes care of the fade down (to avoid clicks) and processing 
until the kill event actually occured.

CU
Christian