Subject: Re: [linux-audio-user] virtual sampler
From: Juan Linietsky (coding_AT_reduz.com.ar)
Date: Thu Mar 21 2002 - 19:05:09 EET
On Thu, 21 Mar 2002 11:52:07 +0100
Frank Barknecht <barknech_AT_ph-cip.uni-koeln.de> wrote:
> Juan Linietsky hat gesagt: // Juan Linietsky wrote:
> > On Wed, 20 Mar 2002 16:11:20 +0000
> > matt yee-king <yeeking_AT_no-future.com> wrote:
> > > Hello
> > >
> > > wondering if anyone knows of a midi controllable virtual sampler?
> > >
> > > cheers
> > >
> > > matthew
> > >
> > Hello! i was wondering the same a while ago. After searching around
> > and asking dave, i found that the only "usable" ones were timidity (which
> > works really nice but it's very limited) and iiwusynth (which sounds
> > really good and uses soundfont, so you can combine it with smurf). The
> > problem with iiwusynth is that it's written enterily using floats, which
> > makes it extremely slow (I tried compiling with pgcc and all possible
> > optimizations and it still cant pull out more than a few channels on my
> > duron 800). Still, the accuracy is great, if you dont mind such enormous
> > cpu usage.
> I cannot find the behaviour you mention in iiwusynth here, and I only have
> an Athlon 900 which is not -that- much faster than your Duron 800. In
> fact, I helped bringing the newer iiwu to PD (like I did with your dx7 ;).
> I have successfully used both the iiwu~ and the rx7~ PD-objects based on
> your rx/saturno in a single patch (iiwu mostly for drums, rx7~ for the melodic
> parts) without seeing one being a big CPU hog. I ran the midisequencer MusE
> in parallel and some other PD objects in that patch as well. And PD itself
> is highly float based, as all number computations use floats, and the dsp
> objects generate signals normally scaled to the range -1 to +1
Ah, let me clarify on the "cpu hog" usage. When I make music, i tend to
mix several voices together, like a drum track, bass, some low pad, strings or normal pads, arrangements such as trumpets, brasses, resonators, bellpads, waves, etc, maybe lead instruments, or a piano,
organ,rhodes,etc. This needs a huge amount of simultaneous channel playback, around 50,60 or so if i'm using proper falloff in the envelopes (not abrupt). Trackers have been supporting this (since impulse tracker), for many years now, but usually people that writes some kind of synth engine (wavetable or oscilator based) never seems to go for speed (poliphony). The result is that you end up with a limited number of channels that you can mix and getting used to that.
I'm also used to a tracker, where the limited concept of ADSR envelopes does not exist, I love to do all kind of effects with the envelopes (which are custom in a tracker, you basically draw them and select loop or sustain points). Also the wonderful thing that you can set sample loop points in bidirectional way (something trackers got from the gravis cards), which is extremely useful. Also remember that you can make those nice envelopes for anything you want, like volume, pan, pitch, cutoff, or Q. This is why I want to make a wavetable engine based on cheesetracker.
This archive was generated by hypermail 2b28 : Thu Mar 21 2002 - 18:51:30 EET