[Stk] voicer->tick() returns clipped audio

[Perry Cook]

That wasn’t in voicer’s original design, so it’s up to you
to manage gains.  Turns out it’s not necessary on Macs
(this is built into the audio subsystem in OS X), but Windoze 
and Linux it is necessary.

Actually what you’re doing is clipping (forcing square tops
onto anything over 1.0).  The audio subsystem is probably
wrapping (making large negative numbers out of numbers
greater than 32767, and visa versa), which totally sounds 
horrible.  Clipping sounds nicer, but something softer than
that is even better.

For this reason, we added Dyno (dynamic range processor
object) to ChucK.  You might consider going to the ChucK
source and grabbing that object and adding it to your STK
build.  (Gary, we should do that as well…).

PRC

> On Aug 22, 2015, at 11:07 AM, Patrick J. Collins <patrick at collinatorstudios.com> wrote:
> 
> Another problem I have found is that the voicer is not doing anything to
> stop the mixing of multiple instruments' audio from clipping.
> 
> For example, if I output a whole bunch of random noteOn events, I see
> voicer->tick() will actually return values greater than 1.0f or less
> than -1.0f...
> 
> In order to prevent clipping in my app, I am having to do:
> 
>        float sample = voicer->tick();
>        if (sample > 0.99f) {
>            sample = 0.99f;
>        } else if (sample < -0.99f) {
>            sample = -0.99f;
>        }
>        buffer[i] = sample * (1 << 15);
> 
> I would think the voicer should be taking care of this internally and
> never return anything that would potentially be distored.  Is there any
> reason for this?
> 
> Patrick J. Collins
> http://collinatorstudios.com
> 
> _______________________________________________
> Stk mailing list
> Stk at ccrma.stanford.edu
> https://cm-mail.stanford.edu/mailman/listinfo/stk
>