Re: [AD] Sound API

[Chris <kcat@xxxxxxxxxx>]

Korval wrote:
> Is there no reasonable heuristic to deal with the clipping issue? Like, if
> you knew what the maximum range of a group of 5 samples was, you would know
> what volume would guarentee no clipping? Or would guarentee it some
> percentage of the time?
>
> Is it possible to do some analysis of the samples to deal with this issue
> automatically?

You *could* analyze the post-mix buffer and decrease the volume it so 
the maximum sample is within the 24-bit spectrum (before scaling), and 
save the divisor.. then for everytime after that, decrease the divisor a 
little and check that it doesn't need to be increased (so the volume 
gradually rises and isn't so jumpy). That seems a bit evil to do in a 
real interrupt though (which is what it is in DOS), or a signal 
interrupt (which it could be under Linux without pthreads). Or really 
even a timer in general since there's only one thread for all the 
installed timers. A timer is supposed to be as fast as possible, don't 
forget.

- Kitty Cat