Re: [Ccrtp-devel] ccRTP audio mixer

[Michel de Boer]

Summing the samples of two or more audio streams that are linear encoded 
indeed
mixes the audio. The problem is however that you have to scale back the 
result
of the summation somehow. If for example you add 2 linear 16 bits samples
then you get a 17 bits value. You have to scale this back to 16 bits again.
With mixing only 2 samples you might get away with just dividing the result
by 2 but you will here the volume go down (by 3 dB I believe). If you want
to do it nice you have to use somethling like an S-curve to scale the value
back again (like is used for G.711 compression).

Another thing you might want to take care of when you mix multiple stream
is that just summing sound samples will increase the level of noise that's
in the original streams.

Regards,
Michel

Federico Montesino Pouzols wrote:

> Hi,
>
> On Wed, Mar 02, 2005 at 06:18:51PM +0100, address@hidden wrote:
>  
>
> > Thx in advance for your answeres.
> >
> > I tried to figure out what is meant with "sum the samples". (As I
> > mentioned within another posting I have only basic knowledge of RTP)
> >
> > So if I have linear samples is it possible to mix the packets with the
> > ccRTP class outgoingDataQueue (putData) ?
> >
> >    
>
> Yes, in principle, it should be simple. As far as the audio codecs are
> simple or you have a good library for codecs processing, a few lines
> of code will make the job.
>
>
> _______________________________________________
> Ccrtp-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/ccrtp-devel
>
>
>