In MSHFSJ, the mixer is the component analogous to a mixer you would find in a conventional digital audio workstation. The task of the mixer is to sum all the tracks together into one signal and handle any global effects processing.
The mixer layer thinly abstracts away the computation portion for tracks, summing them all to a stero signal. The reason for this abstraction is for global effects such as reverb or delay using what are known as sends.
Each track is able to contribute some amount to the global effects. This is known as send, and it is represented internally as a multidimensional array of SPFLOATS. The following function is used to control how much a track contributes to a particular send channel:
void whisper_mixer_send(int track, int send, SPFLOAT amt);
Where track is a zero-index track number, send is a zero index send number, and amt is the amount in dB to send. A value of 0 is considered "fullscale", amplification is above 0, and attenuation is anything below 0. If a value goes below -100dB, it is automatically clipped to be 0, effectively muting the effect send for that track.
The mixer has a variety of global effects. An effect in whisper has a special kind of interface for computation. There are ultimately 3 kinds of callbacks needed for computation:
-
initialization: this function allocates and initializes memory for the effect. The sole parameter needed here is sampling rate. Effects using soundpipe can either use their own soundpipe data structure or use the global soundpipe data struct via
whisper_sp_data() -
computation: an interface for computing a single frame of audio. The hard coded assumption here is that inputs are mono sources, and outputs are stereo. This also passes the void pointer containing any allocated data.
-
destruction: free any memory previously allocated.
In addition to these core callbacks, an additional function will be needed for binding user data.