< Back to IRCAM Forum

KNN vs DBAP vs...?


For a live improvisation setup with Max, I have imagined a 2D sepaker setup combining a circle of 8 with a stereo pair in the centre (to be expanded later as a square). The idea is to put the musicians in the centre and the audience around, in between the central pair and the circle.

8 2 avec public et musiciens

The loudspeakers of the central pair are homemade models spreading sound over 360° angle on the xy plane.

The sources played will be mostly stereo.

The setup is irregular. Also, the size of the circle might change. Therefore I plan to use the DBAP algorithm. However could other ones be (more) relevant? What about KNN? Other ones?

1 Like


DBAP and KNN are, in essence, very similar : they use the source(s)/speaker(s) distances to compute gains. However KNN is more flexible (it has more options for you to control), so I would recommend to use it instead of DBAP.
I guess that should work for your setup.

Maybe you could complement with a VBAP2D for the ring of 8 speakers ? for some of the sounds ?


1 Like

Indeed, I’ve just tried the setup in studio and the exponent parameter from KNN is quite helpful, especially as the loudspeakers are quite close to each other.