[LAU] ambisonics and number of speakers

On 02/25/2010 11:01 PM, Arnold Krille wrote: it depends on the order of your source signal. when i got the chance to play with a second order mix in the sala bianca in parma (a four-wall wavefield synthesis rig) with fons, our first test was to create six virtual speakers in a hexagon. sounded very nice, even better than discrete speakers because you can place them further away to better resemble a plane wave. funny thing was when we went to eight speakers, the localisation deteriorated, very slightly but noticeably. generally, it seems to be advisable not to use too many speakers for any given order. 8 is great for 3rd order horizontal, ok for 2nd and probably too many for first. there is a sound scientific explanation for this which i have forgotten... let me try (and i hope fons will jump in if i'm getting stuff badly wrong): i believe it has to do with the blurring of the energy vector at HF - ideally, you want just one speaker to play a discrete source (which means rE equals one, for a source coming from where the speaker is). the higher the number of speakers that contribute (and if your signal is first order, all speakers will, to some extent), the lower the energy vector, which means you only get good directional cues in the LF band. additionally, HF phasing effects become more pronounced since a number of more-or-less coherent sources interfere. you can work around that by "de-tuning" the delay compensation, but then you trade localisation precision for phasiness. these issues also explain why a 3d rig gives worse 2d reproduction than a horizontal-only rig. it's all not as bad as it sounds, but your investment might not pay off the way you're hoping. best, jörn