[LAU] ambisonics and number of speakers

[Jörn Nettingsmeier]

disclaimer: this user is getting seriously confused.

On 02/28/2010 01:56 PM, fons at kokkinizita.net wrote:

> To reproduce the field of a real source, you need *all*
> spatial harmonics, in theory (for a real point source)
> up to infinite order. So An AMB system of order N must
> not just recreate the harmonics up to order N, but *all*
> of them. It will control some of them (but no more than
> the number of speakers), and the rest has to be generated
> by aliasing.

i can see how a soundfield mike would capture (and thus include in the 
first-order signal)those aliased components, but then what about a 
panner? following your argument, shouldn't we add higher-order aliased 
components into a panner matrix?

> A first order horizontal decoder with 4 speakers will
> exactly control the levels of the 0, 1st and 2nd order
> components, and the rest is filled in by spatial aliasing.
> The aliased components are not exact of course, and that's
> why localisation will be impaired if you move away from
> the sweet spot.
>
> Now if you use e.g. 8 speakers to reproduce horizontal first
> order the system controls components up to 4th order. If the
> input is just first order, the 2nd and 3rd order components
> will be zero, as will be all those that alias from them (5,6,
> 10,11,13,14, etc.) The result is a field that does not match
> well to that of a real source.

hmm. i still don't see from this explanation why 1st order horizontal 
sounds worse over 8 speakers (or 12, for the sake of argument) than over 
just six, even if i'm in the sweet spot.

best,

jörn

(who is about to crack open a can of fizzy InstaMath(tm) extra-strong)