I am currently in the process of building an
eight-channel ambisonics setup in
my home-office. Size is about 3x3x2.5 meters.
To be clear---
You do mean a three channel first-order setup with eight speakers, correct?
To answer questions a few others asked, It's not that a bunch of
channels are needed, rather that more speakers are needed to fill gaps
in the wavefront imaging. quad/5.1/7.1/etc takes an ad-hoc approach
to this by adding more and more fully discrete channels to 'plug
holes' while Ambisonics takes a methodical approach that simply adds
more speakers to the already constructed/encoded model.
("Why didn't Ambisonics win then?" you ask... well, it requires signal
processing that was hideously expensive at the time of its
introduction, and the 'add another full channel for each speaker
approach' was far cheaper and more practical at the time. Today, the
average cereal box contains more computing power than used to land on
the Moon, so I think the Ambisonics approach is suddenly the
easier/cheaper way to do things. Excepting of course that the discrete
channel method has a huge installed base. For that reason, Ambisonics
is still 'weird' and 'fringe',)
Cheers!
Monty