Very interesting read, thanks Serenity. Nice takeaways here:
Four subwoofers are enough to get
the best results of any configuration
tried. Two subwoofers is very nearly
as good and has very good low
frequency support as well.
One subwoofer at each wall midpoint is the best in terms of
Std, Max-ave and Max-min but does not support low
frequencies particularly well. Two subwoofers, at opposing
wall midpoints, performs very nearly as well as four at the
midpoints and gives a much better LF factor. One
subwoofer in each corner also has good low frequency
support, but does not perform quite as well as one
subwoofer at each wall midpoint, in terms of Std, Max-ave
and Max-min. If cost and aesthetics are considered,
subwoofers at 2 wall midpoints is preferred.
Indeed, it backs up Nick's comments.
Interesting note regarding front corner placement canceling odd harmonic modes (because opposing walls are both at pressure maxima, but out of phase). But, presumably this is at the cost of reinforcing even harmonics, and scoring only a very average point-to-point standard deviation score (configuration 4). Its a delicate balance ... modes help with output and coupling of the room/driver, but also introduce coloration to the response.
It is remarkable how much lower the point to point variation is in configuration 6 (center walls, opposite), while it does just as well in the net low frequency output. In this case I would suppose that it is tending to excite the even (symmetric spatial distribution & even number multiples of the lowest resonant frequency) axial modes associated with the side-to-side room dimension (which are the same to survive in the corner placement), while exciting all modes in the direction of the sub's separation, but again with cancellation of the odd harmonics. Not obvious why this is better. Hmm ...
Also remarkable is how steady these results are with regards to changes in the room dimension. I guess this makes sense ... the specific modal frequencies will scale (inversely) with room dimensions, but the nodal locations will keep their same relative spatial distribution. Not to say room dimension is irrelevant, but rather just that this metric doesn't capture it ... the even distribution of modes in frequency space may be much better/worse as certain dimensions become more independent or redundant, respectively.
Last thought ... one of the explicit assumptions of this test was that the transducers were in phase. I am curious if it might be possible to essentially invert the conclusions by tinkering with phase, giving better flexibility in placement.
