Originally Posted by blazar
Question regarding a new house theater design with MSO in mind:
We are building a room where the ceiling increases in height as you get to the back of the room. The floor is otherwise rectangular.
Since this would be a room where you really cant model the FR for speaker placement, this is my strategy in the design:
4 subs behind the AT screen. one pair is 1/4 length from the side walls, and 1/4 length from the floor. The second pair is 1/4 length from the side walls and 1/4 length from the ceiling.
I am thinking we will need another subwoofer or two that we can move (in-ceiling perhaps?). we want all subs to be hidden.
The placement of the 1-2 subs to maximally take advantage of MSO optimization is what I am trying to figure out. I figure more unique positions (center of the room in the ceiling?) will give you unique permutations to correct for deficiencies in other speakers.
Does this seem like a valid approach given the difficulty of modeling non-rectangular rooms?
This is not an easy question to answer. It's worth considering some of the multi-sub techniques that are used with conventional EQ (all subs EQ'ed the same), and asking the question of whether such techniques are also best when using mode manipulation software such as MSO.
One such technique in its simplest form involves two subs, putting one, say, in the lower-left corner in front, and the other in the lower-right corner in front. Since the odd-order width modes have opposite polarities in these two corners, driving the subs in phase and with equal amplitudes excites these modes, but in opposite polarity, giving an effect that some, including Welti, call mode cancellation
. More generally, if you were to drive these subs with signals whose amplitudes were not necessarily equal, and not necessarily in phase (for whatever reason), the result might be called (and is called by Toole) mode manipulation
. This happens with systems like MSO and SFM.
Another technique is to place the subs in locations corresponding to, say, 1/4 and 3/4 of the length, width or height dimensions in a rectangular room. This places them in the nulls of the even-order axial modes. I call this technique mode avoidance
, because it achieves its ends by avoiding the excitation of these modes altogether (in the lossless case). The ultimate mode avoidance system is the Double Bass Array, AKA "DBA". But forgetting about DBA for a moment, using more mundane mode avoidance techniques gives mode manipulation software like MSO fewer ways to fix the problems it may encounter. This is because you can't manipulate modes that you don't excite to begin with
. So on the one hand, you've fixed some problems by avoiding the excitation of these modes, but on the other, you've degraded the ability of MSO to manipulate all modes. So what's the end result of this? It's hard to say. In order to attack such a problem, you'd need to combine MSO somehow with room simulation software, moving sources around, doing optimizations, and comparing results. I've never done this and probably never will. Having a non-rectangular room adds yet further complexity.
Interestingly, there's a paper by Sarris, Stefanakis and Cambourakis
that deals with what they call "multichannel sound equalisation", which is just another phrase for individual EQ of multiple sources to achieve the desired responses at multiple listening positions. They assumed a rectangular room, and evaluated multiple combinations of source locations to achieve the most robust solution. They concluded that having eight sources, all of them in the corners in both floor and ceiling, were the best. See figure 3 of that paper for the configurations they simulated. Since corner locations excite all modes, this seems to suggest that for mode manipulation software, the ability to excite all modes may give better results than using mode avoidance techniques, even though such mode avoidance techniques may be best in the case for which all subs have the same EQ applied.