This is kind of a branch from a thread in the Audio by Van Alstine circle.
The original thread is here:
http://www.audiocircle.com/index.php?topic=23787I was explaining how Ohm Walsh radiators, while being omnidirectional themselves, are not used in that way in current Ohm speakers. (They place damping inside the "can" in a large arc around the driver.) They actually resemble Constant Directivity speakers in some ways.
The claim for CD and the "even power response" camp like Linkwitz is that you're going to have sound directed out into the room and then reflected, so you want an even frequency response going in all directions so the reflections have the same tonal character as the direct sound. (That seems to somehow assume that all frequencies will be evenly reflected by the room.) Still, there's a certain amount of sense to it.
Unless you're listening in an anechoic chamber, your room is going to reflect some of the sound. Your job is to then reduce this as much as possible, and especially to avoid early reflections.
I suspect that reflections from the wall behind the speakers, from the side walls, and from the floor or ceiling all have different deleterious effects on what you hear from the speakers. I don't know if anyone has studied this or not.
I feel that speakers should be as close to point sources as possible. By definition that's going to give a wide radiation pattern.
If you want very directional speakers, you're looking at line arrays or big panel speakers. (Perhaps an ESL with a large box filled with damping behind it.) I feel this is flawed because you get high frequencies radiated from widely spaced sources, and this will screw up transients and imaging.
(Line arrays do have their uses though. They work rather well for very large rooms like a church. Perhaps not for music, but they're great for spoken voice.)
Unless you toe-in your speakers, you're normally listening off-axis anyway. I see most speakers not being toed-in. So you _need_ some off-axis response. Radiation, unless controlled, is symmetrical, so the same thing will be pointed at the side walls and floor and ceiling as is pointed at your ears.
So, in most cases, you're going to get a wide radiation pattern anyway, and I'm not sure that the Ohm speakers are any worse in that respect.
Some designers _design for_ frequency response or power response (flat or whatever they want to achieve). I feel it's better to design for good transient response at the listening position, and then the frequency response at the listening position will be acceptable too.
You could design a speaker system consisting of a bunch of high Q resonators. (Perhaps with organ pipes...) When fed with white noise, the overall response might well be flat, but there's no way a transient will survive a trip through those speakers.
We're going for kind of a minimalist approach. Concentrate on direct radiation from the speakers and hope we can fix the room reflections. That gets us to where most people are or are hoping to be.
The next step is to eliminate crosstalk between the speakers. (Hafler and Carver worked on this, and there was also the M.A.R.S system, but none of them worked very well. We can now do it digitally, and I've played with that.)
Finally, the microphones weren't completely directional either. They "heard" sound coming from various directions. (Indeed, a pair of omni mics are often used for stereo.) Perhaps we can recover that information and replay it from different points in the room. There's a lot of work being done on this type of thing, but it has unfortunately mostly been used to provide sound effects rather than reality.
It gets even trickier with the various stereo mic techniques, and likely impossible for multi-mono recordings. With multi-mono (most studio recordings) and spot mics in a concert hall, the best you can hope for is that the amplitude cues will be enough to provide some sort of imaging.
The other cues we use to localize sounds are time and frequency response. Time = phase, and FR varies due to our head ears (the external part). The whole thing is called HRTF and there's some good material here:
http://interface.cipic.ucdavis.edu/CIL_tutorial/3D_HRTF/3D_HRTF.htmThe time/phase aspect tells us that we need to make sure our speakers can properly present the waveform at our listening position, maintaining the time/phase relationships.. That probably lets out line arrays, long ribbons, panel speakers, and speakers with high-order, uncorrected x-overs.
First-order crossovers can solve much of the problem, but they create a new one. There will be significant radiation from both drivers in a region surrounding the crossover point. That gives us multiple sources for the same sound and will likely affect our perception of the location of a sound.
We may also find increased IM distortion.
There isn't firm agreement on the audibility of many of these things. (My, is that an understatement...)
I suspect that there won't be too much of a problem with a closely spaced pair of drivers. It becomes a bit more of a problem with the 2 M's in an MTM arrangement. It gets worse with line arrays and panels. I've noticed that plain large ESLs don't appear to have any distinct treble. There's something going on up there, but it's pretty muddled and doesn't convey useful information.
Single full-range drivers eliminate many problems, but they tend to display lots of distortion at the low end and at the high end, plus a dollop of IM all the way around.
I feel that things like DEQX are the direction of the future. It eliminates the drawbacks of standard crossovers, eliminates the problems of passive (post power amp) crossovers, and allows for the use of things like metal diaphragm drivers within their linear area of operation. (Drivers with metal diaphragms are often very well-behaved within a limited passband, but they're extremely poorly-behaved outside that region.)
Yes, that leads towards the NHT Xds.
I have yet to hear them, but the theory sounds good.
That's enough for now I think. Cuss 'n discuss.
