Okay, fellas, what are the basic speaker designs that promote good imaging?

I'm sure one is tweeter, mid and woofer on a vertical plane.

But how about mirroring them? Or slight offset of drivers? Or are they better "centered"?

Seems the DCMs with coaxial mounted speakers image well. Why?

And even more puzzling, why do the Sonotube DCMs with drivers facing off at funny angles do so well?

Now, what about cab width? Should the early BAs, like the A100s, A150s, A200s and A400s, wide as thet are, have issues with dispersion?

It´s a widely accepted fact that narrow speaker boxes will image better than wide ones, due to less diffraction occurrences.

Full-Range and Dual-Concentrics also fare better in the imaging department when compared to conventional multi-unit speakers.

Jorge

Ooh, difficult question. The answers are beyond me, so I hope you don't mind my explanations.

Jorge has nailed one of the most common ways to improve imaging, narrow baffles and, better yet, those that have angled panels or large curvature rounded edges to the side. This approach dominates current high end speaker offerings and is why mini-monitors are the audiophile's preferred choice with large speakers that are just a few inches across and several feet deep. The angled panels and large curvature rounded edges help defeat an interesting diffraction effect where baffle edges act like mini-line source radiators along sharp, squared edges.

Beyond that, having the drivers relatively aligned in the vertical plane (most all have the drivers slightly offset to avoid response bumps again due to the baffle) works very well due to stereo information being presented in the horizontal plane. Basically, the fewer radians occupied by the sound source means the finer resolution imaging may take place, all else equal. This is mostly why coaxials image so very well even when placed in large baffle enclosures, as they simulate a point source (sound emanating from an infinitely small point).

When a driver is set into a baffle, frequencies whose wavelengths coincide in some ratio to the distance between driver and edge tend to by either boosted slightly or cut. When the driver is set equidistant from either edge, these aberrations in the response coincide and act additively making otherwise subtle-enough-to-overlook effects very much audible. So, designers tend to set the driver slightly off center so the distance to each edge is a bit different. This spreads these bumps and dips out thus making for a flatter and more accurate response. It also helps with similar resonances in the baffle board itself if the designer gets cheap on materials. (Felt on the baffle also help reduce baffle diffraction at higher frequencies again helping centralize the sound source and improve imaging.)

Many manufacturers in the past and still today usually ignore this offset (sometimes great with the tweeter) and make all the speakers similar to keep costs down. The problem exposes itself when it comes to imaging. Non-mirror imaged speakers (the drivers are offset the same in both speakers, to the right or left) the radians the sound source occupies are the same, but the positioning of the drivers to the listener varies as the frequencies shift through the drivers. The result is a smearing of the image by that difference of position and the center image itself can seemingly move back and forth by the same amount as the music shifts around in the drivers. The effect is a singer's voice coming from a point that may be, say, 2" wide at some point, then 12" wide a millisecond later and that may move back and forth the same amount depending upon the note.

Mirror-imaging greatly helps in this case by keeping the driver positioning consistent with both channels. The result is a stabler image that doesn't shift with the frequencies and is also a fair bit sharper.

So, with conventional drivers, a slight offset is desirable and mirror-imaging all the more so.

The design approach with such models is to broaden the dispersion to widen the sweetspot, yet with less of an impact on imaging than one would suspect. It's an attempt to emulate an omnidirectional speaker to some degree and regardless, omnidirectionals can do a pretty good job of imaging with more depth than a more conventional design. I believe this is a derivative hearkening back to the old school pulsating sphere/column approach. That idea was that the ideal sound source was, as the name implies, a sphere that expanded and contracted linearly to reproduce the music. (For stereo, a pulsating column would work just the same, which is why line arrays work and image so well.) Though the DCMs aren't truly representative of this approach, their funky driver positioning recreates a small segment of the sphere and will image well enough provided the drivers don't fight one another (controlled dispersion helps reduce comb filtering that would hinder the wave's uniformity). It's the same reason why my dodecas image pretty well.

Other approaches to improving imaging tend to be more directional treble drivers, such as planars, ribbons, and horns. By directing most of the treble energy at the listener rather than seeking broad dispersion, reflections and interactions between the channels may be reduced thus improving imaging all the more. This "beaming" when utilized thus is sometimes called the "headphone effect" as it reduces room interactions at these higher frequencies making the imaging similar to that of a pair of headphones, but at the cost of a narrow sweet spot.

To briefly touch on the pulsating column of line sources again and combine it with beaming and dipole radiation, you'll find a reason for why planar designs can image incredibly well. Even 8' tall and 4' wide Sound Labs or 2' wide Maggies can out image a mini-monitor. The dipole backwave also helps add to the depth of sound and the boxless design an extra openness that makes the resultant soundstage realistically three-dimensional.

There are also a couple of "cheater" ways to improve imaging. The most common is to shelve the tweeters so they run a few dB hotter than the rest of the drivers. As a significant portion of imaging information is found in the treble, this can seemingly improve the effect, but it's far from accurate and makes for more fatigue. The other way is to flip the phase on an upper midrange or treble driver. The interactions can make the imaging seem incredibly tight, but at the expense of many other attributes of good sound. The Wilson WATT Puppy 7 is like this, with inverted phase midranges that makes the singer's voice come out of a point the size of a quarter though the speakers are 12' apart. It also makes the sweet spot the size of a narrow head and moving slightly out of it where the phase effects cancel each other out, the sound stage collapses to either speaker.

Hope that helps some. 'Bout all I can say on it.

- JP

Crap, that's exactly what I was going to say.

I knew I had read it before somewhere.

- JP