The first speaker I heard that really "disappeared" was the narrow-baffle Meridian M2, a compact MTM that predated Joseph D'Appolito's landmark article in SpeakerBuilder. The midwoofers were KEF 5.25" drivers and the tweeter was a 2" (!) KEF dome, and the speaker was active (internal electronic crossover and amplifiers). I have yet to hear another MTM speaker throw a soundstage like that little beastie.
In a more conventional driver arrangement, in my experience Green Mountain Audio is king of the imaging department. Roy Johnson does a whole bunch of things extremely well. If I hadn't started up my own speaker line, I'd be a dealer for him today.
The sadly discontinued Snell Type A was also amazing in the imaging department, and was an all-around superb loudspeaker in my opinion.
Single-drivers speakers usually image very well. Omega, Cain & Cain, Lowther, and such. The best imaging I have ever heard was from an 8" Supravox field-coil driver, which is a full-range unit with no wizzer cone. Small sweet spot, unfortunately.
Many coaxial units also image very well, and in my opinion coaxials also offer timbral advantages over more conventional two-way configurations.
As mentioned by others, electrostats also tend to image very well. The concentric-ringed Quad 63/989/2809 is probably the imaging standout in this fine crowd.
Horn speakers tend to not image as well as other types; relatively few disappear as the apparent sound source. I'm a dealer for some that do, so I won't mention their name.
I'm sure I've left out many worthy contenders - frankly I don't place a very high priority on imaging. It's nice if it's there, but in my opinion imaging is secondary to both timbre and liveliness.
Many have noticed that narrow speakers generally image better than wide speakers. This has to do with how the ear/brain system localizes sound. Localization cues come primarily from the first .68 milliseconds of a sound impulse, which is the time it takes for a sound wave to travel about 8.5 inches (roughly the distance around the head from one ear to the other). A small time delay from the tweeter to a diffractive or reflective surface (say .2 milliseconds) is interpreted by the ear/brain system as a small angular image discrepancy. A time delay close to (but less than) .68 milliseconds is interpreted by the ear/brain system as a large angular image discrepancy. Applying this to image width, narrow speakers image better than wide ones unless the wide ones are wide enough to push the edge diffraction beyond .68 milliseconds, or unless the wide ones have large radius baffle edges so that there is no edge diffraction.
Looking back on the Meridian M2, the cabinet edges were closer to the tweeter's edge than with most modern designs because the midwoofs were very small (allowing a very narrow baffle) and the tweeter's dome was very large (not only having a narrower radiation pattern in the crossover region but also placing the edge of the tweeer closer to the cabinet edge). So it was a very good physical configuration from an imaging standpoint. The Green Mountain Audio speakers are even better - take a look at how minimialist the baffles are around their tweets and mids.
Duke
