The design of these speakers takes the form of a woofer with a horn loaded tweeter. The tweeter is generally a compression driver, but does not have to be. There are a few different types of horns, but the one we are interested in is the constant directivity horn. This type of horn is actually often called a waveguide and is just a larger version of those found on forward firing cone / dome speakers. Geddes has written a nice paper that explains what a waveguide is and how it differs in theory from a horn.
The larger the waveguide the lower in frequency the directional control extends to. These speakers are often two ways with a relatively large woofer. The woofer is run up much higher than it would be in a cone/dome speaker and is purposely used to the point at which its response beams. The narrowed response is then matched to the controlled directivity provided by the waveguided compression driver. The theoretical directivity design model for this kind of speaker is wide directivity at low frequencies that narrows as the woofer starts to beam and then stays constant throughout the rest of the frequency range.Pro Audio Technology (formerly Professional Home Cinema), Procella (the brand we love and recommend) and the JTR Noesis line. If you want to DIY check out DIYSoundGroup, who provide flat packs and SEOS CD horns. Seaton speakers do use compression drivers but they don't use true waveguides (they fit into the coaxial driver category). story behind the M2 that is worth reading. CD waveguided speakers, like the JBL M2 and Procella P8, have some important implications for room acoustics. The off axis sound, at least in the range the waveguide is working, has similar spectral content but it lower in level than the direct sound. Reflections from major boundaries in the room are therefore lower in level than they would be with a wide directivity speaker such as a cone/dome speaker and have fewer perceptual effects. Below the waveguide transition frequency the spectral content of the reflected sound will have a different composition to the direct sound, which may cause unwanted effects. However Floyd Toole's research shows that image shift, source broadening and timbral coloration from reflections are primarily related to frequencies above 1kHz so maybe the CD design is solid if large enough to provide control down to a low enough frequency.
We would expect the frequency response at the listening position to exhibit a slope from bass frequencies to the point at which the waveguide takes over at which point it should flatten out. This kind of frequency response will look very different to that from a cone / dome speaker. Automated room correction routines that attempt to correct to an arbitrary target curve do not take differing speaker directivities into account, and this is a good reason not to use them unless you can manually draw the target curve. If you find one that lets you set the target curve, such as Dirac Live, a good approach is to measure the uncorrected speakers, use a high level of smoothing such as one octave to show the shape of the frequency response at the listening position, and then fit the room correction target curve to this.
Nyal Mellor, Founder, Acoustic Frontiers