We've uncovered a possible flaw in Dunlavy's otherwise excellent design. Dunlavy uses first order (6 dB slope) crossovers to best preserve phase (minumum phase) and to generally keep things simple and apply less processing to the signal. All of this makes possible the excellent phase coherence and therefore the unusually clean square wave and impulse performance. The downside of low order crossovers is that more out of band energy is delivered to each driver. But since a multitude of drivers covering a wide range of sizes are used, each driver is only asked to cover a fairly limited frequency band. Because the frequency range each driver must handle at full power is thus limited, most of the drivers can cope with the rest of the out of band energy caused by the low order crossovers pretty well.
The exception is the tweeter. Most tweeters can handle only tiny excursions. This means they don't like getting low frequency energy. But with 6 dB crossover slopes, they're going to get quite a bit whether they like it or not. Unfortunately, they don't like it, and the result seems to be that they occasionally get overloaded by excessive energy from outside (below) their crossover's passband. What we've heard is that sibilants, violins, horns, etc. with a lot of high frequency energy seem to be overloading the tweeter and causing it to distort, leading to some harsh high frequency noise. Our current theory is that this is due to excess energy being sent to the tweeter, especially from below its full power passband. Perhaps this is causing the tweeter to run out of excursion and mechnically clip. It may be possible to tune out this problem with upstream electronics or cables, or even some gentle high frequency rolloff. Another solution would be to use a high excursion tweeter like the Scan-Speaks, or to use more high fequency drivers to distribute the load, but any of these would require a thorough redesign.
Dispersion, and its opposite beaminess, is one of the tradeoffs to consider in good sound reproduction. I propose there's a fundamental philosophical difference between tight and broad dispersion. The question boils down to wishing away the listening room or including it as part of the performance. Perhaps the answer depends on the particular music since you can fit a small jazz combo or string trio in your listening room, but you can't fit a symphony hall or rock arena. On the other hand, room-miked acoustic recordings already have the original room's ambience recorded. Should we really be adding another room, ours, to the equation? By limiting early reflections, tight dispersion can take away much of the perception of the listenening room, in a way somewhat reminiscent of headphones. Canadian research seems to suggest that wide and smooth dispersion with gentle high frequency rolloffs generally sound good. Nature usually favors smoothness in all domains. My suggestion is to use smooth and wide dispersion speakers combined with absorbtive and diffusive sound treatment to dampen early reflections and scramble later ones, with absorbers on the speaker end and diffusers on the back wall. This also matches what most recording monitoring facilities use, which ought to be a good thing to duplicate at home if we want to hear what the recording engineers heard. In untreated rooms or with some recordings, beamy, narrow dispersion speakers may work better.
Bob Stout has created a web site listing lots of DIY speaker resources called DIY Loudspeaker Driver Selection Guide (the LDSG). Section 16 of the guide is A list of recommended system kits some of the best of which are mentioned below.
North Creek is George E. Short III's operation. George's designs look sound, drivers top-notch Scan-Speaks, crossovers lavishly refined and built. Drivers and crossover components are burned in and hand selected and matched to truly obsessive levels. Fully finished cabinets for these kits are available separately from Lee Taylor and Co. Both companies supply a majority of their products to other high end speaker companies. In other words they are mostly OEMs. The top Rhythm/Revelator model looks interesting. Rhythm is in D'Appolito configuration. George's latest design called Manifest uses an Aurum Cantus G1 ribbon tweeter flanked above and below by two custom 7 inch Scan Speak woofers with a G2 ribbon at the back for ambience. Manifest is an evolution of Rhythm, joining D'Appolito woofers with these top-quality ribbon tweeters. A small-footprint, QB3 vented box is utilized, resulting in a high system sensitivity of 92 dB.
Philip E. Bamberg's BESL speakers are also well-known for their crossovers. The Series 5 MTM which uses SEAS drivers in D'Appolito configuration looks interesting, though it may not have built cabinets available yet.
Well-respected speaker scientist Joe D'Appolito is the main force behind Orca Design which designs and distributes Focal, Raven, Cabasse and other high end drivers, components and speakers in America. The Orca Array, Aria 5R and others look interesting. These are licensed as DIY projects yet are fully designed and tested. D'Appolito is best known for his research on MTM driver arrays which Dunlavy and others incorporate in their designs. Drivers in these arrays are symmetrical about a horizontal axis and feature greater sensitivity and carefully controlled vertical and horizontal dispersion. Orca kits are available from several sources. For example, Zalytron sells the Orca designs under license and reportedly builds nice cabinets for them. They also have sell other designers' kits.
A more recent design from D'Appolito called Thor uses two SEAS Magnesium cone woofers and a soft dome tweeter. It differs from other sealed or vented MTMs in using an offset, tapered transmission line to load the bass drivers. It is described, measured and favorably reviewed in the May 2002 audioXpress. Among other strong designed and measured merits, Thor has very even horizontal dispersion and should therefore be well suited to relatively untreated rooms. Thor kits are available from Solen in Canada, Zalytron and Madisound. Madisound and Zalytron offer cabinets built for Thor. Full plans are available at SEAS. This appears to be of very high quality. Thor or a top North Creek may be my next speakers.
Here is Siegfried Linkwitz's independent review of linear and nonlinear distortion components of some of the top quality drivers used in the above designs. SEAS and Scan Speak drivers used in them measure the best. Specifically the SEAS W18E001 used in Thor had the best non-linear distortion at 150 Hz and the best linear distortion at 800 Hz. The SEAS T25CF002-06 tweeter also used in Thor had the lowest non-linear distortion. Scan Speak drivers such as those used in the North Creek designs were nearly as good. These independent tests by a highly respected engineer again suggest that Thor and North Creek are worth looking into, at least in terms of their driver selections.
Hegeman had produced his speaker for a very enthusiastic audience since the late 1960s, but eventually sold his technology to Don Morrison of Toronto. Morrison has refined Hegeman's speaker through several versions has the result is reportedly very good. These may be worth checking out.
Don also makes a purist dual mono preamp based on the Analog Devices AD797 and OP177 op amps. The resulting preamp is favoribly reviewed and has very good measured performance. It may be worth checking out these op-amps.