Thanks in particular to Earl Geddes, BButterfield and FoLLgoTT, also to several people who've posted about MEH designs, though ultimately I could not find an MEH that suits.
For a decade GedLee Harper speakers were my mains, with various (sub-)woofers. This arrangement gave years of pleasure - thanks to Earl, twice over. I bought the Harpers as the Abbeys and Summas were beyond my reach, with a plan of later upgrading to larger GedLee speakers (bad plan as it turned out).
Background: I use a computer for playback, and am content with running convolution filters for crossovers, and very limited room correction at LF, etc., and am familiar with relevant theory (physicist). I also have DCX2496s. I agree with Earl Geddes about the trouble arising from trying to fix acoustic problems electronically.
I've considered options to replace the Harpers to obtain directivity control down to lower frequency, i.e., an extra octave roughly, as Summas would have yielded. I've also considerd narrower vertical directivity to avoid floor and ceiling reflections in my relatively low-ceiling room. That these qualities are not essential allowed me to be content with the Harpers. Designs and ideas that I considered include large conical MEHs; Taipuu-like large-fronted co-ax speakers; Grimm LS-1 shaped speakers, but with a "more-coaxial" layout; and various arrays. For some of these ideas I built prototypes or test baffles, including some MEH before I got the Harpers.
I rejected conventional co-ax designs due to throat problems - I tried a few experiments at correcting reflections in an old B&C drive - not bad, but really hard to get better than the Harpers (measurement or listening) and it would really need to be a 15" or a coax with an array of other emitters round it. I wonder if I ever hear the Taipuu speakers or similar will I be surprised?
My attempts at conical MEH were killed by what I later came to understand as HOMs. My conclusion was that throats must be OS, and ports are tricky. I considered learning 3D printing for a throat section, but rejected that, and I cannot find any commercial waveguide solution to modify.
I'm not sure a waveguide solution is possible, given the desire for vertical directivity - "pattern flip" is unacceptable which rules out almost everything, except a very few options e.g. XT1464, which is rather narrow horizontally ... Moving to a house with higher ceilings, and fitting modifiers is out of the question.
Progress picked-up when Horbach-Keele arrays came to my attention, and shortly after that FoLLgoTTs excellent design
Pseudo-coaxial with narrow directivity (and Horbach-Keele filters)
Fortunately, I read (enough) German and spent a while considering and simulating this approach. My new speakers benefit hugely from that. I was, however, unsatisfied with the side-woofer idea, and general box shape, having convinced myself that wide baffles should work better, and could never quite find the right solution.
I was circling around this when I first read BButterfield's thread*
Fractal Array Straight CBT with Passive XOs and no EQ and this immediately started to open up opportunities, though again the final speakers are not copies of his "prototype" described elsewhere (not linked on this site by him for some reason, so I won't do so either, but not too hard to find), even if at a quick glance they look quite similar.
I desgined and built wedge-shaped speakers: 172cm (h), 66cm (w). The wedge is from about 6 to 14cm deep, with the outer edge larger and rounded. The driver layout is more symmetrical than BButterfield's design with the "centres of mass" of the tweeter, and upper-mid and lower-mid arrays aligned (co-axial, though that's a minor tweak indeed). I also put all the voice-coils in a plane.
The tweeters are AMTPRO-4 in waveguides not unlike FoLLgoTT's, but made from wood filled with modelling "clay" (DAS) using a crude custom shaper, not CNC. The AMT has a highly damped rear enclosure.
The six upper-mids are Peerless TC7FT04-04. They sit behind holes in the baffle that forn low-pass acoustic filters. With a bit of DCX2496 correction, these filters dominate the crossover. The Peerless drives are more sensitive than FoLLgoTT's choice, and were readily available. I went for an asymmetrical layout (three not four pairs - somewhere between the two sources of inspiration). There's a complicated box enclosing the upper-mids and the AMT enclosure. This is also braced to the back (with some lossy compliance for damping).
There are six Peerless FSL-0512R01-08 distributed according to the "fractal array" principle, taking up most of the baffle height. These are conventionally flush-mounted, and the baffle thickness of 30mm puts all 13 voicecoils in a plane. I did not add the extra mid-bass drivers right at the foot of the baffle, as seen in BButterfield's solution, but then was not aiming for as low a cutoff frequency. The main volume is connected, with a bit of lossy-bracing here and there.
I already had separate 12" woofers (old Precision Devices PD12SB30) to go with the Harpers, and these are now repurposed in sufficiently-shallow boxes that sit against the wall just "outside" the main speakers (which sit tight against the wall). The mains are about 1.5m from side walls and about 3.5m apart.
In terms of min-phase filters and delays, a DCX is used to tweak the various subs and woofers and another is for the crossovers. The four (sub-)woofers basically run flat, with some small delays/phase adjustments, most of the work was in finding the right places to put them. So, ignoring detail below 100Hz - entirely room dependent, the crossovers are at 130Hz, 800Hz and 2.8kHz, inspired from the BButterfield arrangement (i.e., min-phase filters for min-phase drivers, with very little EQ). The directivity is intended to be gone by about 250Hz where the array is one wave tall tapering to up to the narrow vertical CD by no more than 750Hz.
I'm still working on minor tweaks of the crossovers while looking at lobes and reflections (my target is 10dB down for anything at all bad radiated over any important range of angles as measured at the listening zone). I am not disclosing them in detail as that would be pointless. Almost all the information that would be needed is given by BButterfield in any case.
Based on impulse measurements, the result meets my goals, at least broadly. Initial measurements suggest that floor and ceiling reflections are reduced by about 10dB at least in the upper-vocal frequency range and above. Side-wall reflections are delayed over 3m and look relatively neutral (measurements continuing). That's not as much delay as with the Harpers, but the directivity is controlled to lower frequency so the reflection is cleaner over the band. The response on/near the listening axis is as desired, without unjustified corrections - there's a min-phase notch for the Helmholtz resonators on the upper mids and some gentle correction of the AMTs above 8 kHz, that's all so far.
The speakers are driven actively in the four bands, with the drives wired in sets to achieve impedance of 3-5 ohms for each input (three parallel pairs, two parallel triples and the tweeter). Consideration of voicecoil thermal mass was important for low compression, even though these very sensitive speakers won't ever be fed much power at all, I was scared to lose the benefits of the B&C drives in the Harpers - six 91dB (1W) drives with 1.2" VCs should do well enough above 130Hz (drives working coherently): more thanks to Earl.
For completeness, there are two 12" woofers, a 12" subwoofer (commercial BK Electronics) and a 3m long 15" closed subwoofer with built-in Helmholtz resonant low-pass filter (kills high order HD). That started as a tapped horn, but I could not brace it well enough to kill a couple of bad resonances on harmonics that were audible at times.
The Harpers now take surround duties in the rear room corners, used for a few suitable recordings (20-30m delay seems reasonably convincing with old church/cathedral sounds, for example). The main amplification comes from Hypex UcD (a mix of 180W and 30W) and a couple of other cheap class D chips for the surrounds and 15" sub.
It astonished me that, after only a little refinement, the tonal impression was rather similar to the Harpers, differences are mainly in the "imaging" - that's so recording dependent that I won't even try to describe it, beyond saying that many recordings work better and a few definitely don't! As measurements suggest, there's less "room" with most recordings. Further comment would be pointless, but I agree with some of the impressions recorded by BButterfield and friends, elsewhere.
If this (rather long) post, is seen by those whom I thank, I'll be more than happy. If it inspires someone else to follow their ideas and have as much fun, and some frustration struggling to find the right design, that's a bonus.
I won't post in-room measurements (need too much interpretation), and these speakers won't be moving outside, ever. I won't post crossover details, at least until I've done very many more measurements. I'm spending too much time listening to music for that to be any time soon.
I kept a couple of alternative designs in reserve, but these don't fit with BButterfield's suggestion - to avoid reflections from the rear (aka front) wall: "baffles should be much wider than the front is from the wall behind", so unless I start a wall-building project, that's the end for a while.
Ken
*apologies for the error in this URL as originally posted. I hope now corrected.