I have 4 NS15s I would like to use for dipole midbass. I have been following StigErik's thread as well as a few others and would like to achieve something similar.

However, I'm not sure how high in frequency they should be taken. The primary cone resonance is ~2.2KHz, which corresponds with the first impedance aberration. Frequency response is flat to 500Hz and is down 6dB by 1KHz. Distortion decreases with frequency up until about 500Hz (-55dB), but is below 1% until ~8KHz.

http://www.aurasound.com/public/pdf/NS15-992-4A.pdf
My intuition is suggesting a ~2nd order FIR filter at ~150Hz.

How would you "suspend" these woofers? I was thinking of using a vertical arrangement with the lower woofer inverted. These really are beautiful woofers, so it should be relatively aesthetically pleasing. Could I mount the woofers to a minimalist frame and suspend this? I have a feeling their frames should be coupled in order to maximize harmonic distortion cancellation.

I'd like to incorporate them into a linear phase multi-way system. A Duelund XO with a=2 could be nice. If the interdriver spacing is minized, a=sqrt(3) could be very interesting.

I've been trying to understand the minimum driver layout constraints in order to be able to assume quasi constant directivity. If we force the lower range driver to be -20dB as its interdriver spacing approaches distance/wavelength=.5, we might get some interesting results. Any thoughts?

Honestly though, a 3-way simply will not be able to work without significant sacrifices if constant directivity dipole behavior is desired. A 5-way could be very interesting, but a 4-way would probably be much easier (monopole subwoofer <50Hz would make it a 4+1 or 5+1 way). Has anybody used a higher order Duelund?

Also, I'd like to avoid symmetrical driver layouts (ie WMTMW, MTM, etc). Waves expand as spheres. If a linesource is used, the wavefront will distort as it transforms into a spherical wavefront in the nearfield. If we assume the listener is in the far field, then by going with a symmetrical layout we have reduced the amount of possible surface area within the plane where the waves sum constructively.

Between ~800Hz and ~1.6KHz, human hearing transitions from ITD to ILD. As a result, we might compromise (ie minor lobes) the response above 4KHz in order to allow for some design flexibility below 1.6KHz by extending the response of an upper midrange driver into the lower treble. John K appears to have achieved success in using this technique in his Nao Note loudspeaker.

Phase is not that significant above 4KHz, so a BG Neo3 might be added as a supertweeter. The SS 10F will afford the opportunity for 1 wavelength driver spacing at ~3.8-4KHz. If we machine the drivers frame off and mount it by the magnet, we might be able to push this up to ~4.2-4.5KHz.

I have access to a machine shop, so driver modifications and metal fabrication are possible. A custom aluminum dipole waveguide would be within reason. I'm not sure if a torus (controlled response) would be superior to no baffle (closer spacing)? I assume there is some ratio between the ID/OD of the torus and the driver diameter which minimizes the dipole peak, of course the depth of the waveguide should be <1/4WL at the highest frequency.

Push/Pull NS15 --> W22EX --> W15CH --> SS 10F ------> BG Neo3

A waveguide for the SS 10F could be interesting since it will assume a relatively wide bandwidth. John K seems to think so. However, I'm not sure how to optimize the waveguide. The velocity distribution at the throat must be known in order to determine and optimize the far field response. Any thoughts on how to estimate or measure this? I'd assume it will be non-trivial considering the modal contributions of the cone have to be considered and the fiberglass cone isn't exactly homogeneous.

Push/Pull NS15 --> TD15M --> W22EX? --> SS10F w/ WG ------> BG Neo3 ???

I have a pair of TD15M Apollo lower midrange drivers, but they don't quite fit into this constant directivity dipole design if the NS15s are used. However, I'd really like to incorporate it (linear phase must be satisfied) into the system if its possible. It really needs to be highpassed at 150Hz, since I'd like to limit RMS excursion to <1mm, but it can't be used above 500Hz due to the dipole peak. Perhaps some insight can be gained by simulating a higher order Duelund.

I'm quite fascinated with the Duelund crossover topology, but the math is a bit terrible. Could anyone recommend any links so I can solve these equations myself? Also, what is the usual mathematics software that is used for crossovers?