Lumped parameters are in the range of PA subwoofers. Generous overhang suggests high excursion performance for the category. This woofer faces a competitive market space.
Do you know what makes IMD? How does harmonic distortion relate to IMD? For example if we have a driver with higher 2nd HD and lower 3rd HD, how does the IMD compare to driver with lower 2nd HD and higher 3rd HD?
Having said that, the essential relationship as I understand it
Any model of nonlinear distortion that produces nonlinear harmonics will create intermodulations with additional tones. The simplest model is a polynomial series (a₁x + a₂x² + a₃x³...). The coefficient is the particular characteristic of the system, the term exponentiated is the input tone(s). For one tone, we have:
a₁(b*cosωt) + a₂(b*cosωt)² + a₃(b*cosωt)³...
a₁b(cosωt) + a₂b²(1+cos(2ωt))/2 + a₃b³(3cosωt+cos3ωt)/4...
You may notice the static offset associated with the even ordered term. Higher terms change more rapidly (per 10dB input change, 20 dB second order, 30 dB third order, and so on).
For two-tone inputs, second-order only:
a₂(b*cosω₁t + c*cosω₂t)²
a₂(b²(1 + cos2ω₁t)/2 + b*c(cos(ω₁t+ω₂t) + cos(ω₁t-ω₂t)) + c²(1 + cos2ω₂t)/2)
Filling in the other orders is a tedious exercise in trigonometric identities and the multinomial theorem. The mixed middle terms, representing intermodulation, are capable of presenting the sum and difference associated with periodic inputs.
What was meant was that the elliptical shape of the voice coil and dust cap is "asymmetrical" to the circular cone, surround and circularly arranged magnets.
The result should be (as I understand it) that distortions of odd order are suppressed, at the price of a more distortion of even order, as asymmetrical nonlinearity generates high even order distortions.
For example when the cone or dust cap breaks up into partial oscillations.
The geometric arrangement here does not directly relate to harmonic products but the linear frequency response. The goal is to decompose the VC former coupling to individually strong axially symmetric resonances into separate lesser resonances. This is not to make the driver behave more rigidly, which Scan-Speak has usually avoided in the past.