>>106318207
Posted while half asleep. Slide from a presentation by Wolfgang Klippel around 76 minutes in.
https://mega.nz/file/2pohyTAb#DNaoGwo_jI60-ZhtjLVYu39SZXhoA4PtFl3sVieHLvU

Electrodynamic drivers (moving coil, planar dynamic) move by the Lorentz Force. Force = B-field * coil Length * Current. Or
F = B*L*I

Increasing BL can increase available force, although it's not that simple at all frequencies (see Faraday's Law and its interaction near the driver's mass-spring frequency Fs).
>that pic
It shows BL plotted against voice coil position. BL is central to how the driver works, and modulating it with voice coil movement is a central cause of audio distortion. By shortening (or lengthening) the top plate past the voice coil, BL can be stabilized over some small displacements. But this does not ideally focus the magnetic field onto the coil.

>more efficient
Even hyper-compressed music still has a 6dB crest factor: peak power level is four times greater than the average. A driver spends most of its time near the center point, so you could extract more driver efficiency by sacrificing driver linearity over stroke length. Most loudspeaker drivers are somewhere between 0.1 to 1% efficient. Small telephone speakers and headphones are even worse, so any efficiency increase is appreciated.
This doesn't work if the driver needs available displacement for low frequencies. Klippel's goal is to incorporate the driver into a feedback and prediction system to control the distortion products.