First video of MFB correction on a planar magnetic driver 

I experimented about 20 years ago with mfb on ESLs. The thing is, like in your case, only a small piece of foil would be sensored, less than 1%. And not all of the surface is doing the same, of course, the resonances varied with the position of the surface, especially if it is divided in different parts. So it is a total different task in comparison to the movement of a stiff Bass membran....at the end I come to the conclusion to work with a speaker management system, herewith is it easier to compensate the system frequency and resonances.

You are much more interesting than watching the Eclipse today. The improvement at 50Hz is great. I think you improved the linearity of the sensor by moving farther from the panel. The motion of the panel is now a smaller fraction of the total distance to the sensor so you are operating on a smaller segment of the 1/x curve, so a bit more linear. A very good idea. Can you share the components used in the sensor? Maybe they are/ are not special photo transistor & LED? I expect the response with the microphone very close, 3 cm, from the panel would now be very flat with feedback. I assume the variation in your measured response is due to the open baffle / baffle step effects.

If your goal is to improve distortion, you should really consider using either a first or second order highpass on the IR sensor to approximate the membrane acceleration instead of the position. Above resonance, the panel terminal voltage is proportional to driver acceleration, so you're trying to correct acceleration using a position reading. The only functional loudspeaker MFB systems I know of are either using accelerometers in the voice coil or Rythmik's design uses a secondary "sensing coil" to detect driver velocity. I think velocity sensing (first order highpass on your IR sensor) is a good compromise between distortion cancellation and loop stability. A positional sensor like you've got right now should really only lower distortion below resonance where the panel terminal voltage is proportional to position. Using an IR distance sensor can work to boost the bass, but if your goal is to correct for distortion, you're gonna need more loop gain in the midbass. Using MFB on a planar driver is an amazing idea though, I wish I had thought of it. Good luck!

Hexagonal shaped planar panels might be helpful when observing cymatic science. Bigger is better when using panels is right especially lower Hz. Bass clarity.

Hi Joppe, I like very much your work! Have you ever tryed to use electrically conductive spray paint instead of aluminium tape?

For a long time I have been wondering that a compensating signal should be sent to a separate, isolated emitter, there will be a radiator of a smaller area, which will be loaded only with this signal. Maybe I'm wrong. There was no way to verify it.

IDK. In headphone world planars are considered electomechanically stiff. The diaphragm behaviour is pistonic. I think in your loudspeaker, as long as you're aiming at the trace, you should have a proper representation of the rest of the diaphragm.