[roamingryan]

I was left wondering the same.

My best guess would be that slope detection could be converting the FM signal into AM. This happens in circuits where the response is strongly dependent on frequency (see [1]). A specifically designed filter is typically used, but a resonant circuit like the headphone wires could do something similar. But then you still need to get from AM to audible, and to do that you still need some sort of non-linearity in the system. Some possibility non-linearies could be a poor connection resulting in a whisker-style rectifier junction, or perhaps a magnetic permeability non-linearity in the headphone speakers. I'm sure there are others.

Related, but probably not relevant: Stereo FM broadcast signals in most parts (all parts?) of the world also contain an AM subcarrier. The primary FM carrier has the L+R audio content and is used by both monaural and stereo receivers. Stereo receivers make use of an additional AM modulated sub-carrier (+/- 38 kHz from primary carrier) to obtain L-R audio content. I don't believe this AM carrier is directly rectifiable because it does not manifest as envelope modulation in the full signal.

[1] https://www.electronics-notes.com/articles/radio/modulation/...

[jjoonathan]

Slope conversion is the hard part, once it's AM just about any energy conversion will do. abs(x) and |x|^2 are plenty nonlinear.

https://youtu.be/uo9nGzIzSPw?t=19