Asymmetrical AM (Amplitude Modulation).

[Mike Watterson]

..., it's rare audio that is net asymmetric as that would be a DC offset. The AC coupling will remove it.

I'd have to disagree there. Although it may be rare audio, it's quite easy to have a waveform that is asymmetric, one peak much higher than the other, with no DC offset - so AC coupling does nothing to remove it.

Waveforms with even-order harmonics will often be asymmetric. The waveform I sketched is an example - no DC offset, but one peak much higher than the other.

It's trivial to create asymmetric waveforms and maybe irrelevant what the initial DC was once it goes through an audio chain.

You are agreeing actually with what I meant to write and essentially what I did write.

BUT:
1) How common is such asymmetric audio and what if it's inverted?
2) What does an AM modulator with gain adjusted so carrier is never less than minimum (1%? 5%?) do with such a signal?
3) Is adaptive carrier power (normal on all but low power AM) going to affect it.
4) Surely it's only instantaneous peak power that's affected and not average power? So would the more than 100% positive when the carrier is never off be only an instantaneous value?

Certainly I think can actually generate suitable audio (I have a DDS based arbitrary waveform generator and physically measure effects on modulation. I have a second function generator with AM in and two valve based signal generators. At least one uses a transformer Audio modulator to supply HT to RF anode, so unlike a multiplier with injected carrier, it can in theory do more positive than negative modulation. However I'm sceptical that is what is done deliberately. I'm sure the AM transmissions use a regular symmetrical compressor, if "loudness" is wanted and adaptive carrier power based on the average audio level after compression.

Maybe someone expert with LT Spice or Labview can simulate it?