There are various methods for producing AM modulated signals. Each one may require its own preferred audio processing limiter.
Focusing on transformer plate modulated transmitters with abundant audio power, the limiting factor is that the anode voltage of the RF class C output tube cannot receive negative voltage. It is indeed a fact that this voltage should not go below about 40 V for tetrodes, otherwise the output AM envelope is flat zero.
Obviously a pure high voltage resistor and diode network cannot fully correct this kind of overmodulation, it will only mitigate the effects, but the envelope will become distorted if overmodulation occurs.
Low level compressors do indeed work, but they have to be tuned with the oscilloscope and the tuning will depend on the loading on the RF output tube. In fact the high level modulation voltage depends on the anode impedance of the RF tube as seen by the modulation transformer and the audio amplifier.
The clever solution is to "detect" and "partially correct" overmodulation with a diode-resistor high level network and use some signal derived form this networ for controlling a low level audio compressor.
The end result is that it works. It does exactly what it should do. If the audio amplifier does not saturate by its own limitations, modulation will never exceed 99% (if you set it so), for any loading condition of the RF output stage.
This is the result of discussion, design and experiments.
Follow the story here:http://amfone.net/Amforum/index.php?topic=29655.0
Go to the technical proposal here:http://www.ing.unitn.it/~fontana/AM%20ALC.pdf
The proposal has been fully tested with success. It was designed with a "vintage mind"; using a filament lamp (like the fist H.P. audio oscillator) instead of a LED, it could certainly be a 1960 design...
Old Italian Buzzards net - www.ik0lrg.it
and the boatanchors-net google group