Some interesting links about superregens.
AM to FM conversion. Quote: "AM received modulation definitely ends up by frequency modulating the quench waveform". http://www.amalgamate2000.com/radio-hobbies/radio/fm%20super%20regen1.html
Superregen at HF: http://www.amalgamate2000.com/radio-hobbies/radio/super%20regen%20at%20HF.htm
I once saw a French design post WW2 where the coil was periodically shorted by a forward biased switching diode. The idea as I understood it was to remove the need to vary the detector gain, its operating point, therefore any internal capacitance changes, thereby reducing the frequency spread and RX bandwidth. The seperate quench oscillator was applied to the diode to stop main oscillation. The nearest I ever got was to use an ECH84 basically as an overdriven mixer, a LF squarewave quench was applied to the third grid, effectively gating the valve on and off. But, even here with very good grid isolation, I noticed no reduction in bandwidth. These things are fascinating.....
I posted this on the "old" TRB forums (without response), but here it is again: there's an interesting expired patent about an "inverted superregenerative receiver" that tries to keep the oscillator stage always close to the critical threshold (not allowed to be quenched all the way back down to zero) so that the tank is in a high-Q condition for a longer period of time, leading to better overall selectivity. Rather complicated to do, but a very interesting idea. Here's the link: https://patents.google.com/patent/US6668165 .
I have often described a superregen as electrical "chaos". The article you reference has nothing to do with FM detection. The quench frequency here is reduced into the audible range and will "pull" slightly higher or lower from a center freq as it tunes through a signal, either AM or FM modulated, or even if an external tuned circuit resonates within the SR freq range, this will also pull the SR and change the audible quench tone..... the change in tone used only to audibly indicate a change has taken place, as with a normal GDO with meter indication. A strategically placed finger will also give a freq shift, albeit not very useful Im guessing, unless you are desperately needing to detect a finger?. The fundamental limitation with NFM and the superregen is the frequency difference between the carrier and the modulation shift as already explained. The superhet artificially amplifies the difference due to the mixing down process, a 1kHz shift at 100 mHz is still retained as a 1kHz shift at 455kHz for example. The SR cannot resolve this difference at the signal frequency UNLESS the RX bandwidth could be reduced to that necessary to slope detect as with WFM. This is not possible. The "Q" of the SR tank is constantly varying , from a minus when not oscillating at a quench null, slowly increases to a maximum until oscillation takes place , and then reduces as oscillations die away. The bandwidth is directly related to "Q" and "Q" is only at a maximum during a very small part of the quench cycle. The RF signal sees the "Q" as an average figure, the value dependent on quench wave shape and amplitude, and to a lesser degree, quench frequency. That explains why these parameters have so much effect on performance and why SRs have a "repeatability" problem, and why "squirelly" ones are likely to be more sensitive than the well behaved . Like the old nursery rhyme goes......, when she was good, she was very very good, and when she was bad she was horrid.......
Greetings Dayle! I remember seeing you around on the "old" TRB forums. Your comments are enlightening about the limitations of the FM modulation in a superregen. Even if limited, the FM modulation can be put to use, for example to make an audible dip meter, as done here: http://vk2zay.net/article/133 . As Alan Yates writes in that article:
I keep wanting to do more experiments with superregens -- it's such a seemingly simple circuit, yet filled with interesting complexity.
Sadly there is no mechanism in a superregen to demodulate FM, wide or narrow. Although it is very dangerous to say never, my experiments over many years have had zero success in reliably demodulating NFM with the SR. Slope detection can only give sufficient audio recovery with wide FM signals. Roughly put, a WFM signal at say 100mHz will easily slope detect with the wide bandwidth of a SR at that frequency..... 100mHz /150 kHz deviation has a modulation index of 666.. A NFM signal at the same 100 mHz with a deviation of 15kHz will give an index of 6666, or a recovered audio output of at least 20dB less . This explains why in the usual superhet rx NFM demod, a low IF frequency, usually the standard 455kHz is used . This then gives a modulation index of 30 under the same circumstances . For the same reasons, any SR quench frequency variations caused by FM modulation are limited by the same constraints, and therefore cannot be used as any audio frequency determining mechanism. One idea that crops up occasionally is to use a BFO at the signal center frequency giving a difference signal with NFM, and somehow eliminating the audio mirror image. This would no doubt work, but increases complexity beyond the simple elegance of the SR, Another method may involve the fremodyne concept, and processing the resultant IF, but again, a complex mess i think, better served with a superhet design.