I was spending a lot of time recently trying to figure out why, in LTspice, my audio amplifiers always seemed to be giving distorted sine wave output from my regenerative detectors. It turns out that the regenerative detector itself was causing the distortion, when the modulation depth of the input signal is 100%.
Here's a simplest-possible reproduction of the problem. This is a Polyakov-style regenerative detector, adjusted somewhat short of the oscillation threshold for a bandwidth of about 1000 Hz between the -3 dB points (500 Hz between the peak and either -3 dB point). An LTspice AC analysis (frequency domain analysis) was used to generate these plots while adjusting the regeneration potentiometer.
The regenerated, resonant frequency of the tank is the frequency of the peak, about 2 MHz.
Keeping the circuit in this almost-oscillating condition, next I fed in an input signal at the exact resonant frequency, from a 50-ohm source into a low-impedance link winding on the tank, and ran a transient (time-domain) simulation. The results are shown below.
The top blue trace is the input signal, a 100% modulated sine wave with 1 uV peak amplitude.
The green trace is the signal across the tank inductor. It is correctly larger in amplitude than the input signal, due to the regenerative amplification.
But there is distortion already visible! This is at the tank inductor, before any detection or AF amplification has taken place. Notice where the input signal drops to zero, the regenerated tank signal can't drop down to zero because the stored energy in the tank is still ringing, preventing the tank energy from decaying as quickly as the input signal. In other words, the smooth, round peak of the sine wave is becoming flattened out and angular as a consequence of the tank ringing due to the high-Q regenerative amplification.
So any detection or AF amplification that we apply to this 100%-modulated signal will also contain this distortion.
I verified that the same thing happened in another regenerative circuit (the 2-transistor cross-coupled or differential pair oscillator), and also at higher voltages. It just seems to be the way things are in regenerative circuits or high-Q tanks. I spent a long time thinking the distortion was caused by my AF amplifiers, but it wasn't.
Now, if we reduce the modulation depth of the input signal, so that the input signal doesn't die down all the way to zero, the distortion seems to go away to possibly negligible levels. I would suppose that as the modulation depth gets closer to 100%, the distortion will gradually increase.
So, lesson learned: use less than 100% modulation depth when testing regnerative circuits.
I don't know if this pertains, but in my simple regens, using op-amp amplifiers and class A transistor amps affects the performance of the regen to the point where it doesn't work. LTspice is beyond my abilities right now, so I build and experiment with physical circuits. The best performance I've achieved, which actually improved my regens, is with this class AB circuit. I can receive band noise dips between broadcasts, similar to a commercial radio. As you increase the volume, there's a positive effect. Adjusting the regeneration also makes it more sensitive. However, higher volume, (P1 below) can cause the regen circuit to oscillate, so I suspect it's affecting the tank circuit. Perhaps it's the combination of components in this specific amp that makes the difference, including input impedance. I've observed this across all the regen circuits I've tried. Additionally, the AB amp can operate at very low voltages, possibly down to 1.5V. I use a voltage regulator, MCP1702 330 for 3.3 volts. it uses is very little current with this regulator, but delivers crushing headphone volume. It also seems to be able to drive other amps line level input directly from the output really well. Maybe it can be adapted to low voltage. This is the cleaned up and reduced to the least parts design I use with 2N3904 2N3906