http://www.valontechnology.com/images/REGEN.PDF

Receiver Design

Superregens are clearly unique and probably the least understood of all the receiver topologies. Much has been written on the subject of regenerative and superregenerative receivers in previous works and the theory is covered in detail in the references provided2.
However, a good engineering analysis of the circuit designs concepts remains scarce. I believe this is due as much to the highly competitive nature of the commercial applications, as it is to the complexity and dynamics of the detector's operation.
The superregenerative detector stage operates simultaneously at three frequencies: one is typically a VHF radio frequency, another is the super sonic quench frequency, and the third is the audio or baseband data frequency. The RF frequency of interest is amplitude modulated and will have a signal level -30 to -100dBm. The quench frequency will be simultaneously changing the bias, and as a result the gain, over a very large range including off. All of these disparate frequency and amplitude consideration defy simulation with most current software tools making circuit design and performance prediction
extremely difficult.

The simulation tools that the modern radio designer has available fall into three categories:

1. Linear simulators such as Eagleware3
2. non-linear simulators such as PSPICE4 and
3. advanced work station non-linear simulators such as Hewlett Packards MDS or Microwave Harmonica

Linear simulators will not be able to deal with the bias changes cause by the quenching action nor will they be able to predict the onset of the RF oscillation. SPICE simulators should be able to model the function but a huge number of samples would be required with very short sample periods in order to capture all three frequencies of interest. It is doubtful that the data obtained with SPICE methods could be used productively. The work station simulators, while processing the capability, require a great deal of knowledge and experience to set up properly for accurate results. In any event I highly doubt that even the best methods available could predict the circuit operation with the degree of certainty that would be required in order to optimize a design for maximum sensitivity, lowest power, or frequency stability.