6 Transistor regenerative SW RX with modified feedback.
This radio from my dusty archives remains the best performing transistorised regen I have built to date. It is only suitable for SW AM listening, for this it is very stable and tunes easily.
The RF signal arrives at the input buffer via the simple attenuator control. This reduces overload on strong signals and acts as a volume control.
The 40673 detector transistor is a historic relic from the dinosaur era, I see these are still available at a ridiculous price, but any DG Mosfet will suffice.
The signal from the input buffer , once resonated in the tuned circuit is introduced into gate 1 of the mosfet. The gain of the mosfet is varied by adjusting the voltage on gate 2, a similar situation to adjusting screen voltage on a pentode valve. This reduces any frequency shift with gain.
The feedback is from source to the tapped tuning coil.
The diode is connected directly to the DC bias on the drain, and is lightly bypassed to ground via a small capacitor, and forward biased by the 470k to ground..
It is not often realised, but this capacitor connects the Drain back to the tuned circuit via the ground path, and is part of the regenerative feedback circuit.
As regeneration increases, the Drain current increases, as a result the drain voltage drops slightly, decreasing the minute current through the diode. Ideally, this diode is sitting on its "knee" of its turn on conduction angle. As the current reduces , its dynamic impedance increases, reducing this feedback path. The 1N4148 is a switching diode with a very sharp turn on. Also, as the detected audio waveform modulates the dynamic resistance, negative feedback increases on modulation peaks, giving a cycle by cycle correction, the small bypass capacitor holding the diode cathode at an average level, rather like the action of an automatic diode noise limiter, that clips any spike above the average audio level.... this reduces gain above any average RF level.
The amplified RF at the drain is also rectified by this diode, its operating point continually kept at the knee by the interaction of the Mosfet and the charge on the bypass capacitor, and also the first audio coupling capacitor, the latter is effectively DC grounded by the forward biased transistor junction. The net result is a secondary negative feedback loop tending to stabilise the operating point.
Its not perfect, but seems to work well for AM, SSB will drive the detector away from oscillation on carrier peaks, preventing good demodulation.
The rest of the radio is just audio stuff, high gain required as the recovered audio from the diode rectifier is of necessity at a very low level.
I fiddled with the 470k diode load for a very long time to get the best results.
A 78L05 regulator supplies the RF and front end audio circuits, current drain is a few milliwatts.
With the coil as shown , and a paralleled polyvaricon AM type tuning type, coverage is just over 4:1, or in this example, 2.5 to 12 megs. A large tuning reduction is required.
I'm coming back to review this circuit again, and must again say, this is a very unique and interesting design. As I understand it, the design can be summed up as follows:
Design a regen that uses a throttle capacitor for feedback. This requires taking the inverted output from the collector or drain and connecting the throttle capacitor there (because what is actually happening is that the inverted output is getting fed back through Miller capacitance to the base/gate input, which damps the tank; the amount of negative feedback through the Miller capacitance is controlled by the throttle capacitor at the collector/drain).
Replace the throttle capacitor with a forward-biased diode in series with a parallel RC network. This provides a "dynamic throttle capacitor" that, amazingly, automatically reduces the feedback as the oscillation becomes stronger.
Increase the amount of feedback (via coil tap or tickler coil) such that the effect of this "dynamic throttle capacitor" is as large as possible, to make it sensitive to small changes in the oscillation level and the collector current.
Use device bias as a manual regeneration control.
I have some ideas about how to do this with a BJT circuit. Currently pondering how to test such a circuit design in LTspice.