Now then, where was I?
This project evolved out of discussions with Mike, AA1TJ. What I thought would be a simple diversion from the FETish receiver turned into a marathon. In the end I decided I'd made too many assumptions, jumped in too far, and had to go back and start over. Hams getting started in home-brewing their own equipment may mistakenly believe experienced builders can whip up working circuits in no time at all. When their own attempts are fraught with chaos and confusion, they may get discouraged and buy a kit.
Over time, if you keep good notes, you develop your own cookbook of circuits that work, but every project is a new experience. Perhaps we'd do more to encourage homebrewing if we published our failures as well as our successes! These pages are a record of my efforts to make a regenerative 40-meter receiver tuned with a variable-crystal oscillator (VXO) instead of an LC circuit. As the saying goes, sometimes you learn more from your failures than from your successes.
Regenerative Receiver Theory
Regenerative receivers ("regens") are a simple way to achieve high gain and reasonable selectivity in a single stage. As the positive feedback of the stage is increased, at some point the tuned circuit resistance goes negative, massively increasing its Q (Q = X/R). When the stage is teetering on the verge of oscillation, the tuned-circuit Q is increased by several orders of magnitude.
In radio's early days, vacuum tubes were inefficient and expensive. Many sets operated from batteries, and fewer tubes meant longer battery life. It was in the designer's interest to get as much gain from each stage as possible. In a regen receiver, gain increases with tuned-circuit Q. Coils, though, are always lossy, while capacitors have little practical loss. W9VES notes that high-Q LC tank circuits give less frequency stability when receiving strong signals. With modern devices, gain is cheap, and the Q multiplication effect of a regen makes the Q of even modest LC tank circuits look superb.
If LC is good, is a crystal better?
The theory behind this project was to use the higher Q of a crystal compared with an LC circuit, to achieve even higher Q, and thus better selectivity. This idea seemed so simple, I wondered why it was so rarely used. A solidly-built VXO operating at low supply voltage is inherently stable. Crystals have inherently high Q. Why not replace the lumped inductance and capacitance of a tuned circuit with a crystal?
Enter the VXO
Crystals are used in oscillators where it isn't necessary to change frequency. Hams discovered that crystals could be "pulled" by combinations of inductance and capacitance, to form a variable crystal oscillator, or VXO. Perhaps the ultimate VXO is described by 7N3WVM, the "Super VXO." A key element of the Super VXO is its use of paralleled crystals of the same frequency. So, why not a Super VXO regen?
The Saga, truncated version
On the off chance that my missteps will be useful to someone, I'll briefly mention them. I started out with a Super VXO circuit I'd used several times. I used a J310 FET, though I generally used devices with higher Idss ratings in other VXOs.
The VXO worked great, tuning from 7039 to 7019 kHz. With a reduction drive turning the variable capacitor, the 20-kHz range would provide a reasonable tuning rate without a fine-tuning control. Now, where to inject the incoming signal? The typical regen uses a parallel-resonant tuned circuit, with high impedance at resonance. I tried injecting the signal into the FET gate, forgetting that the crystals and LC formed a series-resonant circuit, with low impedance at resonance. For CW, I thought the necessary offset to get an audible beat note might overcome that problem, but signal gain was poor.
AA1TJ reminded me of a Ham Radio article by W8YFB, which shows a Colpitts crystal oscillator configured as a regenerative detector.1
W8YFB Crystal Regenerative Detector
W8YFB didn't use a VXO, so his circuit wasn't tunable. He also adjusted regeneration by varying the gate-to-source feedback capacitor. I didn't want to deal with a variable cap floating above ground, which requires insulated mounting and coupling. But I tried his method of feeding the input signal to the source. After some experimenting with source and gate resistances, I still didn't feel the circuit was providing enough gain. If you've used a regen, you know that as you reduce feedback the gain rises dramatically. I wasn't seeing this effect.
N1TEV and WB2UID (now WU2D) described a simple regenerative W1AW receiver that used a 3.579-MHz color-burst crystal as the resonant element.2
N1TEV and WB2UID (WU2D) Crystal Regenerative Detector
This circuit is interesting because it uses a feedback transformer, like traditional LC regens. The collector terminal serves as signal input and output. The authors used a diode detector to demodulate the detected signal. One of the authors later referred to this circuit as "a direct conversion regenerative receiver," in which "[t]he Germanium diode is in effect, a product detector." Regens are direct-conversion receivers, or "autodynes," as they were once called.3
While not widely tunable, the 'W1AW' receiver gave me an idea. Why not couple the signal in series with the crystals and VXO-tuning components, with a transformer? My first attempt, using a ferrite-core transformer, failed. Even one turn on a small binocular core was enough to kill oscillation. No adjustment of biasing would make the stage oscillate. What did work was reducing the feedback caps. Unfortunately, a VXO requires larger feedback caps to achieve wide frequency swing. While reducing the caps from 150 to 47 pF got the stage oscillating again, it would only tune about 5 kHz.
The junkbox produced a T50-2 iron-powder core with two, 15-turn windings in place. Eureka! Now the stage worked as before, with 150-pF feedback caps and full tuning range. And this arrangement gave me more signal gain than the other methods. To encourage the circuit I paralleled the secondary with a 470-pF cap, which resonates it at about 7 MHz. I'm not sure this improved performance very much, but it didn't drastically affect tuning range and it made me feel better.
How do I get out of here?
Having solved the signal-input puzzle, I now had two options for extracting the signal: Source or drain. N1TEV favors the source, and I tried extracting the signal there for most of these experiments.4 At this point it's worth mentioning that the drain voltage at which the stage oscillates is about 2.5 V. Experiments with the source resistor showed that a value around 1500 ohms gave reliable oscillator starting and smooth control; the oscillator slides into and out of oscillation, it doesn't jump.
Original KR1S VXO Regenerative Detector
So far, so good, but I couldn't verify that I had a true regenerative detector, where gain and selectivity peak just below the point where oscillation begins. What's different about this circuit, compared with a typical regen, is the VXO arrangemen. You'll note that I am using a total of 44 µH of inductance, and the inductors are molded rf chokes, intentionally made to have low Q. Then there is the mass of two crystals in parallel. Was a VXO self-defeating? The only way to find out was to build a single-crystal oscillator with no VXO components, and try to optimize it as a regenerative detector.
1 References appear at the end of page 2.