This is a recovered file. The images in this post may be out of order, and there may be duplicates.
Varactor-tuned, hybrid-feedback, low-voltage BJT regen Post by qrp-gaijin » Mon Oct 14, 2013 I've successfully built a low-voltage, varactor-tuned, hybrid-feedback Vackar-Hartley regenerative detector/oscillator and have gathered some data on the frequency-dependence of the threshold oscillation level. For those not familiar with the topic, please see the following documents (by TheRadioBoard member vladn) for background reading: viewtopic.php?f=3&t=3714 and http://www.kearman.com/vladn/hybrid_feedback.pdf The basic idea is that the threshold oscillation level can be equalized over the oscillator tuning range, a procedure referred to below as "balancing the tilt". One goal of mine from the start was to attempt to apply the hybrid feedback approach to a varactor-tuned oscillator. My initial attempts were rather ambitious and failed due to my lack of experience with the hybrid feedback adjustment method and with varactor-tuned oscillators. This time, I decided to start with a low-voltage BJT hybrid-feedback oscillator that I had successfully gotten to work with an air-dielectric variable capacitor. There were two main problems with varactors that concerned me: 1. If the tank voltages are too high, the RF can get rectified and modulate the varactor capacitance, causing an unclean oscillator signal. 2. The varactor Q will change with frequency, which could make it difficult or impossible to balance the tilt due to the varactor-induced frequency-dependent losses. Regarding problem 1, I had hoped that the low oscillator supply voltage would keep the tank voltages low enough to prevent the RF rectification problem. After building the circuit (shown below), and attempting to monitor the battery-powered oscillator's radiated signal on my station recever, I at first had some hum-related problems; the oscillator's signal on the monitoring receiver sounded like it was hum-modulated. I initially incorrectly thought this was again a problem with high tank voltages causing the oscillator to de-tune itself. But it turned out to be something else, some sort of household noise affecting my monitoring receiver and possibly somehow getting back into my oscillator circuit. I noticed this because if I happened to touch a metal part of my desk, perhaps grounding the desk against my body, the oscillator signal was suddenly clean! I "solved" the issue by shutting off all electronic devices near my circuit and running my monitoring receiver off of batteries. After taking these measures, the regen's radiated signal as monitored on the station receiver sounds like a clean sine wave, though without a scope it is hard to know for sure about the signal purity and how high the tank voltages are. I'm still not exactly clear on the design procedure to ensure that a particular oscillator's tank voltage never exceeds a specified limit; if anyone has any advice, please share. Regarding problem 2, I was pleasantly surprised to discover that the tilt is almost balanced in my prototype. The data are presented below. I will continue to attempt to improve the tilt balance. Here is the successful circuit: hybrid-varactor.png (72.66 KiB) Viewed 6377 times Note there is a 0.1 uF decoupling capacitor (not shown) from the positive terminal of the 3V supply to ground.
The circuit analysis is as follows. If we choose Cv to have a modestly high capacitive reactance and choose Tr1.L2 to have a very low inductive reactance, the circuit becomes a Vackar oscillator, with Tr1.L1 being the main tank inductor, Cv being the Vackar feedback capacitor and Tr1.L2 being a short-circuit. Alternatively, if we choose Cv to have a very low capacitive reactance and choose Tr1.L2 to have a modestly high inductive reactance, the circuit becomes a Hartley oscillator, with Tr1.L2 being the Hartley feedback inductor and Cv being a short circuit for RF. Balancing the values of Cv and Tr1.L2 forms a hybrid-feedback oscillator. Tr1.L3 is a low-turn-count link winding designed to prevent the collector from loading down the tank. Earlier, I had tried connecting the collector directly to the hot end of the tank inductor Tr1.L1, and the oscillator signal was very unclean at the high-end of the tuning range (5-6 MHz). The signal was unclean even when using an air-dielectric variable capacitor, indicating the unclean signal was due not to the varactor, but instead to the BJT loading down the tank. POT3 controls regeneration by raising the emitter potential either higher (closer to Vcc) or lower (closer to ground), which in turn alters the amount of collector current that can flow. The purpose of R2 is to leave some of the RF unbypassed so that some RF can be taken off of the Q1 emitter. This is for a future experiment to take the Hartley feedback off of the emitter (which might allow easier tilt balancing with a potentiometer). It was a bit tricky getting this circuit to oscillate; the gain seems to be marginal. I could not get the circuit to oscillate in Vackar-only mode when the collector was connected to the link winding Tr1.L3. When the collector was connected directly to the top of the tank inductor Tr1.L1, the circuit could oscillate in Vackar mode, but was very noisy at higher frequencies, as mentioned above. To measure the tilt, I observed the voltage at the POT3 wiper at the barely-detectable onset of oscillation. One very confusing thing was that at lower frequencies, the regeneration control seems to have a "dead spot": the set will oscillate with the POT3 wiper voltage set both below AND above this dead spot. In other words, when adjusting the wiper voltage from low to high, the oscillation pattern, at lower frequencies, is: strong oscillation -> sudden stop of oscillation -> sudden start of weaker oscillation -> gradually weaker oscillation -> soft stop of oscillation. The "soft stop of oscillation" is the proper point to define the oscillation threshold. The dead spot problem did not appear at higher frequencies. The following graph shows the dependence of the POT3 wiper voltage, required to reach the oscillation threshold, versus frequency in kHz. varactor-hybrid-data-1.png (19.06 KiB) Viewed 6375 times I took many more measurements at the low-end of the tuning range because there initially appeared to be large non-linearity at the low-end, but that turned out to be a measurement error: the apparent low-end non-linearity was caused by incorrectly believing the "dead spot" signaled the oscillation threshold, when actually the oscillation threshold occurred at a higher POT3 wiper voltage. So we can see that the threshold-vs-frequency dependence is actually rather small, and mostly monotonic. I had feared that the threshold level would be jumping up and down unpredictably due to the varying varactor Q, but it seems not to be that big of a problem. Next, I will tweak Cv (make it smaller) to see if I can reduce the tilt further, and take further measurements. Some other miscellaneous notes: 1. Small Cv? The varactor maximum capacitance should be around 450 pF (assuming an inductor of 14.40 uH and a lowest frequency of 2000 kHz). In theory, the Vackar capacitor Cv should be at least 10 times the maximum tuning capacitance, or 4.5 nF; currently, is 1.6 nF. Nevertheless, the current tilt shows a deficiency of Vackar-style feedback, which should require yet more Vackar feedback and an even smaller Cv. So I believe Cv should be decreased, not increased. However, it is conceivable that non-linear varactor losses are being offset by non-linearity caused by an overly-small Cv, which would make it difficult to predict the effects of changing Cv. 2. Applicability to higher frequencies? I tried to get the circuit to oscillate in Vackar-only mode at higher frequencies. I used an inductor consisting of 15-turns on a T50-6 (yellow) toroidal core and a 150 pF air-dielectric variable capacitor. I connected the collector directly to the hot end of the new inductor. No oscillation was observed. Then, the collector was connected to a 6-turn coupling link instead of directly to the inductor. Again, no oscillation was observed. It may be necessary to reduce Cv for higher frequencies. Or, there may simply not be enough gain at higher frequencies with my choice of transistor, Vcc, and biasing arrangement. Conclusion: It does appear possible to largely balance the tilt in a varactor-tuned oscillator. However, the Cv value is smaller than would normally be expected. Experiments continue. Proof of existence: hybrid-circuit.jpg (151.76 KiB) Viewed 6377 times --- Edit: I tried removing one tickler turn from the Hartley feedback portion of the coil to reduce the Hartley-style feedback (instead of decreasing Cv to increase the Vackar-style feedback). This shifted the tuning frequency range upwards, resulting in oscillation from about 4 MHz to 11 MHz. (This in itself is rather startling: removal of one turn shifts the oscillator frequency up several MHz? This needs further investigation.) The upper half of the tuning range was essentially completely tilt-balanced. However the lower half of the tuning range often exhibited an aggravated version of the previously-mentioned problem of the "dead zone" where the regeneration control would snap out of oscillation at some low-wiper-voltage setting and snap back in, with weaker oscillation, at a higher-voltage setting, finally fading out properly at a yet higher-voltage setting. At some lower frequencies the weaker oscillation region "above" the dead zone (higher wiper voltage) did not seem to exist, meaning that when slowly increasing the pot wiper voltage, the set would suddenly snap out of oscillation at some low-wiper-voltage setting, and would only snap back into oscillation (with hysteresis) when the voltage was again reduced beyond the point where oscillation had suddenly stopped. This whole "dead zone"/hysteresis issue leaves me feeling a bit uneasy and has me scratching my head. What may be happening is that for whatever reason the set is losing gain (at certain frequencies) when the regeneration control is set into the dead zone, which would indicate a need for increased gain, which could be had by increasing the turns on the collector link winding Tr1.L3. Another hunch is that R2 may need to be tweaked. Finally, it may turn out that the tilt can only be balanced over a portion of the varactor's tuning range due to frequency-dependent varactor losses. As I recall, varactor Q decreases with decreasing frequency (reference: 1SV149 varactor datasheet). Edit 2: It turns out I was listening to the second harmonic of the oscillating regen, which explains why I thought the frequency had shifted up so drastically. It hadn't. Edit 3: The "dead zone" issue may simply have been caused by a physically defective potentiometer which gave unpredictable jumps in resistance when adjusted to a certain position.