1.2-Volt Regenerative Superheterodyne, Part 1: A well-working circuit

This is a receiver project that I've been working on sporadically for several years. I'm still working on it, but it's reached a stable state that I wanted to share.

The circuit is a shortwave regenerative superheterodyne that runs off of 1.2 volts. I discovered a number of pitfalls while working on this receiver, which I will summarize below.

Completed circuit

Here is the circuit. There's not anything really special here, but the devil is in the details.

Circuit description

The signal enters the receiver at the resonant circuit L3/VC2, where L3 is a ferrite rod antenna. 

Q1 is a single-transistor mixer, whose gain is modulated by the cross-coupled oscillator formed by Q2 and Q3.  

The IF output, about 2 MHz, appears at IFT1, where it is them amplified by 4 common-base IF amplifiers Q4, Q5, Q6, and Q7. 

The base bias of the second IF amp Q5 is controlled by AF-derived AGC; Q5's base bias is pulled down by the variable current through Q15, whose average current in turn is determined by the amount of AF signal fed into the Q15 base; the higher the AF signal, the more current through Q15, which diverts more of Q5's base current to ground, reducing Q5's IF gain.

The last IF amplifier's output tank, IFT5, is regenerated by the cross-coupled oscillator (held slightly below oscillation by regeneration control VR2) formed by Q8 and Q9. The receiver has sufficient gain without regeneration, but regeneration allows narrowing the bandwidth for better selectivity. Thanks to the AGC applied to Q5, the signals reaching the regenerative stage are small enough to prevent overloading the regenerative stage and to preserve selectivity. The regeneration can be kept at a fixed, below-threshold level for the desired level of selectivity.

The amplified IF signal is sent to the non-linearly biased detector transistor Q10.

The AF output from the detector (with RF filtered out by C7) is sent to the 3-stage AF amplifier Q11/Q12/Q13, then finally sent to the power amplifier Q14, which has a low output impedance and can drive 32-ohm headphones to a very loud listening volume.

Pitfalls

The circuit sounds simple enough as described above, but several pitfalls were encountered:

1. Voltage stability: Unwanted AF oscillation can easily occur when a regenerative IF amplifier (Q8/Q9) is used. The increased current flow through the regenerative stage at or near the oscillation threshold can cause a slight sag in the supply voltage, and this ripple on the supply voltage can then again be amplified and cause AF oscillation. One or even two RC low-pass filters in the Vcc line, at various locations, were insufficient to remove these oscillations. However, adding diode D1 in the Vcc line sufficiently stabilized the supply voltage; the more-or-less constant voltage drop across the diode provides a "stabilized" 0.7-volt Vcc_2 line. This "stabilized" Vcc_2 line is then used to power the most voltage-sensitive stages: the LO, and the regenerative amplifier. Detector transistor Q10 also needs to be powered from the stabilized Vcc_2 line, or else AF oscillation can still occur.
2. Fixed regeneration level: If the regenerative stage is connected directly to the mixer, then the impedance changes caused by tuning the LO and RF tanks will cause variable loading on the regenerative stage and will pull the regenerative stage above or below the oscillation threshold, requiring occasional or frequent adjustment of regeneration as the set is tuned. But by placing several IF amplifiers ahead of the regenerative stage, this sufficiently isolates the regenerative stage from the LO/RF impedance changes and allows the regeneration level to remain fixed as the set is tuned.
3. AGC and frequency pulling: Applying AGC to the first IF amplifier Q4 will cause variable loading on the mixer as AGC is applied, which in turn will cause a pulling of the LO frequency. This means that when listening to SSB or CW, the LO frequency will vary on signal peaks, causing chirp on CW signals and distortion on SSB signals. Applying AGC to the second IF amplifier Q5, instead of the first IF amplifier Q4, helps to minimize the LO frequency pulling.
4. Common-base and AGC: Common-base amplifiers are convenient for implementing AGC, because the base is already grounded at RF, and hence the base bias can be adjusted by an AGC mechanism (in my case, the non-linearly-biased Q15) without needing to worry about how the changing biasing affects the input signal.
5. Common-base and stability: Common-base IF amplifiers do not need to worry about neutralizing unwanted output-input feedback through Miller capacitance, because the Miller capacitance is grounded by grounding the base. If, on the other hand, a common-emitter amplifier were used, then neutralization might be required. An earlier version of this receiver used common-emitter IF amplifiers, and had problems with unwanted oscillation of the IF strip.
6. Requirement for tuned IF amplifiers: Because the mixer is unbalanced, there is a strong LO component at the first IF output tank IFT1. If we were to try to use untuned IF amplifiers instead of tuned IF amplifiers to amplify the mixer output at IFT1, then these untuned amplifiers would amplify both the unwanted and large LO signal along with the wanted and much smaller IF signal. This would lead to the smaller IF signal getting drowned out by the LO signal. In practice, untuned IF amplifiers were tried, and while one untuned IF amplifier seemed to offer some marginal amount of amplification, two untuned IF amplifiers resulted in a great increase in noise -- the amplifiers were likely being driven into clipping, as they were attempting to amplify the huge LO signal, and once the amplifier starts to clip off the peaks of the signal, it is clipping off the tiny amplitude variations at IF that we wanted to amplify. An untuned amplifier might be made to work, if a high dynamic-range amplifier and a higher supply voltage were used, but the fundamental problem still remains, that the unbalanced mixer is outputting a very large and unwanted LO signal. By using several tuned IF amplifiers, we simultaneously filter out the LO signal and amplify only the desired IF signal, staying within the amplifier's limited dynamic range, which is especially important at this low supply voltage of 1.2 volts.
   
7. Final power amp to drive 32-ohm headphones: An emitter-follower was initially tried for Q14, with poor results; best results and highest gain were obtained with the common-emitter configuration shown for Q14.
8. RF coupling to mixer: Connecting the ferrite rod antenna to the mixer transistor Q1 is intentionally done with coupling capacitor C17, instead of using an inductive link winding on the ferrite rod antenna L3. If a link winding is used on L3, its small inductance will resonate at FM broadcast frequencies and allow strong, local FM stations to be audible at many points on the tuning dial. Using capacitive coupling with C17 reduces, but does not eliminate, the problem of strong FM station breakthrough. FM stations are audible when a harmonic of the LO frequency lies near a FM broadcast frequency. Somewhat surprisingly, connecting the hot end of the RF tank L3/VC2 through C17 to the mixer's base does not seem to appreciably load down the RF tank. This is probably due to a combination of the mixer's low current biasing arrangement, plus the fact that the mixer is not conducting much of the time, during those portions of the LO cycle when the LO prevents the mixer current from flowing.
   

Performance

The circuit works well, and can hear down to the mixer noise. This can be confirmed by shorting out L3 (removing any RF input from the mixer), then shorting out L9 (input to first IF amp Q4). When L9 is shorted, the receiver noise decreases; when L9 is again in-circuit, the receiver noise increases, which is the noise from the mixer Q1 (without any noise from the RF input, because the ferrite antenna L3 is shorted to ground).

The regenerative stage can be set at a fixed, below-threshold level for AM reception and need not be adjusted as the set is tuned.

The set is stable enough when oscillating for intelligible SSB and CW reception, though some frequency chirp due to AGC is noticeable.

Problems and next steps

With an IF at 2 MHz, image signals are only 4 MHz away. While listening to a shortwave broadcast band, image signals from another shortwave broadcast band (4 MHz away) can often be audible. This is because the single-tuned RF front end has poor selectivity, and because the image signals can be extremely powerful when they come from another shortwave broadcast band. Strong image signals might also beat with strong in-band signals to create heterodyne whistles.

The easiest way to solve this problem is with a higher IF of 12 MHz, which will cause image frequencies to be 24 MHz away from the tuned frequency. The selectivity of a single-tuned RF circuit is then sufficient to attenuate the image signals (because they are so far away from the desired signal), and furthermore, this high-frequency choice of IF prevents image frequencies from any shortwave band from overlapping with any other shortwave band. For instance, when listening to a station at 7.2 MHz in the 41-meter shortwave broadcast band, the image frequency will lie at 7.2 + 24 = 31.2 MHz, which is outside of any shortwave broadcast band.

However, creating an IF strip at 12 MHz will be slightly more difficult than the current 2-MHz IF strip, because of the increased tendency for unwanted feedback and oscillations at this higher frequency of 12 MHz. 

Future experiments and articles will focus on building the 12 MHz IF strip. Also, future articles might show some LTspice simulations of various parts of this circuit, such as the mixer frequency conversion process, the IF strip gain, the detector performance, or the AF strip gain.