However, in my receiver, I am still getting unwanted RF-LO interaction. Specifically, when I tune to higher frequencies like above 15 MHz or so, then tuning the RF tank (which is not currently ganged to the LO tuning) can pull the LO frequency, because my single-transistor mixer has minimal, if any, isolation between the RF tank and the LO tank. You can avoid this by using a buffer stage between LO and the mixer. A good buffer stage can be a cascode amplifier, or a Long Tailed Pair amplifier. Use an emitter follower to buffer the output from these two aforementioned amplifier types to the emitter of the mixer BJT. Or you can use a shunt feedback amplifier as a buffer stage between LO and mixer. In order to achieve a good conversion gain of any BJT mixer stage, the emitter of the mixer stage must see a low impedance at IF, therefore an emitter follower is needed to feed the emitter of a mixer BJT with the LO signal. Otherwise degeneration at IF will occur, causing a low conversion gain.

https://www.youtube.com/watch?v=8I0xQOIYcas "Trying a Franklin regenerative receiver for HF". In this receiver a J-FET is used as a separated infinite impedance detector.

Here is a self-oscillating mixer with a differential pair oscillator. The tail resistor R2 can be substituted by a BJT current source on which you connect the RF tank circuit, in stead of connecting it onto the base of T1. The base of T1 is then connected to the Vcc supply rail. Here is another LTP mixer with an oscillating Hartley current source. This circuit diagram below comes from Elektuur July/August 1974. Disadvantage of these two circuits is that the Local Oscillator can become injection-locked upon strong RF carriers from the antenna tank circuit, especially when the difference signal (IF signal) between the RF signal and the LO signal is small, for example 455 kHz. The LO can also become detuned when you peak the RF tank circuit at the RF signal frequency. This is caused by the parasitic junction capacitances of the BJTs in these circuits, these capacitances cause that the two tank circuits (RF and LO) can 'see each other'.

To stabilize the quiescent point of Q3, an emitter resistor of several kOhms is needed, which must be bypassed by a capacitor of several uFs. But this will introduce an RC time constant which can cause squegging/motorboating when a a weakly oscillating Q-Multiplier is used to improve the Q of L1/C2 and detecting SSB/CW.

You can also use a potmeter to adjust the bias of the detector stage. A potmeter in stead of the LED D1 and R1 connected onto the wiper of the pot.

I thinkt it's a good practice to include C2 over Q3, otherwise the impedance of Q3 forms together with the base-emitter capacitance a low pass filter which causes a gain roll-off at shortwave frequencies.

Dynamic signal envelope related changes in input impedance of any detector stage can also cause regeneration backlash.

Here is an improved version of the noisy regen which consists of an AC-linked differential pair Q-Multiplier in which the regen control is done by changing the base bias current of both Q1 and Q2. R5 is the course regen control, R6 is the fine regen control. The current mirror detector circuit Q3-Q5 can also be connected at the top of the tank circuit L1/C4. Circuit has been drawn in DigiKey's Scheme-it at https://www.digikey.nl/en/schemeit/project. Advantages of this circuit: • Q1 and Q2 don't have to be equally matched, because each BJT has its own independent bias. This will give a smooth and backlash-free regen control; • The base-collector junctions of both Q1 and Q2 will always be reverse biased, which reduces loading of the tank circuit L1/C4 and causes less phase noise when the Q-Multiplier is oscillating while detecting SSB and CW; • The tank circuit L1/C4 has been lossely conected to the base of Q2 by using a small coupling capacitor C5, which also reduces loading of the tank circuit and decreases phase noise.

Isn't it better to put a collector load resistor between the collector of Q2 and the positive supply rail?

How high is the input impedance of Q3, does it remain constant while adjusting the regeneration level and does the detector stage Q3-Q5 have some AF noise which buries weak AM signals? Is it better to use a collector load resistor between the collector of Q4 and the positive supply rail and is this detector circuit also usable at higher shortwave frequencies up to 20 MHz?

While the Noisy Regen may be a poor AM detector, but this circuit is very suitable for using as a synchronous FM detector to detect wideband FM broadcast stations. See https://www.electroschematics.com/small-fm-receiver/ "Simple FM Receiver Circuit". Circuit diagram (Source: https://www.electroschematics.com/small-fm-receiver)

Is this the early effect of any BJT? See https://en.wikipedia.org/wiki/Early_effect more about this.

Does the input impedance of the detector Q3-Q5 change when adjusting the regeneration level of the Q-Multiplier Q1-Q2? If that's true, regeneration backlash will occur.

To avoid any bad interaction (fringe howl/squegging, excessive threshold noice, regeneration backlash) between detector and Q-Multiplier stage, a cascode or a differential pair RF stage between tank circuit and detector must be used. Shifting imput impedances of the AM (voltage doubling diode) detector circuit with signal envelope/regeneration level are not reflected back to the RF tank circuit in such RF amplifier stages. In any cascode or differential pair amplifier stage the output is isolated from the input by using a common base stage.

@Sean O'Connor: The tricky question is how to link a proper AM detector to the circuit. What ususally happens is AM detector input impedance shifts with signal evelope disturbe the delicate balance of forces that make the circuit a good Q multiplier.Taking a signal from the emitters is generally safe as that is at low impedance. However low impedance means low signal voltage. Taking a signal from the tuned circuit to feed into an AM detector requires great care about how the AM detector is designed.Taking a signal from the emitters is generally safe as that is at low impedance. However low impedance means low signal voltage. You can also connect an RF interim amplifier between the emitters of the MC1648-based differential pair regenerative stage and a voltage doubling diode detector. See circuit diagram below. Best pratice is using a cascode or a Long Tailed Pair RF amplifier to isolate the diode detector from the tank circuit in any regenerative receiver.

You can transform this receiver into a small superhet by removing the source decoupling resistor of 10 nF, injecting the LO signal from a Local Oscillator onto the source of the FET and replace the Drain resistor of 1 k by an LC circuit tuned at an IF of 455 kHz. Another option is to connect a Q-Multiplier stage to the areal tank circuit, transforming this TRF receiver into a regenerative one.

@QRP-Gaijin: 1. Connecting a BJT detector (non-linearly biased amplifier) to the top of a regenerated IF tank has a high risk of causing instabilities, distortion, and noise at critical threshold. Any tiny variation in circuit parameters -- such as variable loading caused by a detector stage -- can affect the regeneration level, and the dynamic effect of such variations might be AF oscillation or excessive noise around critical threshold. 3. Excessive noise at or above critical threshold may be a sign of improper and rapidly-varying loading on the regenerated tank. Connecting a BJT detector onto the top of a regenerative tank circuit can even cause regeneration backlash due to impedance changes of the detector stage when the Q-Multiplier goes into and out of oscillating.

@qrp-gaijin: I'm starting to think that this LO distortion is causing an excess of energy at LO harmonics, which is then mixing with strong local FM signals, frequently causing these strong FM broadcast signals to be audible when I am tuning across shortwave frequencies. This problem is somewhat present in the single-conversion design, and seemed much worse in the double-conversion design (which used two cross-coupled oscillators, so that both LO signals probably had significant harmonic content). This problem with strong local FM broadcast signals is caused by the link coil of the antenna tank circuit. That link coil forms with the stray capacitances of the mixer stage a parasitic tank circuit which is resonant at the FM broadcast band. To avoid this problem, use an emitter follower between antenna tank circuit and mixer stage by which only a two-terminal antenna tank coil without link coils/coil taps is needed. Put a base stopper resistor of 1k between antenna tank circuit and base of emitter follower to avoid the emitter follower acting as an unwanted Colpitts oscillator. Or use a capacitive voltage divider to connect the antenna tank circuit to the base of the mixer stage.

@qrp-gaijin: The regenerative stage cannot be used in weakly-oscillating mode for hunting for weak signals because the AF output in oscillating mode (including the threshold noise) is so high that it overloads the high-gain AF stages. The high-gain AF stages are optimised for the low AF output when the regenerative stage is in non-oscillating, below-threshold mode. Why don't you place a potentiometer between the AF preamp Q13 and AF amp Q7 in order to control the AF level feeding into the AF amp? I see that Q7 acts as an envelope detector by the RF bypass cap C27 of 10 nF. There is also RF at the emitter node of the regenerative stage Q14-Q15.

You can use an emitter follower as a buffer stage between the LO and the emitter of the mixer stage. See https://sites.google.com/site/linuxdigitallab/rf-ham-radio/superheterodyne-am-receiver-use-discrete-transistor about a BJT superhet in which a cross-coupled BJT oscillator with an emitter follower buffer is used between LO tank circuit and the emitter of the mixer stage. This will making the LO tank coil more easily, because you don't need any link coil between LO tank coil and mixer stage anymore.