High Performance 2 transistor TRF
The goal I set for myself was to create, modify, or discover a high-quality TRF (Tuned Radio Frequency) radio circuit that could perform at least as well as the MK484 - TA7642 AM radio chip, while using as few parts as possible. My aim was to learn about the constraints of early tube designs from the twenties (1920s).
Back then, designing a TRF radio presented numerous challenges. The tubes of that era had limited gain, necessitating the use of multiple stages, which, in turn, led to additional issues requiring filters and workarounds. Moreover, these designs lacked sensitivity, paving the way for Regenerative Radios and the Superhets. Tube gains hovered around 10, whereas modern transistors, like the 2n3904, range from 200 to 800. However, even with this improvement, the gain remained insufficient without regeneration (superheterodyne).
I tried all sorts of designs, using FET's, Darlingtons, etc. They were all pretty awful.
This led me to think of the existence of a transistor with a gain of 5000 to 10000, a concept nonexistent in the 20s. Eventually, I stumbled upon a straightforward design employing multiple MPSA13 transistors (Darlington). I streamlined it to the detector stage, it worked really well! Adding a preamp stage on the output and connecting it to my headphone amp yielded good sound quality. By adjusting the voltage to the collector, you can enhance the gain, and it even performs OK up to about 17 MHz (Needs an antenna of a few feet at least). The preamp section using the 2N3904 is a great audio preamp design you can use anywhere.
Utilizing a ferrite core antenna wound with litz wire, I achieved satisfactory results on the AM broadcast band without an antenna, but you can use anything at hand. I am able to pick up AM Broadcast stations 600 miles away. It will even work great with an aircore winding on a 1 inch diameter core of plastic or cradboard, but you will need an antenna of some sort, even a 2 foot length of wire. Earth ground is not required if using a ferrite antenna core
I'm planning to explore the MPSA14 Darlington, which boasts twice the gain of the MPSA13, to further enhance sensitivity.
Despite its success, there are a few drawbacks:
1. Driving it too hard with higher voltage can induce oscillations.
2. The extreme high selectivity makes it challenging to use poly variable capacitors effectively.
3. It whistles and oscillates around a station until it is tuned. Going forward, maybe I can figure out an AGC circuit.
This is an easy project to build. It Works.
Sure, here's the link to the February, 1933 QST paper by Robinson (W3LW) on optimizing regenerative (tube) detectors for sensitivity, complete with measured data: https://archive.org/details/sim_qst_1933-02_17_2/page/26/mode/1up . It might be interesting to try to reproduce his results with modern transistors and measurement equipment like a NanoVNA.
In reviewing that article again, I see that Robinson does indeed assert that higher L/C ratios lead to higher regenerative amplification. But again, this is in the context of tube detectors, which have very high input impedance. It might be different for BJTs; it might be similar for JFETs and MOSFETs -- I don't know.
He also has a lot of useful data like a measured maximum amplification of 7000 for 30% modulated reception. It seems that this "amplification" is is the ratio of the AF voltage out (Eb in his diagram), over the RF voltage in (Ea in his diagram). Therefore, this "amplification" represents the end result of two processes: (1) the regenerative amplification of the RF signal in the tank and (2) the conversion/detection of that RF signal to give an AF output. Therefore, this "7000" amplification figure should not be interpreted as the Q of the tank. However, Robinson also provided measured selectivity curves for his experiments, which allow computing the regenerated tank Q as frequency/bandwidth.
I did some LTspice simulations some years ago (reference: https://qrp-gaijin.blogspot.com/2016/12/spice-simulations-of-regenerative.html) where the simulator results showed that a 1.2-volt regenerative detector (BJT) gave a maximum voltage gain of about 100 or 20 dB. My results were:
My results (AF out / RF in = 100 at 100% modulation) vs. Robinson's results (AF out / RF in = 7000 at 30% modulation) show that my detector could still be optimized, though my choice of a low supply voltage (1.2 volts) and a BJT may be the limiting factors. But, in this kind of case where there are strange constraints on the circuit, it's probably best -- as you said -- to simply separate the Q-multiplier and the detector into separate active devices, and simply to optimize the bias of each separately -- bias the Q multiplier for easy and smooth regeneration, and bias the detector for optimum sensitivity. I sometimes still try to minimize transistor count in my circuits as a fun optimization target, but for actual ease of construction with fewest problems, I've come to accept that spreading out the functionality over more transistors is actually simpler.
If one were sufficiently motivated (and I am not, at the moment 😀), it should be possible to download vacuum tube models for LTspice, to create a regenerative detector simulation in LTspice using those tube models, and to reproduce Robinson's results in LTspice. Interestingly, if I'm interpreting Robinson's results correctly, he indicated that a triode detector in his experiments gave only about 33% gain compared to screen-grid tubes. I wonder why the screen-grid tubes performed better than the triode in his experiment. And since modern transistors are basically triodes, I wonder if they also all perform as poorly as triode tubes, giving only 33% AF output compared to the screen-grid tubes.
And here we have a YouTube video where one gentleman is analyzing the performance of different tubes in a regenerative receiver.
Thanks to Jim Kearman, KR1S, from the old TheRadioBoard forums for pointing out the 1933 Robinson article to me. By the way, does anyone know what happened to Jim after TheRadioBoard went down?