Thursday, September 30, 2010

A New Tunnel Diode Regenerative Receiver

I was excited to find a simple, tunnel diode regenerative short-wave receiver circuit in a book some thirty years ago. Unfortunately, I found the receiver to be practically useless for CW reception. It suffered greatly from intermodulation distortion. The audio distortion was simply awful and to make matters worse, the frequency would injection-lock to any but the weakest of signals. Try as I might, I couldn't improve upon the design. Over the years I have revisited the problem a number of times without any real success. While I've built and operated several reasonably useful tunnel diode direct-conversion receivers, a worthwhile regenerative receiver design has eluded me.

Only, my luck started to change last week, due, no doubt, to an ongoing discussion with DL3PB (thanks Peter!). At last it seems that a simple, well-behaved, tunnel diode regenerative receiver is within my grasp. The drawing shown above indicates the current state of affairs.

On account of their restricted (quasi) linear operating range, I think it's best to keep tunnel diodes out of the signal path as much as possible. This imperative pushed me towards a common regenerative receiver configuration in which the Q-multiplication and product detection functions are performed by separate devices.

My recent experiments began with a negative differential resistance (NDR) driven, LC Q-multiplier. I began by using an adjustable source of NDR. A circuit made of three fixed resistors, a variable resistor and a high-speed op-amp allows one to set the NDR for virtually any desired value. Tunnel diodes, on the other hand, generate what is an essentially fixed-value NDR, depending on the diode specification and bias setting. An adjustable NDR source allows one to simply dial-in, and subsequently measure the value required to drive an LC tank circuit beyond the threshold of sustained oscillation. In this way the NDR required to drive an LC tank, plus the antenna and detector loads can be precisely determined.

Right away I had an old problem reappear. Only moderately strong incoming RF signals were sufficient to pull, synchronize, or injection-lock the oscillator frequency. So long as this happens it's virtually impossible to receive CW or SSB signals. Injection-locking can be seen in phenomenon as diverse as clock and metronome pendulums, fire-flies, cicadas and electronic oscillators. It's a fascinating subject!

Alder's equation tells us that to avoid injection-locking we must: 1) insure our LC tank resonator has a high loaded-Q, and; 2) don't allow the ratio of the oscillator to input RF signal amplitude fall too low.

In order to obtain a high loaded resonator Q we must begin with a high unloaded resonator Q, and then take care that external loading (here; the antenna and detector circuits) is held to a minimum.

Tunnel diode receivers tend to fail the second prerequisite; namely, they work with fairly low self-oscillation amplitudes. A recent experimental regenerative receiver using a back, or backward, diode pointed this out all too clearly. It seemed this minute oscillator signal would frequency-synchronize with nearly every signal to appear within the resonator bandpass!

Conversely, a properly coupled tunnel diode having a higher peak-forward current (Ip) will better resist frequency pulling. I decided that an Ip of 10mA might be high enough to avoid excessive injection-locking, and yet not so high that frequency drifting due to self-heating is an issue. What's more, I happened to have two 1N3718 tunnel diodes in my junkbox. Under typical operating conditions, these 10mA peak-forward current diodes will produce a fixed NDR of 13 Ohms.

I next bread-boarded a 3.5MHz LC resonator along with the coupled antenna load. My adjustable op-amp-based NDR allowed me to determine the number of coupling turns needed in order to drive this loaded resonator into oscillation. I found that a four-turn coupling required an NDR of 369 Ohms. Since my 1N3718 has a fixed NDR of 13 Ohms, it will easily drive this loaded tank into fact, too easily! The resulting oscillation would swing the load-line far beyond the quasi-linear range, thus producing excessive harmonic energy.

I reduced the coupling to 2 turns. This required an NDR of 86 Ohms for the onset of oscillation; a value that's still too high. Finally, a single-turn coupling required an NDR of 26 Ohms. This is high enough to insure reliable oscillator start-up and yet the value is low enough to help insure the signal produced contains little harmonic energy. Replacing the adjustable op-amp NDR source with the 1N3718 tunnel diode produced the desired results. A fairly high spectral-purity oscillation was maintained across the CW portion of the 80m band without having to twiddle a "regeneration" knob...which is a good thing, given that regeneration knobs are hard to come by in a two-terminal oscillator!

As for the results; so far so good! Injection-locking only occurs on strongest of incoming signals. The IMD appears to be no worse than one might expect from a simple, triode or pentode-based regenerative set. The receiver is quiet with low signal distortion. Stronger evening signals can be copied with the headphones lying on the operating table. No SWBCI has been heard in three evenings of operation.

Two nights ago I heard VQ9LA on this setup with a very nice signal. Tonight I copied S59A working VE6WZ; again, with perfectly readable signals. I'll also mention that Mike, WA3SLN, was booming into Vermont this evening with his 75w Heathkit DX-60B and inverted-vee antenna. It's always a pleasure to hear those wonderful rigs of our youth.


  1. I have put a collection of negative dynamic resistance circuits here:
    Any of those circuit could replace the tunnel diode I suppose. However regenerative receivers are very sensitive to the exact gain verses amplitude curve of the amplifier. Ideally the amplifier should have exhibit a slight decrease in gain as the signal increases. An AC coupled differential amplifier with a tanh gain/amplitude curve might be one of the better possibilities.

  2. Dear Hoa Xuan,
    Thank you for your message. Oh yes, a two-terminal NDR function may be realized by any number of circuits based on three-terminal devices. The common "Franklin Oscillator" is but one example.

    Another example is the "transconductance," or "cross-coupled pair." I use it here in a 40m regenerative receiver.

    Also, this two-terminal NDR circuit was used in the old Megasonic tuning-fork wristwatch used a PNP/NPN pair.

    The op-amp circuit (NIC) that I talk about in my post is very well known

    I use a high-speed op-amp and set R2=R3. My R1 is a PCB-type, low inductance, Cermet variable resistor. With a resonator attached to the input port (you must have a DC return to ground) simply adjust R1 until the threshold of oscillation is reached. Then pull out R1 and measure the resistance. That is the value of NDR required for the onset of oscillation. Also, knowing the value of L or C, you can now approximate the Q and the Rp of your tuned circuit.

    I should also mention the value of R2=R3 should be on the order of magnitude as R1 (for example, don't use R2=R3=100k Ohms with R1 ~ 100 Ohms)

    BTW, the NIC alone does not exhibit a decrease in gain for a small increase in oscillation amplitude. It does not make a stable sinusoidal oscillator without the addition of extra non-linear circuity (back-to-back diodes, typically).

    Thanks again and Happy New Year!
    Mike, AA1TJ

  3. This comment has been removed by a blog administrator.