Bert,

Yeah I should give that a try, must revisit some RF stuff soon!

Regards,

Alan

Perhaps you can use a ceramic filter instead of crystal for wider bandwidth?

If this crystal-controlled super-regen circuit can demodulate audio at HF, does that mean it can transmit voice (presumably FM) as well? Might make for a simple stable voice transceiver (like a walkie-talkie). With self-quenching it might be reduced to a single transistor.

I look forward to hearing about how the topology works for you.

qrp-gaijin,

Indeed a very interesting circuit!

I agree. I suspect the crystal is operating at its low-impedance series resonance to provide the emitter-collector feedback with the collector tank being the main resonator. This should sharpen the bandwidth over which is has significant power gain.

It is possible to allow the circuit to oscillate continuously I suppose, chopping the gain on and off and watching the rate of gain in oscillation amplitude when it is on. The high-Q crystal resonators seem to just keep ringing between high-frequency quench cycles in my experience, the gain drops enormously as raise regeneration frequency. If the oscillation isn't fully quenched the forcing signal you wish to measure is quite tiny compared to the remaining previous cycle's energy sloshing around in the resonator. I am unsure if you can force-quench a mechanical resonator like a crystal effectively? With LC resonators you can kill their Q and dump out the energy pretty quickly.

I'll have to build this circuit and see how it operates. Thanks for showing it to me.

Regards,

Alan

Hi,

You wrote, "It is often said that you can not super-regenerate crystal oscillators. In fact you can, but because of their high Q the quench frequencies needed are too low to carry high bandwidth signals like voice."

What do you think of this crystal-"controlled" super-regen, which apparently can demodulate audio at HF?

http://www.seekic.com/forum/22_circuit_diagram/26084_CRYSTAL_CONTROLLED_SUPERREGENERATIVE.html

I say crystal-"controlled" with "controlled" in quotes, because the circuit has an LC tank which is I suspect the resonator, while the crystal serves I believe only to limit feedback energy to the crystal frequency. An interesting topology if I am understanding it properly.

Glen,

The simple and robust nature of the Pierce oscillator is what makes this particular circuit practical. It is much harder to do the same with an LC oscillator. Simply replacing the crystal with a LC Pi network doesn't lend itself to easy replication.

I have found it easier to use two IO pins to control LC versions, typically keying the bias and detecting the signal envelope with different pins. You still need to discharge the capacitor in the envelope detector, so it must be an IO capable pin... Of course you can add extra circuitry to enable the use of just one control line or add a diode to discharge the cap via the keyline, etc.

Regards,

Alan

Had tried a super-regen-to-MCU interface awhile back. Was a LC-based superregen. Instead of a diode amplitude detector, tried a more bare-bones approach where the RF was attached directly to the MCU I/O pin. When the RF grew high enough, it would trigger an interrupt on the peak of a cycle. Then the I/O was switched to "output" to quench the oscillator. Found that releasing the quench line (switching back to "input") introduced enough transient energy to excite the LC, reducing sensitivity greatly.

Seems your circuit overcomes this problem.

Hi Alan,this is good work.

I use 11.0592 Mhz crystal in Rx & TX ckt.

On Rx side there is no change in rectified Vout,but few mV(1-3)change at collecor voltage even when both RX-TX very very close.

Plese can you help me to work this ckt.

Thinking about how this works a little more - as you say the receiver essentially samples the received energy just before oscillation starts.

Once oscillations start to build up the received signal no longer has much impact.

I think this is equivalent to sampling the baseband bit stream with a noise bandwidth equal to the receiver bandwidth.

For an ideal demod we would like to filter at approximately the bit rate, for example integrating over the bit period or using a matched filter. For example averaging over an entire bit would reduce the effect of an impulse.

So if your bit rate is 10Hz and bandwidth is 400 Hz, thats 40x (I think) more noise power than an ideal demod. This would have a large effect on the Bit Error Rate (BER) versus Rx power curve.

Using a bit rate at something approaching the receiver bandwidth and sampling once per bit might give better overall results (the extra bits could be used for error correction). This would require single sample/bit synchronization schemes.

Cheers,

David

Fascinating work.

I am planning a HF messaging system (SMS/small email size messages) based on HF mesh networking. It's for people in the developing world so each terminal has to be cheap. It doesn't get much better than this sort of design!

I was wondering if these receivers would also be useful for PSK or FSK? Rather than amplitude detection the uC input could be used as a one bit A/D to sample the frequency and phase of any mixing products. DFTs actually work quite well with low amplitude resolution, and at these bandwidths can be implemented on a uC.

The way the RX signal builds up and is sampled reminds me a lot of "integrate and dump" coherent demodulation.

I of course assuming the super-regen process preserves frequency and phase at the RX.

Cheers,

David, VK5DGR