Finding the point where oscillation is about to start.

I've got an idea for a receiver project, it's not commercial, just as a hobby....

Part of the idea needs a single LC tuned circuit to work at maximum 'Q'. The frequency would be around 150MHz but it has to be tunable say +/- 10MHz using a varactor diode. What I would ideally do is show the tuned frequency on an LCD display but a superhet with local oscillator and PLL is not an option. The idea is to use a microcontroller to produce a PWM signal which would then be filtered to produce a tuning voltage for the varactor, this sets the frequency. A 'Q multiplier' is built around the tuned circuit and under voltage control from the micro, the feedback is increased for just a few cycles to allow it to oscillate so the frequency can be measured. The rest of the time the voltage is set to just below oscillation point so highest Q is achieved. The PWM would be adjusted to maintain the desired tuning based on the frequency that was measured.

Has anyone tried this?

Brian.

I think that this is called "squegging" and is used in the superegenerative receivers, only no one has ever measured the carrier frequency. the "super" as in superegenerative is so called because the squegging frequency is super sonic and only annoys dogs and not the user I think if you were cunning you could use the squegging frequency to gate a suitable timing pulse (say 1 microsec) which opens the frequency counter for counting the HF. Otherwise, as in your suggestion, you would have to be careful about what you were actually counting.
Frank

Thanks Frank.

I can see where the similarity with superregeneration lies but it isn't quite the same thing I had in mind. With superregenative receivers, the LF 'squegging' is done to deliberately start and stop a HF oscillator with the intention that any other signal will preempt the start of oscillation. What I was thinking of was a system that was very close to oscillation all the time but only achieving it under control of a micro, and then only for a few cycles maybe once or twice per second. The tricky part is stabilizing the gain just below oscillation threshold and the only way I can think of to do that is to crank up the gain until oscillation is detected, measure the frequency then back the gain off again until no frequency can be measured again. I had in mind to use a negative resistance circuit as a Q multiplier because it allows smooth feedback control. A conventional oscillator would show a tendency to stay oscillating once it had started until the gain was backed off too much for me to use. It's a tricky problem for which there may be no simple solution.

Brian.

You want to maximize your SNR using narrow band filtering with single stage high Q, yet you did not specify the signal bandwidth.
Normally you can make a circuit almost oscillate with positive feedback and loop gain just <1.
However this does not equate to high Q, so noise can trigger the oscillation and that is what you want to filter out.
So what are the specs?
step 1: define bandwidth and signal level and modulation type.
step 2: define output level of circuit then calculate gain and Q required to exceed input thermal noise.
step 3: step check that Q required does not restrict bandwidth of signal in order to reject noise.

I can make a xtal filter with a Q of 10,000 with no oscillation and an RC filter with feedback that will oscillate.

What makes you think a sub-stable oscillator will filter noise? It just amplifies it according to loop bandwidth. Does that make sense? So you need very high Q which is easier said than done.

Single ended oscillators frequently show hysteresis when it comes to control current or voltage to provoke oscillation and to stop it. If you make a long tail pair, you can avoid this.

The two transistors (emitters connected) have a symmetrical transfer function wtih decreasing gain with increasing amplitude (soft clipping). You can control the gain with the current source transistor.

Because of the constant current and symmetrical soft clipping behavior, you can have stable weak oscillation. For determining the frequency of the circuit, you need weak oscillation to avoid that device parameters change significantly resulting in (small) frequency shift between Q enhanced resonating frequency and oscillation frequency.

I assume you will also use at least two varactors to get symmetrical behavior and reduce distortion.

Brian, perhaps some thing to cogitate on: the voltage across a tuned circuit decays by 1/Q per cycle once the maintaining voltage is removed. So if your high Q LC circuit has a Q of 2000, then after 1000 cycles the voltage across it will have decayed to ~60% of the original!!. So I reckon an extra phase is required, after the oscillations are stopped the Q is lowered to 100?, then raised to the magical 2000 for reception. Perhaps thats what the squegging used to do?
Frank

I would imagine the only way you could get a Q of 2000 with an LC tuned circuit @150MHz is an extremely precise helical resonator with tolerances on winding spacing and angle of 0.01%

But you can get a Q of 200 with some decent Coax and 1/4 wave shorted cct termination resonator, 2000.. hmmm .

@chuckey I agree decay time is 1/Q related as is BW of tuned circuit, and BW of signal has yet to be defined.

I used to try injection locked loops (ILL) for 5Mbps data receivers, they worked but not as well as PLL's with better data filtering. TV's also used ILLs for horizontal and vertical sync.and the user had to sometimes center the free running oscillator because the capture range was also reduced as loop bandwidth was reduced. So these were abandoned.

Thanks for the comments everyone.
It isn't for a commercial project, just some ideas bouncing around in my head (where there's lot of empty space!). I was curious to know if anyone had ever tried to raise the Q of a tuned circuit by holding it just under oscillation threshold and in a situation where it was also tunable. At a single frequency it's easy, but in a situation where the amount of feedback varies according to other parameters it needs some sort of intelligent control. The receiver idea was born from playing with regenerative designs 50 years ago where from the UK I could regularly listen to US and Canadian local broadcasts on MW using a single transistor radio. It had a tuning control of course and a regeneration control which had to be adjusted to just before oscillation for best reception. I considered it might be possible to do this automatically but couldn't see a simple method of determining optimum feedback without letting it oscillate then backing the gain off a little.

Noise within the bandwidth isn't particularly an issue but obviously it could influence the circuit behavior. If I had to specify a bandwidth it would be "just enough for acceptable audio" so something in the region of 3KHz would be about right. It's only an idea, nothing will probably ever be built.

Brian.

so ideal Q= 150MHz/3KHz = 50,000 ... good luck that would take precisely tuned xtal, saw or baw resonators and LC for prefiltering harmonics.

I think from a theoretical aspect what you are proposing is very interesting. I think the real problem is in the construction to keep microphony at bay and its general stability. Once the law relating "just not oscillating" to the tuning frequency have been established then an electronic compensator could be used, driven from the tuning diode volts.
When I wer' a lad, I played with 954 acorn valve VHF super regen RXs, and a few months ago at a general auction there was a lot which included a few 954s, 955s, old rx chassis, cor!, was I tempted but then I thought that I have enough junk (precious electronic components ) at home , so I let it pass.
Frank

Likewise, in years gone by I would have donated a vital organ some of that stuff but I already have enough projects in the pipeline, and junk to build them to last more than my lifetime. I think my first VHF super-regen used a 6J6 so I guess I was quite a latecomer to that kind of technology!

Point taken about the super high Q SunnySkyguy but sometimes you just HAVE to prove the theory wrong!

Brian.