Amplitude noise in LC oscillators

[Teleno]

Not much info on this subject to be found in Google, so I decided to pose the question to you guys.

I want to be able to detect tiny variations in the amplitude (envelope) of an LC oscillator.

My 90KHz oscillator produces a 40Vpp sine wave. Before applying the peak detector I subtract 37V DC (using a resistor and a reference current) so the peak is 3V maximum. This is like "zooming in" into the peak, which allows to detect small amplitude variations of the order of a millivolts in the 40V range.

You would assume that the amplitude of the sine wave would be constant if the power supply is well regulated. Well, this is not the case, instead, the amplitude behaves like a "random walk" (Brownian motion) drifting up and down in what appears to be huge "pink noise" also known as "flicker noise" or "1/f" noise. In this case the drift is in the order of the 100's of millivolts, so large it buries the signal.

Google taught me that flicker noise is the integral of white noise, and since the LC circuit is in fact an integrator, any noise injected by the power supply or the components of the oscillator will result in 1/f noise.

I've tried to remedy this by using a battery as a power supply + big capacitor, since regulators are noisy. The drift is a bit less but still remains.

Any ideas on how to minimize this annoying effect?

Attached: LTSpice simulation. I've added the white noise source BV to show what the effect looks like (plot V(peak)) as it happens in the real circuit without BV.

[waltr]

Have you looked at the Spectrum of the oscillator?
You may also have 'phase noise' which is small variation of the frequency which can show as amplitude variations.

Read about the 'Q' of oscillators - can depend of L to C ratios and the quality of the parts as well as physical lay-out and mounting.

Some of the noise may simply be Resistor Thermal (Boltzman's constant) noise.

Last, it may be possible to run the oscillator through a High Q filter but suspect not.

[Teleno]

Well waltr, what I've read (mostly related to microwave CMOS integrated oscillators) is that phase noise does translate into amplitude noise, and that it can be remedied by a high Q.

The problem is that high Q requires a higher inductance, which in turn - for the same power - causes a higher output voltage, which in turn amplifies the voltage noise.

Another solution is to use a pulse-driven oscillator such as Colpitts. It appears that the shorter the bias pulses, the shorter the time that noise from the active components gets integrated by the LC tank, resulting in less overall noise. The problem with Colpitts is inefficiency, in order to achieve a peak current of 300mA anf 40Vpp in the tank - which is my target - I would have to waste a lot of watts.

There might be some pulse-driven LC oscillator designs out there that are efficient half-bridges, C-class or push-pull, but I wasn't able to find any. Perhaps someone can provide a pointer?

[Polymer]

Hello Teleno,

Have you tried the Metasearchengine http://www.dogpile.com/
I find better results there than in Google for technical stuff.

I have a 1-3Hz "flickering" noise problem on the PI's and interested in this topic myself.

[Skippy]

A few ideas worth considering:

Try a parallel matched-pair of transistors in place of the single device. Try a BC337 and it's complement in place of the 557's. Change the 4148 diodes to higher current parts, I'm not sure what to suggest, UF4002 ( the fast equiv of the 1N4002) might be a first try.

Injection-locking the oscillator to a quartz-derived reference would probably be one way of making it less jittery.

[Davor]

I don't have my reference library at the moment, so I can't give you a precise reference, but from my memory ...
The vast majority of references on oscillators deal with phase noise, not amplitude. That's because the vast majority of oscillators are in fact limited by power rails, and the amplitude noise is mostly defined with power rail noise.
However, it is not the case with your oscillator, which is limited by current.
1/f phase noise in oscillators is mostly due to the off resonance circuit being fed by a transistor that supplies flicker noise amplitude and phase modulated drive. Because of the amplitude limitation provided by the rails, it manifests itself as a phase noise only. If I remember correctly, the most cited author on this is Thomas Lee. So far so good. Now, the only solution to phase noise beside phase locking is emitter degeneration, which introduces some negative feedback in the source of the problem. The first reference to that, if I remember correctly, was an article by U.L. Rohde, back around 2000.

Because your oscillator's amplitude is not tethered to the power rail, and I warned you about it before, you have only two options:
- introduce some kind of limiter, say, a graetz bridge driving a zener
- emitter degeneration in driving transistors

Because you expect your oscillator to respond to targets by amplitude variation, your only option is introducing emitter degeneration.

[Teleno]

Because you expect your oscillator to respond to targets by amplitude variation, your only option is introducing emitter degeneration.

Yeah I thought about that, will try it. In any case the resistors have to be rather

Do you think that driving the transistors with a PLL would improve the amplitude noise or worsen it because of the phase jiter of the PLL?

[Davor]

PLL will do nothing of a kind. The noise generated by active components is simultaneously in phase (amplitude noise) and quadrature (phase noise), so even if you quiet the phase noise by applying a PLL, you'll still have unabated amplitude noise.

Introducing emitter degeneration at, 10 ohm should help a lot, but feel free to forget about amplitude stability. In fact, you'd lose sensitivity if you insist too much on amplitude stability. If you don't intend to use it in static operation, you could introduce a HPF after amplitude detector, and eliminate the most of the random walk. This would increase sensitivity, but you'll lose static operation. You may also introduce some kind of variable SAT, and benefit the both worlds - motion and quasi-static.

[Teleno]

If you don't intend to use it in static operation, you could introduce a HPF after amplitude detector, and eliminate the most of the random walk. This would increase sensitivity, but you'll lose static operation. You may also introduce some kind of variable SAT, and benefit the both worlds - motion and quasi-static.

I'm doing band-pass filtering now and sensitivity is good.

What does "SAT" stand for? I can't find any references in Google.

By the way, I've shielded the coil to GND and the noise went substantially down. Now I've got a new problem, the level of white noise is too low for oversampling and decimating the 10 bit ADC. The 13 bit data shows discrete jumps on a 10 bit scale. Will have to add about 7mV of noise.

[Qiaozhi]

What does "SAT" stand for? I can't find any references in Google.

Self-Adjusting Threshold

By the way, I've shielded the coil to GND and the noise went substantially down. Now I've got a new problem, the level of white noise is too low for oversampling and decimating the 10 bit ADC. The 13 bit data shows discrete jumps on a 10 bit scale. Will have to add about 7mV of noise.

It's always seems a little ironic to me how oversampling and decimation requires the signal to be noisy in order to increase resolution.

[Teleno]

It's always seems a little ironic to me how oversampling and decimation requires the signal to be noisy in order to increase resolution.

Random noise is the simplest way to shift the signal along the resolution threshold. It comes with a noise penalty though.

A better but more complicated approach would be to add a triangle wave to the signal, having a 1 bit amplitude and a period equal to the oversampling period. But this would defeat the purpose of oversampling which is to have a simple ADC converter. See https://electronics.stackexchange.co...google_rich_qa

[Davor]

I take back what I said about PLL not helping.
When oscillator is built with a PLL driving transistors in switching regime, the 1/f noise mostly disappears because transistors are not in a noisy linear regime. I made a circuit that seem to be quite stable, which runs a Musketeer coil (spice model) at over 80V. It draws 400mA as well.
I'll post the schematic tomorrow.

[Teleno]

I take back what I said about PLL not helping.
When oscillator is built with a PLL driving transistors in switching regime, the 1/f noise mostly disappears because transistors are not in a noisy linear regime. I made a circuit that seem to be quite stable, which runs a Musketeer coil (spice model) at over 80V. It draws 400mA as well.
I'll post the schematic tomorrow.

So you actually build the circuit rather than a simulation?

I'm looking forward to seeing it!

[Davor]

You'll have to be satisfied with a simulation for now. I had a few PLL oscillator builds that proved to work identical to the simulation, so I trust this one will work well just the same.

The thing I invented in a process is something I call "anti-Colpitts", where a capacitive divider is used instead of a series capacitor. A few things happen in a process.
- the tank oscillates with almost full natural Q-factor, and the losses are small;
- the amplitude is somewhat controllable without a series resistor;
- there is a current peak related with the switchover, and it resides at the top and bottom of a coil voltage, but its position can be adjusted;
- high Q tanks will produce sinus of high purity, and with little current consumption.

Because the switchover travels right of the sinus maximum in your oscillator due to the lag in driver transistors, I identified it as the main cause of phase and amplitude noise. For optimum results the switchover should be placed precisely on sinus maximum. I can achieve it by lag compensation using a PLL as a tank driver.
The PLL locks to the phase defined by coil voltage zero crossing. Because this phase is 90° off, the PLL's VCO generates 4x the frequency, and 90° phase for a driver is generated by 2 flip-flops. Adjusting delay of a VCO feedback compensates the delay generated by driver circuitry, and switchover happens at precisely sinus maximum.

Driver losses effects, and the crossover, are diminished by the tank capacitive divider, and because the tank Q-factor is marginally affected by this oscillator, the signal purity is surprisingly good.

Ask whatever you need.

[Teleno]

Thank you Davor. I was ale to run the simulation. The current ad voltage at the tank are very much in accordance with my needs.

However, the ESR should be added to capacitors C3 and C4 to get a more realistic simulation.

By the way, the original circuit (first post) can also be adapted to work as "inverse Colpitts" as well.