I want to drive a quartz crystal resonator at its resonant frequency so I need to stay locked on to its resonant frequency as its resonant frequency changes. I'm using an FPGA to do this.

I want to use a PLL to stay locked on to the resonant frequency. My quandary is, how do I capture and compare the phase shift that occurs when the resonant frequency changes?

Suppose my setup is like this: enter image description here

A normal PLL loop looks like this: enter image description here

I would assume that in the typical PLL loop, my "VCO" is the DDS (since the frequency output of the DDS is determined by a constant value input just like a DC signal controls an AC output of a VCO) and the "reference signal" would be the response signal of the resonator (ie. the desired frequency).

Now suppose at time = 0 I know what the resonant frequency is (maybe I found it by doing a frequency sweep and looking for amplitude response). Then, some time later, the resonant frequency changes.

The "new" resonant frequency would cause the "old" frequency to phase shift. But how would I quantify this?

Typical methods involve comparing the signals in the time domain which would mean I would somehow need to save one period of the "old" frequency I'm driving the resonator at and then "overlay" that with the correct period in time with the "new" phase shifted signal such that the phase difference is apparent (they must be synchronized).

My design would look like this:

enter image description here

This seems very convoluted and I believe I'm overcomplicating things.

  • 1
    \$\begingroup\$ Yes you are ... You need to define your resonator transfer function as either series or parallel and simply make it oscillate satisfying the Barkhausen criterion. \$\endgroup\$ Commented Oct 7, 2021 at 3:19
  • \$\begingroup\$ If you are able to detect the resonant frequency during a sweep using peak amplitude, maybe you can just keep sweeping the frequency (slowly) up and down to stay right near the peak as the peak moves. If the amplitude is increasing as you sweep, keep sweeping. If the amplitude is getting smaller, sweep in the other direction. Obviously this would cause it to hunt back and forth near the peak which might not be desireable. And it might not work if the resonant frequency changes very rapidly. \$\endgroup\$
    – user57037
    Commented Oct 7, 2021 at 3:52
  • \$\begingroup\$ Doesn’t Barkhausen stability criterion apply to oscillators? My understanding is that a DC voltage applied to an oscillator alters it’s frequency. In this case, my VCO would be the DDS (not the resonator itself) and it’s “DC” signal would be the controlling value that determines the frequency of the DDS’s outputted signal. Is this analysis correct? \$\endgroup\$
    – Andrew
    Commented Oct 7, 2021 at 3:55
  • 1
    \$\begingroup\$ You have a resonator. Why can't you let it hunt for it's own frequency and then just measure it? A Pierce oscillator is easy to implement even with an FPGA and then you can just measure the frequency. en.wikipedia.org/wiki/Pierce_oscillator \$\endgroup\$ Commented Oct 7, 2021 at 5:46
  • 1
    \$\begingroup\$ Andrew, you are the one asking for help. People asking for help are the ones who have to go out of their way to be polite and accommodating. No one is required to help you. That having been said: in all of this it is STILL not clear what you are trying to actually accomplish. Resonators are generally part of a larger system and the fact that you are refusing to explain the larger system is frustrating to the people trying to help you. \$\endgroup\$ Commented Oct 13, 2021 at 8:17

3 Answers 3


You don't need a PLL, rather you need a series and parallel oscillator and, you must choose which mode is preferable.

It could also be an Injection Locked Loop which could be an offset biased and forced oscillation with mutual coupling, and is a common phenomena of nature.

Your resonant frequency will change with every environmental variable from temperature, pressure, vibration, shock, and equivalent parasitic capacitance and equivalent inductance ESL.

When you make an oscillator, you can easily measure f, and tune it with a known C. It can be pulled or drift within in the BW determined by Q and possible a harmonic with a known capacitance and gain to satisfy the Barkhausen criterion of >=1 loop gain at 0 Deg.

enter image description here

The above is a simulation of the impedance ratio of a 1kHz resonator similar to any Piezo or Xtal resonators. fs= 969 Hz, fp= 1.02 kHz . By adding reactance, one could tune the frequency to exactly 1.000 kHz, if you wanted or make it into an oscillator.

  • Higher voltage low current, or high current low voltage where current is proportional to force like ultrasonic exciters.

  • Series mode is in phase or 0 deg. and parallel mode is anti-phase or 180 deg., and thus requires an inverter to oscillate with negative feedback for DC as well to self-bias.

  • Series mode requires a low source impedance to drive it.

  • The Qs in series mode is \$Q_s= \dfrac{X_{(\omega_0)}}{R~~~~~}= \dfrac{\omega_0 L}{R} = \dfrac{1}{\omega_0 RC}\$

  • The Qp in parallel mode is the inverse, \$Q_p=\dfrac{R~~~~}{X_{(\omega_{_0)}}}=\dfrac{R}{\omega_0 L} = {\omega_0 RC}\$

  • also damping factor \$\zeta=\dfrac{1}{2Q}\$

Q is relevant to all harmonic and oscillatory behaviours from an electron to a gyrotron to a piezo crystal to a structural resonance to an RF induction heater.

Q is also defined by the impedance ratios to determine the resonant frequency to the half-power full bandwidth, BW ratio.

How you choose to oscillate your high Q resonant material depends on your expectations and assumptions or to be defined, (TBD) "SPECS"

i.e. Define your specs then a simple solution is possible.

But you don't need a complex PLL, although that is possible, if you have a signal from the resonating signal and a mixer that shifts the phase by 90 deg or 180 deg or 360 deg, by design.

The simplest solution for a sinusoidal Hartley Oscillator using a 6 MHz Series Xtal is as follows. If you need lower harmonic content or higher voltage or lower impedance, then you need a spec. for V, Z, f, tolerances, bias etc.

But if you only need it to oscillate, this is all you need.

enter image description here

Changing the 150 pF ratios can amplify the signal to 10Vpp. I chose std values for a 6 MHz series Xtal.

  • \$\begingroup\$ I will go over your answer tomorrow. You originally said I need a transfer function - if this is true then I can make one using Simulink and experimental data. I don’t entirely understand your answer at first glance. Thanks for your time \$\endgroup\$
    – Andrew
    Commented Oct 9, 2021 at 5:44
  • \$\begingroup\$ It all depends what you want out of this Xtal, Sine square VCXO, TCVXO, PLL VCXO DTXO etc \$\endgroup\$ Commented Oct 9, 2021 at 5:49
  • \$\begingroup\$ You needed to define both input and output characteristics. \$\endgroup\$ Commented Oct 9, 2021 at 5:58
  • \$\begingroup\$ Fundamentally, I want to use this as a sensor to tell me something like what the ambient temperature is or what mass is on the resonator itself based on a shift in resonant frequency (but remained locked onto the ever changing resonance). As to what method is best to do that, I’m not sure. All I know is that amplitude response is insufficient so I’m using phase now (I need a fully fledged vector network analyzer). I don’t care what the shape of the output or driving signal is as long as it works to stay on resonance. I’m sorry for my shallow understanding. I’ll look over your answer right now \$\endgroup\$
    – Andrew
    Commented Oct 9, 2021 at 6:06
  • \$\begingroup\$ Frequency and pressure voltage sensitivity will be greatest with smallest load capacitance which shifts frequency up. Measure frequency with a stable reference then mixing it is the best method if your sensing is linear or as sensitive as you expect. \$\endgroup\$ Commented Oct 9, 2021 at 17:11

Assuming you want to see a large signal amplification in your resonator, you need to load it so that its unloaded Q is dominating the loaded Q. That means that losses in the resonator dominate losses in the drive and sense impedances.

You either need to drive and sense from short circuits, or from open circuits. Which is easier to do will depend on the impedance level, Q and frequency of your resonator, which you haven't disclosed. That's for an LC resonator where you get the choice of configuring it in series or parallel. For a mechanical resonator, you may be stuck with just one choice.

If you choose short circuits, then a voltage output DDS is easy. You'd sense the current at the ADC with either a virtual ground amplifier, or a very small resistor to ground that you sense the voltage across. If you choose open circuits, then a high impedance buffer for the ADC is easy, and you'd need to current drive from your DDS, either making a high impedance active current drive like a Howland, or an open collector, or simply a high value resistor in series with a voltage source.

When a resonator is in resonance, the voltage across and current through are in phase.

So now the PLL is straightforward. Pick a drive frequency. Measure the current phase with respect to drive phase. Adjust the drive frequency up or down, rinse and repeat until you get zero phase.

The response to a change in frequency will be slow, and the higher the Q of the device, the slower it will be. This places an unavoidable lowpass filter in the loop. For loop stability, you need to arrange for the loop bandwidth to be somewhat less than this.



simulate this circuit – Schematic created using CircuitLab

What you call it PID is actually a compensator, PI like. You do measure voltage and current and compare their phases.

But this circuit would be too complex, you could simplify it by making a zero cross detectors for voltage and current.


simulate this circuit

In All Digital PLL (ADPLL) the phase detector measures pulses, not analog values. It's much simpler and robust than using ADC. Further you can add a digital lag to compensate the voltage/current measurement phase difference



simulate this circuit

Note, that you don't even need a DDS and the analog amplifier for resonant circuit, a H-bridge is enough. You can find similar implementations in inductive heating furnaces, resonant Tesla coil drivers. They all have one issue, at the light load it would blow off due to accumulating energy. So there are many techniques to avoid this by skipping some periods or modulating the output with Enhanced Pulse-Density-Modulation (EPDM).

  • \$\begingroup\$ Can you not find a simpler solution? \$\endgroup\$ Commented Oct 7, 2021 at 16:55
  • \$\begingroup\$ @TonyStewartEE75 The solution is based on OPs wishes, Complex wishes, complex solutions. But those are used in large induction heating furnaces. \$\endgroup\$ Commented Oct 7, 2021 at 18:43
  • \$\begingroup\$ I have helped others make a large induction furnace with a simple oscillator and tuned filter to select the fundamental. \$\endgroup\$ Commented Oct 7, 2021 at 19:09
  • \$\begingroup\$ You have made assumptions not mentioned , regarding mode of operation and phase assumptions for max current or max voltage or matched power transfer with conjugate matching without any reference. \$\endgroup\$ Commented Oct 7, 2021 at 23:18
  • \$\begingroup\$ If you consider the OP's question might be XY, then you can consider better solutions. \$\endgroup\$ Commented Oct 8, 2021 at 18:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.