Do crystal oscillators synchronize by themselves when coupled by supply voltage?

Question

I watched an excellent video about The Surprising Secret of Synchronization

There are examples of pendulum clocks or metronomes synchronizing when coupled by a swinging platform. Other examples are flashing fireflies in a tree or an applauding audience in a theatre.

What about electronic crystal oscillators? Do they synchronize when powered by the same power source with a high internal resistance and only a small decoupling capacitor?

I am thinking of crystal oscillators of the same type and the same nominal frequency but with the usual small deviations of some 10 to 100 ppm.

^{simulate this circuit – Schematic created using CircuitLab}

P.s. Of course this is nothing that a circuit designer should do. If several 10 MHz clocks are needed, they should be derived from the same master clock.

Answer 1

They can do.

When you want oscillators to lock to others like this, it's called 'injection locking', and is a useful but little-known technique. You would apply a bit more signal for injection than just running them side by side though. It's perhaps most frequently used with harmonic or sub-harmonic locking to make a very low phase noise multiplier or divider.

Commercial packaged crystal oscillators will usually have sufficient internal power supply decoupling so that any interference of one oscillator on another will not cause such a level of pulling that the two would be pulled into synchronisation.

However, if you are building a low noise synthesiser, you can guarantee that the effect of one oscillator on another will be visible as frequency modulation. It usually takes some serious effort with decoupling, screening and buffering to get the influence below the -100dB level that separates a high quality synthesiser from a smelly heap of circuits.

Answer 2

Seen it happen. Back in the days when multi-megabit per second digital links were exotica, I tried using an early implementation to connect a scientific instrument to a computer. It would only work when the instrument was right next to the computer. After much debugging, I found that it only worked if the transmitter's and receiver's crystal oscillators were phase-locked. It was supposed to work asynchronously, but it didn't.

Didn't even need a common power supply to get enough coupling.

Answer 3

I've never tried injection locking, per Neil_UKs answer, but it occurs to me that you are taking entirely the wrong approach here, assuming you're not thinking of doing this just to see if it can be done.

Why have, for instance, 4 oscillators which you want to be synchronized, rather than 1 oscillator and a single hex buffer chip? The buffer will be far, far cheaper than 3 oscillators, both in terms of cost and board real estate.

If, for instance, you claim that you want the oscillators spread out across a large board so you don't need the clock signal running all over the place, you should be aware that, for your approach, the injection locking signal will be a grossly unbypassed power supply which will contain a large clock component. This will affect all the other chips, and probably worse than long clock lines.