Hi,

After some search on forums and other sites, I have come to a conclusion that the oscillator circuit shown in the attached schematic could be a good implementation for robustness against temperature drifts and manufacturing spreads. My goal is that the oscillator circuit should bring out the series-resonance properties of the resonator as clearly as possible, and suppress influence from the rest of the oscillator circuit as much as possible. I hope there is no obstacles such as patents, etc, prohibiting free use of this circuit. If so, I would be highly greatful getting informed.

I have divided the drifts/spreads into two groups, coming from the two circuit parts: the circuit *around* the resonator, and the resonator itself. The resonator can be of any series-resonant type, such as a series LC circuit, a ceramic resonator, or a crystal, etc.

To make the resonator work as cleanly as possible based on its series-resonance properties only, and as little as possible based on its parallel parasitics (incl circuit parallel parasitics), I have looked for oscillator circuits that work at (or very close to) zero-phase operation between the ac voltage and current in the resonator. The circuit driving the resonator can simply be a fast in-phase transimpedance amplifier, giving up to a couple of degrees of phase lag, where the resonator ac current is the amplifier input current. This amplifier must not give a phase lag e.g. in the order of 90 degrees, which then is phase compensated back to about zero degrees before its output voltage is fed back to the resonator, because (non-matched) drifts of e.g. 10% of 90 degrees, giving in the order of 10 degrees resonator i-v phase error, could cause a significant oscillation-frequency shift, whereas, 10% of 1 degree is likely to give negligible frequency shift. Furthermore, I believe that this oscillator circuit has the potential to utilize the inherent resonator Q value very well, since the only terminal that the resonator is connected to is a low-impedance emitter, which will degrade the total circuit Q value negligibly.

So, if you have any kind of comments on this circuit or oscillator reasoning in general, or happen to know even better solutions to the task of designing high precision/stable oscillators, and can consider to share them, I'd be highly greatful!

Regards,
Carl F

Well, the intention is to select R10 high enough to be an open circuit (e.g. 1Mohm), and to select C10 high enough to be a short circuit (e.g. 100nF), at the oscillation frequency. The only purpose of C10 and R10 is to ensure zero DC voltage across the crystal, since this oscillator circuit is supposed to deliver the specified, guaranteed performance of the crystal (which is not guaranteed if the crystal is strongly DC biased). If an LC-tank resonator is used C10 and R10 can be omitted.

Note that almost all crystal oscillators work in the inductive region of the crystal impedance characteristic, such as basically all one-transistor oscillators and the classical inverter-based oscillator. Even with a crystal ESR below 100 ohm, its (inductive) impedance can very well rise to the order of 1kohm or even significantly higher, in that "parallel-resonance" mode (often using parallel tuning capacitances to ground forming a pi-network with the crystal).

However, the two-transistor oscillator discussed here makes the crystal work right on its series resonance (v_ac and i_ac are in-phase), so the crystal total impedance is close to its ESR (often 10-100 ohm). A common maximum current drive level for crystals in the range 1-100 MHz is in the order of 1 mA_ac, which gives about 10-100 mV_ac across the crystal, and thus a power loss in the crystal of about 10-100 uW. Therefore, if R10 is going to affect the Q value of this resonance circuit significantly, it needs to dissipate in the same order of power or more (the same power dissipation would roughly halve the Q value, which would happen at R10=~ESR_crystal). To set R10 to that low a value would create other problems, such as that the base drive of Q1 might not be able to drive a strong enough signal into another 10-100 ohm resistive load, in parallel with the crystal, to maintain any oscillation at all. Anyhow, this would not be of much interest to figure out, since R10 is set to in the order of 1 Mohm in practice. An observation, in any case, is that this oscillator topology is highly robust (oscillation-frequency stable) against fairly large parallel parasitics in the crystal node (Q1 emitter node).

Best Regards,
Carl F