I'm experimenting with a 32.768kHz oscillator using a watch crystal (no specs) and a CMOS inverter. The current limiting resistor is 330k and the caps are both 15pF. It oscillates with
4.7-10Meg feedback resistor but not with 3.3M. I increased the Rf in steps from 3.3M and it reluctantly starts up with 3.8M, taking about 2 seconds to build up to full amplitude. Do you think 4.7M will be enough to ensure oscillation with different samples of the same type of crystal and under different environmental conditions? The power supply is regulated.(I have both practical considerations and an academic interest for wanting to know if 4.7M is enough).