I'm using a 7.3728 MHz HC-49 crystal with 22pF disc ceramic load capacitors. I'm doing this on a breadboard, but I've had prior experience with using a breadboard for this and it has worked well, so I'm hoping this is not the problem. On the crystal, the following is written: "ACT Fo 7.3728SCA" and I've tracked the datasheet down to this one. The load capacitance is correct it seems - well within the specified 12pF to 32pF. I'm using with a PIC24FJ64GA002 microcontroller and the microcontroller is set to HSPLL mode, to generate a 29.4912MHz clock. The clock is a baud multiple, and it's very important it's stable. The expected clock is 7.3728 MHz within 50ppm, but the achieved clock is unstable and I can just about measure 20 MHz using my scope, but even that's tricky because it won't trigger well on such an unstable waveform.
Here's the waveform I'm talking about. Notice how unstable it is. Since I tried last night, it changed frequency from 20 MHz down to 10 MHz. What is going on??
My scope only has a limited sample memory of 1K and doesn't sample nearly as fast as a modern one (it does 10 MS/a real time, 10 GS/a equivalent time), but even so I should still be able see a clock frequency.