Some considerations :
Note below depends on your design goals for spectral content (if you care), T and V
goals.....so read with one eye closed
1) The bypass cap controls the amount of - fdbk generated from the emitter bias R.
That in turn affects the spectral response/purity of the waveform generated. Also
- fdbk affects oscillator startup time, additionally too much can prevent oscillation.
So one might experiment with that, eg. size of cap. I would start with a reactance of
1/10 of the emitter R at the osc frequency.
2) The coupling cap does not want to have a lot of parasitic L in it, so care here again
matters. You want most of the collector signal delivered to the tank. But there is a tradeoff,
the transistor T and V characteristics get coupled as well so too small a C and osc does not
start, too large and you affect fine T and V osc frequency and amplitude. The Tank Q I would
think would be a good staring point, eg maintain it by not over coupling the transistor
output Z, yet not too little fdbk that would compromise adequate start up of osc. Also
aggravating this is you will typically run the transistor "hot" so it has plenty of G but that
means its Zout drops, so over coupling that onto tank circuit affects stability.
Of course pay attention to type of C you use, at RF typically ceramics, micas. Critical in the actual
tuning circuit, but can affect (via their "reflected C) in coupling applications. But ceramics
in turn have T drift issues so select accordingly. There are variants of the C vs T slope used to
compensate for osc drift in many designs.
There are IEEE papers, back in the 60's and 70's, that discuss a lot of these considerations.
Don't get too brain fried on this, non linear oscillators, which this is to some extent, can soak
up your life juices essence for many years ? The Skeksis will be looking for you.....
Regards, Dana.