Some people regard the design of RF Oscillators to be something akin to a "black art" and after many years of swearing at "cranky" oscillators I'm not all too sure they are all that wrong. I suggest you ensure you remember this old saying:
"Amplifiers oscillate and oscillators amplify" - unknown
When I was a kid, yes I can remember back to the late 1940's, we collected all manner of junk. Cool was anything remotely electrical and, of course bicycle dynamos, lamps or motors were even "extra cool".
We as precious little seven year olds conceived - all budding nuclear physicists that we were - of this real smart idea, obviously nobody had ever thought of this before.
"Why don't we connect a motor to a generator, so the motor drives the generator, providing electricity for the motor, which continues to drive the generator and it'll go on, and on, and on for a hundred years and we'll become world famous!"
Of course we had no concept of frictional losses (I think that's right) way back then. Nor had the words "perpetual motion" passed our ears.
The whole point of that little story is to crudely demonstrate the principle of how an oscillator works. If you can follow that childishly naive concept then you will kill them in this.
Principles of Oscillator operation
Every oscillator has at least one active device (smarties don't complicate matters for me - just read on) be it a transistor or even the old valve. This active device and, for this tutorial we'll stick to the humble transistor, acts as an amplifier. There is nothing flash about that. For this first part of the discussion we will confine ourselves to LC Oscillators and I'll keep the maths to an absolute minimum.
At turn on, when power is first applied, random noise is generated within our active device and then amplified. This noise is fed back positively through frequency selective circuits to the input where it is amplified again and so on, a bit like my childhood project. Ultimately a state of equilibrium is reached where the losses in the circuit are made good by consuming power from the power supply and the frequency of oscillation is determined by the external components, be they inductors and capacitors (L.C.) or a crystal. The amount of positive feedback to sustain oscillation is also determined by external components.
Frequency or Phase Stability
Frequency or phase stability of an oscillator is customarily considered in the long term stability case where frequency changes are measured over minutes, hours, days even years. Of interest here are the effects of the components changes, with ambient conditions, on the frequency of oscillation. These might be caused by changes in the input voltage, variations in temperature, humidity and ageing of our components.
Never underestimate the effects of these variations on the frequency of operation. I've gone nuts working on so called precision designs, with precision components, where the frequency wandered at random over several kilohertz over several minutes. Needless to say I'd "messed up".
Short term stability is also of great interest and, again I could lay some real heavy maths on you but I won't. I'll simply say it can be mathematically proven that the higher the circuit Q, the higher this stability factor becomes. The higher the circuit Q, the better the ability the tuned circuit can filter out undesired harmonics AND noise.
Reducing Phase Noise
1. Maximize the Qu of the resonator.
2. Maximize reactive energy by means of a high RF voltage across the resonator. Use a low LC ratio.
3. Avoid device saturation and try to use anti parallel (back to back) tuning diodes.
4. Choose your active device with the lowest NF.
5. Choose a device with low flicker noise, this can be reduced by RF feedback. A bipolar transistor with an unby-passed emitter resistor of 10 to 30 ohms can improve flicker noise by as much as 40 dB.
6. The output circuits should be isolated from the oscillator circuit and take as little power as possible.
Effects of ambient changes on stability
A frequency change of a few tens of hertz back and forth over a couple of minutes would mean nothing to an entertainment receiver designed for the FM Radio band. Such a drift in an otherwise contest grade receiver designed to receive CW (morse code) would be intolerable. It's a question of relativity.
Minimizing Frequency drift
These are random and not in any particular order.
1. Isolate the oscillator from succeeding stages with a well designed buffer stage followed by a stage of amplification. Large signals can often then be reduced by a 3 or 6 dB attenuator which also has the benefit of presenting a well defined load impedance to the amplifier. If the stage is feeding a mixer, as is most often the case, then another benefit is the mixer (you are using double balanced mixers?), also see a source impedance of 50 ohms.
2. Ensure the mechanical stability of your oscillator is such that mechanical vibration can have no effect on components, especially those frequency determining components.
3. Supply the oscillator with a clean well regulated supply. If using varactor tuning, doubly ensure the tuning DC voltage is as clean as possible, a few hundred micro volts of noise can be imposed on the oscillator signal. Use back to back diodes for the variable element. Air variables are hard to come by although they offer far superior Q figures. DC tuning tends to be more versatile.
4. Minimize circuit changes from ambient variations by using NPO capacitors, polystyrene are dearer but excellent, silvered mica in my opinion are not what many people believe and are highly over rated.
5. The inductor should be air wound on a coil form with a configuration to maximize Qu. If you must use a toroid, where possible try to use the 6 type as it offers the best Q. Sometimes, for other reasons you might have to use a slug tuned form.
6. Parallel a number of smaller value NPO capacitors rather than using one large one in frequency determining components. For trimmers try and use an air variable. Keep an eye out for small value N750, N1500 capacitors, < 15 pF, when available and are found to be dirt cheap. These are sometimes useful in taming drift in an oscillator.
7. Bipolar or FETS for active device seems to be a matter of personal preference and I've seen some ferocious arguments over that one. Consensus seems to come down in favour of FETS. Me, I'm a bipolar man because FETS hate me pure and simple.
So there are some of the things to keep in mind. Hopefully at this point you have discovered the broad idea of an oscillator, I've outlined broadly two types Hartley and Colpitts. I spoken about frequency stability and listed ways to combat phase noise and reducing frequency drift. Now let's proceed to the main course.