xpyttl wrote:
. . .
Varactors do have some temperature coefficient, and they are often coupled
with toroid inductors, which also have some considerable temperature
coefficient. Most of the designs you see out there are for CW rigs in the
ham bands, where temperature stability is extremely important. The maze of
capacitors around the varactor are there to balance the temperature
coefficients. Usually there is a polystyrene capacitor which has a
temperature coefficient opposite to the toroid and varactor, but you can
never get exactly the right value for that, so it is a question of getting
the right combination of positive and negative temperature coefficients AND
the right value of capacitance. For AM in the broadcast band, you can
probably come close enough with one or two capacitors. For a rig with a 200
Hz CW filter at 15 meters, it can be a real bear keeping the frequency to
within the 0.0001% that you need for comfortable operation.
. . .

All toroid inductors aren't equal, and neither are capacitors.

I routinely build VFOs with no temperature compensation which have about
200 Hz total warmup drift on 40 meters. The trick is to use components
which have inherently low temperature coefficients rather than try to
make ones with high coefficients compensate each other. Polystyrene
capacitors have a fairly high temperature coefficient, but it's in the
opposite direction than a typical poor inductor. Sometimes people get
lucky and the combination works ok, but often they don't and it doesn't.

The other thing you have to do is design your oscillator so that its
frequency depends almost solely on the tank components and not the
active device.

I found that good quality NPO ceramic capacitors have the lowest
temperature coefficient of any commonly available parts, and inductors
wound on type 6 powdered iron cores were the best. It's the inductor
which dominates the drift in my VFOs, and that small amount can easily
be compensated if desired by replacing part of the tank C with a
capacitor with controlled temperature coefficient.

I described these techniques (except for compensation) in more detail in
"An Optimized QRP Transceiver", in August 1980 QST.

Roy Lewallen, W7EL