I'm astonished at the tsunami of answers this arose!.
- Tim: yes: detector + error amplifier (minimum gain, even a single
NPN) + FET bias shifting.
- Bill: I love old ARRL publications, 80's QSTs are my favorites, but
3:1 (continuous) coverage is not common on ham bands. I am looking to
build, and design (oh, well, guesstimate).
- Artie and Barry: I knew the heterodyne way of getting broad coverage
thanks to the old Wavetek 2000 Sweep/Signal Generator. This is also
used in tracking generators for spectrum analyzers. But keeping
unwanted mixing products low is an art. Maybe later.
But, the first answer from Tim made me think: what's the difference in
amplitude stabilizing methods on a Wien bridge and my Hartley?. In a
RC oscillator you control amplitude thru some voltage dependent
resistance (a lamp, a FET channel, etc.). But in my oscillator I vary
it thru gate bias, which varies the average transconductance of a
device working in a (nonlinear) class C. So, I Spice'd a FET damping
the coil (source) tap with its Rds being controlled by oscillations
amplitude, an voilà!, I got ~1Vp-p at the gate (and the varicaps), the
gate's 1N4148 almost does not conduct. I will try a more decent of
controlling the feedback (without ruining Q) and let you know.