See PDF for the rest of the text and images
Anyone who have designed a capacitively tuned VFO/VCO with the wide tuning range (>1.5-2x) has noticed that the oscillation threshold changes substantially with the tuning frequency. The usual solution to the problem is a combination of: 1. the AGC circuit, most frequently via the amplifier bias that changes the amplifier conduction angle; 2. shunting the LC circuit with a damping resistor or by a low impedance amplifier terminal; Both methods have a common drawback - amplifier gain must be increased to maintain the oscillation condition margin on both sides of the tuning range and/or to compensate for the damping resistor losses. Increasing gain requires either more complex amplifier or tighter coupling of the resonant circuit to the amplifier. The later increases the influence of amplifier parasitic elements on the resonant circuit. The more complex amplifier can increase the phase jitter. Both solutions have a negative impact on the oscillator frequency stability as well as the phase noise. There is a passive solution to the problem that requires neither damping the Q-factor of the resonant circuit nor a large dynamic range AGC. As far as I know it first appeared in the research paper by Vackar [1] published back in 1949. Vackar is best known for his oscillator topology shown on fig.5 in [1] (see below). However it is the "extended range" Vackar oscillator shown on fig.6 that introduces the first oscillator arrangement with nearly flat oscillation threshold across the wide tuning range: