Colpitts Oscillator
A Colpitts oscillator builds its positive feedback network from a single inductor and a capacitive voltage divider — two capacitors in series across the tank, with the junction between them tapped for feedback — rather than tapping the inductor itself, which is the one structural choice that separates it from its mirror-image cousin, the Hartley oscillator, and gives it the lower phase noise and better high-frequency behavior that make it the default LC oscillator topology in RF synthesizers and local oscillators above a few hundred megahertz.
Tapping a capacitor pair instead of an inductor is the whole structural difference from a Hartley oscillator, and it has real electrical consequences, not just topological ones. A Hartley oscillator gets its feedback fraction from a tapped inductor — physically dividing a coil's turns — while a Colpitts oscillator gets the same job done by two capacitors in series across the tank; inductors are harder to build with precise, low-loss taps and carry more parasitic resistance and stray capacitance at the tap point than two discrete capacitors do, so the Colpitts arrangement generally achieves a higher tank Q and correspondingly lower phase noise for the same component quality, which is the single biggest practical reason Colpitts dominates over Hartley in modern RF design.
The series combination of the two feedback capacitors sets the actual resonant frequency, not either capacitor alone, which is a detail that trips up a lot of first analyses. Because C1 and C2 sit in series as seen by the inductor across the tank, the frequency-determining capacitance is the series combination C1×C2/(C1+C2), always smaller than either individual capacitor — so increasing either C1 or C2 alone doesn't simply scale frequency the way it would in a single-capacitor LC tank, and a correct Colpitts design has to account for both the series capacitance (which sets f0) and the C1/C2 ratio (which sets the feedback fraction and therefore how much gain the active device needs to sustain oscillation) as two separate, not fully independent, design variables.
Reaching gigahertz frequencies reliably is where the Colpitts topology's capacitive-only feedback path really earns its keep. At the frequencies used in cellular and Wi-Fi local oscillators, a tapped inductor's stray inductance and capacitance at the tap become a significant fraction of the resonant tank's total reactance, distorting the intended frequency and feedback ratio in ways that are hard to predict and correct for; a capacitive divider's parasitics scale more predictably and stay a smaller fraction of the intended capacitance values at these frequencies, which is why essentially every modern RF synthesizer and crystal-oscillator overtone circuit built from discrete LC elements defaults to a Colpitts (or closely related Clapp) configuration rather than a Hartley one.
The Clapp oscillator is a one-component variation that fixes Colpitts's main weakness: frequency stability when the active device's own capacitance drifts. Adding a third capacitor in series with the inductor (rather than across it) lets a designer make that series capacitor small relative to C1 and C2, which dominates the frequency-setting math and makes the resonant frequency far less sensitive to the transistor's own junction capacitance changing with temperature or bias — a genuinely useful refinement for applications needing better frequency stability than a bare Colpitts provides, without changing the core capacitive-feedback mechanism at all.
Read a Colpitts oscillator through a *which-reactive-element-divides-the-feedback* lens rather than a *generic LC oscillator* lens: the choice to split the tank with two capacitors instead of a tapped inductor is a single structural decision with cascading consequences — better achievable Q, lower phase noise, and cleaner scaling to gigahertz frequencies — which is exactly why Colpitts (and its Clapp refinement) displaced Hartley as the default LC oscillator topology once RF frequencies climbed high enough to expose a tapped inductor's real-world limitations.