Class-A burns power continuously. Class-AB conducts for 180°+ of each cycle. Class-C goes further: the transistor conducts for less than 180° — typically 90° to 150° — and a resonant tank circuit reconstructs the missing portion of the sinewave. The payoff is efficiency above 80%, but only at a single frequency and only for constant-envelope signals.

How it works: The transistor is biased below cutoff (negative V_BE for an NPN, or V_GS well below V_th for a MOSFET). RF drive only pushes it into conduction at the peaks of each input cycle, producing narrow current pulses into a parallel LC tank tuned to the fundamental. The tank's flywheel action stores energy during the off-time and rings out the full sinewave — exactly like pushing a swing once per cycle.

Why the efficiency wins: When the transistor conducts, V_CE is near zero (high current, low voltage). When V_CE swings high, current is zero. The overlap of V and I — which is where power dissipation lives — is minimized. Theoretical peak efficiency hits 100% as conduction angle → 0°, but output power also drops to zero. Practical designs target a 120°–150° conduction angle for ~75–85% efficiency.

The killer limitation: Class-C is brutally nonlinear. The output current pulses contain enormous harmonic content — the tank filter is mandatory, not optional. This means Class-C works only for FM, CW, and constant-envelope modulation. Try to amplify AM or SSB with Class-C and you'll destroy the envelope.

Real-world example: The final stage of a 50W FM broadcast transmitter at 100 MHz. A single RF power MOSFET (e.g., MRF101AN) is biased at V_GS = 0V (well below the 2.5V threshold), driven hard with ~5W of RF. A pi-network tank (L ≈ 80nH, two ~30pF caps) tuned to 100 MHz extracts the fundamental and matches to a 50Ω antenna. Drain efficiency: ~82%.

Rule of thumb — conduction angle vs efficiency:

• 180° (Class-B): η_max = 78.5%
• 150°: η ≈ 85%
• 120°: η ≈ 90%
• 90°: η ≈ 94%, but output power drops to ~40% of Class-B

Design gotchas: Tank Q must be high enough (loaded Q ≥ 10) to suppress harmonics below FCC limits — typically −60 dBc for broadcast. Drive level is critical: too little and the transistor never turns on; too much and you saturate and create even more harmonics. And always add a harmonic filter (low-pass) after the tank — Q=10 isn't enough to meet spurious emission specs alone.