Conclusion: a power amplifier output match should not be designed only as a small-signal conjugate match. Start from the required fundamental load impedance for the transistor at the current package plane, then separately define harmonic terminations, stability constraints, and the loss budget of the matching network. Treating these as one broad “50-ohm match” problem usually explains why bench tuning changes gain, efficiency, output power, and ruggedness in conflicting directions.

The first limitation is that the load-pull target is conditional. It depends on supply voltage, quiescent class, compression level, pulse or CW operation, device temperature, and whether the target is maximum Pout, best PAE, linearity, or a compromise. Package and bond-wire parasitics must be included before choosing topology: an impedance that looks reasonable at the PCB connector may become a very different termination at the intrinsic current generator. A useful workflow is to simulate or measure load-pull at the package reference plane, draw acceptable Pout/PAE/return-loss contours, then synthesize a network whose component tolerances and thermal drift keep the load inside those contours rather than exactly at one point.

The second distinction is fundamental versus harmonic matching. For class AB, the fundamental load usually dominates delivered power, while second- and third-harmonic terminations strongly shape voltage waveform, efficiency, and peak voltage stress. In narrowband designs, harmonic traps or controlled open/short terminations may be useful; in broadband designs, those same traps can become bandwidth limiters or create unexpected resonances. Also check out-of-band terminations down to video/baseband frequencies, because envelope impedance and bias-network impedance can affect memory effects and linearization even when the RF match looks correct.

Finally, separate matching loss from impedance transformation. A high-Q matching network may present the right Smith-chart point but dissipate enough power to erase the efficiency benefit. Compare candidate topologies using component Q, transmission-line loss, current/voltage stress, thermal rise, and manufacturability, not just simulated S-parameters. Validate with large-signal measurements such as load-pull, pulsed RF if applicable, harmonic power, and stability under mismatch. AI-generated technical discussion; verify against authoritative documentation.