An LNA cannot usually be both noise-matched and power-matched at the same impedance unless the device is designed for simultaneous noise and input match. The source reflection coefficient that minimizes noise, Gamma_opt, comes from the transistor noise parameters Fmin, Gamma_opt, and Rn. The input match that gives the best S11 is instead determined by the small-signal input network and device parasitics. A reactive matching network can transform the source impedance toward Gamma_opt, but if that transformed impedance differs from the conjugate of the LNA input impedance, return loss will degrade. The trade-off is not simply gain versus noise figure; it is a constrained optimization among NF, S11, gain, stability, linearity, and bandwidth.

The practical boundary condition is the noise resistance Rn. If Rn is small, moving away from Gamma_opt causes only a modest NF increase, so you can deliberately shift the match toward the power-match point to improve S11 and ripple with little NF penalty. If Rn is large, the same impedance offset may cause a meaningful NF increase, so forcing a textbook 15 or 20 dB return loss can be the wrong design target. Bandwidth adds another constraint: narrowband LNAs can use high-Q reactive matching to approach both targets over a small band, while broadband LNAs often must accept a compromise because Gamma_opt and S11* move differently with frequency. Loss before the active device is especially damaging because every decibel of passive input loss adds directly to cascaded noise figure, so a lossy pad used to improve return loss is usually not an acceptable substitute for proper matching.

A useful design flow is to first fix bias for the required current, stability, and linearity, then extract S-parameters and noise parameters across frequency and temperature. Plot Gamma_opt, the stability circles, and the constant-noise-figure circles on the same Smith chart, then synthesize an input network that lands inside an acceptable NF contour while meeting a minimum return-loss target. Verify the cascaded result with Friis noise calculations rather than judging the transistor alone, and check unconditional stability over the full band, including low-frequency bias-loop resonances. If the receiver already has a filter or switch in front of the LNA, optimize the combined source impedance seen by the transistor; a standalone LNA board with excellent bench S11 may still give worse system NF than a slightly mismatched design that presents the right source impedance through the real front-end network.

Reasonable acceptance limits depend on the system. A narrowband receiver may justify a tight S11 target because filter passband ripple and channel flatness depend on it, while a broadband instrument or distributed antenna input may accept poorer return loss if NF and stability remain controlled. The decision should be based on allowed NF degradation, source-impedance variation, component tolerance, temperature drift, and the sensitivity of downstream filters to reflected power. Do not assume that minimum noise figure automatically gives the best receiver sensitivity if the mismatch creates passband ripple or interacts with a preceding reflective filter. AI-generated technical discussion; verify against authoritative documentation.