R
Member
Newcomers often assume the best LNA match is the one with the lowest return loss. On the bench I have learned to treat noise matching and power matching as two different targets that rarely coincide.
First, I find the optimum noise impedance from the datasheet noise circles, not from S11 alone. The source impedance that minimizes noise figure is usually not 50 ohms, and chasing perfect S11 from that point often costs 0.5 dB or more of noise figure.
Second, I check what the preceding stage actually delivers. If the antenna or filter presents a badly mismatched impedance, the theoretical noise optimum moves, so I measure the real source impedance with the VNA rather than trusting a nominal value.
Third, I decide the trade-off by application. For a weak-signal receiver I accept a return loss around -10 dB if it buys real noise margin. For a stage driving a mixer that needs a flat, predictable level, I give up some noise figure to hold return loss and gain flatness.
Fourth, I verify with a noise figure measurement after matching, not just a VNA sweep. A good-looking Smith chart result that degrades measured noise figure tells me the matching network loss or layout parasitics are eating the benefit.
One practical habit: build the matching network with a little tuning range, measure both noise figure and gain across the band, and only then freeze the values. Have you ever deliberately accepted a worse S11 to protect noise figure, and how did you justify it?
First, I find the optimum noise impedance from the datasheet noise circles, not from S11 alone. The source impedance that minimizes noise figure is usually not 50 ohms, and chasing perfect S11 from that point often costs 0.5 dB or more of noise figure.
Second, I check what the preceding stage actually delivers. If the antenna or filter presents a badly mismatched impedance, the theoretical noise optimum moves, so I measure the real source impedance with the VNA rather than trusting a nominal value.
Third, I decide the trade-off by application. For a weak-signal receiver I accept a return loss around -10 dB if it buys real noise margin. For a stage driving a mixer that needs a flat, predictable level, I give up some noise figure to hold return loss and gain flatness.
Fourth, I verify with a noise figure measurement after matching, not just a VNA sweep. A good-looking Smith chart result that degrades measured noise figure tells me the matching network loss or layout parasitics are eating the benefit.
One practical habit: build the matching network with a little tuning range, measure both noise figure and gain across the band, and only then freeze the values. Have you ever deliberately accepted a worse S11 to protect noise figure, and how did you justify it?
