Once you saturate an LNA does the noise figure begin to increase rapidly?
I was working on a project which utilized an LNA in an RFD path to ultimately feed a GPS receiver. While watching the GPS L1 center frequency (1575.42 MHz) at the input to the GPS receiver with a spectrum analyzer we began to radiate INMARSAT transmissions (1626.5 to 1660.5 MHz) in the direction of the GPS receiver. The noise floor at 1575.42 increased as the INMARSAT transmission began to transmit, but slowly at first. The INMARSAT transmitter varies the power output to maintain a fixed data rate with a fixed modulation during clear days or in rain and at different altitudes. So varying the output power is a real world test. When we closed in on a typical maximum RF output from the INMARSAT transmitter the noise floor at 1575.42 MHz quickly doubled.
This leads me to believe that once we hit the compression point of the LNA and the gain ceased to become linear, the SNR kept going down and the noise factor kept increasing as the INMARSAT transmission power increased. I would go on to guess that as the gain leveled off and the power continued to increase, the additional RF input power was translated to mostly heat thus increasing the thermal noise of the LNA and causing the behavior we saw.
It's probably obvious by now that I don't design LNAs for a living. So am I in the right ballpark?