Here's a seldom-talked about issue: "effective sensitivity" . The receiver sensitivity you get in the shop on a 50-ohm dummy load (say, 0.3 microvolts) is the sensitivity as limited _only_ by the receiver's own "self noise" specified by it's noise figure--that's mostly strongly dependent on the noise performance of the first active device in the rx, the first RF amp or the mixer in the absence of an RF amp.
Now you get to a repeater site that has, say, 2 microvolts of steady noise coming in from the antenna (not unusual). Where's that 0.3 microvolts of signal which quieted the rx in the shop? It's buried in the 2 uV of noise & unreadable, nay, completely unnoticeable. What then is the actual, effective sensitivity of the rx in that environment? Certainly NOT the 0.3 uV found in the shop! It will be somewhat above 2 uV, as this is noise level which sets the "starting" point of the quieting curve of an FM rx, not the receiver's own self noise. In this case it's antenna noise, not the rx's self noise, which sets the actual sensitivity in use. We need to remember that in a _system_ (antenna (galactic or other noise), feedline (losses), RF selectivity (losses), & rx (self noise)) the receiver's intrinsic noise level ("noise figure") is further degraded by feedline loss and any further losses due to RF selectivity in front of the receiver, e.g. duplexer/cavity filter, etc. That's why I think it's important to be able to routinely monitor the noise level at the site (see my previous note on S-meters). --John --John