Tom is right. Negative feedback can only reduce the noise introduced by the amplifier itself. If the system works properly, the input will be perfectly amplified - with no noise added by the amplifier. But if noise is part of the input, then that noise is amplified by exactly the same amount as the intended input signal.
Hey Tom,
I once heard the argument that negative feedback could eliminate noise and it was supported by live measurements. Actually, more than once from more than one source. It usually went something like this...
Here's a noisy input signal (shown by live measurement - very fuzzy) and here's the amplified output with a larger amplitude, but less noise content (also shown on the scope - not fuzzy at all). It was clear that the output had *less* noise than the input. Therefore, the negative feedback must be reducing the noise! Many attendees just accepted the result...and continued to propagate the idea. The problem of course was the input noise signal was high frequency, beyond the bandwidth of the system. It was just being filtered out by the inherently lowpass universe. Have you seen any talks like this?
I think there are a lot of similar situations out there that contribute to common misconceptions like this one. And this one is really common.
