As I also did in the past, you are mixing two different things:
- What you say that you find in the papers has to do with the SNR calculation: first the output noise power is calculated or obtained from a simulation, and then you divide it by the (square) of the gain that the amplifier has at the frequency of interest (usually the gain in the pass-band), to obtain the SNR.
-What SPICE gives you is an entirelly different thing: it is the noise spectrum that would have to be generated by a noise source at the input of the amplifier, to have the same output noise spectrum. THIS HAS NO PHYSICAL MEANING, I.E. THIS NOISE DOES NOT EXIST. For example, if the amplifier has a low-pass characteristic, the input noise given by SPICE increases with frequency and, if you integrate it, you get ∞.