Quote:It is my guess that is the noise sources are uncorelated than (2) should be used.
Correlation has nothing to do with. RMS is still RMS. It is always the first definition when you have discrete points. If the signal is continuous, e.g. a sine wave, then the definition is 1/N*∫f(x)
2dx. For the case of sine wave, the integral simplifies to sqrt(2)*peak.