Power spectral density

Power spectral density (PSD) and autocorrelation are like two sides of the same coin (autocorrelation can be recovered from PSD -- see page 575 ). PSD aims at explaining the amount of fluctuation in a random process $$X[n]$$. PSD is like the equivalent of Fourier transform for probabilistic sequences: Just like the Fourier transform explains the fluctuations in a deterministic sequence, so does PSD describe the fluctuations in a random process. If a random process fluctuates a lot, then its PSD will die out slowly, whereas if it depends a lot on the past, then its PSD will die out quickly.