Pink (1/f) noise has equal energy per octave and models the long-range correlations found in many natural signals including music

Pink noise (also called 1/f noise) has a power spectral density proportional to 1/frequency. This means each octave (doubling of frequency) contains the same energy, resulting in a perceptually balanced sound. By contrast, white noise has equal energy per hertz (too bright to the ear) and Brownian noise (1/f squared, random walk) is too correlated. Voss and Clarke (1975) found that musical pitch and loudness sequences have 1/f spectra, suggesting that music’s balance of novelty and return corresponds to 1/f statistics. Algorithmic composers use 1/f generation as a middle ground between random white noise generators (too chaotic) and overly correlated processes (too monotonous).

Examples

Generation based on 1/f noise has been used in algorithmic composition as an alternative to overly random white noise. Voss and Clarke analysed radio broadcasts and found 1/f loudness fluctuations.

Assessment

Why does white noise feel less musical than 1/f noise when used as a pitch or rhythm generator? What makes 1/f noise a better model of natural musical variation than Brownian noise?