13.4 What is white noise? What is pink noise?
Description
This article is from the rec.audio.* FAQ, by with numerous contributions by Bob Neidorff others.
13.4 What is white noise? What is pink noise?
"White noise" is characterized by the fact that its value
at any two different moments in time are uncorrelated.
This leads to such noise having a flat power spectral
density (in signal power per hertz of bandwidth), and is
loosely analogous to "white light" which has a flat power
spectral density with respect to wavelength.
Pink noise has flat power spectral density per PERCENTAGE
of bandwidth, which leads to a rolloff of -3 dB/octave
compared with white noise.
There are many reasons for using pink noise in audio testing.
One is that music has an average spectral content much closer
to pink noise than white noise. Another is that pink noise
can be readily measured with constant Q bandpass filters and
naturally leads to flat plots on logarithmic frequency scales
- which correspond to the equally tempered musical scale.
Pink noise is often used with 1/3 octave band filters to
measure room acoustics. This idea has merit since 1/3 octave
is a convenient number near the limit of our ears ability to
detect frequency response irregularities, and because
averaging measurements over 1/3 octave bands smooths out the
numerous very narrow peaks and dips that arise due to
standing waves in rooms.
Another term you'll hear about is Gaussian noise - this is
noise with a Gaussian amplitude probability density.
Gaussian noise has the amazing property that linearly
filtering it preserves its Gaussian amplitude density and
that sums of Gaussian random variables are again Gaussian.
The two terms shouldn't be confused. It is possible to have
Gaussian white or pink noise.
Continue to:
My Books
