The Colors of Noise Explained: White, Pink, Brown, Blue, and Violet Spectra

In colloquial speech, “noise” means unwanted acoustic interference. In physics and audio engineering, noise is a continuous stochastic (random) signal containing a dense statistical distribution of frequencies.

Just as white light contains all visible wavelengths, white noise contains all audible acoustic frequencies. By tilting the statistical Power Spectral Density (PSD) across the spectrum, engineers define the distinct “colors” of noise. The audible band spans log2(20,000 / 20) = 9.97 octaves, so a slope quoted per octave compares the energy in each doubling of frequency with the octave below it.

---

1. White Noise (Flat Power per Hertz)

White noise is the theoretical foundation of acoustic noise. Its power spectral density is completely flat: PSD(f) = constant, which is 0 dB per octave.

• Energy distribution: equal acoustic energy per Hertz of bandwidth. The 100 Hz band from 100 Hz to 200 Hz carries the same energy as the 100 Hz band from 10,000 Hz to 10,100 Hz.
• Perception: harsh, bright, hissing, like a loud waterfall. Because hearing is logarithmic, white noise sounds overwhelmingly treble-heavy: about 95% of its energy sits above 1 kHz.
• Uses: testing sound isolation and transmission loss, acting as a reference when a measurement is defined in hertz, and generating dither for quantisation, which the dithering guide covers.

---

2. Pink Noise (Flat Power per Octave)

Pink noise (1/f noise, or flicker noise) falls at −3 dB per octave (−10 dB per decade): PSD(f) is proportional to 1/f, which is −3.01 dB per octave.

• Energy distribution: equal acoustic energy per percentage bandwidth (octave). The octave from 100 Hz to 200 Hz carries the same energy as the octave from 10,000 Hz to 20,000 Hz.
• Perception: warm, balanced, natural, like steady rainfall. Because hearing scales logarithmically, pink noise sounds subjectively flat.
• Uses: the workhorse for loudspeaker and room calibration, because with octave or third-octave banding it reads as a horizontal line, so any deviation from flat is the room or the speaker rather than the signal. The Room Acoustic Test uses that principle to expose modal peaks and nulls.

---

3. Brown / Red Noise (Brownian Motion)

Brown noise (named after botanist Robert Brown and Brownian motion, not the color) rolls off at −6 dB per octave (−20 dB per decade): PSD(f) is proportional to 1/f², which is −6.02 dB per octave.

• Energy distribution: heavily weighted toward sub-bass. Each octave up the band carries a quarter of the power of the one below it.
• Perception: deep, rumbling, heavy, like distant thunder or an airliner cabin.
• Uses: masking low-frequency rumble from traffic and ventilation, and exciting room modes for subwoofer checks. It is also popular as a sleep and focus aid, though the sleep claims are marketing rather than measurement, and masking noise only masks.

---

4. Blue Noise (Treble Emphasized)

Blue noise (azure noise) is the inverse of pink noise, rising by +3 dB per octave (+10 dB per decade): PSD(f) is proportional to f, which is +3.01 dB per octave.

• Energy distribution: minimal low-frequency energy, high energy in the upper octaves.
• Perception: a high, sharp hissing sizzle, like steam escaping a valve.
• Uses: tweeter and high-frequency driver tests, because a dead tweeter is obvious when the signal carries almost no bass to hide behind; masking low-frequency disturbances while leaving the bass clear; and blue-noise dither and halftoning.

---

5. Violet Noise (Extreme High End)

Violet noise (purple noise) is the inverse of brown noise, rising at +6 dB per octave (+20 dB per decade): PSD(f) is proportional to f², which is +6.02 dB per octave.

• Energy distribution: acoustic energy quadruples for every octave increase, so bass and midrange are effectively absent.
• Perception: an extremely sharp, piercing hiss, closer to escaping air than to sound.
• Uses: top-octave driver work, checking that a tweeter, super-tweeter or ultrasonic transducer produces output at all, and high-frequency attenuation measurements where a rising spectrum compensates for the roll-off of the device under test.

---

6. Noise Spectrum Reference Matrix

Color	PSD Slope (per hertz)	Amplitude Slope	Reading on a 1/1-Octave RTA	Energy Distribution	Primary Engineering Use
Brown	−6.02 dB/octave (1/f²)	−3.0 dB/octave	Falls 3 dB per octave	Massive sub-bass emphasis	Masking rumble, sub checks
Pink	−3.01 dB/octave (1/f)	−1.5 dB/octave	Flat	Equal energy per octave	Speaker and room calibration
White	0 dB/octave (f⁰)	0 dB/octave	Rises 3 dB per octave	Equal energy per hertz	Isolation testing, dither
Blue	+3.01 dB/octave (f¹)	+1.5 dB/octave	Rises 6 dB per octave	Treble weighted	Tweeter tests, halftoning
Violet	+6.02 dB/octave (f²)	+3.0 dB/octave	Rises 9 dB per octave	Extreme treble	Top-octave and ultrasonic work

The two slope columns describe the same spectrum from different axes: each octave band is twice as wide as the one below it, so a flat-per-hertz spectrum collects twice the power in each successive octave and rises on an octave-band analyser, while pink noise cancels that growth exactly.

---

7. Crest Factor and Headroom

Noise is random, so its peaks sit well above its average level. Gaussian noise is commonly quoted with a crest factor near 12 dB, meaning peaks about four times the RMS value, and the expected maximum over N samples is about σ × √(2 × ln N): over one second at 48 kHz that is 4.64σ, or 13.3 dB.

The practical consequence is headroom. A noise signal with an RMS of −20 dBFS peaks near −8 dBFS, and that 12 dB gap is what a generator must preserve to avoid clipping. The Noise Generator makes the same point with its level slider: 30% reads −10.5 dBFS because 20 × log10(0.3) = −10.5, and that reading is a digital gain figure, not an SPL calibration.

---

8. Choosing a Color for the Job

• Room and speaker checks: pink noise, because equal energy per octave matches the way we hear and reads flat on an octave-band analyser.
• Low-frequency masking and subwoofer work: brown noise, with the corner-frequency caveat in mind.
• Tweeter and high-frequency driver tests: blue or violet noise, at a low level, on wired output.
• Sleep and focus sound: brown or grey noise from Sleep Sounds, played quietly.

---

9. Limits, Mistakes and Troubleshooting

• The channels are identical. Most browser generators build one channel and copy it to the outputs, so the stereo image sits dead centre with no decorrelation.
• Device processing reshapes the extremes. Operating system enhancements, Bluetooth codecs and speaker DSP all alter the top and bottom octaves, so listen on wired headphones before blaming the generator.
• Hearing care comes first. High-frequency colors concentrate energy where hearing is most sensitive, so start low with blue and violet and never raise the level just to make a sound audible. If you are concerned about your hearing, see an audiologist.