The Science of Noise

Not all noise is created equal. Different "colors" of noise have distinct frequency distributions that affect how we perceive them and how they influence our brain. Understanding these differences can help you choose the right noise for your needs.

What is Sound Color?

Just as light has colors determined by wavelength, sound has "colors" determined by how energy is distributed across frequencies. White light contains all visible wavelengths equally; white noise contains all audible frequencies at equal power. Other noise colors have different frequency weightings that change their character and usefulness.

White Noise

Contains all frequencies at equal intensity, creating a consistent "hiss" sound. White noise is excellent for masking sudden sounds like traffic or conversation because it covers the entire audible spectrum. It's commonly used in offices, for baby sleep, and in audio testing.

Character: Bright, hissy, energetic
Best for: Sound masking, concentration, audio calibration

Pink Noise

Has equal energy per octave, meaning lower frequencies are louder than higher ones (3dB/octave rolloff). This creates a more balanced, natural sound that many find more pleasant than white noise. Pink noise mimics the frequency distribution of many natural sounds like rainfall and rustling leaves.

Character: Balanced, natural, soothing
Best for: Sleep, relaxation, studying, tinnitus relief

Brown Noise

Also called Brownian or red noise, it has even more bass emphasis (6dB/octave rolloff). Named after Robert Brown who described Brownian motion, not the color. It produces a deep, rumbling sound like a waterfall, distant thunder, or strong wind. Many people find it the most relaxing noise color.

Character: Deep, rumbly, powerful
Best for: Deep relaxation, meditation, blocking low-frequency sounds

Blue Noise

The opposite of pink noise - higher frequencies are louder (+3dB/octave). Also called azure noise. The power density increases with frequency, creating a crisp, bright sound. Blue noise is used in audio dithering and some find it helpful for maintaining alertness.

Character: Crisp, bright, energizing
Best for: Alertness, audio production, high-frequency tinnitus masking

Violet Noise

The opposite of brown noise - even more high-frequency emphasis (+6dB/octave). Also called purple noise. It has a sharp, airy quality due to the strong treble content. Some people find it helpful for masking high-pitched tinnitus sounds.

Character: Sharp, airy, bright
Best for: High-frequency tinnitus, audio testing, differentiation effects

Research-Backed Benefits

Peer-reviewed scientific studies have examined the effects of noise on sleep, cognition, and well-being:

Sleep Quality: A 2021 study in Sleep Medicine found that white noise significantly improved subjective and objective sleep measurements in participants living in high-noise environments ^[1]

ADHD and Focus: A 2024 meta-analysis in the Journal of the American Academy of Child & Adolescent Psychiatry (k=13, N=335) found a small but statistically significant benefit of white and pink noise on task performance for individuals with ADHD (g=0.249, p<0.0001) ^[2]

Sleep in Infants: Research shows white noise can extend total sleep time, improve sleep efficiency, and reduce awakenings in infants and toddlers ^[3]

Pink Noise and Deep Sleep: A systematic review found that 81.9% of pink noise studies showed positive effects on sleep outcomes, compared to 33% for white noise ^[4]

Tinnitus Masking: Sound therapy has been used since the 1920s for tinnitus relief. The American Tinnitus Association recommends sound enrichment as a key management strategy ^[5]

Creativity: Research by Mehta et al. in the Journal of Consumer Research found that moderate ambient noise (~70dB) enhances creative cognition ^[6]

References

Messineo L, et al. (2021). "The effects of white noise on sleep and duration in individuals living in a high noise environment." Sleep Medicine. PMID: 34049045
Nigg JT, et al. (2024). "Systematic Review and Meta-Analysis: Do White Noise or Pink Noise Help With Task Performance in Youth With ADHD?" J Am Acad Child Adolesc Psychiatry. PMID: 38428577
Impact of white noise on sleep quality across age groups. (2025). Sleep Medicine. Meta-analysis of randomized controlled trials.
Riedy SM, et al. (2021). "Noise as a sleep aid: A systematic review." Sleep Medicine Reviews. PMID: 33007706
American Tinnitus Association. "Sound Therapy." ata.org/about-tinnitus/sound-therapy/
Mehta R, et al. (2012). "Is Noise Always Bad? Exploring the Effects of Ambient Noise on Creative Cognition." Journal of Consumer Research.

Choosing the Right Noise

The best noise color depends on your personal preference and intended use:

For Sleep: Most people prefer pink or brown noise. Pink noise is gentler, while brown noise provides a deeper, more enveloping sound

For Focus/Studying: Try pink noise first. Some prefer white noise for more effective sound masking in noisy environments

For Relaxation: Brown noise is often the most relaxing due to its deep, soothing quality

For Tinnitus: Start with pink noise, then experiment. The goal is to match the frequency range of your tinnitus

For Babies: White noise is traditional, but pink noise may be safer for developing ears due to lower high-frequency content

For ADHD: Research supports white or pink noise for improved attention. Individual response varies - experiment to find what works for you

Tinnitus Relief

Tinnitus (ringing, buzzing, or hissing in the ears) affects millions worldwide. Sound therapy is one of the most accessible management approaches:

Masking: External sound covers the perceived tinnitus, providing immediate relief. Match the noise frequency to your tinnitus pitch

Habituation: Low-level background noise helps the brain learn to ignore tinnitus over time. Use at barely audible levels

Sound Enrichment: Consistent audio stimulation may help compensate for hearing loss that often accompanies tinnitus

Recommended approach: Start with pink noise at a low volume. Adjust the low/high cut filters to focus on your tinnitus frequency range

Safe Listening Practices

While noise generators are generally safe, follow these guidelines:

Keep Volume Moderate: Listen at the lowest effective volume. If you need to raise your voice to be heard over the noise, it's too loud. Research suggests some devices can exceed 91dB at max volume ^[7]

Use Quality Speakers/Headphones: Cheap speakers can distort at certain frequencies, reducing effectiveness and potentially causing fatigue

Take Breaks: If using for extended periods, give your ears periodic rest

Distance Matters: Place speakers at a reasonable distance rather than right next to your head

For Infants: Place the noise source away from the crib and use the lowest effective volume. Consult your pediatrician for personalized guidance

The Mathematics Behind Noise Generation

Understanding the mathematical foundations of noise generation reveals why different colors sound so different. Each noise type follows a specific power spectral density function that determines how energy is distributed across frequencies.

Power Spectral Density

The "color" of noise is defined by its power spectral density S(f), which describes how power is distributed across frequencies. For colored noise, this follows the general form:

S(f) = 1 / f a Where f is frequency and a determines the noise color

White Noise (a = 0)

S(f) = constant

Algorithm: Pure random sampling

for each sample:
    output[i] = random(-1, 1)

White noise has flat spectral density - equal power at all frequencies. Generated by sampling a uniform random distribution, where each sample is independent. The power per Hz is constant, but since there are more high frequencies in each octave, it sounds "bright" or "hissy".

Spectral slope: 0 dB/octave
Autocorrelation: Delta function (no memory)

Pink Noise (a = 1)

S(f) = 1/f

Algorithm: Paul Kellet's refined method (IIR filter bank)

b0 = 0.99886*b0 + white*0.0555179
b1 = 0.99332*b1 + white*0.0750759
b2 = 0.96900*b2 + white*0.1538520
b3 = 0.86650*b3 + white*0.3104856
b4 = 0.55000*b4 + white*0.5329522
b5 = -0.7616*b5 - white*0.0168980
pink = b0+b1+b2+b3+b4+b5+b6+white*0.5362
b6 = white * 0.115926

Pink noise has equal power per octave. The Kellet algorithm uses 6 first-order IIR filters in parallel, each tuned to produce the 1/f slope. The coefficients were empirically optimized to achieve <0.05dB error from ideal pink noise across the audio spectrum.

Spectral slope: -3 dB/octave
Self-similarity: Scale-invariant (fractal)

Brown Noise (a = 2)

S(f) = 1/f 2

Algorithm: Integrated white noise (random walk)

lastOut = 0
for each sample:
    white = random(-1, 1)
    output[i] = (lastOut + 0.02*white) / 1.02
    lastOut = output[i]

Brown noise is the integral of white noise, equivalent to Brownian motion in one dimension. Each sample depends on the previous, creating a "random walk". The leaky integrator (division by 1.02) prevents DC drift while maintaining the 1/f² characteristic.

Spectral slope: -6 dB/octave
Physical analog: Brownian particle motion

Blue Noise (a = -1)

S(f) = f

Algorithm: Differentiated pink noise

// High-pass filter approach
for each sample:
    blue[i] = white[i] - white[i-1]
    // Apply light smoothing to reduce harshness

Blue noise is the derivative of pink noise, boosting high frequencies by +3dB per octave. Also called "azure noise," it's commonly used in audio dithering and error diffusion in image processing.

Spectral slope: +3 dB/octave
Applications: Audio dithering, halftoning

Violet Noise (a = -2)

S(f) = f 2

Algorithm: Differentiated white noise

lastSample = 0
for each sample:
    white = random(-1, 1)
    violet[i] = white - lastSample
    lastSample = white

Violet noise (also called purple noise) is the derivative of white noise, with power increasing proportionally to f². This creates strong high-frequency emphasis, useful for testing treble response and masking high-pitched tinnitus.

Spectral slope: +6 dB/octave
Mathematical: First derivative of white noise

The Filter Bank Approach

Pink noise generation typically uses parallel filter banks because the 1/f spectrum cannot be created with a simple finite-order filter. The Kellet method approximates the ideal response using cascaded first-order sections:

H(z) = sum of [ g k / (1 - p k z -1) ] Where each section contributes a different frequency band

The pole positions (p_k) are logarithmically spaced to cover the audio range, and the gains (g_k) are chosen to sum to the correct 1/f response. This approach achieves excellent accuracy with minimal computation.

Frequency Domain Relationships

The noise colors form a mathematical family related by integration and differentiation:

Violet (a=-2) d/dt

Blue (a=-1) d/dt

White (a=0) integrate

Pink (a=1) integrate

Brown (a=2)

Differentiating a signal boosts high frequencies (+6 dB/octave), while integrating boosts low frequencies (-6 dB/octave). This means brown noise is white noise integrated once, and pink noise sits exactly halfway between them.

Digital Implementation Considerations

Sample Rate: Higher sample rates provide better high-frequency accuracy. At 44.1kHz, the Nyquist limit is 22.05kHz, covering the full audible range
Bit Depth: 32-bit float provides sufficient dynamic range for noise generation without quantization artifacts
Buffer Looping: Pre-computed buffers (2+ seconds) are looped for efficiency. The buffer must be long enough to avoid perceptible repetition
Random Number Quality: JavaScript's Math.random() is sufficient for audio noise, though cryptographic randomness isn't required
Gain Compensation: Different noise colors have different RMS levels; brown noise requires ~3.5x gain boost to match perceived loudness

Why Pink Noise Sounds "Natural"

Pink noise (1/f noise) appears remarkably often in nature and human perception:

Natural Sounds: Rainfall, wind, ocean waves, and rustling leaves all exhibit approximate 1/f spectra
Human Hearing: Our perception of loudness roughly follows a logarithmic scale, matching the octave-based energy distribution of pink noise
Music: Analysis of diverse musical genres shows their long-term spectral content approximates pink noise
Biology: Heart rate variability, neural activity, and many biological signals show 1/f characteristics

This ubiquity suggests that 1/f noise represents a fundamental pattern in complex systems operating at the boundary between order and randomness.

Focus Hum

Use Cases

Pure Noise Types

Noise Mixer

Volume & Timer

Your Saved Presets

Frequency Shaping

Modulation