Pink noise generator

What pink noise is

Pink noise has a power spectral density that falls in inverse proportion to frequency — the relationship usually written 1/f. Double the frequency and the power halves.

The consequence is the useful part. That falling curve exactly cancels the fact that octaves get wider as you go up, so every octave ends up carrying the same total energy: as much between 100 and 200 Hz as between 5,000 and 10,000 Hz. Human hearing is roughly organised by octaves, so a spectrum that is flat per octave is the one that sounds flat to us. White noise is flat on paper; pink noise is flat to the ear.

Why people prefer it

The high-frequency emphasis that makes white noise an effective speech masker is also what makes it tiring after a while. Pink noise trades some of that masking power for comfort, and for long sessions most people find the trade worth making. It is the colour listeners tend to converge on when they have tried all three over a period of nights rather than minutes.

The resemblance to rain, wind, and moving water is not a coincidence either. Many natural broadband sounds fall roughly along a 1/f slope, and 1/f patterns show up well beyond acoustics — in heartbeat intervals, in river flow rates, in the statistics of music. Whether that familiarity is why pink noise feels comfortable is speculation, but the familiarity itself is real.

What the research supports

This is where it is worth being careful, because pink noise attracts confident claims it cannot support.

There are small studies suggesting that pink noise delivered in short bursts, timed to the slow brain waves of deep sleep, may affect sleep depth and next-day memory. They are interesting. They are also small, short, mostly conducted in labs with equipment that tracks the sleeper’s brain activity in real time, and they are not a finding about leaving a phone playing noise on a nightstand. Broader reviews of continuous noise for sleep report low-certainty evidence and inconsistent results.

What can be said plainly: pink noise raises the background level, and a raised background level genuinely does mask intrusive sound. That mechanism is not in dispute. Everything past it is worth trying rather than believing.

How this generator makes it

Filtering random noise into a 1/f slope is harder than it looks, and the straightforward approaches are either expensive or drift out of balance.

This player uses Paul Kellet’s refined method, the same one in the Android app: six one-pole filters running in parallel over a white noise source, each tuned to a different corner frequency, summed with a small direct contribution. Together they trace the 1/f curve closely across the audible range at a cost of a few multiplications per sample.

The filter state carries across audio blocks continuously. That detail matters more than it sounds — resetting those six values between blocks would put a tick into the output at a steady rate, and a periodic tick is precisely the thing generated noise is supposed to avoid.

Pink, white, or brown

Pink sits in the middle by design. If it is not masking enough, go up to white; if it is still too bright, or the problem is traffic and bass rather than voices, go down to brown. The swatches switch colours without stopping playback, so the comparison takes seconds.