How we measure

What the tuner does under the hood, which thresholds it decides with, and how far off it is on signals of known frequency. The table is written by an automated test every time this site is published.

Online tuner →

The method

• Pitch detection with MPM (McLeod Pitch Method) on the normalized squared difference function. It takes the first peak above 0.9 times the highest, which is what keeps it from mistaking a note for its lower octave on harmonic-rich sounds.
• Minimum clarity of 0.75: below that, the tuner shows nothing. It would rather stay silent than make something up.
• Median of 3 readings and 2 consistent readings before showing a note; then smoothing that opens fully on a 20-cent jump and closes as the note settles.
• 5 cents of hysteresis between neighbouring notes, and “in tune” within ±5 cents.
• The same logic, with the same numbers, in the iPhone and Apple Watch app (Swift), the Android app (Kotlin) and the online tuner (JavaScript). Audio is analysed on your device and never leaves it.

The results

Synthetic signals at 48 kHz with six harmonics and a little noise, analysed over twelve consecutive windows, as the page does. Largest error in this table: 0.74 cents; the test fails if any goes above 1.

What this test does not prove

• A synthetic signal is not a microphone. On a real instrument the note comes out a few cents sharp on the attack and drops as it decays, and low strings have inharmonicity (explained in what cents are).
• With a lot of noise —around 4 dB SNR— the tuner goes quiet: clarity drops below 0.75. That is on purpose.
• The phone's microphone and the room matter. Moving the instrument closer helps more than any setting.

Check it yourself

With a 440 Hz tuning fork or a tone generator, the tuner should read A4 inside the in-tune zone (±5 cents), very close to 0. If your fork is 442 Hz, change the reference in the app (415 to 466 Hz) and it should sit at 0 again.