The FAQ says "better than ±1 cent on a clean ringing string". That is an easy sentence to write. So we checked it the only way that means anything: we took the detection code straight out of this page, ran it outside the browser, and fed it notes whose pitch we had set ourselves. If you know the answer before you ask, you can mark the tuner's homework.
A quick unit first. A cent is a hundredth of a semitone. On the open A string (110 Hz) one cent is about 0.06 Hz. Most trained ears start to notice a difference somewhere around five cents.
Each test note is a synthetic pluck: a fundamental plus seven harmonics, each fading a little faster than the one below, sampled at 48,000 times a second. The tuner looks at 4,096 samples at a time (about 85 milliseconds of sound), keeps only readings it is at least 88% confident in, and shows the median of the last eight. The test does exactly the same, so the numbers below are what the readout would have shown, not one lucky frame.
First, an idealised note where every harmonic is an exact multiple of the fundamental. This measures the detector on its own, with nothing else in the way.
| String | True pitch | Shown | Error |
|---|---|---|---|
| Low E (E2) | 82.407 Hz | 82.407 Hz | 0.00 ¢ |
| A (A2) | 110.000 Hz | 110.000 Hz | 0.00 ¢ |
| D (D3) | 146.832 Hz | 146.832 Hz | 0.00 ¢ |
| G (G3) | 195.998 Hz | 195.998 Hz | 0.00 ¢ |
| B (B3) | 246.942 Hz | 246.942 Hz | 0.00 ¢ |
| High E (E4) | 329.628 Hz | 329.627 Hz | 0.00 ¢ |
On a clean tone the detector is exact to the hundredth of a cent. That part of the FAQ holds.
Real strings are stiff, so their upper harmonics sit a little sharp of perfect multiples. Wound bass strings do it more, thin plain strings less. When we gave the test plucks a typical amount of stretch, every guitar string read between 1.33 and 1.37 cents sharp of its fundamental. The 5-string bass's low B (30.868 Hz), given a gentler stretch, read 30.881 Hz: 0.73 cents sharp.
Two things make this harmless. The offset is the same on every string, so a guitar tuned with it agrees with itself, and chords ring true. And your ear hears those sharp harmonics too, so the note you perceive is also a touch above the bare fundamental. We also detuned the A string on purpose: set 3 cents sharp, it showed 4.4 sharp; set 7 cents flat, it showed 5.6 flat. Turn the peg until the needle centres and you land within about a cent and a half of the textbook pitch.
We added random hiss to the low E at four levels. At the first three the reading barely moved: 1.46, 1.53 and −1.42 cents. At the loudest level the detector's confidence fell as low as 0.78, none of the eight readings reached 0.88, and the readout showed a dash. That is the behaviour we want: when the sound is too messy to read, it says so instead of inventing a number.
A very soft pluck and a note heard two seconds after it was struck both produced a dash. Neither got past the volume gate, so the tuner did not even try to guess. If you see a dash with a ringing string, pluck a little harder and closer to the phone, then read it straight away.
The FAQ used to say that several notes at once gives a dash. The test says otherwise. We played the open A and the E above it together: the tuner showed 54.997 Hz, an A an octave below the A string, with 99.9% confidence. A strummed E major chord showed 41.264 Hz, an E an octave below the low E.
It is not a malfunction. Two notes a fifth apart combine into a wave that repeats at a lower pitch, and the detector faithfully finds that repeat. It just is not the note you meant. So the rule is simple: one string at a time, and rest a finger on the others so they stop ringing. We have corrected the FAQ answer to match.
These are computed notes, not a microphone in a real room. Your phone's microphone, its noise suppression, and how the instrument is played all add their own error. The tuner's device report tells you if noise suppression is on, which is the most common reason a phone struggles with long notes.