Steady Pitch / for singers

Why your tuner says you are singing the wrong note, and why it is usually the octave

You sing a note that feels right. The tuner names something an octave away, or calls you fifty cents flat on a note you are certain you hit. Three quite different things cause that, and only one of them is you.

The three causes, in the order they actually happen

I have read a lot of threads that start with somebody wondering whether they are tone deaf because of a tuner reading. In almost all of them the tuner was the thing that was wrong, or nothing was wrong at all.

You matched the note an octave down, and that is correct. Somebody plays a note in the middle of the piano and asks you to sing it. If you are a man, or a woman with a low voice, you land the same note one octave below. That is a correct match. Musically it is the same note, it is what almost everybody does, and it is what a choir does every time the tenors and the sopranos sing a unison line. A tuner that names the note will say the right letter. A tuner that shows you a target and a needle will show you nowhere near it.

The tuner picked the wrong octave. This is the interesting one. A sung note is not a clean tone. It is a fundamental plus a stack of harmonics, and low in the voice the fundamental is frequently not the loudest thing in the sound. On many vowels the second harmonic, one octave up, carries more energy than the fundamental does. A detector that finds the strongest periodicity and calls it the pitch will therefore name a note an octave above what you sang, and it will look completely confident doing it.

Your ear is not fooled by this, which is why the reading feels so wrong. Human hearing reconstructs a fundamental that is weak or entirely absent from the sound, an effect that has a name, the missing fundamental. Your ear hears the note you sang. The tuner heard the loudest partial.

You really are off. It happens, and it is worth knowing about, but it is the third thing to check rather than the first, and it does not usually arrive as an octave. Being genuinely out of tune looks like twenty or forty cents in one direction, not twelve hundred.

How to tell which one you are in, in about a minute

Sing it louder and more open. Go from a closed vowel to an open ah, and add some volume without pushing. That puts more energy into the fundamental. If the tuner's answer suddenly drops an octave, the tuner was wrong and you were not. Nothing about your singing changed by an octave in two seconds.

Give it a full second of steady tone. Almost every tuner averages over a window, so the first fraction of a note is attack and transient and the reading during it is meaningless. Sing, wait, then read. Half the wild readings people report are the tuner being asked to name a note that had not settled yet.

Look at the note name, not the needle. If the needle says you are forty cents sharp but the letter is a fifth away from the target, the needle is answering a question you did not ask. The letter tells you whether the tuner and you are even discussing the same note.

Check it against a key you can see. Play a piano key, sing along, and look at what the tuner names. If it names the key you are pressing while you sing an octave below, you have just watched it choose an octave.

Sing a slow scale upward and watch for a jump. A voice climbing a scale moves in semitones. If the reading leaps a whole octave between two adjacent notes and comes straight back, that is the detector changing its mind, not your voice. The one honest exception is a real register change, where a singer genuinely flips up an octave into head voice or falsetto, and that one does not come back down on the next note.

Why phone tuners in particular get voices wrong

Most tuners were built for guitars, and it shows.

A plucked string is close to the textbook case. Its fundamental is strong, its harmonics fall away above it, and the signal is stable for a second or more. Almost any pitch detection method works on it. The human voice is the opposite of a convenient signal. The fundamental can be weak or missing, the harmonic balance changes completely with the vowel, there is vibrato, and every singer slides into a note before settling on it.

So a tool that is genuinely accurate on a guitar can be routinely wrong on a voice, and the manufacturer is not lying to you. You are simply outside what it was measured on.

What this costs people, and it is not a small thing

The harm is not the wrong reading. It is the conclusion drawn from it.

Somebody sings, the tuner says a note they know is not what they sang, and they conclude that their ear cannot be trusted. That is precisely backwards. Their ear reconstructed a weak fundamental correctly, which is a sophisticated thing that hearing does automatically, and the machine did not. I have watched people describe themselves as tone deaf on the strength of a reading that was an artefact of harmonic balance on the vowel they happened to be singing.

If a tool tells you something surprising about your own voice, make it prove it before you believe it.

What I did about it in mine

This is the failure I spent most of my time on, so treat the following as interested rather than neutral.

The detector is YIN, and the part that matters here is a small bias that favours the longer lag when two candidate periods compete. The longer lag is the lower octave. A plain peak picker takes whichever lobe is tallest, which on a sung low note is often the harmonic rather than the fundamental. Biasing the choice, and then refining the position inside the chosen lobe on the unbiased curve, keeps the octave right without moving the cents.

Below 65 Hz it stops and says it cannot hear you, rather than guessing. That is deliberate. A confident wrong answer is worse than an admission.

And the measurements are published rather than described, both for the raw detector and for what the page actually shows, because the first thing anyone sensible wants from a tool like this is to know where it stops working.

The range test draws the raw pitch curve it heard underneath your result, which exists for exactly this reason. If it ever misreads an octave on you, you can see the jump instead of taking a number on trust.