This page is the prerequisite primer for all SonaLab FAQ documents. You need no acoustics knowledge — just 10–15 minutes of hands-on play. We will start from "what is sound" and work our way to every line on the spectrum, and in the end you will discover: every number in the app is a part of your voice.

When you speak or sing, your vocal folds vibrate, opening and closing rapidly in your throat. Each closure gives the air a push, like dropping pebbles into a pond — ripples spread outward and finally hit your eardrum, and that is when you "hear" sound.

If your vocal folds open and close 200 times per second, that is 200 Hz. The faster they open and close, the "sharper" (higher) the sound; the slower, the "duller" (lower). Human ears hear roughly 20 Hz – 20,000 Hz. When you sing, your pitch is the vibration frequency of your vocal folds — in vocal terminology, F0 (fundamental frequency).

Say "ah—" softly and then belt it: the difference is amplitude — the air is pushed with more force. Loudness is measured in dB (decibels), a logarithmic scale — multiplying amplitude by 10 only adds 20 dB. This matches how our ears perceive loudness (it is not linear).

If sound were a single sine wave, every instrument and every voice would sound identical — a monotonous "beep". Real sound is many frequencies vibrating at once: besides the fundamental F0 there are its integer multiples 2×F0, 3×F0… which we call harmonics H1 (=F0), H2 (=2×F0), H3 (=3×F0). How strong each harmonic is decides the timbre.

White light passes through a prism and becomes a rainbow — each color is a different component of the white light. Similarly, FFT "takes apart" a complex waveform and tells us "how strong each harmonic is". The result, drawn out, is the spectrum: x-axis = frequency, y-axis = energy (dB). The waveform and the spectrum are two views of the same sound.

Remember? Harmonics are integer multiples of F0: H1 = 1×F0, H2 = 2×F0, H3 = 3×F0… So when F0 changes, all the spectral lines shift together (like an accordion). The app uses the YIN algorithm to find this "most basic vibration period" in the waveform — in other words, it finds F0.

The height of a spectral line = how strong that harmonic is. Let's single out H1 (at F0) and H2 (at 2×F0): strong H1 → round, thick waveform (like male voices and chest voice); strong H2 → sharp, bright waveform (like female voices and head voice). On the spectrum it simply comes down to which of the two lines is taller.

This is a core metric in vocal analysis. Why? Thick vocal folds (TA contraction) → low harmonics are strong → H1 above H2 → H1−H2 is negative; thin vocal folds (CT stretch) → high harmonics are strong → H2 above H1 → H1−H2 is positive. The ideal range in the app is −3 dB to +8 dB (the green zone) — too thick or too thin both lose points.

As the sound from the vocal folds passes through the mouth and pharynx it gets "colored": formants are the frequency regions amplified by the vocal tract (F1, F2, F3…), appearing as "humps" on the spectral envelope. The key rule: lower larynx → longer vocal tract → lower F1. A low F1 means a "deep", grounded sound — this is the core of the app's resonance depth score (30% weight).

In a trained singer's voice there is a concentrated "singer's formant" around 2.5–3.5 kHz — an acoustic "lens" created by opening the pharynx. SPR quantifies it: SPR = 10×log₁₀(high-band energy / low-band energy). The closer to 0, the brighter and more "metallic" the voice. An orchestra's energy happens to dip right at 2.5–3.5 kHz — so the singer's formant lets a voice cut through the band.

The previous parameters can all be "read" off the spectrum. But NAQ (Normalized Amplitude Quotient) cannot — it describes how the vocal folds close. The sound is colored by the vocal tract before reaching the mic, so the app models the tract with LPC, then performs inverse filtering to "peel off" the tract and recover the glottal airflow pulse, then measures: total airflow (A_AC) ÷ closure speed (d_peak).

Each chapter so far looked at one parameter. But in real sound they exist together and affect each other. This workbench puts all 6 parameters back on one spectrum: pick a phonation mode, or freely drag every slider, and watch how the spectrum and the 4 dimension scores (depth / compression / thickness / throatiness) interact.