loading...
Press Play or choose a preset.
running...
How to use
Choose a preset vowel or move the tongue and lip sliders. The formants in the readouts (and the green markers in the first graph) change with the tube shape; the white dot in the vowel chart shows where the sound sits among the reference vowels. Press Sound to hear it, or Sing through vowels for a glide.
- Raise the pitch f0: the harmonics spread apart while the filter shape stays the same.
- Shorten the tract (child preset): all formants rise.
- Choose which resonance the standing pressure wave shows (F1, F2, F3 or f0).
Drag to orbit, wheel or pinch to zoom, or press Guided tour. Quality switches High/Low (also ?q=low).
The science
The source-filter theory of speech says a vowel is a voiced source (the vocal folds) shaped by a filter (the vocal tract). The folds release air in pulses at the pitch f0, so the source spectrum is a comb of harmonics at multiples of f0 that falls with frequency. The tract is modelled as 44 short cylindrical sections of area A_i, each of length dx = L/44, with plane waves inside and a complex wavenumber k(1 - j eps) that includes wall and viscous loss. Each section is a 2x2 chain matrix; the product from glottis to lips is closed by a piston-in-a-baffle radiation impedance:
[P; U]_glottis = prod_i [[cos k dx, j Z_i sin k dx],[j sin k dx / Z_i, cos k dx]] [P; U]_lips, Z_i = rho c / A_iH(f) = U_lips / U_glottis = 1 / (C Z_rad + D), Z_rad = (rho c/A_l) ( (ka)^2/2 + j 0.6133 ka )
The peaks of |H| are the formants F1, F2, F3 and so on. A uniform tube closed at the glottis and open at the lips has F_n = (2n - 1) c / (4 L), about 500, 1500 and 2500 Hz for 17.5 cm. Narrowing the tract changes the resonances: F1 rises with mouth opening and falls with a high tongue or protruded lips; F2 follows the position of the constriction (high for front vowels such as /i/, low for back vowels such as /u/ and /c/). Lip rounding and protrusion lengthen the tube and lower every formant. The area function is a Gaussian dip (the tongue) in a 5 cm2 tube with a lip-opening taper, so one constriction stands for the whole tongue. The radiated sound is the source spectrum times |H| times a radiation term that rises with frequency; the Sound button plays that harmonic sum.
Honest simplifications. Plane-wave (one-dimensional) acoustics, a bent circular-section tube (real cross sections are not round and the nasal tract and sinus cavities are absent), a single tongue constriction (so the mid vowels /E/ and /ae/ are only approximated), frequency-independent loss, a Rosenberg-type glottal pulse without source-tract interaction, and no noise. The area-to-formant map is not unique: different shapes can give the same F1 and F2. Reference vowel formants are the Peterson and Barney (1952) male averages.
In-page checks. The neutral tube F1 and F2 are compared with the quarter-wave formula including the lip end correction; a two-tube /a/ model (1 cm2 back, 8 cm2 front, ideal open end) is compared with the exact root of tan(k L1) tan(k L2) = A2/A1; and at 20 Hz the volume velocity at the lips must equal that at the glottis (|H| = 1).