Arau-Puchades treats dB/m as m and ignores temperature/humidity

[rdmiller]

Arau-Puchades adds 4 * airAttenuation([freq])[0] * V — raw dB/m times volume — and calls airAttenuation with default 20 °C / 40 % RH, ignoring this.temperature / this.humidity. Sabine/Eyring at least convert (wrongly, # companion). Arau mixes dB/m with m² in the denominator.

Found on varese-dev. arau-puchades-air-absorption.spec.ts checks that 4*m*V is the shape of the term with a hand-waved m = 0.01; it never reads index.ts.

Cause

const airabsterm = 4 * airAttenuation([freq])[0] * v; // dB/m · m³
const rt =
  ((K*v) / (-A * Math.log(1-αx[f]) + airabsterm)) ** (Ax/A) *
  ((K*v) / (-A * Math.log(1-αy[f]) + airabsterm)) ** (Ay/A) *
  ((K*v) / (-A * Math.log(1-αz[f]) + airabsterm)) ** (Az/A);

-A ln(1-α) is m² (sabins). 4 α_dB V is m²·dB. At 4 kHz, α_dB ≈ 0.02 m⁻¹·dB, so 4αV ≈ 0.08 V — about 4.3× the correct 4mV (m = α_dB / 4.343). High-frequency Arau RT is too short vs Sabine/Eyring on the same room (and those are already too long from the /20).

The three axial factors each include the full 4mV, then geometric-mean. For a cube that is equivalent to adding it once; for a long hall it is not the usual “add 4mV to the total equivalent absorption once.”

Projection uses local surface.triangles + orthographic P, not matrixWorld. Rotated rooms shear the axial areas.

Suggested fix

Share one mValues[i] (intensity Np/m, temp/RH) with Sabine/Eyring. Add 4mV once to each axial denominator, or document the geometric-mean choice. Test a 10×10×10 m box, α = 0.2, air off: Arau = Eyring = Sabine (uniform α, cube). Air on: all three drop, Arau not 2× more than Sabine.

Acceptance criteria

• Arau airAttenuation is called with this.temperature, this.humidity
• Units of the air term are m² (same as -A ln(1-α))
• Cube + uniform α, air off: |T_Arau − T_Eyring| / T_Eyring < 1 %