Refuerzo SonoroDesign Lab
Prediction, not measurement

Reference: the physics behind every number

The platform is not a black box. These are all the models the simulation uses, with worked examples computed by the same code, and the assumptions that limit them.

Free-field propagation (point source)

Lp₂ = Lp₁ − 20·log₁₀(r₂ / r₁)
Lp₁level at reference distance r₁ (dB SPL)
r₂distance of interest (m)

Why?

Sound energy from a small source spreads over a sphere whose area grows with r². Intensity ∝ 1/r², pressure ∝ 1/r, so level falls 20·log₁₀ of the distance ratio: −6.02 dB every time the distance doubles.

Example

100 dB @ 1 m
2 m: 100 − 20·log₁₀(2) ≈ 94.0 dB
4 m: 100 − 20·log₁₀(4) ≈ 88.0 dB
8 m: 100 − 20·log₁₀(8) ≈ 81.9 dB
16 m: 100 − 20·log₁₀(16) ≈ 75.9 dB

Assumptions & limits

  • Far field of the source (r ≫ box size)
  • No reflections (outdoor free field, or direct sound only indoors)
  • No air absorption

Electrical power and level

ΔL = 10·log₁₀(P₂ / P₁)
P₁, P₂electrical power (W)

Why?

Power is an energy-like quantity, so it uses 10·log₁₀. Pressure, voltage and distance ratios are field quantities and use 20·log₁₀ (because power ∝ pressure²: 10·log₁₀(p²) = 20·log₁₀(p)). Doubling power is only +3 dB; doubling distance is −6 dB.

Example

1 W → reference
2 W → +3.0 dB
4 W → +6.0 dB
10 W → +10.0 dB
100 W → +20.0 dB

Assumptions & limits

  • Linear loudspeaker (no power compression)
  • Same impedance

SPL from sensitivity

SPL(r) ≈ S + 10·log₁₀(P) − 20·log₁₀(r)
Ssensitivity, dB SPL @ 1 W / 1 m
Pinput power (W)
rdistance (m)

Why?

Sensitivity gives the level for 1 W at 1 m. Add the power gain (10·log₁₀) and subtract the distance loss (20·log₁₀).

Example

S = 97 dB, P = 400 W
1 m: 97 + 10·log₁₀(400) ≈ 123.0 dB
20 m: 123.0 − 20·log₁₀(20) ≈ 97.0 dB

Assumptions & limits

  • Simplified model — not a measurement
  • Real drivers compress several dB near rated power
  • Sensitivity may be half-space or 2.83 V: check the datasheet conditions

Combining sources (energetic sum)

L_total = 10·log₁₀( Σ 10^(Lᵢ/10) )
Lᵢlevel from each source at the listener (dB)

Why?

Decibels are logarithmic and cannot be added arithmetically. Converting to energy, adding and converting back gives the average level of uncorrelated (incoherent) sources.

Example

90 dB + 90 dB = 93.0 dB (not 180 dB)
90 dB + 84 dB = 91.0 dB
90 dB + 80 dB = 90.4 dB
4 × 90 dB = 96.0 dB

Assumptions & limits

  • Does NOT model phase, interference, comb filtering or coherence
  • Two coherent in-phase sources could reach +6 dB at some frequencies and cancel at others
  • The map is a spatial average estimate, not a frequency-resolved prediction

Nominal directivity model (estimate)

A(u) = max(−6·u², floor), u = γ / b(ψ), 1/b² = cos²ψ/(H/2)² + sin²ψ/(V/2)²
γangle between the aim axis and the listener
ψdirection around the axis (0 = horizontal plane)
H, Vnominal coverage angles (−6 dB points)
floorlowest attenuation used off-axis / behind (default −20 dB)

Why?

A datasheet “90° × 60°” means the level is 6 dB below on-axis at ±45° horizontally and ±30° vertically. With only those numbers we interpolate a smooth elliptical pattern. When the manufacturer publishes beamwidth per frequency, those values replace H and V.

Example

u = 0: 0.0 dB
u = 0.5: -1.5 dB
u = 1: -6.0 dB
u = 1.5: -13.5 dB
u = 2: -20.0 dB

Assumptions & limits

  • NOMINAL COVERAGE — not measured polar data
  • Real patterns change with frequency and have lobes
  • Front-to-back ratio unknown: floor is an assumption

Coverage width

w = 2·d·tan(θ / 2)
θcoverage angle (deg)
ddistance along the axis (m)

Why?

Simple trigonometry of the −6 dB coverage triangle. At the edge of that width the level is ~6 dB below the on-axis level at the same distance.

Example

θ = 90°, d = 10 m → w = 20.0 m
θ = 60°, d = 10 m → w = 11.5 m
θ = 40°, d = 30 m → w = 21.8 m

Assumptions & limits

  • Measured perpendicular to the aim axis
  • Nominal angle, broadband

Speed of sound vs temperature

c ≈ 331 + 0.6·T (exact ideal gas: c = 331.3·√(1 + T/273.15))
Tair temperature (°C)

Why?

Sound travels faster in warmer air. 343 m/s corresponds to about 20 °C. The simulation uses the ideal-gas expression; the linear one is an easy approximation.

Example

0 °C: 331.0 m/s (linear), 331.3 m/s (exact)
20 °C: 343.0 m/s (linear), 343.2 m/s (exact)
35 °C: 352.0 m/s (linear), 351.9 m/s (exact)

Assumptions & limits

  • Dry air; humidity adds < 0.5 %

Delay speakers

delay(ms) = Δd / c × 1000 (+ precedence offset)
Δdpath difference main → listener minus delay → listener (m)
cspeed of sound (m/s)

Why?

A delay speaker is closer to the listener than the main system. Without delay its sound would arrive first and the image would pull toward it; with too little delay you hear echoes. Delaying it by the travel-time difference (plus ~5–15 ms) lets the main system arrive first (precedence / Haas effect) while the delay adds level.

Example

Δd = 17.2 m, c = 343 m/s → 50.1 ms
Δd = 40 m → 116.6 ms

Assumptions & limits

  • Only exact at the reference listening position
  • Temperature changes c and therefore the ideal delay

Line sources: cylindrical vs spherical spreading

ideal line: −10·log₁₀(r₂/r₁) point: −20·log₁₀(r₂/r₁) transition r_t ≈ H²·f / (2c)
Harray height (m)
ffrequency (Hz)

Why?

An infinitely long coherent line radiates a cylinder: −3 dB per doubling of distance. A real array is finite, so it behaves like a line only close to it (near field) and like a point source (−6 dB/dd) beyond the transition distance, which grows with frequency and with the square of the array height.

Example

H = 4 m, 1 kHz → r_t ≈ 23.3 m
H = 4 m, 125 Hz → r_t ≈ 2.9 m
H = 8 m, 4 kHz → r_t ≈ 373 m

Assumptions & limits

  • Order-of-magnitude estimate for a straight, uniformly driven array
  • Curved / shaded real arrays must be predicted with the manufacturer’s data
  • This platform does not apply −3 dB/dd to real arrays

Obstacles: barrier diffraction (Maekawa)

ΔL = 10·log₁₀(3 + 20·N), N = 2δ / λ
δextra path length over/around the obstacle edge (m)
λwavelength c/f (m)

Why?

Sound bends around obstacles. The shadow is deeper when the detour is long compared with the wavelength: high frequencies are strongly shadowed, low frequencies much less. The map uses the selected band (1 kHz by default).

Example

δ = 0.1 m, 1 kHz: N ≈ 0.6 → 11.7 dB
δ = 0.1 m, 125 Hz: N ≈ 0.1 → 6.5 dB
δ = 1 m, 4 kHz: → 24.0 dB (capped at 24 dB)

Assumptions & limits

  • Empirical fit for a thin screen
  • Single-edge path; reflections ignored
  • Capped at 24 dB

Air absorption (ISO 9613-1)

A_air = α(f, T, RH) · r
αattenuation coefficient (dB/m), ISO 9613-1

Why?

Air converts sound energy to heat, mostly at high frequency. Negligible in a classroom, several dB at 4–8 kHz across a festival field.

Example

1 kHz, 20 °C, 50 % RH: 0.5 dB per 100 m
4 kHz, 20 °C, 50 % RH: 3.0 dB per 100 m
8 kHz, 20 °C, 50 % RH: 10.5 dB per 100 m

Assumptions & limits

  • Pure-tone coefficient at band centre
  • Applied only when a band is selected

Distributed (ceiling) speaker spacing

D = 2·h·tan(θ/2); s = D (edge-to-edge), D/√2 (minimum overlap), D/2 (centre-to-centre)
hceiling height − ear height (m)
θcoverage angle (deg)

Why?

Each ceiling speaker covers a circle at ear height. Closer spacing trades more units for more uniform level (±1–2 dB instead of ±6 dB).

Example

Ceiling 3.5 m, seated 1.2 m, 90°: D = 4.6 m
edge-to-edge 4.6 m, min. overlap 3.3 m, centre-to-centre 2.3 m

Assumptions & limits

  • Square grid
  • Nominal conical coverage

What this platform is — and isn’t

Predictions are simplified estimates (direct sound, energetic summation, nominal directivity). They do not replace the manufacturer’s own prediction software, room-acoustic modelling or on-site measurement. Manufacturers provide their own prediction tools for their systems, and room-acoustic software models reflections and reverberation. Use this lab to explore options, understand trade-offs and communicate a concept; verify the final design with the manufacturer’s data and on-site measurement.

Precision

Levels are shown with one decimal at most (none in summaries), distances with one decimal. Inputs like datasheet max SPL are typically ±1–3 dB between measurement methods; more digits would be false precision.