Sounding

Make an acoustic portrait of a place instead of a perspective drawing. Build a curved room — a cylinder, an ellipse with its two foci, a Serra torqued ellipse or torqued spiral in weathering steel, a dome on a drum with or without an oculus, a barrel vault, a tapered drum, or a shoebox for comparison — on gravel, grass, a plaza or a floor, open to the sky or roofed, with a slot entrance. Place sounds in it (a voice, a clap, footsteps, a choir, a struck steel plate, rain, your own microphone or a file) and a listener you walk with the arrow keys. A curvature-aware ray tracer (it reflects off the true curve, not the facets) renders a binaural impulse response per sound and paints a floor map of where the room gathers sound; closed-form findings say why — focal distances, the ellipse's constant path, the dome's centre, the whispering-gallery condition, the corridor's note — and Bessel-zero modes give the drum's low end. Mix it all on a full board you can drive without looking: numbered strips, faders that speak in dB, meters read back in words, mute/solo/EQ/room-send by key or by voice. Record the portrait as a stereo WAV, export the impulse response, the written portrait, the room, the Rhino mesh. Reach for Sounding to build a room and walk a listener through it with the sounds you choose; reach for Caustic (/caustic) to read a room's acoustics against theory, run its wave tank or bring in your own Rhino model. The Serra rooms are studies, not surveyed dimensions. Client-side; Claude can optionally rewrite the portrait as prose (sign-in for that button only).

Method: A ready-made room: a shape, its dimensions and materials. The Serra presets are at the scale of the Dia Beacon Torqued Ellipses (about 13 ft tall, 36 × 22 ft). Choosing one in this list changes nothing yet: press Load this room (or Enter on the list). Every number stays editable below.
Method: The plan family. Round plans are lofted as parametric surfaces with analytic normals, so the trace reflects off the true curve, not the facets.
ft
Method: Half the plan dimension along east–west, from the centre to the wall.
ft
Method: Half the plan dimension along north–south. Equal to the east–west semi-axis, an ellipse is a circle.
ft
Method: Floor to the top of the wall (the rim of an open piece, the springing of a roof).
°
Method: Serra's torque: the top ellipse is the bottom one turned by this angle; the wall is the ruled surface lofted between them.
ft
Method: The rim is pushed outward (positive) or inward (negative) by this much, radially. A wall that leans out throws reflections up and, with the top open, away.
°
Method: A gap cut through the wall on the east side, as degrees of plan angle. Sound escapes through it and the ground outside reflects some of it back.
Method: Open: the top is sky (α = 1 in Sabine/Eyring; rays that miss every surface escape). Flat, dome and cone roofs are traced as surfaces with their own material.
segs
Method: Facets around the plan. The trace reflects off the interpolated normal, so even 32 segments focus like the smooth surface; higher values only sharpen the drawing and the floor map's edges.

Floor / ground 3,574 sq ft (332 m²) · Walls 1,270 sq ft (118 m²) · Sky 817 sq ft (76 m²) · Air 8,133 cu ft (230 m³)

Method: Per-octave absorption and mid-band scattering coefficients for this surface. Coefficients source: Metal-panel rows after Egan, Architectural Acoustics (1988) tables & Vorländer, Auralization (2008); scattering per Cox & D'Antonio (2009) for a plane plate.
Method: Per-octave absorption and mid-band scattering coefficients for this surface. Coefficients source: Outdoor ground rows after Harris, Handbook of Acoustical Measurements and Noise Control (1991) & Embleton (1996), 'Tutorial on sound propagation outdoors'.
×
Method: Multiplies every surface's scattering coefficient. At each wall hit a ray scatters (Lambert) with that probability, else reflects specularly. Smooth steel ≈ 0.03 — nearly all specular, which is why it focuses.
Method: Stochastic rays per channel. More rays mean a smoother late tail and a steadier floor map; the seed is fixed, so the same room and setting give the same impulse response every time.

Each step re-runs a quick trace so the room changes under your feet. Off, press Run after moving.

Not traced yet
Map method: Per listening point: traced reflected energy arriving within 50 ms of the direct sound, in dB over the direct energy at that distance (mid band, 500 Hz–1 kHz). A concave wall's focus is where this is highest.
Plan of torqued ellipse. No floor map yet. Listener at -13 ft east, 0 ft north, facing 270 degrees. Channel 1 Voice at 13 ft east, 0 ft north. Channel 2 Clap at 13 ft east, 0 ft north. Channel 3 Footsteps at 5 ft east, -2 ft north. Channel 4 Choir at -4 ft east, -5 ft north.Two foci1Voice2Clap3Footsteps4ChoirListenerN10 ft (3.0 m)

The floor map traces Voice to — ear-height points. Click a channel on the plan to analyse it instead. Open sky and the entrance let sound out; the ground outside gives some back.

Walk the listener. With this panel focused: up and down arrows step forward and back 3 feet; left and right arrows turn 45 degrees; W A S D step north, west, south, east; Shift triples a step; L speaks where you stand.

Listener

Listener 13 ft west, 0 ft north of centre, facing west. Nearest wall 1.8 ft (0.6 m).

facing west (270°)

Each step is 3 ft (0.9 m). Keys: arrows, W A S D, L.

Numbers

Whole room by formula; at the listener by the trace of Voice. The top is open: the formulas count the sky as absorption and give a room-average that means little here — read the trace.

RT60 mid, Eyring0.08 s — method: Eyring: T = 0.161·V / (−S·ln(1 − ᾱ) + 4mV), mean of the 500 Hz and 1 kHz bands, over the mesh's measured areas (526 m² incl. any sky) and 230 m³.
RT60 mid, Sabine0.12 s — method: Sabine: T = 0.161·V / (A + 4mV), A = Σ S·α. Over-estimates in a room with one very absorbent surface (an open top).
Schroeder frequency38 Hz — method: f_s = 2000·√(T/V). Above about twice this the ray trace is the authority; below it the room rings in modes.
T30 at listener— — method: Schroeder backward integration of the traced centre-of-head impulse response, −5 to −35 dB, ×2 (ISO 3382-1), mean of 500 Hz and 1 kHz.
Early decay time— — method: EDT: the first 10 dB of the Schroeder decay, ×6 (ISO 3382-1).
Clarity C50— — method: 10·log10 of energy in the first 50 ms after the direct over the rest (speech clarity; > 0 dB is clear).
Clarity C80— — method: As C50 with an 80 ms window (music clarity).
Definition D50— — method: Share of the energy arriving within 50 ms of the direct sound.
Room gain— — method: 10·log10((direct + all reflected energy) / direct), mid band, from the trace's receiver histogram vs the analytic direct tap.
Early gain (focus)— — method: 10·log10((direct + reflected energy in the 50 ms after the direct) / direct). A concave wall aimed at you pushes this up; open air is 0 dB.
Crept along a wall— — method: Share of the reflected energy whose ray made four or more consecutive grazing specular hits on one curved surface before arriving — the whispering-gallery path.
Left to the sky— — method: Share of the launched sound energy that left through an opening before being absorbed.
Direct path— — method: Straight-line distance talker → listener and its travel time at the set temperature; blocked when a wall lies between.
Echo— — method: Dietsch-style heuristic: a 1 ms spike after 50 ms whose energy stands out against the early sum. An interface flag, not an ISO number.
Per-octave-band reverberation, Eyring and Sabine for the whole room, and the traced T30 at the listener.
Band1252505001k2k4k
Eyring0.20.10.10.10.10.1
Sabine0.20.10.10.10.10.1

What the curves do

  • Two foci. The ellipse's foci sit 14.2 ft (4.3 m) either side of the centre. Every path from one focus to the wall and on to the other is exactly 36 ft (11 m) long, so a word spoken at one arrives at the other as one loud, late echo 32 ms after the direct sound. This holds at floor level only — the torqued wall leans and turns above it, so the focus smears upward and weakens with height. The talker IS at a focus and the listener at the other: this is the whisper-gallery stand. — method: Ellipse with semi-axes a, b: foci at ±√(a² − b²); reflected path length from focus to focus = 2a for every wall point (the defining property), delay = 2a/c.
  • Open to the sky. 817 sq ft (76 m²) of the room is sky, which absorbs everything it is given. Reflections that climb the walls leave; what you hear is the first bounce or two off the steel and the ground, then quiet — bright, close and dry, not reverberant. The floor material decides most of what remains. — method: The opening enters Sabine/Eyring as an area of α = 1 (the open-window sabin); the trace simply lets rays that miss every surface escape.
  • Not a whispering gallery — yet. The wall is smooth and hard enough, but the talker stands 4.9 ft (1.5 m) from it. Move within about 3 ft (1 m) of the wall and aim the beam along it, and the wall will carry a whisper round the room. — method: Whispering-gallery condition (Rayleigh, 1878): grazing specular hops along a smooth concave wall; needs low absorption and scattering and a source within a wavelength or two of the wall (here < 3 ft). Measured by the trace as the share of reflected energy with ≥4 consecutive grazing hits on one curved surface.

Low end

No closed form for an ellipse: these are the modes of the same-area cylinder (radius 4.3 m). An ellipse splits each lobed pair into two frequencies; expect the real rows to straddle these.

  • 2 lobes · ring 123 Hz
  • 4 lobes · ring 139 Hz
  • floor–ceiling 143 Hz
  • ring 148 Hz
  • 2 lobes · ring 1 · height 149 Hz
  • 6 lobes · ring 153 Hz
  • 4 lobes · ring 1 · height 158 Hz
  • 0 lobes · ring 1 · height 165 Hz
  • 8 lobes · ring 167 Hz
  • 2 lobes · ring 267 Hz
  • 6 lobes · ring 1 · height 168 Hz
  • 8 lobes · ring 1 · height 180 Hz
  • 2 lobes · ring 2 · height 180 Hz
  • 10 lobes · ring 181 Hz

The portrait

An acoustic portrait of torqued ellipse (serra)

This is a torqued ellipse 36 ft (11 m) by 22 ft (6.8 m) at the ground, its rim turned 55 degrees and leaning 2 ft (0.6 m) outward, 13 ft (4 m) tall, open to the sky. Its walls are weathering steel plate (cor-ten), 2 in; underfoot is gravel, loose, 4 in. It holds 8,133 cubic feet of air (230 m³), and 817 square feet (76 m²) of its surface is sky.

The listener stands 13 ft west, 0 ft north of centre, 4.9 ft up facing west.

Two foci. The ellipse's foci sit 14.2 ft (4.3 m) either side of the centre. Every path from one focus to the wall and on to the other is exactly 36 ft (11 m) long, so a word spoken at one arrives at the other as one loud, late echo 32 ms after the direct sound. This holds at floor level only — the torqued wall leans and turns above it, so the focus smears upward and weakens with height. The talker IS at a focus and the listener at the other: this is the whisper-gallery stand.

Open to the sky. 817 sq ft (76 m²) of the room is sky, which absorbs everything it is given. Reflections that climb the walls leave; what you hear is the first bounce or two off the steel and the ground, then quiet — bright, close and dry, not reverberant. The floor material decides most of what remains.

Not a whispering gallery — yet. The wall is smooth and hard enough, but the talker stands 4.9 ft (1.5 m) from it. Move within about 3 ft (1 m) of the wall and aim the beam along it, and the wall will carry a whisper round the room.

Below about 0 Hz the room stops behaving like rays and rings as a whole; its lowest standing waves sit at 23 Hz (2 lobes · ring 1), 39 Hz (4 lobes · ring 1), 43 Hz (floor–ceiling 1). No closed form for an ellipse: these are the modes of the same-area cylinder (radius 4.3 m). An ellipse splits each lobed pair into two frequencies; expect the real rows to straddle these.

The portrait is mixed from 4 channels: Voice (speaking voice, 26 ft to the east, 5.2 ft above the floor, 0 dB); Clap (hand clap, 26 ft to the east, 5.2 ft above the floor, 0 dB); Footsteps (footsteps, 18 ft to the east, 0.3 ft above the floor, minus 6 dB); Choir (choir chord, 10 ft to the south-east, 4.6 ft above the floor, minus 3 dB).

Written by the tool from its own numbers. Claude can rewrite it as prose:

Mixing board. Keys while the board has focus: 1 – 8 select channel 1–8 (speaks its strip); ↑ / ↓ fader of the selected channel ±1 dB (Shift: ±3 dB); M mute / unmute the selected channel; S solo / unsolo the selected channel; R read the meters; D describe the selected channel; Space play / stop the mix; A audition the selected channel alone through the room; 0 select the master strip; ? read this key list.

Mixing board

Method: A Rosenberg glottal pulse train at ~118 Hz with syllable-by-syllable pitch declination, 5.3 Hz vibrato and per-cycle jitter, differentiated and fed in parallel through two RBJ band-pass formant resonators whose centres glide through the vowels ah, ee, oo, eh, ah (F1/F2 = 730/1090, 270/2290, 300/870, 530/1840 Hz) in five 0.35 s syllables, with a little aspiration noise through the same formants.
dB
Method: Channel level in decibels, after the EQ and before the room. 0 dB plays the sound at the level the trace gives it at the listener's distance.
%
Method: How much of the traced reflections this channel gets. The direct sound always plays; 100 % adds every reflection the trace found, 0 % is the voice alone at its distance.
Method: White noise with a 0.3 ms attack and ~9 ms exponential decay, band-passed through two RBJ biquads at 1.6 and 2.9 kHz, mixed with two 1–3 ms delayed copies of itself for the cupped-hand resonance, plus a quiet 70–210 Hz pitch-dropping sine thump.
dB
Method: Channel level in decibels, after the EQ and before the room. 0 dB plays the sound at the level the trace gives it at the listener's distance.
%
Method: How much of the traced reflections this channel gets. The direct sound always plays; 100 % adds every reflection the trace found, 0 % is the voice alone at its distance.
Method: Five impacts 0.55 s apart with ±25 ms jitter; each is an 8 ms band-passed (2.5–4.5 kHz) noise heel click followed by a sine thud gliding from ~180 Hz to 80 Hz and decaying in ~60 ms, with a low-passed noise body, alternating heavy and light steps.
dB
Method: Channel level in decibels, after the EQ and before the room. 0 dB plays the sound at the level the trace gives it at the listener's distance.
%
Method: How much of the traced reflections this channel gets. The direct sound always plays; 100 % adds every reflection the trace found, 0 % is the voice alone at its distance.
Method: Four voices at 174.6, 220, 261.6 and 349.2 Hz, each the mono sum of four unison copies detuned by −7 to +8.5 cents; every copy is a band-limited sawtooth (harmonics with 1/n amplitude up to 4 kHz, computed by phasor recurrence) with its own 5–6 Hz ±6-cent vibrato, run through an 'ah' formant pair (RBJ band-passes near 730/1090 Hz), with a 0.3 s onset, 0.6 s release and a trace of breath noise.
dB
Method: Channel level in decibels, after the EQ and before the room. 0 dB plays the sound at the level the trace gives it at the listener's distance.
%
Method: How much of the traced reflections this channel gets. The direct sound always plays; 100 % adds every reflection the trace found, 0 % is the voice alone at its distance.
Master
dB
Method: The whole mix's level before the limiter.
%
Method: Scales every channel's room send together — one knob to hear the room come and go.

Headphones. Off plays the centre-of-head impulse response to both ears.

A 20:1 compressor at −3 dBFS so a focused clap cannot clip.

Method: Adds a sound to the board at the listener's position; move it from its strip or by clicking the plan.

Headphones for the two-ear rendering. The trace is seeded (seed 20261002), so the same room and stand give the same impulse response every time; a live re-trace uses 6,000 rays and the Run button uses your chosen quality.

Sounding · Good Tools