Sign in

Caustic

Open in new tab

Build an acoustic portrait of a curved place instead of a perspective drawing. Load a rotunda, dome, whispering ellipse or a torqued-steel ring in the manner of Serra — or import your own Rhino model — then stand in it: a 3-D ray tracer that reflects off the true curve of the rooms it builds (analytic normals, not facets) shows where sound gathers and what echoes it throws, a 2-D wave tank shows the diffraction and standing waves rays cannot, and a full binaural mixing board lets you place and mix a four-part choir in the room, by keyboard, by voice or by screen reader. Prose portrait, wave tank, impulse-response export. Reach for Caustic to read a room's acoustics against theory — where sound gathers, the echoes it throws, a wave tank, your own Rhino model — and to mix a choir in it; reach for Sounding (/sounding) to build a room from its ground to its roof and walk a listener through it. The Serra-manner rooms are studies, not surveyed dimensions. The engines run in your browser and need no sign-in; Voice mode uses your browser's own speech recognition, which in Chrome and Edge sends speech to the browser maker.

on Channel 5, Hand clap
Method: The seat whose ears the board plays and the numbers describe.
Method: How many rays are traced for the impulse responses (6,000, 24,000 or 80,000) and how long the tail is followed. More rays make the numbers steadier and take longer.

Starting the first render…

Method: A ready-made room with the singers and seats already placed. Choosing one in this list changes nothing yet: press Load this study (or Enter on the list) to replace the room and the positions. The mixing board is left as you set it.

A spherical dome whose centre of curvature sits near the floor — a domed hall.

Listen for: The dome's centre of curvature is at about floor level, so a voice anywhere near the middle is gathered by the whole ceiling and thrown back to the same spot. Loading this room puts you at the centre with the portrait pointed at the clap (channel 5, which starts muted and is brought in once the room is rendered), so the focus is the first thing you hear: the clap comes back about 50 ms later, far above the sound around it. Stand in the centre and the choir will sound huge and a little late; step aside and it clears. Try scattering the dome surface (the roughness slider) and hear the focus fall apart — that is what coffers are for. (The Pantheon's own dome is a hemisphere, so its centre of curvature is up at the springing of the dome, about 71 ft (22 m) above the floor, not down at the floor as it is here.)

Method: The wall's shape from above. A circle focuses sound at its centre; an ellipse at two points; a stadium and a squircle in between.
ft
Method: Size sets how long every echo takes to come back: a wall 20 ft away returns sound after about 35 milliseconds, one 60 ft away after about 106.
Method: The curve is built from flat facets, and sound reflects off a smoothed normal so a facet does not scatter it. Below about 64 facets a tight curve starts to behave like a polygon.
ft
Method: Height decides how much sound goes up to the ceiling instead of staying between the walls — and how long the ceiling's echo takes.
%
Method: Leans the wall by making the whole top outline bigger or smaller than the bottom, by the same proportion east–west and north–south, so a room whose top is already different in the two directions keeps that difference; the dial shows the east–west size. A wall that leans in is an overhang: it throws its echo down onto the people below it. One that leans out throws it up and away, toward the sky or the ceiling. Once the top is more than 2% bigger or smaller than the bottom, the tool stops offering its closed-form echo predictions for a round room, because no single radius is true at every height.
°
Method: Rotates the top outline against the bottom. It changes the room even on a circle: the wall becomes a pinched surface whose middle is narrower than its floor, by a factor cos(twist ÷ 2) — 87% at 60°, 71% at 90° — so the air volume, the echo times and the focusing all change. On an ellipse the twist also makes the echoes wander. Past a twist of about 23° the tool stops offering its closed-form echo predictions for a round room, because no single radius is true at every height.
ft
Method: Slides the top outline sideways, so the wall leans one way: an overhang above you on one side.
ft
Method: As above, along the other axis.
Method: The ceiling's surface. This is the biggest single change you can make to how a room sounds.

A spherical dome focuses sound to a point at its centre of curvature. If you stand near that point, the dome returns your own voice to you, and so loud it can feel like the room is talking.

ft
Method: How high the middle of the ceiling is above the top of the wall. For a spherical dome the centre of curvature, where it gathers sound, is at (wall height + rise − (radius² + rise²) ÷ (2 × rise)) above the floor: far below the floor for a shallow rise (harmless), climbing as the rise grows until it reaches the top of the wall when the rise equals the radius, a hemisphere. Whether it passes through head height on the way depends on how tall the wall is.
%
Method: Removes the crown, where a dome's focus is strongest. Sound that goes up through it leaves and does not come back.

A hymn-style chorale in C major for soprano, alto, tenor and bass, 80 bpm, sung on “ah”. The default.

Method: The same five presets as the Acoustician's, four parts of one 12-second piece each, synthesised in the browser (no recordings), one part on each of channels 1 to 4 — the four people the plan stages. Choosing in this list changes nothing yet: press Load this preset (or Enter on the list) to change what those four channels play, their names and their two groups. Faders, EQ, sends, mutes, the room and the four probe channels stay as they are. The chant singers beam forward like a talker; the instruments use the default instrument directivity, which is not measured for them.
Method: Pick a singer or a seat; the sliders below move whichever you pick. In the plan you can also drag it, or Tab to it and use the arrow keys.
ft
Method: Plan position, east is positive. Moving someone changes every path of sound between them and everyone else, so the impulse responses are re-rendered when you stop.
ft
Method: Plan position, north is positive.
ft
Method: A standing singer's mouth is about 5 ft up and a seated listener's ears about 4 ft. Height matters most under a dome, where it decides whether you are near the focus.
°
Method: Which way the singer faces. A voice is close to omnidirectional at low frequencies and quite forward-directed above 1 kHz, so turning around changes the bright part of the sound much more than the bass.
Method: The seat the impulse responses and the numbers are for: the board's room sound is what your ears would hear sitting here.

The room: plan, section, rays, wave tank and 3-D model

A circular room 49 ft across and 13 ft high, under a spherical dome rising 15 ft. You are at Centre — under the dome, at the centre, 4.9 ft up, facing 90 degrees counter-clockwise from east. Soprano: 20 ft away, at your 8 o'clock; 18 ft west and 10 ft south of the centre. Alto: 20 ft away, at your 7 o'clock; 7 ft west and 19 ft south of the centre. Tenor: 20 ft away, at your 5 o'clock; 7 ft east and 19 ft south of the centre. Bass: 20 ft away, at your 4 o'clock; 18 ft east and 10 ft south of the centre. Hand clap: right at your seat; at the centre. Hand bell: 16 ft away, straight ahead; 16 ft north of the centre. Speech: 16 ft away, at your 3 o'clock; 16 ft east of the centre. Pink burst: 16 ft away, at your 9 o'clock; 16 ft west of the centre.

singers (numbered discs) · listening seats (squares) · drag, or Tab into the plan and use the arrow keys; in 3D, click a surface to put the selected person there. Colour is never the only code: the map has contour lines and a legend, and everything on it is also in the portrait.

The acoustic portrait

The first render is under way. The portrait appears here, in words, when it finishes.

Reach is read from the first finished render, which has not come back yet. It says which channels are heard directly from the seat and which are behind the steel.

To see it on the plan, switch on "Reach: who is behind the steel" under Reading the picture: the hatched area is out of the seat's sight, and a line through a disc marks a channel behind the steel.

Mixing board

Console cursorCH 5 · Hand clap · Fader · 0.0 dB← → channel · PgUp/PgDn control · ↑ ↓ turn · M S P mute solo polarity · B dry/room · ? keys
Hand clap, fader is zero decibels

dB
Method: Type a number, then Enter to set it or Escape to cancel. It is held to this control's range and step.
Type a channel and a control, like “tenor room send”. Enter goes there.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.

  • dB
    Method: The channel's level into its group or the master, after the processing. 0 dB is unity; at the bottom it is off.
dB
Method: Gain at the very top of the channel, before the filters and compressor — set it so the loudest part just touches the meter's top.

Flips the waveform upside-down.

Hz
Method: Removes everything below this frequency — rumble, breath, and the low modes a curved steel room loves to ring. On the console cursor, Up and Down step through the third-octave centres (Control: three at a time); Shift takes the fine step; the slider takes any whole hertz.
dB
Method: How much of this channel is heard THROUGH the room — convolved with the impulse response from where this singer stands to where you sit.
dB
Method: How loud this channel is when you hear it DRY, as if recorded in an anechoic room and panned. It is only heard when Hear says dry or both — that switch, not this level, decides which path plays.
%
Method: Left–right position of the DIRECT (un-roomed) signal, constant-power. It does nothing to the room path: in the room, where the singer stands is what pans them.

The curve is flat.

dB
Method: Boost or cut everything below the shelf frequency.
Hz
Method: Corner frequency of the low shelf. On the console cursor, Up and Down step through the third-octave centres (Control: three at a time); Shift takes the fine step; the slider takes any whole hertz.
dB
Method: Boost or cut around the low-mid frequency — where a room's boominess and a voice's warmth both live.
Hz
Method: Centre frequency of the low-mid bell. On the console cursor, Up and Down step through the third-octave centres (Control: three at a time); Shift takes the fine step; the slider takes any whole hertz.
Method: Narrowness of the low-mid bell: low Q is broad and gentle, high Q is a surgical notch.
dB
Method: Boost or cut around the high-mid frequency — presence and the edge of consonants.
Hz
Method: Centre frequency of the high-mid bell. On the console cursor, Up and Down step through the third-octave centres (Control: three at a time); Shift takes the fine step; the slider takes any whole hertz.
Method: Narrowness of the high-mid bell.
dB
Method: Boost or cut everything above the shelf frequency — air and brightness.
Hz
Method: Corner frequency of the high shelf. On the console cursor, Up and Down step through the third-octave centres (Control: three at a time); Shift takes the fine step; the slider takes any whole hertz.
dB
Method: Level of the whole group — move all its channels together.
dB
Method: Level of the whole group — move all its channels together.
dB
Method: Level of the whole group — move all its channels together.
dB
Method: Level of the whole group — move all its channels together.
dB
Method: Overall level to your headphones. It starts at −6 dB: the room can be much louder than the singers alone.

Leave it on: a focused room can make sudden loud peaks. The master also has a soft ceiling just under full scale that is always in the path.

Channel 5, Hand clap: fader zero decibels, pan centre, in the room zero decibels, muted.

Help and keys

  1. Press Play the choir (or Space on the board). You will hear the four voices in the room you chose. The four probe sounds — a hand clap, a hand bell, a speech sample and a burst of noise, on channels 5 to 8 — start muted, except in the four lesson rooms (Rotunda, Dome, Dome with an oculus, Parabolic bowl), where the clap on channel 5 is brought in for you once the room has rendered. To bring any probe in yourself, with the console focused (next step) press 5 to jump to channel 5, then M: M flips mute, and the console says which way it went.
  2. Tab to the console. It is one stop on the Tab key: Left and Right choose a channel, Page Down chooses a control, Up and Down turn it. Every move is spoken.
  3. Press B to flip between the dry voice and the same voice in the room. That is the whole idea of the tool in one key.
  4. Press A to hear the whole board, or R for the channel you are on.
  5. Press P on a channel to hear what a polarity flip does to a room; press S to solo a channel into the room.
  6. Choose Say it in the portrait to hear a short description of the room, your seat, how it rings, its echoes and where sound gathers; use the plan’s arrow keys to move a singer (the square brackets, or Page Up and Page Down, choose who), and each move says where they stand in feet east or north of the room’s centre and how far from the nearest wall. Clock positions from the listener come from the portrait and from asking “where is the soprano” with Voice on.
  7. Turn on Voice in the corner of any page and you can say all of this: “bring up the basses by three”, “pan the sopranos thirty percent right”, “play the mix”.
Left / Right
Choose the channel, then the four groups, then the master.
Shift + Left / Right
The same, but keep the control: from the Tenor's EQ low-mid to the Bass's EQ low-mid; a fader or a mute goes with you from a channel to a group and back (the master has a fader and no mute). A control the next place has not got (a group has no EQ) is named, and you land on its first.
Page Up / Page Down
Choose the control on that channel: fader, pan, trim, room send, direct send, high-pass, the four EQ bands, the compressor, group.
Shift + Page Up / Page Down
Hop a whole section: level and sends, input and filter, each EQ band, compressor and group. Down goes to the next section, Up to the start of this one, then the one before.
Up / Down
Turn the control: one step (on a frequency, a third of an octave — the standard centres, from any value); Shift for a fine step (a quarter of one, never finer than the control's own smallest step); Ctrl for three steps. Home puts it back. Switching the safety limiter off takes a second press of Down.
1 – 8, 0
Jump to channel 1 to 8, or to the master.
M · S · P
Mute, solo, polarity (phase flip) on the chosen channel.
E · C · H
Equaliser, compressor, high-pass: in or out.
B
Flip between the dry voice and the voice in the room.
R · A · L
Read this channel (every setting that is off where a fresh channel starts; one that a switch has turned off says it is not in effect) · read the whole board · find the loudest channel.
W
Say where the cursor is: the channel, the control and its value.
Shift + 1–4 · Shift + Alt + 1–4
Recall a scene · store a scene.
Space
Play and stop.
?
Say the keys; press it again, within a minute and a half and with no other key between, for the rest, and what each sound means.
D
Say what has changed since the board was opened, a project or room loaded, a scene stored, you marked the mix, or just before a scene was recalled (so right after a recall it says what the recall changed). It only reads: ask again and it says the same.
U · Shift + U
Undo the last change to the mixing board, and put it back again (Shift + U), saying what was given back. A run of nudges to one control is one step; a scene stored or recalled, Home and Reset the board are a step each. It covers the board only, not the room or where anyone stands, it never opens or closes a microphone or an audio file, and it never switches the safety limiter off. Control or Command + Z, Control + Shift + Z and Control + Y do the same from a slider or a button, and so do the Undo and Redo buttons above the console.
K · N · J
Keep the room as it is now (K, with a short rising pair of soft notes, not the scene chime), change the design, then flip what you hear between the kept room and the new one (N): the board's own crossfade, a low pair of notes for the kept room, a high pair for the new one, and four words. J says how the two differ. The kept room is scaled so the voices you can hear carry the same sound energy in both, counted for the sources on the board: that is energy, not loudness, so your ears can still find one room louder. The True level switch next to the buttons plays each room at the level the room itself gives instead. One room is kept at a time, it lasts only until the page closes and is never saved with the project, and with Hear on Dry a flip says the room is off. Spoken: keep this room, unpin the room, hear the kept room, hear the new room, flip kept and new.
=
Type an exact value for the control the cursor is on (a control with a number: a switch has none, and the safety limiter is never set this way). The cursor moves to the typing box named Exact, with the number selected; type it, Enter sets it and puts you back on the console, Escape cancels, a blank box changes nothing, and Tab sets it and goes on. It says the value that landed, whether it was held to the control's range or step, and what it replaced.
G
Jump to a control by name: the typing cursor moves to the Jump to a control box. Type a channel and a control (tenor room send), the best match is already chosen, Up and Down choose another, Enter puts the cursor there and says its value, Escape closes the list, then clears the box, then returns to the console.

Console keys. Left and right choose the channel, then groups, then master. Page up and down choose the control. Up and down turn it: one step, a third of an octave on a frequency; shift for a fine step, control for three. Home resets it. M mute, S solo, P polarity, E equaliser, C compressor, H high-pass. Keys 1 to 8 jump to a channel, zero to master. B flips dry and in the room. R reads this channel, A the whole board, L finds the loudest. Shift with 1 to 4 recalls a scene; shift and alt store one. Space plays and stops. Limiter off takes a second press of down.

Shift left or right keeps the control; shift page up or down hops sections. W where, D what changed, equals types a value, G goes to a control. U undo, shift U redo. K keep, N flip kept and new, J difference. The sounds: click, zero decibels; thud, the end; falling chime, refused; rising chime, scene; single note, room ready.

Caustic · Good Tools