You came in through the Feel door — listen first, measure second. Play the room below and decide with your own ears how it feels; the meters (RT60, room modes, flutter) wait one scroll down for when you want the why. Design guidance, not a measurement.
Edit
RoomLoaded. The demo: Fraser's whole round barn with three speakers and two listeners — the room on screen now. Load the Whole barn for the same barn with one speaker and one listener.
Place
Simulate
Listen
Play live
Treat
Keep
Sign in to archive this analysis. The archive is kept under your account.
This is a demo: Fraser's round barn, the barn this page has always opened on, with four speakers playing the Jazz, after Mingus sound preset (bass, piano, horns and ride cymbal, synthesised in the browser — the same presets Caustic offers, so no student's audio is in it) and two listeners under the dome. It traces itself as it opens (a few seconds); then press Play to hear it. Importing a model, opening a setup or loading a room from Start from replaces it.
Opening on Fraser’s round barn: Round barn — the whole barn, stable to dome — the 16-faceted drum and its attached shed, with the gambrel dome above the mow as real sloped surfaces. A hard cone over a round floor gathers reflections back toward the middle; the talker and the listener start under the apex so you can hear it. Run the simulation, then Play this room. Not a shoebox: RAP Studio draws the plan and the storeys, Barn Jig solves the section. Load a preset from Start from or Import a room… (both in the toolbar) to work on your own.
What this model does and doesn’t say (14 notes)
THE WHOLE BARN — both storeys in one model. The mow deck is a solid floor here: neither tool models Fraser's hay chutes or the stair, so no sound passes between the stable and the mow. That makes the single Sabine/Eyring figure a WHOLE-BUILDING average rather than any one room's — the simulation, from your two points, is the authority. For an honest one-room number, load the Mow or the Stable on its own.
Stable floor: Ceiling height 9 ft — the underside of the Mow floor plate above.
Wing floor: Ceiling height 9 ft — the wall height; no roof or ceiling is modeled above it.
Two coupled rooms — Sabine assumes one diffuse field; the simulation is the authority here.
The portal between the Stable floor and the Wing floor is left as a real hole — the ray tracer carries energy through it, room to room.
Window sill heights aren't in RAP's schema — window patches sit mid-wall.
Doors are modeled closed — a plywood leaf fills each doorway.
The Silo stands inside the Stable floor as a sealed floor-to-ceiling band; its footprint is subtracted from the air volume.
Room use is set to “general” — RAP's schema has no room-use field.
The dome is real geometry, not a flat lid: a gambrel of two slopes off a 26′ plate, breaking at half the radius and running in 21′5″ above the eave until it meets the silo — which is where a round barn's roof actually stops. It is traced as 16 sloped facets per band.
A hard conical roof over a round room GATHERS reflections back toward the middle of the floor. That focusing is what you are here to hear, and it is why both the talker and the listener start under the apex. Read it as a lower bound: 16 flat facets focus less sharply than a true cone would.
RAP Studio models the plan and the storeys; its schema has no cone or arc, so the section above the plate comes from Barn Jig's own solver, built on RAP's own drum ring. Both read the one set of Fraser numbers — this is one barn, not two.
The air volume is analytic throughout — plan footprints × heights below the plate, exact frustum volumes for the two roof slopes above it, minus the silo — and it overrides the mesh volume.
Neither tool models finishes, so every surface wears a declared timber stand-in — sheathing on frame for walls and roof, a plank deck on joists for the mow floor, a sealed concrete stable floor, and a hard sealed drum for the stave silo (a wooden stave would take a little more bass). Swap any of them in the layer list; the numbers update instantly.
Imported room
Round barn — the whole barn, stable to dome — 13 layers, 109 ft × 59 ft × 47 ft envelope, 21805 ft² of surface.
Open mesh — the volume below is an estimate you can correct; escaped sound shortens the simulated tail. Statistics run on the real surface areas; the simulation traces the real triangles. Each layer’s material travels with the model in Save, Export → Setup file and Share (a model too big for a Work save is left out of it, and the save says so). Start from → Back to the shoebox, then Load, in the toolbar leaves the mesh.
Target 0.76–1.14 s for this volume. DIN 18041:2016 class A3 — education and communication: T = 0.32·lg V − 0.17 at 110,653 ft³ (3,133 m³) = 0.95 s; band 0.8–1.2 T (0.76–1.14 s), 125 Hz 0.65–1.45 T, 4 kHz down to 0.65 T (Figure 2).
Method: Each seated person is counted as free-standing absorption of 2.2, 3.8, 4.8, 5.4, 5.4, 5.4 sabins (ft² of perfect absorption) from 125 Hz to 4 kHz — 0.20, 0.35, 0.45, 0.50, 0.50, 0.50 m² a head after Kuttruff (Room Acoustics, Table 6.3) and Beranek (2004), rounded. It enters Sabine, Eyring and Arau as a uniform rise in the mean coefficient, and the trace the same way; the untreated A/B room keeps the people. The slider's ceiling is one person per 7.5 ft² (0.7 m²) of floor.
°F⇎
Method: Sets the speed of sound, c = 331.3·√(1 + T/273.15) m/s (343.2 m/s now) — so the Sabine constant 24·ln10/c, every room mode's frequency and every reflection's delay follow it — and, with the humidity, the air absorption per octave band (ISO 9613-1) that enters each reverberation time as 4·m·V and attenuates every ray in the trace.
%⇎
Method: Air absorption (ISO 9613-1, from the oxygen and nitrogen relaxation frequencies) is strongest in DRY air at high frequency: in a large room at 20 % humidity the 4 kHz reverberation can be a third shorter than at 70 %. It barely matters below 1 kHz or in a small room. Read by Sabine, Eyring and Arau (the 4·m·V term) and by every ray of the trace.
Recordings (0)
Drop recordings here, or choose several at once: WAV, MP3, M4A, OGG, FLAC. They are decoded in your browser, played whole, and never leave this device. A file named like a speaker (ignoring case, spaces, hyphens, underscores and the extension) attaches to it.
A recording still carries some echo of where it was made. The model's echo is heard on top of it, so a close, dry recording gives the truest result.
What each speaker plays
stated 70 dB at 3.3 ft (1 m)
Method: What this speaker plays: one of your recordings, a built-in sound or the live microphone. Changing it changes only what is heard and its level, never the trace.
stated 70 dB at 3.3 ft (1 m)
Method: What this speaker plays: one of your recordings, a built-in sound or the live microphone. Changing it changes only what is heard and its level, never the trace.
stated 70 dB at 3.3 ft (1 m)
Method: What this speaker plays: one of your recordings, a built-in sound or the live microphone. Changing it changes only what is heard and its level, never the trace.
stated 70 dB at 3.3 ft (1 m)
Method: What this speaker plays: one of your recordings, a built-in sound or the live microphone. Changing it changes only what is heard and its level, never the trace.
At 50 ft east, 79 ft north, 13 ft up, facing Bass, 120° (south-east).
Method: The listener whose two ears the board, the before/after A/B and the voice play. Each listener keeps its own trace: choosing one heard before plays its kept trace at once; one not traced yet, or moved since, is traced first, and the page says how far along it is.
Height cannot be heard: the head model gives left and right, front and back, but no up or down, so a sound above you sounds the same as one at ear height.
Would you want a conversation here?
That is the question every meter in this tool is only approximating — and your ears get the first vote. Put on headphones, press Play, flip between Dry and In the room, and decide: could two people talk comfortably, or would the room swallow the words? When you can say how it feels, the numbers will tell you why.
Room: approximate — until you run the trace every speaker goes through the same response rebuilt from the reverberation time: no positions, the same in both ears. Run in the toolbar for the real thing.
Mixing board — one strip per speaker
Console cursorCH 1 · Bass · Fader · 0.0 dB← → channel · PgUp/PgDn control · ↑ ↓ turn · M S P mute solo polarity · B dry/room · ? keys
Bass, fader is zero decibels
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: 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 1, Bass: fader zero decibels, pan centre, in the room zero decibels.
Sounds. Each strip’s source picks what its speaker plays: one of your recordings (dropped into Recordings — decoded on this computer, never uploaded, folded to mono and played whole), thirteen synthesized dry sounds (a clap, a snap, a pink-noise burst, a sine sweep, a bell, a plucked string, a drum, a whisper, a tone and the four parts of a chorale — made in the browser), a dry speech recording, your microphone (headphones only: through speakers it howls), your MIDI keyboard, played live (Chrome, Edge or Firefox; a note sounds the moment you play it, even with the board stopped), or a line input — Ableton, Max or an audio interface, chosen by device. Every one plays at a real level: a recording at its loudness tag, a built-in sound, the microphone or the keyboard at its stated loudness, each at 3.3 ft (1 m), and the room carries it to the listener. The faders are a trim on top. A speaker whose recording is missing is silent — nothing plays in its place. A recording already carries the room it was made in, and this adds the model’s room on top of it.
The head. The listener is a rigid sphere 3.4 in (8.75 cm) in radius facing the way the microphone faces: left–right cues (time and level difference between the ears) are modelled, front–back only crudely, and there is no elevation cue — a reflection from a dome overhead is heard at ear height. Turn the head with the microphone’s Facing under Simulate; it re-renders without tracing again.
Pan moves only the DRY signal; in the room, where the speaker stands is what places it. The board starts at −6 dB on the master with the limiter on: full scale stands for 80 dB at 0 dB on the master, and a mix that would go over it is held down by the limiter, which the page says each time. Export → Room render (.wav) renders exactly this board offline, in stereo; the per-speaker impulse responses are there too. This is design guidance from a geometric model, not a recording of the room.
Hearing readout
Press Play (or Run) to trace the model from Under the dome; the readout is written from that trace.
Viewports
View
Loading 3D view…
Run the simulation first: the paths it finds carry the clock.
Surfaces & materials
Each Rhino layer is one material surface — the auto-guess read the layer names; correct it here. RT60 updates instantly.
Where the absorption numbers come from
Concrete, sealed: After Vorländer (2008) & standard handbook tables; cf. pyroomacoustics DB (MIT)
Glass, window (⅛ in, 3 mm single pane): After Vorländer (2008) & standard handbook tables; cf. pyroomacoustics DB (MIT)
Plywood panel on frame (air gap): After Vorländer (2008) & standard handbook tables; cf. pyroomacoustics DB (MIT)
Wood floor on joists: After Vorländer (2008) & standard handbook tables; cf. pyroomacoustics DB (MIT)
80 bpm swing: walking bass, rootless piano voicings, a three-horn section and the ride cymbal. Fm9 · D♭7♯11 · C7♯9 · B♭13 and three dominants sliding down to F.
Method: The same five presets as Caustic's, four parts of one 12-second piece each, synthesised in the browser (no recordings). Choosing in this list changes nothing yet: press Load this preset (or Enter on the list) and speakers 1 to 4 take the four parts — each speaker's name, its built-in sound and its beam (a talker for a sung part, a round point for an instrument; neither is measured for an instrument). Where each speaker stands, its trim and speakers 5 to 8 stay as they are; a part with no speaker yet gets one, beside the selected speaker. Each plays at the stated loudness of a built-in sound (70 dB at 3.3 ft (1 m), an assumption), so the four keep their balance.
Speakers & listeners
Each speaker gets its own impulse response to the listener you hear from. Pick a speaker, press Place (or Place mic for the listener you hear from) in the toolbar at the top, then click a floor in the 3D room (it lands at ear height) — or drag a speaker or a listener in the 3D room (level in the plan, perspective and axon; up and down in a section or an elevation), or open its Position to type it in feet east, north and height. Clicking a speaker in the room selects it; clicking a listener hears from it. + Add speaker, Remove the selected speaker, Remove mic and Undo are up there too; listeners are added below, up to 4. In an imported model every point is tested against the model's own surfaces — a point inside a wall, a column or outside the room is flagged, not moved.
Position 82.6 east · 60 north · 14 high, ft — Bass; open to type it
ft
plays jazz upright bass
dB⇎
Method: This IS the speaker's channel fader on the mixing board (move either, both move): a gain on its dry sound before it is convolved with its own binaural impulse response to the microphone. The impulse responses keep their traced relative levels (one normalisation for the whole run), so 0 dB is the dry sound at its levelled peak and a speaker far from the mic is quieter by the distance the trace found. It does not change the simulation; −60 dB is off.
Pattern Omni — Bass; open to change how it beams and where it aims
Method: The same in every direction at every frequency: a textbook point source, the way every speaker was traced before 2026-10-09. A first-order pattern g(θ) = (1 − β) + β·cosθ with β per octave band (125 Hz to 4 kHz), after published voice directivity (Marshall & Meyer 1985; Monson et al. 2012) — rounded stand-ins, not their data. The pattern is traced: changing it asks for a new trace.
Position 71.6 east · 79.1 north · 14 high, ft — Piano; open to type it
ft
plays jazz piano
dB⇎
Method: This IS the speaker's channel fader on the mixing board (move either, both move): a gain on its dry sound before it is convolved with its own binaural impulse response to the microphone. The impulse responses keep their traced relative levels (one normalisation for the whole run), so 0 dB is the dry sound at its levelled peak and a speaker far from the mic is quieter by the distance the trace found. It does not change the simulation; −60 dB is off.
Pattern Omni — Piano; open to change how it beams and where it aims
Method: The same in every direction at every frequency: a textbook point source, the way every speaker was traced before 2026-10-09. A first-order pattern g(θ) = (1 − β) + β·cosθ with β per octave band (125 Hz to 4 kHz), after published voice directivity (Marshall & Meyer 1985; Monson et al. 2012) — rounded stand-ins, not their data. The pattern is traced: changing it asks for a new trace.
Position 38.6 east · 60 north · 14 high, ft — Horns; open to type it
ft
plays jazz horns
dB⇎
Method: This IS the speaker's channel fader on the mixing board (move either, both move): a gain on its dry sound before it is convolved with its own binaural impulse response to the microphone. The impulse responses keep their traced relative levels (one normalisation for the whole run), so 0 dB is the dry sound at its levelled peak and a speaker far from the mic is quieter by the distance the trace found. It does not change the simulation; −60 dB is off.
Pattern Omni — Horns; open to change how it beams and where it aims
Method: The same in every direction at every frequency: a textbook point source, the way every speaker was traced before 2026-10-09. A first-order pattern g(θ) = (1 − β) + β·cosθ with β per octave band (125 Hz to 4 kHz), after published voice directivity (Marshall & Meyer 1985; Monson et al. 2012) — rounded stand-ins, not their data. The pattern is traced: changing it asks for a new trace.
Position 49.6 east · 40.9 north · 15 high, ft — Ride; open to type it
ft
plays jazz ride cymbal
dB⇎
Method: This IS the speaker's channel fader on the mixing board (move either, both move): a gain on its dry sound before it is convolved with its own binaural impulse response to the microphone. The impulse responses keep their traced relative levels (one normalisation for the whole run), so 0 dB is the dry sound at its levelled peak and a speaker far from the mic is quieter by the distance the trace found. It does not change the simulation; −60 dB is off.
Pattern Omni — Ride; open to change how it beams and where it aims
Method: The same in every direction at every frequency: a textbook point source, the way every speaker was traced before 2026-10-09. A first-order pattern g(θ) = (1 − β) + β·cosθ with β per octave band (125 Hz to 4 kHz), after published voice directivity (Marshall & Meyer 1985; Monson et al. 2012) — rounded stand-ins, not their data. The pattern is traced: changing it asks for a new trace.
Listeners
○ 1Position 49.6 east · 79.1 north · 13 high, ft — Under the dome; open to type it
ft
heard from
Method: Which way this listener's nose points. Toward a speaker follows that speaker when it moves; a compass bearing stays put. Turning a listener is heard without a new trace: its kept trace is re-rendered for the new facing.
Under the dome is facing Bass, 120° (south-east).
○ 2Position 71.6 east · 40.9 north · 13 high, ft — Across the mow; open to type it
ft
Method: Which way this listener's nose points. Toward a speaker follows that speaker when it moves; a compass bearing stays put. Turning a listener is heard without a new trace: its kept trace is re-rendered for the new facing.
Across the mow is facing Horns, 300° (north-west).
Two ways of measuring this room
The formulas average the whole room; the simulation traces YOUR two points. They should disagree — where they do is the finding.
Not run yet — press Run simulation in the toolbar. It traces the actual geometry from the source to the listener — the number the formulas can’t give you.
On a mixing board — through an approximate room rebuilt from the reverberation time until you run the trace.
▶ Play, Solo, Hear dry, the sounds and the before/after A/B are in the toolbar’s Listen group above; faders, EQ and sends are on the board. Headphones.
Simulate — from your seat
The formulas treat the whole room as one average. This traces the actual geometry from each speaker (the numbered dark spheres) to the microphone (the open sphere) — a real two-ear impulse response for every speaker → microphone pair, the way ODEON or Pachyderm would. The trace runs backwards from the microphone, so eight speakers cost the same as one, and every number below is read off the impulse response you hear on the board. Run, its quality, + Add speaker and Place for a speaker or the mic are in the toolbar at the top; where each speaker and the mic stand is in Speakers & microphone.
Showing Bass → microphone — not run yet
The formulas — whole room
2.86 sRTmid · Too live · target 0.76–1.14 s
Too live for general / multipurpose — RTmid ≈ 2.86 s; aim 0.76–1.14 s. Add absorption. How: the mean of Sabine 2.91 s and Eyring 2.80 s. Sabine T = 0.161·V/(ΣSα plus the air's 4mV (68°F, 50 % RH)) = 2.91 s; Eyring T = 0.161·V/(−S·ln(1−ᾱ) plus the air's 4mV (68°F, 50 % RH)) = 2.80 s; Arau-Puchades T = Π(Tᵢ)^(Sᵢ/S) over the three axis pairs = — s. The headline is the mean of Sabine and Eyring (Sabine overstates a well-damped room, Eyring is exact for a uniform one; the unevenness index is 0.29, under the 0.5 at which Arau takes over). DIN 18041:2016 class A3 — education and communication: T = 0.32·lg V − 0.17 at 110,653 ft³ (3,133 m³) = 0.95 s; band 0.8–1.2 T (0.76–1.14 s), 125 Hz 0.65–1.45 T, 4 kHz down to 0.65 T (Figure 2).
Confidence: low — the headline could be anywhere from 2.05 to 4.33 s with every coefficient at its published uncertainty. Design guidance — not a measurement.
Why low
Each published coefficient is known to about ±0.02 plus 15 % of its value (the spread between laboratories in ISO 354 round-robins). With every coefficient at the top of that band the room has 714 more sabins at mid band and the headline is 2.05 s; at the bottom (no lower than half the table's value) it is 4.33 s — a range of ±40 %.
The imported mesh is not closed, so its volume is a figure typed in or the bounding box, not the air inside — every formula scales with it.
Reverberation by frequency
Representative RT60 in each octave band against the room-use target.
on target near off modal-influencedshaded = target zone · dashed = f_s
Dashed line = Schroeder frequency f_s ≈ 60 Hz: above it the room behaves diffusely; below ~2·f_s it is modal.
Reverberation time by octave band — text read-back of the chart above.
Frequency
RT (s) · formulas
Status
125 Hz
1.44
On target
250 Hz
1.84
Too live
500 Hz
2.19
Too live
1 kHz
3.52
Too live
2 kHz
3.08
Too live
4 kHz
2.09
Too live
Room modes
The lowest standing-wave resonances of the empty box. Nominal frequencies — real rooms shift and broaden them.
Imported model: these modes are computed on its BOUNDING BOX — scale guidance only. The real shape shifts and splits them; the simulation panel is the geometry-aware read.
Lowest 16 modes — tap one to focus it, or ♪ to hear it.
Show the 16 lowest modes — lowest 5 Hz (axial)
A mode is a single frequency — ♪ plays it, so the number has a sound. Below about 60 Hz most laptop speakers reproduce only their own distortion; use headphones for the bass modes, which are the ones that matter.
4 bass pile-ups between 36 and 57 Hz · 1 wide bass gap (~7 Hz). Axial modes closer than a quarter-tone stack into a resonant bass peak. A wide gap between neighbouring axial modes leaves uneven, “one-note” bass. Only pairs below the Schroeder frequency (60 Hz) are counted: above it modes overlap and no single pair rings alone.
Bass gap (~7 Hz): Wide gap between 5 and 10 Hz — uneven, "one-note" bass.
Bass pile-up (~36 Hz): Axial modes at 36 and 36 Hz nearly coincide — a resonant bass peak.
Bass pile-up (~47 Hz): Axial modes at 47 and 48 Hz nearly coincide — a resonant bass peak.
Bass pile-up (~48 Hz): Axial modes at 48 and 48 Hz nearly coincide — a resonant bass peak.
Bass pile-up (~57 Hz): Axial modes at 57 and 57 Hz nearly coincide — a resonant bass peak.
Echo & flutter
Parallel hard surfaces can trap a buzzy repeating echo.
No obvious flutter-echo risk — no parallel pair of hard, reflective surfaces was found.
Clarity & speech
Mid-band clarity, definition and (if noise was given) intelligibility.
Clarity and definition proxies at the mid octave bands, estimated from RT60.
Metric
500 Hz
1 kHz
EDT — early decay
2.19 s
3.52 s
C50 — speech clarity
-4.3 dB
-6.6 dB
C80 — music clarity
-1.8 dB
-4.3 dB
D50 — definition
27 %
18 %
Estimated from RT60 — guidance, not measurement. Real clarity, definition and intelligibility need a measured or simulated decay curve.
Treat — reach your target
About 3441 sabins (ft² of perfect absorption) more puts the headline on 0.95 s, the target’s centre line. Method: bisection on the engine's own headline (Sabine, Eyring and Arau as the page grades them) for the uniform free-standing absorption that puts it on the target's centre line. Any one of these does it — press it to apply (Treat → Reach target in the toolbar has the same list):
Assistant
Sign in to ask the assistant. It answers on the studio's AI key, so it needs an account. Everything else on this page works without one: the room, the trace, the mixing board and Export.