About Room Acoustics
Why does a clap ring for seconds in an empty hall but die at once in a living room? Why is a tiled bathroom so flattering to singers, and why can you hear the bass of next door's music but not the words? This page works out what a room does to a sound and lets you hear it: pick the room, what its walls, floor and ceiling are made of, whether it is furnished and whether its doors are open, put yourself and the sound anywhere on the plan, and press CLAP, VOICE or NOTE.
THE ROOM: the plan, drawn from above, with the sound (S) and you (YOU): drag either, or use the arrow keys on the plan (Shift moves the sound). a door to open or shut it. The analyser below shows the ECHOGRAM: every arrival at your ears, in decibels against time. The tall line first is the direct sound; the coloured lines after it are the early reflections, each one a single bounce path (step through them with ◀ and ▶ and the plan draws the path, with its mirror-image source); the grey area is the reverberation, the thousands of later reflections run together. DECAY shows the same sound as a falling curve (Schroeder's backward sum), with the straight lines Sabine's and Eyring's formulas predict. The meter gives the reverberation time (the time to fall 60 dB) three ways, the early decay time, C50 and STI for speech, and the level. The tape window draws what reaches you, so everything works with the sound off.
Two ears (wear headphones): YOU has a nose, two ears and an arrow showing the way you face. Turn by dragging the round handle on the arrow, with the mouse wheel over your head, or with Q and E (Shift for 5° steps). What you hear is worked out for each ear separately, and the tape draws the left ear above the line and the right below. EAR on the analyser shows the echogram at either ear. Spokes round your head point to the 12 loudest early arrivals, longer for louder, teal when your left ear hears it louder, amber for the right. Turn your back on a talker and the voice goes duller and a little quieter; stand by a wall and turn, and its echo moves from ear to ear.
WHY IT RINGS does the classic sums for the room the sound is in, written out: how much each surface soaks up, Sabine's formula, Eyring's, and the simulation beside them. THROUGH THE WALL: how much sound a wall, window or door stops, why doubling its weight gains only about 6 dB, where a wall's own bending lets sound through (the coincidence dip), what a small gap does, and why it is the bass that gets through. Pick the wall and what is happening next door, and hear it in their room and through the wall (turned up, so you can hear what the wall does to it).
THE STORY
In 1895 Wallace Clement Sabine, a young assistant professor of physics at Harvard, was asked to fix the new Fogg Art Museum lecture room, where speech was impossible to follow. Over several years of nights (the building had to be quiet) he timed how long the sound of an organ pipe stayed audible, carrying hundreds of seat cushions in and out of the room from the nearby Sanders Theatre, and found that the time was proportional to the room's volume divided by the total absorption. He published it in 1900, and the formula still carries his name; the unit of absorption, one square metre (or foot) of open window, is the sabin. Boston's Symphony Hall, opened in 1900, was the first hall designed with it, and is still one of the best-liked halls in the world. Carl Eyring at Bell Laboratories showed in 1930 that Sabine's formula gives too long a time in very absorbent rooms (a room with every surface open window would still ring a little by Sabine) and gave the corrected form. Image sources, mirror images of the source in each wall, go back to optics; Jont Allen and David Berkley's 1979 paper made them the standard way to compute a rectangular room's echoes. Ray tracing for rooms was first done by Asbjørn Krokstad, Staffan Strøm and Svein Sørsdal in 1968. Manfred Schroeder's backward-integrated decay curve (1965) is how reverberation times are measured today (ISO 3382). The Speech Transmission Index came from Tammo Houtgast and Herman Steeneken in the Netherlands in the 1970s and 1980s.
The mass law (a wall's insulation rises by about 6 dB for each doubling of its weight or of the frequency) goes back to the early 20th century. Lothar Cremer explained the coincidence dip in 1942: at one frequency the wall's bending waves travel at the same speed as the sound sweeping along it, and the wall lets far more through. On 3 June 2012 Trevor Cox and Allan Kilpatrick measured a reverberation time of 75 seconds (112 seconds at 125 Hz) in an empty Inchindown oil tank in the Scottish Highlands; Guinness World Records recognised it in 2013, and it was announced in January 2014.
HOW IT WORKS
Plans: every room is a box: a flat plan of rectangular rooms with one ceiling height (set with HEIGHT); the same materials are used throughout. The sound is 1.5 m above the floor, your ears 1.2 m. Doors are 2.05 m high; an open door is a hole, a door to outside lets the sound out; windows are 6 mm glass. Materials: octave-band absorption coefficients from 125 Hz to 4 kHz, the classic textbook table of typical values (as reproduced in Sengpiel's table of absorption coefficients; the same figures appear in Egan's and Everest's textbooks): plaster smooth on brick, unglazed brick, concrete, 12.5 mm plasterboard on studs, 10 mm plywood panelling, window glass, heavy velour curtains, carpet on foam underlay, wood floor on joists, ceramic tiles, suspended acoustic ceiling tiles, wood boards, and empty fabric-upholstered seats. Doors use the plywood panel figures (and glazed doors the glass figures). Real products vary: these are typical. WITH FURNITURE makes the small room a bathroom: a bath along one wall (enamelled steel, taken as reflective as glazed tile, but scattering), a bath mat (the table's carpet on concrete) and two bath towels on a rail (both faces, using the table's heavy molleton curtain as the nearest figure for heavy cotton). In the other rooms it covers part of the floor (25% in the medium room, 60% in the hall, 10% in the corridor and 30% in the suite) with seats and sofas, using the upholstered seats' published figures per square metre of floor, and makes the walls scatter more. How much each surface scatters (throws sound off in random directions rather than like a mirror) is the page's own round figure: about 0.1 for smooth surfaces, 0.4 for curtains, 0.7 for seats, at least 0.3 for walls with furniture. Air absorbs high notes: the page uses ISO 9613-1 at 20 °C and 50% humidity (4.7 dB per km at 1 kHz, 30 dB per km at 4 kHz).
The echoes: the early reflections (the first 80 ms) by image sources, worked out exactly for the box the sound is in, each path checked for open doors and with each surface's absorbed and scattered share taken off. After that, and for everything scattered, a ray tracer: tens of thousands of rays (4,000 to 60,000, depending on the room) leave the sound in random directions, lose energy at each surface, bounce like a mirror or (with the scattering share) off in a random direction, and pass through open doorways; a sphere round your head counts the energy passing through it in 1 ms steps. Sound through walls and shut doors is too faint for rays to carry, so it comes from a statistical model of rooms joined together (each room's sound energy evenly spread, flowing through the wall and through the door, with a 10 mm gap underneath each shut door). What you hear: from those, an impulse response is built in octave bands (spikes for the direct sound and early reflections, shaped noise for the rest) and the dry sound is played through it with the browser's convolver (Web Audio's ConvolverNode). The sounds are real dry recordings: a single hand clap, a voice saying "hello" and a C6 note (the note's own fading is the instrument's). Each room's level is played compressed (a room 30 dB quieter plays 10 dB quieter) so nothing is deafening or silent; DRY plays the recording as it is.
The two ears: every arrival (the direct sound, each image source and each ray reaching your head) keeps the direction it came from. How much later the far ear hears it follows Woodworth's spherical head (1938): (a / c)(θ + sin θ), with a head 8.75 cm in radius, at most 0.66 ms, placed between samples. How much quieter the shadowed ear hears it follows Brown and Duda's (1998) spherical-head shadow filter (ears at ±100°): little difference in the bass, up to about 15 dB in the treble with the sound fully to one side. Sound from behind also has the treble turned down (by up to 6 dB at 4 kHz directly behind, and 1 dB off everything), a page's own simple stand-in for the way the outer ear shadows sound from behind. Real front and back hearing depends on the shape of each person's own ears (their head-related transfer function, HRTF; see Blauert's Spatial Hearing): a simple model like this can only hint at it, and many people will still hear some sounds inside their head or in front when they are behind. The reverberant tail is shaped noise for each ear from that ear's own share of the rays, partly alike in the two ears as in a real room (alike in the bass, independent in the treble). Averaged over every direction, each ear gets the same energy as the single point in the middle of the head that the meter, Sabine and Eyring use.
The numbers: reverberation time from the decay curve between −5 and −35 dB (T30), early decay time from 0 to −10 dB (EDT, closer to what you hear), C50 (the sound in the first 50 ms after the direct sound against everything later: above 0 dB, early sound outweighs late, which usually helps speech), STI from the room alone with no background noise (IEC 60268-16's weights; the 8 kHz band takes the 4 kHz figure), level in decibels against the direct sound 1 m from the source. Sabine: T = 0.161 V / A; Eyring: T = 0.161 V / (−S ln(1 − ᾱ) + 4mV). Sabine and Eyring assume sound visits every surface equally (a 'diffuse' field). When most of the absorption is on the floor or ceiling and the walls are bare and parallel, sound bouncing sideways misses it and the simulation rings longer than either formula, as real rooms do.
What it leaves out: this is a teaching model. Sound is treated as rays, so it doesn't bend round corners or door edges (diffraction): with no straight line through a doorway there is no direct sound at all, where in a real building some sound would bend round. A room's own resonances (modes), which shape the bass in small rooms below the Schroeder frequency (shown on WHY IT RINGS), aren't worked out; the page's Ripple Tank shows waves doing this in two dimensions. Walls are simple: no sound going round the wall through the floor and ceiling (flanking), and a gap simply lets everything through. The scattering figures, the walls' weights, stiffness and loss figures and the 10 mm gap are the page's own; the absorption figures are typical, not for any product.
Walls: the 'field incidence' mass law (sound arriving from every angle up to 78°), and, near and above the wall's coincidence frequency, Sharp's method (1978), with the wall's losses into its edges added as EN ISO 12354-1 gives them. That makes 215 mm plastered brick come out at Rw 54 dB, 6 mm glass at 30 dB and a a 44 mm solid-core door at 30 dB and a hollow-core door at 24 dB before their gaps, close to measured figures. Two-leaf constructions (the stud wall, plywood panelling on studs, the hollow door) follow Sharp's double-wall method: below the 'mass-air-mass' resonance the two leaves move together, above it the gap helps a lot, up to the limit the studs (or the door's core) allow because they tie the leaves together; the stud wall comes out at Rw 40 with mineral wool in it. Rw is the octave-band version of ISO 717-1's single number. THROUGH THE WALL: two rooms 5 × 4 × 2.6 m side by side, a wall of 10.4 m² between them, each room furnished (reverberation time 0.5 s); the levels next door and a quiet bedroom's background (about the NR 25 curve) are the page's own round figures; the threshold of hearing is ISO 226:2003's, for a pure tone at each band's centre. The level heard through the wall is L1 − R + 10 log(S/A).
Sounds: single clap by rajatchoudhary, "hello" and C6 note by freesound_community, via Pixabay (Pixabay Content Licence).
FURTHER WATCHING AND READING