Room Reverb Meter
Clap once or play a sweep and read the reverberation time of the room in every octave band, bass included, with the decay curve behind each figure and the absorption it would take to reach a target.
Measure
Uses your microphone; the sound is analysed on your device and is not recorded or uploaded.
Press Start listening, keep quiet for a second, then clap once, hard. Wait for the result before you clap again.
| Band | EDT | T20 | T30 | Range |
|---|
Measurements
- None yet. Every clap or sweep you measure is listed here and goes into the averages above.
Decay curves
- decay curve
- energy in 5 ms steps
- EDT fit
- T20 fit
- T30 fit
- within 10 dB of the noise
Room size and target
| Band | T (s) | A (m²) | Mean α | Add (m²) | Absorber (m²) |
|---|
Room modes
Tall lines are axial modes, middle ones tangential, short ones oblique; the dashed line is the Schroeder frequency.
| Hz | Length, width, height | Type |
|---|
Runs on your device. Uses your microphone; the sound is processed on your device. Nothing you type or record is uploaded. Keeps working without a connection while the page is open. Works with touch. Saves the room size, target, absorber values and sweep settings on this device; measurements are cleared when you reload.
Added
How to measure
- Quiet room. Switch off fans and close the windows. After Start listening the meter spends a second learning the room noise.
- Clap. Put the phone or laptop where you sit or record, stand 2 to 3 m away and clap once, hard, with cupped hands; a popping balloon has more bass. The result appears half a second after the sound has died away. Then clap again.
- Average. Three to five claps from one spot steady the figures; claps from two or three more spots give a room average. Remove any measurement that went wrong (a cough, a door).
- Or a sweep. With a loudspeaker in the room, choose Sine sweep and press Play sweep and measure: a 3 s tone glides from 30 Hz to 16 kHz and the page works out the room’s impulse response from what the microphone hears, with far more bass than a clap. Or save the sweep as WAV, play it from another device, record the room and open the recording.
The page asks the browser to switch off echo cancellation, noise suppression and automatic gain, which cut the quiet end of a decay. Some phones keep them on; the page says so under the status line when the browser reports it. A recording from an app that records unprocessed sound avoids this: open it with Open a recording (WAV, FLAC or MP3 of claps, or a WAV impulse response, first channel).
Reading the results
- EDT (early decay time): the slope of the first 10 dB of the decay, scaled to 60 dB. It follows what you hear while sound goes on and is shorter close to the source.
- T20 and T30: the slope from 5 to 25 dB and from 5 to 35 dB below the start, scaled to 60 dB, which is what RT60 means in practice. On a straight decay all three agree.
- Range: how far the start of the decay stands above the room noise. A full value needs 30 dB for EDT, 38 dB for T20 and 45 dB for T30. ISO 3382-1 suggests the 45 dB for T30; the other two are the ranges above which the page stayed within 5 % in its tests with strong bass noise. Up to 10 dB less, the value is in brackets and left out of the averages; below that, a dash.
- Averages include every measurement in the list; ± is the standard deviation. The mid value is the mean of 500 Hz and 1 kHz, T30 where both have one, otherwise T20.
- Notes flag a decay that bends (T30 more than 10 % longer than T20: coupled spaces such as an open door to a hall, or absorption on one surface only), a decay that is not straight, flutter echo, a clipped input and measurements that disagree.
How the meter works
Finding the clap: a rise of at least 10 dB within 20 ms, out of near silence, to 20 dB or more above the background. The decay starts at the first sample within 20 dB of the peak.
Octave bands: a 10th-order Butterworth band-pass filter per band, edges at fc × 10±0.15, run backwards in time so that its own ringing falls before the clap and does not lengthen short decays in the bass.
Noise: the method of Lundeby and colleagues (1995) finds the noise level of the band, the crosspoint tc where the decay sinks into it and a line through the late decay. The decay curve is Schroeder’s backward integral (1965) of the squared band signal h2 minus the noise level N, stopped at the crosspoint and topped up with the energy C the late line holds after it:
E(t) = ∫ttc [h2(τ) − N] dτ + C, L(t) = 10 log10(E(t) / E(0)) dB
A least-squares line through L(t) over each range has the slope s (dB per second), and T = −60 / s.
Quality: curvature 100 × (T30 / T20 − 1) % and non-linearity 1000 × (1 − r2) ‰, both listed in the annex of ISO 3382-2 on decay quality. Flutter echo shows as a level pattern repeating at a fixed interval in the 2 and 4 kHz bands during the first 30 dB; surfaces d apart send sound back every 2d / c.
Sweep: x(t) = sin(2π f1 K (et/K − 1)), K = T / ln(f2 / f1), with f1 = 30 Hz, f2 = 16 kHz and T = 3 s: the exponential sweep Angelo Farina described in 2000. Dividing the spectrum of the recording by that of the sweep leaves the room’s impulse response; distortion of the loudspeaker lands before it and is cut off.
Room size, absorption and modes
Sabine’s formula links the time to the volume V (m³) and the equivalent absorption area A (m²), the area of a perfect absorber that would swallow as much sound as the room: T = 0.161 V / A, so A = 0.161 V / T, and the mean absorption coefficient is A / S with S the area of all surfaces (0.161 s/m = 24 ln 10 / c, c = 343 m/s at 20 °C).
To bring a band down to the target Tt you add ΔA = 0.161 V (1/Tt − 1/T). An absorber with coefficient α laid over a surface with αs adds α − αs per square metre, so it takes ΔA / (α − αs) square metres. The table goes band by band because α changes with frequency: a few centimetres of foam or fibre absorb well above 500 Hz and little at 125 Hz, where rooms usually need the most.
Below the Schroeder frequency fS = 2000 √(T / V) Hz (with the mid value of T) a small room is ruled by its modes, standing waves at f = (c / 2) √((p/L)2 + (q/W)2 + (r/H)2) for whole numbers p, q, r. Axial modes (one number not zero) are the strongest, tangential (two) weaker, oblique (three) the weakest. Where modes bunch or leave gaps, some bass notes boom and others vanish, and the 63 and 125 Hz times become an average over a few ringing modes. The page also checks the proportions ITU-R BS.1116-3 recommends for listening rooms: 1.1 w/h ≤ l/h ≤ 4.5 w/h − 4, with l/h < 3 and w/h < 3.
Targets
| Room | Target | Averaged over | Source |
|---|---|---|---|
| Listening or mixing room | 0.25 × (V / 100 m³)⅓ s, ±0.05 s | 200 Hz to 4 kHz | ITU-R BS.1116-3 |
| Recording control room | up to 0.5 s | 500 Hz to 2 kHz | BB93 |
| Music practice room | up to 0.6 s; above 30 m³ up to 0.8 s | 500 Hz to 2 kHz | BB93 |
| Ensemble or live room | 0.6 to 1.2 s | 500 Hz to 2 kHz | BB93 |
| Primary school classroom | up to 0.6 s | 500 Hz to 2 kHz | BB93 |
| Secondary school classroom | up to 0.8 s | 500 Hz to 2 kHz | BB93 |
| Meeting or video conference room | up to 0.8 s | 500 Hz to 2 kHz | BB93 |
BB93 covers new school buildings in England, furnished but unoccupied. BS.1116-3 is written for stereo listening rooms of 20 to 60 m² floor area; the page averages its octave bands from 250 Hz to 4 kHz. My own target is compared with the mean of 500 Hz, 1 kHz and 2 kHz.
Example
Try an example room builds three claps in a 5 × 4 × 2.5 m room from octave bands of noise that die away in 1.10 s at 63 Hz, 0.95 s at 125 Hz, 0.70, 0.60, 0.55, 0.50 and 0.45 s from 250 Hz to 4 kHz and 0.38 s at 8 kHz, weak in the bass like a real clap, with some room noise. The table reads a T30 of 0.55 s at 1 kHz and 0.50 s at 2 kHz, the set values, and 0.62 and 0.72 s at 500 and 250 Hz, about 3 % long. At 125 Hz the claps stand only 34 dB above the noise, so T20 (0.92 s, 3 % short) and T30 are in brackets; at 63 Hz only an EDT in brackets is left. For a recording control room (up to 0.5 s) the room card asks for about 1.6 m² of absorption at mid frequencies (2.1 m² of absorber with α 0.8) and, in brackets, 7.3 m² at 125 Hz.
Limits
- A guide, not a certified measurement. ISO 3382 asks for an omnidirectional source, several positions and a measuring microphone; a clap and a phone are an estimate, not a test report.
- Little bass in a clap. A hand clap puts most of its energy above 500 Hz, and many phone and laptop microphones pick up little below 100 Hz, so the 63 and 125 Hz rows often show brackets or dashes. A balloon or the sweep helps.
- Sabine’s formula assumes sound spread evenly through the room. With a lot of absorption (mean α above about 0.3) it gives a longer time than the room has, so the absorption it asks for is an upper bound.
- Room modes are calculated for a rectangular room with hard walls; furniture, openings and soft walls move them a little.
- Tested on 14 November 2026 on synthetic impulse responses with a different decay time in every band, from 0.2 to 1.6 s. From a file, through a simulated microphone and from a recording of three sweeps, every full EDT, T20 and T30 came within 0.3 % of the set value; eight claps of decaying noise averaged within 3.3 %. In strong bass noise the full values stayed within 3.2 %, while values in brackets were up to 14 % off.
Sources
- International Telecommunication Union. Recommendation ITU-R BS.1116-3 (2015), section 8.2: the listening room target 0.25 (V/V0)1/3 s, V0 = 100 m³, from 200 Hz to 4 kHz, the floor area and the proportions.
- Department for Education. Building Bulletin 93, Acoustic design of schools: performance standards (2015), table 6: mid-frequency reverberation times for classrooms, music and meeting rooms.
- A. Venturi, A. Farina and L. Tronchin. On the effects of pre-processing of impulse responses in the evaluation of acoustic parameters on room acoustics, Proceedings of Meetings on Acoustics 19 (2013): Schroeder’s backward integration, the start 20 dB below the peak, Lundeby’s noise method step by step and synthetic responses with a known decay per band.