Record how a room answers. A sine sweep from the speakers, a microphone in the room, and the sweep’s inverse filter turn a few seconds of recording into the space’s impulse response: the click of the direct sound, the first reflections off the nearest surfaces, and the tail. The WAV it saves is the file the convolution reverb attaches to put a recording in that room.
Running a measurement
Set the sweep length and start. The microphone opens with the browser’s voice processing off, because echo cancellation exists to remove the speakers from the microphone, which is the whole signal here. The sweep plays after half a second, then the recording continues for the response length plus a second so the decay is caught. Keep still until the status returns to idle; a chair creak during the sweep is a reflection the room never made.
The sweep
An exponential sweep from the chosen low frequency to 20 kHz, or to 45 percent of the sample rate where that is lower. The inverse filter is the sweep reversed with a 6 dB per octave envelope, which undoes the longer time an exponential sweep spends at the top. Linear convolution by FFT, then the response is cut from 2 ms before the loudest point after the loop’s zero, normalised to a peak of 0.9, and faded over its last 10 ms.
What the result shows
The large figure is the mid-frequency T20, the mean of the 500 Hz and 1 kHz bands, measured on the captured response by the same backward integration the reverb time analyzer uses. The decay plot is the response’s envelope in dB; the frequency response is third-octave smoothed and centred on its 200 Hz to 5 kHz mean. The number of decibels above the floor says how much of the decay is clean: T20 needs 35, T30 needs 45.
What is in the response
The room through this loudspeaker and this microphone. A laptop’s speaker is in the frequency response as much as the walls are, and a phone’s microphone adds its own tilt. For a measurement of the room alone the chain needs a flat loudspeaker and a measurement microphone, which is what REW and a calibrated capsule are for. For a reverb to use, the chain’s colour is part of the sound and often the point.
The distortion products
A loudspeaker driven by a sweep makes harmonics, and in an exponential sweep each harmonic is the same sweep started earlier. After deconvolution they land at negative times, before the click, in a row. The cut starts at the click, so they are left out, which is the reason to sweep rather than to clap.