The exported points are resampled onto a 48-points-per-octave logarithmic grid and smoothed by power-averaging over ±1/24 octave (the 1/12-octave curve) and ±1/6 octave (the 1/3-octave curve): a rectangular fractional-octave average, similar to but not the same as REW's smoothing, which uses a Gaussian-weighted kernel, so curves can differ slightly. The reference level is the average of the 1/3-octave curve from 200 Hz to 10 kHz; flatness is the largest excursion from it in that band. Bass extension is read downward from 200 Hz on the 1/3-octave curve (−3 dB and −10 dB). Modal peaks and dips are the 1/12-octave curve against a two-octave trend below 300 Hz (+6 dB and −8 dB, a rule of thumb). The treble slope is a least-squares fit of the 1/3-octave curve from 1 kHz to 10 kHz in dB per octave.
Impulse responses: the energy-time curve is 20 log of the sample magnitude relative to the peak; the decay is Schroeder backward integration of the squared response, and EDT, T20 and T30 are the 0 to −10, −5 to −25 and −5 to −35 dB fits of it (ISO 3382), broadband. The magnitude response is an FFT of 500 ms from the peak with a Hann tail; its absolute level is unknown, so it is drawn around 75 dB.
Waterfall: a 200 ms Hann window of the impulse response stepped every 15 ms over 300 ms, each slice labelled by where its window starts (the first 2 ms before the direct sound, so a slice is centred about 100 ms after its label), FFT per slice, 15 Hz – 2 kHz, levels relative to the loudest bin of the first slice; a bass bin still within 25 dB of its start at the end of the span, or decaying at less than half the typical rate, is reported as ringing. Clarity: C50 and C80 are 10 log of the energy before and after 50 ms and 80 ms, D50 the early share, Ts the energy centroid (ISO 3382-1). Distortion: THD is recomputed from the stored harmonics, 10 log Σ10^(Hk/10) minus the fundamental, with harmonics below the noise floor ignored. Group delay is REW's stored value, or −dφ/df from unwrapped phase, lightly smoothed; the Blauert & Laws thresholds are interpolated between their published points. EQ filters use REW's own biquad coefficients from an .mdat, or the RBJ cookbook formulas for a text export; the prediction is the measurement plus the filters' response.
Spectrogram: the whole range, 20 Hz to 20 kHz, in 1/12-octave steps. Each frequency is weighted by a Gaussian window seven of its own periods long (a Morlet wavelet, so about 56 ms at 20 Hz and 1 ms at 1 kHz), evaluated every 10 ms over 300 ms after the direct sound; levels are relative to the loudest point. The time to fall 30 dB is read per frequency from its own loudest slice; a frequency that never rises 35 dB above its level at the end of the recording gets no figure.
Topt and curvature: Topt is the longest straight-line fit of each band's Schroeder decay that starts between −3 and −10 dB, spans at least 15 dB above the usable end and has a non-linearity ξ = 1000 (1 − r²) of 10 ‰ or less (ISO 3382-2 Annex B); when no stretch is that straight, the straightest one, marked as such. REW has its own Topt rules, so the figures can differ slightly. Curvature C = 100 (T30/T20 − 1) %; above 10 % the decay is not a single slope (ISO 3382-2).
Share links carry the analysed summary (the 1/3-octave curve at 1/6-octave points, the headline figures, decay and clarity) compressed in the part of the address after #, which browsers do not send to the server; the notes and the measurement hardware are left out.
RT60 tables: the mid-band figure averages the 500 Hz to 1.25 kHz bands and is compared with CEDIA/CTA-RP22's 0.2–0.5 s recommendation for home cinemas; the bass ratio divides the bands below 250 Hz by it.
Biggest issues: the findings for the first ticked measurement ranked by size (bass peaks first, then a crossover dip, level, dips, flatness, treble slope and the decay time), at most three, each with what to try first. The confidence comes from the measurement check: Low with a noisy impulse response (under the 35 dB above the noise that ISO 3382-1 asks for T20), Medium with any warning such as a missing microphone calibration, a sweep starting above 25 Hz, coarse or heavily smoothed data, otherwise High. Room modes: every mode of the box (axial f = n × 343 ÷ (2 × dimension), tangential and oblique) is predicted; a measured peak or dip is pinned on a mode only below about half the Schroeder frequency (2000 × √(RT60 ÷ volume), with 0.4 s when no RT60 is given; at most 120 Hz), and only when exactly one predicted mode lies within ±3 % of it and that mode is axial, or tangential with no other mode within ±6 %. Below half the Schroeder frequency the modes overlap less than one bandwidth on average; higher up a feature is the sum of several, so none is named (a rule of thumb). A crossover dip is a dip of more than 6 dB in the 1/3-octave curve between 60 and 120 Hz that climbs back at least 3 dB between 40 and 55 Hz, in a measurement that reaches below 60 Hz within 3 dB; a speaker's own roll-off is not one. Subwoofer integration: the measurements are resampled from 15 to 300 Hz with 1/24-octave smoothing; the crossover region is where the subwoofer alone and the speaker alone are within 10 dB of each other around the frequency where they are closest. The measured combination is compared with each alone and with their power sum; with phase, the complex sum is predicted for both polarities and for delays from −10 to +10 ms in 0.1 ms steps, scored against the perfect in-phase sum, and the smallest change within 0.25 dB of the best is suggested. The prediction is trusted only when it matches the measured combination within 2 dB rms or the files say a timing reference was used.
Thresholds marked "rule of thumb" or "common practice" are exactly that. The named standards are ISO 3382 (decay definitions), CEDIA/CTA-RP22 (home cinema reverberation), the Dolby and DTS LFE bandwidth, and the Harman in-room target research; the SMPTE X-curve is mentioned as the large-room reference. Nothing here replaces a calibrator's judgement or REW's own analysis tools. Files stay in your browser.