Measurements

Das-Meter claims its numbers, not a display standard. The Loudness Meter measures to ITU-R BS.1770-5 and EBU Tech 3341/3342 (momentary, short-term and integrated LUFS, LRA, true peak), tested on those standards' cases (see How it's checked). It does not claim "EBU Mode", because that also requires the EBU +9/+18 display scales, which Das-Meter doesn't have yet.

These definitions are fixed, because you compare readings across meters, and changing them later would silently move everyone's numbers.

Definitions

• LUFS, LRA, true peak: the ebur128 crate, pinned to an exact version. True peak oversamples 4× below 96 kHz and 2× at 96 kHz.
• Sample peak: 20·log10(max|x|) per channel.
• Scales (the bars' VU, PPM and K-System ballistics): defined in ADR 0007. VU reaches 99% in 300 ms with 1.25% overshoot; PPM Type I (DIN, Nordic) reads 80% of a 5 ms burst and falls 20 dB in 1.7 s, Type II (BBC, EBU) 80% of a 10 ms burst and 24 dB in 2.8 s; alignment and 0 VU are a sine peaking at −18 dBFS.
• Counters: per channel, since the last reset. Clips: runs of samples with |x| ≥ 1.0 (0 dBFS), each run counted once. TP over: runs of true peak above the ceiling (−1 dBTP by default), looked at in 10 ms blocks, so overs less than 10 ms apart count once.
• RMS: AES17 by default (+3.01 dB, so a 0 dBFS sine reads 0 dB), labelled "RMS (AES17)"; plain RMS is an option. A sliding rectangular window, 300 ms by default.
• Correlation: Pearson Σ(L·R)/√(Σ(L²)·Σ(R²)), no DC removal, the sums averaged exponentially with τ = 300 ms. With both channels below −90 dBFS it reads 0 and dims as "no signal"; with only one channel carrying signal it reads 0.
• Band correlation (Stereo Bands): the Correlation above, per octave band (31 Hz to 16 kHz; each band Left and Right through the same 4th-order Butterworth high- and low-pass at its edges), averaged with τ = 300 ms by default. Width is S²/(M²+S²) with M = (L+R)/2, S = (L−R)/2: 0 mono, ½ unrelated, 1 opposite polarity. Below −80 dBFS a band has neither.
• Cycle correlation (the Phase Scope's number, not a measurement this page claims): Pearson between the main and Overlay Source's lows, below the Cycle correlation below frequency, about each one's mean, over the last whole Cycle. It answers "do kick and bass fit?", so it's named apart from the Correlation above.
• Spectrum: sine-calibrated (a 0 dBFS sine on a bin reads 0 dB, corrected for the window's coherent gain), the slope pivoted at 1 kHz, a Hann window by default.

How it's checked

• Every change: automated tests on signals we generate ourselves at 44.1, 48, 96 and 192 kHz. Tolerances: LUFS ±0.1 LU, LRA ±1 LU, true peak +0.2/−0.4 dB, sample peak ±0.01 dB, RMS ±0.05 dB, correlation ±0.02; a Spectrum sine on a bin ±0.1 dB.
• Before each release (from v0.1.0 on): the official EBU loudness test set (Tech 3341 and 3342) at 48 kHz, every stereo case with the tolerances above. The EBU files may not be redistributed, so they're downloaded by hand and never committed. Case 6 of Tech 3341 is 5.0 and 5.1 surround, which the stereo Loudness Meter doesn't measure.

There's no official certification, so there's no badge.

Known limits

• 192 kHz: true peak isn't oversampled there, so it's only a sample peak. The Loudness Meter then labels it Peak instead of TP, so it never claims more than it measures.
• System Capture measures what the system mixer plays. For levels straight from a DAW track, use a Send Plugin.
• A Spectrum sine half a bin off-centre reads up to 1.42 dB low (the Hann window's scalloping loss).