Reading Mastering Meters

Mastering meters are useful only when read together. A single number cannot tell you whether a master is good.

The browser demo shows loudness, peak level, crest factor, correlation, phase scope, and stereo image. These meters answer different questions: how loud it is, how safe it is, how dense it is, and whether the stereo image will translate.

Read meters in groups, then trust your ears

No single reading proves a master is good. Loudness + peak answer "is it on target and safe"; crest factor + correlation answer "is it still alive and mono-safe". Use the meters to catch problems, then confirm them by listening.

The Main Readings

Loudness is reported in LUFS (Loudness Units relative to Full Scale), which tracks perceived loudness rather than raw sample level.

Meter	Question it answers
Output LUFS	Is the master near the chosen delivery target?
Peak / True Peak	Is there enough peak safety for playback and conversion?
Crest Factor	Does the track still have peak-to-average movement?
Correlation	Is the stereo image likely to survive mono playback?
Phase Scope	Is the stereo field balanced or unstable?
Stereo Image	Is width coming from useful side energy or risky spread?

The demo below covers the first two rows on a real clip — loudness and peak safety. Correlation and the phase scope are a different measurement and are shown on Mono Compatibility.

METERS · LOUDNESSIDLE

Loudness metering — LUFS, true-peak, and range

Scroll to load

The bar tracks momentary loudness as the clip plays; the panel is the loudness over time. Integrated LUFS is the single overall number, true-peak estimates peaks between samples, and LRA captures how much the loudness moves. Switch the window to compare the fast momentary meter with the smoother short-term one — each reports only once its window has filled, so the short-term contour starts three seconds into the clip.

Window

Loudness and Peak Readings in Detail

"Loudness" and "peak" each come in several forms, and meters label them separately. The loudness readings follow ITU-R BS.1770 and EBU R128; the loudness range follows EBU Tech 3342.

Reading	How it is measured	What it tells you
Momentary loudness	400 ms window, updated every 100 ms	Moment-to-moment level; jumps on every hit
Short-term loudness	3 s window	The level of a phrase; the one to watch while playing
Integrated loudness	The whole track in 400 ms blocks: blocks below −70 LUFS are dropped, then blocks more than 10 LU below the average of the rest	The single number delivery targets refer to
Loudness range (LRA)	Short-term readings across the track, ignoring those more than 20 LU below their average; the spread between the 10th and 95th percentile, in LU	How much the loudness moves between quiet and loud sections
Sample peak	The largest stored sample value, in dBFS	Can miss peaks that fall between samples
True peak	The peak of the waveform reconstructed at 4x oversampling, in dBTP	Never below the sample peak; what delivery ceilings are written against
PLR (peak-to-loudness ratio)	True peak minus integrated loudness, in dB	Small means dense and heavily limited; large means open and dynamic

The gap between sample peak and true peak is easiest to see on a waveform:

LOUDNESS · TRUE PEAKIDLE

Inter-sample peaks — why a master clips on playback

Scroll to load

The dots are samples of a sine wave; the curve shows its analytic waveform. A crest can fall between samples, above every stored value. The gap depends on frequency and phase, so it does not increase smoothly toward Nyquist. Compare the analytic peak with the oversampled meter reading. Raising the sample peak toward 0 dBFS can leave an inter-sample over that clips a converter or processor without sufficient headroom.

Sample peak

-0.3 dBFS

Frequency (×Nyquist)

0.4

What To Check First

Start with loudness and peak safety. A master that misses the target or clips is not ready.

Then check crest factor and correlation. A track can hit the target and still feel lifeless if the crest factor is too low. A track can sound wide in headphones and still collapse badly in mono if correlation swings negative.

Do Not Chase Perfect Numbers

Different genres behave differently. Dense EDM, sparse acoustic music, speech, and AI-generated music should not land on identical crest factor or stereo values.

Use the meters to catch problems, then confirm them by listening:

1. Compare before and after with loudness matching enabled.
2. Listen for lost punch, harshness, pumping, and stereo imbalance.
3. Check whether the reference track is in the same range, not whether the numbers are identical.

Implementation notes

The demo uses two measurement paths:

Path	Purpose	Where it appears
Lightweight UI metrics	Immediate feedback while interacting	Vue components
Authoritative render metrics	Values to keep in the report	libsonare render result and JSON report

Peak in the UI is scanned across every sample of the source or rendered buffer, because a peak readout is a safety number and a decimated scan can step straight over a single-sample clip. RMS, crest factor, and correlation are averages, so they are computed from a strided subset — roughly 200,000 points — to keep the interface responsive. The phase scope and stereo image are visualization aids.

Because the UI metrics and the post-render metrics come from separate paths, the numbers can differ slightly. Treat the JSON report as the processing record.

Related: LUFS, True Peak, Crest Factor, Mono Compatibility