What K-weighting is and is not
EBU R128 is a recommendation, published as a technical standard, that says how programme loudness should be measured so that different broadcasters agree. It defines a measurement procedure. It does not define taste, and it never claimed to.
The weighting filter is called K-weighting and it has two stages. The first is a shelving filter giving roughly four decibels of lift above about 1.5kHz. The second is a high-pass that removes the very low end, around 38Hz for the loudness range version and higher for the RLB variant. Together they approximate average equal-loudness contours measured with loudness matching.
Be honest about what that is. It is a fixed curve derived from panel listening tests, and it bakes in the assumption that the listener is average and that the programme is continuous. It is a reasonable model and it is reproducible, which is the property that makes it useful. It is not a claim that a 4kHz hiss is twice as loud as a 1kHz hiss to you.
| Measurement | Window | What it tells you |
|---|---|---|
| Integrated (LUFS-I) | Gated, whole programme | One number for delivery targets |
| Short-term (LUFS-S) | 3 seconds, sliding | How loud each section is |
| Momentary (LUFS-M) | 400 ms, sliding | Frame-to-frame loudness movement |
| True peak (dBTP) | Instantaneous | Peak level after intersample reconstruction |
Three numbers, not one
The most common confusion is treating LUFS as a single value. It is not. The integrated figure is a gated average across the entire piece, and the gating exists because a piece with ten seconds of silence in it would otherwise drag the average down misleadingly.
Short-term loudness is a three-second moving window. It is the number that tells you whether your chorus is six decibels hotter than your verse, which is a musical fact the integrated value deliberately hides. Momentary loudness is a 400 millisecond window and it is what you read while the track plays.
If your limiter is moving constantly and the momentary value is pinned, you have told your listeners where the ceiling is. If the momentary value floats across a ten decibel range, you have kept the drama.
- Integrated is a delivery target and belongs at the end of the mastering chain.
- Short-term is a production tool, useful for checking that sections track each other.
- Momentary is a performance tool, useful for catching abrupt jumps that create pumping.
- True peak is a separate measurement entirely, and passing it says nothing about loudness.
Targets and ceilings are different constraints
Broadcast R128 targets around minus 23 LUFS integrated. Streaming platforms normalise nearer minus 14, and some sit slightly lower still. These are convenience agreements so that a listener pressing a volume button gets a consistent result, not quality judgements.
True peak is measured separately and in different units. It reconstructs the waveform between samples, because a summed or filtered signal can peak above the highest digital sample value that was stored. A common delivery ceiling is minus one dBTP, chosen to leave a margin for whatever the player does next.
A master can sit at minus 14 LUFS and peak at minus 1 dBTP, or sit at minus 14 LUFS and peak at minus 3.5. Nothing in the loudness number tells you which. That is the whole argument for looking at crest factor alongside LUFS.
What LUFS is not
The strongest claim made on behalf of loudness targets is that matching them makes a track sound better or more professional. That does not survive a question. Matching a target makes a track behave predictably on a platform. Whether it is good is entirely a separate question that no meter answers.
There is a real risk here too. Once a target exists, a compressor can be set to hit it, and if the limiter is doing the work instead of the arrangement the track will be loud and flat and the integrated number will look perfect. That is a measurement of a problem, not a solution to it.
The defensible position is to treat LUFS the way you treat any other delivery specification: hit it because it removes an argument with the platform, and leave every creative decision to a level-matched listening session.
- Set the arrangement's dynamics first, with the loudness meter muted.
- Measure integrated, short-term and momentary to see what the piece is actually doing.
- Apply limiting only as far as the delivery ceiling requires.
- Check true peak separately, because it is a different measurement with a different purpose.
The bottom line
LUFS is a measurement with a stated purpose: it predicts how loud a piece will sound to a listener adjusting a device to a target level. It is not a quality score, and two masters with identical LUFS can differ enormously in dynamics and impact. Use the number to keep your delivery predictable and use your ears for everything else.
Match levels before you judge anything
Loudness comparison only works when both versions are level-matched. 432Hz MASTER runs at matched level so a retune is heard as tuning, not as volume.
Frequently asked questions
What does K-weighting actually do?
It is a fixed two-stage filter: a high shelf adding roughly four decibels above 1.5kHz, followed by a high-pass. It is a simplified model of average human equal-loudness contours, measured decades ago. It is a defensible approximation of an average listener, not a truth about how any individual perceives volume.
Is a higher LUFS target always better?
No. Higher integrated loudness means less remaining headroom and less room for the platform's own processing. It does not mean more impact. Dynamics and transient shape are what survive a raise, and a compressor can remove both.
What does -14 LUFS mean?
That the gated integrated loudness of the whole programme measures minus fourteen on the K-weighted scale, which is roughly 14 dB quieter than a 0 dBFS sine at 1kHz would register. Streaming platforms normalise to targets in that region because it leaves headroom for their own encoders.
Does LUFS guarantee a master sounds good?
It guarantees nothing about quality. It measures a level using a defined weighting over a defined period. Two masters can share a LUFS figure and differ by several decibels in crest factor, and the quieter one will usually sound larger and more open.