Integrated LUFS and the Floating Threshold
A lesser-known detail of loudness normalization on streaming platforms
Most streaming platforms apply loudness normalization based on Integrated LUFS (LUFS-I) so that all tracks are played back at a similar perceived level.
The rule is simple:
the higher the measured LUFS-I, the more the track is attenuated during playback.
This is now widely understood.
What is far less discussed is how Integrated LUFS is actually calculated, and why the result can sometimes be counter-intuitive.
Integrated LUFS is an average, but not a simple one
Integrated LUFS represents an average loudness value measured over the full programme duration.
However, it is not a simple arithmetic mean.
The measurement relies on perceptual frequency weighting, designed to reflect human hearing, and on a gating mechanism that excludes certain parts of the signal entirely.
Both aspects matter, but gating is often overlooked.
Frequency weighting and perception
Most audio engineers are aware that LUFS uses frequency weighting to approximate human loudness perception.
This ensures that energy in different frequency bands does not contribute equally to the loudness value.
However, frequency weighting alone does not explain some of the surprising behaviors observed with LUFS-I.
For that, we need to look at gating.
The two gates used in LUFS measurement
Integrated loudness measurement uses two distinct gates.
The first one is an absolute gate, fixed at –70 LUFS.
Any signal below this level is excluded from the calculation. Its role is mainly to remove silence and inaudible background noise.
The second one is a relative gate, often referred to as a floating threshold.
The relative gate and the floating threshold
view of iZotope RX Loudness Optimize plugin
The relative gate is set 10 LU below the current integrated loudness level.
Any section that is more than 10 LU quieter than the integrated level does not contribute to the LUFS-I value.
Quiet intros, breakdowns, or transitions often fall below this threshold.
When they do, they are simply ignored by the loudness measurement.
From the meter’s perspective, they do not exist.
When quiet sections do not affect loudness at all
Those sections below the relative gate are not included in the Integrated LUFS calculation.
This explains why very quiet passages may have no impact on the integrated loudness value.
They can be present musically, but completely absent from the LUFS-I calculation.
This leads to a result that feels paradoxical at first.
When making something louder results in lower LUFS-I
Before and after gently raising the level of the quiet section
If those quiet sections are slightly increased in level so that they just exceed the relative gate, they suddenly become part of the measurement.
Because these sections are quieter than the rest of the track, they pull the average down.
In practice:
the integrated LUFS value decreases
streaming platforms interpret the track as less loud
less attenuation is applied during playback
The master can therefore sound louder after normalization, even though the LUFS-I number is lower.
A deliberate trade-off between micro and macro dynamics
This approach preserves micro-dynamics, but can reduce macro-dynamic contrast between sections.
It is not a trick, and not something to apply systematically.
It is a deliberate trade-off.
A case-by-case mastering decision
In practice, this is always a case-by-case decision.
Sometimes the original macro-dynamics are fully preserved.
Sometimes a very subtle level automation is applied so that certain sections remain above the relative gating threshold.
Understanding LUFS gating helps making more informed mastering decisions for streaming platforms.
LUFS-I alone never tells the full story.
What matters is understanding how the system behaves — and using that knowledge in service of the music.