Skip to content

K-weighting and gating: how LUFS meters work

Send feedback

Last updated 7 min readTags:MasteringLoudnessLUFS

Learn how BS.1770 uses K-weighting, energy averaging and gating to measure LUFS, and why bass, quiet intros and stereo width affect the result.

K-weighting is the frequency filter a LUFS meter uses before it calculates loudness. It reduces the contribution of sub-bass and gives more weight to higher frequencies. Gating excludes silence and sufficiently quiet passages from integrated loudness.

Both are part of ITU-R BS.1770, the standard used by LUFS meters. They explain why adding bass can push your limiter harder without much change in LUFS, and why a longer quiet intro may barely affect the integrated reading.

This page explains the calculation and what it means for your mix. For an explanation of loudness targets and normalisation, see LUFS and streaming.

How BS.1770 calculates loudness

For integrated loudness, the calculation has four stages:

Stage What the meter does What it means for your mix
K-weighting Filters each channel before measurement Equal signal levels at different frequencies can contribute different amounts to LUFS
Mean square Calculates average energy in each channel Sustained sound contributes more than brief peaks at the same peak level
Channel summation Adds the channel energies with fixed weights Stereo loudness does not reveal cancellation between left and right
Gating Excludes blocks below two loudness thresholds Silence and sufficiently quiet passages do not lower the integrated result

The result is expressed in LUFS, also called LKFS. These are equivalent units. Differences between readings are expressed in LU; a difference of 1 LU corresponds to 1 dB.

1. K-weighting changes how frequencies contribute

K-weighting combines two filters: a high shelf, which raises higher frequencies, and a high-pass filter, which attenuates low frequencies. These filters operate inside the meter. They do not change the sound of your mix.

The following values show the combined response relative to 1 kHz, calculated from the standard's published filter coefficients at 48 kHz:

Frequency Weighting relative to 1 kHz
20 Hz −14.0 dB
40 Hz −6.3 dB
60 Hz −3.6 dB
100 Hz −1.8 dB
1 kHz 0.0 dB
4 kHz +3.3 dB
10 kHz +3.3 dB

At equal input levels, a 40 Hz component contributes less to the weighted energy than a 1 kHz component. A 4 kHz component contributes more.

Why extra sub-bass may barely move LUFS

A sub-bass boost can increase peaks and make a limiter work harder. Its effect on LUFS may be small because K-weighting attenuates that region before measurement.

The result depends on how much bass the track already contains, the EQ settings and any processing after the EQ. There is no fixed conversion from a bass boost to a change in LUFS. A 3 dB boost at 40 Hz does not necessarily raise the whole track by 3 LU.

The reverse can happen with upper mids and treble. Reducing a harsh region around 3–5 kHz may lower the reading even though the mix sounds better.

Use EQ to correct the sound. The weighting curve is not a recommended tonal balance, and brightening a mix just to raise LUFS can leave it harsh. Check true peak separately to see what happens to your peak level.

2. The meter averages energy, not decibel readings

After K-weighting, the meter squares the samples and averages the results. This is the mean-square calculation. It measures average signal energy rather than peak level.

Suppose a track contains:

  • 120 seconds at a steady −10 LUFS;
  • a 60-second breakdown at a steady −16 LUFS.

A time-weighted average of those decibel values would be −12 LUFS. The energy average is approximately −11.25 LUFS. Both sections pass the relative gate in this example, so gating does not change that result. These figures ignore the small effect of measurement blocks across section boundaries.

The louder section contributes more energy per second. Its level has more influence on integrated loudness, but the duration and level of the quieter section still matter. Integrated loudness is not simply the loudness of the chorus.

This also explains why limiting followed by make-up gain can raise LUFS: the average level rises relative to the peaks. Whether that improves the track depends on what happens to its transients, punch and distortion. Crest factor explains the relationship between peaks and average level.

3. Stereo channels are measured separately

For stereo, the meter gives left and right equal weight and adds their measured energies. It does not combine the two waveforms into mono before measuring them.

If you invert the polarity of the right channel, its energy stays the same. The stereo LUFS reading therefore stays the same too, even though the mix may sound very different when summed to mono.

This distinction matters when you use widening effects. A widener can change LUFS if it changes the energy in either channel. What LUFS cannot tell you is whether those channels reinforce or cancel each other in mono.

Check the mix in mono and use stereo correlation alongside the loudness reading. An unchanged LUFS figure does not confirm that the stereo image is sound.

4. Gating excludes sufficiently quiet passages

For integrated loudness, the meter measures overlapping 400 ms blocks, starting a new block every 100 ms. It then applies two thresholds:

  1. Absolute gate: blocks below −70 LUFS are excluded.
  2. Relative gate: the meter calculates the energy average of the blocks that passed the absolute gate, then sets a second threshold 10 LU below that average. Blocks below this threshold are excluded as well.

The final integrated reading is the energy average of the blocks that pass both gates. The relative threshold is based on the average before the relative gate is applied. It is not set 10 LU below the chorus or the final integrated reading.

Does a quiet intro lower integrated LUFS?

It depends on whether its blocks pass the gates. Consider a 180-second song body at a steady −10 LUFS:

Addition to the song Approximate integrated loudness Why
None −10.0 LUFS The song body sets the result
A 30-second intro at −30 LUFS −10.0 LUFS The intro falls below the relative gate
A 30-second intro at −18 LUFS −10.6 LUFS The intro passes the relative gate and lowers the average
20 seconds of digital silence at the end −10.0 LUFS The silence falls below the absolute gate

These are simplified calculations. Actual readings can vary slightly because overlapping blocks include audio from either side of a section boundary.

In the −30 LUFS intro example, the average after the absolute gate is about −10.7 LUFS. The relative threshold is therefore about −20.7 LUFS, which excludes the intro.

The threshold also changes as more audio is measured. A quiet intro may count when you play it alone, then be excluded once the louder song body has played. Measure the whole track before comparing integrated readings.

Are momentary and short-term loudness gated too?

In EBU Mode, momentary loudness uses a 400 ms window and short-term loudness uses a three-second window. Neither uses these gates. They show local changes in loudness that the integrated reading may conceal.

Loudness range, or LRA, uses a separate calculation with different gating rules.

How to use this when checking a mix

Measure the full export, then investigate any unexpected reading before changing the mix:

  • Peaks rise but LUFS barely moves: check whether extra sub-bass is driving the peaks or limiter. Compare the sound at matched playback levels before reducing it.
  • The integrated reading looks reasonable but a section sounds too loud: check that section's short-term loudness and listen to the transition. A whole-track average can conceal an awkward level change.
  • A longer intro makes little difference: check whether it falls below the relative gate. Its musical effect still matters, even when the meter excludes it.
  • The mix loses bass or instruments in mono: investigate phase and stereo processing. Stereo LUFS will not identify the cancellation.

Sonalix reports integrated loudness alongside other measurements. Read it with true peak, tonal balance and stereo correlation when deciding what to change. Two tracks with the same LUFS can differ substantially in bass weight, punch and clarity.

    Type to search titles, headings and summaries.