dBFS vs dBTP vs LUFS: What Each Level Measures
Last updated 6 min readTags:MasteringLoudnessStreaming
Understand sample peak, true peak and loudness. Learn how to read dBFS, dBTP and LUFS, check your export and set an appropriate limiter ceiling.
A track can read −0.3 dBFS on a sample-peak meter, +0.4 dBTP on a true-peak meter and −8.5 LUFS on a loudness meter. All three readings can be correct: each measures a different aspect of the signal.
Use sample peak to check the level of the digital samples, true peak to check for peaks between them, and LUFS to assess loudness over time.
| Reading | What it measures | What to use it for |
|---|---|---|
| Sample peak, in dBFS | The highest sample level relative to digital full scale | Checking sample headroom at recording, processing and export |
| True peak, in dBTP | An estimate of the reconstructed waveform's highest level, including peaks between samples | Checking peak headroom for playback and encoding |
| Loudness, in LUFS | Frequency-weighted average energy over a defined period | Comparing loudness and checking delivery targets |
dBFS: sample level and digital headroom
dBFS means decibels relative to full scale. On a sample-peak meter, 0 dBFS is the maximum sample level for a fixed-point file, such as a 16-bit or 24-bit PCM export. A peak of −6 dBFS is 6 dB below that limit.
If the signal exceeds the available range when it is recorded or exported to fixed point, the samples clip. Floating-point processing inside a DAW can accommodate levels above 0 dBFS, but the signal still needs to fit the range of the final fixed-point export.
A sample-peak reading does not tell you whether distortion occurred earlier. A clipped recording turned down by 6 dB will peak below full scale and retain its distortion.
Nor does sample peak measure loudness. A brief snare hit and a sustained synth can reach the same peak, while the synth sounds much louder overall.
Check which measurement your meter displays. dBFS is a unit, not a measurement method. An RMS meter can also use dBFS; its reading describes average level rather than the highest sample.
dBTP: peaks between the samples
Playback reconstructs a continuous waveform from the digital samples. That waveform can peak above the highest sample, so a file whose samples remain below 0 dBFS can still exceed full scale during reconstruction.
A true-peak meter estimates these peaks, usually through oversampling, and reports the result in dBTP, or decibels true peak. The method is specified in ITU-R BS.1770.
For example, −0.3 dBFS sample peak and +0.4 dBTP true peak mean the samples remain below full scale, but the estimated waveform peak exceeds it by 0.4 dB. Whether that produces audible distortion depends on the playback equipment and processing that follow.
Lossy encoding can also change peak levels. A true-peak ceiling below 0 dBTP leaves some margin for those changes, although the appropriate margin depends on the delivery format and specification.
See true peak for more on interpreting these readings.
LUFS: loudness over time
LUFS means loudness units relative to full scale. The measurement uses K-weighting, a frequency filter designed for loudness assessment, then averages the weighted signal energy. It is more useful than peak level for comparing overall loudness, though equal LUFS does not guarantee that two tracks sound equally loud.
Loudness meters commonly offer three readings:
- Integrated: all audio measured since the meter was started or reset, with gating that excludes silence and sufficiently quiet passages.
- Short-term: the previous three seconds.
- Momentary: the previous 400 milliseconds.
These time windows follow EBU Tech 3341. For whole-track integrated loudness, reset the meter and measure the entire song, or analyse the complete exported file. A reading taken from the chorus alone describes that chorus.
Less negative values mean louder: −9 LUFS is louder than −14 LUFS. A clean 2 dB gain reduction lowers loudness by approximately 2 LU and lowers both peak readings by 2 dB. Compression, limiting and EQ can change the relationship between loudness and peaks.
The LUFS reference explains the measurement in more detail. You may also see LKFS, which is equivalent to LUFS.
What the three readings tell you together
Consider the example from the opening:
| Measurement | Reading | Interpretation |
|---|---|---|
| Sample peak | −0.3 dBFS | The highest sample is below full scale |
| True peak | +0.4 dBTP | The estimated reconstructed peak exceeds full scale |
| Integrated loudness | −8.5 LUFS | The track measures louder than Spotify's Normal playback level |
The loudness figure alone is not a fault. With Spotify's Normal setting active, an individually normalised track at −8.5 LUFS would receive about 5.5 dB of attenuation to reach −14 LUFS. Album playback and other settings can behave differently. See Spotify's normalisation guidance.
The positive true peak deserves a separate check. If you wanted this illustrative file below −2 dBTP, a clean 2.5 dB gain reduction would bring it to approximately −2.1 dBTP, −2.8 dBFS sample peak and −11 LUFS integrated. That changes the level without adding compression.
If you need to retain more loudness while reducing peaks, a true-peak limiter may be appropriate. Listen for changes to drums, bass and other transients as you adjust it. The input gain and ceiling interact, so remeasure loudness and peaks after either changes.
Which targets should you use?
Delivery guidance distinguishes loudness from peak level:
| Destination | Loudness | True peak |
|---|---|---|
| Spotify | −14 LUFS for Normal playback | Recommends below −1 dBTP, or below −2 dBTP for masters louder than −14 LUFS |
| EBU R128 broadcast | −23.0 LUFS programme loudness | Maximum −1 dBTP during linear-audio production; distribution requirements may be lower |
Spotify's figures are platform guidance, rather than a universal requirement for music mastering. Choose the track's loudness according to its sound and intended delivery; normalisation does not undo compression or limiting already applied.
EBU R128 permits ±1.0 LU where the target cannot practically be met, and ±0.2 LU for measurement error in quality-control workflows. A broadcaster's delivery specification may include further requirements. Check the completed deliverable against that specification.
For a limiter, check whether true-peak detection is enabled. The unit printed beside its ceiling control is not sufficient to establish how it controls peaks. Measure the export to confirm the result.
How to check your export
- Finish the mix and choose the delivery requirements. Check any loudness target and true-peak ceiling supplied by the recipient.
- Export the complete track. Include the intended intro, ending and effect tails.
- Measure that file. Note integrated LUFS, maximum true peak and maximum sample peak. Do not use a loop of the loudest section as the whole-track measurement.
- Address the specific problem. A gain reduction can resolve excess peak level if lower loudness is acceptable. Use limiting when you need peak control with less reduction in overall loudness, and judge its effect by ear.
- Remeasure after changes. Check the final file after sample-rate conversion and other export processing. Use an encode-and-decode check if you need to assess peaks in a particular lossy format.
None of these readings tells you whether the vocal is clear, the bass is balanced or the chorus has enough impact. Use them to check level and headroom, then assess the mix by listening. For feedback on your own track, analyse your mix with Sonalix.