Loudness Measurement: ITU-R BS.1770, EBU R128 and ATSC A/85
Loudness jumps between programs and commercials have long been among the most common viewer complaints. Broadcasters solved this by moving from peak-based level control to loudness normalization: every program is measured with the same algorithm and brought to the same target. This guide explains how that measurement works under ITU-R BS.1770, what the EBU R 128 and ATSC A/85 targets are, how the two differ, and which mistakes come up most often. All values are taken from the documents themselves: ITU-R BS.1770-5 (11/2023), EBU R 128 version 5.0 (November 2023) with EBU Tech 3341 and Tech 3342 (version 3.0, January 2016), and ATSC A/85:2026-07 (approved 8 July 2026).
Why loudness normalization?
EBU R 128 states the problem directly: peak normalization of audio signals “has led to considerable loudness differences between programmes and between broadcast channels”, and the quasi-peak meter (QPPM) used to read peaks does not reflect loudness. A heavily compressed commercial and a dynamic drama can both peak at the same level and still sound very different in loudness. Normalizing to a measured average loudness, and limiting peaks separately with a true-peak limit, removes that mismatch while leaving room for dynamics.
How BS.1770 measures loudness
ITU-R BS.1770 “Algorithms to measure audio programme loudness and true-peak audio level” is the basis of every loudness standard in broadcasting. Its Annex 1 describes four stages:
- K-weighting. A two-stage pre-filter: a shelving filter that models the acoustic effect of the head, followed by a high-pass filter. The combination is called K-weighting.
- Mean square of each channel.
- Channel-weighted sum. Left, right and center have a weight of 1.0; the left and right surround channels 1.41 (about +1.5 dB). The LFE channel is not included. BS.1770-5 Annex 3 extends the weights to the loudspeaker layouts of advanced sound systems.
- Gating of 400 ms blocks that overlap by 75%, with two thresholds: an absolute gate at -70 LKFS, and a relative gate 10 dB below the level measured after the absolute gate.
The result is expressed in LKFS (Loudness, K-weighted, relative to nominal Full Scale). An increase of 1 dB in signal level raises the reading by 1 LKFS. A 0 dBFS, 997 Hz sine wave in the left, center or right channel reads -3.01 LKFS.
The gate is what makes the integrated value a measure of “foreground” loudness: silence and very quiet passages are left out, so a program with long quiet sections is not normalized louder than it should be.
The measures
| Measure | Definition | Defined in |
|---|---|---|
| Momentary loudness (M) | Sliding rectangular window of 0.4 s, not gated | EBU Tech 3341, 2.2 |
| Short-term loudness (S) | Sliding rectangular window of 3 s, not gated; live meters update at least 10 times per second | EBU Tech 3341, 2.2 |
| Integrated (program) loudness (I) | Gated measurement over the whole program (absolute gate -70 LUFS, relative gate -10 LU, 400 ms blocks with 75% overlap) | ITU-R BS.1770-5 Annex 1; EBU Tech 3341, 2.3 |
| Loudness Range (LRA) | Difference between the 10th and 95th percentiles of the distribution of short-term (3 s) loudness values, after an absolute gate of -70 LUFS and a relative gate of -20 LU | EBU Tech 3342 |
| True Peak | Maximum of the signal waveform in the continuous time domain, estimated by oversampling (4x at 48 kHz, to at least 192 kHz), in dBTP | ITU-R BS.1770-5 Annex 2 |
Note that LRA uses a relative gate of -20 LU, not the -10 LU used for integrated loudness. EBU R 128 does not recommend LRA for programs shorter than one minute, because there are too few short-term values, and Tech 3341 asks meters to flag the LRA as not yet stable during its first 60 seconds.
For a meter calibration check, Tech 3341 recommends a 1 kHz stereo sine wave at -18 dBFS peak in both channels, which should read -18.0 LUFS.
EBU R 128
EBU R 128 “Loudness normalisation and permitted maximum level of audio signals” (version 5.0, November 2023) recommends:
- Target: program loudness normalized to -23.0 LUFS.
- Tolerance: +/-1.0 LU where the target cannot practically be met, for example in live programs, and +/-0.2 LU in Quality Control workflows to allow for measurement errors.
- Maximum True Peak Level: -1 dBTP during production, measurement tolerance +/-0.3 dB. Lower maximums may apply for particular distribution systems and data-reduction rates (see EBU Tech 3344).
- Measurement: with a meter compliant with BS.1770 (including the gate) and EBU Tech 3341, on the audio signal in its entirety, “without emphasis on specific foreground elements such as speech, music or sound effects”.
- Supplementary measures: Loudness Range, Maximum Momentary Loudness and Maximum Short-term Loudness may be used to characterize a program.
Supplements cover short-form content (R 128 s1), streaming (s2), radio (s3) and cinematic content (s4). The EBU uses the unit LUFS; R 128 notes that it is equivalent to the LKFS of BS.1770.
ATSC A/85
ATSC A/85 “Techniques for Establishing and Maintaining Audio Loudness for Digital Television” is a Recommended Practice. The current edition, A/85:2026-07, replaces A/85:2013, clarifies the measurement guidelines and adds an annex on streaming (Annex L) and a one-page quick reference (Annex M). Many loudness meters and delivery specifications were set up for A/85:2013 and measure the full program mix with BS.1770 and its relative gate, so check which edition and which method a specification refers to.
- Target: for delivery or exchange of content without metadata, and without another agreement, -24 LKFS (section 6).
- Tolerance: a measurement tolerance of up to approximately +/-2 dB is anticipated, and content “should not be targeted to the high or low side of this tolerance”.
- True peak: kept below -2 dBTP, with up to approximately +/-0.5 dB anticipated meter variation.
- Anchor element: A/85 measures the element viewers set their volume by, normally dialogue. For long-form content, the average dialogue loudness is measured over the entire program, with a dialogue-gating algorithm and BS.1770 without the relative gate (which A/85 refers to as BS.1770-1). Short-form content such as commercials is measured on the full-program mix with BS.1770-3 or later.
- Metadata: for AC-3 and E-AC-3, the dialnorm value must match the measured loudness of the content; ATSC A/53 mandates correctly set dialnorm.
- Streaming: Annex L recommends a single target between -23 and -27 LKFS for streaming services that use codecs without loudness metadata.
R 128 and A/85 compared
| EBU R 128 (v5.0, 2023) | ATSC A/85:2026-07 | |
|---|---|---|
| Target | -23.0 LUFS | -24 LKFS |
| Tolerance | +/-1.0 LU (live), +/-0.2 LU (QC measurement) | approx. +/-2 dB measurement tolerance |
| Maximum true peak | -1 dBTP (production) | -2 dBTP |
| What is measured | The whole program, gated per BS.1770 | Long-form: dialogue (anchor element), dialogue-gated; short-form: full mix |
| Relative gate | Yes, -10 LU (BS.1770) | Not for long-form dialogue measurement; yes for short-form |
| Loudness Range | May be used to evaluate a program (Tech 3342) | Should not be enforced in content specifications |
| Unit | LUFS / LU | LKFS / dB |
Common pitfalls
Applying dialogue gating to R 128
R 128 measures the complete mix and explicitly says not to emphasize speech. Dialogue gating belongs to the A/85 long-form method. A meter set to a dialogue-gated mode will not give an R 128 program loudness.
Using a relative-gated meter for A/85 long-form
The reverse also applies: A/85:2026-07 states that the relative gate of BS.1770-2 and later is not a substitute for dialogue gating for long-form content.
Sample peak instead of true peak
The true peak of the reconstructed waveform can lie between samples and be higher than the largest sample value. A sample-peak meter therefore under-reads, and content that looks safe at -1 dBFS can exceed -1 dBTP. A/85 also notes that true-peak values may differ after decoding, which is why it recommends measuring before encoding.
Not measuring per program
Integrated loudness is defined over one program. A measurement that runs across a program and the following commercial break mixes long-form and short-form content. Reset or trigger the measurement at program boundaries, or measure files individually.
Judging by momentary or short-term values
Momentary and short-term loudness are expected to move around the target. The target applies to the integrated value. R 128 offers maximum momentary and short-term loudness only as optional limits.
Loudness metadata that does not match
For AC-3 and E-AC-3, the receiver adjusts playback level using dialnorm. If dialnorm does not match the actual loudness, the audience hears a level jump even when the audio was mixed correctly. R 128 also requires loudness metadata to indicate the actual program loudness.
Testing it in practice
- DVBAnalyzer has a Loudness Viewer that measures according to ITU-R BS.1770-5 and EBU R 128. It shows momentary (400 ms), short-term (3 s) and integrated loudness, Loudness Range per EBU Tech 3342 and true peak in dBTP, with EBU +9 and EBU +18 scales. Up to nine audio components can be metered side by side, the momentary loudness of many audio PIDs can be graphed over up to an hour, and the Manual Recorder measures one program from start to stop. The Loudness Meter tab shows the target in force and the deviation from it; for AC-3 and Dolby E services that signal dialnorm, the target follows the signaled value.
- DVBMonitor monitors loudness (ITU-R BS.1770, EBU R 128) 24/7 on many services from many transport streams, checks it against templates and raises alarms by SNMP and mail. The loudness measurements of the former DVBLoudness product, including program and service loudness, LRA, true peak and control of Audio Level Control equipment, are part of DVBMonitor; see DVBLoudness.
Both products also take OTT and SRT inputs next to broadcast, IP and SDI sources; see the OTT stream monitoring guide. For transport stream health, see the ETSI TR 101 290 guide, and for ad insertion markers the SCTE-35 guide.
Frequently asked questions
What is the difference between LUFS and LKFS?
None in value. LKFS is the unit used in ITU-R BS.1770; the EBU uses LUFS for the same measurement because it fits international naming conventions. -23 LUFS and -23 LKFS are the same level. LU is used for relative values: 1 LU equals 1 dB.
Why is the EBU target -23 LUFS and the ATSC target -24 LKFS?
They are separate recommendations with different measurement methods: R 128 measures the full mix with the BS.1770 relative gate, A/85 measures dialogue for long-form content. The one-unit difference in target cannot be read as a simple offset between the two methods.
What is a good Loudness Range?
Neither R 128 nor A/85 sets a required value. R 128 offers LRA as a tool to evaluate a program and its later dynamic processing; A/85:2026-07 says LRA should not be enforced in content specifications.
Does the LFE channel count toward loudness?
No. BS.1770 excludes the LFE channel from the loudness measurement. It can still clip, so it should still be metered for peaks and overload.