ITU-R BS.1770-5 defines how true peak is measured, but it does not give every producer one universal delivery ceiling. The number you should obey comes from the specification governing the finished file, the distribution path, or the broadcaster receiving it.
This distinction matters because -1 dBTP appears so often online that it starts looking like a law of digital audio. It is not. The same program can face one limit while it remains linear PCM and a different operating margin before a lossy codec.
A proper delivery check therefore starts with the destination, not with a favorite limiter preset. You need to know who receives the file, what format they expect, and whether their document states a hard maximum, a recommended ceiling, or a limiter setting for a later transmission stage.
The EBU adds an actual production requirement in R 128. For generic linear audio, program true peak must not exceed -1 dBTP, with the recommendation explicitly warning that lower permitted levels may apply to particular distribution systems and data-reduction rates. The limit belongs to a defined workflow, not to audio files in general.
This is where the practical difference between dBFS and dBTP becomes useful. Once you know which quantity the meter reports, the next job is identifying which authority sets its acceptable maximum. Confusing those two jobs is how a measurement standard gets mistaken for a delivery specification.
Broadcasters can also add house requirements on top. NRK, for example, currently recommends -3 dBTP for short-form delivery while treating levels above -1 dBTP as a QC failure in its published technical requirements. A file can therefore satisfy the broad EBU ceiling yet still miss the receiving broadcaster's preferred working margin.
The phrase "at the codec input" does real work. It identifies the point in the chain where the recommendation applies. Measuring a master before later filtering, sample-rate conversion, gain changes, or other processing does not prove the signal reaching the encoder still has the same peak.
EBU distribution guidance goes further by separating production level from the levels used around specific broadcast codec systems. Some codec paths call for more conservative end-stage limiting than the linear program limit. A single number copied from a mastering article hides this stage dependence and can leave too little margin where conversion actually occurs.
The engineering reason is not mysterious. Filtering, reconstruction, sample-rate conversion, and data reduction can reshape a waveform enough to raise its peak while reducing or barely changing its overall energy. Measured converter overload behavior documents the broader failure mode, including overload during signal conversion even when stored samples appear to leave room.
The same caution applies when somebody hands you a platform target from a chart. Playback normalization, recommended mastering practice, codec-input protection, and file-rejection limits are different things. A service can turn audio down during playback without requiring every uploaded master to arrive at the normalization target.
Read the receiving specification literally. "Recommended" gives you an operating preference, "shall not exceed" establishes a ceiling, and a QC failure threshold tells you where the asset is rejected. Those values can sit at different points even inside one organization's workflow.
For a television deliverable, use the broadcaster or distributor's current technical sheet first, then trace its references to standards such as R 128, A/85, or BS.1770 where needed. For lossy streaming infrastructure, check the level at the codec input and leave any extra margin the codec, bit rate, or downstream processing requires.
A master prepared for several destinations may need separate deliverables rather than one aggressively optimized file. Keep the clean parent master, create each required version from it, and measure the finished asset against the specification attached to that handoff. The valid ceiling is the one governing the file you are actually sending.
This distinction matters because -1 dBTP appears so often online that it starts looking like a law of digital audio. It is not. The same program can face one limit while it remains linear PCM and a different operating margin before a lossy codec.
A proper delivery check therefore starts with the destination, not with a favorite limiter preset. You need to know who receives the file, what format they expect, and whether their document states a hard maximum, a recommended ceiling, or a limiter setting for a later transmission stage.
Measurement rules and delivery limits do different jobs
BS.1770 provides the true-peak measurement method used by many later specifications. It tells compliant meters how to estimate peaks between stored samples, which makes readings comparable across equipment that follows the recommendation. It does not turn -1 dBTP, -2 dBTP, or any other number into a global mastering target.The EBU adds an actual production requirement in R 128. For generic linear audio, program true peak must not exceed -1 dBTP, with the recommendation explicitly warning that lower permitted levels may apply to particular distribution systems and data-reduction rates. The limit belongs to a defined workflow, not to audio files in general.
This is where the practical difference between dBFS and dBTP becomes useful. Once you know which quantity the meter reports, the next job is identifying which authority sets its acceptable maximum. Confusing those two jobs is how a measurement standard gets mistaken for a delivery specification.
Broadcasters can also add house requirements on top. NRK, for example, currently recommends -3 dBTP for short-form delivery while treating levels above -1 dBTP as a QC failure in its published technical requirements. A file can therefore satisfy the broad EBU ceiling yet still miss the receiving broadcaster's preferred working margin.
Codec input changes what the ceiling is protecting
Lossy delivery complicates the number because encoding can create peaks that were not present in the linear source. AES guidance for internet streaming recommends keeping maximum true peak at or below -1 dBTP at the codec input, while also noting that lower bit rates can require a lower limiting threshold because peak overshoot tends to increase.The phrase "at the codec input" does real work. It identifies the point in the chain where the recommendation applies. Measuring a master before later filtering, sample-rate conversion, gain changes, or other processing does not prove the signal reaching the encoder still has the same peak.
EBU distribution guidance goes further by separating production level from the levels used around specific broadcast codec systems. Some codec paths call for more conservative end-stage limiting than the linear program limit. A single number copied from a mastering article hides this stage dependence and can leave too little margin where conversion actually occurs.
The engineering reason is not mysterious. Filtering, reconstruction, sample-rate conversion, and data reduction can reshape a waveform enough to raise its peak while reducing or barely changing its overall energy. Measured converter overload behavior documents the broader failure mode, including overload during signal conversion even when stored samples appear to leave room.
Current specifications outrank inherited mastering folklore
Delivery documents change, sometimes substantially. ATSC approved a major revision of A/85 on July 8, 2026, its first comprehensive update since 2013, adding modern guidance for broadcast and streaming workflows plus a new loudness and true-peak quick reference. An old preset labeled "ATSC" is not evidence that your current job matches the current document.The same caution applies when somebody hands you a platform target from a chart. Playback normalization, recommended mastering practice, codec-input protection, and file-rejection limits are different things. A service can turn audio down during playback without requiring every uploaded master to arrive at the normalization target.
Read the receiving specification literally. "Recommended" gives you an operating preference, "shall not exceed" establishes a ceiling, and a QC failure threshold tells you where the asset is rejected. Those values can sit at different points even inside one organization's workflow.
For a television deliverable, use the broadcaster or distributor's current technical sheet first, then trace its references to standards such as R 128, A/85, or BS.1770 where needed. For lossy streaming infrastructure, check the level at the codec input and leave any extra margin the codec, bit rate, or downstream processing requires.
A master prepared for several destinations may need separate deliverables rather than one aggressively optimized file. Keep the clean parent master, create each required version from it, and measure the finished asset against the specification attached to that handoff. The valid ceiling is the one governing the file you are actually sending.
